Homotopical Algebraic Geometry II:

arXiv:math/0404373v7 [math.AG] 14 Mar 2006

geometric stacks and applications

Bertrand Toën
Gabriele Vezzosi
Laboratoire Emile Picard UMR CNRS 5580, Université Paul Sabatier,
Toulouse
E-mail address: [email protected]

Dipartimento di Matematica Applicata “G. Sansone”, Università di

Firenze
E-mail address: [email protected]
This work is dedicated to Alexandre Grothendieck
 Contents

Abstract ix

Introduction 1
Reminders on abstract algebraic geometry 1
The setting 2
Linear and commutative algebra in a symmetric monoidal model category 2
Geometric stacks 3
Infinitesimal theory 4
Higher Artin stacks (after C. Simpson) 4
Derived algebraic geometry: D− -stacks 4
Complicial algebraic geometry: D-stacks 6
Brave new algebraic geometry: S-stacks 6
Relations with other works 7
Acknowledgments 8
Notations and conventions 9

Part 1. General theory of geometric stacks 11

Introduction to Part 1 13

Chapter 1.1. Homotopical algebraic context 15

Chapter 1.2. Preliminaries on linear and commutative algebra in an HA context 25

1.2.1. Derivations and the cotangent complex 25
1.2.2. Hochschild homology 29
1.2.3. Finiteness conditions 30
1.2.4. Some properties of modules 35
1.2.5. Formal coverings 36
1.2.6. Some properties of morphisms 37
1.2.7. Smoothness 41
1.2.8. Infinitesimal lifting properties 41
1.2.9. Standard localizations and Zariski open immersions 43
1.2.10. Zariski open immersions and perfect modules 47
1.2.11. Stable modules 49
1.2.12. Descent for modules and stable modules 54
1.2.13. Comparison with the usual notions 57

Chapter 1.3. Geometric stacks: Basic theory 61

1.3.1. Reminders on model topoi 61
1.3.2. Homotopical algebraic geometry context 65
1.3.3. Main definitions and standard properties 76
1.3.4. Quotient stacks 80
1.3.5. Quotient stacks and torsors 83
vii
viii CONTENTS

1.3.6. Properties of morphisms 86

1.3.7. Quasi-coherent modules, perfect modules and vector bundles 88
Chapter 1.4. Geometric stacks: Infinitesimal theory 97
1.4.1. Tangent stacks and cotangent complexes 97
1.4.2. Obstruction theory 105
1.4.3. Artin conditions 109

Part 2. Applications 121

Introduction to Part 2 123
Chapter 2.1. Geometric n-stacks in algebraic geometry (after C. Simpson) 129
2.1.1. The general theory 129
2.1.2. Comparison with Artin’s algebraic stacks 132
Chapter 2.2. Derived algebraic geometry 135
2.2.1. The HA context 135
2.2.2. Flat, smooth, étale and Zariski open morphisms 139
2.2.3. The HAG context: Geometric D− -stacks 151
2.2.4. Truncations 154
2.2.5. Infinitesimal criteria for smooth and étale morphisms 159
2.2.6. Some examples of geometric D− -stacks 164
2.2.6.1. Local systems 164
2.2.6.2. Algebras over an operad 170
2.2.6.3. Mapping D− -stacks 173
Chapter 2.3. Complicial algebraic geometry 177
2.3.1. Two HA contexts 177
2.3.2. Weakly geometric D-stacks 181
2.3.3. Examples of weakly geometric D-stacks 182
2.3.3.1. Perfect modules 182
2.3.3.2. The D-stacks of dg-algebras and dg-categories 183
2.3.4. Geometric D-stacks 187
2.3.5. Examples of geometric D-stacks 188
2.3.5.1. D− -stacks and D-stacks 188
2.3.5.2. CW-perfect modules 188
2.3.5.3. CW-dg-algebras 192
2.3.5.4. The D-stack of negative CW-dg-categories 193
Chapter 2.4. Brave new algebraic geometry 199
2.4.1. Two HAG contexts 199
2.4.2. Elliptic cohomology as a Deligne-Mumford S-stack 203
Appendix A. Classifying spaces of model categories 207
Appendix B. Strictification 211
Appendix C. Representability criterion (after J. Lurie) 215
Bibliography 219
 Abstract

1
This is the second part of a series of papers called “HAG”, and devoted to
develop the foundations of homotopical algebraic geometry. We start by defining
and studying generalizations of standard notions of linear algebra in an abstract
monoidal model category, such as derivations, étale and smooth morphisms, flat and
projective modules, etc. We then use our theory of stacks over model categories,
introduced in [HAGI], in order to define a general notion of geometric stack over a
base symmetric monoidal model category C, and prove that this notion satisfies the
expected properties.
The rest of the paper consists in specializing C in order to give various exam-
ples of applications in several different contexts. First of all, when C = k − M od is
the category of k-modules with the trivial model structure, we show that our notion
gives back the algebraic n-stacks of C. Simpson. Then we set C = sk − M od, the
model category of simplicial k-modules, and obtain this way a notion of geometric
D− -stack which are the main geometric objects of derived algebraic geometry. We
give several examples of derived version of classical moduli stacks, as the D− -stack
of local systems on a space, the D− -stack of algebra structures over an operad, the
D− -stack of flat bundles on a projective complex manifold, etc. We also present the
cases where C = C(k) is the model category of unbounded complexes of k-modules,
and C = SpΣ the model category of symmetric spectra. In these two contexts we give
some examples of geometric stacks such as the stack of associative dg-algebras, the
stack of dg-categories, and a geometric stack constructed using topological modular
forms.

There are more things in heaven and earth, Horatio,

than are dreamt of in our philosophy. But come...

W. Shakespeare, Hamlet, Act 1, Sc. 5.

Mon cher Cato, il faut en convenir, les forces de l’ether nous pénètrent,
et ce fait déliquescent il nous faut l’appréhender coute que coute.

P. Sellers, Quand la panthère rose s’en mêle.

12000 Mathematics Subject Classification: 14A20, 18G55, 18F10, 55U40, 55P42, 55P43, 18F20,
18D10, 18E30 18G35, 18G30, 13D10, 55N34.
Keywords: Algebraic stacks, higher stacks, derived algebraic geometry
Received by the Editor June 14 2005

ix
 Introduction

This is the second part of a series of papers called “HAG”, devoted to start the
development of homotopical algebraic geometry. The first part [HAGI] was concerned
with the homotopical generalization of sheaf theory, and contains the notions of model
topologies, model sites, stacks over model sites and model topoi, all of these being
homotopical versions of the usual notions of Grothendieck topologies, sites, sheaves
and topoi. The purpose of the present work is to use these new concepts in some spe-
cific situations, in order to introduce a very general notion of geometric stacks, a far
reaching homotopical generalization of the notion of algebraic stacks introduced by
P. Deligne, D. Mumford and M. Artin. This paper includes the general study and the
standard properties of geometric stacks, as well as various examples of applications
in the contexts of algebraic geometry and algebraic topology.

Reminders on abstract algebraic geometry. A modern point of view on

algebraic geometry consists of viewing algebraic varieties and schemes through their
functors of points. In this functorial point of view, schemes are certain sheaves of
sets on the category of commutative rings endowed with a suitable topology (e.g.
the Zariski topology). Keeping this in mind, it turns out that the whole theory
of schemes can be completely reconstructed starting from the symmetric monoidal
category Z − M od of Z-modules alone. Indeed, the category of commutative rings
is reconstructed by taking the category of commutative monoids in Z − M od. Flat
morphisms can be recognized via the exactness property of the base change functor on
the category of modules. Finitely presented morphisms are recognized via the usual
categorical characterization in terms of commutation of mapping sets with respect to
filtered colimits. Finally, Zariski open immersion can be defined as flat morphisms of
finite presentation A −→ B such that B ≃ B ⊗A B. Schemes are then reconstructed
as certain Zariski sheaves on the opposite category of commutative rings, which are
obtained by gluing affine schemes via Zariski open immersions (see for example the
first chapters of [Dem-Gab]).
The fact that the notion of schemes has such a purely categorical interpretation
has naturally lead to the theory of relative algebraic geometry, in which the base
symmetric monoidal category Z − M od is replaced by an abstract base symmetric
monoidal category C, and under reasonable assumptions on C the notion of schemes
over C can be made meaningful as well as useful (see for example [Del1, Ha] for some
applications).
The key observation of this work is that one can generalize further the theory of
relative algebraic geometry by requiring C to be endowed with an additional model
category structure, compatible with its monoidal structure (relative algebraic geom-
etry is then recovered by taking the trivial model structure), in such a way that the
notions of schemes and more generally of algebraic spaces or algebraic stacks still
have a natural and useful meaning, compatible with the homotopy theory carried
1
2 INTRODUCTION

by C. In this work, we present this general theory, and show how this enlarges the
field of applicability by investigating several examples not covered by the standard
theory of relative algebraic geometry. The most important of these applications is the
existence of foundations for derived algebraic geometry, a global counter part of the
derived deformation theory of V. Drinfel’d, M. Kontsevich and al.

The setting. Our basic datum is a symmetric monoidal model category C (in
the sense of [Ho1]), on which certain conditions are imposed (see assumptions 1.1.0.1,
1.1.0.2, 1.1.0.3 and 1.1.0.4). We briefly discuss these requirements here. The model
category C is assumed to satisfy some reasonable additional properties (as for example
being proper, or that cofibrant objects are flat for the monoidal structure). These
assumptions are only made for the convenience of certain constructions, and may
clearly be omitted. The model category C is also assumed to be combinatorial (see
e.g. [Du2]), making it reasonably behaved with respect to localization techniques.
The first really important assumption on C states that it is pointed (i.e. that the
final and initial object coincide) and that its homotopy category Ho(C) is additive.
This makes the model category C homotopically additive, which is a rather strong
condition, but is used all along this work and seems difficult to avoid (see however
[To-Va2]). Finally, the last condition we make on C is also rather strong, and states
that the theory of commutative monoids in C, and the theory of modules over them,
both possess reasonable model category structures. This last condition is of course far
from being satisfied in general (as for example it is not satisfied when C is the model
category of complexes over some ring which is not of characteristic zero), but all the
examples we have in mind can be treated in this setting2. The model categories of
simplicial modules, of complexes over a ring of characteristic zero, and of symmetric
spectra are three important examples of symmetric monoidal model category satisfy-
ing all our assumptions. More generally, the model categories of sheaves with values
in any of these three fundamental categories provide additional examples.

Linear and commutative algebra in a symmetric monoidal model cate-

gory. An important consequence of our assumptions on the base symmetric monoidal
model category C is the existence of reasonable generalizations of general construc-
tions and results from standard linear and commutative algebra. We have gathered
some of these notions (we do not claim to be exhaustive) in §1.1. For example, we
give definitions of derivations as well as of cotangent complexes representing them,
we define formally étale morphisms, flat morphisms, open Zariski immersions, for-
mally unramified morphisms, finitely presented morphism of commutative monoids
and modules, projective and flat modules, Hochschild cohomology, etc. They are
all generalizations of the well known notions in the sense that when applied to the
case where C = Z − M od with the trivial model structure we find back the usual
notions. However, there are sometimes several nonequivalent generalizations, as for
example there exist at least two, nonequivalent reasonable generalizations of smooth
morphisms which restrict to the usual one when C = Z − M od. This is why we have
tried to give an overview of several possible generalizations, as we think all definitions
could have their own interest depending both on the context and on what one wants
to do with them. Also we wish to mention that all these notions depend heavily on
the base model category C, in the sense that the same object in C, when considered
2Alternatively, one could switch to E -algebras (and modules over them) for which useful
∞
model and semi-model structures are known to exist, thanks to the work of M. Spitzweck [Sp], in
much more general situations than for the case of commutative monoids (and modules over them).
 INTRODUCTION 3

in different model categories structures on C, might not behave the same way. For
example, a commutative ring can also be considered as a simplicial commutative ring,
and the notion of finitely presented morphisms is not the same in the two cases. We
think that keeping track of the base model category C is rather important, since play-
ing with the change of base categories might be very useful, and is also an interesting
feature of the theory.
The reader will immediately notice that several notions behave in a much better
way when the base model category satisfies certain stability assumptions (e.g. is a
stable model category, or when the suspension functor is fully faithful, see for ex-
ample Prop. 1.2.6.5, Cor. 1.2.6.6). We think this is one of the main features of
homotopical algebraic geometry: linear and commutative algebra notions tend to be
better behaved as the base model category tend to be “more” stable. We do not claim
that everything becomes simpler in the stable situation, but that certain difficulties
encountered can be highly simplified by enlarging the base model category to a more
stable one.

Geometric stacks. In §1.3 we present the general notions of geometric stacks

relative to our base model category C. Of course, we start by defining Af fC , the model
category of affine objects over C, as the opposite of the model category Comm(C) of
commutative monoids in C. We assume we are given a model (pre-)topology τ on
Af fC , in the sense we have given to this expression in [HAGI, Def. 4.3.1] (see also
Def. 1.3.1.1). We also assume that this model topology satisfies certain natural
assumptions, as quasi-compactness and the descent property for modules. The model
category Af fC together with its model topology τ is a model site in the sense of
[HAGI, Def. 4.3.1] or Def. 1.3.1.1, and it gives rise to a model category of stacks
Af fC∼,τ . The homotopy category of Af fC∼,τ will simply be denoted by St(C, τ ). The
Yoneda embedding for model categories allows us to embed the homotopy category
Ho(Af fC ) into St(C, τ ), and this gives a notion of representable stack, our analog of
the notion of affine scheme. Geometric stacks will result from a certain kind of gluing
representable stacks.
Our notion of geometric stack is relative to a class of morphisms P in Af fC ,
satisfying some compatibility conditions with respect to the topology τ , essentially
stating that the notion of morphisms in P is local for the topology τ . With these
two notions, τ and P, we define by induction on n a notion of n-geometric stack (see
1.3.3.1). The precise definition is unfortunately too long to be reproduced here, but
one can roughly say that n-geometric stacks are stacks F whose diagonal is (n − 1)-
` (n − 1)-geometric stacks),
representable (i.e. its fibers over representable stacks are
and which admits a covering by representable stacks i Ui −→ F , such that all
morphisms Ui −→ F are in P.
The notion of n-geometric stack satisfies all the expected basic properties. For ex-
ample, geometric stacks are stable by (homotopy) fiber products and disjoint unions,
and being an n-geometric stack is a local property (see Prop. 1.3.3.3, 1.3.3.4). We
also present a way to produce n-geometric stacks as certain quotients of groupoid
actions, in the same way that algebraic stacks (in groupoids) can always be presented
as quotients of a scheme by a smooth groupoid action (see Prop. 1.3.4.2). When a
property Q of morphisms in Af fC satisfies a certain compatibility with both P and τ ,
there exists a natural notion of Q-morphism between n-geometric stacks, satisfying
all the expected properties (see Def. 1.3.6.4, Prop. 1.3.6.3). We define the stack of
quasi-coherent modules, as well its sub-stacks of vector bundles and of perfect mod-
ules (see Thm. 1.3.7.2, Cor. 1.3.7.3). These also behave as expected, and for example
4 INTRODUCTION

the stack of vector bundles is shown to be 1-geometric as soon as the class P contains
the class of smooth morphisms (see Cor. 1.3.7.12).

Infinitesimal theory. In §1.4, we investigate the infinitesimal properties of geo-

metric stacks. For this we define a notion of derivation of a stack F with coefficients
in a module, and the notion of cotangent complex is defined via the representability of
the functor of derivations (see Def. 1.4.1.4, 1.4.1.5, 1.4.1.7). The object representing
the derivations, the cotangent complex, is not in general an object in the base model
category C, but belongs to the stabilization of C (this is of course related to the well
known fact that cotangent spaces of algebraic stacks are not vector spaces but rather
complexes of vector spaces). This is why these notions will be only defined when
the suspension functor of C is fully faithful, or equivalently when the stabilization
functor from C to its stabilization is fully faithful (this is again an incarnation of the
fact, mentioned above, that homotopical algebraic geometry seems to prefer stable
situations). In a way, this explains from a conceptual point of view the fact that
the infinitesimal study of usual algebraic stacks in the sense of Artin is already part
of homotopical algebraic geometry, and does not really belong to standard algebraic
geometry. We also define stacks having an obstruction theory (see Def. 1.4.2.1), a
notion which controls obstruction to lifting morphisms along a first order deformation
in terms of the cotangent complex. Despite its name, having an obstruction theory
is a property of a stack and not an additional structure. Again, this notion is really
well behaved when the suspension functor of C is fully faithful, and this once again
explains the relevance of derived algebraic geometry with respect to infinitesimal de-
formation theory. Finally, in the last section we give sufficient conditions (that we
called Artin’s conditions) insuring that any n-geometric stack has an obstruction the-
ory (Thm. 1.4.3.2). This last result can be considered as a far reaching generalization
of the exixtence of cotangent complexes for algebraic stacks as presented in [La-Mo].

Higher Artin stacks (after C. Simpson). As a first example of application,

we show how our general notion of geometric stacks specializes to C. Simpson’s alge-
braic n-stacks introduced in [S3]. For this, we let C = k − M od, be the symmetric
monoidal category of k-modules (for some fixed commutative ring k), endowed with
its trivial model structure. The topology τ is chosen to be the étale (ét) topology,
and P is chosen to be the class of smooth morphisms. We denote by St(k) the corre-
sponding homotopy category of ét-stacks. Then, our definition of n-geometric stack
gives back the notion of algebraic n-stack introduced in [S3] (except that the two n’s
might differ); these stacks will be called Artin n-stacks as they contain the usual alge-
braic stacks in the sense of Artin as particular cases (see Prop. 2.1.2.1). However, all
our infinitesimal study (cotangent complexes and obstruction theory) does not apply
here as the suspension functor on k − M od is the zero functor. This should not be
viewed as a drawback of the theory; on the contrary we rather think this explains
why deformation theory and obstruction theory in fact already belong to the realm
of derived algebraic geometry, which is our next application.

Derived algebraic geometry: D− -stacks. Our second application is the so-

called derived algebraic geometry. The base model category C is chosen to be sk−M od,
the symmetric monoidal model category of simplicial commutative k-modules, k being
some fixed base ring. The category of affine objects is k − D− Af f , the opposite
model category of the category of commutative simplicial k-algebras. In this setting,
 INTRODUCTION 5

our general notions of flat, étale, smooth morphisms and Zariski open immersions
all have explicit descriptions in terms of standard notions (see Thm. 2.2.2.6). More
precisely, we prove that a morphism of simplicial commutative k-algebras A −→ B is
flat (resp. smooth, resp. étale, resp. a Zariski open immersion) in the general sense
we have given to these notions in §1.2, if and only if it satisfies the following two
conditions
• The induced morphism of affine schemes
Spec π0 (B) −→ Spec π0 (A)
is flat (resp. smooth, resp. étale, resp. a Zariski open immersion) in the
usual sense.
• The natural morphism
π∗ (A) ⊗π0 (A) π0 (B) −→ π∗ (B)
is an isomorphism.
We endow k − D− Af f with the étale model topology, a natural extension of
the étale topology for affine schemes; the corresponding homotopy category of D− -
stacks is simply denoted by D− St(k). The class P is taken to be the class of smooth
morphisms. The n-geometric stacks in this context will be called n-geometric D− -
stacks, where the notation D− is meant to bring to mind the negative bounded derived
category3. An important consequence of the above descriptions of étale and smooth
morphisms is that the natural inclusion functor from the category of k-modules to the
category of simplicial k-modules, induces a full embedding of the category of Artin
n-stacks into the category of n-geometric D− -stack. This inclusion functor i has
furthermore a right adjoint, called the truncation functor t0 (see Def. 2.2.4.3), and
the adjunction morphism it0 (F ) −→ F provides a closed embedding of the classical
Artin n-stack it0 (F ) to its derived version F , which behaves like a formal thickening
(see Prop. 2.2.4.7). This is a global counterpart of the common picture of derived
deformation theory of a formal classical moduli space sitting as a closed sub-space in
the corresponding formal derived moduli space.
We also prove that our general conditions for the existence of an obstruction the-
ory are satisfied, and so any n-geometric D− -stack has an obstruction theory (see
Prop. 2.2.3.3). An important particular case is when this result is applied to the
image of an Artin n-stack via the natural inclusion functor i; we obtain in this way
the existence of an obstruction theory for any Artin n-stack, and in particular the
existence of a cotangent complex. This is a very good instance of our principle that
things simplifies when the base model category C becomes more stable: the infinitesi-
mal study of classical objects of algebraic geometry (such as schemes, algebraic stacks
or Artin n-stacks) becomes conceptually clearer and behaves much better when we
consider these objects as geometric D− -stacks.
Finally, we give several examples of D− -stacks being derived versions of some well
known classical moduli problems. First of all the D− -stack of local systems on a topo-
logical space, and the D− -stack of algebra structures over a given operad are shown to
be 1-geometric (see Lem. 2.2.6.3, Prop. 2.2.6.8). We also present derived versions of
the scheme of morphisms between two projective schemes, and of the moduli stack of
flat bundles on a projective complex manifold (Cor. 2.2.6.14 and Cor. 2.2.6.15). The
proofs that these last two stacks are geometric rely on a special version of J. Lurie’s

3Recall that the homotopy theory of simplicial k-modules is equivalent to the homotopy theory
of negatively graded cochain complexes of k-modules. Therefore, derived algebraic geometry can
also be considered as algebraic geometry over the category of negatively graded complexes.
6 INTRODUCTION

representability theorem (see [Lu1] and Appendix C).

Complicial algebraic geometry: D-stacks. What we call complicial alge-

braic geometry is an unbounded version of derived algebraic geometry in which the
base model category is C(k) the category of complexes over some commutative ring
k (of characteristic zero), and is presented in §2.3. It turns out that linear algebra
over C(k) behaves rather differently than over the category of simplicial k-modules
(corresponding to complexes in non-positive degrees). Indeed, the smooth, étale and
Zariski open immersion can not be described using a simple description on homotopy
groups anymore. For example, a usual ring A may have Zariski open localizations
A −→ A′ in the context of complicial algebraic geometry such that A′ is not cohomo-
logically concentrated in degree 0 anymore. Also, a usual non affine scheme might be
affine when considered as a scheme over C(k): for example any quasi-compact open
subscheme of a usual affine scheme is representable by an affine scheme over C(k)
(see example 2.3.1.5).
This makes the complicial theory rather different from derived algebraic geom-
etry for which the geometric intuition was instead quite close to the usual one, and
constitutes a very interesting new feature of complicial algebraic geometry. The cate-
gory Af fC is here the opposite of the category of unbounded commutative differential
graded algebras over k. It is endowed with a strong étale topology, and the correspond-
ing homotopy category of D-stacks is simply denoted by DSt(k). A new feature is
here the existence of several interesting choices for the class P. We will present two
of them, one for which P is taken to be the class of perfect morphisms, a rather
weak notion of smoothness, and a second one for which P is taken to be the class of
fip-smooth morphism, a definitely stronger notion behaving similarly to usual smooth
morphisms with respect to lifting properties. We check that such choices satisfy the re-
quired properties in order for n-geometric stacks (called weakly n-geometric D-stacks
and n-geometric D-stacks, according to the choice of P) to make sense. Furthermore,
for our second choice, we prove that Artin’s conditions are satisfied, and thus that
n-geometric D-stacks have a good infinitesimal theory. We give several examples of
weakly geometric D-stacks, the first one being the D-stack of perfect modules Perf .
We also show that the D-stack of associative algebra structures Ass is a weakly 1-
geometric D-stack. Finally, the D-stack of connected dg-categories Cat∗ is shown
to be weakly 2-geometric. It is important to note that these D-stacks can not be
reasonably described as geometric D− -stacks, and provide examples of truly “exotic”
geometric objects.
Suitable slight modifications of the D-stacks Perf , Ass and Cat∗ are given and
shown to be geometric. This allows us to study their tangent complexes, and show
in particular that the infinitesimal theory of a certain class of dg-algebras and dg-
categories is controlled by derivations and Hochschild cohomology, respectively (see
Cor. 2.3.5.9 and Cor. 2.3.5.12). We also show that Hochschild cohomology does
not control deformations of general dg-categories in any reasonable sense (see Cor.
2.3.5.13 and Rem. 2.3.5.14). This has been a true surprise, as it contradicts some
of the statements one finds in the existing literature, including some made by the
authors themselves (see e.g. [To-Ve2, Thm. 5.6]).

Brave new algebraic geometry: S-stacks. Our last context of application,

briefly presented in §2.4, is the one where the base symmetric monoidal model cate-
gory is C = SpΣ , the model category of symmetric spectra ([HSS, Shi]), and gives
rise to what we call, after F. Waldhausen, brave new algebraic geometry. Like in the
 INTRODUCTION 7

complicial case, the existence of negative homotopy groups makes the general theory
of flat, smooth, étale morphisms and of Zariski open immersions rather different from
the corresponding one in derived algebraic geometry. Moreover, typical phenomena
coming from the existence of Steenrod operations makes the notion of smooth mor-
phism even more exotic and rather different from algebraic geometry; to give just a
striking example, Z[T ] is not smooth over Z in the context of brave new algebraic
geometry. Once again, we do not think this is a drawback of the theory, but rather
an interesting new feature one should contemplate and try to understand, as it might
reveal interesting new insights also on classical objects. In brave new algebraic ge-
ometry, we also check that the strong étale topology and the class P of fip-smooth
morphisms satisfy our general assumptions, so that n-geometric stacks exists in this
context. We call them n-geometric S-stacks, while the homotopy category of S-stacks
for the strong étale topology is simply denoted by St(S). As an example, we give a
construction of a 1-geometric S-stack starting from the “sheaf” of topological modu-
lar forms (Thm. 2.4.2.1).

Relations with other works. It would be rather long to present all related
works, and we apologize in advance for not mentioning all of them.
The general fact that the notion of geometric stack only depends on a topology
and a choice of the class of morphisms P has already been stressed by Carlos Simpson
in [S3], who attributes this idea to C. Walter. Our general definition of geometric
n-stacks is a straightforward generalization to our abstract context of the definitions
found in [S3].
Originally, derived algebraic geometry have been approached using the notion of
dg-schemes, as introduced by M. Kontsevich, and developed by I. Ciocan-Fontanine
and M. Kapranov. We have not tried to make a full comparison with our theory.
Let us only mention that there exists a functor from dg-schemes to our category of
1-geometric D− -stacks (see [To-Ve2, §3.3]). Essentially nothing is known about this
functor: we tend to believe that it is not fully faithful, though this question does
not seem very relevant. On the contrary, the examples of dg-schemes constructed in
[Ci-Ka1, Ci-Ka2] do provide examples of D− -stacks and we think it is interesting to
look for derived moduli-theoretic interpretations of these (i.e. describe their functors
of points).
We would like to mention that an approach to formal derived algebraic geometry
has been settled down by V. Hinich in [Hin2]. As far as we know, this is the first
functorial point of view on derived algebraic geometry that appeared in the literature.
There is a big overlap between our Chapter 2.2 and Jacob Lurie’s thesis [Lu1].
The approach to derived algebraic geometry used by J. Lurie is different from ours as
it is based on a notion of ∞-category, whereas we are working with model categories.
The simplicial localization techniques of Dwyer and Kan provide a way to pass from
model categories to ∞-categories, and the “strictification” theorem of [To-Ve1, Thm.
4.2.1] can be used to see that our approach and Lurie’s approach are in fact equivalent
(up to some slight differences, for instance concerning the notion of descent). We
think that the present work and [Lu1] can not be reasonably considered as totally
independent, as their authors have been frequently communicating on the subject
since the spring 2002. It seems rather clear that we all have been influenced by these
communications and that we have greatly benefited from the reading of the first drafts
of [Lu1]. We have to mention however that a huge part of the material of the present
paper had been already announced in earlier papers (see e.g. [To-Ve4, To-Ve2]),
and have been worked out since the summer 2000 at the time were our project has
8 INTRODUCTION

started. The two works are also rather disjoint and complementary, as [Lu1] contains
much more materials on derived algebraic geometry than what we have included in
§2.2 (e.g. a wonderful generalization of Artin’s representability theorem, to state
only the most striking result). On the other hand, our “HAG” project has also been
motivated by rather exotic contexts of applications, as the ones exposed for example
in §2.3 and §2.4 , and which are not covered by the framework of [Lu1].
The work of K. Behrend on dg-schemes and dg-stacks [Be1, Be2] has been done
while we were working on our project, and therefore §2.2 also has some overlaps
with his work. However, the two approaches are rather different and nonequivalent,
as K. Behrend uses a 2-truncated version of our notions of stacks, with the effect
of killing some higher homotopical information. We have not investigated a precise
comparison between these two approaches in this work, but we would like to mention
that there exists a functor from our category of D− -stacks to K. Behrend’s category
of dg-sheaves. This functor is extremely badly behaved: it is not full, nor faithful, nor
essentially surjective, nor even injective on isomorphism classes of objects. The only
good property is that it sends 1-geometric Deligne-Mumford D− -stacks to Deligne-
Mumford dg-stack in Behrend’s sense. However, there are non geometric D− -stacks
that become geometric objects in Behrend’s category of dg-sheaves.
Some notions of étale and smooth morphisms of commutative S-algebras have
been introduced in [MCM], and they seem to be related to the general notions we
present in §1.2. However a precise comparison is not so easy. Moreover, [MCM] con-
tains some wrong statements like the fact that thh-smoothness generalizes smoothness
for discrete algebras (right after Definition 4.2) or like Lemma 4.2 (2). The proof of
Theorem 6.1 also contains an important gap, since the local equivalences at the end
of the proof are not checked to glue together.
Very recentely, J. Rognes has proposed a brave new version of Galois theory,
including brave new notions of étaleness which are very close to our notions (see
[Ro]).
A construction of the moduli of dg-algebras and dg-categories appears in [Ko-So].
These moduli are only formal moduli by construction, and we propose our D-stacks
Ass and Cat∗ as their global geometrical counterparts.
We wish to mention the work of M. Spitzweck [Sp], in which he proves the exis-
tence of model category structures for E∞ -algebras and modules in a rather general
context. This work can therefore be used in order to suppress our assumptions on
the existence of model category of commutative monoids. Also, a nice symmetric
monoidal model category of motivic complexes is defined in [Sp], providing a new in-
teresting context to investigate. It has been suggested to us to consider this example
of algebraic geometry over motives by Yu. Manin, already during spring 2000, but we
do not have at the moment interesting things to say on the subject.
Finally, J. Gorski has recently constructed a D− -stack version of the Quot functor
(see [Go]), providing this way a functorial interpretation of the derived Quot scheme
of [Ci-Ka1]. A geometric D− -stack classifying objects in a dg-category has been
recently constructed by the first author and M. Vaquié in [To-Va1].

Acknowledgments. First of all, we are very grateful to C. Simpson for sharing

with us all his ideas on higher stacks, and for encouraging us to pursue our project.
We owe him a lot.
We are very grateful to J. Lurie for various communications on the subject, and
for sharing with us his work [Lu1]. We have learned a lot about derived algebraic
geometry from him.
 INTRODUCTION 9

We thank M. Vaquié for reading a former version of the present work, and for his
comments and suggestions.
For various conversations on the subject we thank M. Anel, J. Gorski, A. Hirschowitz,
A. Joyal, M. Kontsevich, L. Katzarkov, H. Miller, T. Pantev, C. Rezk, J. Rognes, S.
Schwede, B. Shipley, M. Spitzweck, N. Strickland and J. Tapia.
Finally, we thank both referees for their careful reading of the manuscript and
for their interesting and useful remarks and suggestions.

Notations and conventions. We will use the word universe in the sense of
[SGA4-I, Exp.I, Appendice]. Universes will be denoted by U ∈ V ∈ W . . . . For any
universe U we will assume that N ∈ U. The category of sets (resp. simplicial sets,
resp. . . . ) belonging to a universe U will be denoted by SetU (resp. SSetU , resp. . . . ).
The objects of SetU (resp. SSetU , resp. . . . ) will be called U-sets (resp. U-simplicial
sets, resp. . . . ). We will use the expression U-small set (resp. U-small simplicial set,
resp. . . . ) to mean a set isomorphic to a set in U (resp. a simplicial set isomorphic
to a simplicial set in U, resp. . . . ). A unique exception concerns categories. The
expression U-category refers to the usual notion of [SGA4-I, IDef.1.2], and denotes
a category C such that for any two objects x and y in C the set HomC (x, y) is U-small.
In the same way, a category C is U-small is it is isomorphic to some element in U.
Our references for model categories are [Ho1] and [Hi]. By definition, our model
categories will always be closed model categories, will have all small limits and colimits
and the functorial factorization property. The word equivalence will always mean weak
equivalence and will refer to a model category structure.
The homotopy category of a model category M is W −1 M (see [Ho1, Def. 1.2.1]),
where W is the subcategory of equivalences in M , and it will be denoted as Ho(M ).
The sets of morphisms in Ho(M ) will be denoted by [−, −]M , or simply by [−, −]
when the reference to the model category M is clear. We will say that two objects
in a model category M are equivalent if they are isomorphic in Ho(M ). We say that
two model categories are Quillen equivalent if they can be connected by a finite string
of Quillen adjunctions each one being a Quillen equivalence. The mapping space
of morphisms between two objects x and y in a model category M is denoted by
M apM (x, y) (see [Ho1, §5]), or simply M ap(x, y) if the reference to M is clear. The
simplicial set depends on the choice of cofibrant and fibrant resolution functors, but
is well defined as an object in the homotopy category of simplicial sets Ho(SSet). If
M is a U-category, then M apM (x, y) is a U-small simplicial set.
The homotopy fiber product (see [Hi, 13.3, 19.5], [DHK, Ch. XIV] or [DS,
10] ) of a diagram x /zo y in a model category M will be denoted by
h /y
x ×z y. In the same way, the homotopy push-out of a diagram x o z
`L
will be denoted by x z y. For a pointed model category M , the suspension and loop
functors functor will be denoted by
S : Ho(M ) −→ Ho(M ) Ho(M ) ←− Ho(M ) : Ω.
`L
Recall that S(x) := ∗ x ∗, and Ω(x) := ∗ ×hx ∗.
When a model category M is a simplicial model category, its simplicial sets of
morphisms will be denoted by HomM (−, −), and their derived functors by RHomM
(see [Ho1, 1.3.2]), or simply Hom(−, −) and RHom(−, −) when the reference to M
is clear. When M is a symmetric monoidal model category in the sense of [Ho1, §4],
the derived monoidal structure will be denoted by ⊗L .
For the notions of U-cofibrantly generated, U-combinatorial and U-cellular model
category, we refer to [Ho1, Hi, Du2], or to [HAGI, Appendix], where the basic
definitions and crucial properties are recalled in a way that is suitable for our needs.
10 INTRODUCTION

As usual, the standard simplicial category will be denoted by ∆. The category

op
of simplicial objects in a category C will be denoted by sC := C ∆ . In the same way,
the category of co-simplicial objects in C will be denoted by csC. For any simplicial
object F ∈ sC in a category C, we will use the notation Fn := F ([n]). Similarly, for
any co-simplicial object F ∈ C ∆ , we will use the notation Fn := F ([n]). Moreover,
when C is a model category, we will use the notation
|X∗ | := Hocolim[n]∈∆op Xn
for any X∗ ∈ sC.
A sub-simplicial set K ⊂ L will be called full is K is a union of connected
components of L. We will also say that a morphism f : K −→ L of simplicial sets is
full if it induces an equivalence bewteen K and a full sub-simplicial set of L. In the
same way, we will use the expressions full sub-simplicial presheaf, and full morphisms
of simplicial presheaves for the levelwise extension of the above notions to presheaves
of simplicial sets.
For a Grothendieck site (C, τ ) in a universe U, we will denote by P r(C) the
op
category of presheaves of U-sets on C, P r(C) := C SetU . The subcategory of sheaves on
(C, τ ) will be denoted by Shτ (C), or simply by Sh(C) if the topology τ is unambiguous.
All complexes will be cochain complexes (i.e. with differential increasing the
degree by one) and therefore will look like
dn
... / En / E n+1 / ... / E0 / E1 / ....
The following notations concerning various homotopy categories of stacks are de-
fined in the main text, and recalled here for readers’ convenience (see also the Index
at the end of the book).

St(C, τ ) := Ho(Af fC∼,τ )

St(k) := Ho(k − Af f ∼,ét)

D− St(k) := Ho(k − D− Af f ∼,ét)

DSt(k) := Ho(k − DAf f ∼,s-ét )

St(S) := Ho(SAf f ∼,s-ét )

 Part 1

General theory of geometric stacks

 Introduction to Part 1

In this first part we will study the general theory of stacks and geometric stacks
over a base symmetric monoidal model category C. For this, we will start in §1.1
by introducing the notion of a homotopical algebraic context (HA context for short),
which consists of a triple (C, C0 , A) where C is our base monoidal model category, C0 is
a sub-category of C, and A is a sub-category of the category Comm(C) of commutative
monoids in C; we also require that the triple (C, C0 , A) satisfies certain compatibility
conditions. Although this might look like a rather unnatural and complicated defi-
nition, it will be shown in §1.2 that this data precisely allows us to define abstract
versions of standard notions such as derivations, unramified, étale, smooth and flat
morphisms. In other words a HA context describes an abstract context in which the
basic notions of linear and commutative algebra can be developed.
The first two chapters are only concerned with purely algebraic notions and the
geometry only starts in the third one, §1.3. We start by some reminders on the notions
of model topology and of model topos (developed in [HAGI]), which are homotopical
versions of the notions of Grothendieck topology and of Grothendieck topos and which
will be used all along this work. Next, we introduce a notion of a homotopical algebraic
geometry context (HAG context for short), consisting of a HA context together with
two additional data, τ and P, satsfying some compatiblity conditions. The first
datum τ is a model topology on Af fC , the opposite model category of commutative
monoids in C. The second datum P consists of a class of morphisms in Af fC which
behaves well with respect to τ . The model topology τ gives a category of stacks
over Af fC (a homotopical generalization of the category of sheaves on affine schemes)
in which everything is going to be embedded by means of a Yoneda lemma. The
class of morphisms P will then be used in order to define geometric stacks and more
generally n-geometric stacks, by considering successive quotient stacks of objects of
Af fC by action of groupoids whose structural morphisms are in P. The compatibility
axioms between τ and P will insure that this notion of geometricity behaves well, and
satisfies the basic expected properties (stability by homotopy pullbacks, gluing and
certain quotients).
In §1.4, the last chapter of part I, we will go more deeply into the study of geo-
metric stacks by introducing infinitesimal constructions such as derivations, cotangent
complexes and obstruction theories. The main result of this last chapter states that
any geometric stack has an obstruction theory (including a cotangent complex) as
soon as the HAG context satisfies suitable additional conditions.

13
 CHAPTER 1.1

Homotopical algebraic context

The purpose of this chapter is to fix once for all our base model category as well
as several general assumptions it should satisfy.

All along this chapter, we refer to [Ho1] for the general definition of monoidal
model categories, and to [Schw-Shi] for general results about monoids and modules
in monoidal model categories.

From now on, and all along this work, we fix three universes U ∈ V ∈ W (see, e.g.
[SGA4-I, Exp.I, Appendice]). We also let (C, ⊗, 1) be a symmetric monoidal model
category in the sense of [Ho1, §4]. We assume that C is a V-small category, and that
it is U-combinatorial in the sense of [HAGI, Appendix].

We make a first assumption on the base model category C, making it closer to an

additive category. Recall that we denote by Q a cofibrant replacement functor and
by R a fibrant replacement functor in M .

Assumption 1.1.0.1. (1) The model category C is proper, pointed (i.e. the
final object is also an initial object) and for any two object X and Y in C
the natural morphisms
a a
QX QY −→ X Y −→ RX × RY

are all equivalences.

(2) The homotopy category Ho(C) is an additive category.

Assumption 1.1.0.1 implies in particular that finite homotopy coproducts are also
finite homotopy products in C. It is always satisfied when C is furthermore a stable
model category in the sense of [Ho1, §7]. Note that 1.1.0.1 implies in particular that
for any two objects x and y in C, the set [x, y] has a natural abelian group structure.

As (C, ⊗, 1) is a symmetric monoidal category, which is closed and has U-small

limits and colimits, all the standard notions and constructions of linear algebra makes
sense in C (e.g. monoids, modules over monoids, operads, algebra over an operad . . . ).
The category of all associative, commutative and unital monoids in C will be denoted
by Comm(C). Objects of Comm(C) will simply be called commutative monoids in C,
or commutative monoids if C is clear. In the same way, one defines Commnu (C) to
be the category of non-unital commutative monoids in C. Therefore, our convention
will be that monoids are unital unless the contrary is specified.
The categories Comm(C) and Commnu (C) are again U-categories which are V-
small categories, and possess all U-small limits and colimits. They come equipped
with natural forgetful functors

Comm(C) −→ C Commnu (C) −→ C,

15
16 1.1. HOMOTOPICAL ALGEBRAIC CONTEXT

possessing left adjoints

F : C −→ Comm(C) Fnu : C −→ Commnu (C)
sending an object of C to the free commutative monoid it generates. We recall that
for X ∈ C one has a
F (X) = X ⊗n /Σn
n∈N
a
Fnu (X) = X ⊗n /Σn ,
n∈N−{0}
⊗n
where X is the n-tensor power of X, Σn acts on it by permuting the factors and
X ⊗n /Σn denotes the quotient of this action in C.

Let A ∈ Comm(C) be a commutative monoid. We will denote by A − M od the

category of unital left A-modules in C. The category A − M od is again a U-category
which is a V-small category, and has all U-small limits and colimits. The objects in
A − M od will simply be called A-modules. It comes equiped with a natural forgetful
functor
A − M od −→ C,
possessing a left adjoint
A ⊗ − : C −→ A − M od
sending an object of C to the free A-module it generates. We also recall that the
category A − M od has a natural symmetric monoidal structure − ⊗A −. For two
A-modules X and Y , the object X ⊗A Y is defined as the coequalizer in C of the two
natural morphisms
X ⊗ A ⊗ Y −→ X ⊗ Y X ⊗ A ⊗ Y −→ X ⊗ Y.
This symmetric monoidal structure is furthermore closed, and for two A-modules X
and Y we will denoted by HomA (X, Y ) the A-module of morphisms. One has the
usual adjunction isomorphisms
Hom(X ⊗A Y, Z) ≃ Hom(X, HomA (Y, Z)),
as well as isomorphisms of A-modules
HomA (X ⊗A Y, Z) ≃ Hom(X, HomA (Y, Z)).
We define a morphism in A − M od to be a fibration or an equivalence if it is so
on the underlying objects in C.
Assumption 1.1.0.2. Let A ∈ Comm(C) be any commutative monoid in C. Then,
the above notions of equivalences and fibrations makes A−M od into a U-combinatorial
proper model category. The monoidal structure − ⊗A − makes furthermore A − M od
into a symmetric monoidal model category in the sense of [Ho1, §4].
Using the assumption 1.1.0.2 one sees that the homotopy category Ho(A − M od)
has a natural symmetric monoidal structure ⊗LA , and derived internal Hom′ s associ-
ated to it RHomA , satisfying the usual adjunction rule
[X ⊗LA Y, Z] ≃ [X, RHomA (Y, Z)].
Assumption 1.1.0.3. Let A be a commutative monoid in C. For any cofibrant
object M ∈ A − M od, the functor
− ⊗A M : A − M od −→ A − M od
preserves equivalences.
 1.1. HOMOTOPICAL ALGEBRAIC CONTEXT 17

Let us still denote by A a commutative monoid in C. We have categories of

commutative monoids in A − M od, and non-unital commutative monoids in A − M od,
denoted respectively by A−Comm(C) and A−Commnu (C), and whose objects will be
called commutative A-algebras and non-unital commutative A-algebras. They come
equipped with natural forgetful functors
A − Comm(C) −→ A − M od A − Commnu (C) −→ A − M od,
possessing left adjoints
FA : A − M od −→ A − Comm(C) FAnu : A − M od −→ A − Commnu (C)
sending an object of A − M od to the free commutative monoid it generates. We recall
that for X ∈ A − M od one has
a
FA (X) = X ⊗A n /Σn
n∈N
a
FAnu (X) = X ⊗A n /Σn ,
n∈N−{0}
⊗A n
where X is the n-tensor power of X in A − M od, Σn acts on it by permuting the
factors and X ⊗A n /Σn denotes the quotient of this action in A − M od.
Finally, we define a morphism in A − Comm(C) or in A − Commnu (C) to be
a fibration (resp. an equivalence) if it is so as a morphism in the category C (or
equivalently as a morphism in A − M od).
Assumption 1.1.0.4. Let A be any commutative monoid in C.
(1) The above classes of equivalences and fibrations make the categories A −
Comm(C) and A − Commnu (C) into U-combinatorial proper model cate-
gories.
(2) If B is a cofibrant object in A − Comm(C), then the functor
B ⊗A − : A − M od −→ B − M od
preserves equivalences.
Remark 1.1.0.5. One word concerning non-unital algebras. We will not really
use this notion in the sequel, except at one point in order to prove the existence of
a cotangent complex (so the reader is essentially allowed to forget about this unfre-
quently used notion). In fact, by our assumptions, the model category of non-unital
commutative A-algebras is Quillen equivalent to the model category of augmented
commutative A-algebra. However, the categories themselves are not equivalent, since
the category C is not assumed to be strictly speaking additive, but only additive up
to homotopy (e.g. it could be the model category of symmetric spectra of [HSS]).
Therefore, we do not think that the existence of the model structure on A−Comm(C)
implies the existence of the model structure on A − Commnu (C); this explains why
we had to add condition (1) on A − Commnu (C) in Assumption 1.1.0.4. Furthermore,
the passage from augmented A-algebras to non-unital A-algebras will be in any case
necessary to construct a certain Quillen adjunction during the proof of Prop. 1.2.1.2,
because such a Quillen adjunction does not exist from the model category of aug-
mented A-algebras (as it is a composition of a left Quillen functor by a right Quillen
equivalence).
An important consequence of assumption 1.1.0.4 (2) is that for A a commutative
monoid in C, and B, B ′ two commutative A-algebras, the natural morphism in Ho(A−
18 1.1. HOMOTOPICAL ALGEBRAIC CONTEXT

M od)
L
a
B B ′ −→ B ⊗LA B ′
A
is an isomorphism (here the object on the left is the homotopy coproduct in A −
Comm(C), and the one on the right is the derived tensor product in A − M od).
An important remark we will use implicitly very often in this paper is that the
category A − Comm(C) is naturally equivalent to the comma category A/Comm(C),
of objects under A. Moreover, the model structure on A−Comm(C) coincides through
this equivalence with the comma model category A/Comm(C).

We will also fix a full subcategory C0 of C, playing essentially the role of a kind
of “t-structure” on C (i.e. essentially defining which are the “non-positively graded
objects”, keeping in mind that in this work we use the cohomological grading when
concerned with complexes). More precisely, we will fix a subcategory C0 ⊆ C satisfying
the following conditions.
Assumption 1.1.0.6. (1) 1 ∈ C0 .
(2) The full subcategory C0 of C is stable by equivalences and by U-small homo-
topy colimits.
(3) The full subcategory Ho(C0 ) of Ho(C) is stable by the monoidal structure
− ⊗L − (i.e. for X and Y in Ho(C0 ) we have X ⊗L Y ∈ Ho(C0 )).
Recall that as C is a pointed model category one can define its suspension functor
S : Ho(C) −→ Ho(C)
`L
x 7→ ∗ x ∗
left adjoint to the loop functor
Ω : Ho(C) −→ Ho(C)
x 7→ := ∗ ×hx ∗.
We set C1 to be the full subcategory of C consisting of all objects equivalent to the
suspension of some object in C0 . The full subcategory C1 of C is also closed by
equivalences, homotopy colimits and the derived tensor structure. We will denote
by Comm(C)0 the full subcategory of Comm(C) consisting of commutative monoids
whose underlying C-object lies in C0 . In the same way, for A ∈ Comm(C) we denote by
A − M od0 (resp. A − M od1 , resp. A − Comm(C)0 ) the full subcategory of A − M od
consisting of A-modules whose underlying C-object lies in C0 (resp. of A − M od
consisting of A-modules whose underlying C-object lies in C1 , resp. of A − Comm(C)
consisting of commutative A-algebras whose underlying C-object lies in C0 ).
An important consequence of Assumption 1.1.0.6 is that for any morphism A −→
B in Comm(C)0 , and any M ∈ A− M od0 , we have B ⊗LA M ∈ Ho(B − M od0 ). Indeed,
any such A-module can be written as a homotopy colimit of A-modules of the form
L
A⊗ n ⊗L M , for which we have
L L
B ⊗LA A⊗ n
⊗L M ≃ A⊗ (n−1)
⊗L M.
In particular the full subcategory Comm(C)0 is closed by homotopy push-outs in
Comm(C). Passing to the suspension we also see that for any M ∈ A − M od1 , one
also has B ⊗LA M ∈ Ho(B − M od1 ).
Remark 1.1.0.7. (1) The reason for introducing the subcategory C0 is to
be able to consider reasonable infinitesimal lifting properties; these infini-
tesimal lifting properties will be used to develop the abstract obstruction
 1.1. HOMOTOPICAL ALGEBRAIC CONTEXT 19

theory of geometric stacks in §1.4. It is useful to keep in mind that C0 plays

a role analogous to a kind of t-structure on C, in that it morally defines
which are the non-positively graded objects (a typical example will appear
in §2.3 where C will be the model category of unbounded complexes and C0
the subcategory of complexes with vanishing positive cohomology). Differ-
ent choices of C0 will then give different notions of formal smoothness (see
§1.2.8), and thus possibly different notions of geometric stacks. We think
that playing with C0 as a degree of freedom is an interesting feature of our
abstract infinitesimal theory.
(2) Assumptions 1.1.0.1, 1.1.0.2 and 1.1.0.3 are not really serious, and are only
useful to avoid taking too many fibrant and cofibrant replacements. With
some care, they can be omitted. On the other hand, the careful reader
will probably be surprised by assumption 1.1.0.4, as it is known not to be
satisfied in several interesting examples (e.g. when C is the model category of
complexes over some commutative ring k, not of characteristic zero). Also,
it is well known that in some situations the notion of commutative monoid is
too strict and it is often preferable to use the weaker notion of E∞ -monoid.
Two reasons has led us to assume 1.1.0.4. First of all, for all contexts of
application of the general theory we will present in this work, there is always
a base model category C for which this condition is satisfied and gives rise
to the correct theory. Moreover, if one replaces commutative monoids by
E∞ -monoids then assumption 1.1.0.4 is almost always satisfied, as shown in
[Sp], and we think that translating our general constructions should then
be a rather academic exercise. Working with commutative monoids instead
of E∞ -monoids simplifies a lot the notations and certain constructions, and
in our opinion this theory already captures the real essence of the subject.
Finally, in partial defense of our choice, let us also mention that contrary
to what one could think at first sight, working with E∞ -monoids would not
strictly speaking increase the degree of generality of the theory. Indeed, one
of our major application is to the category of simplicial k-modules, whose
category of commutative monoids is the category of simplicial commutative
k-algebras. However, if k has non-zero characteristic, the homotopy the-
ory of simplicial commutative k-algebras is not equivalent to the homotopy
theory of E∞ -monoids in simplicial k-modules (the latter is equivalent to
the homotopy theory of E∞ -k-algebras in non positive degrees). Therefore,
using E∞ -monoids throughout would prevent us from developing derived
algebraic geometry as presented in §2.2.

We list below some important examples of symmetric monoidal model categories C

satisfying the four above assumptions, and of crucial importance for our applications.
(1) Let k be any commutative ring, and C = k − M od be the category of U-
k-modules, symmetric monoidal for the tensor product ⊗k , and endowed
with its trivial model structure (i.e. equivalences are isomorphisms and
all morphisms are fibrations and cofibrations). Then assumptions 1.1.0.1,
1.1.0.2, 1.1.0.3 and 1.1.0.4 are satisfied. For C0 one can take the whole C for
which Assumption 1.1.0.6 is clearly satisfied.
(2) Let k be a commutative ring of characteristic 0, and C = C(k) be the
category of (unbounded) complexes of U-k-modules, symmetric monoidal
for the tensor product of complexes ⊗k , and endowed with its projective
model structure for which equivalences are quasi-isomorphisms and fibra-
tions are epimorphisms (see [Ho1]). Then, the category Comm(C) is the
20 1.1. HOMOTOPICAL ALGEBRAIC CONTEXT

category of commutative differential graded k-algebras belonging to U. For

A ∈ Comm(C), the category A − M od is then the category of differen-
tial graded A-modules. It is well known that as k is of characteristic zero
then assumptions 1.1.0.1, 1.1.0.2, 1.1.0.3 and 1.1.0.4 are satisfied (see e.g;
[Hin1]). For C0 one can take either the whole C, or the full subcategory of
C consisting of complexes E such that H i (E) = 0 for any i > 0, for which
Assumption 1.1.0.6 is satisfied.
A similar example is given by non-positively graded complexes C =
C − (k).
(3) Let k be any commutative ring, and C = sM odk be the category of U-
simplicial k-modules, endowed with the levelwise tensor product and the
usual model structure for which equivalences and fibrations are defined on
the underlying simplicial sets (see e.g. [Goe-Ja]). The category Comm(C) is
then the category of simplicial commutative k-algebras, and for A ∈ sM odk ,
A − M od is the category of simplicial modules over the simplicial ring A.
Assumptions 1.1.0.1, 1.1.0.2, 1.1.0.3 and 1.1.0.4 are again well known to be
satisfied. For C0 one can take the whole C.
Dually, one could also let C = csM odk be the category of co-simplicial
k-modules, and our assumptions would again be satisfied. In this case, C0
could be for example the full subcategory of co-simplicial modules E such
that π−i (E) = H i (E) = 0 for any i > 0. This subcategory is stable under
homotopy colimits as the functor E 7→ H 0 (E) is right Quillen and right
adjoint to the inclusion functor k − M od −→ csM odk .
(4) Let SpΣ be the category of U-symmetric spectra and its smash product,
endowed with the positive stable model structure of [Shi]. Then, Comm(C)
is the category of commutative symmetric ring spectra, and assumptions
1.1.0.1, 1.1.0.2, 1.1.0.3 and 1.1.0.4 are known to be satisfied (see [Shi, Thm.
3.1, Thm. 3.2, Cor. 4.3]). The two canonical choices for C0 are the whole C
or the full subcategory of connective spectra.
It is important to note that one can also take for C the category of Hk-
modules in SpΣ for some commutative ring k. This will give a model for the
homotopy theory of E∞ -k-algebras that were not provided by our example
2.
(5) Finally, the above three examples can be sheafified over some Grothendieck
site, giving the corresponding relative theories over a base Grothendieck
topos.
In few words, let S be a U-small Grothendieck site, and C be one the
three symmetric monoidal model category C(k), sM odk , SpΣ discussed
above. One considers the corresponding categories of presheaves on S,
P r(S, C(k)), P r(S, sM odk ), P r(S, SpΣ ). They can be endowed with the
projective model structures for which fibrations and equivalences are de-
fined levelwise. This first model structure does not depend on the topology
on S, and will be called the strong model structure: its (co)fibrations and
equivalences will be called global (co)fibrations and global equivalences.
The next step is to introduce notions of local equivalences in the model
categories P r(S, C(k)), P r(S, sM odk ), P r(S, SpΣ ). This notion is defined
by first defining reasonable homotopy sheaves, as done for the notion of local
equivalences in the theory of simplicial presheaves, and then define a mor-
phism to be a local equivalence if it induces isomorphisms on all homotopy
sheaves (for various choice of base points, see [Jo1, Ja1] for more details).
The final model structures on P r(S, C(k)), P r(S, sM odk ), P r(S, SpΣ ) are
 1.1. HOMOTOPICAL ALGEBRAIC CONTEXT 21

the one for which equivalences are the local equivalences, and cofibrations
are the global cofibrations. The proof that this indeed defines model cate-
gories is not given here and is very similar to the proof of the existence of
the local projective model structure on the category of simplicial presheaves
(see for example [HAGI]).
Finally, the symmetric monoidal structures on the categories C(k), sM odk
and SpΣ induces natural symmetric monoidal structures on the categories
P r(S, C(k)), P r(S, sM odk ), P r(S, SpΣ ). These symmetric monoidal struc-
tures make them into symmetric monoidal model categories when S has
finite products. One can also check that the symmetric monoidal model
categories P r(S, C(k)), P r(S, sM odk ), P r(S, SpΣ ) constructed that way all
satisfy the assumptions 1.1.0.1, 1.1.0.2, 1.1.0.3 and 1.1.0.4.
An important example is the following. Let O be a sheaf of commutative
rings on the site S, and let HO ∈ P r(S, SpΣ ) be the presheaf of symmetric
spectra it defines. The object HO is a commutative monoid in P r(S, SpΣ )
and one can therefore consider the model category HO − M od, of HO-
modules. The category HO − M od is a symmetric monoidal model category
and its homotopy category is equivalent to the unbounded derived category
D(S, O) of O-modules on the site S. This gives a way to define all the
standard constructions as derived tensor products, derived internal Hom′ s
etc., in the context of unbounded complexes of O-modules.
Let f : A −→ B be a morphism of commutative monoids in C. We deduce an
adjunction between the categories of modules
f ∗ : A − M od −→ B − M od A − M od ←− B − M od : f∗ ,
∗
where f (M ) := B ⊗A M , and f∗ is the forgetful functor that sees a B-module as an
A-module through the morphism f . Assumption 1.1.0.2 tells us that this adjunction
is a Quillen adjunction, and assumption 1.1.0.3 implies it is furthermore a Quillen
equivalence when f is an equivalence (this is one of the main reasons for assumption
1.1.0.3).
The morphism f induces a pair of adjoint derived functors
Lf ∗ : Ho(A−M od) −→ Ho(B−M od) Ho(A−M od) ←− Ho(B−M od) : Rf∗ ≃ f∗ ,
and, as usual, we will also use the notation
Lf ∗ (M ) =: B ⊗LA M ∈ Ho(B − M od).
Finally, let
f
A /B
p p′
 
A′ / B′
f′

be a homotopy cofiber square in Comm(C). Then, for any A′ -module M we have the
well known base change morphism
Lf ∗ p∗ (M ) −→ (p′ )∗ L(f ′ )∗ (M ).
Proposition 1.1.0.8. Let us keep the notations as above. Then, the morphism
Lf ∗ p∗ (M ) −→ (p′ )∗ L(f ′ )∗ (M )
is an isomorphism in Ho(B − M od) for any A′ -module M .
22 1.1. HOMOTOPICAL ALGEBRAIC CONTEXT

Proof. As the homotopy categories of modules are invariant under equivalences

of commutative monoids, one can suppose that the diagram
f
A /B
p p′
 
A′ / B′
f′

is cocartesian in Comm(C), and consists of cofibrations in Comm(C). Then, using

1.1.0.3 and 1.1.0.4 (2) one sees that the natural morphisms
M ⊗LA B −→ M ⊗A B M ⊗LA′ B ′ −→ M ⊗A′ B ′ ,
are isomorphism in Ho(B −M od). Therefore, the proposition follows from the natural
isomorphism of B-modules
M ⊗A′ B ′ ≃ M ⊗A′ (A′ ⊗A B) ≃ M ⊗A B.

Remark 1.1.0.9. The above base change formula will be extremely important in
the sequel, and most often used implicitly. It should be noticed that it implies that
the homotopy coproduct in the model category of commutative monoids is given by
the derived tensor product. This last property is only satisfied because we have used
commutative monoids, and is clearly wrong for simply associative monoids. This is
one major reason why our setting cannot be used, at least without some modifications,
to develop truly non-commutative geometries. Even partially commutative structures,
like En -monoids for n > 1 would not satisfy the base change formula, and one really
needs E∞ -monoids at least. This fact also prevents us to generalize our setting by
replacing the category of commutative monoids by more general categories, like some
category of algebras over more general operads.
We now consider A ∈ Comm(C) and the natural inclusion A−M od0 −→ A−M od.
We consider the restricted Yoneda embedding
op ∧
Rh− op
0 : Ho(A − M od ) −→ Ho((A − M od0 ) ),

sending an A-module M to the functor

M ap(M, −) : A − M odop
0 −→ SSetV .

We recall here from [HAGI, §4.1] that for a model category M , and a full sub-
category stable by equivalences M0 ⊂ M , M0∧ is the left Bousfield localization of
SP r(M0 ) along equivalences in M0 .
Definition 1.1.0.10. We will say that A ∈ Comm(C) is good with respect to C0
(or simply C0 -good) if the functor
Ho(A − M odop ) −→ Ho((A − M odop ∧
0 ) )

is fully faithful.
In usual category theory, a full subcategory D ⊂ C is called dense if the restricted
Yoneda functor C −→ P r(D) := Hom(Dop , Ens) is fully faithful (in [SGA4-I] this
notion is equivalent to the fact that D generates C through strict epimorphisms).
This implies for example that any object of C is the colimit of objects of D, but is
a slightly stronger condition because any object x ∈ C is in fact isomorphic to the
colimit of the canonical diagram D/x −→ C (see e.g. [SGA4-I, ExpI-Prop. 7.2]).
 1.1. HOMOTOPICAL ALGEBRAIC CONTEXT 23

Our notion of being good (Def. 1.1.0.10) essentially means that A − M odop
0 is homo-
topically dense in A − M odop . This of course implies that any object in A − M od
is equivalent to a homotopy limit of objects in A − M od0 , and is equivalent to the
fact that any cofibrant object M ∈ A − M od is equivalent to the homotopy limit of
the natural diagram (M/A − M od0 )c −→ A − M od, where (M/A − M od0 )c denotes
the category of cofibrations under M . Dually, one could say that A being good with
respect to C0 means that A − M od0 cogenerates A − M od through strict monomor-
phisms in a homotopical sense.

We finish this first chapter by the following definition, gathering our assumptions
1.1.0.1, 1.1.0.2, 1.1.0.3, 1.1.0.4 and 1.1.0.6 all together.
Definition 1.1.0.11. A Homotopical Algebraic context (or simply HA context)
is a triplet (C, C0 , A), consisting of a symmetric monoidal model category C, two full
sub-categories stable by equivalences
C0 ⊂ C A ⊂ Comm(C),
such that any A ∈ A is C0 -good, and assumptions 1.1.0.1, 1.1.0.2, 1.1.0.3, 1.1.0.4,
1.1.0.6 are satisfied.
 CHAPTER 1.2

Preliminaries on linear and commutative algebra in

an HA context

All along this chapter we fix once for all a HA context (C, C0 , A), in the sense
of Def. 1.1.0.11. The purpose of this chapter is to show that the assumptions of
the last chapter imply that many general notions of linear and commutative algebra
generalize in some reasonable sense in our base category C.

1.2.1. Derivations and the cotangent complex

This section is nothing else than a rewriting of the first pages of [Ba], which stay
valid in our general context.

Let A ∈ Comm(C) be a commutative monoid in C, and M be an A-module.

We define a new commutative monoid` A ⊕ M in the following way. The underlying
object of A ⊕ M is the coproduct A M . The multiplicative structure is defined by
the morphism
a a a a a a
(A M ) ⊗ (A M) ≃ A ⊗ A A⊗M A⊗M M ⊗ M −→ A M
given by the three morphisms
a a
µ ∗ : A ⊗ A −→ A M
a a
∗ ρ : A ⊗ M −→ A M
∗ : M ⊗ M −→ M,
where µ : A⊗A −→ A is the multiplicative structure of A, and ρ : A⊗M −→ M is the
module structure of M . The monoid A⊕M is commutative and unital, and defines an
object in Comm(C).
` It comes furthermore with a natural morphism
` of commutative
monoids id ∗ : A ⊕ M −→ A, which has a natural section id ∗ : A −→ A ⊕ M .
Now, if A −→ B is a morphism in Comm(C), and M is a B-module, the morphism
B ⊕ M −→ B can be seen as a morphism of commutative A-algebras. In other words,
B⊕M can be seen as an object of the double comma model category A−Comm(C)/B.
Definition 1.2.1.1. Let A −→ B be a morphism of commutative monoids, and
M be a B-module. The simplicial set of derived A-derivations from B to M , is the
object
DerA (B, M ) := M apA−Comm(C)/B (B, B ⊕ M ) ∈ Ho(SSetU ).
Clearly, M 7→ DerA (B, M ) defines a functor from the homotopy category of B-
module Ho(B − M od) to the homotopy category of simplicial sets Ho(SSet). More
precisely, the functoriality of the construction of mapping spaces implies that one can
also construct a genuine functor
DerA (B, −) : B − M od −→ SSetU ,
25
26 1.2. PRELIMINARIES ON LINEAR AND COMMUTATIVE ALGEBRA IN AN HA CONTEXT

lifting the previous functor on the homotopy categories. This last functor will be
considered as an object in the model category of pre-stacks (B − M odop )∧ as defined
in [HAGI, §4.1]. Recall from [HAGI, §4.2] that there exists a Yoneda embedding
Ho(B − M od)op −→ Ho((B − M odop )∧ )
sending a B-module M to the simplicial presheaf N 7→ M apB−Mod (M, N ), and ob-
jects in the essential image will be called co-representable.
Proposition 1.2.1.2. For any morphism A −→ B in Comm(C), there exists a
B-module LB/A , and an element d ∈ π0 (DerA (B, LB/A )), such that for any B-module
M , the natural morphism obtained by composing with d
d∗ : M apB−Mod (LB/A , M ) −→ DerA (B, M )
is an isomorphism in Ho(SSet).
Proof. The proof is the same as in [Ba], and uses our assumptions 1.1.0.1,
1.1.0.2, 1.1.0.3 and 1.1.0.4. We will reproduce it for the reader convenience.

We first consider the Quillen adjunction

−⊗A B : A−Comm(C)/B −→ B−Comm(C)/B A−Comm(C)/B ←− B−Comm(C)/B : F,
where F is the forgetful functor. This induces an adjunction on the level of homotopy
categories
− ⊗LA B : Ho(A − Comm(C)/B) −→ Ho(B − Comm(C)/B)

Ho(A − Comm(C)/B) ←− Ho(B − Comm(C)/B) : F.

We consider a second Quillen adjunction
K : B−Commnu (C) −→ B−Comm(C)/B B−Commnu (C) ←− B−Comm(C)/B : I,
where B − Commnu (C) is the category of non-unital commutative B-algebras (i.e.
non-unital commutative monoids in B − M od). The functor I takes a diagram of
s / p
commutative monoids B C / B to the kernel of p computed in the cat-
egory of non-unital commutative B-algebras. In the other direction, the functor K
takes a non-unital commutative B-algebra C to the trivial extension of B by C (de-
fined as our B ⊕ M but taking into account the multiplication on C). Clearly, I is
a right Quillen functor, and the adjunction defines an adjunction on the homotopy
categories
LK : Ho(B − Commnu (C)) −→ Ho(B − Comm(C)/B)
Ho(B − Commnu (C)) ←− Ho(B − Comm(C)/B) : RI.
Lemma 1.2.1.3. The adjunction (LK, RI) is an equivalence.
Proof. This follows easily from our assumption 1.1.0.1. Indeed, it implies that
for any fibration in C, f : X −→ Y , which has a section s : Y −→ X, the natural
morphism
a a
i s:F Y −→ X,
where i : F −→ X is the fiber of f , is `an equivalence.
` It also implies that the
homotopy fiber of the natural morphism id ∗ : X Y −→ X is naturally equivalent
to Y . These two facts imply the lemma. 
 1.2.1. DERIVATIONS AND THE COTANGENT COMPLEX 27

Finally, we consider a third adjunction

Q : B − Commnu (C) −→ B − M od B − Commnu (C) ←− B − M od : Z,
where Q of an object C ∈ B − Commnu (C) is the push-out of B-modules
µ
C ⊗B C /C

 
• / Q(C),

and Z sends a B-module M to the non-unital B-algebra M endowed with the zero
multiplication. Clearly, (Q, Z) is a Quillen adjunction and gives rise to an adjunction
on the homotopy categories
LQ : Ho(B − Commnu (C)) −→ Ho(B − M od)
Ho(B − Commnu (C)) ←− Ho(B − M od) : Z.
We can now conclude the proof of the proposition by chaining up the various adjunc-
tion to get a string of isomorphisms in Ho(SSet)
DerA (B, M ) ≃ M apA−Comm(C)/B (B, F (B⊕M )) ≃ M apB−Comm(C)/B (B⊗LA B, B⊕M )
≃ M apB−Commnu(C) (RI(B⊗LA B), RI(B⊕M )) ≃ M apB−Commnu(C) (RI(B⊗LA B), Z(M ))
≃ M apB−Mod (LQRI(B ⊗LA B), M ).
Therefore, LB/A := LQRI(B ⊗LA B) and the image of id ∈ M apB−Mod (LB/A , LB/A )
gives what we were looking for. 
Remark 1.2.1.4. Proposition 1.2.1.2 implies that the two functors
M 7→ M apB−Mod (LB/A , M ) M 7→ DerA (B, M )
are isomorphic as objects in Ho((B − M odop )∧ ). In other words, Prop. 1.2.1.2 implies
that the functor DerA (B, −) is co-representable in the sense of [HAGI].
Definition 1.2.1.5. Let A −→ B be a morphism in Comm(C).
(1) The B-module LB/A ∈ Ho(B − M od) is called the cotangent complex of B
over A.
(2) When A = 1, we will use the following notation
LB := LB/1 ,
and LB will be called the cotangent complex of B.
Using the definition and proposition 1.2.1.2, it is easy to check the following facts.
Proposition 1.2.1.6. (1) Let A −→ B −→ C be two morphisms in Comm(C).
Then, there is a homotopy cofiber sequence in C − M od
LB/A ⊗LB C −→ LC/A −→ LC/B .
(2) Let
A /B

 
A′ / B′
be a homotopy cofiber square in Comm(C). Then, the natural morphism
LB/A ⊗LB B ′ −→ LB ′ /A′
28 1.2. PRELIMINARIES ON LINEAR AND COMMUTATIVE ALGEBRA IN AN HA CONTEXT

is an isomorphism in Ho(B ′ − M od). Furthermore, the natural morphism

a
LB/A ⊗LB B ′ LA′ /A ⊗LA′ B ′ −→ LB ′ /A

is an isomorphism in Ho(B ′ − M od).

(3) Let
A /B

 
A′ / B′
be a homotopy cofiber square in Comm(C). Then the following square is
homotopy cocartesian in B ′ − M od

LA ⊗LA B ′ / LB ⊗L B ′
B

 
LA′ ⊗LA′ B ′ / LB ′ .

(4) For any commutative monoid A and any A-module M , one has a natural
isomorphism in Ho(A − M od)
a
LA⊕M ⊗LA⊕M A ≃ LA LQZ(M ),
where
Q : A − Commnu (C) −→ A − M od A − Commnu (C) ←− A − M od : Z
is the Quillen adjunction used during the proof of 1.2.1.2.
Proof. (1) to (3) are simple exercises, using the definitions and that for any
morphism of commutative monoids A −→ B, and any B-module M , the following
square is homotopy cartesian in Comm(C) (because of our assumption 1.1.0.1)
A⊕M / B⊕M

 
A / B.

(4) We note that for any commutative monoid A, and any A-modules M and N ,
one has a natural homotopy fiber sequence
M apA−Comm(C)/A(A ⊕ M, A ⊕ N ) −→ M apComm(C)/A (A ⊕ M, A ⊕ N ) →

−→ M apComm(C)/A (A, A ⊕ N ),
or equivalently using lemma 1.2.1.3
M apA−Mod (LQZ(M ), N ) −→ Der(A ⊕ M, N ) −→ Der(A, N ).
This implies the existence of a natural homotopy cofiber sequence of A-modules
LA −→ LA⊕M ⊗LA⊕M A −→ LQZ(M ).
Clearly this sequence splits in Ho(A − M od) and gives rise to a natural isomorphism
a
LA⊕M ⊗LA⊕M A ≃ LA LQZ(M ).

 1.2.2. HOCHSCHILD HOMOLOGY 29

The importance of derivations come from the fact that they give rise to infinites-
imal extensions in the following way. Let A −→ B be a morphism of commutative
monoids in C, and M be a B-module. Let d : LB/A −→ M be a morphism in
Ho(B − M od), corresponding to a derivation d ∈ π0 (DerA (B, M )). This derivation
can be seen as a section d : B −→ B ⊕ M of the morphism of commutative A-algebras
B ⊕ M −→ B. We consider the following homotopy cartesian diagram in the category
of commutative A-algebras
C /B

d
 
B / B⊕M
s
where s : B −→ B ⊕ M is the natural section corresponding to the zero morphism
LB/A −→ M . Then, C −→ B is a morphism of commutative A-algebras such that
its fiber is a non-unital commutative A-algebra isomorphic in Ho(A − Commnu (C))
to the loop A-module ΩM := ∗ ×hM ∗ with the zero multiplication. In other words, C
is a square zero extension of B by ΩM . It will be denoted by B ⊕d ΩM . The most
important case is of course when A = B, and we make the following definition.
Definition 1.2.1.7. Let A be a commutative monoid, M and A-module and
d ∈ π0 Der(A, M ) be a derivation given by a morphism in d : A −→ A ⊕ M in
Ho(Comm(C)/A). The square zero extension associated to d, denoted by A ⊕d ΩM ,
is defined as the homotopy pullback diagram of commutative monoids
A ⊕d ΩM /A

d
 
A / A ⊕ M,
s

where s is the natural morphism corresponding to the zero derivation. The top hori-
zontal morphism A ⊕d ΩM −→ A will be called the natural projection.

1.2.2. Hochschild homology

For a commutative monoid A ∈ Comm(C), we set
T HH(A) := S 1 ⊗L A ∈ Ho(Comm(C)),
where S 1 ⊗L denotes the derived external product of object in Comm(C) by the sim-
plicial circle S 1 := ∆1 /∂∆1 . Presenting the circle S 1 has the homotopy push-out
L
a
∗ ∗
`
∗ ∗

one gets that

T HH(A) ≃ A ⊗LA⊗L A A.
The natural point ∗ −→ S 1 induces a natural morphism in Ho(Comm(C))
A −→ T HH(A)
making T HH(A) as a commutative A-algebra, and as a natural object in Ho(A −
Comm(C)).
Definition 1.2.2.1. Let A be a commutative monoid in C. The topological
Hochschild homology of A (or simply Hochschild homology) is the commutative A-
algebra T HH(A) := S 1 ⊗L A.
30 1.2. PRELIMINARIES ON LINEAR AND COMMUTATIVE ALGEBRA IN AN HA CONTEXT

More generally, if A −→ B is a morphism of commutative monoids in C, the

relative topological Hochschild homology of B over A (or simply relative Hochschild
homology) is the commutative A-algebra
T HH(B/A) := T HH(B) ⊗LT HH(A) A.
By definition, we have for any commutative monoid B,
M apComm(C)(T HH(A), B) ≃ M apSSet (S 1 , M apComm(C)(A, B)).
This implies that if f : A −→ B is a morphism of commutative monoids in C, then
we have
M apA−Comm(C) (T HH(A), B) ≃ Ωf M apComm(C) (A, B),
where Ωf M apComm(C) (A, B) is the loop space of M apComm(C) (A, B) at the point f .
More generally, if B and C are commutative A-algebras, then
M apA−Comm(C)(T HH(B/A), C) ≃ M apSSet (S 1 , M apA−Comm(C) (B, C)),
and for a morphism f : B −→ C of commutative A-algebras
M apB−Comm(C) (T HH(B/A), C) ≃ Ωf M apA−Comm(C)(B, C).
Proposition 1.2.2.2. (1) Let A −→ B −→ C be two morphisms in Comm(C).
Then, the natural morphism
T HH(C/A) ⊗LT HH(B/A) B −→ T HH(C/B)
is an isomorphism in Ho(B − Comm(C)).
(2) Let
A /B

 
A′ / B′
be a homotopy cofiber square in Comm(C). Then, the natural morphism
T HH(B/A) ⊗LA T HH(A′ /A) −→ T HH(B ′ /A)
is an isomorphism in Ho(A − Comm(C)).
Proof. Exercise. 

1.2.3. Finiteness conditions

We present here two different finiteness conditions for objects in model categories.
The first one is valid in any model category, and is a homotopy analog of the notion
of finitely presented object in a category. The second one is only valid for symmetric
monoidal categories, and is a homotopy generalization of the notion of rigid objects
in monoidal categories.
Definition 1.2.3.1. A morphism x −→ y in a proper model category M is finitely
presented (we also say that y is finitely presented over x) if for any filtered diagram
of objects under x, {zi }i∈I ∈ x/M , the natural morphism
Hocolimi∈I M apx/M (y, zi ) −→ M apx/M (y, Hocolimi∈I zi )
is an isomorphism in Ho(SSet).
Remark 1.2.3.2. If the model category M is not proper the definition 1.2.3.1 has
to be modified by replacing M apx/M with M apQx/M , where Qx is a cofibrant model
for x. By our assumption 1.1.0.1 all the model categories we will use are proper.
Proposition 1.2.3.3. Let M be a proper model category.
 1.2.3. FINITENESS CONDITIONS 31

(1) Finitely presented morphisms in M are stable by equivalences. In other

words, if one has a commutative diagram in M
f
x /y
p q
 
x′ / y′
f′

such that p and q are equivalences, then f is finitely presented if and only if
f ′ is finitely presented.
(2) Finitely presented morphisms in M are stable by compositions and retracts.
(3) Finitely presented morphisms in M are stable by homotopy push-outs. In
other words, if one has a homotopy push-out diagram in M
f
x /y
p q
 
x′ / y′
f′

then f ′ is finitely presented if f is so.

Proof. (1) is clear as M apx/M (a, b) only depends on the isomorphism class of a
and b as objects in the homotopy category Ho(x/M ).
(2) Let x −→ y −→ z be two finitely presented morphisms in M , and let {zi }i∈I ∈
x/M be a filtered diagrams of objects. Then, one has for any object t ∈ x/M a
fibration sequence of simplicial sets

M apy/M (z, t) −→ M apx/M (z, t) −→ M apx/M (y, t).

As fibration sequences are stable by filtered homotopy colimits, one gets a morphism
of fibration sequences

Hocolimi∈I Mapy/M (z, zi ) / Hocolimi∈I Mapx/M (z, zi ) / Hocolimi∈I Mapx/M (y, zi )

  
Mapy/M (z, Hocolimi∈I zi ) / Mapx/M (z, Hocolimi∈I zi ) / Mapx/M (y, Hocolimi∈I zi ),

and the five lemma tells us that the vertical arrow in the middle is an isomorphism
in Ho(SSet). This implies that z is finitely presented over x.
The assertion concerning retracts is clear since, if x −→ y is a retract of x′ −→ y ′ ,
for any z ∈ x/M the simplicial set M apx/M (y, z) is a retract of M apx′ /M (y ′ , z).

(3) This is clear since we have for any object t ∈ x′ /M , a natural equivalence
M apx′ /M (y ′ , t) ≃ M apx/M (y, t). 

Let us now fix I, a set of generating cofibrations in M .

Definition 1.2.3.4. (1) An object X is a strict finite I-cell object, if there

exists a finite sequence

X0 = ∅ / X1 / ... / Xn = X,
32 1.2. PRELIMINARIES ON LINEAR AND COMMUTATIVE ALGEBRA IN AN HA CONTEXT

and for any 0 ≤ i < n a push-out square

XO i / Xi+1
O

A / B,
ui

with ui ∈ I.
(2) An object X is a finite I-cell object (or simply a finite cell object when I is
clear) if it is equivalent to a strict finite I-cell object.
(3) The model category M is compactly generated if it satisfies the following
conditions.
(a) The model category M is cellular (in the sense of [Hi, §12]).
(b) There exists a set of generating cofibrations I, and generating trivial
cofibrations J whose domains and codomains are cofibrant, ω-compact
(in the sense of [Hi, §10.8]) and ω-small with respect to the whole cat-
egory M .
(c) Filtered colimits commute with finite limits in M .
The following proposition identifies finitely presented objects when M is com-
pactly generated.
Proposition 1.2.3.5. Let M be a compactly generated model category, and I be a
set of generating cofibrations whose domains and codomains are cofibrant, ω-compact
and ω-small with respect to the whole category M .
(1) A filtered colimit of fibrations (resp. trivial fibrations) is a fibration (resp. a
trivial fibration).
(2) For any filtered diagram Xi in M , the natural morphism
Hocolimi Xi −→ Colimi Xi
is an isomorphism in Ho(M ).
(3) Any object X in M is equivalent to a filtered colimit of strict finite I-cell
objects.
(4) An object X in M is finitely presented if and only if it is equivalent to a
retract, in Ho(M ), of a strict finite I-cell object.
Proof. (1) By assumption the domain and codomain of morphisms of I are ω-
small, so M is finitely generated in the sense of [Ho1, §7]. Property (1) is then proved
in [Ho1, §7].

(2) For a filtered category A, the colimit functor

Colim : M A −→ M
is a left Quillen functor for the levelwise projective model structure on M . By (1) we
know that Colim preserves trivial fibrations, and thus that it also preserves equiva-
lences. We therefore have isomorphisms of functors Hocolim ≃ LColim ≃ Colim.

(3) The small object argument (e.g. [Ho1, Thm. 2.1.14]) gives that any object
X is equivalent to a I-cell complex Q(X). By ω-compactness of the domains and
codomains of I, Q(X) is the filtered colimit of its finite sub-I-cell complexes. This
implies that X is equivalent to a filtered colimit of strict finite I-cell objects.

(4) Let A be a filtered category, and Y ∈ M A be a A-diagram. Let c(Y ) −→

R∗ (Y ) be a Reedy fibrant replacement of the constant simplicial object c(Y ) with
 1.2.3. FINITENESS CONDITIONS 33

values Y (in the model category of simplicial objects in M A , see [Ho1, §5.2]). By
(2), the induced morphism
c(Colima∈A Ya ) −→ Colima∈A R∗ (Ya )
is an equivalence of simplicial objects in M A . Moreover, (1) and the exactness of
filtered colimits implies that Colima∈A R∗ (Ya ) is a Reedy fibrant object in the model
categroy of simplicial objects in M (as filtered colimits commute with matching ob-
jects for the Reedy category ∆op , see [Ho1, §5.2]). This implies that for any cofibrant
and ω-small object K in M , we have
Hocolima∈A M ap(K, Ya ) ≃ Colima∈A M ap(K, Ya ) ≃ Colima∈A Hom(K, R∗ (Ya )) ≃
Hom(K, Colima∈A R∗ (Ya )) ≃ M ap(K, Colima∈AYa ).
This implies that the domains and codomains of I are homotopically finitely presented.
As filtered colimits of simplicial sets preserve homotopy pull-backs, we deduce that
any finite cell objects is also homotopically finitely presented, as they are constructed
from domains and codomains of I by iterated homotopy push-outs (we use here that
domains and codomains of I are cofibrant). This implies that any retract of a finite
cell object is homotopically finitely presented. Conversely, let X be a homotopically
finitely presented object in Ho(M ), and by (3) let us write it as Colimi Xi , where
Xi is a filtered diagram of finite cell objects. Then, [X, X] ≃ Colimi [X, Xi ], which
implies that this identity of X factors through some Xi , or in other words that X is
a retract in Ho(M ) of some Xi . 
Now, let M be a symmetric monoidal model category in the sense of [Ho1, §4]. We
remind that this implies in particular that the monoidal structure on M is closed, and
therefore possesses Hom’s objects HomM (x, y) ∈ M satisfying the usual adjunction
rule
Hom(x, HomM (y, z))) ≃ Hom(x ⊗ y, z).
The internal structure can be derived, and gives on one side a symmetric monoidal
structure − ⊗L − on Ho(M ), as well as Hom’s objects RHomM (x, y) ∈ Ho(M )
satisfying the derived version of the previous adjunction
[x, RHomM (y, z))] ≃ [x ⊗L y, z].
In particular, if 1 is the unit of the monoidal structure of M , then
[1, RHomM (x, y)] ≃ [x, y],
and more generally
M apM (1, RHomM (x, y)) ≃ M apM (x, y).
Moreover, the adjunction between −⊗L and RHomM extends naturally to an ad-
junction isomorphism
RHomM (x, RHomM (y, z))) ≃ RHomM (x ⊗L y, z).
The derived dual of an object x ∈ M will be denoted by
x∨ := RHomM (x, 1).
Definition 1.2.3.6. Let M be a symmetric monoidal model category. An object
x ∈ M is called perfect if the natural morphism
x ⊗L x∨ −→ RHomM (x, x)
is an isomorphism in Ho(M ).
Proposition 1.2.3.7. Let M be a symmetric monoidal model category.
34 1.2. PRELIMINARIES ON LINEAR AND COMMUTATIVE ALGEBRA IN AN HA CONTEXT

(1) If x and y are perfect objects in M , then so is x ⊗L y.

(2) If x is a perfect object in M , then for any objects y and z, the natural
morphism
RHomM (y, x ⊗L z) −→ RHomM (y ⊗L x∨ , z)
is an isomorphism in Ho(M ).
(3) If x in perfect in M , and y ∈ M , then the natural morphism
RHomM (x, y) −→ x∨ ⊗L y
is an isomorphism in Ho(M ).
(4) If 1 is finitely presented in M , then so is any perfect object.
(5) If M is furthermore a stable model category then perfect objects are sta-
ble by homotopy push-outs and homotopy pullbacks. In other words, if
x /y / z is a homotopy fiber sequence in M , and if two of the
objects x, y, and z are perfect then so is the third.
(6) Perfect objects are stable by retracts in Ho(M ).
Proof. (1), (2) and (3) are standard, as perfect objects are precisely the strongly
dualizable objects of the closed monoidal category Ho(M ) (see for example [May2]).

(4) Let x be a perfect object in M , and {zi }i∈I be a filtered diagram of objects
in M . Let x∨ := RHom(x, 1) the dual of x in Ho(M ). Then, we have
M apM (x, Hocolimi zi ) ≃ M apM (1, x∨ ⊗L Hocolimi zi ) ≃ M apM (1, Hocolimi x∨ ⊗L zi )
≃ Hocolimi M apM (1, x∨ ⊗L zi ) ≃ Hocolimi M apM (x, zi ).
(5) Let x /y / z be a homotopy fiber sequence in M . It is enough to
prove that if y and z are perfect then so is x. For this, let x∨ , y ∨ and z ∨ the duals of
x, y and z.
One has a morphism of homotopy fiber sequences

x ⊗L z ∨ / x ⊗L y ∨ / x ⊗L x∨

  
RHomM (z, x) / RHomM (y, x) / RHomM (x, x).

The five lemma and point (3) implies that the last vertical morphism is isomorphism,
and that x is perfect.

(6) If x is a retract of y, then the natural morphism

x ⊗L x∨ −→ RHomM (x, x)
is a retract of
y ⊗L y ∨ −→ RHomM (y, y).

The following corollary gives a condition under which perfect and finitely pre-
sented objects are the same.
Corollary 1.2.3.8. Suppose that M is a stable and compactly generated sym-
metric monoidal model category with 1 being ω-compact and cofibrant. We assume
that the set I of morphisms of the form
S n ⊗ 1 −→ ∆n+1 ⊗ 1
 1.2.4. SOME PROPERTIES OF MODULES 35

is a set of generating cofibration for M . Then an object x in M is perfect if and only

if it is finitely presented, and if and only if it is a retract of a finite I-cell object.
Proof. This essentially follows from Prop. 1.2.3.5 and Prop. 1.2.3.7, the only
statement which remains to be proved is that retract of finite I-cell objects are perfect.
But this follows from the fact that 1 is always perfect, and from Prop. 1.2.3.7 (5)
and (6). 
Remark 1.2.3.9. The notion of finitely presented morphisms and perfect objects
depend on the model structure and not only on the underlying category M . They
specialize to the corresponding usual categorical notions when M is endowed with the
trivial model structure.

1.2.4. Some properties of modules

In this Section we give some general notions of flatness and projectiveness of
modules over a commutative monoid in C.
Definition 1.2.4.1. Let A ∈ Comm(C) be a commutative monoid, and M be an
A-module.
(1) The A-module M is flat if the functor
− ⊗LA M : Ho(A − M od) −→ Ho(A − M od)
preserves homotopy pullbacks.
`L
(2) The A-module M is projective if it is a retract in Ho(A − M od) of E A
for some U-small set E.
Proposition 1.2.4.2. Let A −→ B be a morphism of commutative monoids in
C.
(1) The free A-module An of rank n is flat. Moreover, if infinite direct sums in
Ho(A − M od) commute with homotopy pull-backs, then for any U-small set
`L
E, the free A-module E A is flat.
(2) Flat modules in Ho(A − M od) are stable by derived tensor products, finite
coproducts and retracts.
(3) Projective modules in Ho(A − M od) are stable by derived tensor products,
finite coproducts and retracts.
(4) If M is a flat (resp. projective) A-module, then M ⊗LA B is a flat (resp.
projective) B-module.
(5) A perfect A-module is flat.
(6) Let us suppose that 1 is a finitely presented object in C. Then, a projective
`L
A-module is finitely presented if and only if it is a retract of E A for some
finite set E.
(7) Let us assume that 1 is a finitely presented object in C. Then, a projective
A-module is perfect if and only it is finitely presented.
Proof. (1) to (4) are easy and follow from the definitions.

(5) Let M be a perfect A-module, and M ∨ := RHomA (M, A) be its dual. Then,
for any homotopy cartesian square of A-modules
P /Q

 
P′ / Q′ ,
36 1.2. PRELIMINARIES ON LINEAR AND COMMUTATIVE ALGEBRA IN AN HA CONTEXT

the diagram
P ⊗LA M / Q ⊗L M
A

 
P ′ ⊗LA M / Q′ ⊗L M
A
is equivalent to
RHomA (M ∨ , P ) / RHom (M ∨ , Q)
A

 
RHomA (M ∨ , P ′ ) / RHom (M ∨ , Q′ ),
A
which is again homotopy cartesian by the general properties of derived internal Hom’s.
This shows that − ⊗L M preserves homotopy pullbacks, and hence that M is flat.
`
(6) Clearly, if E a finite set, then LE A is finitely presented object, as for any
A-module M one has
aL
M apA−Mod ( A, M ) ≃ M apC (1, M )E .
E
`L
Therefore, a retract of E A is also finitely presented.
Conversely, let M be a projective A-module which is also finitely presented. Let
` `
i : M −→ LE A be a morphism which admits a retraction. As LE A is the colimit
`
of LE0 A, for E0 running over the finite subsets of E, the morphism i factors as
L
a L
a
M −→ A −→ A
E0 E
`L
for some finite subset E0 ⊂ E. This shows that M is in fact a retract of E A.

(7) Using Prop. 1.2.3.7 (4) one sees that if M is perfect then it is finitely presented.
Conversely, let M be a finitely presented projective A-module. By (6) we know that
`L `L
M is a retract of E A for some finite set E. But, as E is finite, E A is perfect,
and therefore so is M as a retract of a perfect module. 

1.2.5. Formal coverings

The following notion will be highly used, and is a categorical version of faithful
morphisms of affine schemes.
Definition 1.2.5.1. A family of morphisms of commutative monoids {fi : A →
Bi }i∈I is a formal covering if the family of functors
{Lfi∗ : Ho(A − M od) −→ Ho(Bi − M od)}i∈I
is conservative (i.e. a morphism u in Ho(A − M od) is an isomorphism if and only if
all the Lfi∗ (u) are isomorphisms).
The formal covering families are stable by equivalences, homotopy push-outs and
compositions and therefore do form a model topology in the sense of [HAGI, Def.
4.3.1] (or Def. 1.3.1.1).
Proposition 1.2.5.2. Formal covering families form a model topology (Def. 1.3.1.1)
on the model category Comm(C).
 1.2.6. SOME PROPERTIES OF MORPHISMS 37

Proof. Stability by equivalences and composition is clear. The stability by

homotopy push-outs is an easy consequence of the transfer formula Prop. 1.1.0.8. 

1.2.6. Some properties of morphisms

In this part we review several classes of morphisms of commutative monoids in C,
generalizing the usual notions of Zariski open immersions, unramified, étale, smooth
and flat morphisms of affine schemes. It is interesting to notice that these notions
make sense in our general context, but specialize to very different notions in various
specific cases (see the examples at the beginning of each section §2.1, §2.2, §2.3 and
§2.4).

Definition 1.2.6.1. Let f : A −→ B be a morphism of commutative monoids in

C.
(1) The morphism f is an epimorphism if for any commutative A-algebra C the
simplicial set M apA−Comm(C)(B, C) is either empty or contractible.
(2) The morphism f is flat if the induced functor
Lf ∗ : Ho(A − M od) −→ Ho(B − M od)
commutes with finite homotopy limits.
(3) The morphism f is a formal Zariski open immersion if it is flat and if the
functor
f∗ : Ho(B − M od) −→ Ho(A − M od)
is fully faithful.
(4) The morphism f is formally unramified if LB/A ≃ 0 in Ho(B − M od).
(5) The morphism f is formally étale if the natural morphism
LA ⊗LA B −→ LB
is an isomorphism in Ho(B − M od).
(6) The morphism f is formally thh-étale if the natural morphism
B −→ T HH(B/A)
is an isomorphism in Ho(Comm(C)).
Remark 1.2.6.2. One remark concerning our notion of epimorphism of commu-
tative monoids is in order. First of all, in a category C (without any model structure)
having fiber products, a morphism x −→ y is a monomorphism if and only the diag-
onal morphism x −→ x ×y x is an isomorphism. A natural generalization of this fact
gives a notion of monomorphism in any model category M , as a morphism x −→ y
whose diagonal x −→ x ×hy x is an isomorphism in Ho(M ). Equivalently, the mor-
phism x −→ y is a monomorphism if and and only if for any z ∈ M the induced
morphism of simplicial sets M apM (z, x) −→ M apM (z, y) is a monomorphism in the
model category SSet. Furthermore, it is easy to check that a morphism f : K −→ L
is a monomorphism in the model category SSet if and only if for any s ∈ L the
homotopy fiber of f at s is either empty or contractible. Therefore we see that a mor-
phism A −→ B in Comm(C) is an epimorphism in the sense of Def. 1.2.6.1 (1) if and
only if it is a monomorphism when considered as a morphism in the model category
Comm(C)op , or equivalently if the induced morphism B ⊗LA B −→ B is an isomor-
phism in Ho(Comm(C)). This justifies our terminology, and moreover shows that
when the model structure on the category M is trivial, an epimorphism in the sense
of our definition is nothing else than an epimorphism in M in the usual categorical
sense.
38 1.2. PRELIMINARIES ON LINEAR AND COMMUTATIVE ALGEBRA IN AN HA CONTEXT

Proposition 1.2.6.3. Epimorphisms, flat morphisms, formal Zariski open im-

mersions, formally unramified morphisms, formally étale morphisms, and formally
thh-étale morphisms, are all stable by compositions, equivalences and homotopy push-
outs.
Proof. This is a simple exercise using the definitions and Propositions 1.1.0.8,
1.2.1.6, 1.2.2.2, and 1.2.3.7. 
The relations between all these notions are given by the following proposition.
Proposition 1.2.6.4. (1) A morphism f : A −→ B is an epimorphism if
and only if the functor
f∗ : Ho(B − M od) −→ Ho(A − M od)
is fully faithful.
(2) A formal Zariski open immersion is an epimorphism. A flat epimorphism
is a formal Zariski open immersion.
(3) A morphism f : A −→ B of commutative monoids is formally thh-étale if
and only if for any commutative A-algebra C the simplicial set
M apA−Comm(C)(B, C)
is discrete (i.e. equivalent to a set).
(4) A formally étale morphism is formally unramified.
(5) An epimorphism is formally unramified and formally thh-étale.
(6) A morphism f : A −→ B in Comm(C) is formally unramified if and only
the morphism
B ⊗LA B −→ B
is formally étale.
Proof. (1) Let f : A −→ B be a morphism such that the right Quillen functor
f∗ : B−M od −→ A−M od induces a fully faithful functor on the homotopy categories.
Therefore, the adjunction morphism
M ⊗LA B −→ M
is an isomorphism for any M ∈ Ho(B − M od). In particular, the functor
f∗ : Ho(B − Comm(C)) −→ Ho(A − Comm(C))
is also fully faithful. Let C be a commutative A-algebra, and let us suppose that
M apA−Comm(C)(B, C) is not empty. This implies that C is isomorphic in Ho(A −
Comm(C)) to some f∗ (C ′ ) for C ′ ∈ Ho(B − Comm(C)). Therefore, we have
M apA−Comm(C)(B, C) ≃ M apA−Comm(C)(f∗ (B), f∗ (C ′ )) ≃ M apB−Comm(C)(B, C ′ ) ≃ ∗,
showing that f is a epimorphism.
Conversely, let f : A −→ B be epimorphism. For any commutative A-algebra C,
we have
M apA−Comm(C)(B, C) ≃ M apA−Comm(C)(B, C) × M apA−Comm(C)(B, C),
showing that the natural morphism B ⊗LA B −→ B is an isomorphism in Ho(A −
Comm(C)). This implies that for any B-module M , we have
M ⊗LA B ≃ M ⊗LB (B ⊗LA B) ≃ M,
or in other words, that the adjunction morphism M −→ f∗ Lf ∗ (M ) ≃ M ⊗LA B is an
isomorphism in Ho(B − M od). This means that f∗ : Ho(B − M od) −→ Ho(A − M od)
is fully faithful.
 1.2.6. SOME PROPERTIES OF MORPHISMS 39

(2) This is clear by (1) and the definitions.

(3) For any morphism of commutative A-algebras, f : B −→ C we have

M apB−Comm(C) (T HH(B/A), C)) ≃ Ωf M apA−Comm(C)(B, C).
Therefore, B −→ T HH(B/A) is an equivalence if and only if for any such f : B −→ C,
the simplicial set Ωf M apA−Comm(C)(B, C) is contractible. Equivalently, f is formally
thh-étale if and only if M apA−Comm(C)(B, C) is discrete.

(4) Let f : A −→ B be a formally étale morphism of commutative monoids in C.

By Prop. 1.2.1.6 (1), there is a homotopy cofiber sequence of B-modules
LA ⊗LA B −→ LB −→ LB/A ,
showing that if the first morphism is an isomorphism then LB/A ≃ ∗ and therefore
that f is formally unramified.

(5) Let f : A −→ B an epimorphism. By definition and (3) we know that f

is formally thh-étale. Let us prove that f is also formally unramified. Let M be a
B-module. As we have seen before, for any commutative B-algebra C, the adjunction
morphism C −→ C ⊗LA B is an isomorphism. In particular, the functor
f∗ : Ho(B − Comm(C)/B) −→ Ho(A − Comm(C)/B)
is fully faithful. Therefore we have
DerA (B, M ) ≃ M apA−Comm(C)/B (B, B ⊕ M ) ≃ M apB−Comm(C)/B (B, B ⊕ M ) ≃ ∗,
showing that DerA (B, M ) ≃ ∗ for any B-module M , or equivalently that LB/A ≃ ∗.

(6) Finally, Prop. 1.2.1.6 (2) shows that the morphism B ⊗LA B −→ B is formally
étale if and only if the natural morphism
a
LB/A LB/A −→ LB/A

is an isomorphism in Ho(B − M od). But this is equivalent to LB/A ≃ ∗. 

In order to state the next results we recall that as C is a pointed model category
one can define a suspension functor (see [Ho1, §7])
S : Ho(C) −→ Ho(C)
`
X 7→ S(X) := ∗ LX ∗.
This functor possesses a right adjoint, the loop functor
Ω : Ho(C) −→ Ho(C)
X 7→ Ω(X) := ∗ ×hX ∗.
Proposition 1.2.6.5. Assume that the base model category C is such that the
suspension functor S : Ho(C) −→ Ho(C) is fully faithful.
(1) A morphism of commutative monoids in C is formally étale if and only if it
is formally unramified.
(2) A formally thh-étale morphism of commutative monoids in C is a formally
étale morphism.
(3) An epimorphism of commutative monoids in C is formally étale.
40 1.2. PRELIMINARIES ON LINEAR AND COMMUTATIVE ALGEBRA IN AN HA CONTEXT

Proof. (1) By the last proposition we only need to prove that a formally un-
ramified morphism is also formally étale. Let f : A −→ B be such a morphism. By
Prop. 1.2.1.6 (1) there is a homotopy cofiber sequence of B-modules
LA ⊗LA B −→ LB −→ ∗.
This implies that for any B-module M , the homotopy fiber of the morphism
Der(B, M ) −→ Der(A, M )
is contractible, and in particular that this morphism induces isomorphisms on all
higher homotopy groups. It remains to show that this morphism induces also an
isomorphism on π0 . For this, we can use the hypothesis on C which implies that the
suspension functor on Ho(B − M od) is fully faithful. Therefore, we have
π0 (Der(B, M )) ≃ π0 (Der(B, ΩS(M ))) ≃ π1 (Der(B, SM )) ≃
≃ π1 (Der(A, SM )) ≃ π0 (Der(A, M )).
(2) Let A −→ B be a formally thh-étale morphism in Comm(C).
As M apA−Comm(C) (B, C) is discrete for any commutative A-algebra C, the simplicial
set DerA (B, M ) is discrete for any B-module M . Using the hypothesis on C we get
that for any B-module M
π0 (DerA (B, M )) ≃ π0 (DerA (B, ΩS(M ))) ≃ π1 (DerA (B, SM )) ≃ 0,
showing that DerA (B, M ) ≃ ∗, and therefore that LB/A ≃ ∗. This implies that f is
formally unramified, and therefore is formally étale by the first part of the proposition.

(3) This follows from (2) and Prop. 1.2.6.4 (5). 

The hypothesis of Proposition 1.2.6.5 saying that the suspension is fully faithful
will appear in many places in the sequel. It is essentially equivalent to saying that the
homotopy theory of C can be embedded in a stable homotopy theory in such a way
that homotopy colimits are preserved (it could be in fact called left semi-stable). It is
a very natural condition as many statements will then simplify, as shown for example
by our infinitesimal theory in §3.
Corollary 1.2.6.6. Assume furthermore that C is a stable model category, and
let f : A −→ B be a morphism of commutative monoids in C. The following are
equivalent.
(1) The morphism f is a formal Zariski open immersion.
(2) The morphism f is an epimorphism.
(3) The morphism f is a formally étale epimorphism.
Proof. Indeed, in the stable case all base change functors commute with limits,
so Prop. 1.2.6.4 (1) and (2) shows that formal Zariski open immersions are exactly the
epimorphisms. Furthermore, by Prop. 1.2.6.4 (5) and Prop. 1.2.6.5 all epimorphisms
are formally étale. 
Definition 1.2.6.7. A morphism of commutative monoids in C is a Zariski open
immersion (resp. unramified, resp. étale, resp. thh-étale) if it is finitely presented (as
a morphism in the model category Comm(C)) and is a formal Zariski open immersion
(resp. formally unramified, resp. formally étale, resp. formally thh-étale).
Clearly, using what we have seen before, Zariksi open immersions, unramified
morphisms, étale morphisms, and thh-étale morphisms are all stable by equivalences,
compositions and push-outs.
 1.2.8. INFINITESIMAL LIFTING PROPERTIES 41

1.2.7. Smoothness
We define two general notions of smoothness, both different generalizations of the
usual notion, and both useful in certain contexts. A third, and still different, notion
of smoothness will be given in the next section.
Definition 1.2.7.1. Let f : A −→ B be a morphism of commutative algebras.
(1) The morphism f is formally perfect (or simply fp) if the B-module LB/A is
perfect (in the sense of Def. 1.2.3.6).
(2) The morphism f is formally smooth if the B-module LB/A is projective (in
the sense of Def. 1.2.4.1) and if the morphism
LA ⊗LA B −→ LB
has a retraction in Ho(B − M od).
Definition 1.2.7.2. Let f : B −→ C be a morphism of commutative algebras.
The morphism f is perfect, or simply p, (resp. smooth) if it is finitely presented (as
a morphism in the model category Comm(C)) and is fp (resp. formally smooth).
Of course, (formally) étale morphisms are (formally) smooth morphisms as well
as (formally) perfect morphisms.
Proposition 1.2.7.3. The fp, perfect, formally smooth and smooth morphisms
are all stable by compositions, homotopy push outs and equivalences.
Proof. Exercise. 

1.2.8. Infinitesimal lifting properties

While the first two notions of smoothness only depend on the underlying symmet-
ric monoidal model category C, the third one, to be defined below, will also depend
on the HA context we are working in.
Recall from Definition 1.1.0.11 that an HA context is a triplet (C, C0 , A), consisting of
a symmetric monoidal model category C, two full sub-categories stable by equivalences
C0 ⊂ C A ⊂ Comm(C),
such that:
• 1 ∈ C0 , C0 is closed under by U-small homotopy colimits, and X ⊗L Y ∈
Ho(C0 ) if X and Y in Ho(C0 ).
• any A ∈ A is C0 -good (i.e. the functor
Ho(A − M od) −→ Ho((A − M odop ∧
0 ) )

is fully faithful);
• assumptions 1.1.0.1, 1.1.0.2, 1.1.0.3, 1.1.0.4, 1.1.0.6 are satisfied.
Recall also that C1 is the full subcategory of C consisting of all objects equivalent to
suspensions of objects in C0 , Comm(C)0 the full subcategory of Comm(C) consisting
of commutative monoids whose underlying C-object is in C0 , and, for A ∈ Comm(C),
A − M od0 (resp. A − M od1 , resp. A − Comm(C)0 ) is the full subcategory of A − M od
consisting of A-modules whose underlying C-object is in C0 (resp. of A − M od consist-
ing of A-modules whose underlying C-object is in C1 , resp. of A−Comm(C) consisting
of commutative A-algebras whose underlying C-object is in C0 ).

Definition 1.2.8.1. Let f : A −→ B be a morphism in Comm(C).

42 1.2. PRELIMINARIES ON LINEAR AND COMMUTATIVE ALGEBRA IN AN HA CONTEXT

(1) The morphism f is called formally infinitesimally smooth relative to the

HA context (C, C0 , A) (or simply formally i-smooth when the HA context is
clear) if for any R ∈ A, any morphism A −→ R of commutative monoids,
any M ∈ R − M od1 , and any d ∈ π0 (DerA (R, M )), the natural morphism



π0 M apA−Comm(C)(B, R ⊕d ΩM ) −→ π0 M apA−Comm(C) (B, R)
is surjective.
(2) The morphism f is called i-smooth if it is formally i-smooth and finitely
presented.
The following proposition is immediate from the definition.
Proposition 1.2.8.2. Formally i-smooth and i-smooth morphisms are stable by
equivalences, composition and homotopy push-outs.
The next result provides a criterion for formally i-smooth morphisms in terms of
their cotangent complexes.
Proposition 1.2.8.3. A morphism f : A −→ B is formally i-smooth if and only
if for any morphism B −→ R with R ∈ A, and any R-module M ∈ R − M od1 , the
natural morphism
[LR/A , M ] ≃ π0 (DerA (R, M ))) −→ π0 (DerA (B, M )) ≃ [LB/A , M ]B−Mod
is zero.
Proof. Let us first assume that the condition of the proposition is satisfied.
Let us consider R ∈ A, a morphism A −→ R, an R-module M ∈ R − M od1 and
d ∈ π0 (DerA (R, M )). The homotopy fiber of the natural morphism
M apA−Comm(C)(B, R ⊕d ΩM ) −→ M apA−Comm(C)(B, R)
taken at some morphism B −→ R in Ho(Comm(C)), can be identified with the path
space P ath0,d′ DerA (B, M ) from 0 to d′ in DerA (B, M ), where d′ is the image of d
under the natural morphism
π0 (DerA (R, M )) −→ π0 (DerA (B, M )).
By assumption, 0 and d′ belong to the same connected component of DerA (B, M ),
and thus we see that P ath0,d′ DerA (B, M ) is non-empty. We have thus shown that
M apA−Comm(C)(B, R ⊕d ΩM ) −→ M apA−Comm(C)(B, R)
has non-empty homotopy fibers and therefore that



π0 M apA−Comm(C)(B, R ⊕d ΩM ) −→ π0 M apA−Comm(C) (B, R)
is surjective. The morphism f is therefore formally i-smooth.
Conversely, let R ∈ A, A −→ R a morphism, M ∈ R−M od1 and d ∈ π0 (DerA (R, M )).
Let A → R be a morphism of commutative monoids. We consider the diagram

A / R[Ωd M ] /R

  
B /R / R ⊕ M.
s

The homotopy fiber of

M apA−Comm(C)(B, R ⊕d ΩM ) −→ M apA−Comm(C)(B, R)
 1.2.9. STANDARD LOCALIZATIONS AND ZARISKI OPEN IMMERSIONS 43

is non-empty because f is formally i-smooth. By definition this means that the image
of d by the morphism
π0 (DerA (R, M )) −→ π0 (DerA (B, M ))
is zero. As this is true for any d, this finishes the proof of the proposition. 
Corollary 1.2.8.4. (1) Any formally unramified morphism is formally i-
smooth.
(2) Assume that the suspension functor S is fully faithful, that C1 = C (so that
in particular C is stable), and A = Comm(C). Then formally i-smooth
morphisms are precisely the formally étale morphisms.
Proof. It follows immediately from 1.2.8.3. 
Corollary 1.2.8.5. We assume that for any M ∈ C1 one has [1, M ] = 0. Then
any formally smooth morphism is formally i-smooth.
Proof. By Prop. 1.2.8.3 and definition of formal smoothness, it is enough to
show that for any commutative monoid A ∈ C, any A-module M ∈ A− M od1 and any
projective A-module P we have [P, M ] = 0. By assumption this is true for P = A,
and thus also true for free A-modules and their retracts. 

1.2.9. Standard localizations and Zariski open immersions

For any object A ∈ C, we can define its underlying space as
|A| := M apC (1, A) ∈ Ho(SSetU ).
The model category C being pointed, the simplicial set |A| has a natural base point
∗ ∈ |A|, and one can therefore define the homotopy groups of A
πi (A) := πi (|A|, ∗).
When A is the underlying object of a commutative monoid A ∈ Comm(C) one has
by adjunction
π0 (A) = π0 (M apC (1, A)) ≃ [1, A]C ≃ [A, A]A−Mod .
Since A is the unit of the monoidal structure on the additive category Ho(A − M od),
the composition of endomorphisms endows π0 (A) with a multiplicative structure mak-
ing it into a commutative ring. More generally, the category Ho(A − M od) has a nat-
ural structure of a graded category (i.e. has a natural enrichment into the symmetric
monoidal category of N-graded abelian groups), defined by
[M, N ]∗ := ⊕i [S i (N ), M ] ≃ ⊕i [N, Ωi (M )],
where S i is the i-fold iterated suspension functor, and Ωi the i-fold iterated loop func-
tor. Therefore, the graded endomorphism ring of the unit A has a natural structure
of a graded commutative ring. As this endomorphism ring is naturally isomorphic to
π∗ (A) := ⊕i πi (A),
we obtain this way a natural structure of a graded commutative ring on π∗ (A). This
clearly defines a functor
π∗ : Ho(Comm(C)) −→ GComm,
from Ho(Comm(C)) to the category of N-graded commutative rings.
In the same way, for a commutative monoid A ∈ Comm(C), and a A-module M ,
one can define
π∗ (M ) := π∗ (|M |) = π∗ (M apA−Mod (A, M )),
44 1.2. PRELIMINARIES ON LINEAR AND COMMUTATIVE ALGEBRA IN AN HA CONTEXT

which has a natural structure of a graded π∗ (A)-module. This defines a functor

π∗ : Ho(A − M od) −→ π∗ (A) − GM od,
from Ho(A − M od) to the category of N-graded π∗ (A)-modules.
Proposition 1.2.9.1. Let A ∈ Comm(C) be a commutative monoid in C, and
a ∈ π0 (A). There exists an epimorphism A −→ A[a−1 ], such that for any commutative
A-algebra C, the simplicial set M apA−Comm(C) (A[a−1 ], C) is non-empty (and thus
contractible) if and only if the image of a in π0 (C) by the morphism π0 (A) → π0 (C)
is an invertible element.
Proof. We represent the element a as a morphism in Ho(A − M od) a : A −→ A.
Taking the image of a by the left derived functor of the free commutative A-algebra
functor (which is left Quillen)
FA : A − M od −→ A − Comm(C)
we find a morphism in Ho(A − Comm(C))
a : LFA (A) −→ LFA (A).
As the model category A − Comm(C) is a U-combinatorial model category, one can
apply the localization techniques of in order to invert any U-small set of morphisms
in A − Comm(C) (see [Sm, Du2]). We let La A − Comm(C) be the left Bousfield
localization of A − Comm(C) along the set of morphisms (with one element)
Sa := {a : LFA (A) −→ LFA (A)}.
−1
We define A[a ] ∈ Ho(A − Comm(C)) as a local model of A in La A − Comm(C),
the left Bousfield localization of A − Comm(C) along Sa .
First of all, the Sa -local objects are the commutative A-algebras B such that the
induced morphism
a∗ : M apA−Mod (A, B) −→ M apA−Mod (A, B)
is an equivalence. Equivalently, the multiplication by a ∈ π0 (A)
×a : π∗ (B) −→ π∗ (B)
is an isomorphism. This shows that the Sa -local objects are the commutative A-
algebras A −→ B such that the image of a by π0 (A) −→ π0 (B) is invertible.
Suppose now that C ∈ A − Comm(C) is such that M apA−Comm(C) (A[a−1 ], C) is
not empty. The morphism π0 (A) −→ π0 (C) then factors through π0 (A[a−1 ]), and
thus the image of a is invertible in π0 (C). Therefore, C is a Sa -local object, and thus
M apA−Comm(C)(A[a−1 ], C) ≃ M apA−Comm(C) (A, C) ≃ ∗.
This implies that A −→ A[a−1 ] is an epimorphism. It only remain to prove that
if C ∈ A − Comm(C) is such that the image of a is invertible in π0 (C), then
M apA−Comm(C)(A[a−1 ], C) is non-empty. But such a C is a Sa -local object, and
therefore
∗ ≃ M apA−Comm(C) (A, C) ≃ M apA−Comm(C) (A[a−1 ], C).


Definition 1.2.9.2. Let A ∈ Comm(C) and a ∈ π0 (A). The commutative A-

algebra A[a−1 ] is called the standard localization of A with respect to a.
 1.2.9. STANDARD LOCALIZATIONS AND ZARISKI OPEN IMMERSIONS 45

A useful property of standard localizations is given by the following corollary of

the proof of Prop. 1.2.9.1. In order to state it, we will use the following notations. For
any A ∈ Comm(C) and a ∈ π0 (A), we represent a as a morphism in Ho(A − M od),
A −→ A. Tensoring this morphism with M gives a morphism in Ho(A − M od),
denoted by
×a : M −→ M.
Corollary 1.2.9.3. Let A ∈ Comm(C), a ∈ π0 (A) and let f : A −→ A[a−1 ] be
as in Prop. 1.2.9.1.
(1) The functor
Ho(A[a−1 ] − Comm(C)) −→ Ho(A − Comm(C))
is fully faithful, and its essential image consists of all commutative B-algebras
such that the image of a is invertible in π0 (B).
(2) The functor
f∗ : Ho(A[a−1 ] − M od) −→ Ho(A − M od)
is fully faithful and its essential image consists of all A-modules M such that
the multiplication by a
×a : M −→ M
is an isomorphism in Ho(A − M od).
Proof. (1) The fact that the functor f∗ is fully faithful is immediate as f is an
epimorphism. The fact that the functor f∗ takes its values in the required subcategory
is clear by functoriality of the construction π∗ .
Let B be a commutative A-algebra such that the image of a is invertible in π0 (B).
Then, we know that B is a Sa -local object. Therefore, one has
M apA−Comm(C) (A[a−1 ], B) ≃ M apA−Comm(C) (A, B) ≃ ∗,
showing that B is in the image of f∗ .

(2) The fact that f∗ is fully faithful follows from Prop. 1.2.6.4 (1). Let M ∈
A[a−1 ] − M od and let us prove that the morphism ×a : M −→ M is an isomorphism
in Ho(A − M od). Using that A −→ A[a−1 ] is an epimorphism, one finds M ≃
M ⊗LA A[a−1 ], which reduces the problem to the case where M = A[a−1 ]. But then,
the morphism ×a : A[a−1 ] −→ A[a−1 ], as a morphism in Ho(A[a−1 ] − M od) lives in
[A[a−1 ], A[a−1 ]] ≃ π0 (A[a−1 ]), and correspond to the image of a by the morphism
π0 (A) −→ π0 (A[a−1 ]), which is then invertible. In other words, ×a : A[a−1 ] −→
A[a−1 ] is an isomorphism.
Conversely, let M be an A-module such that the morphism ×a : M −→ M is an
isomorphism in Ho(A − M od). We need to show that the adjunction morphism
M −→ M ⊗LA A[a−1 ]
is an isomorphism. For this, we use that the morphism A −→ A[a−1 ] can be con-
structed using a small object argument with respect to the horns over the morphism
a : LFA (A) −→ LFA (A) (see [Hi, 4.2]). Therefore, a transfinite induction argument
shows that it is enough to prove that the morphism induced by tensoring
a ⊗ Id : LFA (A) ⊗LA M ≃ LFA (M ) −→ LFA (A) ⊗LA M ≃ LFA (M )
is an isomorphism in Ho(A − M od). But this morphism is the image by the functor
LFA of the morphism ×a : M −→ M , and is therefore an isomorphism. 
46 1.2. PRELIMINARIES ON LINEAR AND COMMUTATIVE ALGEBRA IN AN HA CONTEXT

Proposition 1.2.9.4. Let A ∈ Comm(C), a ∈ π0 (A) and A −→ A[a−1 ] the

standard localization with respect to a. Assume that the model category C is finitely
generated (in the sense of [Ho1]).
(1) The morphism A −→ A[a−1 ] is a formal Zariski open immersion.
(2) If 1 is a finitely presented object in C, then A −→ A[a−1 ] is a Zariski open
immersion.
Proof. (1) It only remains to show that the morphism A −→ A[a−1 ] is flat.
In other words, we need to prove that the functor M 7→ M ⊗LA A[a−1 ], preserves
homotopy fiber sequences of A-modules.
The model category A − M od is U-combinatorial and finitely generated. In par-
ticular, there exists a U-small set G of ω-small cofibrant objects in A − M od, such
that a morphism N −→ P is an equivalence in A − M od if and only if for any X ∈ G
the induced morphism M apA−Mod (X, N ) −→ M apA−Mod (X, P ) is an isomorphism
in Ho(SSet). Furthermore, filtered homotopy colimits preserve homotopy fiber se-
quences. For any A-module M , we let Ma be the transfinite homotopy colimit
×a ×a
M /M / ...
where the morphism ×a is composed with itself ω-times.
The functor M 7→ Ma commutes with homotopy fiber sequences. Therefore, it
only remain to show that Ma is naturally isomorphic in Ho(A−M od) to M ⊗LA A[a−1 ].
For this, it is enough to check that the natural morphism M −→ Ma induces an
isomorphism in Ho(A − M od)
M ⊗LA A[a−1 ] ≃ (Ma ) ⊗LA A[a−1 ],
and that the natural morphism
(Ma ) ⊗LA A[a−1 ] −→ Ma
is an isomorphism in Ho(A − M od). The first assumption follows easily from the
fact that − ⊗LA A[a−1 ] commutes with homotopy colimits, and the fact that ×a :
A[a−1 ] −→ A[a−1 ] is an isomorphism in Ho(A − M od). For the second assumption
we use Cor. 1.2.9.3 (2), which tells us that it is enough to check that the morphism
×a : Ma −→ Ma
is an isomorphism in Ho(A − M od). For this, we need to show that for any X ∈ G
the induced morphism
M apA−Mod (X, Ma ) −→ M apA−Mod (X, Ma )
is an isomorphism in Ho(SSet). But, as the objects in G are cofibrant and ω-small,
M apA−Mod (X, −) commutes with ω-filtered homotopy colimits, and the morphism
M apA−Mod (X, Ma ) −→ M apA−Mod (X, Ma )
is then obviously an isomorphism in Ho(SSet) by the construction of Ma .

(2) When 1 is finitely presented, one has for any filtered diagram of commutative
A-algebras Bi an isomorphism
Colimi π∗ (Bi ) ≃ π∗ (Hocolimi Bi ).
Using that M apA−Comm(C)(A[a−1 ], B) is either empty or contractible, depending
whether of not a goes to a unit in π∗ (B), we easily deduce that
Hocolimi M apA−Comm(C) (A[a−1 ], Bi ) ≃ M apA−Comm(C)(A[a−1 ], Hocolimi Bi ).

 1.2.10. ZARISKI OPEN IMMERSIONS AND PERFECT MODULES 47

We can also show that the natural morphism A −→ A[a−1 ] is a formally étale
morphism in the sense of Def. 1.2.6.1.
Proposition 1.2.9.5. Let A ∈ Comm(C) and a ∈ π0 (A). Then, the natural
morphism A −→ A[a−1 ] is formally étale.
Proof. Let M be any A[a−1 ]-module. We need to show that the natural mor-
phism
M apA−Comm(C)/A[a−1] (A[a−1 ], A[a−1 ]⊕M ) −→ M apA−Comm(C)/A[a−1] (A, A[a−1 ]⊕M )
is an isomorphism in Ho(SSet). Using the universal property of A −→ A[a−1 ] given
by Prop. 1.2.9.1 we see that it is enough to prove that for any B ∈ Comm(C), and
any B-module M the natural projection π0 (B ⊕ M ) −→ π0 (B) reflects invertible
elements (i.e. an element in π0 (B ⊕ M ) is invertible if and only its image in π0 (B) is
so). But, clearly, π0 (B ⊕ M ) can be identified with the trivial square zero extension
of the commutative ring π0 (B) by π0 (M ), which implies the required result. 

Corollary 1.2.9.6. Assume that the model category C is finitely presented, and
that the unit 1 is finitely presented in C. Then for any A ∈ Comm(C) and a ∈ π0 (A),
the morphism A −→ A[a−1 ] is an étale, flat epimorphism.
Proof. Put 1.2.9.4 and 1.2.9.5 together. 

1.2.10. Zariski open immersions and perfect modules

Let A be a commutative monoid in C and K be a perfect A-module in the sense
of Def. 1.2.3.6. We are going to define a Zariski open immersion A −→ AK , which
has to be thought as the complement of the support of the A-module K.
Proposition 1.2.10.1. Assume that C is stable model category. Then there exists
a formal Zariski open immersion A −→ AK , such that for any commutative A-algebra
C, the simplicial set
M apA−Comm(C) (AK , C)
is non-empty (and thus contractible) if and only if K ⊗LA C ≃ ∗ in Ho(C − M od). If
the unit 1 is furthermore finitely presented, then A −→ AK is finitely presented and
thus is a Zariski open immersion.
Proof. The commutative A-algebra is constructed using a left Bousfield local-
ization of the model category A − Comm(C), as done in the proof of Prop. 1.2.9.1.
We let I be a generating U-small set of cofibrations in A − Comm(C), and K ∨ :=
RHomA−Mod (K, A) be the dual of K in Ho(A − M od). For any morphism X −→ Y
in I, we consider the morphism of free commutative A-algebras
LFA (K ∨ ⊗LA X) −→ LFA (K ∨ ⊗LA Y ),
where FA : A − M od −→ A − Comm(C) is the left Quillen functor sending an A-
module to the free commutative A-algebra it generates. When X −→ Y varies in I
this gives a U-small set of morphisms denoted by SK in A − Comm(C). We consider
LK A − Comm(C), the left Bousfield localization of A − Comm(C) along the set SK .
By definition, A −→ AK is an SK -local model of A in the localized model category
LK A − Comm(C).
Lemma 1.2.10.2. The SK -local objects in LK A− Comm(C) are precisely the com-
mutative A-algebras B such that K ⊗LA B ≃ ∗ in Ho(A − M od).
48 1.2. PRELIMINARIES ON LINEAR AND COMMUTATIVE ALGEBRA IN AN HA CONTEXT

Proof. First of all, one has an adjunction isomorphism in Ho(SSet)

M apA−Comm(C) (LFA (K ∨ ⊗LA X), B) ≃ M apA−Mod (X, K ⊗LA B).
This implies that an object B ∈ A − Comm(C) is SK -local if and only if for all
morphism X −→ Y in I the induced morphism
M apA−Mod (Y, K ⊗LA B) −→ M apA−Mod (X, K ⊗LA B)
is an isomorphism in Ho(SSet). As I is a set of generating cofibrations in A − M od
this implies that B is SK -local if and only if K ⊗LA C ≃ ∗ in Ho(A − M od). 

We now finish the proof of proposition 1.2.10.1. First of all, Lem. 1.2.10.2 implies
that for any commutative A-algebra B, if the mapping space M apA−Comm(C) (AK , B)
is non-empty then
B ⊗LA K ≃ B ⊗LAK (AK ⊗LA K) ≃ ∗,
and thus B is an SK -local object. This shows that if M apA−Comm(C) (AK , B) is
non-empty then one has
M apA−Comm(C)(AK , B) ≃ M apA−Comm(C)(A, B) ≃ ∗.
In other words A −→ AK is a formal Zariski open immersion by Cor. 1.2.6.6 and the
stability assumption on C. It only remains to prove that A −→ AK is also finitely
presented when 1 is a finitely presented object in C.
For this, let {Ci }i∈I be a filtered diagram of commutative A-algebras and C be its
homotopy colimit. By the property of A −→ AK , we need to show that if K ⊗LA C ≃ ∗
then there is an i ∈ I such that K ⊗LA Ci ≃ ∗. For this, we consider the two elements
Id and ∗ in [K ⊗LA C, K ⊗LA C]C−Mod . As K is perfect and 1 is a finitely presented
object, K is a finitely presented A-module by Prop. 1.2.3.7 (4). Therefore, one has
∗ ≃ [K ⊗LA C, K ⊗LA C] ≃ Colimi∈I [K, K ⊗LA Ci ]A−Mod .
As the two elements Id and ∗ becomes equal in the colimit, there is an i such that
they are equal as elements in
[K, K ⊗LA Ci ]A−Mod ≃ [K ⊗LA Ci , K ⊗LA Ci ]Ci −Mod ,
showing that K ⊗LA Ci ≃ ∗ in Ho(Ci − M od). 

Corollary 1.2.10.3. Assume that C is a stable model category, and let f : A −→

AK be as in Prop. 1.2.10.1.
(1) The functor
Ho(AK − Comm(C)) −→ Ho(A − Comm(C))
is fully faithful, and its essential image consists of all commutative B-algebras
such that B ⊗LA K ≃ ∗.
(2) The functor
f∗ : Ho(AK − M od) −→ Ho(A − M od)
is fully faithful and its essential image consists of all A-modules M such that
M ⊗LA K ≃ ∗.
Proof. As the morphism A −→ AK is an epimorphism, we know by Prop.
1.2.6.4 that the functors
Ho(AK − Comm(C)) −→ Ho(A − Comm(C)) Ho(AK − M od) −→ Ho(A − M od)
 1.2.11. STABLE MODULES 49

are both fully faithful. Furthermore, a commutative A-algebra B is in the essential

image of the first one if and only if M apA−Comm(C)(AK , B) ≃ ∗, and therefore if and
only if B ⊗LA K ≃ ∗. For the second functor, its clear that if M is a AK -module, then
M ⊗LA K ≃ M ⊗LAK AK ⊗LA K ≃ ∗.
Conversely, let M be an A-module such that M ⊗LA K ≃ ∗.
Lemma 1.2.10.4. For any A-module M , K ⊗LA M ≃ ∗ if and only if K ∨ ⊗LA M ≃ ∗.
Proof. Indeed, as K is perfect, K ∨ is a retract of K ∨ ⊗LA K ⊗LA K ∨ in Ho(A −
M od). This implies that K ∨ ⊗LA M is a retract of K ∨ ⊗LA K ⊗LA K ∨ ⊗LA M , showing
that


K ⊗LA M ≃ ∗ ⇒ K ∨ ⊗LA M ≃ ∗ .
By symmetry this proves the lemma. 
We need to prove that the adjunction morphism
M −→ M ⊗LA AK
is an isomorphism in Ho(A − M od). For this, we use the fact that the morphism
A −→ AK can be constructed using a small object argument on the set of horns on
the set SK (see [Hi, 4.2]). By a transfinite induction we are therefore reduced to show
that for any morphism X −→ Y in C, the natural morphism
LFA (K ∨ ⊗LA X) ⊗LA M −→ LFA (K ∨ ⊗LA Y ) ⊗LA M
is an isomorphism in Ho(C). But, using that M ⊗LA K ≃ ∗ and lemma 1.2.10.4, this
is clear by the explicit description of the functor FA . 

1.2.11. Stable modules

We recall that as C is a pointed model category one can define a suspension
functor (see [Ho1, §7])
S : Ho(C) −→ Ho(C)
`L
X 7→ S(X) := ∗ X ∗.
This functor possesses a right adjoint, the loop functor
Ω : Ho(C) −→ Ho(C)
X 7→ Ω(X) := ∗ ×hX ∗.
We fix, once for all an object SC1 ∈ C, which is a cofibrant model for S(1) ∈ Ho(C).
For any commutative monoid A ∈ Comm(C), we let
1
SA := SC1 ⊗ A ∈ A − M od
be the free A-module on SC1 . It is a cofibrant object in A − M od, which is a model
1
for the suspension S(A) (note that SA is cofibrant in A − M od, but not in C unless
A is itself cofibrant in C). The functor
1
SA ⊗A − : A − M od −→ A − M od
has a right adjoint
1
HomA (SA , −) : A − M od −→ A − M od.
1
Furthermore, assumption 1.1.0.2 implies that SA ⊗A − is a left Quillen functor. We
can therefore apply the general construction of [Ho2] in order to produce a model
1
category SpSA (A − M od), of spectra in A − M od with respect to the left Quillen
1
endofunctor SA ⊗A −.
50 1.2. PRELIMINARIES ON LINEAR AND COMMUTATIVE ALGEBRA IN AN HA CONTEXT

Definition 1.2.11.1. Let A ∈ Comm(C) be a commutative monoid in C. The

1
model category of stable A-modules is the model category SpSA (A − M od), of spectra
in A − M od with respect to the left Quillen endo-functor
1
SA ⊗A − : A − M od −→ A − M od.
It will simply be denoted by Sp(A − M od), and its objects will be called stable A-
modules.
Recall that objects in the category Sp(A − M od) are families of objects Mn ∈
1
A − M od for n ≥ 0, together with morphisms σn : SA ⊗A Mn −→ Mn+1 . Morphisms
in Sp(A − M od) are simply families of morphisms fn : Mn → Nn commuting with
the morphisms σn . One starts by endowing Sp(A − M od) with the levelwise model
structure, for which equivalences (resp. fibrations) are the morphisms f : M∗ −→ N∗
such that each morphism fn : Mn −→ Nn is an equivalence in A − M od (resp. a
fibration). The definitive model structure, called the stable model structure, is the
left Bousfield localization of Sp(A − M od) whose local objects are the stable modules
M∗ ∈ Sp(A − M od) such that each induced morphism
1
Mn −→ RHomA (SA , Mn+1 )
is an isomorphism in Ho(A − M od). These local objects will be called Ω-stable A-
modules. We refer to [Ho2] for details concerning the existence and the properties of
this model structure.
There exists an adjunction
SA : A − M od −→ Sp(A − M od) A − M od ←− Sp(A − M od) : (−)0 ,
where the right adjoint sends a stable A-module M∗ to M0 . The left adjoint is
1 ⊗A n
defined by SA (M )n := (SA ) ⊗A M , with the natural transition morphisms. This
adjunction is a Quillen adjunction, and can be derived into an adjunction on the level
of homotopy categories
LSA ≃ SA : Ho(A − M od) −→ Ho(Sp(A − M od))
Ho(A − M od) ←− Ho(Sp(A − M od)) : R(−)0 .
Note that by 1.1.0.2, SA preserves equivalences, so LSA ≃ SA . On the contrary, the
functor (−)0 does not preserve equivalences and must be derived on the right. In
particular, the functor SA : Ho(A − M od) −→ Ho(Sp(A − M od)) is not fully faithful
in general.
Lemma 1.2.11.2. (1) Assume that the suspension functor
S : Ho(C) −→ Ho(C)
is fully faithful. Then, for any commutative monoid A ∈ Comm(C), the
functor
SA : Ho(A − M od) −→ Ho(Sp(A − M od))
is fully faithful.
(2) If furthermore, C is a stable model category then SA is a Quillen equivalence.
Proof. (1) As the adjunction morphism M −→ SA (M )0 is always an isomor-
phism in A − M od, it is enough to show that for any M ∈ A − M od the stable
A-module SA (M ) is a Ω-stable A-module. For this, it is enough to show that for any
M ∈ A − M od, the adjunction morphism
1 1
M −→ RHomA (SA , SA ⊗A M )
 1.2.11. STABLE MODULES 51

is an isomorphism in Ho(C). But, one has natural isomorphisms in Ho(C)

1
SA ⊗A M ≃ SC1 ⊗ M ≃ S(M )
1 1
RHomA (SA , SA ⊗A M ) ≃ RHom1 (S 1 , S(M )) ≃ Ω(S(M )).
This shows that the above morphism is in fact isomorphic in Ho(C) to the adjunction
morphism
M −→ Ω(S(M ))
which is an isomorphism by hypothesis on C.

(2) As C is a stable model category, the functor SC1 ⊗ − : C −→ C is a Quillen

1
equivalence. This also implies that for any A ∈ Comm(C), the functor SA : A−
M od −→ A−M od is a Quillen equivalence. We know by [Ho2] that SA : A−M od −→
Sp(A − M od) is a Quillen equivalence. 
The Quillen adjunction
SA : A − M od −→ Sp(A − M od) A − M od ←− Sp(A − M od) : (−)0 ,
is furthermore functorial in A. Indeed, for A −→ B a morphism in Comm(C), one
defines a functor
− ⊗A B : Sp(A − M od) −→ Sp(B − M od)
defined by
(M∗ ⊗A B)n := Mn ⊗A B.
The transitions morphisms are given by
1 1
SB ⊗B (Mn ⊗A B) ≃ (SA ⊗A Mn ) ⊗A B −→ Mn+1 ⊗A B.
Clearly, the square of left Quillen functors
SA
A − M od / Sp(A − M od)

−⊗A B −⊗A B
 
B − M od / Sp(B − M od)
SB

commutes up to a natural isomorphism. So does the square of right Quillen functors

B − M od / Sp(B − M od)

 
A − M od / Sp(A − M od).

Finally, using techniques of symmetric spectra, as done in [Ho2], it is possible to

show that the homotopy category of stable A-modules inherits from Ho(A − M od) a
symmetric monoidal structure, still denoted by − ⊗LA −. This makes the homotopy
category Ho(Sp(A − M od)) into a closed symmetric monoidal category. In particular,
for two stable A-modules M∗ and N∗ one can define a stable A-modules of morphisms
RHomSp
A (M∗ , N∗ ) ∈ Ho(Sp(A − M od)).

We now consider the category (A − M odop ∧ op

0 ) of pre-stacks over A − M od0 , as
defined in [HAGI]. Recall it is the category of V-simplicial presheaves on A − M odop
0 ,
and that its model structure is obtained from the projective levelwise model structure
by a left Bousfield localization inverting the equivalences in A − M od. The homotopy
52 1.2. PRELIMINARIES ON LINEAR AND COMMUTATIVE ALGEBRA IN AN HA CONTEXT

category Ho((A − M odop ∧

0 ) ) can be naturally identified with the full subcategory of
op
Ho(SP r(A − M od0 )) consisting of functors
F : A − M od0 −→ SSetV ,
sending equivalences of A-modules to equivalences of simplicial sets.
We define a functor
h−
s : Sp(A − M od)
op
−→ (A − M odop
0 )
∧
M∗
M∗ 7→ hs
by
hM
s
∗
: A − M od0 −→ SSetV
N 7→ Hom(M∗ , Γ∗ (SA (N ))),
where Γ∗ is a simplicial resolution functor on the model category Sp(A − M od).
Finally we will need some terminology. A stable A-module M∗ ∈ Ho(Sp(A −
M od)) is called 0-connective, if it is isomorphic to some SA (M ) for an A-module
M ∈ Ho(A − M od0 ). By induction, for an integer n > 0, a stable A-module M∗ ∈
Ho(Sp(A − M od)) is called (−n)-connective, if it is isomorphic to Ω(M∗′ ) for some
−(n − 1)-connective stable A-module M∗′ (here Ω is the loop functor on Ho(Sp(A −
M od))). Note that if the suspension functor is fully faithful, connective stable modules
are exactly connective objects with respect to the natural t-structure on A-modules.
Proposition 1.2.11.3. For any A ∈ Comm(C), the functor h−
s has a total right
derived functor
Rh−
s : Ho(Sp(A − M od))
op
−→ Ho((A − M odop ∧
0 ) ),

which commutes with homotopy limits 1. If the suspension functor

S : Ho(C) −→ Ho(C)
is fully faithful, and if A ∈ A, then for any integer n ≥ 0, the functor Rh− s is fully
faithful when restricted to the full subcategory of (−n)-connective objects.
Proof. As SA : A − M od0 −→ Sp(A − M od) preserves equivalences, one checks
easily that hsM∗ is a fibrant object in (A − M odop ∧
0 ) when M∗ is cofibrant in Sp(A −
QM∗
M od). This easily implies that M∗ 7→ hs is a right derived functor for h−
s , and the
standard properties of mapping spaces imply that it commutes with homotopy limits.
We now assume that the suspension functor
S : Ho(C) −→ Ho(C)
is fully faithful, and that A ∈ A. Let n ≥ 0 be an integer.
Let S n : A − M od −→ A − M od be a left Quillen functor which is a model for the
suspension functor iterated n times (e.g. S n (N ) := SA
n n
⊗A N , where SA 1 ⊗A n
:= (SA ) ).
There is a pullback functor
(S n )∗ : Ho((A − M odop ∧ op ∧
0 ) ) −→ Ho((A − M od0 ) )

defined by (S n )∗ (F )(N ) := F (S n (N )) for any N ∈ A − M od0 (note that S n stablizes

the subcategory A−M od0 because of our assumption 1.1.0.6). For an (−n)-connective
object M∗ , we claim there exists a natural isomorphism in Ho((A−M odop ∧
0 ) ) between

1This makes sense as the functor Rh− is naturally defined on the level of the Dwyer-Kan
s
simplicial localizations with respect to equivalences
LSp(A − M od)op −→ L((A − M odop ∧
0 ) ).
 1.2.11. STABLE MODULES 53

RhM
s
∗
and (S n )∗ (RhM ), where RhM is the value at M of the restricted Yoneda em-
bedding
op ∧
RhM
0 : Ho(A − M od) −→ Ho((A − M od0 ) ).

Indeed, let us write M as Ωn (SA (M )) for some object M ∈ Ho(A − M od), where
n
Ω is the loop functor of Sp(A − M od), iterated n times. Then, using our lemma
Lem. 1.2.11.2, for any N ∈ Ho(A − M od), we have natural isomorphisms in Ho(SSet)

M apSp(A−Mod) (M∗ , SA (N )) ≃ M apSp(A−Mod) (SA (M ), S n (SA (N ))) ≃

≃ M apA−Mod (M, S n (N )),

where S n denotes the suspension functor iterated n-times. Using that A is C0 -good,
this shows that
RhsM∗ ≃ (S n )∗ (RhM
0 ).

Moreover, for any (−n)-connective objects M∗ and N∗ in Ho(Sp(A − M od)) one has
natural isomorphisms in Ho(SSet)

M apSp(A−Mod) (M∗ , N∗ ) ≃ M apA−Mod (M, N )

where M∗ ≃ Ωn (SA (M )) and N∗ ≃ Ωn (SA (N )). We are therefore reduced to show

that for any A-modules M and N in Ho(A−M od), the natural morphism in Ho(SSet)
n ∗ N n ∗ M
M apA−Mod (M, N ) −→ M ap(A−Modop
0 )
∧ ((S ) (Rh0 ), (S ) (Rh0 ))

is an isomorphism. To see this, we define a morphism in the opposite direction in the

following way. Taking the n-th loop functor on each side gives a morphism
n ∗ N n ∗ M
M ap(A−Modop ∧ ((S ) (Rh0 ), (S ) (Rh0 )) →
0 )

n n ∗ N n n ∗ M
−→ M ap(A−Modop
0 )
∧ (Ω (S ) (Rh0 ), Ω (S ) (Rh0 ).

Moreover, there are isomorphisms

Ωn M apA−Mod (M, S n (P )) ≃ M apA−Mod (M, Ωn (S n (P ))) ≃ M apA−Mod (M, P ) ≃ RhM

0 (P ),

showing that there exists a natural isomorphism in Ho((A − M odop ∧

0 ) ) between
M M
Ω (S ) (Rh0 ) and Rh0 . One therefore gets a morphism
n n ∗

n ∗ N n ∗ M
M ap(A−Modop ∧ ((S ) (Rh0 ), (S ) (Rh0 )) →
0 )

n n ∗ N n n ∗ M
−→ M ap(A−Modop
0 )
∧ (Ω (S ) (Rh0 ), Ω (S ) (Rh0 ) ≃

N M
≃ M ap(A−Modop ∧ (Rh0 , Rh0 ).
0 )

Using that A is C0 -good we get the required morphism

n ∗ N n ∗ M
M ap(A−Modop ∧ ((S ) (Rh0 ), (S ) (Rh0 )) −→ M apA−Mod (M, N ),
0 )

and it is easy to check it is an inverse in Ho(SSet) to the natural morphism

n ∗ N n ∗ M
M apA−Mod (M, N ) −→ M ap(A−Modop
0 )
∧ ((S ) (Rh0 ), (S ) (Rh0 )).


54 1.2. PRELIMINARIES ON LINEAR AND COMMUTATIVE ALGEBRA IN AN HA CONTEXT

1.2.12. Descent for modules and stable modules

In this last section we present some definitions concerning descent for modules and
stable modules in our general context. In a few words, a co-augmented co-simplicial
object A −→ B∗ in Comm(C), is said to have the descent property for modules (resp.
for stable modules) if the homotopy theory of A-modules (resp. of stable A-modules)
is equivalent to the homotopy theory of certain co-simplicial B∗ -modules (resp. stable
B∗ -modules). From the geometric, dual point of view, the object A −→ B∗ should be
thought as an augmented simplicial space Y∗ −→ X, and having the descent property
essentially means that the theory of sheaves on X is equivalent to the theory of certain
sheaves on Y∗ (see [SGA4-II, Exp. V bis ] for more details on this point of view).
The purpose of this section is only to introduce the basic set-up for descent that
will be used later to state that the theory of quasi-coherent modules is local with
respect to the topology, or in other words that hypercovers have the descent property
for modules. This is a very important property allowing local-to-global arguments.

Let A∗ be a co-simplicial object in the category Comm(C) of commutative monoids.

Therefore, A∗ is given by a functor
∆ −→ Comm(C)
[n] 7→ An .
A co-simplicial A∗ -module M∗ is by definition the following datum.
• A An -module Mn ∈ An − M od for any n ∈ ∆.
• For any morphism u : [n] → [m] in ∆, a morphism of An -modules αu :
u / v
Mn −→ Mm , such that αv ◦ αu = αv◦u for any [n] [m] / [p] in
∆.
In the same way, a morphism of co-simplicial A∗ -modules f : M∗ −→ N∗ is the data
of morphisms fn : Mn −→ Nn for any n, commuting with the α’s
αN M
u ◦ fn = fm ◦ αu

for any u : [n] → [m].

The co-simplicial A∗ -modules and morphisms of A∗ -modules form a category, de-
noted by csA∗ − M od. It is furthermore a U-combinatorial model category for which
the equivalences (resp. fibrations) are the morphisms f : M∗ −→ N∗ such that each
fn : Mn −→ Nn is an equivalence (resp. a fibration) in An − M od.

Let A be a commutative monoid, and B∗ be a co-simplicial commutative A-

algebra. We can also consider B∗ as a co-simplicial commutative monoid together
with a co-augmentation
A −→ B∗ ,
which is a morphism of co-simplicial objects when A is considered as a constant co-
simplicial object. As a co-simplicial commutative monoid A possesses a category of
co-simplicial A∗ -modules csA − M od which is nothing else than the model category
of co-simplicial objects in A − M od with its projective levelwise model structure.
For a co-simplicial A-module M∗ , we define a co-simplicial B∗ -module B∗ ⊗A M∗
by the formula
(B∗ ⊗A M∗ ) := Bn ⊗A Mn ,
and for which the transitions morphisms are given by the one of B∗ and of M∗ . This
construction defines a functor
B∗ ⊗A − : csA − M od −→ csB∗ − M od,
 1.2.12. DESCENT FOR MODULES AND STABLE MODULES 55

which has a right adjoint

csB∗ − M od −→ csA − M od.
This right adjoint, sends a B∗ -module M∗ to its underlying co-simplicial A-module.
Clearly, this defines a Quillen adjunction. There exists another adjunction
ct : Ho(A−M od) −→ Ho(csA−M od) Ho(A−M od) ←− Ho(csA−M od) : Holim,
where Holim is defined as the total right derived functor of the functor lim (see [Hi,
8.5], [DS, 10.13]).
Composing these two adjunctions gives a new adjunction
Z
B∗ ⊗LA − : Ho(A−M od) −→ Ho(csB∗ −M od) Ho(A−M od) ←− Ho(csB∗ −M od) : .

Definition 1.2.12.1. (1) Let B∗ be a co-simplicial commutative monoid

and M∗ be a co-simplicial B∗ -module. We say that M∗ is homotopy cartesian
if for any u : [n] → [m] in ∆ the morphism, induced by αu ,
Mn ⊗LBn Bm −→ Mm
is an isomorphism in Ho(Bm − M od).
(2) Let A be a commutative monoid and B∗ be a co-simplicial commutative A-
algebra. We say that the co-augmentation morphism A −→ B∗ satisfies the
descent condition, if in the adjunction
Z
B∗ ⊗LA − : Ho(A−M od) −→ Ho(csB∗ −M od) Ho(A−M od) ←− Ho(csB∗ −M od) :

the functor B∗ ⊗LA − is fully faithful and induces an equivalence between

Ho(A − M od) and the full subcategory of Ho(csB∗ − M od) consisting of
homotopy cartesian objects.
Remark 1.2.12.2. If a morphism A −→ B∗ satisfies the descent condition, then
so does any co-simplicial object equivalent to it (as a morphism in the model category
of co-simplicial commutative monoids).
Lemma 1.2.12.3. Let A −→ B∗ be a co-augmented co-simplicial commutative
monoid in Comm(C) which satisfies the descent condition. Then, the natural mor-
phism
A −→ Holimn Bn
is an isomorphism in Ho(Comm(C)).
R
Proof. This is clear as B∗ = Holimn Bn . 
We now pass to descent for stable modules. For this, let again A∗ be a co-
simplicial object in the category Comm(C) of commutative monoids. A co-simplicial
stable A∗ -module M∗ is by definition the following datum.
• A stable An -module Mn ∈ Sp(An − M od) for any n ∈ ∆.
• For any morphism u : [n] → [m] in ∆, a morphism of stable An -modules
u / v
αu : Mn −→ Mm , such that αv ◦ αu = αv◦u for any [n] [m] / [p]
in ∆.
In the same way, a morphism of co-simplicial stable A∗ -modules f : M∗ −→ N∗ is the
data of morphisms fn : Mn −→ Nn in Sp(An − M od) for any n, commuting with the
α’s
αN M
u ◦ fn = fm ◦ αu
for any u : [n] → [m].
56 1.2. PRELIMINARIES ON LINEAR AND COMMUTATIVE ALGEBRA IN AN HA CONTEXT

The co-simplicial stable A∗ -modules and morphisms of A∗ -modules form a cate-

gory, denoted by Sp(csA∗ −M od). It is furthermore a U-combinatorial model category
for which the equivalences (resp. fibrations) are the morphisms f : M∗ −→ N∗ such
that each fn : Mn −→ Nn is an equivalence (resp. a fibration) in Sp(An − M od).
We note that Sp(csA∗ − M od) is also naturally equivalent as a category to the cate-
1
gory of SA -spectra in csA∗ M od, hence the notation Sp(csA∗ −M od) is not ambiguous.

Let A be a commutative monoid, and B∗ be a co-simplicial commutative A-

algebra. We can also consider B∗ as a co-simplicial commutative monoid together
with a co-augmentation
A −→ B∗ ,
which is a morphism of co-simplicial objects when A is considered as a constant co-
simplicial object. As a co-simplicial commutative monoid A possesses a category of
co-simplicial stable A∗ -modules Sp(csA − M od) which is nothing else than the model
category of co-simplicial objects in Sp(A − M od) with its projective levelwise model
structure.
For a co-simplicial stable A-module M∗ , we define a co-simplicial stable B∗ -
module B∗ ⊗A M∗ by the formula
(B∗ ⊗A M∗ ) := Bn ⊗A Mn ,
and for which the transitions morphisms are given by the one of B∗ and of M∗ . This
construction defines a functor
B∗ ⊗A − : Sp(csA − M od) −→ Sp(csB∗ − M od),
which has a right adjoint
Sp(csB∗ − M od) −→ Sp(csA − M od).
This right adjoint, sends a co-simplicial stable B∗ -module M∗ to its underlying co-
simplicial stable A-module. Clearly, this is a Quillen adjunction. One also has an
adjunction
ct : Ho(Sp(A − M od)) −→ Ho(Sp(csA − M od))
Ho(Sp(A − M od)) ←− Ho(Sp(csA − M od)) : Holim,
where Holim is defined as the total right derived functor of the functor lim (see [Hi,
8.5], [DS, 10.13]).
Composing these two adjunctions gives a new adjunction
B∗ ⊗LA − : Ho(Sp(A − M od)) −→ Ho(Sp(csB∗ − M od))
Z
Ho(Sp(A − M od)) ←− Ho(Sp(csB∗ − M od)) : .

Definition 1.2.12.4. (1) Let B∗ be a co-simplicial commutative monoid

and M∗ be a co-simplicial stable B∗ -module. We say that M∗ is homotopy
cartesian if for any u : [n] → [m] in ∆ the morphism, induced by αu ,
Mn ⊗LBn Bm −→ Mm
is an isomorphism in Ho(Sp(Bm − M od)).
(2) Let A be a commutative monoid and B∗ be a co-simplicial commutative A-
algebra. We say that the co-augmentation morphism A −→ B∗ satisfies the
stable descent condition, if in the adjunction
B∗ ⊗LA − : Ho(Sp(A − M od)) −→ Ho(Sp(csB∗ − M od))
Z
Ho(Sp(A − M od)) ←− Ho(Sp(csB∗ − M od)) :
 1.2.13. COMPARISON WITH THE USUAL NOTIONS 57

the functor B∗ ⊗LA − is fully faithful and induces an equivalence between

Ho(Sp(A−M od)) and the full subcategory of Ho(Sp(csB∗ −M od)) consisting
of homotopy cartesian objects.
The following proposition insures that the unstable descent condition implies the
stable one under certain conditions.
Proposition 1.2.12.5. Let A be a commutative monoid and B∗ be a co-simplicial
commutative A-algebra, such that A −→ B∗ satisfies the descent condition. Assume
that the two following conditions are satisfied.
(1) The suspension functor S : Ho(C) −→ Ho(C) is fully faithful.
(2) For any n, the morphism A −→ Bn is flat in the sense of Def. 1.2.4.1.
Then A −→ B∗ satisfies the stable descent property.
Proof. The second condition insures that the diagram
Ho(Sp(A − M od)) / Ho(Sp(csB∗ − M od)

 
Ho(A − M od) / Ho(csB∗ − M od)

commutes up to a natural isomorphism (here the vertical functors are the right adjoint
to the suspensions inclusion functors, and send a spectrum to its 0-th level). So, for
any M ∈ Sp(A − M od), the adjunction morphism
Z
M −→ (B∗ ⊗L M )

is easily seen to induces an isomorphism on the 0-th level objects in Ho(A − M od). As
this is true for any M , and in particular for the suspensions of M , we find that this
adjunction morphism induces isomorphisms on each n-th level objects in Ho(A−M od).
Therefore it is an isomorphism. In the same way, one proves that for any cartesian
object M∗ ∈ Ho(Sp(csB∗ − M od)) the adjunction morphism
Z
B∗ ⊗A (M∗ ) −→ M∗

is an isomorphism. 

1.2.13. Comparison with the usual notions

In this last section we present what the general notions introduced before give,
when C is the model category of k-modules with the trivial model structure. The
other non trivial examples will be given at the beginning of the various chapters §2.2,
§2.3 and §2.4 where the case of simplicial modules, complexes and symmetric spectra
will be studied.

Let k be a commutative ring in U, and we let C be the category of k-modules

belonging to the universe U, endowed with its trivial model structure for which equiv-
alences are isomorphisms and all morphisms are fibrations and cofibrations. The cat-
egory C is then a symmetric monoidal model category for the tensor product − ⊗k −.
Furthermore, all of our assumption 1.1.0.1, 1.1.0.3, 1.1.0.2 and 1.1.0.4 are satisfied.
The category Comm(C) is of course the category of commutative k-algebras in U,
endowed with its trivial model structure. In the same way, for any commutative k-
algebra A, the category A − M od is the usual category of A-modules in U together
with its trivial model structure. We set C0 = C and A = Comm(C). The notions we
58 1.2. PRELIMINARIES ON LINEAR AND COMMUTATIVE ALGEBRA IN AN HA CONTEXT

have presented before restrict essentially to the usual notions of algebraic geometry
with some remarkable caveats.
• For any morphism of commutative k-algebras A −→ B and any B-module
M , the simplicial set DerA (B, M ) is discrete and naturally isomorphic to
the set DerA (B, M ) of derivations from B to M over A. Equivalently, the
B-module LB/A (defined in Def. 1.2.1.5) is simply the usual B-module of
Kähler differentials Ω1B/A . Note that this LB/A is not the Quillen-Illusie
cotangent complex of A → B.
• For any morphism of commutative k-algebras A −→ B the natural morphism
B −→ T HH(B/A) is always an isomorphism. Indeed

T HH(A) ≃ A ⊗A⊗k A A ≃ A.

• A morphism of commutative k-algebras A −→ B is finitely presented in the

sense of Def. 1.2.3.1 if and only if it B is a finitely presented A-algebra in
the usual sense. In the same way, finitely presented objects in A − M od in
the sense of Def. 1.2.3.1 are the finitely presented A-modules. Also, perfect
objects in A − M od are the projective A-modules of finite type.
• A morphism A −→ B of commutative k-algebras is a formal covering if and
only if it is a faithful morphism of rings.
• Let f : A −→ B be a morphism of commutative k-algebras, and Spec B −→
Spec A the associated morphism of schemes. We have the following compar-
ison board.

In the sense of Def . 1.2.6.1, 1.2.6.7, 1.2.7.1 As a morphism of affine schemes

epimorphism monomorphism
f lat f lat
f ormal Zariski open immersion f lat monomorphism
Zariski open immersion open immersion
(f ormally) unramif ied (f ormally) unramif ied
f ormally thh − etale always satisf ied
f ormally i − smooth always satisf ied
thh − etale f initely presented

The reader will immediately notice the absence of (formally) smooth

and (formally) étale maps in the previous table. This is essentially due
to the fact that there are no easy general characterizations of this maps in
terms of the module of Kähler differentials alone (which in this trivial model
structure context is our cotangent complex). On the contrary such charac-
terizations do exist in terms of the “correct” cotangent complex which is
the Quillen-Illusie one. But this correct cotangent complex of a morphism
of usual commutative k-algebras A → B will appear as the cotangent com-
plex according to our definition 1.2.1.5 only if we consider this morphism
in the category of simplicial k-modules, i.e. if we replace the category C of
k-modules with the category of simplicial k-modules. In other words our
definitions of (formally) smooth and (formally) étale maps reduce to the
usual ones (e.g. [EGAIV]) between commutative k-algebras only if we con-
sider them in the context of derived algebraic geometry (see §2.2). This is
consistent with the general philosophy that some aspects of usual algebraic
 1.2.13. COMPARISON WITH THE USUAL NOTIONS 59

geometry, especially those related to infinitesimal lifting properties and de-

formation theory, are conceptually more transparent in (and actually already
a part of) derived algebraic geometry. See also Remark 2.1.2.2 for another
instance of this point of view.
• For commutative monoid A, one has π∗ (A) = A, and for any a ∈ A the
commutative A-algebra A −→ A[a−1 ] is the usual localization of A inverting
a.
• One has that AK ≃ 0 for any projective A-module K of constant finite rank
n > 0 (for example if the scheme Spec A is connected and K is non zero).
 CHAPTER 1.3

Geometric stacks: Basic theory

In this chapter, after a brief reminder of [HAGI], we present the key definition
of this work, the notion of n-geometric stack. The definition we present here is a
generalization of the original notion of geometric n-stack introduced by C. Simpson
in [S3]. As already remarked in [S3], the notion of n-geometric stack only depends on
a topology on the opposite category of commutative monoids Comm(C)op , and on a
class P of morphisms. Roughly speaking, geomeric stacks are the stacks obtained by
taking quotient of representable stacks by some equivalence relations in P. By choos-
ing different classes P one gets different notions of geometric stacks. For example, in
the classical situation where C = Z − M od, and the topology is chosen to be the étale
topology, Deligne-Mumford algebraic stacks correspond to the case where P is the
class of étale morphisms, whereas Artin algebraic stacks correspond to the case where
P is the class of smooth morphisms. We think it is important to leave the choice
of the class P open in the general definition, so that it can be specialized differently
depending of the kind of objects one is willing to consider.

From the second section on, we will fix a HA context (C, C0 , A), in the sense of
Def. 1.1.0.11.

1.3.1. Reminders on model topoi

To make the paper essentially self-contained, we briefly summarize in this sub-
section the basic notions and results of the theory of stacks in homotopical contexts
as exposed in [HAGI]. We will limit ourselves to recall only the topics that will be
needed in the sequel; the reader is addressed to [HAGI] for further details and for
proofs.

Let M be a U-small model category, and W (M ) its class of weak equivalences.

op
We let SP r(M ) := SSetsM be the category of simplicial presheaves on M with its
projective model structure, i.e. with equivalences and fibrations defined objectwise.
The model category of prestacks M ∧ on M is the model category obtained as the
left Bousfield localization of SP r(M ) at {hu | u ∈ W (M )}, where h : M → P r(M ) ֒→
SP r(M ) is the (constant) Yoneda embedding. The homotopy category Ho(M ∧ ) can
be identified with the full subcategory of Ho(SP r(M )) consisting of those simplicial
presheaves F on M that preserve weak equivalences ([HAGI, Def. 4.1.4]); any such
simplicial presheaf (i.e. any object in Ho(M ∧ )) will be called a prestack on M . M ∧
is a U-cellular and U-combinatorial ([HAGI, App. A]) simplicial model category and
its derived simplicial Hom’s will be denoted simply by RHom (denoted as Rw Hom
in [HAGI, §4.1]).

If Γ∗ : M → M ∆ is a cofibrant resolution functor for M ([Hi, 16.1]), and we

define
h : M → M ∧ : x 7−→ (hx : y 7→ HomM (Γ∗ (y), x)),
61
62 1.3. GEOMETRIC STACKS: BASIC THEORY

we have that h preserves fibrant objects and weak equivalences between fibrant objects
([HAGI, Lem. 4.2.1]). Therefore we can right-derive h to get a functor Rh := h ◦ R :
Ho(M ) → Ho(M ∧ ), where R is a fibrant replacement functor in M ; Rh is in fact fully
faithful ([HAGI, Thm. 4.2.3]) and is therefore called the (model ) Yoneda embedding
for the model category M (Rh, as opposed to h, does not depend, up to a unique
isomorphism, on the choice of the cofibrant resolution functor Γ∗ ).
We also recall that the canonical morphism hx → Rhx is always an isomorphism
in Ho(M ∧ ) ([HAGI, Lem. 4.2.2]), and that with the notations introduced above for
the derived simplicial Hom’s in M ∧ , the model Yoneda lemma ([HAGI, Cor. 4.2.4])
is expressed by the isomorphisms in Ho(SSets)
RHom(Rhx , F ) ≃ RHom(hx , F ) ≃ F (x)
for any fibrant object F in M ∧ .

A convenient homotopical replacement of the notion of a Grothendieck topology

in the case of model categories, is the following ([HAGI, Def. 4.3.1])
Definition 1.3.1.1. A model (pre-)topology τ on a U-small model category M ,
is the datum for any object x ∈ M , of a set Covτ (x) of subsets of objects in Ho(M )/x,
called τ -covering families of x, satisfying the following three conditions
(1) (Stability) For all x ∈ M and any isomorphism y → x in Ho(M ), the
one-element set {y → x} is in Covτ (x).
(2) (Composition) If {ui → x}i∈I ∈ Covτ (x), and for any i ∈ I, {vij →
ui }j∈Ji ∈ Covτ (ui ), the family {vij → x}i∈I,j∈Ji is in Covτ (x).
(3) (Homotopy base change) Assume the two previous conditions hold. For any
{ui → x}i∈I ∈ Covτ (x), and any morphism in Ho(M ), y → x, the family
{ui ×hx y → y}i∈I is in Covτ (y).
A U-small model category M together with a model pre-topology τ will be called a
U-small model site.
By [HAGI, Prop. 4.3.5] a model pre-topology τ on M induces and is essentially
the same thing as a Grothendieck topology, still denoted by τ , on the homotopy cat-
egory Ho(M ).

Given a model site (M, τ ) we have, as in [HAGI, Thm. 4.6.1], a model cate-
gory M ∼,τ (U-combinatorial and left proper) of stacks on the model site, which is
defined as the left Bousfield localization of the model category M ∧ of prestacks on M
along a class Hτ of homotopy τ -hypercovers ([HAGI, 4.4, 4.5]). To any prestack F
we can associate a sheaf π0 of connected components on the site (Ho(M ), τ ) defined
as the associated sheaf to the presheaf x 7−→ π0 (F (x)). In a similar way ([HAGI,
Def. 4.5.3]), for any i > 0, any fibrant object x ∈ M , and any s ∈ F (x)0 , we can
define a sheaf of homotopy groups πi (F, s) on the induced comma site (Ho(M/x), τ ).
The weak equivalences in M ∼,τ turn out to be exactly the π∗ -sheaves isomorphisms
([HAGI, Thm. 4.6.1]), i.e. those maps u : F → G in M ∧ inducing an isomorphism
of sheaves π0 (F ) ≃ π0 (G) on (Ho(M ), τ ), and isomorphisms πi (F, s) ≃ πi (G, u(s)) of
sheaves on (Ho(M/x), τ ) for any i ≥ 0, for any choice of fibrant x ∈ M and any base
point s ∈ F (x)0 .

The left Bousfield localization construction defining M ∼,τ yields a pair of adjoint
Quillen functors
Id : M ∧ −→ M ∼,τ M ∧ ←− M ∼,τ : Id
 1.3.1. REMINDERS ON MODEL TOPOI 63

which induces an adjunction pair at the level of homotopy categories

a := LId : Ho(M ∧ ) −→ Ho(M ∼,τ ) Ho(M ∧ ) ←− Ho(M ∼,τ ) : j := RId
where j is fully faithful.
Definition 1.3.1.2. (1) A stack on the model site (M, τ ) is an object F ∈
SP r(M ) whose image in Ho(M ∧ ) is in the essential image of the functor j.
(2) If F and G are stacks on the model site (M, τ ), a morphism of stacks is
a morphism F → G in in Ho(SP r(M )), or equivalently in Ho(M ∧ ), or
equivalently in Ho(M ∼,τ ).
(3) A morphism of stacks f : F −→ G is a covering (or a cover or an epimor-
phism) if the induced morphism of sheaves
π0 (f ) : π0 (F ) −→ π0 (G)
is an epimorphism in the category of sheaves.
Recall that a simplicial presheaf F : M op → SSetV is a stack if and only if
it preserves weak equivalences and satisfy a τ -hyperdescent condition (descent, i.e.
sheaf-like, condition with respect to the class Hτ of homotopy hypercovers): see
[HAGI, Def. 4.6.5 and Cor. 4.6.3]. We will always consider Ho(M ∼,τ ) embedded in
Ho(M ∧ ) embedded in Ho(SP r(M )), omitting in particular to mention explicitly the
functor j above. With this conventions, the functor a above becomes an endofunctor of
Ho(M ∧ ), called the associated stack functor for the model site (M, τ ). The associated
stack functor preserves finite homotopy limits and all homotopy colimits ([HAGI,
Prop. 4.6.7]).
Definition 1.3.1.3. A model pre-topology τ on M is sub-canonical if for any
x ∈ M the pre-stack Rhx is a stack.
M ∼,τ is a left proper (but not right proper) simplicial model category and its de-
rived simplicial Hom’s will be denoted by Rτ Hom (denoted by Rw, τ Hom in [HAGI,
Def. 4.6.6]); for F and G prestacks on M , there is always a morphism in Ho(SSet)
Rτ Hom(F, G) → RHom(F, G)
which is an isomorphism when G is a stack ([HAGI, Prop.4.6.7]).
Moreover M ∼,τ is a t-complete model topos ([HAGI, Def. 3.8.2]) therefore possesses
important exactness properties. For the readers’ convenience we collect below (from
[HAGI]) the definition of (t-complete) model topoi and the main theorem character-
izing them (Giraud-like theorem).

For a U-combinatorial model category N , and a U-small set S of morphisms in

N , we denote by LS N the left Bousfield localization of N along S. It is a model
category, having N as underlying category, with the same cofibrations as N and
whose equivalences are the S-local equivalences ([Hi, Ch. 3]).
Definition 1.3.1.4. (1) An object x in a model category N is truncated if
there exists an integer n ≥ 0, such that for any y ∈ N the mapping space
MapN (y, x) ([Hi, §17.4]) is a n-truncated simplicial set
(i.e. πi (MapN (x, y), u) = 0 for all i > n and for all base point u).
(2) A model category N is t-complete if a morphism u : y → y ′ in Ho(N ) is an
isomorphism if and only if the induced map u∗ : [y ′ , x] → [y, x] is a bijection
for any truncated object x ∈ N .
Recall that for any U-small S-category T (i.e. a category T enriched over sim-
plicial sets, [HAGI, Def. 2.1.1]), we can define a U-combinatorial model category
64 1.3. GEOMETRIC STACKS: BASIC THEORY

SP r(T ) of simplicial functors T op −→ SSetU , in which equivalences and fibrations

are defined levelwise ([HAGI, Def. 2.3.2]).
Definition 1.3.1.5. ([HAGI, §3.8]) A U-model topos is a U-combinatorial model
category N such that there exists a U-small S-category T and a U-small set of mor-
phisms S in SP r(T ) satisfying the following two conditions.
(1) The model category N is Quillen equivalent to LS SP r(T ).
(2) The identity functor
Id : SP r(T ) −→ LS SP r(T )
preserves homotopy pullbacks.
A t-complete model topos is a U-model topos N which is t-complete as a model
category.
We need to recall a few special morphisms in the standard simplicial category ∆.
For any n > 0, and 0 ≤ i < n we let
σi : [1] −→ [n]
0 7→ i
1 7→ i + 1.
Definition 1.3.1.6. Let N be a model category. A Segal groupoid object in N
is a simplicial object
X∗ : ∆op −→ N
satisfying the following two conditions.
(1) For any n > 0, the natural morphism
Y
σi : Xn −→ X1 ×hX0 X1 ×hX0 · · · ×hX0 X1
0≤i<n
| {z }
n times

is an isomorphism in Ho(N ).
(2) The morphism
d0 × d1 : X2 −→ X1 ×hd0 ,X0 ,d0 X1
is an isomorphism in Ho(N ).
The homotopy category of Segal groupoid objects in N is the full subcategory of
op
Ho(N ∆ ) consisting of Segal groupoid objects. It is denoted by Ho(SeGpd(N )).
The main theorem characterizing model topoi is the following analog of Giraud’s
theorem.
Theorem 1.3.1.7. ([HAGI, Thm. 4.9.2]) A model category N is a model topos
if and only if it satisfies the following conditions.
(1) The model category N is U-combinatorial.
(2) For any U-small family of objects {xi }i∈I in N , and any i 6= j in I the
following square
∅ / xi

 
xj / `L xk
k∈I

is homotopy cartesian.
 1.3.2. HOMOTOPICAL ALGEBRAIC GEOMETRY CONTEXT 65

(3) For any U-small category I, any morphism y → z and any I-diagram x :
I −→ N/z, the natural morphism
Hocolimi∈I (xi ×hz y) −→ (Hocolimi∈I xi ) ×hz y
is an isomorphism in Ho(N ).
(4) For any Segal groupoid object (in the sense of Def. 1.3.1.6)
X∗ : ∆op −→ N,
the natural morphism
X1 −→ X0 ×h|X∗ | X0
is an isomorphism in Ho(N ).
An important consequence is the following
Corollary 1.3.1.8. For any U-model topos N and any fibrant object x ∈ N , the
category Ho(N/x) is cartesian closed.
The exactness properties of model topoi will be frequently used all along this
work. For instance, we will often use that for any cover of stacks p : F −→ G (over
some model site (M, τ )), the natural morphism
|F∗ | −→ G
is an isomorphism of stacks, where F∗ is the homotopy nerve of p (i.e. the nerve of
a fibration equivalent to p, computed in the category of simplicial presheaves). This
result is also recalled in Lem. 1.3.4.3.

1.3.2. Homotopical algebraic geometry context

Let us fix a HA context (C, C0 , A).

We denote by Af fC the opposite of the model category Comm(C): this will be

our base model category M to which we will apply the [HAGI] constructions recalled
in §1.3.1.
An object X ∈ Af fC corresponding to a commutative monoid A ∈ Comm(C)
will be symbolically denoted by X = Spec A. We will consider the model category
Af fC∧ of pre-stacks on Af fC as described in §1.3.1 above. By definition, it is the
Af f op
left Bousfield localization of SP r(Af fC ) := SSetV C (the model category of V-
simplicial presheaves on Af fC ) along the (V-small) set of equivalences of Af fC , and
the homotopy category Ho(Af fC∧ ) will be naturally identified with the full subcate-
gory of Ho(SP r(Af fC )) consisting of all functors F : Af fCop −→ SSetV preserving
weak equivalences. Objects in Ho(Af fC∧ ) will be called pre-stacks, and the derived
simplicial Hom’s of the simplicial model category Af fC∧ will be denoted by RHom.
We will fix once for all a model pre-topology τ on Af fC (Def. 1.3.1.1), which
induces a Grothendieck topology on Ho(Af fC ), still denoted by the same symbol. As
recalled in §1.3.1.1, one can then consider a model category Af fC∼,τ , of stacks on the
model site (Af fC , τ ). A morphism F −→ G of pre-stacks is an equivalence in Af fC∼,τ
if it induces isomorphisms on all homotopy sheaves (for any choice of X ∈ Af fC and
any s ∈ F (X)).

To ease the notation we will write St(C, τ ) for the homotopy category Ho(Af fC∼,τ )
of stacks.
66 1.3. GEOMETRIC STACKS: BASIC THEORY

The Bousfield localization construction yields an adjunction

a : Ho(Af fC∧ ) −→ St(C, τ ) Ho(Af fC∧ ) ←− St(C, τ ) : j
where j is fully faithful.
Definition 1.3.2.1. (1) A stack is an object F ∈ SP r(Af fC ) whose image
in Ho(Af fC∧ ) is in the essential image of the functor j above.
(2) A morphism of stacks is a morphism between stacks in Ho(SP r(Af fC )), or
equivalently in Ho(Af fC∧ ), or equivalently in St(C, τ ).
(3) A morphism of stacks f : F −→ G is a covering (or a cover) if the induced
morphism of sheaves
π0 (f ) : π0 (F ) −→ π0 (G)
is an epimorphism in the category of sheaves.
We will always omit mentioning the functor j and consider the category St(C, τ )
as embedded in Ho(Af fC∧ ), and therefore embedded in Ho(SP r(Af fC )). With these
conventions, the endofunctor a of Ho(Af fC∧ ) becomes the associated stack functor,
which commutes with finite homotopy limits and arbitrary homotopy colimits.
A functor F : Af fCop −→ SSetV is a stack (§1.3.1) if and only if it preserves equiv-
alences and possesses the descent property with respect to homotopy τ -hypercovers.
The derived simplicial Hom’s in the model category Af fC∼,τ of stacks will be denoted
by Rτ Hom. The natural morphism
RHom(F, G) −→ Rτ Hom(F, G)
is an isomorphism in Ho(SSet) when G is a stack.
The model category Af fC∼,τ is a t-complete model topos (Def. 1.3.1.5). We warn
the reader that neither of the model categories Af fC∧ and Af fC∼,τ is right proper,
though they are both left proper. Because of this certain care has to be taken when
considering homotopy pullbacks and more generally homotopy limit constructions,
as well as comma model categories of objects over a base object. Therefore, even
when nothing is specified, adequate fibrant replacement may have been chosen before
considering certain constructions.

The model Yoneda embedding (§1.3.1)

h : Af fC −→ Af fC∧
has a total right derived functor
Rh : Ho(Af fC ) −→ Ho(Af fC∧ ),
which is fully faithful. We also have a naive Yoneda functor
h : Af fC −→ Af fC∧
sending an object X ∈ Af fC to the presheaf of sets it represents (viewed as a simplicial
presheaf). With these notations, the Yoneda lemma reads
RHom(RhX , F ) ≃ RHom(hX , F ) ≃ F (X)
for any fibrant object F ∈ Af fC∧ . The natural morphism hX −→ RhX is always an
isomorphism in Ho(Af fC∧ ) for any X ∈ Af fC .
We will also use the notation Spec A for hSpec A . We warn the reader that Spec A
lives in Af fC whereas Spec A is an object of the model category of stacks Af fC∼,τ .

We will assume the topology τ satisfies some conditions. In order to state them,
recall the category sAf fC of simplicial objects in Af fC is a simplicial model category
 1.3.2. HOMOTOPICAL ALGEBRAIC GEOMETRY CONTEXT 67

for the Reedy model structure. Therefore, for any object X∗ ∈ sAf fC and any U-
RK
small simplicial set K we can define an object X∗ ∈ Ho(sAf fC ), by first taking a
RK
Reedy fibrant model for X∗ and then the exponential by K. The zero-th part of X∗
will be simply denoted by
X∗RK := (X∗RK )0 ∈ Ho(Af fC ).
We also refer the reader to [HAGI, §4.4] for more details and notations.

The following assumption on the pre-topology will be made.

Assumption 1.3.2.2. (1) The topology τ on Ho(Af fC ) is quasi-compact.
In other words, for any covering family {Ui −→ X}i∈I in Af fC there exists
a finite subset I0 ⊂ I such that the induced family {Ui −→ X}i∈I0 is a
covering.
(2) For any finite family of objects {Xi }i∈I in Af fC (including the empty family)
the family of morphisms
L
a
{Xi −→ Xj }i∈I
j∈I
`L
form a τ -covering family of j∈I Xj .
(3) Let X∗ −→ Y be an augmented simplicial object in Af fC , corresponding to a
co-augmented co-simplicial object A −→ B∗ in Comm(C). We assume that
for any n, the one element family of morphisms
n
Xn −→ X∗R∂∆ ×hY R∂∆n Y
form a τ -covering family in Af fC . Then the morphism
A −→ B∗
satisfies the descent condition in the sense of Def. 1.2.12.1.
The previous assumption has several consequences on the homotopy theory of
stacks. They are subsumed in the following lemma.
Lemma 1.3.2.3. (1) For any finite family of objects Xi in Af fC the natural
morphism
a
RhXi −→ Rh` L Xi
i
i

is an equivalence in Af fC∼,τ .
(2) Let H be the (V-small) set of augmented simplicial objects X∗ −→ Y in
Af fC such that for any n the one element family of morphisms
n
Xn −→ X∗R∂∆ ×hY R∂∆n Y
is a τ -covering family in Af fC . Then, the model category Af fC∼,τ is the left
Bousfield localization of Af fC∧ along the set of morphisms
a
|RhX∗ | −→ RhY RhUi −→ Rh` Li Ui
i

where X∗ → Y runs in H and {Ui } runs through the set of all finite families
of objects in Af fC .
68 1.3. GEOMETRIC STACKS: BASIC THEORY

Proof. (1) The case where the set of indices I is empty follows from our as-
sumption 1.3.2.2 (2) with I empty, as it states that the empty family covers the
initial object on Af fC .
Let us assume that the set of indices is not empty. By induction, it is clearly
enough to treat the case where the finite family consists of two objects X and Y . Our
assumption 1.3.2.2 (2) then implies that the natural morphism
a
p : RhX RhY −→ RhX ` L Y
is a covering. Therefore, RhX ` L Y is naturally equivalent to the homotopy colimit of
the homotopy nerve of the morphism p. Using this remark we see that it is enough
to prove that
RhX ×hRh ` L RhY ≃ ∅,
X Y

as then
` the homotopy nerve of p will be a constant simplicial object with values
RhX RhY . As the functor Rh commutes with homotopy pullbacks, it is therefore
enough to check that
A ⊗LA×h B B ≃ 0,
for A and B two commutative monoids in C such that X = Spec A and Y = Spec B
(here 0 is the final object in Comm(C)). For this we can of course suppose that A
and B are fibrant objects in C.
We define a functor
F : A × B − Comm(C) −→ A − Comm(C) × B − Comm(C)
by the formula
F (C) := (C ⊗A×B A, C ⊗A×B B).
The functor F is left Quillen for the product model structures on the right hand side,
and its right adjoint is given by G(C, D) := C × D for any (C, D) ∈ A − Comm(C) ×
B − Comm(C). For any C ∈ A × B − Comm(C), one has



C ≃ C ⊗LA×B (A × B) ≃ C ⊗LA×B A × C ⊗LA×B B ,
because of our assumptions 1.1.0.1 and 1.1.0.4, which implies that the adjunction
morphism
C −→ RG(LF (C))
is an isomorphism in Ho(A × B − Comm(C)). As the functor G reflects equivalences
(because of our assumption 1.1.0.1) this implies that F and G form a Quillen equiv-
alence. Therefore, the functor RG commutes with homotopy push outs, and we have
 
aL
A ⊗LA×B B ≃ RG(A, 0) ⊗LRG(A,B) RG(0, B) ≃ RG (A, 0) (0, B) ≃ RG(0) ≃ 0.
(A,B)

(2) We know by [HAGI] that Af fC∼,τ is the left Bousfield localization of Af fC∧
along the set of morphisms |F∗ | −→ hX , where F∗ −→ hX runs in a certain V-small set
of τ -hypercovers. Recall that for each hypercover F∗ −→ hX in this set, each simplicial
presheaf Fn is a coproduct of some hU . Using the quasi-compactness assumption
1.3.2.2 (1) one sees immediately that one can furthermore assume that each Fn is a
finite coproduct of some hU . Finally, using the part (1) of the present lemma we see
that the descent condition of [HAGI] can be stated as two distinct conditions, one
concerning finite coproducts and the other one concerning representable hypercovers.
From this we deduce part (2) of the lemma. 
Lemma 1.3.2.3 (2) can be reformulated as follows.
 1.3.2. HOMOTOPICAL ALGEBRAIC GEOMETRY CONTEXT 69

Corollary 1.3.2.4. A simplicial presheaf

F : Comm(C) −→ SSetV
is a stack if and only if it satisfies the following three conditions.
• For any equivalence A −→ B in Comm(C) the induced morphism F (A) −→
F (B) is an equivalence of simplicial sets.
• For any finite family of commutative monoids {A}i∈I in C (including the
empty family), the natural morphism
Y Y
F ( h Ai ) −→ F (Ai )
i∈I i∈I

is an isomorphism in Ho(SSet).
• For any co-simplicial commutative A-algebra A −→ B∗ , corresponding to a
τ -hypercover
Spec B∗ −→ Spec A
in Af fC , the induced morphism
F (A) −→ Holim[n]∈∆F (Bn )
is an isomorphism in Ho(SSet).
Another important consequence of lemma 1.3.2.3 is the following.
Corollary 1.3.2.5. The model pre-topology τ on Af fC is sub-canonical in the
sense of Def. 1.3.1.3.
Proof. We need to show that for any Z ∈ Af fC the object G := RhZ is a stack,
or in other words is a local object in Af fC∼,τ . For this, we use our lemma 1.3.2.3
(2). The descent property for finite coproducts is obviously satisfied because of the
Yoneda lemma. Let X∗ −→ Y be a simplicial object in Af fC such that
RhX∗ −→ RhY
is a τ -hypercover. By lemma 1.2.12.3 the natural morphism
Hocolimn Xn −→ Y
is an isomorphism in Ho(Af fC ). Therefore, the Yoneda lemma implies that one has
RHom(hY , G) ≃ M ap(Y, Z) ≃ Holimn M ap(Xn , Z) ≃ Holimn RHom(F∗ , G),
showing that G is a stack. 
The corollary 1.3.2.5 implies that Rh provides a fully faithful functor
Rh : Ho(Af fC ) −→ St(C, τ ).
Objects in the essential image of Rh will be called representable objects. If such an
object corresponds to a commutative monoid A ∈ Ho(Comm(C)), it will also be
denoted by RSpec A ∈ St(C, τ ). In formula
RSpec A := RhSpec A ,
for any A ∈ Comm(C) corresponding to Spec A ∈ Af fC . As Rh commutes with U-
small homotopy limits, we see that the subcategory of representable stacks is stable
by U-small homotopy limits. The reader should be careful that a V-small homotopy
limit of representable stacks is not representable in general. Lemma 1.3.2.3 (1) also
implies that a finite coproduct of representable stacks is a representable stack, and
we have a
RhUi ≃ Rh` L Ui .
i
i
70 1.3. GEOMETRIC STACKS: BASIC THEORY

Also, by identifying the category Ho(Comm(C))op with the full subcategory of

St(C, τ ) consisting of representable stacks, one can extend the notions of morphisms
defined in §1.2 (e.g. (formally) étale, Zariski open immersion, flat, smooth . . . ) to
morphisms between representable stacks. Indeed, they are all invariant by equiva-
lences and therefore are properties of morphisms in Ho(Comm(C)). We will often use
implicitly these extended notions. In particular, we will use the expression τ -covering
families of representable stacks to denote families of morphisms of representable stacks
corresponding in Ho(Comm(C))op to τ -covering families.
We will use the same terminology for the morphisms in Ho(Comm(C)) and for
the corresponding morphisms of representable stacks, except for the notion of epimor-
phism (Def. 1.2.6.1), which for obvious reasons will be replaced by monomorphism in
the context of stacks. This is justified since a morphism A −→ B is an epimorphism
in the sense of Def. 1.2.6.1 if and only if the induced morphism of stacks
RSpec B −→ RSpec A
is a monomorphism in the model category Af fC∼,τ (see Remark 1.2.6.2).
Remark 1.3.2.6. The reader should be warned that we will also use the expression
epimorphism of stacks, which will refer to a morphism of stacks that induces an
epimorphism on the sheaves π0 (see Def. 1.3.1.2 or [HAGI], where they are also
called coverings). It is important to notice that a τ -covering
` family of representable
stacks {Xi −→ X} induces an epimorphism of stacks Xi −→ X. On the contrary,
there might`very well exist families of morphisms of representable stacks {Xi −→ X}
such that Xi −→ X is an epimorphism of stacks, but which are not τ -covering
families (e.g. a morphism between representable stacks that admits a section).
Corollary 1.3.2.7. Let {ui : Xi = Spec Ai −→ X = Spec A}i∈I be a covering
family in Af fC . Then, the family of base change functors
{Lu∗i : Ho(A − M od) −→ Ho(Ai − M od)}i∈I
is conservative. In other words, a τ -covering family in Af fC is a formal covering in
the sense of Def. 1.2.5.1.
Proof. By the quasi-compactness assumption on τ we can assume that the cov-
Qh
ering family is finite. Also, the morphism A −→ i Ai = B is a covering. Therefore,
the descent assumption 1.3.2.2 (3) implies that the base change functor
B ⊗LA − : Ho(A − M od) −→ Ho(B − M od)
is conservative. Finally, we have seen during the proof of 1.3.2.3 (1) that the product
of the base change functors
Y Y
Ai ⊗LB − : Ho(B − M od) −→ Ho(Ai − M od)
i i

is an equivalence. Therefore, the composition

Y Y
Ai ⊗LA − : Ho(A − M od) −→ Ho(Ai − M od)
i i

is a conservative functor. 

We also recall the Yoneda lemma for stacks, stating that for any A ∈ Comm(C),
and any fibrant object F ∈ Af fC∼,τ , there is a natural equivalence of simplicial sets
RHom(RSpec A, F ) ≃ Rτ Hom(RSpec A, F ) ≃ F (A).
 1.3.2. HOMOTOPICAL ALGEBRAIC GEOMETRY CONTEXT 71

For an object F ∈ St(C, τ ), and A ∈ Comm(C) we will use the following notation
RF (A) := Rτ Hom(RSpec A, F ).
Note that RF (A) ≃ RF (A), where R is a fibrant replacement functor on Af fC∼,τ .
Note that there is always a natural morphism F (A) −→ RF (A), which is an equiva-
lence precisely when F is a stack.

Finally, another important consequence of assumption 1.3.2.2 is the local charac-

ter of representable stacks.
Proposition 1.3.2.8. Let G be a representable stack and F −→ G be any mor-
phism. Assume there exists a τ -covering family of representable stacks
{Gi −→ G},
such that each stack F ×hG Gi is representable. Then F is a representable stack.
Proof. Let X ∈ Af fC be a fibrant object such that G ≃ hX . We can of course
assume that G = hX . We can also assume that F −→ G is a fibration, and therefore
that G and F are fibrant objects.
By choosing a refinement of the covering family {Gi −→ G}, one can suppose that
the covering family is finite and that each morphism Gi −→ G is the image by h of a
`L
fibration Ui −→ X in Af fC . Finally, considering the coproduct U = Ui −→ X in
Af fC and using lemma 1.3.2.3 (1) one can suppose that the family {Gi −→ G} has
only one element and is the image by h of a fibration U −→ X in Af fC .
We consider the augmented simplicial object U∗ −→ X in Af fC , which is the
nerve of the morphism U → X, and the corresponding augmented simplicial object
hU∗ −→ hX in Af fC∼,τ . We form the pullback in Af fC∼,τ
FO / hX
O

F∗ / hU
∗

which is a homotopy pullback because of our choices. In particular, for any n, Fn is

a representable stack.
Clearly F∗ is the nerve of the fibration F ×hX hU −→ F . As this last morphism
is an epimorphism in Af fC∼,τ the natural morphism
|F∗ | := Hocolimn Fn −→ F
is an isomorphism in St(C, τ ). Therefore, it remains to show that |F∗ | is a representable
stack.
We will consider the category sAf fC∼,τ of simplicial objects in Af fC∼,τ , endowed
with its Reedy model structure (see [HAGI, §4.4] for details and notations). In the
same way, we will consider the Reedy model structure on sAf fC , the category of
simplicial objects in Af fC .
Lemma 1.3.2.9. There exists a simplicial object V∗ in Af fC and an isomorphism
RhV∗ ≃ F∗ in Ho(sAf fC∼,τ ).
Proof. First of all, our functor h extends in the obvious way to a functor on the
categories of simplicial objects
h : sAf fC −→ sAf fC∼,τ ,
by the formula
(hX∗ )n := hXn
72 1.3. GEOMETRIC STACKS: BASIC THEORY

for any X∗ ∈ sAf fC . This functor possesses a right derived functor

Rh : Ho(sAf fC ) −→ Ho(sAf fC∼,τ ),
which is easily seen to be fully faithful.
We claim that the essential image of Rh consists of all simplicial objects F∗ ∈
Ho(sAf fC∼,τ ) such that for any n, Fn is a representable stack. This will obviously
imply the lemma. Indeed, as Rh commutes with homotopy limits, one sees that this
essential image is stable by (U-small) homotopy limits. Also, any object F∗ ∈ sAf fC∼,τ
can be written as a homotopy limit
F∗ ≃ Holimn RCoskn (F∗ )
of its derived coskeleta (see [HAGI, §4.4]). Recall that for a fibrant object F∗ in
sAf fC∼,τ one has
p
Coskn (F∗ )p ≃ RCoskn (F∗ )p ≃ (F∗ )Skn ∆ ,
where, for a simplicial set K, (F∗ )K is defined to be the equalizer of the two natural
maps Y Y
(Fn )Kn ⇉ (Fm )Kn .
[n] [n]→[m]
Now, for any simplicial set K, and any integer n ≥ 0, there is a homotopy push out
square
SknO K / Skn+1 K
O

` `
K ∂∆n+1 ∂∆n+1 /
K ∆n+1 ∆n+1 .
Using this, and the fact that K 7→ F∗K sends homotopy push outs to homotopy
pullbacks when F∗ is fibrant, we see that for any fibrant object F∗ ∈ sAf fC∼,τ , and
any n, we have a homotopy pullback diagram in sAf fC∼,τ
RCoskn (F∗ ) / RCoskn−1 (F∗ )
O O

An∗ / B∗n .
Here, An∗ and B∗n are defined by the following formulas
An∗ : ∆op −→ Af fC∼,τ
Q ∆n+1
[p] 7→ (∆p )∆n+1 F∗

B∗n : ∆op −→ Af fC∼,τ

Q ∂∆n+1
[p] 7→ (∆p )∂∆n+1 F∗ .
Therefore, by induction on n, it is enough to see that if F∗ is fibrant then
Cosk0 (F∗ ), An∗ and B∗n all belongs to the essential image of Rh.
The simplicial object Cosk0 (F∗ ) is isomorphic to the nerve of F0 −→ ∗, and
therefore is the image by Rh of the nerve of a fibration X −→ ∗ in Af fC representing
F0 . This shows that Cosk0 (F∗ ) is in the image of Rh.
n+1
We have F∗∆ = Fn+1 , which is a representable stack. Let Xn+1 ∈ Af fC be
a fibrant object such that Fn+1 is equivalent to hXn+1 . Then, as h commutes with
limits, we see that An∗ is equivalent the image by h of the simplicial object
Y
[p] 7→ Xn+1 .
(∆p )∆n+1
 1.3.2. HOMOTOPICAL ALGEBRAIC GEOMETRY CONTEXT 73

n+1
In the same way, F∗∂∆ can be written as a finite homotopy limit of Fi ’s, and
therefore is a representable stack. Let Yn+1 be a fibrant object in Af fC such that
n+1
F∗∂∆ is equivalent to hYn+1 . Then, B∗n is equivalent to the image by h of the
simplicial object
Y
[p] 7→ Yn+1 .
(∆p )∂∆n+1

This proves the lemma. 

We now finish the proof of Proposition 1.3.2.8. Let V∗ be a simplicial object in
sAf fC such that F∗ ≃ RhV∗ . The augmentation F∗ −→ RhU∗ gives rise to a well
defined morphism in Ho(sAf fC∼,τ )
q : RhV∗ −→ RhU∗ .
We can of course suppose that V∗ is a cofibrant object in sAf fC . As U∗ is the nerve
of a fibration between fibrant objects it is fibrant in sAf fC . Therefore, as Rh is fully
faithful, we can represent q, up to an isomorphism, as the image by h of a morphism
in sAf fC
r : V∗ −→ U∗ .
On the level of commutative monoids, the morphism r is given by a morphism B∗ −→
C∗ of co-simplicial objects in Comm(C). By construction, for any morphism [n] → [m]
in ∆, the natural morphism
Fm −→ Fn ×hRhU RhUm
n

is an isomorphism in St(C, τ ). This implies that the underlying co-simplicial B∗ -

module of C∗ is homotopy cartesian in the sense of Def. 1.2.12.1. Our assumption
1.3.2.2 (3) implies that if Y := Hocolimn V∗ ∈ Af fC , then the natural morphism
V∗ −→ U∗ ×hX Y
is an isomorphism in Ho(sAf fC ). As homotopy colimits in Af fC∼,τ commute with
homotopy pullbacks, this implies that
|F∗ | ≃ |RhV∗ | ≃ |RhU∗ | ×hRhX RhY .
But, as RhU∗ −→ RhX is the homotopy nerve of an epimorphism we have
|RhU∗ | ≃ RhX ,
showing finally that |F∗ | is isomorphic to RhY and therefore is a representable stack.

Finally, we finish this first section by the following description of the comma model
category Af fC∼,τ /hX , for some fibrant object X ∈ Af fC . This is not a completely
trivial task as the model category Af fC∼,τ is not right proper.
For this, let A ∈ Comm(C) such that X = Spec A, so that A is a cofibrant object
in Comm(C). We consider the comma model category A − Comm(C) = (Af fC /X)op ,
op
which is also the model category Af fA−Mod = Comm(A − M od). The model pre-
topology τ on Af fC induces in a natural way a model pre-topology τ on Af fC /X =
Af fA−Mod . Note that there exists a natural equivalence of categories, compatible
with the model structures, between (Af fC /X)∼,τ and Af fC∼,τ /hX . We consider the
natural morphism hX −→ hX . It gives rise to a Quillen adjunction
Af fC∼,τ /hX −→ Af fC∼,τ /hX Af fC∼,τ /hX ←− Af fC∼,τ /hX
where the right adjoint sends F −→ hX to F ×hX hX −→ hX .
74 1.3. GEOMETRIC STACKS: BASIC THEORY

Proposition 1.3.2.10. The above Quillen adjunction induces a Quillen equiva-

lence between Af fC∼,τ /hX ≃ (Af fC /X)∼,τ and Af fC∼,τ /hX .
Proof. For F −→ hX a fibrant object, and Y ∈ Af fC /X, there is a homotopy
cartesian square
(F ×hX hX )(Y ) / hX (Y )

 
F (Y ) / h (Y ).
X

As the morphism hX (Y ) −→ hX (Y ) is always surjective up to homotopy when Y is

cofibrant, this implies easily that the derived pullback functor
Ho(Af fC∼,τ /hX ) −→ Ho(Af fC∼,τ /hX )
is conservative. Therefore, it is enough to show that the forgetful functor
Ho(Af fC /X ∼,τ ) −→ Ho(Af fC∼,τ /hX )
is fully faithful.
Using the Yoneda lemma for Af fC /X, we have
M ap(Af fC /X)∼,τ (hY , hZ ) ≃ M apAf fC /X (Y, Z)
for any two objects Y and Z in Af fC /X. Therefore, there exists a natural fibration
sequence
M ap(Af fC /X)∼,τ (hY , hZ ) / M apAf fC (Y, Z) / M apAf fC (Z, X).

In the same way, the Yoneda lemma for Af fC implies that there exists a fibration
sequence
M apAf fC∼,τ /hX (hY , hZ ) / M apAf fC (Y, Z) / M apAf fC (Z, X).

This two fibration sequences implies that the forgetful functor induces equivalences
of simplicial sets
M apAf fC /X ∼,τ (hY , hZ ) ≃ M apAf fC∼,τ /hX (Y, Z).
In other words, the functor
Ho(Af fC /X ∼,τ ) −→ Ho(Af fC∼,τ /hX )
is fully faithful when restricted to the full subcategory of representable stacks. But,
any object in Ho(Af fC /X ∼,τ ) is a homotopy colimit of representable stacks. Fur-
thermore, as the derived pullback
Ho(Af fC∼,τ /hX ) −→ Ho(Af fC∼,τ /hX )
commutes with homotopy colimits (as homotopy pullbacks of simplicial sets do), this
implies that the functor
Ho(Af fC /X ∼,τ ) −→ Ho(Af fC∼,τ /hX )
is fully faithful on the whole category. 
The important consequence of Prop. 1.3.2.10 comes from the fact that it allows
to see objects in Ho(Af fC∼,τ /hX ) as functors
A − Comm(C) −→ SSet.
This last fact will be used implicitly in the sequel of this work.
 1.3.2. HOMOTOPICAL ALGEBRAIC GEOMETRY CONTEXT 75

As explained in the introduction to this chapter, we will need to fix a class P of

morphisms in Af fC . Such a class will be then used to glue representable stacks to get
a geometric stack. In other words, geometric stacks will be the objects obtained by
taking some quotient of representable stacks by equivalence relations whose structural
morphisms are in P. Of course, different choices of P will lead to different notions of
geometric stacks. To fix his intuition the reader may think of P as being the class of
smooth morphisms (though in some applications P can be something different).

From now, and all along this section, we fix a class P of morphism in Af fC , that
is stable by equivalences. As the Yoneda functor
Rh : Ho(Af fC )op −→ St(C, τ )
is fully faithful we can extend the notion of morphisms belonging to P to its essential
image. So, a morphism of representable objects in Ho(SP r(Af fC∼,τ )) is in P if by
definition it correspond to a morphism in Ho(Af fC ) which is in P. We will make
the following assumtions on morphisms of P with respect to the topology τ , making
“being in P” into a τ -local property.
Assumption 1.3.2.11. (1) Covering families consist of morphisms in P i.e.
for any τ -covering family {Ui −→ X}i∈I in Af fC , the morphism Ui −→ X
is in P for all i ∈ I.
(2) Morphisms in P are stable by compositions, equivalences and homotopy pull-
backs.
(3) Let f : X −→ Y be a morphism in Af fC . If there exists a τ -covering family
{Ui −→ X}
such that each composite morphism Ui −→ Y lies in P, then f belongs to
P.
(4) For any two objects X and Y in Af fC , the two natural morphisms
L
a L
a
X −→ X Y Y −→ X Y
are in P.
The reader will notice that assumptions 1.3.2.2 and 1.3.2.11 together imply the
following useful fact.
Lemma 1.3.2.12. Let {Xi −→ X} be a finite family of morphisms in P. The total
morphism
L
a
Xi −→ X
i
is also in P.
Proof. We consider the family of natural morphisms
L
a
{Xj −→ Xi } j .
i

According to our assumption 1.3.2.2 it is a τ -covering family in Af fC . Moreover, each

`L
morphism Xj −→ i Xi and Xj −→ X is in P, so assumption 1.3.2.11 (3) implies
`L
that so is i Xi −→ X. 
We finish this section by the following definition.
76 1.3. GEOMETRIC STACKS: BASIC THEORY

Definition 1.3.2.13. A Homotopical Algebraic Geometry context (or simply

HAG context) is a 5-tuple (C, C0 , A, τ, P), where (C, C0 , A) is a HA context in the
sense of Def. 1.1.0.11, τ is a model pre-topology on Af fC , P is a class of morphism
in Af fC , and such that assumptions 1.3.2.2 and 1.3.2.11 are satisfied.

1.3.3. Main definitions and standard properties

From now on, we fix a HAG context (C, C0 , A, τ, P) in the sense of Def. 1.3.2.13.
We will consider the model category of stacks Af fC∼,τ as described in the previous
section, and introduce the notion of geometric and n-geometric stacks, which are
objects in Af fC∼,τ satisfying certain properties.
The basic geometric idea is that a stack is geometric if it is obtained by taking
the quotient of a representable stack X (or more generally of a disjoint union of
representable stacks) by the action of a groupoid object X1 acting on X, such that X1
is itself representable, and such that the source morphism X1 −→ X is a morphism in
the chosen class P. If one thinks of P as being the class of certain smooth morphisms,
being geometric is thus equivalent of being a quotient by a smooth groupoid action.
It turns out that this notion is not enough for certain applications, as some natu-
rally arising stacks are obtained as quotients by an action of a groupoid in geometric
stacks rather than in representable stacks (e.g. the quotients by a group-stack action).
We will therefore also introduce the notion of n-geometric stack, which is defined in-
ductively as a stack obtained as a quotient by an action of a groupoid object in
(n − 1)-geometric stacks whose source morphism is in P. Of course, for this definition
to make sense one must also explain, inductively on n, what are the morphisms in P
between n-geometric stacks.
The inductive definition we give below uses a different (though equivalent) point
of view, closer to the original definition of algebraic stacks due to Deligne-Mumford
and M. Artin. It says that a stack F is n-geometric if for any pair of points of F , the
stack of equivalences between them is (n − 1)-geometric, and if moreover it receives a
morphism, which is surjective and in P, from a representable stack (or from a disjoint
union of representable stacks). The equivalence of this definition with the previously
mentioned quotient-by-groupoids point of view will be established in the next section
(see Prop. 1.3.4.2).

Definition 1.3.3.1. (1) A stack is (−1)-geometric if it is representable.

(2) A morphism of stacks f : F −→ G is (−1)-representable if for any rep-
resentable stack X and any morphism X −→ G, the homotopy pullback
F ×hG X is a representable stack.
(3) A morphism of stacks f : F −→ G is in (−1)-P if it is (−1)-representable,
and if for any representable stack X and any morphism X −→ G, the in-
duced morphism
F ×hG X −→ X
is a P-morphism between representable stacks.

Now let n ≥ 0.
(1) Let F be any stack. An n-atlas for F is a U-small family of morphisms
{Ui −→ F }i∈I such that
(a) Each Ui is representable.
(b) Each morphism Ui −→ F is in (n − 1)-P.
 1.3.3. MAIN DEFINITIONS AND STANDARD PROPERTIES 77

(c) The total morphism

a
Ui −→ F
i∈I

is an epimorphism.
(2) A stack F is n-geometric if it satisfies the following two conditions.
(a) The diagonal morphism F −→ F ×h F is (n − 1)-representable.
(b) The stack F admits an n-atlas.
(3) A morphism of stacks F −→ G is n-representable if for any representable
stack X and any morphism X −→ G, the homotopy pullback F ×hG X is
n-geometric.
(4) A morphism of stacks F −→ G is in n-P (or has the property n-P, or is
a n-P-morphism) if it is n-representable and if for any representable stack
X, any morphism X −→ G, there exists an n-atlas {Ui } of F ×hG X, such
that each composite morphism Ui −→ X is in P.
Remark 1.3.3.2. In the above definition, condition (2a) follows from condition
(2b). This is not immediate now but will be an easy consequence of the description
of geometric stacks as quotients by groupoids given in the next section. We prefer to
keep the definition of n-geometric stacks with the two conditions (2a) and (2b) as it
is very similar to the usual definition of algebraic stacks found in the literature (e.g.
in [La-Mo]).
The next Proposition gives the fundamental properties of geometric n-stacks.
Proposition 1.3.3.3.
(1) Any (n − 1)-representable morphism is n-representable.
(2) Any (n − 1)-P-morphism is a n-P-morphism.
(3) n-representable morphisms are stable by isomorphisms, homotopy pullbacks
and compositions.
(4) n-P-morphisms are stable by isomorphisms, homotopy pullbacks and com-
positions.
Proof. We use a big induction on n. All the assertions are easily verified for
n = −1 using our assumptions 1.3.2.11 on the morphisms in P. So, we fix an integer
n ≥ 0 and suppose that all the assertions are true for any m < n; let’s prove that
they all remain true at the level n.

(1) By definition 1.3.3.1 it is enough to check that a (n − 1)-geometric stack F is

n-geometric. But an (n − 1)-atlas for F is a n-atlas by induction hypothesis (which
tells us in particular that a (n − 2)-P-morphism is a (n − 1)-P-morphism), and more-
over the diagonal of F is (n − 2)-representable thus (n − 1)-representable, again by
induction hypothesis. Therefore F is n-geometric.

(2) Let f : F −→ G be an (n − 1)-P-morphism. By definition it is (n − 1)-

representable, hence n-representable by (1). Let X −→ G be a morphism with X
representable. Since f is in (n−1)-P, there exists an (n−1)-atlas {Ui } of F ×hG X such
that each Ui −→ X is in P. But, as already observed in (1), our inductive hypothesis,
shows that any (n − 1)-atlas is an n-atlas, and we conclude that f is also in n-P.

(3) Stability by isomorphisms and homotopy pullbacks is clear by definition. To

prove the stability by composition, it is enough to prove that if f : F −→ G is an
n-representable morphism and G is n-geometric then so is F .
78 1.3. GEOMETRIC STACKS: BASIC THEORY

Let {Ui } be an n-atlas of G, and let Fi := F ×hG Ui . The stacks Fi are n-geometric,
so we can find an n-atlas {Vi,j }j for Fi , for any i. By induction hypothesis (telling
us in particular that (n − 1)-P-morphisms are closed under composition) we see that
the family of morphisms {Vi,j −→ F } is an n-atlas for F . It remains to show that
the diagonal of F is (n − 1)-representable.
There is a homotopy cartesian square

F ×hG F / F ×h F

 
G / G ×h G.

As G is n-geometric, the stability of (n− 1)-representable morphisms under homotopy

pullbacks (true by induction hypothesis) implies that F ×hG F −→ F ×h F is (n − 1)-
representable. Now, the diagonal of F factors has F −→ F ×hG F −→ F ×h F, and
therefore by stability of (n − 1)-representable morphisms by composition (true by
induction hypothesis), we see that it is enough to show that F −→ F ×hG F is (n − 1)-
representable. Let X be a representable stack and X −→ F ×hG F be any morphism.
Then, we have
F ×hF ×h F X ≃ X ×h(F ×h X)×h (F ×h X) (F ×hG X).
G G G

As by hypothesis F ×hG X is n-geometric this shows that F ×hF ×h F X is (n − 1)-

G
geometric, showing that F −→ F ×hG F is (n − 1)-representable.

(4) Stability by isomorphisms and homotopy pullbacks is clear by definition. Let

F −→ G −→ H be two n-P-morphisms of stacks. By (3) we already know the
composite morphism to be n-representable. By definition of being in n-P one can
assume that H is representable. Then, there exists an n-atlas {Ui } for G such that
each morphism Ui −→ H is in P. Let Fi := F ×hG Ui , and let {Vi,j }j be an n-atlas
for Fi such that each Vi,j −→ Ui is in P. Since by induction hypothesis (n − 1)-P-
morphisms are closed under composition, we see that {Vi,j } is indeed an n-atlas for
F such that each Vi,j −→ H is in P. 

An important consequence of our descent assumption 1.3.2.2 and Prop. 1.3.2.8 is

the following useful proposition.
Proposition 1.3.3.4. Let f : F −→ G be any morphism such that G is an n-
geometric stack. We suppose that there exists a n-atlas {Ui } of G such that each stack
F ×hG Ui is n-geometric. Then F is n-geometric.
If furthermore each projection F ×hX Ui −→ Ui is in n-P, then so is f .
Proof. Using the stability of n-representable morphisms by composition (see
Prop. 1.3.3.3 (3)) we see that it is enough to show that f is n-representable. The
proof goes by induction on n. For n = −1 this is our corollary Prop. 1.3.2.8. Let us
assume n ≥ 0 and the proposition proved for rank less than n. Using Prop. 1.3.3.3
(3) it is enough to suppose that G is a representable stack X.
Let {Ui } be an n-atlas of X as in the statement, and let {Vi,j } be an n-atlas for
Fi := F ×hX Ui . Then, the composite family Vi,j −→ F is clearly an n-atlas for F . It
remains to prove that F has an (n − 1)-representable diagonal.
The diagonal of F factors as
F −→ F ×hX F −→ F ×h F.
 1.3.3. MAIN DEFINITIONS AND STANDARD PROPERTIES 79

The last morphism being the homotopy pullback

F ×hX F / F ×h F

 
X / X ×h X

is representable and therefore (n−1)-representable. Finally, let Z be any representable

stack and Z −→ F ×hX F be a morphism. Then, the morphism
Z ×hF ×h F F −→ X
X

satisfies the conditions of the proposition 1.3.3.4 for the rank (n − 1). Indeed, for any
i we have
(Z ×hF ×h F F ) ×hX Ui ≃ (Z ×hX Ui ) ×hFi ×h Fi Fi .
X Ui

Therefore, using the induction hypothesis we deduce that the stack Z ×hF ×h F F is
X
(n − 1)-geometric, proving that F −→ F ×hX F is (n − 1)-representable.
The last part of the proposition follows from the fact that any n-atlas {Vi,j } of
Fi is such that each morphism Vi,j −→ X is in n-P by construction. 
Corollary 1.3.3.5. The full subcategory of n-geometric stacks in St(C, τ ) is
stable by homotopy pullbacks, and by U-small disjoint union if n ≥ 0.
Proof. Let F /H o G be a diagram of stacks. There are two homo-
topy cartesian squares
F ×h G /G F ×hH G /H

   
F /• F ×h G / H ×h H,

showing that the stability under homotopy pullbacks follows from the stability of
n-representable morphisms under compositions and homotopy pullbacks.
` corollary, concerning U-small disjoint union.
Let us prove the second part of the
Suppose now tha n ≥ 0 and let F be i Fi with each Fi an n-geometric stack. Then,
we have a
F ×h F ≃ Fi ×h Fj .
i,j
For any representable stack X, and any morphism X −→ F ×h F , there exists a
0-atlas {Uk } of X, and commutative diagrams of stacks
Uk / Fi(k) ×h Fj(k)

 ` 
X /
i,j Fi ×h Fj .
We apply Prop. 1.3.3.4 to the morphism
G := F ×hF ×h F X −→ X
and for the covering {Uk }. We have
G ×hX Uk ≃ ∅ if i(k) 6= j(k)
and
G ×hX Uk ≃ Fi(k) ×hFi(k) ×h Fi(k) Uk
80 1.3. GEOMETRIC STACKS: BASIC THEORY

otherwise. Prop. 1.3.3.4 implies that G is (n − 1)-geometric, and therefore that the
diagonal of F is (n − 1)-representable. Finally, the same argument and assumption
1.3.2.11 show that the disjoint union of n-atlases of the Fi ’s will form an n-atlas for
F. 
Finally, let us mention the following important additional property.
Proposition 1.3.3.6. Let f : F −→ G be an n-representable morphism. If f is
in m-P for m > n then it is in n-P.
Proof. By induction on m it is enough to treat the case m = n + 1. The proof
goes then by induction on n. For n = −1 this is our assumption 1.3.2.11 (3). For
n ≥ 0 we can by definition assume G is a representable stack and therefore that F is
n-geometric. Then, there exists an (n + 1)-atlas {Ui } for F such that each Ui −→ G
is in P. By induction, {Ui } is also an n-atlas for F , which implies that the morphism
f is in fact in P. 
The last proposition implies in particular that for an n-representable morphism
of stacks the property of being in n-P does not depend on n. We will therefore give
the following definition.
Definition 1.3.3.7. A morphism in St(C, τ ) is in P if it is in n-P for some
integer n.

1.3.4. Quotient stacks

We will now present a characterization of geometric stacks in terms of quotient
by groupoid actions. This point of view is very much similar to the presentation of
manifolds by charts, and much less intrinsic than definition Def. 1.3.3.1. However,
it is sometimes more easy to handle as several stacks have natural presentations as
quotients by groupoids.

Let X∗ be a Segal groupoid object in a model category M (Def. 1.3.1.6). Inverting

the equivalence
X2 −→ X1 ×hX0 X1
and composing with d1 : X2 −→ X1 gives a well defined morphism in Ho(M )
µ : X1 ×hX0 X1 −→ X1 ,
that is called composition. In the same way, inverting the equivalence
X2 −→ X1 ×hd0 ,X0 ,d0 X1 ,
and composing with
Id ×h s0 : X1 −→ X1 ×hd0 ,X0 ,d0 X1
d2 : X2 −→ X1
gives a well defined isomorphism in Ho(M )
i : X1 −→ X1 ,
called inverse. It is easy to check that d1 ◦ i = d0 as morphisms in Ho(M ), showing
that the two morphisms d0 and d1 are always isomorphic in Ho(M ). Finally, using
condition (1) of Def. 1.3.1.6 we see that for any i > 0, all the face morphisms
Xi −→ Xi−1
of a Segal groupoid object are isomorphic in Ho(M ).
 1.3.4. QUOTIENT STACKS 81

Definition 1.3.4.1. A Segal groupoid object X∗ in Ho(SP r(Af fC∼,τ )) is an n-P

Segal groupoid if it satisfies the following two conditions.
(1) The stacks X0 and X1 are disjoint unions of n-geometric stacks.
(2) The morphism d0 : X1 −→ X0 is in n-P.
As n-geometric stacks are stable by homotopy pullbacks, Xi is a disjoint union
of n-geometric stacks for any i and any n-P Segal groupoid X∗ . Furthermore, the
condition (2) of Def. 1.3.1.6 implies that the two morphisms
d0 , d1 : X1 −→ X0
are isomorphic as morphisms in St(C, τ ). Therefore, for any n-P Segal groupoid X∗ ,
all the faces Xi −→ Xi−1 are in P.
Proposition 1.3.4.2. Let F ∈ Ho(SP r(Af fC∼,τ )) be a stack and n ≥ 0. The
following conditions are equivalent.
(1) The stack F is n-geometric.
(2) There exists an (n − 1)-P Segal groupoid object X∗ in SP r(Af fC∼,τ ), such
that X0 is a disjoint union of representable stacks, and an isomorphism in
Ho(SP r(Af fC∼,τ ))
F ≃ |X∗ | := Hocolim[n]∈∆Xn .
(3) There exists an (n − 1)-P Segal groupoid object X∗ in SP r(Af fC∼,τ ), and
an isomorphism in Ho(SP r(Af fC∼,τ ))
F ≃ |X∗ | := Hocolim[n]∈∆Xn .
If these conditions are satisfied we say that F is the quotient stack of the (n − 1)-P
Segal groupoid X∗ .
Proof. We have already seen that a 0-geometric stack is n-geometric for any n.
Therefore, (2) ⇒ (3). It remains to show that (1) implies (2) and (3) implies (1).

(1) ⇒ (2) Let F be an n-geometric stack, and {Ui } be an n-atlas for F . We let
a
p : X0 := Ui −→ F
i

be the natural projection. Up to an equivalence, we can represent p by a fibration

X0 −→ F between fibrant objects in SP r(Af fC∼,τ ). We define a simplicial object X∗
to be the nerve of p
Xn := X0 ×F X0 ×F · · · ×F X0 .
| {z }
n times
Clearly, X∗ is a groupoid object in SP r(Af fC∼,τ ) in the usual sense, and as p is a
fibration between fibrant objects it follows that it is also a Segal groupoid
` object in
the sense of Def. 1.3.1.6. Finally, as F is n-geometric, one has X1 ≃ i,j Ui ×hF Uj
` an (n − 1)-geometric stack by Cor. 1.3.3.5. The morphism d0 :
which is therefore
X1 −→ X0 ≃ i Ui is then given by the projections Ui ×hF Uj −→ Ui which are in
(n − 1)-P as {Ui } is an n-atlas. This implies that X∗ is an (n − 1)-P Segal groupoid
such that X0 is a disjoint union of representable.
Lemma 1.3.4.3. The natural morphism
|X∗ | −→ F
is an isomorphism in Ho(SP r(Af fC∼,τ )).
82 1.3. GEOMETRIC STACKS: BASIC THEORY

Proof. For any fibration of simplicial sets f : X −→ Y , we know that the

natural morphism from the geometric realization of the nerve of f to Y is equivalent
to an inclusion of connected components. This implies that the morphism |X∗ | −→ F
induces an isomorphisms on the sheaves πi for i > 0 and an injection on π0 . As
the morphism p : X0 −→ F is an epimorphism and factors as X0 → |X∗ | → F , the
morphism |X∗ | −→ F is also an epimorphism. This shows that it is surjective on the
sheaves π0 and therefore is an isomorphism in Ho(SP r(Af fC∼,τ ). 

The previous lemma finishes the proof of (1) ⇒ (2).

(3) ⇒ (1) Let X∗ be an (n − 1)-P Segal groupoid and F = |X∗ |. First of all, we
recall the following important fact.
Lemma 1.3.4.4. Let M be a U-model topos, and X∗ be a Segal groupoid object in
M with homotopy colimit |X∗ |. Then, for any n > 0, the natural morphism
Xn −→ X0 ×h|X∗ | X0 ×h|X∗ | · · · ×h|X∗ | X0
| {z }
n times

is an isomorphism in Ho(M ).
Proof. This is one of the standard properties of model topoi. See Thm. 1.3.1.7.


Let {Ui } be an (n − 1)-atlas for X0 , and let us consider the composed morphisms
fi : Ui −→ X0 −→ F.
`
Clearly, i Ui −→ F is a composition of epimorphism, and is therefore a epimorphism.
In order to prove that {Ui } form an n-atlas for F it is enough to prove that the
morphism X0 −→ F is in (n − 1)-P.
Let X be any representable stack, X −→ F be a morphism, and let G be X0 ×hF X.
As the morphism X0 −→ F is an epimorphism, we can find a covering family {Zj −→
X}, such that each Zj is representable, and such that there exists a commutative
diagram in St(C, τ )
X0Z4 /F
44 O
44
44
44
44 XO
44
44
`
j Zj .

By assumption 1.3.2.11 (1) we can also assume that each morphism Zj −→ X is in

P, and therefore that {Zj −→ X} is a 0-atlas of X.
In order to prove that G −→ X is in (n − 1)-P it is enough by Prop. 1.3.3.4 to
prove that each stack Gj := G×hX Zj ≃ X0 ×hF Zj is (n−1)-geometric, and furthermore
that each projection Gj −→ Zj is in P. We have
Gj ≃ (X0 ×hF X0 ) ×hX0 Zj .
Therefore, lemma 1.3.4.4 implies that Gj ≃ X1 ×hX0 Zj , showing that each Gj is
(n − 1)-geometric and finally that G is (n − 1)-geometric. Furthermore, this also
shows that each projection Gj −→ Zj is of the form X1 ×hX0 Zj −→ Zj which is in P
by hypothesis on the Segal groupoid object X∗ . 
 1.3.5. QUOTIENT STACKS AND TORSORS 83

Corollary 1.3.4.5. Let f : F −→ G be an epimorphism of stacks and n ≥ 0. If

F is n-geometric and f is (n − 1)-representable and in P, then G is n-geometric.

Proof. Indeed, let U −→ F be an n-atlas, and let g : U −→ G be the compo-

sition. The morphism g is still an epimorphism and (n − 1)-representable, and thus
we can assume that F is 0-representable. The morphism f being an epimorphism, G
is equivalent to the quotient stack of the Segal groupoid X∗ which is the homotopy
nerve of f . By assumption this Segal groupoid is an (n − 1)-Segal groupoid and thus
G is n-geometric by Prop. 1.3.4.2 (1). 

1.3.5. Quotient stacks and torsors

Writing a stack as a quotient stack of a Segal groupoid is also useful in order
to describe associated stacks to certain pre-stacks. Indeed, it often happens that a
pre-stack is defined as the quotient of a Segal groupoid, and we are going to show
in this section that the associated stack of such quotients can be described using an
adequate notion of torsor over a Segal groupoid. This section is in fact completely
independent of the notion of geometricity and concern pure stacky statements that
are valid in any general model topos. We have decided to include this section as it
helps understanding the quotient stack construction presented previously. However,
it is not needed for the rest of this work and proofs will be more sketchy than usual.

We let M be a general U-model topos in the sense of Def. 1.3.1.5. The main case
of application will be M = Af fC∼,τ but we rather prefer to state the results in the
most general setting (in particular we do not even assume that M is t-complete).
Let X∗ be a Segal groupoid object in M in the sense of Def. 1.3.1.6, and we
assume that each Xn is a fibrant object in M . We will consider sM , the category
of simplicial objects in M , which will be endowed with its levelwise projective model
structure, for which fibrations and equivalences are defined levelwise. We consider
sM/X∗ , the model category of simplicial objects over X∗ . Finally, we let Z be a
fibrant object in M , and X∗ −→ Z be a morphism in sM (where Z is considered as
a constant simplicial object), and we assume that the induced morphism

|X∗ | −→ Z

is an isomorphism in Ho(M ).
We define a Quillen adjunction

φ : sM/X∗ −→ M/Z sM/X∗ ←− M/Z : ψ

in the following way. For any Y −→ Z in M , we set

ψ(Y ) := Y ×Z X∗ ∈ sM,

or in other words ψ(Y )n = Y ×Z Xn and with the obvious transitions morphisms.

The left adjoint to ψ sends a simplicial object Y∗ −→ X∗ to its colimit in M

φ(Y∗ ) := Colimn∈∆op Yn −→ Colimn Xn∈∆op −→ Z.

It is easy to check that (φ, ψ) is a Quillen adjunction.

Proposition 1.3.5.1. The functor

Rψ : Ho(M/Z) −→ Ho(sM/X∗ )
84 1.3. GEOMETRIC STACKS: BASIC THEORY

is fully faithful. Its essential image consists of all Y∗ −→ X∗ such that for any
morphism [n] → [m] in ∆ the square
Ym / Yn

 
Xm / Xn

is homotopy cartesian.
Proof. Let Y → Z in M/Z. Proving that Rψ is fully faithful is equivalent to
prove that the natural morphism
|Y ×hZ X∗ | −→ Y
is an isomorphism in Ho(M ). But, using the standard properties of model topoi (see
Thm. 1.3.1.7), we have
|Y ×hZ X∗ | ≃ Y ×hZ |X∗ | ≃ Y
as |X∗ | ≃ Z. This shows that Rψ is fully faithful.
By definition of the functors φ and ψ, it is clear that Rψ takes its values in the
subcategory described in the statement of the proposition. Conversely, let Y∗ −→ X∗
be an object in Ho(M ) satisfying the condition of the proposition. As X∗ is a Segal
groupoid object, we know that X∗ is naturally equivalent to the homotopy nerve of
the augmentation morphism X0 −→ Z (see Thm. 1.3.1.7). Therefore, the object
RψLφ(Y∗ ) is by definition the homotopy nerve of the morphism
|Y∗ | ×hZ X0 −→ |Y∗ |.
But, we have
|Y∗ | ×hZ X0 ≃ |Y∗ ×hZ X0 | ≃ Y0
by hypothesis on Y∗ . Therefore, the object RψLφ(Y∗ ) is naturally isomorphic in
Ho(sM/X∗ ) to the homotopy nerve of the natural
Y0 −→ |Y∗ |.
Finally, as X∗ is a Segal groupoid object, so is Y∗ by assumption. The standard
properties of model topoi (see Thm. 1.3.1.7) then tell us that Y∗ is naturally equivalent
to the homotopy nerve of Y0 −→ |Y∗ |, and thus to RψLφ(Y∗ ) by what we have just
done. 

The model category sM/X∗ , or rather its full subcategory of objects satisfying
the conditions of Prop. 1.3.5.1, can be seen as the category of objects in M together
with an action of the Segal groupoid X∗ . Proposition 1.3.5.1 therefore says that the
homotopy theory of stacks over |X∗ | is equivalent to the homotopy theory of stacks
together with an action of X∗ . This point of view will now help us to describe the
stack associated to |X∗ |.
For this, let F be a fixed fibrant object in M . We define a new model category
sM/(X∗ , F ) in the following way. Its objects are pairs (Y∗ , f ), where Y∗ → X∗ is
an object in sM/X∗ and f : Colimn Yn −→ F is a morphism in M . Morphisms
(Y∗ , f ) → (Y∗′ , f ′ ) are given by morphisms Y∗ −→ Y∗′ in sM/X∗ , such that
ColimnIYn / Colimn Yn′
II u
II uu
II uuu
f III uu ′
$ uz u f
F
 1.3.5. QUOTIENT STACKS AND TORSORS 85

commutes in M . The model structure on sM/(X∗ , F ) is defined in such a way that

fibrations and equivalences are defined on the underlying objects in sM . The model
category sM/(X∗ , F ) is also the comma model category sM/(X∗ × F ) where F is
considered as a constant simplicial object in M .
Definition 1.3.5.2. An object Y∗ ∈ sM/(X∗ , F ) is a X∗ -torsor on F if it satisfies
the following two conditions.
(1) For all morphism [n] → [m] in ∆, the square
Ym / Yn

 
Xm / Xn
is homotopy cartesian.
(2) The natural morphism
|Y∗ | −→ F
is an isomorphism in Ho(M ).
The space of X∗ -torsors over F , denoted by T orsX∗ (F ), is the nerve of the sub category
of fibrant objects sM/(X∗ , F )f , consisting of equivalences between X∗ -torsors on F .
Suppose that f : F −→ F ′ is a morphism between fibrant objects in M . We get
a pullback functor
sM/(X∗ , F ′ ) −→ sM/(X∗ , F ),
which is right Quillen, and such that the induced functor on fibrant objects
sM/(X∗ , F ′ )f −→ sM/(X∗ , F )f
sends X∗ -torsors over F to X∗ -torsors over F ′ (this uses the commutation of homotopy
colimits with homotopy pullbacks). Therefore, restricting to the sub categories of
equivalences, we get a well defined morphism between spaces of torsors
f ∗ : T orsX∗ (F ′ ) −→ T orsX∗ (F ).
By applying the standard strictification procedure, we can always suppose that (f ◦
g)∗ = g ∗ ◦f ∗ . This clearly defines a functor from (M f )op , the opposite full subcategory
of fibrant objects in M f , to SSet
T orsX∗ : (M f )op −→ SSet.
This functor sends equivalences in M f to equivalences of simplicial sets, and therefore
induces a Ho(SSet)-enriched functor (using for example [D-K1])
T orsX∗ : Ho(M f )op ≃ Ho(M )op −→ Ho(SSet).
In other words, there are natural morphisms in Ho(SSet)
M apM (F, F ′ ) −→ M apSSet (T orsX∗ (F ′ ), T orsX∗ (F )),
compatible with compositions.
The main classification result is the following. It gives a way to describe the stack
associated to |X∗ | for some Segal groupoid object X∗ in M .
Proposition 1.3.5.3. Let X∗ be a Segal groupoid object in M and Z be a fibrant
model for |X∗ | in M . Then, there exists an element α ∈ π0 (T orsX∗ (Z)), such that
for any fibrant object F ∈ M , the evaluation at α
α∗ : M apM (F, Z) −→ T orsX∗ (F )
is an isomorphism in Ho(SSet).
86 1.3. GEOMETRIC STACKS: BASIC THEORY

Proof. We can of course assume that X∗ is cofibrant in sM , as everything is

invariant by changing X∗ with something equivalent. Let Colimn Xn −→ Z be a
morphism in M such that Z is fibrant and such that the induced morphism
|X∗ | −→ Z
is an equivalence. Such a morphism exists as Colimn Xn is cofibrant in M and com-
putes the homotopy colimit |X∗ |. The element α ∈ π0 (T orsX∗ (Z)) is defined to be
the pair (X∗ , p) ∈ Ho(sM/(X∗ , Z)), consisting of the identity of X∗ and the natural
augmentation p : Colimn Xn −→ Z. Clearly, α is a X∗ -torsor over Z.
Applying Cor. A.0.5 of Appendix A, we see that there exists a homotopy fiber
sequence
M apM (F, Z) / N ((M/Z)f ) / N (M f )
W W

where the (M/Z)fW (resp. f

MW )is the subcategory of equivalences between fibrant
objects in M/Z (resp. in M ), and the morphism on the right is the forgetful functor
and the fiber is taken at F . In the same way, there exists a homotopy fiber sequence
T orsX∗ (F ) / N (((sM/X∗ )f )cart ) / N (M f )
W W

where (sM/X∗ )fW )cart is the subcategory of sM/X∗ consisting of equivalence between
fibrant objects satisfying condition (1) of Def. 1.3.5.2, and where the morphism on
the right is induced my the homotopy colimit functor of underlying simplicial objects
and the fiber is taken at F . There exists a morphism of homotopy fiber sequences of
simplicial sets
M apM (F, Z) / N ((M/Z)f ) / N (M f )
W W

α∗ Id
  
T orsX∗ (F ) / N (((sM/X∗ )f )cart ) / N (M f )
W W

and the arrow in the middle is an equivalence because of our proposition 1.3.5.1. 

1.3.6. Properties of morphisms

We fix another class Q of morphisms in Af fC , which is stable by equivalences,
compositions and homotopy pullbacks. Using the Yoneda embedding the notion of
morphisms in Q (or simply Q-morphisms) is extended to a notion of morphisms
between representable stacks.
Definition 1.3.6.1. We say that morphisms in Q are compatible with τ and
P (or equivalently that morphisms in Q are local with respect to τ and P) if the
following two conditions are satisfied:
(1) If f : X −→ Y is a morphism in Af fC such that there exists a covering
family
{Ui −→ X}
with each composite morphism Ui −→ Y in Q, then f belongs to Q.
(2) If f : X −→ Y is a morphism in Af fC and there exists a covering family
{Ui −→ Y }
such that each homotopy pullback morphism
X ×hY Ui −→ Ui
is in Q, then f belongs to Q.
 1.3.6. PROPERTIES OF MORPHISMS 87

For a class of morphism Q, compatible with τ and P in the sense above we can
make the following definition.
Definition 1.3.6.2. Let Q be a class of morphisms in Af fC , stable by equiva-
lences, homotopy pullbacks and compositions, and which is compatible with τ and P
in the sense above. A morphism of stacks f : F −→ G is in Q (or equivalently is a
Q-morphism) if it is n-representable for some n, and if for any representable stack X
and any morphism X −→ G there exists an n-atlas {Ui } of F ×hG X such that each
morphism Ui −→ X between representable stacks is in Q.
Clearly, because of our definition 1.3.6.1, the notion of morphism in Q of definition
1.3.6.2 is compatible with the original notion. Furthermore, it is easy to check, as it
was done for P-morphisms, the following proposition.
Proposition 1.3.6.3. (1) Morphisms in Q are stable by equivalences, com-
positions and homotopy pullbacks.
(2) Let f : F −→ G be any morphism between n-geometric stacks. We suppose
that there exists a n-atlas {Ui } of G such that each projection F ×hX Ui −→ Ui
is in Q. Then f is in Q.
Proof. Exercise. 

We can also make the following two general definitions of morphisms of stacks.
Definition 1.3.6.4. Let f : F −→ G be a morphism of stacks.
(1) The morphism is categorically locally finitely presented if for any repre-
sentable stack X = RSpec A, any morphism X −→ G, and any U-small
filtered system of commutative A-algebras {Bi }, the natural morphism
Hocolimi MapAf fC∼,τ /X (RSpec Bi , F ×hG X) −→ MapAf fC∼,τ /X (RSpec (Hocolimi Bi ), F ×hG X)
is an isomorphism in Ho(SSet).
(2) The morphism f is quasi-compact if for any representable stack X and any
morphism X −→ G there exists a finite family of representable stacks {Xi }
and an epimorphism
a
Xi −→ F ×hG X.
i

(3) The morphism f is categorically finitely presented if it is categorically locally

finitely presented and quasi-compact.
(4) The morphism f is a monomorphism if the natural morphism
F −→ F ×hG F
is an isomorphism in St(C, τ ).
(5) Assume that the class Q of finitely presented morphism in Af fC is compatible
with the model topology τ in the sense of Def. 1.3.6.1. The morphism f is
locally finitely presented if it is a Q-morphism in the sense of Def. 1.3.6.2.
It is a finitely presented morphism if it is quasi-compact and locally finitely
presented.
Clearly, the above notions are compatible with the one of definition 1.2.6.1, in
the sense that a morphism of commutative monoids A −→ B has a certain property
in the sense of Def. 1.2.6.1 if and only if the corresponding morphism of stacks
RSpec B −→ RSpec A has the same property in the sense of Def. 1.3.6.4.
88 1.3. GEOMETRIC STACKS: BASIC THEORY

Proposition 1.3.6.5. Quasi-compact morphisms, categorically (locally) finitely

presented morphisms, (locally) finitely presented morphisms and monomorphisms are
stable by equivalences, composition and homotopy pullbacks.
Proof. Exercise. 
Remark 1.3.6.6. When the class of finitely presented morphisms in Af fC is
compatible with the topology τ , Def. 1.3.6.4 gives us two different notions of locally
finitely presented morphism which are a priori rather difficult to compare. Giving
precise conditions under which they coincide is however not so important as much
probably they are already different in some of our main examples (e.g. for each
example for which the representable stacks are not truncated, as in the complicial,
or brave new algebraic geometry presented in §2.3 and §2.4). In this work we have
chosen to use only the non categorical version of locally finitely presented morphisms,
the price to pay being of course that they do not have easy functorial characterization.

1.3.7. Quasi-coherent modules, perfect modules and vector bundles

For a commutative monoid A in C, we define a category A − QCoh, of quasi-
coherent modules on A (or equivalently on Spec A) in the following way. Its objects
are the data of a B-module MB for any commutative A-algebra B ∈ A − Comm(C),
together with an isomorphism
αu : MB ⊗B B ′ −→ MB ′
for any morphism u : B −→ B ′ in A−Comm(C), such that one has αv ◦(αu ⊗B ′ B ′′ ) =
αv◦u for any pair of morphisms
u / B′ v / B ′′
B
in A−Comm(C). Such data will be denoted by (M, α). A morphism in A−QCoh, from
(M, α) to (M ′ , α′ ) is given by a family of morphisms of B-modules fB : MB −→ MB′ ,
for any B ∈ A − Comm(C), such that for any u : B → B ′ in A − Comm(C) the
diagram
αu
MB ⊗B B ′ / MB ′

fB ⊗B B ′ fB ′
 
(MB′ ) ⊗B B ′ / M′ ′
B
α′u

commutes. As the categories A − M od and Comm(C) are all V-small, so are the
categories A − QCoh.
There exists a natural projection A − QCoh −→ A − M od, sending (M, α) to
MA , and it is straightforward to check that it is an equivalence of categories. In
particular, the model structure on A−M od will be transported naturally on A−QCoh
through this equivalence. Fibrations (resp. equivalences) in A − QCoh are simply the
morphisms f : (M, α) −→ (M ′ , α′ ) such that fA : MA −→ MA′ is a fibration (resp.
an equivalence).
Let now f : A −→ A′ be a morphism of commutative monoids in C. There exists
a pullback functor
f ∗ : A − QCoh −→ A′ − QCoh
defined by f ∗ (M, α)B := MB for any B ∈ A − Comm(C), and for u : B −→ B ′ in
A − Comm(C) the transition morphism
f ∗ (M, α)B ⊗B B ′ = MB ⊗B B ′ −→ f ∗ (M, α)B ′ = MB ′
 1.3.7. QUASI-COHERENT MODULES, PERFECT MODULES AND VECTOR BUNDLES 89

is given by αu . By definition of the model structure on A − QCoh, the functor

f ∗ : A − QCoh −→ A′ − QCoh
is clearly a left Quillen functor, as the natural diagram
f∗
A − QCoh / A′ − QCoh

 
A − M od / A′ − M od
∗
f

commutes up to a natural isomorphism. Furthermore, for any pair of morphisms

f g
A / A′ / A′′

in Comm(C), there is an equality (g ◦ f )∗ = g ∗ ◦ f ∗ . In other words, the rule

A 7→ A − QCoh (f : A → A′ ) 7→ f ∗
defines a U-cofibrantly generated left Quilllen presheaf on Af fC = Comm(C)op in the
sense of Appendix B.
We now consider for any A ∈ Comm(C), the subcategory A − QCohcW of A −
QCoh, consisting of equivalences between cofibrant objects. As these are preserved
by the pullback functors f ∗ , we obtain this way a new presheaf of V-small categories
QCohcW : Comm(C) = Af fCop −→ CatV
A 7→ A − QCohcW .
Composing with the nerve functor
N : CatV −→ SSetV
we get a simplicial presheaf
N (QCohcW ) : Comm(C) = Af fCop −→ SSetV
A 7→ N (A − QCohcW ).
Definition 1.3.7.1. The simplicial presheaf of quasi-coherent modules is
N (QCohcW ) defined above. It is denoted by QCoh, and is considered as an object in
Af fC∼,τ .
It is important to note that for any A ∈ Comm(C), the simplicial set QCoh(A)
is naturally equivalent to the nerve of A − M odcW , the subcategory of equivalences
between cofibrant objects in A−M od, and therefore also to the nerve of A−M odW , the
subcategory of equivalences in A − M od. In particular, π0 (QCoh(A)) is in bijection
with isomorphisms classes of objects in Ho(A − M od) (i.e. equivalence classes of
objects in A − M od). Furthermore, by [D-K3] (see also Appendix A), for any x ∈
QCoh(A), corresponding to an equivalence class of M ∈ A − M od, the connected
component of QCoh(A) containing x is naturally equivalent to BAut(M ), where
Aut(M ) is the simplicial monoid of self equivalences of M in A − M od. In particular,
we have
π1 (QCoh(A), x) ≃ AutHo(A−Mod) (M )
πi+1 (QCoh(A), x) ≃ [S i M, M ]A−Mod ∀ i > 1.
The main result of this section is the following.
Theorem 1.3.7.2. The simplicial presheaf QCoh is a stack.
90 1.3. GEOMETRIC STACKS: BASIC THEORY

Proof. This is a direct application of the strictification theorem B.0.7 recalled

in Appendix B.
More precisely, we use our lemma 1.3.2.3 (2). Concerning finite direct sums, we
have already seen (during the proof of lemma 1.3.2.3) that the natural functor
(− ⊗A×h B A) × (− ⊗A×h B B) : (A ×h B) − M od −→ A − M od × B − M od
is a Quillen equivalence. This implies that for any two objects X, Y ∈ Af fC the
natural morphism
L
a
QCoh(X Y ) −→ QCoh(X) × QCoh(Y )

is an equivalence. It only remains to show that QCoh has the descent property with
respect to hypercovers of the type described in lemma 1.3.2.3 (2). But this is nothing
else than our assumption 1.3.2.2 (3) together with Cor. B.0.8. 

An important consequence of theorem 1.3.7.2 is the following.

Corollary 1.3.7.3. Let A ∈ Comm(C) be a commutative monoid, A − M odW
the subcategory of equivalences in A − M od, and N (A − M odW ) be its nerve. Then,
there exists natural isomorphisms in Ho(SSet)
RQCoh(A) ≃ Rτ Hom(RSpec A, QCoh) ≃ N (A − M odW ).
To finish this section, we will describe two important sub-stacks of QCoh, namely
the stack of perfect modules and the stack of vector bundles.
For any commutative monoid A in C, we let Perf (A) be the sub-simplicial set of
QCoh(A) consisting of all connected components corresponding to perfect objects in
Ho(A − M od) (in the sense of Def. 1.2.3.6). More precisely, if Iso(D) denotes the set
of isomorphisms classes of a category D, the simplicial set Perf (A) is defined as the
pullback
Perf (A) / Iso(Ho(A − M od)perf )

 
QCoh(A) / π0 QCoh(A) ≃ Iso(Ho(A − M od))

where Ho(A − M od)perf is the full subcategory of Ho(A − M od) consisting of perfect
A-modules in the sense of Def. 1.2.3.6.
We say that an A-module M ∈ Ho(A − M od) is a rank n vector bundle, if there
exists a covering family A −→ A′ such that M ⊗LA A′ is isomorphic in Ho(A′ −
M od) to (A′ )n . As we have defined the sub simplicial set Perf (A) of QCoh(A) we
define Vectn (A) to be the sub simplicial set of QCoh(A) consisting of connected
components corresponding to rank n vector bundles.
For any morphism of commutative monoids u : A −→ A′ , the base change functor
Lu∗ : Ho(A − M od) −→ Ho(A′ − M od)
preserves perfect modules as well as rank n vector bundles. Therefore, the sub sim-
plicial sets Vectn (A) and Perf form in fact full sub simplicial presheaves
Vectn ⊂ QCoh Perf ⊂ QCoh.
The simplicial presheaves Vectn and Perf then define objects in Af fC∼,τ .
Corollary 1.3.7.4. The simplicial presheaves Perf and Vectn are stacks.
 1.3.7. QUASI-COHERENT MODULES, PERFECT MODULES AND VECTOR BUNDLES 91

Proof. Indeed, as they are full sub-simplicial presheaves of the stack QCoh, it
is clearly enough to show that being a perfect module and being a vector bundle or
rank n is a local condition for the topology τ . For vector bundles this is obvious from
the definition.
Let A ∈ Comm(C) be a commutative monoid, and P be an A-module, such
that there exists a τ -covering A −→ B such that P ⊗LA B is a perfect B-module.
Assume that A → B is a cofibration, and let B∗ be its co-nerve, considered as a co-
simplicial object in A − Comm(C). Let Q be any A-module, and define two objects
in Ho(csB∗ − M od) by
RHomA (P, Q)∗ := RHomA (P, Q) ⊗LA B∗
RHomA (P, Q∗ ) := RHomA (P, Q ⊗LA B∗ ).
There is a natural morphism
RHomA (P, Q)∗ −→ RHomA (P, Q∗ ).
These co-simplicial objects are both cartesian, and by assumption 1.3.2.2 (3) applied
to the A-modules Q and RHomA (P, Q) the induced morphism in Ho(A − M od)
Z
RHomA (P, Q)∗ ≃ RHomA (P, Q) ≃
Z
≃ Holimn∈∆RHomA (P, Q ⊗LA Bn ) ≃ RHomA (P, Q ⊗LA B∗ )

is an isomorphism. Therefore, assumption 1.3.2.2 (3) implies that the natural mor-
phism
RHomA (P, Q)0 −→ RHomA (P, Q0 )
is an isomorphism. By definition this implies that
RHomA (P, Q) ⊗LA B −→ RHomA (P, Q ⊗LA B)
is an isomorphism in Ho(A − M od). In particular, when Q = A we find that the
natural morphism
P ∨ ⊗LA B −→ RHomB (P ⊗LA B, B)
is an isomorphism in Ho(B − M od). As P ⊗LA B is by assumption a perfect B-module,
we find that the natural morphism
P ∨ ⊗LA Q −→ RHom(P, Q)
becomes an isomorphism after base changing to B, and this for any A-module Q. As
A −→ B is a τ -covering, we see that this implies that
P ∨ ⊗LA Q −→ RHom(P, Q)
is always an isomorphism in Ho(A − M od), for any Q, and thus that P is a perfect
A-module. 
Definition 1.3.7.5. The stack of vector bundles of rank n is Vectn . The stack
of perfect modules is Perf .
The same construction can also been done in the stable context. For a commu-
tative monoid A in C, we define a category A − QCohSp , of stable quasi-coherent
modules on A (or equivalently on Spec A) in the following way. Its objects are the
data of a stable B-module MB ∈ Sp(B − M od) for any commutative A-algebra
B ∈ A − Comm(C), together with an isomorphism
αu : MB ⊗B C −→ MB ′
92 1.3. GEOMETRIC STACKS: BASIC THEORY

for any morphism u : B −→ B ′ in A−Comm(C), such that one has αv ◦(αu ⊗B ′ B ′′ ) =

αv◦u for any pair of morphisms
u / B′ v / B ′′
B
in A − Comm(C). Such data will be denoted by (M, α). A morphism in A − QCohSp ,
from (M, α) to (M ′ , α′ ) is given by a family of morphisms of stable B-modules fB :
MB −→ MB′ , for any B ∈ A − Comm(C), such that for any u : B → B ′ in A −
Comm(C) the diagram
αu
MB ⊗B B ′ / MB ′

fB ⊗B B ′ fB ′
 
(MB′ ) ⊗B B ′ / M′ ′
B
α′u

commutes. As the categories Sp(A − M od) and Comm(C) are all V-small, so are the
categories A − QCohSp .
There exists a natural projection A − QCohSp −→ Sp(A − M od), sending (M, α)
to MA , and it is straightforward to check that it is an equivalence of categories. In
particular, the model structure on Sp(A − M od) will be transported naturally on
A − QCohSp through this equivalence.
Let now f : A −→ A′ be a morphism of commutative monoids in C. There exists
a pullback functor
f ∗ : A − QCohSp −→ A′ − QCohSp
defined by f (M, α)B := MB for any B ∈ A − Comm(C), and for u : B −→ B ′ in
A − Comm(C) the transition morphism
f (M, α)B ⊗B B ′ = MB ⊗B B ′ −→ f (M, α)B ′ = MB ′
is given by αu . By definition of the model structure on A − QCohSp , the functor
f ∗ : A − QCohSp −→ A′ − QCohSp
is clearly a left Quillen functor. Furthermore, for any pair of morphisms
f g
A / A′ / A′′

in Comm(C), there is an equality (g ◦ f )∗ = g ∗ ◦ f ∗ . In other words, the rule

A 7→ A − QCohSp (f : A → A′ ) 7→ f ∗
defines a U-cofibrantly generated left Quilllen presheaf on Af fC = Comm(C)op in the
sense of Appendix B.
We now consider for any A ∈ Comm(C), the subcategory A − QCohSp,c W of A −
QCohSp , consisting of equivalences between cofibrant objects. As these are preserved
by the pullback functors f ∗ , one gets this way a new presheaf of V-small categories
QCohSp,c
W : Comm(C) = Af fCop −→ CatV
A 7→ A − QCohSp,c
W .
Composing with the nerve functor
N : CatV −→ SSetV
one gets a simplicial presheaf
N (QCohSp,c op
W ) : Comm(C) = Af fC −→ SSetV
A 7→ N (A − QCohSp,c
W ).
 1.3.7. QUASI-COHERENT MODULES, PERFECT MODULES AND VECTOR BUNDLES 93

Definition 1.3.7.6. The simplicial presheaf of stable quasi-coherent modules is

N (QCohSp,c Sp
W ) defined above. It is denoted by QCoh , and is considered as an object
∼,τ
in Af fC .
It is important to note that for any A ∈ Comm(C), the simplicial set QCohSp (A)
is naturally equivalent to the nerve of A − M odSp,c W , the subcategory of equiva-
lences between cofibrant objects in Sp(A − M od), and therefore also to the nerve
of Sp(A − M od)W , the subcategory of equivalences in Sp(A − M od). In particular,
π0 (QCohSp (A)) is in bijection with isomorphisms classes of objects in Ho(Sp(A −
M od)) (i.e. equivalence classes of objects in Sp(A − M od)). Furthermore, by [D-K3]
(see also Appendix A), for any x ∈ QCohSp (A), corresponding to an equivalence
class of M ∈ Sp(A − M od), the connected component of QCohSp (A) containing x
is naturally equivalent to BAut(M ), where Aut(M ) is the simplicial monoid of self
equivalences of M in Sp(A − M od). In particular, we have
π1 (QCohSp (A), x) ≃ AutHo(Sp(A−Mod)) (M )
πi+1 (QCohSp (A), x) ≃ [S i M, M ]Sp(A−Mod) ∀ i > 1.
The same proof as theorem 1.3.7.2, but based on Proposition 1.2.12.5 gives the
following stable version.
Theorem 1.3.7.7. Assume that the two conditions are satisfied.
(1) The suspension functor S : Ho(C) −→ Ho(C) is fully faithful.
(2) For all τ -covering family {Ui → X} in Af fC , each morphism Ui → X is
flat in the sense of Def. 1.2.4.1.
Then, the simplicial presheaf QCohSp is a stack.
For any commutative monoid A in C, we let Perf Sp (A) be the sub-simplicial set of
QCohSp (A) consisting of all connected components corresponding to perfect objects
in Ho(Sp(A − M od)) (in the sense of Def. 1.2.3.6). This defines a full sub-prestack
Perf Sp of QCohSp . Then, the same argument as for Cor. 1.3.7.4 gives the following
corollary.
Corollary 1.3.7.8. Assume that the two conditions are satisfied.
(1) The suspension functor S : Ho(C) −→ Ho(C) is fully faithful.
(2) For all τ -covering family {Ui → X} in Af fC , each morphism Ui → X is
flat in the sense of Def. 1.2.4.1.
The simplicial presheaf Perf Sp is a stack.
This justifies the following definition.
Definition 1.3.7.9. Under the condition of Cor. 1.3.7.8, the stack of stable
perfect modules is Perf Sp .
We finish this section by the standard description of Vectn as the classifying stack
of the group stack Gln , of invertible n by n matrices. For this, we notice that the
natural morphism of stacks ∗ −→ Vectn , pointing the trivial rank n vector bundle,
induces an isomorphism of sheaves of sets ∗ ≃ π0 (Vectn ). Therefore, the stack Vectn
can be written as the geometric realization of the homotopy nerve of the morphism
∗ −→ Vectn . In other words, we can find a Segal groupoid object X∗ , such that
X0 = ∗, and with |X∗ | ≃ Vectn . Furthermore, the object X1 is naturally equivalent
to the loop stack Ω∗ Vectn := ∗ ×hVectn ∗. By construction and by Prop. A.0.6, this
loop stack can be described as the simplicial presheaf
Ω∗ Vectn : Comm(C) −→ SSet
A 7→ M ap′C (1n , An ),
94 1.3. GEOMETRIC STACKS: BASIC THEORY

where M ap′C (1n , An ) is the sub simplicial set of the mapping space M apC (1n , An )
consisting of all connected components corresponding to automorphisms in
π0 M apC (1n , An ) ≃ π0 M apA−Mod (An , An ) ≃ [An , An ]A−Mod .
The important fact concerning the stack Ω∗ Vectn is the following result.
Proposition 1.3.7.10. (1) The stack Ω∗ Vectn is representable and the mor-
phism Ω∗ Vectn −→ ∗ is formally smooth.
(2) If Moreover 1 is finitely presented in C, then the morphism Ω∗ Vectn −→ ∗
is finitely presented (and thus smooth by (1)).
Proof. (1) We start by defining a larger stack Mn , of n by n matrices. We set
Mn : Comm(C) −→ SSet
A 7→ M apC (1n , An ).
This stack is representable, as it is isomorphic in St(C, τ ) to RSpec B, where
2 2
B = LF (1n ) is the free commutative monoid generated by the object 1n ∈ C. We
claim that the natural inclusion morphism
Ω∗ Vectn −→ Mn
is (−1)-representable and a formally étale morphism. Indeed, let A be any commu-
tative monoid,
x : X := RSpec A −→ Mn
be a morphism of stacks, and let us consider the stack
F := Ω∗ Vectn ×hMn X −→ X.
The point x corresponds via the Yoneda lemma to a morphism u : An −→ An in
Ho(A − M od). Now, for any commutative monoid A′ , the natural morphism
RF (A′ ) −→ (RSpec A)(A′ )
identifies RF (A′ ) with the sub simplicial set of (RSpec A)(A′ ) ≃ M apComm(C)(A, A′ )
consisting of all connected components corresponding to morphisms A −→ A′ in
Ho(Comm(C)) such that
u ⊗LA A′ : (A′ )n −→ (A′ )n
is an isomorphism in Ho(A′ − M od). Considering u as an element of [An , An ] ≃
Mn (π0 (A)), we can consider its determinant d(u) ∈ π0 (A). Then, using notations
from Def. 1.2.9.2, we clearly have an isomorphism of stacks
F ≃ RSpec (A[d(u)−1 ]).
By Prop. 1.2.9.5 this shows that the morphism
F −→ RSpec A
is a formally étale morphism between representable stacks. As Mn is representable
this implies that Ω∗ Vectn is a representable stack and that the morphism
Ω∗ Vectn −→ Mn
is formally étale (it is also a flat monomorphism by 1.2.9.4).
2
Moreover, we have Mn ≃ RSpec B, where B := LF (1n ) is the derived free
2
commutative monoid over 1n . This implies that LB is a free B-module of rank n2 ,
and therefore that the morphism 1 −→ B is formally smooth in the sense of Def.
1.2.7.1. By composition, we find that Ω∗ Vectn −→ ∗ is a formally smooth morphism
as required.
 1.3.7. QUASI-COHERENT MODULES, PERFECT MODULES AND VECTOR BUNDLES 95

2
(2) This follows from (1), Prop. 1.2.9.4 (2) and the fact that B = LF (1n ) is a
finitely presented object in Comm(C). 
Definition 1.3.7.11. (1) The stack Ω∗ Vectn is denoted by Gln , and is
called the linear group stack of rank n. The stack Gl1 is denoted by Gm ,
and is called the multiplicative group stack.
(2) The stack Mn defined during the proof of 1.3.7.10 (1) is called the stack of
n × n matrices. The stack M1 is denoted by Ga , and is called the additive
group stack.
Being a stack of loops, the stack Gln = Ω∗ Vectn has a natural group structure,
encoded in the fact that it is the X1 of a Segal groupoid object X∗ with X0 = ∗.
Symbolically, we will simply write
BGln := |X∗ |.
Our conclusion is that the stack Vectn can be written as BGln , where Gln is a
formally smooth representable group stack. Furthermore this group stack is smooth
when the unit 1 is finitely presented. As a corollary we get the following geometricity
result on Vectn .
Corollary 1.3.7.12. Assume that the unit 1 is a finitely presented object in
C. Assume that all smooth morphisms in Comm(C) belong to P. Then, the stack
Vectn is 1-geometric, the morphism Vectn −→ ∗ is in P and finitely presented, and
furthermore its diagonal is a (−1)-representable morphism.
Proof. The 1-geometricity statement is a consequence of Prop. 1.3.4.2 and the
fact that the natural morphism ∗ −→ Vectn is a 1-P-atlas. That ∗ is a 1-P-atlas also
implies that Vectn −→ ∗ is in P and finitely presented. The statement concerning
the diagonal follows from the fact that Gln is a representable stack and the locality
of representable objects Prop. 1.3.2.8. 
We finish with an analogous situation for perfect modules. We let K be a perfect
object in C, and we define a stack REnd(K) in the following way. We chose a cofibrant
replacement QK of K, and Γ∗ a simplicial resolution functor on C. One sets
REnd(K) : Comm(C) −→ SSetV
A 7→ Hom(QK, Γ∗ (QK ⊗ A)).
Note that for any A the simplicial set REnd(K)(A) is naturally equivalent to
M apA−Mod (K ⊗L A, K ⊗L A).
Lemma 1.3.7.13. The simplicial presheaf REnd(K)(A) ∈ Af fC∼,τ is a stack. It
is furthermore representable.
Proof. This is clear as K being perfect one sees that there exists an isomorphism
in Ho(SP r(Af fC ))
REnd(K) ≃ RSpec B,
where B := LF (K ⊗L K ∨ ) is the derived free commutative monoid on the object
K ⊗L K ∨ . 
We now define a sub-stack RAut(K) of REnd(K). For a commutative monoid
A ∈ Comm(C), we define RAut(K)(A) to be the union of connected components of
REnd(K)(A) corresponding to isomorphisms in
π0 (REnd(K)(A)) ≃ [K ⊗L A, K ⊗L A]A−Mod .
This clearly defines a full sub-simplicial presheaf RAut(K) of REnd(K), which is a
sub-stack as one can see easily using Cor. 1.3.2.7.
96 1.3. GEOMETRIC STACKS: BASIC THEORY

Proposition 1.3.7.14. Assume that C is a stable model category. Then, the stack
RAut(K) is representable. Furthermore the morphism RAut(K) −→ REnd(K) is a
formal Zariski open immersion, and RAut(K) −→ ∗ is fp. If furthermore 1 is finitely
presented in C then RAut(K) −→ ∗ is a perfect morphism.
Proof. It is the same as 1.3.7.10 but using the construction AK and Prop.
1.2.10.1, instead of the standard localization A[a−1 ]. More precisely, for a repre-
sentable stack X := RSpec A and a morphism x : X −→ REnd(K), the homotopy
pullback
RAut(K) ×hREnd(K) X −→ X
is isomorphic in Ho(Af fC∼,τ /X) to
RSpec AE −→ RSpec A,
where E is the homotopy cofiber of the endomorphism x : K ⊗L A −→ K ⊗L A
corresponding to the point x. 
 CHAPTER 1.4

Geometric stacks: Infinitesimal theory

As in the previous chapter, we fix once for all a HAG context (C, C0 , A, τ, P).

In this chapter we will assume furthermore that the suspension functor

S : Ho(C) −→ Ho(C)
is fully faithful. In particular, for any commutative monoid A ∈ Comm(C), the
stabilization functor
SA : Ho(A − M od) −→ Ho(Sp(A − M od))
is fully faithful (see 1.2.11.2). We will therefore forget to mention the functor SA
and simply consider A-modules as objects in Ho(Sp(A − M od)), corresponding to
0-connective objects.

1.4.1. Tangent stacks and cotangent complexes

We consider the initial commutative monoid 1 ∈ Comm(C). It can be seen as a
module over itself, and gives rise to a trivial square zero extension 1 ⊕ 1 (see §1.2.1).
Definition 1.4.1.1. The dual numbers over C is the commutative monoid 1 ⊕ 1,
and is denoted by 1[ǫ]. The corresponding representable stack is denoted by
Dǫ := RSpec (1[ǫ])
and is called the infinitesimal disk.
Of course, as every trivial square zero extension the natural morphism 1 −→ 1[ǫ]
possesses a natural retraction 1[ǫ] −→ 1. On the level of representable stacks this
defines a natural global point
∗ −→ Dǫ .
We recal that Ho(Af fC∼,τ ) being the homotopy category of a model topos has internal
Hom’s objects Rτ Hom(−, −). They satisfy the usual adjunction isomorphisms
M apAf fC∼,τ (F, Rτ Hom(G, H)) ≃ M apAf fC∼,τ (F ×h G, H).

Definition 1.4.1.2. Let F ∈ St(C, τ ) be a stack. The tangent stack of F is

defined to be
T F := Rτ Hom(Dǫ , F ).
The natural morphism ∗ −→ Dǫ induces a well defined projection
π : T F −→ F,
and the projection Dǫ −→ ∗ induces a natural section of π
e : F −→ T F.
97
98 1.4. GEOMETRIC STACKS: INFINITESIMAL THEORY

For any commutative monoid A ∈ Comm(C), it is clear that there is a natural

equivalence
A ⊗L (1[ǫ]) ≃ A ⊕ A,
where A ⊕ A is the trivial square zero extension of A by itself. We will simply denote
A ⊕ A by A[ǫ], and
DA
ǫ := RSpec (A[ǫ]) ≃ RSpec (A) × Dǫ

the infinitesimal disk over A.

With these notations, and for any fibrant object F ∈ Af fC∼,τ , the stack T F can
be described as the following simplicial presheaf
T F : Comm(C) = Af fCop −→ SSetV
A 7→ F (A[ǫ]).
Note that if F is fibrant, then so is T F as defined above. In other words for any
A ∈ Comm(C) there exists a natural equivalence of simplicial sets
RT F (A) ≃ RF (A[ǫ]).
Proposition 1.4.1.3. The functor F 7→ T F commutes with V-small homotopy
limits.
Proof. This is clear as Rτ Hom(H, −) always commutes with homotopy limits
for any H. 

Let F ∈ St(C, τ ) be a stack, A ∈ Comm(C) a commutative monoid and

x : RSpec A −→ F
be a A-point. Let M be an A-module, and let A ⊕ M be the trivial square zero
extension of A by M . Let us fix the following notations
X := RSpec A X[M ] := RSpec (A ⊕ M ).
The natural augmentation A ⊕ M −→ A gives rise to a natural morphism of stacks
X −→ X[M ].
Definition 1.4.1.4. Let x : X −→ F be as above. The simplicial set of deriva-
tions from F to M at the point x is defined by
DerF (X, M ) := M apX/Af fC∼,τ (X[M ], F ) ∈ Ho(SSetV ).
It will also denoted by DerF (x, M ).
As the construction M 7→ X[M ] is functorial in M , we get this way a well defined
functor
DerF (X, −) : Ho(A − M od) −→ Ho(SSetV ).
This functor is furthermore naturally compatible with the Ho(SSet)-enrichment, in
the sense that there exists natural morphisms
M apA−Mod (M, N ) −→ M apSSet (DerF (X, M ), DerF (X, N ))
which are compatible with compositions. Note that the morphism X −→ X[M ] has
a natural retraction X[M ] −→ X, and therefore that the simplicial set DerF (X, M )
has a distinguished base point, the trivial derivation. The functor DerF (X, −) takes
its values in the homotopy category of pointed simplicial sets.
We can also describe the functor DerF (X, M ) using functors of points in the
following way. Let F ∈ Af fC∼,τ be a fibrant object. For any commutative monoid A,
 1.4.1. TANGENT STACKS AND COTANGENT COMPLEXES 99

any A-module M and any point x ∈ F (A), we consider the standard homotopy fiber1
of
F (A ⊕ M ) −→ F (A)
at the point x. This clearly defines a functor
A − M od −→ SSetV
M 7→ Hof iber (F (A ⊕ M ) → F (A))
which is a lift of the functor considered above
DerF (X, −) : Ho(A − M od) −→ Ho(SSetV ),
where x ∈ F (A) corresponds via the Yoneda lemma to a morphism X = RSpec A −→
F . The fact that the functor DerF (X, −) has a natural lift as above is important, as it
then makes sense to say that it commutes with homotopy limits or homotopy colimits.
The functor DerF (X, −) can be considered as an object in Ho((A − M odop )∧ ), the
homotopy category of pre-stacks on the model category A − M odop , as defined in
[HAGI, §4.1] (see also §1.3.1). Restricting to the subcategory A − M od0 we get an
object
DerF (X, −) ∈ Ho((A − M odop ∧
0 ) ).
In the sequel the functor DerF (X, −) will always be considered as an object in Ho((A−
M odop ∧
0 ) ).

Definition 1.4.1.5. Let F be any stack and let A ∈ A.

(1) Let x : X := RSpec A −→ F be an A-point. We say that F has a cotangent
complex at x if there exists an integer n, an (−n)-connective stable A-module
LF,x ∈ Ho(Sp(A − M od)), and an isomorphism in Ho((A − M odop ∧
0 ) )

DerF (X, −) ≃ RhLs F,x .

(2) If F has a cotangent complex at x, the stable A-module LF,x is then called
the cotangent complex of F at x.
(3) If F has a cotangent complex at x, the tangent complex of F at x is then
the stable A-module
TF,x := RHomSp
A (LF,x , A) ∈ Ho(Sp(A − M od)).

In other words, the existence of a cotangent complex of F at x : X := RSpec A −→

F is equivalent to the co-representability of the functor
DerF (X, −) : A − M od0 −→ SSetV ,
by some (−n)-connective object LF,x ∈ Ho(Sp(A − M od)). The fact that LF,x is well
defined is justified by our Prop. 1.2.11.3.

The first relation between cotangent complexes and the tangent stack is given by
the following proposition.
Proposition 1.4.1.6. Let F be a stack and x : X := RSpec A −→ F be an A-
point with A ∈ A. If F has a cotangent complex LF,x at the point x then there exists
a natural isomorphism in Ho(SSetV )
RHomAf fC∼,τ /F (X, T F ) ≃ M apSp(A−Mod) (A, TF,x ) ≃ M apSp(A−Mod) (LF,x , A).

1The standard homotopy fiber product of a diagram /Zo

X Y of fibrant simplicial
sets is defined for example by
1
∆
X ×h
Z Y := (X × Y ) ×Z×Z Z .
100 1.4. GEOMETRIC STACKS: INFINITESIMAL THEORY

Proof. By definition of T F , there is a natural isomorphism

RHomAf fC∼,τ /F (X, T F ) ≃ RHomAf fC∼,τ /X (X[A], T F ) ≃ M apSp(A−Mod) (LF,x , A).
Moreover, by definition of the tangent complex we have
M apSp(A−Mod) (LF,x , A) ≃ M apSp(A−Mod) (A, TF,x ).

Now, let F be a stack, and u a morphism in Ho(Af fC∼,τ /F )
u /X
Y @
@@ ~~
@@ ~
y @@ ~~~y
 ~~
F,
with X = RSpec A and Y = RSpec B belonging to A. Let M ∈ B − M od0 , which
is also an A-module by the forgetful functor A − M od0 −→ B − M od0 . There is a
commutative diagram of commutative monoids
A⊕M / B⊕M

 
A / B,
inducing a commutative square of representable stacks
X[M ] o Y [M ]
O O

Xo Y.
This implies the existence of a natural morphism in Ho(SSetV )
DerF (Y, M ) −→ DerF (X, M ).
If the stack F has cotangent complexes at both points x and y, Prop. 1.2.11.3 induces
a well defined morphism in Ho(Sp(B − M od))
u∗ : LF,x ⊗LA B −→ LF,y .
Of course, we have (u ◦ v)∗ = v ∗ ◦ u∗ whenever this formula makes sense.
In the same way, the construction of LF,x is functorial in F . Let f : F −→ F ′
be a morphism of stacks, A ∈ A, and x : RSpec A −→ F be an A-point with image
f (x) : RSpec A −→ F ′ . Then, for any A-module M , there is a natural morphism
DerF (X, M ) −→ DerF ′ (X, M ).
Therefore, if F has a cotangent complex at x and F ′ has a cotangent complex at x′ ,
we get a natural morphism in Ho(Sp(A − M od))
dfx : LF ′ ,f (x) −→ LF,x ,
called the differential of f at x. Once again, we have d(f ◦ g)x = dgx ◦ dfx each time
this formula makes sense (this is the chain rule). Dually, we also get by duality the
derivative of f at x
T fx : TF,x −→ TF ′ ,f (x) .
Definition 1.4.1.7. A stack F has a global cotangent complex relative to the
HA context (C, C0 , A) (or simply has a cotangent complex when there is no ambiguity
on the context) if the following two conditions are satisfied.
 1.4.1. TANGENT STACKS AND COTANGENT COMPLEXES 101

(1) For any A ∈ A, and any point x : RSpec A −→ F , the stack F has a
cotangent complex LF,x at x.
(2) For any morphism u : A −→ B in A, and any morphism in Ho(Af fC∼,τ /F )
u
Y := RSpec B / X := RSpec A
LLL rr
LLL rrr
L
y LLL rr
r x
L% ry rr
F,
the induced morphism
u∗ : LF,x ⊗LA B −→ LF,y
is an isomorphism in Ho(Sp(B − M od)).
As a corollary of Prop. 1.2.1.2 and the standard properties of derivations any
representable stack has a cotangent complex.
Proposition 1.4.1.8. Any representable stack F = RSpecA has a global cotan-
gent complex.
Proof. This is nothing else than the existence of a universal derivation as proved
in Prop. 1.2.1.2. 
If X = RSpec A is a representable stack in A, and x : X → X is the iden-
tity, then the stable A-module LX,x is naturally isomorphic in Ho(Sp(A − M od))
to LA . More generally, for any morphism A −→ B with B ∈ A, corresponding to
y : RSpec B −→ X, the B-module LX,y is naturally isomorphic in Ho(Sp(B − M od))
to LA ⊗LA B.

The next proposition explains the relation between the tangent stack and the
global cotangent complex when it exists. It is a globalization of Prop.1.4.1.6.
Proposition 1.4.1.9. Let F be a stack having a cotangent complex. Let x : X =
RSpec A −→ F be any morphism, and
T Fx := T F ×hF X −→ X
the natural projection. Let A −→ B be a morphism with B ∈ A, corresponding to a
morphism of representable stacks X = RSpec A −→ Y = RSpec B. Then, there exists
a natural isomorphism in Ho(SSet)
RT Fx (B) ≃ M apAf fC∼,τ /F (Y, T F ) ≃ M apSp(A−Mod) (LF,x , B).
Proof. This is a reformulation of Prop. 1.4.1.6, and the fact that
M apSp(A−Mod) (LF,x , B) ≃ M apSp(B−Mod) (LF,x ⊗LA B, B).

Remark 1.4.1.10. Of course, the isomorphism of proposition Prop. 1.4.1.9 is
functorial in F .
Proposition 1.4.1.11. Let F be an n-geometric stack. We assume that for any
A ∈ A, and any point x : X := RSpec A −→ F , and any A-module M ∈ A − M od0 ,
the natural morphism
DerF (X, M ) ≃ DerF (X, ΩS(M )) −→ ΩDerF (X, S(M ))
is an isomorphism in Ho(SSet). Then F has a global cotangent complex, which is
furthermore (−n)-connective.
102 1.4. GEOMETRIC STACKS: INFINITESIMAL THEORY

Proof. The proof is by induction on n. For n = −1 this is Prop. 1.4.1.8 and

does not use our exactness condition on the functor DerF (X, −).
Let n ≥ 0 be an integer and F be an n-geometric stack. Let X = RSpec A be a
representable stack in A and x : X −→ F be any morphism in St(C, τ ). We consider
the natural morphisms
d : X −→ X ×h X dF : X −→ X ×hF X.
By induction on n, we see that the stacks X ×h X and X ×hF X both have cotangent
complexes at the point d and dF , denoted respectively by L and L′ . There is moreover
a natural morphism in Ho(Sp(A − M od))
f : L′ −→ L,
induced by
X / X ×h X / X ×h X.
F
We set L′′ as the homotopy cofiber of f in Sp(A − M od). By construction, for any
A-module M ∈ A − M od0 , the simplicial set M apSp(A−Mod) (L′′ , M ) is naturally
equivalent to the homotopy fiber of
DerX (X, M ) ×hDerF (X,M) DerX (X, M ) −→ DerX (X, M ) × DerX (X, M ),
and thus is naturally equivalent to Ω∗ DerF (X, M ). By assumption we have natural
isomorphisms in Ho(SSet)
M apSp(A−Mod) (Ω(L′′ ), M ) ≃ M apSp(A−Mod) (L′′ , S(M ))
≃ Ω∗ DerF (X, S(M )) ≃ DerF (X, Ω∗ S(M )) ≃ DerF (X, M ).
This implies that Ω∗ L′′ ∈ Ho(Sp(A − M od)) is a cotangent complex of F at the point
x. By induction on n and by construction we also see that this cotangent complex is
(−n)-connective.
Now, let
u
Y = RSpec B / X = RSpec A
KKK
KKK ss
KK sssss
y KKK s x
% ysss
F,
be a morphism in Ho(Af fC∼,τ /F ), with A and B in A. We consider the commutative
diagram with homotopy cartesian squares
/ X ×h X / X ×h X
< XO OF O
xxx
xx
xxx
xx
Y / Y ×h Y / Y ×h Y / Y ×h Y.
X F

By the above explicit construction and an induction on n, the fact that the natural
morphism
u∗ : LF,x ⊗LA B −→ LF,y
is an isomorphism simply follows from the next lemma.
Lemma 1.4.1.12. Let
x
X = RSpec A /F
O
/G
O
O
u

Y = RSpec B y
/ F′ / G′
 1.4.1. TANGENT STACKS AND COTANGENT COMPLEXES 103

be a commutative diagram with the right hand square being homotopy cartesian in
Af fC∼,τ . We assume that A and B are in A and that F and G have global cotangent
complexes. Then the natural square
LG′ ,y / LF ′ ,y
O O

LG,x ⊗LA B / LF,x ⊗L B

A

is homotopy cartesian in Sp(B − M od).

Proof. This is immediate from the definition and the homotopy cartesian square
DerF (Y, M ) ≃ DerF (X, M ) / DerG (Y, M ) ≃ DerG (X, M )
O O

DerF ′ (Y, M ) / DerG′ (Y, M )

for any A-module M . 

This finishes the proof of Prop. 1.4.1.11. 

In fact, the proof of Proposition 1.4.1.11 also proves the following

Proposition 1.4.1.13. Let F be a stack such that the diagonal F −→ F ×h F is
(n−1)-representable. We suppose that for an A ∈ A, any point x : X := RSpec A −→
F , and any A-module M ∈ A − M od0 the natural morphism
DerF (X, M ) ≃ DerF (X, ΩS(M )) −→ ΩDerF (X, S(M ))
is an isomorphism in Ho(SSet). Then F has a cotangent complex, which is further-
more (−n)-connective.
We finish this section by the notion of relative cotangent complex and its rela-
tion with the absolute notion. Let f : F −→ G be a morphism of stacks, A ∈ A,
and X := RSpec A −→ F be a morphism. We define an object DerF/G (X, −) ∈
Ho((A − M odop ∧
0 ) ), to be the standard homotopy fiber of the morphism of the natu-
ral morphism
df : DerF (X, −) −→ DerG (X, −).
In terms of functors the object DerF/G (X, −) sends an A-module M ∈ A − M od0 to
the simplicial set
DerF/G (X, M ) = M apX/Af fC∼,τ /G (X[M ], F ).
Definition 1.4.1.14. Let f : F −→ G be a morphism of stacks.
(1) Let A ∈ A, and x : X := RSpec A −→ F be an A-point. We say that f has a
(relative) cotangent complex at x relative to A (or simply f has a (relative)
cotangent complex at x when A is unambiguous) if there exists an integer n,
and an (−n)-connective stable A-module LF/G,x ∈ Ho(Sp(A − M od)), and
an isomorphism in Ho(A − M odop 0 )
∧

L
DerF/G (X, −) ≃ Rhs F/G,x .
(2) If f has a cotangent complex at x, the stable A-module LF,x is then called
the (relative) cotangent complex of f at x.
104 1.4. GEOMETRIC STACKS: INFINITESIMAL THEORY

(3) If f has a cotangent complex at x, the (relative) tangent complex of f at x

is then the stable A-module
TF/G,x := RHomSp
A (LF/G,x , A) ∈ Ho(Sp(A − M od)).

Let now be a morphism of stacks f : F −→ G, and a commutative diagram in

Af fC∼,τ
u /X
Y @
@@ ~~
@@ ~
y @@ ~~y
 ~~~
F,
with X = RSpec A and Y = RSpec B belonging to A. We have a natural morphism
in Ho((A − M odop ∧
0 ) )
DerF/G (Y, ) −→ DerF/G (X, −).
If the morphism f : F −→ G has cotangent complexes at both points x and y, Prop.
1.2.11.3 induces a well defined morphism in Ho(Sp(B − M od))
u∗ : LF/G,x ⊗LA B −→ LF/G,y .
Of course, we have (u ◦ v)∗ = v ∗ ◦ u∗ when this formula makes sense.
Definition 1.4.1.15. A morphism of stacks f : F −→ G has a relative cotangent
complex relative to (C, C0 , A) (or simply has a cotangent complex when the HA context
is clear) if the following two conditions are satisfied.
(1) For any A ∈ A, and any point x : RSpec A −→ F , the morphism f has a
cotangent complex LF/G,x at x.
(2) For any morphism u : A −→ B in A, and any morphism in Ho(Af fC∼,τ /F )
u
Y := RSpec B / X := RSpec A
LLL r
LLL
L rr rrr
y LLL rr x
L% ry rr
F,
the induced morphism
u∗ : LF/G,x ⊗LA B −→ LF/G,y
is an isomorphism in Ho(Sp(B − M od)).
The important remark is the following, relating absolute and relative notions of
cotangent complexes.
Lemma 1.4.1.16. Let f : F −→ G be a morphism of stacks.
(1) If both stacks F and G have cotangent complexes then the morphism f has
a cotangent complex. Furthermore, for any A ∈ A, and any morphism of
stacks X = RSpec A −→ F , there is a natural homotopy cofiber sequence of
stable A-modules
LG,x / LF,x / LF/G,x .

(2) If the morphism f has a cotangent complex then for any stack H and any
morphism H −→ G, the morphism F ×hG H −→ H has a relative cotangent
complex and furthermore we have
LF/G,x ≃ LF ×hG H/H,x
for any A ∈ A, and any morphism of stacks X = RSpec A −→ F ×hG H.
 1.4.2. OBSTRUCTION THEORY 105

(3) If for any A ∈ A and any morphism of stacks x : X := RSpec A −→ F ,

the morphism F ×hG X −→ X has a relative cotangent complex, then the
morphism f has a relative cotangent complex. Furthermore, we have

LF/G,x ≃ LF ×hG X/X,x .

(4) If for any A ∈ A and any morphism of stacks x : X := RSpec A −→ F ,

the stack F ×hG X has a cotangent complex, then the morphism f has a
relative cotangent complex. Furthermore we have a natural homotopy cofiber
sequence
LA −→ LF ×hG X,x −→ LF/G,x .

Proof. (1) and (2) follow easily from the definition. Point (3) follows from (2).
Finally, point (4) follows from (3), (1) and Prop. 1.4.1.8. 

1.4.2. Obstruction theory

Recall from 1.2.1.7 that for any commutative monoid A, any A-module M , and
any derivation d : A −→ A ⊕ M , we can form the square zero extension of A by ΩM ,
denoted by A ⊕d ΩM , as the homotopy cartesian square in Comm(C)
p
A ⊕d ΩM /A

d
 
A / A⊕M
s

where s : A −→ A ⊕ M is the trivial derivation. In the sequel, the morphism p :

A ⊕d ΩM −→ A will be called the natural projection.

Definition 1.4.2.1. (1) A stack F is infinitesimally cartesian relative to

the HA context (C, C0 , A) (or simply inf-cartesian when the HA context is
unambiguous) if for any commutative monoid A ∈ A, any M ∈ A − M od1 ,
and any derivation d ∈ π0 (Der(A, M )), corresponding to a morphism d :
A −→ A ⊕ M in Ho(Comm(C)/A), the square

RF (A ⊕d ΩM ) / RF (A)

d
 
RF (A) s
/ RF (A ⊕ M )

is homotopy cartesian.
(2) A stack F has an obstruction theory (relative to (C, C0 , A)) if it has a
(global) cotangent complex and if it is infinitesimally cartesian (relative to
(C, C0 , A)).

One also has a relative version.

Definition 1.4.2.2. (1) A morphism of stacks F −→ G is infinitesimally

cartesian relative to (C, C0 , A) (or simply inf-cartesian if the HA context is
clear) if for any commutative monoid A ∈ A, any A-module M ∈ A−M od1 ,
and any derivation d ∈ π0 (Der(A, M )), corresponding to a morphism d :
106 1.4. GEOMETRIC STACKS: INFINITESIMAL THEORY

A −→ A ⊕ M in Ho(Comm(C)/A), the square

RF (A ⊕d ΩM ) / RG(A ⊕d ΩM )

 
RF (A) ×hRF (A⊕M) RF (A) / RG(A) ×hRG(A⊕M) RG(A)

is homotopy cartesian.
(2) A morphism of stacks f : F −→ G has an obstruction theory relative to
(C, C0 , A) (or simply has an obstruction theory if the HA context is clear)
if it has a (global) cotangent complex and if it is infinitesimally cartesian
relative (C, C0 , A).
As our HA context (C, C0 , A) is fixed once for all we will from now avoid to men-
tion the expression relative to (C, C0 , A) when referring to the property of having an
obstruction theory. The more precise terminology will only be used when two differ-
ent HA contexts are involved (this will only happen in §2.3).

We have the following generalization of lemma 1.4.1.16.

Lemma 1.4.2.3. Let f : F −→ G be a morphism of stacks.
(1) If both stacks F and G have an obstruction theory then the morphism f has
an obstruction theory.
(2) If the morphism f has an obstruction theory then for any stack H and any
morphism H −→ G, the morphism F ×hG H −→ H has a relative obstruction
theory.
(3) If for any B ∈ Comm(C) and any morphism of stacks y : Y := RSpec B −→
G, the stack F ×hG B has an obstruction theory , then the morphism f has
a relative obstruction theory.
Proof. The existence of the cotangent complexes is done in Lem. 1.4.1.16, and
it only remains to deal with the inf-cartesian property. The points (1) and (2) are
clear by definition.
(3) Let A ∈ A, M ∈ A − M od1 and d ∈ π0 (Der(A, M )). We need to show that
the square
RF (A ⊕d ΩM ) / RG(A ⊕d ΩM )

 
RF (A) ×hRF (A⊕M) RF (A) / RG(A) ×h
RG(A⊕M) RG(A)

is homotopy cartesian. Let z be a point in RG(A ⊕d ΩM ). We need to prove that the

morphism induced on the homotopy fibers of the two horizontal morphisms taken at
z is an equivalence. But this easily follows from the fact that the pullback of f by
the morphism corresponding to z,
F ×hG Xd [ΩM ] −→ Xd [ΩM ],
has an obstruction theory. 

Proposition 1.4.2.4. (1) Any representable stack has an obstruction the-

ory.
(2) Any representable morphism has an obstruction theory.
 1.4.2. OBSTRUCTION THEORY 107

Proof. (1) By Prop. 1.4.1.8 we already know that representable stacks have
cotangent complexes. Using the Yoneda lemma, it is obvious to check that any rep-
resentable stack is inf-cartesian. Indeed, for F = RSpec B we have

RF (A ⊕d ΩM ) ≃ M apComm(C)(B, A ⊕d ΩM ) ≃

M apComm(C)(B, A)×hMapComm(C) (B,A⊕M) M apComm(C)(B, A) ≃ RF (A)×hRF (A⊕M) RF (A).

(2) Follows from (1) and Lem. 1.4.2.3 (3). 

In general, the expression has an obstruction theory relative to (C, C0 , A) is justi-

fied by the following proposition.

Proposition 1.4.2.5. Let F be a stack which has an obstruction theory. Let

A ∈ A, M ∈ A − M od1 and let d ∈ π0 (Der(A, M )) be a derivation with A ⊕d ΩM the
corresponding square zero extension. Let us denote by

X := RSpec A −→ Xd [ΩM ] := RSpec (A ⊕d ΩM )

the morphism of representable stacks corresponding to the natural projection A ⊕d

ΩM −→ A. Finally, let x : X −→ F be an A-point of F .
(1) There exists a natural obstruction

α(x) ∈ π0 (M apSp(A−Mod) (LF,x , M )) = [LF,x , M ]Sp(A−Mod)

vanishing if and only x extends to a morphism x′ in Ho(X/Af fC∼,τ )

X@ / Xd [ΩM ]
@@ v
@@ vv
x @@@ vvv′
v
 {vv x
F.
(2) Let us suppose that α(x) = 0. Then, the simplicial set of lifts of x

RHomX/Af fC∼,τ (Xd [ΩM ], F )

is (non canonically) isomorphic in Ho(SSet) to the simplicial set

M apSp(A−Mod) (LF,x , ΩM ) ≃ ΩM apSp(A−Mod) (LF,x , M ).

More precisely, it is a simplicial torsor over the simplicial group

ΩM apSp(A−Mod) (LF,x , M ).

Proof. First of all, the space of lifts x′ is by definition RHomX/Af fC∼,τ (Xd [ΩM ], F ),
which is naturally equivalent to the homotopy fiber at x of the natural morphism

Rτ Hom(Xd [ΩM ], F ) −→ Rτ Hom(X, F ).

Using that F is inf-cartesian, we see that there exists a homotopy cartesian square

Rτ Hom(Xd [ΩM ], F ) / Rτ Hom(X, F )

d
 
Rτ Hom(X, F ) s
/ Rτ Hom(X[M ], F ).
108 1.4. GEOMETRIC STACKS: INFINITESIMAL THEORY

Therefore, the simplicial set RHomX/Af fC∼,τ (Xd [ΩM ], F ) fits into a homotopy carte-
sian square

RHomX/Af fC∼,τ (Xd [ΩM ], F ) /•

 
• / RHomX/Af f ∼,τ (X[M ], F ).
0 C

As F has a cotangent complex we have

RHomX/Af fC∼,τ (X[M ], F ) ≃ M apSp(A−Mod) (LF,x , M ),

and we see that the image of the right vertical arrow in the last diagram provides
the element α(x) ∈ π0 (M apSp(A−Mod) (LF,x , M )), which clearly vanishes if and only if
RHomX/Af fC∼,τ (Xd [ΩM ], F ) is non-empty. Furthermore, this last homotopy cartesian
diagram also shows that if α(x) = 0, then one has an isomorphism in Ho(SSet)
RHomX/Af fC∼,τ (Xd [ΩM ], F ) ≃ ΩM apSp(A−Mod) (LF,x , M ).


One checks immediately that the obstruction of Prop. 1.4.2.5 is functorial in F

and X in the following sense. If f : F −→ F ′ be a morphism of stacks, then clearly
df (α(x)) = α(f (x)) ∈ π0 (DerF ′ (X, M )).
In the same way, if A −→ B is a morphism in A, corresponding to a morphism of
representable stacks Y −→ X, and y : Y −→ F be the composition, then we have
y ∗ (α(x)) = αy ∈ π0 (DerF (Y, M ⊗LA B)).
Proposition 1.4.2.5 also has a relative form, whose proof is essentially the same.
We will also express it in a more precise way.
Proposition 1.4.2.6. Let f : F −→ G be a morphism of stacks which has an
obstruction theory. Let A ∈ A, M ∈ A − M od1 , d ∈ π0 (Der(A, M )) a deriva-
tion and A ⊕d ΩM the corresponding square zero extension. Let x be a point in
RF (A) ×hRG(A⊕d ΩM) RG(A) with projection y ∈ RF (A), and let L(x) be the homotopy
fiber, taken at x, of the morphism
RF (A ⊕d ΩM ) −→ RF (A) ×hRG(A⊕d ΩM) RG(A).

Then there exists a natural point α(x) in M apA−Mod (LF/G,x , M ), and a natural iso-
morphism in Ho(SSet)
L(z) ≃ Ωα(x),0 M apA−Mod (LF/G,y , M ),
where Ωα(x),0 M apA−Mod (LF/G,y , M ) is the simplicial set of paths from α(x) to 0.
Proof. Essentially the same as for Prop. 1.4.2.5. The point x corresponds to a
commutative diagram in Ho(Af fC∼,τ /G)

X /F ,

 
Xd [ΩM ] /G
 1.4.3. ARTIN CONDITIONS 109

where X := RSpec A and Xd [ΩM ] := RSpec (A⊕d ΩM ). Composing with the natural
commutative diagram
d /X
X[M ]
s
 
X / Xd [ΩM ]

we get a well defined commutative diagram in Ho(Af fC∼,τ /G)

d /X /F
X[M ]
s
 
X / G,

giving rise to a well defined point

α(x) ∈ RHomX/Af fC∼,τ /G (X[M ], F ) = DerF/G (X, M ).
Using that the morphism f is inf-cartesian, we easily see that the simplicial set
Ωα(x),0 DerF/G (X, M ) is naturally equivalent to the space of lifts
L(x) = RHomX/Af fC∼,τ /G (Xd [ΩM ], F ) ≃ Ωα(x),0 DerF/G (X, M ).

Proposition 1.4.2.7. Let F be a stack whose diagonal F −→ F ×h F is n-
representable for some n. Then F has an obstruction theory if and only if it is
inf-cartesian.
Proof. It is enough to show that a stack F that is inf-cartesian satisfies the
condition of proposition 1.4.1.11. But this follows easily from the following homotopy
cartesian square
A⊕M /A

s
 
A / A ⊕ S(M )
s

for any commutative monoid A ∈ A and any A-module M ∈ A − M od1 . 

1.4.3. Artin conditions

In this section we will give conditions on the topology τ and on the class of
morphisms P ensuring that any stack which is geometric for such a τ and P will
have an obstruction theory. We call these conditions Artin’s conditions, though we
warn the reader that these are not the rather famous conditions for a functor to be
representable by an algebraic space. We refer instead to the fact that an algebraic
stack in the sense of Artin (i.e. with a smooth atlas) has a good infinitesimal and
obstruction theory. We think that this has been first noticed by M. Artin, since this
is precisely one part of the easy direction of his representability criterion ([Ar, I, 1.6]
or [La-Mo, Thm. 10.10]).
Definition 1.4.3.1. We will say that τ and P satisfy Artin’s conditions relative
to (C, C0 , A) (or simply satisfy Artin’s conditions if the HA context is clear) if there
exists a class E of morphisms in Af fC such that the following conditions are satisfied.
(1) Any morphism in P is formally i-smooth in the sense of Def. 1.2.8.1.
110 1.4. GEOMETRIC STACKS: INFINITESIMAL THEORY

(2) Morphisms in E are formally étale, are stable by equivalences, homotopy

pullbacks and composition.
(3) For any morphism A −→ B in E with A ∈ A, we have B ∈ A.
(4) For any epimorphism of representable stacks Y −→ X, which is a P-
morphism, there exists an epimorphism of representable stacks X ′ −→ X,
which is in E, and a commutative diagram in St(C, τ )

=Y
|||
|
||
|| 
X′ / X.

(5) Let A ∈ A, M ∈ A − M od1 , d ∈ π0 (Der(A, M )) be a derivation, and

X := RSpec A −→ Xd [ΩM ] := RSpec, (A ⊕d ΩM )
be the natural morphism in St(C, τ ). Then, a formally étale morphism of
representable stacks p : U −→ Xd [ΩM ] is in E if and only if U ×hXd [ΩM]
X −→ X is so. Furthermore, if p is in E, then p is an epimorphism of
stacks if and only if U ×hXd [ΩM] X −→ X is so.
The above definition might seem a bit technical and somehow hard to follow. In
order to fix his intuition, we suggest the reader to think in terms of standard algebraic
geometry with τ being the etale topology, P the class of smooth morphisms, and E
the class of étale morphisms. This is only meant to convey some classical geometric
intuition because this classical situation in algebraic geometry does not really fit into
the above definition; in fact in this case the base category is the category of k-modules
and thus the suspension functor is not fully faithful.

In order to simplify notations we will say that a morphism A −→ B in Comm(C)

is an E-covering if it is in E and if the corresponding morphism of stacks
RSpec B −→ RSpec A
is an epimorphism of stacks.
Theorem 1.4.3.2. Assume τ and P satisfy Artin’s conditions.
(1) Any n-representable morphism of stacks has an obstruction theory.
(2) Let f : F −→ G be an n-representable morphism of stacks. If f is in P then
for any A ∈ A, and any morphism x : X := RSpec A −→ F there exists an
E-covering
x′ : X ′ := RSpec A′ −→ X
such that for any M ∈ A′ − M od1 the natural morphism
[LX ′ /G,x′ , M ] −→ [LF/G,x , M ]A−Mod
is zero.
Proof. Before starting the proof we will need the following general fact, that
will be used all along the proof of the theorem.
Lemma 1.4.3.3. Let D be a pointed model category for which the suspension func-
tor
S : Ho(D) −→ Ho(D)
is fully faithful. Then, a homotopy co-cartesian square in D is also homotopy carte-
sian.
 1.4.3. ARTIN CONDITIONS 111

Proof. When D is a stable model category this is well known since homotopy
fiber sequences are also homotopy cofiber sequences (see [Ho1]). The general case is
proved in the same way. When D is furthermore U-cellular (which will be our case),
one can even deduce the result from the stable case by using the left Quillen functor
D −→ Sp(D), from D to the model category of spectra in D as defined in [Ho2], and
using the fact that it is homotopically fully faithful. 

The previous lemma can be applied to C, but also to the model categories B−M od
of modules over some commutative monoid B. In particular, homotopy cartesian
square of B-modules which are also homotopy co-cartesian will remain homotopy
cartesian after a derived tensor product by any B-module.

Let us now start the proof of theorem 1.4.3.2.

Some topological invariance statements

We start with several results concerning topological invariance of formally étale

morphisms and morphisms in E.
Lemma 1.4.3.4. Let A be a commutative monoid, M an A-module and A⊕M −→
A the natural augmentation. Then, the homotopy push out functor
A ⊗LA⊕M − : Ho((A ⊕ M ) − Comm(C)) −→ Ho(A − Comm(C))
induces an equivalence between the full sub-categories consisting of formally étale com-
mutative A ⊕ M -algebras and formally étale commutative A-algebras.
Proof. We see that the functor is essentially surjective as the morphism A ⊕
M −→ A possesses a section A −→ A⊕M . Next, we show that the functor A⊗LA⊕M −
is conservative. For this, let us consider the commutative square
A⊕M /A

 
A / A ⊕ S(M ),

which, as a commutative square in C, is homotopy cocartesian and homotopy carte-

sian. Therefore, this square is homotopy cocartesian in (A ⊕ M ) − M od, and thus
lemma 1.4.3.3 implies that for any commutative A ⊕ M -algebra B, the natural mor-
phism


 
B −→ A ⊗LA⊕M B ×h(A⊕S(M))⊗L B A ⊗LA⊕M B ≃ A ×hA⊕S(M) A ⊗LA⊕M B
A⊕M

is then an isomorphism in Ho(Comm(C)). This clearly implies that the functor

A ⊗LA⊕M − is conservative.
Now, let A ⊕ M −→ B be a formally étale morphism of commutative monoids.
The diagonal morphism M −→ M ×h M ≃ M ⊕ M in Ho(A − M od), induces a well
defined morphism in Ho(Comm(C)/A ⊕ M )

(A ⊕ M ) ⊕ M
7
p pppp
ppp
ppp 
A⊕M / A ⊕ M,
Id
112 1.4. GEOMETRIC STACKS: INFINITESIMAL THEORY

and therefore a natural element in π0 Der(A ⊕ M, M ). Composing with M −→ M ⊗LA

B, we get a well defined element in π0 Der(A ⊕ M, M ⊗LA B). Using that A ⊕ M −→ B
is formally étale, this element extends uniquely to an element in π0 Der(B, M ⊗LA B).
This last derivation gives rise to a well defined morphism in Ho((A ⊕ M ) − Comm(C))


u : B −→ A ⊗LA⊕M B ⊕ M ⊗LA B.

Furthermore, by construction, this morphism is sent to the identity of A ⊗LA⊕M B by

the functor A ⊗LA⊕M −, and as we have seen this implies that u is an isomorphism in
Ho((A ⊕ M ) − Comm(C)).
We now finish the proof of the lemma by showing that the functor A ⊗LA⊕M − is
fully faithful. For this, let A ⊕ M −→ B and A ⊕ M −→ B ′ be two formally étale
morphisms of commutative monoids. As we have seen, B ′ can be written as
B ′ ≃ (A ⊕ M ) ⊗LA A′ ≃ A′ ⊕ M ′
where A −→ A′ is formally étale, and M ′ := M ⊗LA A′ . We consider the natural
morphism
M ap(A⊕M)−Comm(C)(B, B ′ ) ≃ M ap(A⊕M)−Comm(C) (B, A′ ⊕ M ′ ) −→

M apA−Comm(C) (A ⊗LA⊕M B, A ⊗LA⊕M B ′ ) ≃ M apA−Comm(C)(A ⊗LA⊕M B, A′ ).

The homotopy fiber of this morphism at a point B → A′ is identified with DerA⊕M (B, M ′ ),
which is contractible as B is formally étale over A⊕M . This shows that the morphism
M ap(A⊕M)−Comm(C)(B, B ′ ) −→ M apA−Comm(C) (A ⊗LA⊕M B, A ⊗LA⊕M B ′ )
has contractible homotopy fibers and therefore is an isomorphism in Ho(SSet), and
finishes the proof of the lemma. 

Lemma 1.4.3.5. Let A be a commutative monoid and M an A-module. Then the

homotopy push out functor
A ⊗LA⊕M − : Ho((A ⊕ M ) − Comm(C)) −→ Ho(A − Comm(C))
induces an equivalence between the full sub-categories consisting of E-coverings of
A ⊕ M and of E-coverings of A.
Proof. Using lemma 1.4.3.4 it is enough to show that a formally étale morphism
f : A ⊕ M −→ B is in E (respectively an E-covering) if and only if A −→ A′ :=
A ⊗LA⊕M B is in E (respectively an E-covering). But, as f is formally étale, lemma
1.4.3.4 implies that it can written as
A ⊕ M −→ A′ ⊕ M ′ ≃ (A ⊕ M ) ⊗LA A′ ,
with M ′ := M ⊗LA A′ . Therefore the lemma simply follows from the stability of
epimorphisms and morphisms in E by homotopy pullbacks. 

Lemma 1.4.3.6. Let A ∈ Comm(C), M ∈ A − M od1 , and d ∈ π0 (Der(A, M ))

be a derivation. Let B := A ⊕d ΩM −→ A be the natural augmentation, and let us
consider the base change functor
A ⊗LB − : Ho(B − Comm(C)) −→ Ho(A − Comm(C)).
Then, A ⊗LB − restricted to the full subcategory consisting of formally étale commu-
tative B-algebras is fully faithful.
 1.4.3. ARTIN CONDITIONS 113

Proof. Let
B /A

d
 
A / A⊕M
s

be the standard homotopy cartesian square of commutative monoids, which is also

homotopy co-cartesian in C as M ∈ A − M od1 . We represent it as a fibered square in
Comm(C)
B /A

d
 
A′ / A ⊕ M,
s′

where s′ : A′ −→ A ⊕ M is fibrant replacement of the trivial section s : A −→ A ⊕ M .

We define a model category D whose objects are 5-plets (B1 , B2 , B3 , a, b), where
B1 ∈ A − Comm(C), B2 ∈ A′ − Comm(C), B3 ∈ (A ⊕ M ) − Comm(C), and a and b
are morphisms of commutative (A ⊕ M )-algebras
a b
(A ⊕ M ) ⊗A B1 / B3 o (A ⊕ M ) ⊗A′ B2

(where the co-base change on the left is taken with respect of the morphism s′ : A′ →
A ⊕ M and the one on the right with respect to d : A → A ⊕ M ). For an object
(B1 , B2 , B3 , a, b) in D, the morphisms a and b can also be understood as B1 −→ B3
in A − Comm(C) and B2 −→ B3 in A′ − Comm(C). The morphisms
(B1 , B2 , B3 , a, b) −→ (B1′ , B2′ , B3′ , a′ , b′ )
in D are defined in the obvious way, as families of morphisms {Bi → Bi′ } commuting
with the a’s and b’s. A morphism in D is defined to be an equivalence or a cofibration
if each morphism Bi → Bi′ is so. A morphism in D is defined to be a fibration if each
morphism Bi → Bi′ is a fibration in C, and if the natural morphisms
B1 −→ B1′ ×B3′ B3 B2 −→ B2′ ×B3′ B3
are fibrations in C. This defines a model category structure on D which is a Reedy type
model structure. An important fact concerning D is the description of its mapping
spaces as the following homotopy cartesian square

M apD ((B, a, b), (B ′ , a′ , b′ )) / M apA−Comm(C)(B1 , B1′ ) × M apA′ −Comm(C) (B2 , B2′ )

 
M ap(A⊕M)−Comm(C)(B3 , B3′ ) / M apA−Comm(C)(B1 , B3′ ) × M apA′ −Comm(C) (B2 , B3′ )

where we have denoted B := (B1 , B2 , B3 ) and B ′ := (B1′ , B2′ , B3′ ). There exists a
natural functor
F : B − Comm(C) −→ D
sending a commutative B-algebra B ′ to the object
F (B ′ ) := (A ⊗B B ′ , A′ ⊗B B ′ , (A ⊕ M ) ⊗B B ′ , a, b)
where
a : (A⊕M )⊗A (A⊗B B ′ ) ≃ (A⊕M )⊗B B ′ b : (A⊕M )⊗A′ (A′ ⊗B B ′ ) ≃ (A⊕M )⊗B B ′
114 1.4. GEOMETRIC STACKS: INFINITESIMAL THEORY

are the two natural isomorphisms in (A ⊕ M ) − Comm(C). The functor F has a right
adjoint G, sending an object (B1 , B2 , B3 , a, b) to the pullback in B − Comm(C)
G(B1 , B2 , B3 , a, b) / B1

a
 
B2 / B3 .
b

Clearly the adjunction (F, G) is a Quillen adjunction. Furthermore, lemma 1.4.3.3

implies that for any commutative B-algebra B → B ′ the adjunction morphism
B ′ −→ RGLF (B ′ ) = A ⊗LB B ′ ×h(A⊕M)⊗L B ′ A′ ⊗LB B ′ ≃ (A ×hA⊕M A′ ) ⊗LB B ′
B

is an isomorphism in Ho(B − Comm(C)). This implies in particular that

LF : Ho(B − Comm(C)) −→ Ho(D)
is fully faithful.
We now consider the functor
D −→ A − Comm(C)
sending (B1 , B2 , B3 , a, b) to B1 . Using our lemma 1.4.3.4, and the description of the
mapping spaces in D, it is not hard to see that the induced functor
Ho(D) −→ Ho(A − Comm(C))
becomes fully faithful when restricted to the full subcategory of Ho(D) consisting of
objects (B1 , B2 , B3 , a, b) such that A → B1 and A′ → B2 are formally étale and the
induced morphism
a : (A ⊕ M ) ⊗LA B1 −→ B3 b : (A ⊕ M ) ⊗LA′ B2 −→ B3
are isomorphisms in Ho((A ⊕ M ) − Comm(C)). Putting all of this together we deduce
that the functor
A ⊗LB − : Ho(B − Comm(C)) −→ Ho(A − Comm(C))
is fully faithful when restricted to the full subcategory of formally étale morphisms.


Lemma 1.4.3.7. Let A be a commutative monoid, M an A-module and d : Der(A, M )

be a derivation. Let B := A ⊕d ΩM −→ A be the natural augmentation. Then, there
exists a natural homotopy cofiber sequence of A-modules
/ LA d / LQZ(M ),
LB ⊗LB A
where
Q : A − Commnu (C) −→ A − M od A − Commnu (C) ←− A − M od : Z
is the Quillen adjunction described during the proof of 1.2.1.2.
Proof. When d is the trivial derivation, we know the lemma is correct as by
Prop. 1.2.1.6 (4) we have
a
LA⊕M ⊗LA⊕M A ≃ LA LQZ(M ).
For the general situation, we use our left Quillen functor
F : B − Comm(C) −→ D
 1.4.3. ARTIN CONDITIONS 115

defined during the proof of lemma 1.4.3.6. The commutative monoid A is naturally
an (A ⊕ M )-algebra, and can be considered as a natural object (A, A, A) in D with
the obvious transition morphisms
A ⊗A (A ⊕ M ) ≃ A ⊕ M → A,

A ⊗A′ (A ⊕ M ) → A ⊗A (A ⊕ M ) ≃ A ⊕ M → A.
For any A-module N , one can consider A ⊕ N as a commutative A-algebra, and
therefore as as an object (A ⊕ N, A ⊕ N, A ⊕ N ) in D (with the obvious transition
morphisms). We will simply denote by A the object (A, A, A) ∈ D, and by A ⊕ N
the object (A ⊕ N, A ⊕ N, A ⊕ N ) ∈ D. The left Quillen property of F implies that
Der(B, N ) ≃ M apD/A (F (B), A ⊕ N ).
This shows that the morphism
Der(A, N ) −→ Der(B, N )
is equivalent to the morphism
Der(A, N ) −→ Der(A, N ) ×hDer(A⊕M,N ) Der(A, N ).
This implies the the morphism
LB ⊗LB A −→ LA
is naturally equivalent to the morphism of A-modules
L
a
LA LA −→ LA .
LA⊕M ⊗LA⊕M A

Using the already known result for the trivial extension A ⊕ M we get the required
natural cofiber sequence
LB ⊗LB A −→ LA −→ LQZ(M ).


Lemma 1.4.3.8. Let A be a commutative monoid, M ∈ A − M od1 and d ∈

π0 (Der(A, M )) be a derivation. Let B := A ⊕d ΩM −→ A be the natural augmenta-
tion, and let us consider the base change functor
A ⊗LB − : Ho(B − Comm(C)) −→ Ho(A − Comm(C)).
Then, A ⊗LB − induces an equivalence between the full sub-categories consisting of
formally étale commutative B-algebras and of formally étale commutative A-algebras.
Proof. By lemma 1.4.3.6 we already know that the functor is fully faithful, and
it only remains to show that any formally étale A-algebra A → A′ is of the form
A ⊗LB B ′ for some formally étale morphism B → B ′ .
Let A −→ A′ be a formally étale morphism. The derivation d ∈ π0 Der(A, M )
lifts uniquely to a derivation d′ ∈ π0 Der(A′ , M ′ ) where M ′ := M ⊗LA A′ . We form the
corresponding square zero extension
B ′ := A′ ⊕d ΩM ′ / A′

d′
 
A′ / A′ ⊕ M ′
116 1.4. GEOMETRIC STACKS: INFINITESIMAL THEORY

which comes equipped with a natural morphism B −→ B ′ fitting in a homotopy

commutative square
BO ′ / A′
O

B / A.
We claim that B −→ B is formally étale and that the natural morphism A ⊗LB B ′ −→
′

A′ is an isomorphism in Ho(A − Comm(C)).

There are natural cofiber sequences in C
B ′ −→ A′ −→ M ′
B −→ A −→ M,
as well as a natural morphism of cofiber sequences in C
A ⊗LB B ′ / A ⊗L A′
B
/ A ⊗L M ′
B

  
B ⊗LB A′ ≃ A′ / A ⊗L A′
B
/ M ⊗L A′ ,
B

which by our lemma 1.4.3.3 is also a morphism of fiber sequences. The two right
vertical morphisms are equivalences and thus so is the arrow on the left. This shows
that A ⊗LB B ′ ≃ A′ .
Finally, lemma 1.4.3.7 implies the existence of a natural morphism of cofiber
sequences of A′ -modules
LB ⊗LB A / LA / LQZ(M )

  
LB ′ ⊗LB ′ A′ / LA′ / LQZ(M ′ ) ≃ LQZ(M ) ⊗L A′ .
A

As A → A′ is étale, this implies that the natural morphism

(LB ⊗LB B ′ ) ⊗LB ′ A′ ≃ (LB ⊗LB A) ⊗LA A′ −→ LB ′ ⊗LB ′ A′
is an isomorphism in Ho(A′ − M od). This would show that B −→ B ′ is formally étale
if one knew that the base change functor
A′ ⊗LB ′ − : Ho(B ′ − M od) −→ Ho(A′ − M od)
were conservative. However, this is the case as lemma 1.4.3.3 implies that for any
B ′ -module N we have a natural isomorphism in Ho(B ′ − M od)
N ≃ (A′ ⊗LB ′ N ) ×h(A′ ⊕M ′ )⊗L N (A′ ⊗LB ′ N ).
B′


Lemma 1.4.3.9. Let A ∈ A, M ∈ A − M od1 and d ∈ π0 (Der(A, M )) be a deriva-
tion. Let
A ⊕d ΩM −→ A
be the natural morphism and let us consider the homotopy push-out functor
(A ⊕d ΩM ) ⊗LA − : Ho((A ⊕d ΩM ) − Comm(C)) −→ Ho(A − Comm(C)).
Then, (A ⊕d ΩM ) ⊗LA − induces an equivalence between the full sub-categories con-
sisting of E-covers of A ⊕d ΩM and of E-covers of A.
 1.4.3. ARTIN CONDITIONS 117

Proof. This is immediate from Lem. 1.4.3.8 and condition (4) of Artin’s condi-
tions 1.4.3.1. 

Proof of Theorem 1.4.3.2

We are now ready to prove that F has an obstruction theory. To simplify nota-
tions we assume that F is fibrant in Af fC∼,τ , so F (A) ≃ RF (A) for any A ∈ Comm(C).
We then argue by induction on the integer n. For n = −1 Theorem 1.4.3.2 follows
from Prop. 1.4.2.4, hypothesis Def. 1.4.3.1 (5) and Prop. 1.2.8.3. We now assume
that n ≥ 0 and that both statement of theorem 1.4.3.2 are true for all m < n.

We start by proving Thm. 1.4.3.2 (1) for rank n. For this, we use Lem. 1.4.2.3
and Prop. 1.4.2.7, which show that we only need to prove that any n-geometric stack
is inf-cartesian.

Lemma 1.4.3.10. Let F be an n-geometric stack. Then F is inf-cartesian.

Proof. Let A ∈ A, M ∈ A − M od1 , and d ∈ π0 (Der(A, M )). Let x be a point

in π0 (F (A) ×hF (A⊕M) F (A)), whith projection x1 ∈ π0 (F (A)) on the first factor. We
need to show that the homotopy fiber, taken at x, of the morphism

F (A ⊕d ΩM ) −→ F (A) ×hF (A⊕M) F (A)

is contractible. For this, we replace the homotopy cartesian diagram

A ⊕d ΩM /A

 
A / A⊕M

be an equivalent commutative diagram in Comm(C)

B / B1

 
B2 / B3 ,

in such a way that each morphism is a cofibration in Comm(C). The point x can be
represented as a point in the standard homotopy pullback F (B1 ) ×hF (B3 ) F (B2 ). We
then define a functor

S : B − Comm(C) = (Af fC /Spec B)op −→ SSetV ,

in the following way. For any morphism of commutative monoids B −→ B ′ , the

simplicial set S(B ′ ) is defined to be the standard homotopy fiber, taken at x, of the
natural morphism

F (B ′ ) −→ F (B1 ⊗B B ′ ) ×hF (B3 ⊗B B ′ ) F (B2 ⊗B B ′ ).

Because of our choices on the Bi ’s, it is clear that the simplicial presheaf S is a
stack on the comma model site Af fC /Spec B. Therefore, in order to show that S(B)
is contractible it is enough to show that S(B ′ ) is contractible for some morphism
B −→ B ′ such that RSpec B ′ −→ RSpec B is an epimorphism of stacks. In particular,
we are allowed to homotopy base change by some E-covering of B. Also, using our
118 1.4. GEOMETRIC STACKS: INFINITESIMAL THEORY

lemma 1.4.3.8 (or rather its proof), we see that for any E-covering B → B ′ , the
homotopy cartesian square

B′ / B1 ⊗B B ′

 
B2 ⊗B B ′ / B3 ⊗B B ′ ,

is in fact equivalent to some

A′ ⊕d′ ΩM ′ / A′

d′
 
A′ / A′ ⊕ M ′ ,

for some E-covering A −→ A′ (and with M ′ ≃ M ⊗LA A′ , and where d′ is the unique
derivation d′ ∈ π0 Der(A′ , M ′ ) extending d). This shows that we can always replace
A by A′ , d by d′ and M by M ′ . In particular, for an (n − 1)-atlas {Ui −→ F } F ,
we can assume that the point x1 ∈ π0 (F (A)), image of the point x, lifts to a point in
y1 ∈ π0 (Uj (A)) for some j. We will denote U := Uj .
Sub-lemma 1.4.3.11. The point x ∈ π0 (F (A) ×hF (A⊕M) F (A)) lifts to point y ∈
π0 (U (A) ×hU(A⊕M) U (A))
Proof. We consider the commutative diagram of simplicial sets
f
U (A) ×hU(A⊕M) U (A) / F (A) ×hF (A⊕M) F (A)

p q
 
U (A) / F (A)

induced by the natural projection A ⊕ M → A. Let F (p) and F (q) be the homotopy
fibers of the morphisms p and q taken at y1 and x1 . We have a natural morphism
g : F (p) −→ F (q). Moreover, the homotopy fiber of the morphism f , taken at the
point x, receives a natural morphism from the homotopy fiber of the morphism g.
It is therefore enough to show that the homotopy fiber of g is not empty. But, by
definition of derivations, the morphism g is equivalent to the morphism
Ωd,0 DerU (X, M ) −→ Ωd,0 DerF (X, M ),
where the derivation d is given by the image of the point y1 by d : A → A ⊕ M .
Therefore, the homotopy fiber of the morphism g is equivalent to
Ωd,0 DerU/F (X, M ) ≃ Ωd,0 M ap(LU/F,y1 , M ).
But, using Thm. 1.4.3.2 (2) at rank (n − 1) and for the morphism U −→ F we obtain
that Ωd,0 DerU/F (X, M ) is non empty. This finishes the proof of the sub-lemma. 

We now consider the commutative diagram

a / U (A) ×h
U (A ⊕d ΩM ) U(A⊕M) U (A)

b′ b
 
F (A ⊕d ΩM ) / F (A) ×hF (A⊕M) F (A).
a′
 1.4.3. ARTIN CONDITIONS 119

The morphism a is an equivalence because U is representable. Furthermore, by our

inductive assumption the above square is homotopy cartesian. This implies that the
homotopy fiber of a′ at x is either contractible or empty. But, by the above sub-
lemma the point x lifts, up to homotopy, to a point in U (A) ×hU(A⊕M) U (A), showing
that this homotopy fiber is non empty. 
We have finished the proof of lemma 1.4.3.10 which implies that any n-geometric
stack is inf-cartesian, and thus that any n-representable morphism has an obstruction
theory. It only remain to show part (2) of Thm. 1.4.3.2 at rank n. For this, we
use Lem. 1.4.1.16 (3) which implies that we can assume that G = ∗. Let U −→ F
be an n-atlas, A ∈ A and x : X := RSpec A −→ F be a point. By passing to an
epimorphism of representable stacks X ′ −→ X which is in E, we can suppose that the
point x factors through a point u : X −→ U , where U is representable and U −→ F is
in P. By composition and the hypothesis that morphisms in P are formally i-smooth,
we see that U −→ ∗ is a formally i-smooth morphism. We then have a diagram
LF,x −→ LU,u −→ LX,x ,
which obviously implies that for any M ∈ A − M od1 the natural morphism
[LX,x , M ] −→ [LF,x , M ]
factors through the morphism
[LX,x , M ] −→ [LU,u , M ]
which is itself equal to zero by Prop. 1.2.8.3. 
We also extract from Lem. 1.4.3.8 and its proof the following important corollary.
Corollary 1.4.3.12. Let A ∈ Comm(C), M ∈ A − M od and d ∈ π0 (Der(A, M ))
be a derivation. Assume that the square
B = A ⊕d ΩM /A

 
A / A⊕M
is homotopy co-cartesian in C, then the base change functor
A ⊗LB − : Ho(B − Comm(C)) −→ Ho(A − Comm(C))
induces an equivalence between the full sub-categories of formally étale commutative
B-algebras and formally étale commutative A-algebras. The same statement holds
with formally étale replaced by étale.
Proof. Only the assertion with formally étale replaced by étale requires an
argument. For this, we only need to prove that if a formally étale morphism f :
B −→ B ′ is such that A −→ A ⊗LB B ′ = A′ is finitely presented, then so is f . For this
we use the fully faithful functor
LF : Ho(B − Comm(C)) −→ Ho(D)
defined during the proof of Lem. 1.4.3.6. Using the description of mappping spaces in
D in terms of a certain homotopy pullbacks, and using the fact that filtered homotopy
colimits in SSet commutes with homotopy pullbacks, we deduce the statement. 
 Part 2

Applications
 Introduction to Part 2

In this second part we apply the theory developed in the first part to study the
geometry of stacks in various HAG contexts (Def. 1.3.2.13). In particular we will
specialize our base symmetric monoidal model category C to the following cases:
• C = Z − M od, the category of Z-modules to get a theory of geometric stacks
in (classical) Algebraic Geometry (§2.1);
• C = sM odk , the category of simplicial modules over an arbitrary base com-
mutative ring k to get a theory of derived or D− -geometric stacks (§2.2);
• C = C(k), the category of unbounded cochain complexes of modules over
a characteristic zero base commutative ring k to get a theory of geomet-
ric stacks in complicial algebraic geometry, also called geometric D-stacks
(§2.3);
• C = SpΣ , the category of symmetric spectra ([HSS, Shi]) to get a theory
of geometric stacks in brave new algebraic geometry (§2.4).
In §2.1 we are concerned with classical algebraic geometry, the base category
C = Z − M od being endowed with the trivial model structure. We verify that if
k−Af f := Comm(C)op = (k−Alg)op is endowed with its étale Grothendieck topology
and P is the class of smooth morphisms between (usual) commutative rings then

(C, C0 , A, τ, P) := (k − M od, k − M od, k − Alg, ét, P)

is a HAG-context according to Def. 1.3.2.13. n-geometric stacks in this context will

be called Artin n-stacks, and essentially coincide with geometric n-stacks as defined in
[S3]; their model category will be denoted by k − Af f ∼,ét and its homotopy category
by St(k).
After having established a coherent dictionary (Def. 2.1.1.4), we show in §2.1.2
how the theory of schemes, of algebraic spaces, and of Artin’s algebraic stacks in
groupoids ([La-Mo]) embeds in our theory of geometric stacks (Prop. 2.1.2.1). We
also remark (Rmk. 2.1.2.2) that the general infinitesimal theory for geometric stacks
developed in Part I does not apply to this context. The reason for this is that the cat-
egory C = k −M od (with its trivial model structure) is as unstable as it could possibly
be: the suspension functor S : Ho(C) = k − M od → k − M od = Ho(C) is trivial (i.e.
sends each k-module to the zero k-module). The explanation for this is the following.
Usual infinitesimal theory that applies to schemes, algebraic spaces or to (some classes
of) Artin’s algebraic stacks in groupoids is in fact (as made clear e.g. by the definition
of cotangent complex of a scheme ([Ill, II.2]) which uses simplicial resolutions of ob-
jects Comm(C)), already conceptually part of derived algebraic geometry in the sense
that its classical definition already requires to embed Comm(C) = k − Alg into the
category of simplicial k-algebras. And in fact (as we will show in §2.2) when schemes,
algebraic spaces and Artin algebraic stacks in groupoids are viewed as derived stacks,
then their classical infinitesimal theory can be recovered (and generalized, see Cor.
2.2.4.5) and interpreted geometrically within our general formalism of Chapter 1.4 (in
123
124 INTRODUCTION TO PART 2

particular in Prop. 1.4.1.6).

Therefore we are naturally brought to §2.2 where we treat the case of derived
algebraic geometry, i.e. the case where C := sk − M od, the category of simplicial
modules over an arbitrary commutative ring k.
In §2.2.1 we describe the model categories C = sk−M od and Comm(C) = sk−Alg
whose opposite is denoted by k−D− Af f , in particular finite cell and finitely presented
objects, suspension and loop functors, Postnikov towers and stable modules. We also
show that (C, C0 , A) := (sk − M od, sk − M od, sk − Alg) is a HA context in the sense
of Def. 1.1.0.11.
In §2.2.2 we show how the general definitions of properties of modules (e.g. pro-
jective, flat, perfect) and of morphisms between commutative rings in C (e.g. finitely
presented, flat, (formally) smooth, (formally) étale, Zariski open immersion) given
in Chapter 1.2 translates concretely in the present context. The basic idea here is
that of strongness which says that a module M over a simplicial k-algebra A (re-
spectively, a morphism A → B in sk − Alg) has the property P, defined in the
abstract setting of Chapter 1.2, if and only if π0 (M ) has the corresponding clas-
sical property as a π0 (A)-module and π0 (M ) ⊗π0 (A) π∗ (A) ≃ π∗ (M ) (respectively,
the induced morphism π0 (A) → π0 (B) has the corresponding classical property, and
π0 (B) ⊗π0 (A) π∗ (A) ≃ π∗ (B)). A straightforward extension of the étale topology
to simplicial k-algebras, then provides us with an étale model site (k − D− Af f, ét)
satisfying assumption 1.3.2.2 (Def. 2.2.2.12 and Lemma 2.2.2.13) and with the cor-
responding model category k − D− Af f ∼,ét of ét D− -stacks (Def. 2.2.2.14). The
homotopy category Ho(k − D− Af f ∼,ét ) of ét D− -stacks will be simply denoted by
D− St(k). We conclude the section with two useful corollaries about topological in-
variance of étale and Zariski open immersions (Cor. 2.2.2.9 and 2.2.2.10), stating that
ths small étale and Zariski site of a simplicial ring A is equivalent to the corresponding
site of π0 (A).
In §2.2.3 we describe our HAG context (Def. 1.3.2.13) for derived algebraic ge-
ometry by choosing the class P to be the class of smooth morphisms in sk − Alg and
the model topology to be the étale topology. This HAG context (C, C0 , A, τ, P) :=
(sk − M od, sk − M od, sk − Alg, ét, smooth) is shown to satisfy Artin’s conditions (Def.
2.2.3.2) relative to the HA context (C, C0 , A) := (sk − M od, sk − M od, sk − Alg) in
Prop. 1.4.3.1; as a corollary of the general theory of Part I, this gives (Cor. 2.2.3.3)
an obstruction theory (respectively, a relative obstruction theory) for any n-geometric
D− -stack (resp., for any n-representable morphism between D− -stacks), and in par-
ticular a (relative) cotangent complex for any n-geometric D− -stack (resp., for any
n-representable morphism). We finish the section showing that the properties of
being flat, smooth, ètale and finitely presented can be extended to n-representable
morphisms between D− -stacks (Lemma 2.2.3.4), and by the definition of open and
closed immersion of D− -stacks (Def. 2.2.3.5).
In §2.2.4 we study truncations of derived stacks. The inclusion functor j : k −
Af f ֒→ sk − Alg, that sends a commutative k-algebra R to the constant simplicial
k-algebra R, is Quillen right adjoint to the functor π0 : sk − Alg → k − Alg, and this
Quillen adjunction induces an adjunction
i := Lj! : St(k) −→ D− St(k)
St(k) ←− D− St(k) : Ri∗ =: t0
between the (homotopy) categories of derived and un-derived stacks. The functor i
is fully faithful and commutes with homotopy colimits (Lemma 2.2.4.1) and embeds
the theory of stacks into the theory of derived stacks, while the functor t0 , called the
 INTRODUCTION TO PART 2 125

truncation functor, sends the affine stack corresponding to a simplicial k-algebra A

to the affine scheme Spec π0 (A), and commutes with homotopy limits and colimits
(Lemma 2.2.4.2). Both the inclusion and the truncation functor preserve n-geometric
stacks and flat, smooth, étale morphisms between them (Prop. 2.2.4.4). This gives
a nice compatibility between the theories in §2.1 and §2.2, and therefore between
moduli spaces and their derived analogs.
In particular, we get the that for any Artin algebraic stack in groupoids (actually,
any Artin n-stack) X , i(X ) has an obstruction theory, and therefore a cotangent
complex; in other words viewing Artin stacks as derived stacks simplifies and clarifies
a lot their infinitesimal theory, as already remarked in this introduction. We finish the
section by showing (Prop. 2.2.4.7) that for any geometric D− -stack F , its truncation,
viewed again as a derived stack (i.e. it0 (F )) sits inside F as a closed sub-stack, and
one can reasonably think of F as behaving like a formal thickening of its truncation.
In §2.2.5 we give useful criteria for a n-representable morphism between D− -
stacks being smooth (respectively, étale) in terms of locally finite presentation of the
induced morphism on truncations, and infinitesimal lifting properties (Prop. 2.2.5.1,
resp., Prop. 2.2.5.4) or properties of the cotangent complex (Cor. 2.2.5.3, resp., Cor.
2.2.5.6).
The final §2.2.6 of Chapter 2.2 contains applications of derived algebraic geome-
try to the construction of various derived versions of moduli spaces as D− -stacks. In
2.2.6.1 we first show (Lemma 2.2.6.1) that the stack of rank n vector bundles when
viewed as a D− -stack using the inclusion i : St(k) → D− St(k) is indeed isomorphic to
the derived stack Vectn of rank n vector bundles defined as in §1.3.7. Then, for any
simplicial set K, we define the derived stack RLocn (K) as the derived exponentiation
of Vectn with respect to K (Def. 2.2.6.2), show that when K is finite dimensional
then RLocn (K) is a finitely presented 1-geometric D− -stack (Lemma 2.2.6.3), and
identify its truncation with the usual Artin stack of rank n local systems on K (Lemma
2.2.6.4). Finally we give a more concrete geometric interpretation of RLocn (K) as a
moduli space of derived geometric objects (derived rank n local systems) on the topo-
logical realization |K| (Prop. 2.2.6.5) and show that the tangent space of RLocn (K)
at a global point correponding to a rank n local system E on K is the cohomology
complex C ∗ (K, E ⊗k E ∨ )[1] (Prop. 2.2.6.6). The latter result shows in particular that
the D− - stack RLocn (K) depends on strictly more than the fundamental groupoid
of K (because its tangent spaces can be nontrivial even if K is simply connected)
and therefore carries higher homotopical informations as opposed to the usual (i.e
underived) Artin stack of rank n local systems. In subsection 2.2.6.2 we treat the
case of algebras over an operad. If O is an operad in the category of k-modules, we
consider a simplicial presheaf AlgO −
n on k − D Af f which associates to any simplicial
k-algebra A the nerve of the subcategory O − Alg(A) of weak equivalences in the
category of cofibrant algebras over the operad O ⊗k A (which is an operad in simpli-
cial A-modules) whose underlying A-module is a vector bundle of rank n. In Prop.
2.2.6.8 we show that AlgO −
n is a 1-geometric quasi-compact D -stack, and in Prop.
2.2.6.9 we identify its tangent space in terms of derived derivations. In subsection
2.2.6.3, using a special case of J.Lurie’s representability criterion (see Appendix C),
we give sufficient conditions for the D− -stack Map(X , F ) of morphisms of derived
stacks i(X ) → F (for X ∈ St(k)) to be n-geometric (Thm. 2.2.6.11), and compute its
tangent space in two particular cases (Cor. 2.2.6.14 and 2.2.6.15). The latter of these,
i.e. the case RMDR (X) := Map(i(XDR ), Vectn ), where X is a complex smooth pro-
jective variety, is particularly important because it is the first step in the construction
of a derived version of non-abelian Hodge theory which will be investigated in a future
126 INTRODUCTION TO PART 2

work.

In §2.3 we treat the case of complicial (or unbounded) algebraic geometry,

i.e. homotopical algebraic geometry over the base category C := C(k) of (unbounded)
complexes of modules over a commutative Q-algebra k 2.
Here Comm(C) is the model category k − cdga of commutative differential graded
k-algebras, (called shortly cdga’s) whose opposite model category will be denoted
by k − DAf f . A new feature of complicial algebraic geometry with respect to de-
rived algebraic geometry is the existence of two interesting HA contexts (Lemma
2.3.1.1): the weak one (C, C0 , A) := (C(k), C(k), k − cdga), and the connective one
(C, C0 , A) := (C(k), C(k)≤0 , k − cdga0 ) where C(k)≤0 is the full subcategory of C(k)
consisting of complexes which are cohomologically trivial in positive degrees and
k − cdga0 is the full subcategory of k − cdga of connected algebras (i.e. cohomolog-
ically trivial in non-zero degrees). A related phenomenon is the fact that the notion
of strongness that describes completely the properties of modules and of morphisms
between between simplicial algebras in derived algebraic geometry, is less strictly
connected with standard properties of modules and of morphisms between between
cdga’s. For example, morphisms between cdga’s which have strongly the property P
(e.g. P= flat, étale etc.; see Def. 2.3.1.3) have the property P but the converse is
true only if additional hypotheses are met (Prop. 2.3.1.4). Using the strong version
of étale morphisms, we endow the category k − DAf f with the strongly étale model
topology s-ét (Def. 2.3.1.6 and Lemma 2.3.1.7 ), and define the model category of
D-stacks as k − DAf f ∼,s-ét (Def. 2.3.1.8). The homotopy category of k − DAf f ∼,s-ét
will be simply denoted by DSt(k).
While the notion of stack does not depend on the HA or HAG contexts chosen
but only the base model site, the notion of geometric stack depends on the HA and
HAG contexts. In §2.3.2 we complete the weak HA context above to the HAG context
(C, C0 , A, τ, P) := (C(k), C(k), k − cdga, s-ét, Pw ), where Pw is the class of formally
perfect morphisms (Def. 1.2.7.2), and call the corresponding geometric stacks weakly
geometric stacks (Def. 2.3.2.2).
Section 2.3.3 contains some interesting examples of weakly geometric stacks. We
first show (Prop. 2.3.3.1) that the stack Perf of perfect modules (defined in an ab-
stract context in Def. 1.3.7.5) is weakly 1-geometric, categorically locally of finite
presentation and its diagonal is (−1)-representable. In subsection 2.3.3.2 we first de-
fine the simplicial presheaf Ass sending a cdga A to the nerve of the subcategory of
weak equivalences in the category of associative and unital (not necessarily commu-
tative) cofibrant A-dg algebras whose underlying A-dg module is perfect, and show
that this is a weakly 1-geometric D-stack (Prop. 2.3.3.2 and Cor. 2.3.3.4). Then we
define (using the model structure on dg-categories of [Tab]) a dg-categorical variation
of Ass, denoted by Cat∗ , by sending a cdga A to the nerve of the subcategory of
weak equivalences of the model category of A-dg categories D which are connected
and have perfect and cofibrant A-dg-modules of morphisms, and prove (Prop. 2.3.3.5)
that the canonical classifying functor Ass → Cat∗ is a weakly 1-representable, f p
epimorphism of D-stacks; it follows that Cat∗ is a weakly 2-geometric D-stack (Cor.
2.3.3.5).
In §2.3.4 we switch to the connective HA context (C, C0 , A) := (C(k), C(k)≤0 , k −
cdga0 ) and complete it to the connective HAG context
(C(k), C(k)≤0 , k − cdga0 , s-ét, fip-smooth),
2The case of k of positive characteristic can be treated as a special case of brave new algebraic
geometry (§2.4) over the base Hk, H being the Eilenberg-Mac Lane functor.
 INTRODUCTION TO PART 2 127

where fip-smooth is the class of formally perfect (Def. 1.2.7.1) and formally i-smooth
morphisms (Def. 1.2.8.1). Geometric stacks in this HAG context will be simply
called geometric D-stacks (Def. 2.3.4.2); since fip-smooth morphisms are in Pw , any
geometric D-stack is weakly geometric. As opposed to the weak HAG context, this
connective context indeed satisfies Artin’s condition of Def. 1.4.3.1 (Prop. 2.3.4.3),
and therefore any geometric D-stack has an obstruction theory (and a cotangent
complex).
In §2.3.5 we give some examples of geometric D-stacks. We first observe (sub-
section 2.3.5.1) that the normalization functor N : sk − Alg → k − cdga induces a
fully faithful functor j := LN! : D− St(k) ֒→ DSt(k). This provides us with lots
of examples of (geometric) D-stacks. In subsection 2.3.5.2 we study the D-stack of
CW-perfect modules. After having defined, for any cdga A, the notion of CW-A-dg-
module of amplitude in [a, b] (Def. 2.3.5.2) and proved some stability properties of
this notion (Lemma 2.3.5.3), we define the sub-D-stack Perf CW [a,b] ⊂ Perf , consisting
of all perfect modules locally equivalent to some CW-dg-modules of amplitude con-
tained in [a, b]. We prove that Perf CW[a,b] is 1-geometric (Prop. 2.3.5.4), and that its
tangent space at a point corresponding to a perfect CW-A-dg-module E is given by
the complex (E ∨ ⊗LA E)[1] (Cor. 2.3.5.6). In subsection 2.3.5.3 we define the D-stack
of CW-dg-algebras as the homotopy pullback of Ass → Perf along the inclusion
Perf CW
[a,b] → Perf , prove that it is 1-geometric and compute its tangent space at a
global point in terms of derived derivations (Cor. 2.3.5.9). Finally subsection 2.3.5.4
is devoted to the analysis of the D-stack CatCW ∗, [n,0] of CW-dg-categories of perfect am-
plitude in [n, 0] (Def. 2.3.5.10). As opposite to the weakly geometric D-stack Cat∗ ,
that cannot have a reasonable infinitesimal theory, its full sub-D-stack CatCW ∗, [n,0] is
not only a 2-geometric D-stack but has a tangent space that can be computed in terms
of Hochschild homology (Thm. 2.3.5.11). As corollaries of this important result we
can prove a folklore statement (see e.g. [Ko-So, p. 266] in the case of A∞ -categories
with one object) regarding the deformation theory of certain negative dg-categories
being controlled by the Hochschild complex of dg-categories (Cor. 2.3.5.12), and a
result showing that if one wishes to keep the existence of the cotangent complex, the
restriction to non-positively graded dg-categories is unavoidable (Cor. 2.3.5.13).

In the last, short §2.4, we establish the basics of brave new algebraic ge-
ometry, i.e. of homotopical algebraic geometry over the base category C = SpΣ
of symmetric spectra ([HSS, Shi]). We consider C endowed with the positive model
structure of [Shi], which is better behaved than the usual one when dealing with com-
mutative monoid objects an modules over them. We denote Comm(C) by S − Alg
(and call its objects commutative S-algebras, S being the sphere spectrum, or some-
times bn-rings), and its opposite model category by S − Af f . Like in the case of
complicial algebraic geometry, we consider two HA contexts here (Lemma 2.4.1.1):
(C, C0 , A) := (SpΣ , SpΣ , S − Alg) and (C, C0 , A) := (SpΣ , SpΣ Σ
c , S − Alg0 ) where Spc
is the subcategory of connective symmetric spectra, and S − Alg0 the subcategory
of S-algebras with homotopy groups concentrated in degree zero. After giving some
examples of formally étale and formally thh-étale maps between bn-rings, we define
(Def. 2.4.1.3) strong versions of flat, (formally) étale, (formally) smooth, and Zariski
open immersions, exactly like in chapters 2.2 and 2.3, and give some results relat-
ing them to the corresponding non-strong notions (Prop. 2.4.1.4). An interesting
exception to this relationship occurs in the case of smooth morphism: the Eilenberg-
MacLane functor H from commutative rings to bn rings does not preserve (formal)
smoothness in general, though it preserves (formal) strong smoothness, due to the
128 INTRODUCTION TO PART 2

presence of non-trivial Steenrod operations in characteristic p > 0 (Prop. 2.4.1.5).

We endow the category S − Af f with the strong étale model topology s-ét (Def.
2.4.1.7 and Lemma 2.4.1.8), and define the model category S − Af f ∼, s-ét of S-stacks
(Def. 2.4.1.9). The homotopy category of S − Af f ∼, s-ét will be simply denoted by
St(S). The two HA contexts defined above are completed (Cor. 2.4.1.11) to two
different HAG contexts by choosing for both HA contexts the s-ét model topology,
and the class Ps-ét of strongly étale morphsims for the first HA context (respectively,
the class P of fip-morphisms for the second HA context). We call geometric Deligne-
Mumford S-stacks (respectively, geometric S-stacks) the geometric stacks in the first
(resp, second) HAG context, and observe that both contexts satisfy Artin’s condition
of Def. 1.4.3.1 (Prop. 2.4.1.13).
Finally in §2.4.2, we use the definition of topological modular forms of Hopkins-
Miller to build a 1-geometric Deligne-Mumford stack E S that is a “bn-derivation”
of the usual stack E of generalized elliptic curves 3 (i.e., the truncation of E S is
E), and such that the spectrum tmf coincide with the spectrum of functions on E S .
We conclude by the remark that a moduli theoretic interpretation of E S (or, most
probably some variant of it), i.e. finding out which are the brave new objects that
it classifies, could give not only interesting new geometry over bn rings but also new
insights on classical objects of algebraic topology.

3See e.g. [Del-Rap, IV], where it is denoted by M .

(1)
 CHAPTER 2.1

Geometric n-stacks in algebraic geometry (after C.

Simpson)

All along this chapter we fix an associative commutative ring k ∈ U with unit.

2.1.1. The general theory

We consider C = k − M od, the category of k-modules in the universe U. We
endow the category C with the trivial model structure for which equivalences are
isomorphisms and all morphisms are cofibrations and fibrations. The category C is
furthermore a symmetric monoidal model category for the monoidal structure given by
the tensor product of k-modules. The assumptions 1.1.0.1, 1.1.0.2, 1.1.0.3 and 1.1.0.4
are all trivially satisfied. The category Comm(C) is identified with the category k −
Alg, of commutative (associative and unital) k-algebras in U, endowed with the trivial
model structure. Objects in k − Alg will simply be called commutative k-algebras,
without any reference to the universe U. For any A ∈ k − Alg, the category A − M od
is the usual symmetric monoidal category of A-modules in U, also endowed with its
trivial model structure. Furthermore, we have RHomA (M, N ) ≃ HomA (M, N ) for
any two objects A ∈ k − Alg, and is the usual A-module of morphisms from M to N .
We set C0 := k − M od, and A := k − Alg. The triplet (k − M od, k − M od, k − Alg)
is then a HA context in the sense of Def. 1.1.0.11.
The category Af fC is identified with k − Alg op , and therefore to the category
of affine k-schemes in U. It will simply be denoted by k − Af f and its objects will
simply be called affine k-schemes, without any reference to the universe U. The model
category of pre-stacks k − Af f ∧ = Af fC∧ is simply the model category of V-simplicial
presheaves on k − Af f , for which equivalences and fibrations are defined levelwise.
The Yoneda functor
h : k − Af f −→ k − Af f ∧
is the usual one, and sends an affine k-scheme X to the presheaf of sets it represents
hX (considered as a presheaf of constant simplicial sets). Furthermore we have natural
isomorphisms of functors

h ≃ h ≃ Rh : k − Af f −→ Ho(k − Af f ∧ ),

which is nothing else than the natural composition

h / k − Af f ∧ / Ho(k − Af f ∧ ).
k − Af f

We let τ = ét, the usual étale pre-topology on k − Af f (see e.g. [Mil]). Recall
that a family of morphisms

{Xi = Spec Ai −→ X = Spec A}i∈I

129
130 2.1. GEOMETRIC n-STACKS IN ALGEBRAIC GEOMETRY (AFTER C. SIMPSON)

is an ét-covering family if and only if it contains a finite sub-family {Xi −→ X}i∈J ,

J ⊂ I, such that the corresponding morphism of commutative k-algebras
Y
A −→ Ai
i∈J

is a faithfully flat and étale morphism of commutative rings.

Lemma 2.1.1.1. The étale topology on k − Af f satisfies assumption 1.3.2.2.
Proof. Points (1) and (2) of 1.3.2.2 are clear. Point (3) is induced by the
faithfully flat descent of quasi-coherent modules for affine f f qc-hypercovers (see e.g.
[SGA4-II, Exp. V bis ])). 
The model category of stacks k − Af f ∼,ét is the projective model structure for
simplicial presheaves on the Grothendieck site (k − Af f, ét), as defined for example
in [Bl] (see also [To1, §1] ). Its homotopy category, denoted simply as St(k), is then
identified with the full subcategory of Ho(k−Af f ∧ ) consisting of simplicial presheaves
F : k − Alg = k − Af f op −→ SSetV
satisfying the following two conditions
• For any two commutative k-algebra A and B, the natural morphism
F (A × B) −→ F (A) × F (B)
is an isomorphism in Ho(SSet).
• For any co-augmented co-simplicial commutative k-algebra, A −→ B∗ , such
that the augmented simplicial object
Spec B∗ −→ Spec A
is an étale hypercover, the natural morphism
F (A) −→ Holimn∈∆F (Bn )
is an isomorphism in Ho(SSet).
It is well known (and also a consequence of Cor. 1.3.2.8) that the étale topology
is sub-canonical; therefore there exists a fully faithful functor
h : k − Af f −→ St(k) ⊂ Ho(k − Af f ∼,ét ).
Furthermore, we have
Spec A ≃ Spec A ≃ RSpec A
for any A ∈ k − Alg.
We set P to be the class of smooth morphisms in k−Af f in the sense of [EGAIV,
17.3.1]. It is well known that our assumption 1.3.2.11 is satisfied (e.g. that smooth
morphisms are étale-local in the source and target, see for example [EGAIV, 17.3.3,
17.3.4, 17.7.3]). In particular, we get that (k − M od, k − M od, k − Alg, ét, P) is a
HAG context in the sense of Def. 1.3.2.13. The general definition 1.3.3.1 can then
be applied, and provides a notion of n-geometric stack in St(k). A first important
observation (the lemma below) is that n-geometric stacks are n-stacks in the sense of
[S3]. This is a special feature of standard algebraic geometry, and the same would
be true for any theory for which the model structure on C is trivial: the geometric
complexity is a bound for the stacky complexity.
Recall that a stack F ∈ St(k) is n-truncated if for any X ∈ k − Af f , any s ∈
π0 (F (X)) and any i > n, the sheaf πi (F, s) is trivial. By [HAGI, 3.7], this is
equivalent to say that for any stack G the simplicial set Rτ Hom(G, F ) is n-truncated.
 2.1.1. THE GENERAL THEORY 131

Lemma 2.1.1.2. Let F be an n-geometric stack in St(k). Then F is (n + 1)-

truncated.

Proof. The proof is by induction on n. Representable stacks are nothing else

than affine schemes, and therefore are 0-truncated. Suppose that the lemma is known
for m < n. Let F be an n-geometric stack, X ∈ k − Af f and s ∈ π0 (F (X)). We have
a natural isomorphism of sheaves on k − Af f /X

πi (F, s) ≃ πi−1 (X ×hF X, s),

where X −→ F is the morphisms of stacks corresponding to s. As the diagonal of

F is (n − 1)-representable, X ×hF X ≃ F ×F ×h F X ×h X is (n − 1)-geometric. By
induction we find that πi (F, s) ≃ ∗ for any i > n. 

Lemma 2.1.1.2 justifies the following terminology, closer to the usual terminology
one can find in the literature.

Definition 2.1.1.3. An Artin n-stack is an n-truncated stack which is m-geometric

for some integer m.

The general theory of Artin n-stacks could then be pursued in a similar fashion as
for Artin stacks in [La-Mo]. A part of this is done in [S3] and will not be reproduced
here, as many of these statements will be settled down in the more general context
of geometric D− -stacks (see §2.2). Let us mention however, that as explained in Def.
1.3.6.2, we can define the notions of flat, smooth, étale, unramified, regular, Zariski
open immersion . . . morphisms between Artin n-stacks. These kinds of morphisms are
as usual stable by homotopy pullbacks, compositions and equivalences. In particular
this allows the following definition.

Definition 2.1.1.4. • An Artin n-stack is a Deligne-Mumford n-stack if

there exists an n-atlas {Ui } for F such that each morphism Ui −→ F is an
étale morphism.
• An Artin n-stack is an algebraic space if it is a Deligne-Mumford n-stack,
and if furthermore the diagonal F −→ F ×h F is a monomorphism in the
sense of Def. 1.3.6.4.
• An Artin n-stack F is an scheme if there exists an n-atlas {Ui } for F such
that each morphism Ui −→ F is a monomorphism.

Remark 2.1.1.5. (1) An algebraic space in the sense of the definition above
which is automatically a 1-geometric stack, and is nothing else than an
algebraic space in the usual sense. Indeed, this can be shown by induction on
n: an algebraic space which is also n-geometric is by definition the quotient
of a union of affine schemes X by some étale equivalence relation R ⊂ X ×X
where R is an algebraic space which is (n − 1)-geometric. In particular, R
being a subobject in X × X we see that R is a separated algebraic space,
and thus is a 0-geometric stack. This implies that X/R is a 1-geometric
stack. In the same way, any scheme is automatically a 1-geometric stack.
Moreover, algebraic spaces (resp. schemes) which are 0-geometric stacks are
precisely algebraic spaces (resp. schemes) with an affine diagonal.
(2) Thought there is a small discrepancy between the notion of Artin n-stack
and the notion of n-geometric stack in St(k), our notion of Artin n-stack is
equivalent to the notion of slightly geometric n-stacks of [S3].
132 2.1. GEOMETRIC n-STACKS IN ALGEBRAIC GEOMETRY (AFTER C. SIMPSON)

2.1.2. Comparison with Artin’s algebraic stacks

Artin n-stacks as defined in the last section are simplicial presheaves, whereas
Artin stacks are usually presented in the literature using the theory of fibered cate-
gories (se e.g. [La-Mo]). In this section we briefly explain how the theory of fibered
categories in groupoids can be embedded in the theory of simplicial presheaves, and
how this can be used in order to compare the original definition of Artin stacks to
our definition of Artin n-stacks.

In [Hol], it is shown that there exists a model category Grpd/S, of cofibered

categories in groupoids over a Grothendieck site S. The fibrant objects for this model
structure are precisely the stacks in groupoids in the sense of [La-Mo], and the equiv-
alences in Grpd/S are the morphisms of cofibered categories becoming equivalences
on the associated stacks (i.e. local equivalences). There exists furthermore a Quillen
equivalence
p : P (S, Grpd) −→ Grpd/S P (S, Grpd) ←− Grpd/S : Γ,
where P (S, Grpd) is the local projective model category of presheaves of groupoids
on S. Finally, there exists a Quillen adjunction
Π1 : SP rτ (S) −→ P (S, Grpd) SP rτ (S) ←− P (S, Grpd) : B,
where Π1 is the natural extension of the functor sending a simplicial set to its funda-
mental groupoid, B is the natural extension of the nerve functor, and the SP rτ (S)
is the local projective model structure of simplicial presheaves on the site S (also
denoted by S ∼,τ in our context, at least when S has limits and colimits and thus can
be considered as a model category with the trivial model structure). The functor B
preserves equivalences and the induced functor
B : Ho(P (S, Grpd)) −→ Ho(SP rτ (S))
is fully faithful and its image consists of all 1-truncated objects (in the sense of [HAGI,
§3.7]). Put in another way, the model category P (S, Grpd) is Quillen equivalent to the
S 2 -nullification of SP rτ (S) (denoted by SP rτ≤1 (S) in [HAGI, §3.7]). In conclusion,
there exists a chain of Quillen equivalences
Grpd/S ⇄ P (S, Grpd) ⇄ SP rτ≤1 (S),
and therefore a well defined adjunction
t : Ho(SP rτ (S)) −→ Ho(Grpd/S) Ho(SP rτ (S)) ←− Ho(Grpd/S) : i,
such that the right adjoint
i : Ho(Grpd/S) −→ Ho(SP rτ (S))
is fully faithful and its image consists of all 1-truncated objects.
The category Ho(Grpd/S) can also be described as the category whose objects
are stacks in groupoids in the sense of [La-Mo], and whose morphisms are given by
1-morphisms of stacks up to 2-isomorphisms. In other words, for two given stacks F
and G in Grpd/S, the set of morphisms from F to G in Ho(Grpd/S) is the set of
isomorphism classes of the groupoid Hom(F, G), of morphisms of stacks. This implies
that the usual category of stacks in groupoids, up to 2-isomorphisms, can be identified
through the functor i with the full subcategory of 1-truncated objects in Ho(SP rτ (S)).
Furthermore, the functor i being defined as the composite of right derived functors
and derived Quillen equivalences will commutes with homotopy limits. As homotopy
limits in Grpd/S can also be identified with the 2-limit of stacks as defined in [La-Mo],
 2.1.2. COMPARISON WITH ARTIN’S ALGEBRAIC STACKS 133

the functor i will send 2-limits to homotopy limits. As a particular case we obtain
that i sends the 2-fiber product of stacks in groupoids to the homotopy fiber product.
The 2-categorical structure of stacks in groupoids can also be recovered from the
model category SP rτ (S). Indeed, applying the simplicial localization techniques of
[D-K1] to the Quillen adjunctions described above, we get a well defined diagram of
S-categories
L(Grpd/S) −→ LP (S, Grpd) −→ LSP rτ (S),
which is fully faithful in the sense of [HAGI, Def. 2.1.3]. In particular, the S-category
L(Grpd/S) is naturally equivalent to the full sub-S-category of LSP rτ (S) consisting
of 1-truncated objects. Using [D-K3], the S-category L(Grpd/S) is also equiva-
lent to the S-category whose objects are stacks in groupoids, cofibrant as objects
in Grpd/S, and whose morphisms simplicial sets are given by the simplicial Hom’s
sets of Grpd/S. These simplicial Hom’s sets are simply the nerves of the groupoid
of functors between cofibered categories in groupoids. In other words, replacing the
simplicial sets of morphisms in L(Grpd/S) by their fundamental groupoids, we find
a 2-category naturally 2-equivalent to the usual 2-category of stacks in groupoids on
S. Therefore, we see that the 2-category of stacks in groupoids can be identified, up
to a natural 2-equivalence, as the 2-category obtained from the full sub-S-category of
LSP rτ (S) consisting of 1-truncated objects, by replacing its simplicial sets of mor-
phisms by their fundamental groupoids.

We now come back to the case where S = (k − Af f, ét), the Grothendieck site of
affine k-schemes with the étale pre-topology. We have seen that there exists a fully
faithful functor
i : Ho(Grpd/k − Af f ∼,ét ) −→ St(k),
from the category of stacks in groupoids up to 2-isomorphisms, to the homotopy
category of stacks. The image of this functor consists of all 1-truncated objects and it
is compatible with the simplicial structure (i.e. possesses a natural lifts as a morphism
of S-categories). We also have seen that i sends 2-fiber products of stacks to homotopy
fiber products.
Using the functor i, every stack in groupoids can be seen as an object in our
category of stacks St(k). For example, all examples of stacks presented in [La-Mo]
give rise to stacks in our sense. The proposition below subsumes the main properties
of the functor i, relating the usual notion of scheme, algebraic space and stack to the
one of our definition Def. 2.1.1.4. Recall that a stack in groupoids X is separated
(resp. quasi-separated) if its diagonal is a proper (resp. separated) morphism.
Proposition 2.1.2.1. (1) For any commutative k-algebra A, there exists a
natural isomorphism
i(Spec A) ≃ Spec A.
(2) If X is a scheme (resp. algebraic space, resp. Deligne-Mumford stack, resp.
Artin stack) with an affine diagonal in the sense of [La-Mo], then i(X) is
an Artin 0-stack which is 0-geometric (resp. an Artin 0-stack which is 0-
geometric, resp. a Deligne-Mumford 1-stack which is 0-geometric, resp. an
Artin 1-stack which is 0-geometric) in the sense of Def. 2.1.1.4.
(3) If X is a scheme (resp. algebraic space, resp. Deligne-Mumford stack, resp.
Artin stack) in the sense of [La-Mo], then i(X) is an Artin 0-stack which is
1-geometric (resp. an Artin 0-stack which is 1-geometric, resp. a Deligne-
Mumford 1-stack which is 1-geometric, resp. an Artin 1-stack which is 1-
geometric) in the sense of Def. 2.1.1.4.
134 2.1. GEOMETRIC n-STACKS IN ALGEBRAIC GEOMETRY (AFTER C. SIMPSON)

(4) Let f : F −→ G be a morphism between Artin stacks in the sense of

[La-Mo]. Then the morphism f is flat (resp. smooth, resp. étale, resp. un-
ramified, resp. Zariski open immersion) if and only if i(f ) : i(F ) −→ i(G)
is so.
Proof. This readily follows from the definition using the fact that i preserves
affine schemes, epimorphisms of stacks, and sends 2-fiber products to homotopy pull-
backs. 
Remark 2.1.2.2. If we try to apply the general infinitesimal and obstruction
theory developed in §1.4 to the present HAG-context, we immediately see that this is
impossible because the suspension functor S : Ho(C) = k − M od → k − M od = Ho(C)
is trivial. On the other hand, the reader might object that there is already a well
established infinitesimal theory, at least in the case of schemes, algebraic spaces and
for a certain class of algebraic stacks in groupoids, and that our theory does not
seem to be able to reproduce it. The answer to this question turns out to be both
conceptually and technically relevant. First of all, if we look at e.g. the definition of
the cotangent complex of a scheme ([Ill, 2.1.2]) we realize that a basic and necessary
step is to enlarge the category of rings to the category of simplicial rings in order
to be able to consider free (or more generally cofibrant) resolutions of maps between
rings. In our setup, this can be reformulated by saying that in order to get the
correct infinitesimal theory, even for ordinary schemes, it is necessary to view them
as geometric objects in derived algebraic geometry, i.e. on homotopical algebraic
geometry over the base category C = sk − M od of simplicial k-modules (so that
Comm(C) is exactly the category of commutative simplicial k-algebras). In other
words, the usual infinitesimal theory of schemes is already “secretly” a part of derived
algebraic geometry, that will be studied in detail in the next chapter 2.2. Moreover,
as it will be shown, this approach has, even for classical objects like schemes or Artin
stacks in groupoids, both conceptual advantages (like e.g. the fact that the cotangent
complex of a scheme can be interpreted geometrically as a genuine cotangent space
to the scheme when viewed as a derived stack, satisfying a natural universal property,
while the cotangent complex of a scheme do not have any universal property inside
the theory of schemes), and technical advantages (like the fact, proved in Cor. 2.2.4.5,
that any Artin stack in groupoids has an obstruction theory).
Convention 2.1.2.3. From now on we will omit mentioning the functor i, and
will simply view stacks in groupoids as objects in St(k). In particular, we will allow
ourselves to use the standard notions and vocabulary of the general theory of schemes.
 CHAPTER 2.2

Derived algebraic geometry

All along this chapter k will be a fixed commutative (associative and unital) ring.

2.2.1. The HA context

In this section we specialize our general theory of Part I to the case where C = sk−
M od, is the category of simplicial k-modules in the universe U. The category sk−M od
is endowed with its standard model category structure, for which the fibrations and
equivalences are defined on the underlying simplicial sets (see for example [Goe-Ja]).
The tensor product of k-modules extends naturally to a levelwise tensor product on
sk − M od, making it into a symmetric monoidal model category. Finally, sk − M od
is known to be a U-combinatorial proper and simplicial model category.
The model category sk − M od is known to satisfy assumptions 1.1.0.1, 1.1.0.2 and
1.1.0.3 (see [Schw-Shi]). Finally, it follows easily from [Q1, II.4, II.6] that sk − M od
also satisfies assumption 1.1.0.4.

The category Comm(sk − M od) will be denoted by sk − Alg, and its objects
will be called simplicial commutative k-algebras. More generally, for A ∈ sk − Alg,
the category A − Comm(C) will be denoted by A − Algs . For any A ∈ sk − Alg
we will denoted by A − M ods the category of A-modules in sk − M od, which is
nothing else than the category of simplicial modules over the simplicial ring A. The
model structure on sk − Alg, A − Algs and A − M ods is the usual one, for which
the equivalences and fibrations are defined on the underlying simplicial sets. For an
object A ∈ sk − Alg, we will denote by πi (A) its homotopy group (pointed at 0). The
graded abelian group π∗ (A) inherits a structure of a commutative graded algebra from
A, which defines a functor A 7→ π∗ (A) from sk − Alg to the category of commutative
graded k-algebras. More generally, if A is a simplicial commutative k-algebra, and M
is an A-module, the graded abelian group π∗ (M ) has a natural structure of a graded
π∗ (A)-module.
There exists a Quillen adjunction
π0 : sk − Alg −→ k − Alg sk − Alg ←− k − Alg : i,
where i sends a commutative k-algebra to the corresponding constant simplicial com-
mutative k-algebra. This Quillen adjunction induces a fully faithful functor
i : k − Alg −→ Ho(sk − Alg).
From now on we will omit to mention the functor i, and always consider k − Alg as
embedded in sk −Alg, except if the contrary is specified. Note that when A ∈ k −Alg,
also considered as an object in sk − Alg, we have two different notions of A-modules,
A − M od, and A − M ods . The first one is the usual category of A-modules, whereas
the second one is the category of simplicial objects in A − M od.
For any morphism of simplicial commutative k-algebras A −→ B, the B-module
LB/A constructed in 1.2.1.2 is naturally isomorphic in Ho(B − M od) to D. Quillen’s
cotangent complex introduced in [Q2]. In particular, if A −→ B is a morphism
135
136 2.2. DERIVED ALGEBRAIC GEOMETRY

between (non-simplicial) commutative k-algebras, then we have π0 (LB/A ) ≃ Ω1B/A .

More generally, we find by adjunction
π0 (LB/A ) ≃ Ω1π0 (B)/π0 (A) .
Recall also that A −→ B in k − Alg is étale in the sense of [EGAIV, 17.1.1] if and
only if LB/A ≃ 0 and B is finitely presented as a commutative A-algebra. In the same
way, a morphism A −→ B in k − Alg is smooth in the sense of [EGAIV, 17.1.1] if
and only if Ω1B/A is a projective B-module, πi (LB/A ) ≃ 0 for i > 0 and B is finitely
presented as a commutative A-algebra.. Finally, recall the existence of a natural first
quadrant spectral sequence (see [Q1, II.6 Thm. 6(b)])
T orπp ∗ (A) (π∗ (M ), π∗ (N ))q ⇒ πp+q (M ⊗LA N ),
for A ∈ sk − Alg and any objects M and N in A − M ods .
We set C0 := C = sk − M od, and A := sk − Alg. Then, clearly, assumption
1.1.0.6 is also satisfied. The triplet (sk − M od, sk − M od, sk − Alg) is then a HA
context in the sense of Def. 1.1.0.11. Note that for any A ∈ sk − Alg we have
(A − M ods )0 = A − M od, whereas (A − M ods )1 consists of all A-modules M such
that π0 (M ) = 0, also called connected modules.

For an integer n ≥ 0 we define the n-th sphere k-modules by Skn := S n ⊗ k ∈

sk − M od. The free commutative monoid on Skn is an object k[S n ] ∈ sk − Alg, such
that for any A ∈ sk − Alg there are functorial isomorphisms
[k[S n ], A]sk−Alg ≃ πn (A).
In the same way we define ∆n ⊗ k ∈ sk − M od and its associated free commutative
monoid k[∆n ]. There are natural morphisms k[S n ] −→ k[∆n+1 ], coming from the
natural inclusions ∂∆n+1 = S n ֒→ ∆n+1 . The set of morphisms
{k[S n ] −→ k[∆n+1 ]}n≥0
form a generating set of cofibrations in sk − Alg. The model category sk − Alg is
then easily checked to be compactly generated in the sense of Def. 1.2.3.4. A finite
cell object in sk − Alg is then any object A ∈ sk − Alg for which there exists a finite
sequence in sk − Alg
A0 = k / A1 / A2 / ... / Am = A,

such that for any i there exists a push-out square in sk − Alg

AO i / Ai+1
O

k[S ni ] / k[∆ni +1 ]

Our Prop. 1.2.3.5 implies that an object A ∈ sk − Alg is finitely presented in the
sense of Def. 1.2.3.1 if and only if it is equivalent to a retract of a finite cell object
(see also [EKMM, III.2] or [Kr-Ma, Thm. III.5.7] for other proofs). More generally,
for A ∈ sk − Alg, there exists a notion of finite cell object in A − Algs using the
elementary morphisms
A[S n ] := A ⊗k k[S n ] −→ A[∆n+1 ] := A ⊗k k[∆n+1 ].
In the same way, a morphism A −→ B in sk − Alg is finitely presented in the sense
of Def. 1.2.3.1 if and only if B is equivalent to a retract of a finite cell objects in
A − Algs . Prop. 1.2.3.5 also implies that any morphism A −→ B, considered as
 2.2.1. THE HA CONTEXT 137

an object in A − Algs , is equivalent to a filtered colimit of finite cell objects, so in

particular to a filtered homotopy colimit of finitely presented objects.
The Quillen adjunction between k − Alg and sk − Alg shows that the functor
π0 : sk − Alg −→ k − Alg
does preserve finitely presented morphisms. On the contrary, the inclusion functor
i : k − Alg −→ sk − Alg does not preserve finitely presented objects, and the finite
presentation condition in sk − Alg is in general stronger than in k − Alg.
For A ∈ sk − Alg, we also have a notion of finite cell objects in A − M ods , based
the generating set for cofibrations consisting of morphisms of the form
n
SA := A ⊗k Skn −→ ∆n+1
A := A ⊗k ∆n+1
A .
Using Prop. 1.2.3.5 we see that the finitely presented objects in A − M ods are the
objects equivalent to a retract of a finite cell objects (see also [EKMM, III.2] or
[Kr-Ma, Thm. III.5.7]). Moreover, the functor
π0 : A − M ods −→ π0 (A) − M od
is left Quillen, so preserves finitely presented objects. On the contrary, for A ∈ k−Alg,
the natural inclusion functor A − M od −→ A − M ods from A-modules to simplicial
A-modules does not preserve finitely presented objects in general.
The category sk − M od is also Quillen equivalent (actually equivalent) to the
model category C − (k) of non-positively graded cochain complexes of k-modules,
through the Dold-Kan correspondence ([We, 8.4.1]). In particular, the suspension
functor
S : Ho(sk − M od) −→ Ho(sk − M od)
corresponds to the shift functor E 7→ E[1] on the level of complexes, and is a fully
faithful functor. This implies that for any A ∈ sk − Alg, the suspension functor
S : Ho(A − M ods ) −→ Ho(A − M ods )
is also fully faithful. We have furthermore πi (S(M )) ≃ πi+1 (M ) for all M ∈ A−M ods .
The suspension and loop functors will be denoted respectively by
M [1] := S(M ) M [−1] := Ω(M ).
For any A ∈ sk − Alg, we can construct a functorial tower in sk − Alg, called the
Postnikov tower,
A / ... / A≤n / A≤n−1 / ... / A≤0 = π0 (A)

in such a way that πi (A≤n ) = 0 for all i > n, and the morphism A −→ A≤n induces
isomorphisms on the πi ’s for all i ≤ n. The morphism A −→ A≤n is characterized by
the fact that for any B ∈ sk − Alg which is n-truncated (i.e. πi (B) = 0 for all i > n),
the induced morphism
M apsk−Alg (A≤n , B) −→ M apsk−Alg (A, B)
is an isomorphism in Ho(SSet). This implies in particular that the Postnikov tower
is furthermore unique up to equivalence (i.e. unique as an object in the homotopy
category of diagrams). There exists a natural isomorphism in Ho(sk − Alg)
A ≃ Holimn A≤n .
For any integer n, the homotopy fiber of the morphism
A≤n −→ A≤n−1
is isomorphic in Ho(sk − M od) to S n ⊗k πn (A), and is also denoted by πn (A)[n].
The k-module πn (A) has a natural structure of a π0 (A)-module, and this induces a
138 2.2. DERIVED ALGEBRAIC GEOMETRY

natural structure of a simplicial π0 (A)-module on each πn (A)[i] for all i. Using the
natural projection A≤n−1 −→ π0 (A), we thus see the object πn (A)[i] as an object in
A≤n−1 − M ods . Note that there is a natural isomorphism in Ho(A≤n−1 − M ods )
S(πn (A)[i]) ≃ πn (A)[i + 1] Ω(πn (A)[i]) ≃ πn (A)[i − 1],
where πn (A)[i] is understood to be 0 for i < 0. We recall the following important and
well known fact.
Lemma 2.2.1.1. With the above notations, there exists a unique derivation
dn ∈ π0 (Derk (A≤n−1 , πn (A)[n + 1]))
such that the natural projection
A≤n−1 ⊕dn πn (A)[n] −→ A≤n−1
is isomorphic in Ho(sk − Alg/A≤n−1 ) to the natural morphism
A≤n −→ A≤n−1 .
Sketch of Proof. (See also [Ba] for more details).
The uniqueness of dn follows easily from our lemma 1.4.3.7, and the fact that the
natural morphism
LQZ(πn (A)[n + 1]) −→ πn (A)[n + 1]
induces an isomorphism on πi for all i ≤ n + 1 (this follows from our lemma 2.2.2.7
below). To prove the existence of dn , we consider the homotopy push-out diagram in
sk − Alg
A≤n−1 /B
O
O

A≤n / A≤n−1 .
The identity of A≤n−1 induces a morphism B −→ A≤n−1 , which is a retraction of
A≤n−1 −→ B. Taking the (n + 1)-truncation gives a commutative diagram
A≤n−1 / B≤n+1
O O
s

A≤n / A≤n−1 .

in such a way that s has a retraction. This easily implies that the morphism s
is isomorphic, in a non-canonical way, to the zero derivation A≤n−1 −→ A≤n−1 ⊕
πn (A)[n + 1]. The top horizontal morphism of the previous diagram then gives rise
to a derivation
dn : A≤n−1 −→ A≤n−1 ⊕ πn (A)[n + 1].
The diagram
dn
A≤n−1 / B≤n+1
O O
s

A≤n / A≤n−1 ,
is then easily checked to be homotopy cartesian, showing that A≤n −→ A≤n−1 is
isomorphic to A≤n−1 ⊕dn πn (A)[n] −→ A≤n−1 . 
 2.2.2. FLAT, SMOOTH, ÉTALE AND ZARISKI OPEN MORPHISMS 139

Finally, the truncation construction also exists for modules. For any A ∈ sk−Alg,
and M ∈ A − M ods , there exists a natural tower of morphisms in A − M ods
M / ... / M≤n / M≤n−1 / ... / M≤0 = π0 (M ),

such a way that πi (M≤n ) = 0 for all i > n, and the morphism M −→ M≤n induces
an isomorphisms on πi for i ≤ n. The natural morphism
M −→ Holimn M≤n
is an isomorphism in Ho(A − M ods ). Furthermore, the A-module M≤n is induced by
a natural A≤n -module, still denoted by M≤n , through the natural morphism A −→
A≤n . The natural projection M −→ M≤n is again characterized by the fact that for
any A-module N which is n-truncated, the induced morphism
M apA−Mods (M≤n , N ) −→ M apA−Mods (M, N )
is an isomorphism in Ho(SSet).
For an object A ∈ sk − Alg, the homotopy category Ho(Sp(A − M ods )), of stable
A-modules can be described in the following way. By normalization, the commutative
simplicial k-algebra A can be transformed into a commutative dg-algebra over k, N (A)
(because N is lax symmetric monoidal). We can therefore consider its model category
of unbounded N (A)-dg-modules, and its homotopy category Ho(N (A) − M od). The
two categories Ho(Sp(A−M ods )) and Ho(N (A)−M od) are then naturally equivalent.
In particular, when A is a commutative k-algebra, then N (A) = A, and one finds that
Ho(Sp(A − M ods )) is simply the unbounded derived category of A, or equivalently
the homotopy category of the model category C(A) of unbounded complexes of A-
modules
Ho(Sp(A − M ods )) ≃ D(A) ≃ Ho(C(A)).
Finally, using our Cor. 1.2.3.8 (see also [EKMM, III.7]), we see that the perfect
objects in the symmetric monoidal model category Ho(Sp(A − M ods )) are exactly
the finitely presented objects.

We now let k − D− Af f be the opposite model category of sk − Alg. We use

our general notations, Spec A ∈ k − D− Af f being the object corresponding to A ∈
sk − Alg. We will also sometimes use the notation
t0 (Spec A) := Spec π0 (A).

2.2.2. Flat, smooth, étale and Zariski open morphisms

According to our general definitions presented in §1.2 we have various notions
of projective and perfect modules, flat, smooth, étale, unramified . . . morphisms in
sk − Alg. Our first task, before visiting our general notions of stacks, will be to give
concrete descriptions of these notions.

Any object A ∈ sk−Alg gives rise to a commutative graded k-algebra of homotopy

π∗ (A), which is functorial in A. In particular, πi (A) is always endowed with a natural
structure of a π0 (A)-module, functorially in A. For a morphism A −→ B in sk − Alg
we obtain a natural morphism
π∗ (A) ⊗π0 (A) π0 (B) −→ π∗ (B).
More generally, for A ∈ sk − Alg and M an A-module, one has a natural morphism
of π0 (A)-modules π0 (M ) −→ π∗ (M ), giving rise to a natural morphism
π∗ (A) ⊗π0 (A) π0 (M ) −→ π∗ (M ).
140 2.2. DERIVED ALGEBRAIC GEOMETRY

These two morphisms are the same when the commutative A-algebra B is considered
as an A-module in the usual way.
Definition 2.2.2.1. Let A ∈ sk − Alg and M be an A-module. The A-module
M is strong if the natural morphism
π∗ (A) ⊗π0 (A) π0 (M ) −→ π∗ (M )
is an isomorphism.
Lemma 2.2.2.2. Let A ∈ sk − Alg and M be an A-module.
(1) The A-module M is projective if and only if it is strong and π0 (M ) is a
projective π0 (A)-module.
(2) The A-module M is flat if and only if it is strong and π0 (M ) is a flat π0 (A)-
module.
(3) The A-module M is perfect if and only if it is strong and π0 (M ) is a pro-
jective π0 (A)-module of finite type.
(4) The A-module M is projective and finitely presented if and only if it is
perfect.
Proof. (1) Let us suppose that M is projective.We first notice that a retract
of a strong module A-module is again a strong A-module. This allows us to suppose
that M is free, which clearly implies that M is strong and that π0 (M ) is a free π0 (A)-
module (so in particular projective). Conversely, let M be a strong A-module with
π0 (M ) projective over π0 (A). We write π0 (M ) as a retract of a free π0 -module
i / π0 (A)(I) = ⊕I π0 (A) r / π0 (M ).
π0 (M )
The morphism r is given by a family of elements ri ∈ π0 (M ) for i ∈ I, and therefore
can be seen as a morphism r′ : A(I) −→ M , well defined in Ho(A − M ods ). In the
same way, the projector p = i ◦ r of π0 (A)(I) , can be seen as a projector p′ of A(I) in
the homotopy category Ho(A − M ods ). By construction, this projector gives rise to
a split fibration sequence
K −→ A(I) −→ C,
and the morphism r′ induces a well defined morphism in Ho(A − M ods )
r′ : C −→ M.
By construction, this morphism induces an isomorphisms on π0 , and as C and M are
strong modules, r′ is an isomorphism in Ho(A − M ods ).

(2) Let M be a strong A-module with π0 (M ) flat over π0 (A), and N be any
A-module. Clearly, π∗ (M ) is flat as a π∗ (A)-module. Therefore, the Tor spectral
sequence of [Q2]
T orπ∗ ∗ (A) (π∗ (M ), π∗ (N )) ⇒ π∗ (M ⊗LA N )
degenerates and gives a natural isomorphism
π∗ (M ⊗LA N ) ≃ π∗ (M ) ⊗π∗ (A) π∗ (N ) ≃
≃ (π0 (M ) ⊗π0 (A) π∗ (A)) ⊗π∗ (A) π∗ (N ) ≃ π0 (M ) ⊗π0 (A) π∗ (N ).
As π0 (M ) ⊗π0 (A) − is an exact functor its transform long exact sequences into long
exact sequences. This easily implies that M ⊗LA − preserves homotopy fiber sequences,
and therefore that M is a flat A-module.
Conversely, suppose that M is a flat A-module. Any short exact sequence 0 →
N → P of π0 (A)-modules can also be seen as a homotopy fiber sequence of A-modules,
 2.2.2. FLAT, SMOOTH, ÉTALE AND ZARISKI OPEN MORPHISMS 141

as any morphism N → P is always a fibration. Therefore, we obtain a homotopy fiber

sequence
0 −→ M ⊗LA N −→ M ⊗LA P
which on π0 gives a short exact sequence
0 −→ π0 (M ) ⊗π0 (A) N −→ π0 (M ) ⊗π0 (A) P.
This shows that π0 (M ) ⊗π0 (A) − is an exact functor, and therefore that π0 (M ) is a
flat π0 (A)-module. Furthermore, taking N = 0 we get that for any π0 (A)-module P
one has πi (M ⊗LA P ) = 0 for any i > 0. In other words, we have an isomorphism in
Ho(A − M ods )
M ⊗LA P ≃ π0 (M ) ⊗π0 (A) P.
By shifting P we obtain that for any i ≥ 0 and any π0 (A)-module P we have
M ⊗LA (P [i]) ≃ (π0 (M ) ⊗π0 (A) P )[i].
Passing to Postnikov towers we see that this implies that for any A-module P we have
πi (M ⊗LA P ) ≃ π0 (M ) ⊗π0 (A) πi (P ).
Applying this to P = A we find that M is a strong A-module.

(3) Let us first suppose that M is strong with π0 (M ) projective of finite type over
π0 (A). By point (1) we know that M is a projective A-module. Moreover, the proof
of (1) also shows that M is a retract in Ho(A − M ods ) of some AI with I finite. By
our general result Prop. 1.2.4.2 this implies that A is perfect. Conversely, let M be
a perfect A-module. By (2) and Prop. 1.2.4.2 we know that M is strong and that
π0 (M ) is flat over π0 (A). Furthermore, the unit k of sk − M od is finitely presented,
so by Prop. 1.2.4.2 M is finitely presented object in A − M ods . Using the left Quillen
functor π0 from A − M ods to π0 (A)-modules, we see that this implies that π0 (M ) is
a finitely presented π0 (A)-module, and therefore is projective of finite type.

(4) Follows from (1), (3) and Prop. 1.2.4.2. 

Definition 2.2.2.3. (1) A morphism A −→ B in sk − Alg is strong if the

natural morphism
π∗ (A) ⊗π0 (A) π0 (B) −→ π∗ (B)
is an isomorphism (i.e. B is strong as an A-module).
(2) A morphism A −→ B in sk − Alg is strongly flat (resp. strongly smooth,
resp. strongly étale, resp. a strong Zariski open immersion) if it is strong
and if the morphism of affine schemes
Spec π0 (B) −→ Spec π0 (A)
is flat (resp. smooth, resp. étale, resp. a Zariski open immersion).
We start by a very useful criterion in order to recognize finitely presented mor-
phisms. We have learned this proposition from J. Lurie (see [Lu1]).
Proposition 2.2.2.4. Let f : A −→ B be a morphism in sk − Alg. Then, f is
finitely presented if and only if it satisfies the following two conditions.
(1) The morphism π0 (A) −→ π0 (B) is a finitely presented morphism of commu-
tative rings.
(2) The cotangent complex LB/A ∈ Ho(B − M ods ) is finitely presented.
142 2.2. DERIVED ALGEBRAIC GEOMETRY

Proof. Let us assume first that f is finitely presented. Then (1) and (2) are
easily seen to be true by fact that π0 is left adjoint and by definition of derivations.
Let us now assume that f : A −→ B is a morphism in sk − Alg such that (1) and
(2) are satisfied. Let k ≥ 0 be an integer, and let P (k) be the following property: for
any filtered diagram Ci in A − Algs , such that πj (Ci ) = 0 for all j > k, the natural
morphism
Hocolimi M apA−Algs (B, Ci ) −→ M apA−Algs (B, Hocolimi Ci )
is an isomorphism in Ho(SSet).
We start to prove by induction on k that P (k) holds for all k. For k = 0 this
is hypothesis (1). Suppose P (k − 1) holds, and let Ci be a any filtered diagram in
A − Algs , such that πj (Ci ) = 0 for all j > k. We consider C = Hocolimi Ci , as well
as the k-th Postnikov towers
C −→ C≤k−1 (Ci ) −→ (Ci )≤k−1 .
There is a commutative square of simplicial sets
Hocolimi M apA−Algs (B, Ci ) / Hocolimi M apA−Algs (B, (Ci )k−1 )

 
M apA−Algs (B, C) / M apA−Algs (B, Ck−1 ).

By induction, the morphism on the right is an equivalence. Furthermore, using Prop.

1.4.2.6, Lem. 2.2.1.1 and the fact that the cotangent complex LB/A is finitely pre-
sented, we see that the morphism induced on the homotopy fibers of the horizontal
morphisms is also an equivalence. By the five lemma this implies that the morphism
Hocolimi M apA−Algs (B, Ci ) −→ M apA−Algs (B, C)
is an equivalence. This shows that P (k) is satisfied.
As LB/A is finitely presented, it is a retract of a finite cell B-module (see Prop.
1.2.3.5). In particular, there is an integer k0 > 0, such that [LB/A , M ]B−Mods = 0 for
any B-module M such that πi (M ) = 0 for all i < k0 (one can chose k0 strictly bigger
than the dimension of the cells of a module of which LB/A is a retract). Once again,
Prop. 1.4.2.6 and Lem. 2.2.1.1 implies that for any commutative A-algebra C, the
natural projection C −→ C≤k0 induces a bijection
[B, C]A−Algs ≃ [B, C≤k0 ]A−Algs .
Therefore, as the property P (k0 ) holds, we find that for any filtered system Ci in
A − Algs , the natural morphism
Hocolimi M apA−Algs (B, Ci ) −→ M apA−Algs (B, Hocolimi Ci )
induces an isomorphism in π0 . As this is valid for any filtered system, this shows that
the morphism
Hocolimi M apA−Algs (B, Ci ) −→ M apA−Algs (B, Hocolimi Ci )
induces an isomorphism on all the πi ’s, and therefore is an equivalence. This shows
that f is finitely presented. 
Proposition 2.2.2.5. Let f : A −→ B be a morphism in sk − Alg.
(1) The morphism f is smooth if and only if it is perfect. The morphism f
is formally smooth if it is formally perfect. The morphism f is formally
i-smooth if and only if it is formally smooth.
(2) The morphism f is (formally) unramified if and only if it is (formally) étale.
 2.2.2. FLAT, SMOOTH, ÉTALE AND ZARISKI OPEN MORPHISMS 143

(3) The morphism f is (formally) étale if and only if it is (formally) thh-étale.

(4) A morphism A −→ B in sk − Alg is flat (resp. a Zariski open immersion)
if and only if it is strongly flat (resp. a strong Zariski open immersion).
Proof. (1) The first two assumptions follow from Lem. 2.2.2.2 (4). For the
comparison between formally i-smooth and formally smooth morphism we notice
that a B-module P ∈ B − M od is projective if and only if [P, M [1]] = 0 for any
M ∈ B − M od. This and Prop. 1.2.8.3 imply the statement (note that in our present
context A = sk − Alg and C0 = C).

(2) Follows from the fact that the suspension functor of sk − M od is fully faithful
and from Prop. 1.2.6.5 (1).

(3) By Prop. 1.2.6.5 (2) (formally) thh-étale morphisms are (formally) étale.
Conversely, we need to prove that a formally étale morphism A −→ B is thh-étale.
For this, we use the well known spectral sequence
π∗ (Sym∗ (LB/A [1])) ⇒ π∗ (T HH(B/A))
of [Q2, 8]. Using this spectral sequence we see that B ≃ T HH(B/A) if and only if
LB/A ≃ 0. In particular a formally étale morphism is always formally thh-étale.

(4) For flat morphism this is Lem. 2.2.2.2 (2). Let f : A −→ B be a Zariski open
immersion. By definition, f is a flat morphism, and therefore is strongly flat by what
we have seen. Moreover, for any commutative k-algebra C, considered as an object
C ∈ sk − Alg concentrated in degree 0, we have natural isomorphisms in Ho(SSet)
M apsk−Alg (A, C) ≃ Homk−Alg (π0 (A), C)

M apsk−Alg (B, C) ≃ Homk−Alg (π0 (B), C).

In particular, as f is a epimorphism in sk − Alg the induced morphism of affine
schemes ϕ : Spec π0 (B) −→ Spec π0 (A) is a monomorphism of schemes. This last
morphism is therefore a flat monomorphism of affine schemes. Moreover, as f is
finitely presented so is ϕ, which is therefore a finitely presented flat monomorphism.
By [EGAIV, 2.4.6], a finitely presented flat morphism is open, so ϕ is an open flat
monomorphism and thus a Zariski open immersion. This implies that f is a strong
Zariski open immersion.
Conversely, let f : A −→ B be a strong Zariski open immersion. By (1) it is a flat
morphism. It only remains to show that it is also an epimorphism in sk − Alg and
that it is finitely presented. For the first of these properties, we use the T or spectral
sequence to see that the natural morphism B −→ B ⊗LA B induces an isomorphism
on homotopy groups
π∗ (B ⊗LA B) ≃ π∗ (B) ⊗π∗ (A) π∗ (B) ≃ (π0 (B) ⊗π0 (A) π0 (B)) ⊗π0 (A) π∗ (A) ≃ π∗ (B).

In other words, the natural morphism B ⊗LA B −→ B is an isomorphism in Ho(sk −

Alg), which implies that f is an epinomorphism. It remain to be shown that f is
furthermore finitely presented, but this follows from Prop. 2.2.2.4 and Prop. 1.2.6.5.


Theorem 2.2.2.6. (1) A morphism in sk − Alg is étale if and only if it is

strongly étale.
(2) A morphism in sk − Alg is smooth if and only if it is strongly smooth.
144 2.2. DERIVED ALGEBRAIC GEOMETRY

Proof. We start by two fundamental lemmas.

Recall from §1.2.1 the Quillen adjunction

Q : A − Commnu (C) −→ A − M od A − Commnu (C) ←− A − M od : Z.
Lemma 2.2.2.7. Let A ∈ sk − Alg, and M ∈ A − M ods be a A-module such that
πi (M ) = 0 for all i < k, for some fixed integer k. Then, the adjunction morphism
LQZ(M ) −→ M
induces isomorphisms
πi (LQZ(M )) ≃ πi (M )
for all i ≤ k. In particular
πi (LQZ(M )) ≃ 0 f or i < k πk (LQZ(M )) ≃ πk (M ).
Proof. The non-unital A-algebra Z(M ) being (k − 1)-connected, we can write
it, up to an equivalence, as a CW object in A − Commnu (sk − M od) with cells of
dimension at least k (in the sense of [EKMM]). In other words Z(M ) is equivalent
to a filtered colimit
... / Ai / Ai+1 / ...
where at each step there is a push-out diagram in A − Commnu (sk − M od)
AO i / Ai+1
O

`
k[S ni ]nu / ` k[∆ni +1 ]nu ,

where ni+1 > ni ≥ k − 1, and k[K]nu is the free non-unital commutative simplicial k-
algebra generated by a simplicial set K. Therefore, Q being left Quillen, the A-module
LQZ(M ) is the homotopy colimit of
... / Q(Ai ) / Q(Ai+1 ) / ...

where at each step there exists a homotopy push-out diagram in A − M ods

Q(Ai ) / Q(Ai+1 )
O O

⊕Q(k[S ni ]nu ) / 0.

Computing homotopy groups using long exact sequences, we see that the statement of
the lemma can be reduced to prove that for a free non-unital commutative simplicial
k-algebra k[S n ]nu , we have
πi (Q(k[S n ]nu )) ≃ 0 f or i < n πn (Q(k[S n ]nu )) ≃ k.
But, using that Q is left Quillen we find
Q(k[S n ]nu ) ≃ S n ⊗ k,
which implies the result. 
Lemma 2.2.2.8. Let A be any object in sk − Alg and let us consider the k-th stage
of its Postnikov tower
A≤k −→ A≤k−1 .
There exist natural isomorphisms
πk+1 (LA≤k−1 /A≤k ) ≃ πk (A)
 2.2.2. FLAT, SMOOTH, ÉTALE AND ZARISKI OPEN MORPHISMS 145

πi (LA≤k−1 /A≤k ) ≃ 0 f or i ≤ k.
Proof. This follows from lemma 2.2.1.1, lemma 1.4.3.7, and lemma 2.2.2.7. 
Let us now start the proof of theorem 2.2.2.6.

(1) Let f : A −→ B be a strongly étale morphism. By definition of strongly étale

the morphism f is strongly flat. In particular, the square
A /B

 
π0 (A) / π0 (B)

is homotopy cocartesian in sk − Alg. Therefore, Prop. 1.2.1.6 (2) implies

LB/A ⊗B π0 (B) ≃ Lπ0 (B)/π0 (A) ≃ 0.
Using the Tor spectral sequence
T orπ∗ ∗ (B) (π0 (B), π∗ (LB/A )) ⇒ π∗ (LB/A ⊗B π0 (B)) = 0
we find by induction on k that πk (LB/A ) = 0. This shows that the morphism f
is formally étale. The fact that it is also finitely presented follows then from Prop.
2.2.2.4.
Conversely, let f : A −→ B be an étale morphism. As π0 : sk − Alg −→ k − Alg
is left Quillen, we deduce immediately that π0 (A) −→ π0 (B) is an étale morphism of
commutative rings. We will prove by induction that all the truncations
fk : A≤k −→ B≤k
are strongly étale and thus étale as well by what we have seen before. We thus assume
that fk−1 is strongly étale (and thus also étale by what we have seen in the first part
of the proof). By adjunction, the truncations
fk : A≤k −→ B≤k
are such that
(LB≤k /A≤k )≤k ≃ (LB/A )≤k ≃ 0.
Furthermore, let M be any π0 (B)-module, and d be any morphism in Ho(B≤k −M ods )
d : LB≤k /A≤k −→ M [k + 1].
We consider the commutative diagram in Ho(A≤k /sk − Alg/B≤k )
A≤k / B≤k ⊕d M [k]

 
B≤k / B≤k .

The obstruction of the existence of a morphism

u : B≤k −→ B≤k ⊕d M [k]
in Ho(A≤k /sk − Alg/B≤k ) is precisely d ∈ [LB≤k /A≤k , M [k + 1]]. On the other hand,
by adjunction, the existence of such a morphism f is equivalent to the existence of a
morphism
u′ : B −→ B≤k ⊕d M [k]
in Ho(A/sk−Alg/B≤k ). The obstruction of the existence of u′ itself is d′ ∈ [LB/A , M [k+
1]], the image of d by the natural morphism, which vanishes as A → B is formally
146 2.2. DERIVED ALGEBRAIC GEOMETRY

étale. This implies that u′ and thus u exists, and therefore that for any π0 (B)-module
M we have
[LB≤k /A≤k , M [k + 1]] = 0.
As the object LB≤k /A≤k is already known to be k-connected, we conclude that it is
furthermore is (k + 1)-connected.
Now, there exists a morphism of homotopy cofiber sequences in sk − M od
LA≤k ⊗LA≤k A≤k−1 / LA≤k−1 / LA≤k−1 /A≤k

  
LB≤k ⊗LB≤k B≤k−1 / LB≤k−1 / LB≤k−1 /B≤k .

Base changing the first row by A≤k−1 −→ B≤k−1 gives another morphism of homo-
topy cofiber sequences in sk − M od
LA≤k ⊗LA≤k B≤k−1 / LA≤k−1 ⊗LA
≤k−1
B≤k−1 / LA≤k−1 /A≤k ⊗LA
≤k−1
B≤k−1

  
LB≤k ⊗LB≤k B≤k−1 / LB≤k−1 / LB≤k−1 /B≤k .

Passing to the long exact sequences in homotopy, and using that fk−1 is étale, as well
as the fact that LB≤k /A≤k ⊗LA≤k−1 B≤k−1 is (k + 1)-connected, it is easy to see that
the vertical morphism on the right induces an isomorphism
πk+1 (LA≤k−1 /A≤k ⊗LA≤k−1 B≤k−1 ) ≃ πk+1 (LB≤k−1 /B≤k ).
Therefore, Lemma 2.2.2.8 implies that the natural morphism
πk (A) ⊗π0 (A) π0 (B) −→ πk (B)
is an isomorphism. By induction on k this shows that f is strongly étale.

(2) Let f : A −→ B be a strongly smooth morphism. As the morphism f is

strongly flat, there is a homotopy push-out square in sk − Alg
A /B

 
π0 (A) / π0 (B).

Together with Prop. 1.2.1.6 (2), we thus have a natural isomorphism

LB/A ⊗LB π0 (B) ≃ Lπ0 (B)/π0 (A) .
As the morphism π0 (A) −→ π0 (B) is smooth, we have Lπ0 (B)/π0 (A) ≃ Ω1π0 (B)/π0 (A) [0].
Using that Ω1π0 (B)/π0 (A) is a projective π0 (B)-module, we see that the isomorphism
Ω1π0 (B)/π0 (A) [0] ≃ LB/A ⊗LB π0 (B)
can be lifted to a morphism
P −→ LB/A
in Ho(B − M ods ), where P is a projective B-module such that P ⊗LB π0 (B) ≃
Ω1π0 (B)/π0 (A) [0]. We let K be the homotopy cofiber of this last morphism. By con-
struction, we have
K ⊗LB π0 (B) ≃ 0,
 2.2.2. FLAT, SMOOTH, ÉTALE AND ZARISKI OPEN MORPHISMS 147

which easily implies by induction on k that πk (K) = 0. Therefore, the morphism

P −→ LB/A
is in fact an isomorphism in Ho(B − M ods ), showing that LB/A is a projective B-
module. Moreover, the homotopy cofiber sequence in B − M ods

LA ⊗LA B / LB / LB/A

gives rise to a homotopy cofiber sequence

LB / LB/A / S(LA ⊗L B).

A

But, LB/A being a retract of a free B-module, we see that [LB/A , S(LA ⊗LA B)] is a
retract of a product of π0 (S(LA ⊗LA B)) = 0 and thus is trivial. This implies that the
morphism LB/A −→ S(LA ⊗LA B) is trivial in Ho(B − M ods ), and therefore that the
homotopy cofiber sequence

LA ⊗LA B / LB / LB/A ,

which is also a homotopy fiber sequence, splits. In particular, the morphism LA ⊗LA
B −→ LB has a retraction. We have thus shown that f is a formally smooth mor-
phism. The fact that f is furthermore finitely presented follows from Prop. 2.2.2.4
and the fact that LB/A if finitely presented because Ω1B/A is so (see Lem. 2.2.2.2).
Conversely, let f : A −→ B be a smooth morphism in sk − Alg and let us prove
it is strongly smooth. First of all, using that π0 : sk − Alg −→ k − Alg is left Quillen,
we see that π0 (A) −→ π0 (B) has the required lifting property for being a formally
smooth morphism. Furthermore, it is a finitely presented morphism, so is a smooth
morphism of commutative rings. We then form the homotopy push-out square in
sk − Alg
A /B

 
π0 (A) / C.

By base change, π0 (A) −→ C is a smooth morphism. We will start to prove that the
natural morphism C −→ π0 (C) ≃ π0 (B) is an isomorphism. Suppose it is not, and
let i be the smallest integer i > 0 such that πi (C) 6= 0. Considering the homotopy
cofiber sequence

LC≤i /π0 (A) ⊗LC≤i π0 (C) / Lπ0 (C)/π0 (A) ≃ Ω1π (C)/π (A) [0] / LC≤i /π0 (C) ,
0 0

and using lemma 2.2.2.8, we see that

πi (LC/π0 (A) ⊗LC π0 (C)) ≃ πi (LC≤i /π0 (A) ⊗LC≤i π0 (C)) ≃ πi+1 (LC≤i /π0 (C) ) ≃ πi (C) 6= 0.

But this contradicts the fact that LC/π0 (A) ⊗LC π0 (C) is a projective (and thus strong
by lemma 2.2.2.2 (1)) π0 (C)-module, and thus the fact that π0 (A) −→ C is a smooth
morphism. We therefore have a homotopy push-out diagram in sk − Alg
A /B

 
π0 (A) / π0 (B).
148 2.2. DERIVED ALGEBRAIC GEOMETRY

Using that the bottom horizontal morphism in flat and the Tor spectral sequence, we
get by induction that the natural morphism
πi (A) ⊗π0 (A) π0 (B) −→ πi (B)
is an isomorphism. We thus have seen that f is a strongly smooth morphism. 
An important corollary of theorem 2.2.2.6 is the following topological invariance
of étale morphisms.
Corollary 2.2.2.9. Let A ∈ sk − Alg and t0 (X) = Spec (π0 A) −→ X = Spec A
be the natural morphism. Then, the base change functor
Ho(k − D− Af f /X) −→ Ho(k − D− Af f /t0 (X))
induces an equivalence from the full subcategory of étale morphism Y → X to the full
subcategory of étale morphisms Y ′ → t0 (X). Furthermore, this equivalence preserves
epimorphisms of stacks.
Proof. We consider the Postnikov tower
A / ... / A≤k / A≤k−1 / ... / A≤0 = π0 (A),

and the associated diagram in k − D− Af f

X≤0 = t0 (X) / X≤1 / ... / X≤k−1 / X≤k / ... / X.

We define a model category k − D− Af f /X≤∗ , whose objects are families of objects

Yk −→ X≤k in k − D− Af f /X≤k , together with transitions morphisms Yk−1 −→
Yk ×X≤k X≤k−1 in k − D− Af f /X≤k−1 . The morphisms in k − D− Af f /X≤∗ are
simply the families of morphisms Yk −→ Yk′ in k − D− Af f /X≤k that commute with
the transition morphisms. The model structure on k − D− Af f /X≤∗ is such that a
morphism f : Y∗ −→ Y∗′ in k − D− Af f /X≤∗ is an equivalence (resp. a fibration) if
all morphisms Yk −→ Yk′ are equivalences (resp. fibrations) in k − D− Af f .
There exists a Quillen adjunction
G : k−D− Af f /X≤∗ −→ k−D− Af f /X k−D− Af f /X≤∗ ←− k−D− Af f /X : F,
defined by G(Y∗ ) to be the colimit in k − D− Af f /X of the diagram
Y0 / Y1 / . . . Yk / ... / X.
The right adjoint F sends a morphism Y −→ X to the various pullbacks
F (Y )k := Y ×X X≤k −→ Xk
together with the obvious transition isomorphisms. Using Thm. 2.2.2.6 (1) it is not
hard to check that the derived adjunction
LG : Ho(k − D− Af f /X≤∗) −→ Ho(k − D− Af f /X)
Ho(k − D− Af f /X≤∗ ) ←− Ho(k − D− Af f /X) : RF,
induces an equivalence from the full subcategory of Ho(k − D− Af f /X) consisting of
étale morphisms Y −→ X, and the full subcategory of Ho(k−D− Af f /X≤∗) consisting
of objects Y∗ such that each Yk −→ X≤k is étale and each transition morphism
Yk−1 −→ Yk ×hX≤k X≤k−1
is an isomorphism in Ho(k − D− Af f /X≤k−1 ). Let us denote these two categories
respectively by Ho(k − D− Af f /X)et and Ho(k − D− Af f /X≤∗)cart,et . Using Cor.
1.4.3.12 and Lem. 2.2.1.1, we know that each base change functor
Ho(k − D− Af f /X≤k ) −→ Ho(k − D− Af f /X≤k−1 )
 2.2.2. FLAT, SMOOTH, ÉTALE AND ZARISKI OPEN MORPHISMS 149

induces an equivalence on the full sub-categories of étale morphisms. This easily

implies that the natural projection functor
Ho(k − D− Af f /X≤∗ )cart,et −→ Ho(k − D− Af f /X≤0 )et
is an equivalence of categories. Therefore, by composition, we find that the base
change functor
Ho(k − D− Af f /X)et −→ Ho(k − D− Af f /X≤∗ )cart,et −→ Ho(k − D− Af f /X≤0 )et
is an equivalence of categories.
Finally, the statement concerning epimorphism of stacks is obvious, as a flat
morphism Y −→ X in k − D− Af f induces an epimorphism of stacks if and only if
t0 (Y ) −→ t0 (X) is a surjective morphism of affine schemes. 

A direct specialization of 2.2.2.9 is the following.

Corollary 2.2.2.10. Let A ∈ sk −Alg and t0 (X) = Spec (π0 A) −→ X = Spec A
be the natural morphism. Then, the base change functor
Ho(k − D− Af f /X) −→ Ho(k − D− Af f /t0 (X))
induces an equivalence from full sub-categories of Zariski open immersions Y → X
to the full subcategory of Zariski open immersions Y ′ → t0 (X). Furthermore, this
equivalence preserves epimorphisms of stacks.
Proof. Indeed, using 2.2.2.9 it is enough to see that an étale morphism A −→ B
is a Zariski open immersion if and only if π0 (A) −→ π0 (B) is. But this true by Thm.
2.2.2.6 and Prop. 2.2.2.5. 

From the proof of Thm. 2.2.2.6 we also extract the following more precise result.
Corollary 2.2.2.11. Let f : A −→ B be a morphism in sk − Alg. The following
are equivalent.
(1) The morphism f is smooth (resp. étale).
(2) The morphism f is flat and π0 (A) −→ π0 (B) is smooth (resp. étale).
(3) The morphism f is formally smooth (resp. formally étale) and π0 (B) is a
finitely presented π0 (A)-algebra.
We are now ready to define the étale model topology (Def. 1.3.1.1) on k−D− Af f .
Definition 2.2.2.12. A family of morphisms {Spec Ai −→ Spec A}i∈I in k −
D− Af f is an étale covering family (or simply ét-covering family) if it satisfies the
following two conditions.
(1) Each morphism A −→ Ai is étale.
(2) There exists a finite sub-set J ⊂ I such that the family {A −→ Ai }i∈J is a
formal covering family in the sense of 1.2.5.1.
Using that étale morphisms are precisely the strongly étale morphisms (see Corol-
lary 2.2.2.11) we immediately deduce that a family of morphisms {Spec Ai −→
Spec A}i∈I in k − D− Af f is an ét-covering family if and only if there exists a fi-
nite sub-set J ⊂ I satisfying the following two conditions.
• For all i ∈ I, the natural morphism
π∗ (A) ⊗π0 (A) π0 (Ai ) −→ π∗ (Ai )
is an isomorphism.
150 2.2. DERIVED ALGEBRAIC GEOMETRY

• The morphism of affine schemes

a
Spec π0 (Ai ) −→ Spec π0 (A)
i∈J

is étale and surjective.

Lemma 2.2.2.13. The ét-covering families define a model topology (Def. 1.3.1.1)
on k − D− Af f , which satisfies assumption 1.3.2.2.
Proof. That ét-covering families defines a model topology simply follows from
the general properties of étale morphisms and formal coverings described in propo-
sitions 1.2.5.2 and 1.2.6.3. It only remain to show that the étale topology satisfies
assumption 1.3.2.2.
The étale topology is quasi-compact by definition, so (1) of 1.3.2.2 is satisfied. In
the same way, property (2) of 1.3.2.2 is obviously satisfied according to the explicit
definition of étale coverings given above. Finally, let us check property (3) of 1.3.2.2.
For this, let A −→ B∗ be a co-simplicial object in sk − Alg, corresponding to a
representable étale-hypercover in k − D− Af f in the sense of 1.3.2.2 (3). We consider
the adjunction
B∗ ⊗LA − : Ho(A − M ods ) −→ Ho(csB∗ − M od)
Z
Ho(A − M ods ) ←− Ho(csB∗ − M od) :

defined in §1.2. We restrict it to the full subcategory Ho(csB∗ −M od)cart of Ho(csB∗ −

M od) consisting of cartesian objects
B∗ ⊗LA − : Ho(A − M ods ) −→ Ho(csB∗ − M od)cart
Z
Ho(A − M ods ) ←− Ho(csB∗ − M od)cart :
and we need to prove that this is an equivalence. By definition of formal coverings,
the base change functor
B∗ ⊗LA − : Ho(A − M ods ) −→ Ho(csB∗ − M od)cart
is clearly conservative, so it only remains to show that the adjunction map
Z
Id −→ ◦(B∗ ⊗LA −)

is an isomorphism.
For this, let E∗ ∈ Ho(csB∗ − M od)cart , and let us consider the adjunction mor-
phism
E1 −→ (Holimn En ) ⊗LA B1 .
We need to show that this morphism is an isomorphism in Ho(B1 − M ods ). For this,
we first use that A −→ B1 is a strongly flat morphism, and thus
π∗ ((Holimn En ) ⊗LA B1 ) ≃ π∗ ((Holimn En )) ⊗π0 (A) π0 (B1 ).
We then apply the spectral sequence
H p (T ot(πq (E∗ ))) ⇒ πq−p ((Holimn En ).
The object πq (E∗ ) is now a co-simplicial module over the co-simplicial commutative
ring π0 (B∗ ), which is furthermore cartesian as all the coface morphisms Bn −→ Bn+1
are flat. The morphism π0 (A) −→ π0 (B∗ ) being an étale, and thus faithfully flat,
hypercover in the usual sense, we find by the usual flat descent for quasi-coherent
 2.2.3. THE HAG CONTEXT: GEOMETRIC D− -STACKS 151

sheaves that H p (T ot(πq (E∗ ))) ≃ 0 for p 6= 0. Therefore, the above spectral sequence
degenerates and gives an isomorphism
πp ((Holimn En ) ≃ Ker (πp (E0 ) ⇉ πp (E1 )) .
In other words, πp ((Holimn En )) is the π0 (A)-module obtained by descent from
πp (E∗ ) on πp (B∗ ). In particular, the natural morphism
πp ((Holimn En ) ⊗π0 (A) π0 (B1 ) −→ πp (E1 )
is an isomorphism. Putting all of this together we find that
E1 −→ (Holimn En ) ⊗LA B1
is an isomorphism in Ho(B1 − M ods ). 
We have now the model site (Def. 1.3.1.1) (k − D− Af f, ét), with the étale model
topology, and we make the following definition.
Definition 2.2.2.14. (1) A D− -stack is an object F ∈ k − D− Af f ∼,ét
which is a stack in the sense of Def. 1.3.2.1.
(2) The model category of D− -stacks is k − D− Af f ∼,ét , and its homotopy
category will be simply denoted by D− St(k).
The following result is a corollary of Proposition 2.2.2.4. It states that the prop-
erty of being finitely presented is local for the étale topology defined above.
Corollary 2.2.2.15. Let f : A −→ B be a morphism in sk − Alg.
(1) If there exists an étale covering B −→ B ′ such that A −→ B ′ is finitely
presented, then f is finitely presented.
(2) If there exists an étale covering A −→ A′ , such that
A′ −→ A′ ⊗LA B
is finitely presented, then f is finitely presented.
Proof. This follows from proposition 2.2.2.4. Indeed, it suffices to prove that
both conditions of proposition 2.2.2.4 have the required local property. The first one
is well known, and the second one is a consequence of corollary 1.3.7.8, as finitely
presented objects in Ho(Sp(B − M ods )) are precisely the perfect objects. 

2.2.3. The HAG context: Geometric D− -stacks

We now let P be the class of smooth morphisms in sk − Alg.
Lemma 2.2.3.1. The class P of smooth morphisms and the étale model topology
satisfy assumption 1.3.2.11.
Proof. As étale morphisms are also smooth we see that assumption 1.3.2.11
(1) is satisfied. Assumption 1.3.2.11 (2) is satisfied as smooth morphisms are stable
by homotopy pullbacks, compositions and equivalences. Let us prove that smooth
morphisms satisfy 1.3.2.11 (3) for the étale model topology.
Let X −→ Y be a morphism in k − D− Af f , and {Ui −→ X} a finite étale
covering family such that each Ui −→ X and Ui −→ Y is smooth. First of all, using
that {Ui −→ X} is a flat formal covering, and that each morphism Ui −→ Y is flat,
we see that the morphism X −→ Y is flat. Using our results Prop. 2.2.2.5 and
Thm. 2.2.2.6 it only remain to show that the morphism t0 (X) −→ t0 (Y ) is a smooth
morphism between affine schemes. But, passing to t0 we find a smooth covering
family {t0 (Ui ) −→ t0 (X)}, such that each morphism t0 (Ui ) −→ t0 (Y ) is smooth, and
152 2.2. DERIVED ALGEBRAIC GEOMETRY

we know (see e.g. [EGAIV]) that this implies that t0 (X) −→ t0 (Y ) is a smooth
morphism of affine schemes.
Finally, property (4) of 1.3.2.11 is obvious. 
We have verified our assumptions 1.3.2.2 and 1.3.2.11 for the étale model topology
and P the class of smooth morphisms. We thus have that
(sk − M od, sk − M od, sk − Alg, ét, P)
is a HAG context in the sense of Def. 1.3.2.13. We can therefore apply our general
definitions to obtain a notion of n-geometric D− -stacks in D− St(k), as well as the
notion of n-smooth morphisms. We then check that Artin’s conditions of Def. 1.4.3.1
are satisfied.
Proposition 2.2.3.2. The étale model topology and the smooth morphisms satisfy
Artin’s conditions relative to the HA context (sk − M od, sk − M od, sk − Alg) in the
sense of Def. 1.4.3.1.
Proof. We will show that the class E of étale morphisms satisfies conditions (1)
to (5) of Def. 1.4.3.1.

(1) is clear as P is exactly the class of all i-smooth morphisms.

(2) and (3) are clear by the choice of E and A.

To prove (4), let p : Y −→ X be a smooth morphism in k − D− Af f , and let us

consider the smooth morphism of affine schemes t0 (p) : t0 (Y ) −→ t0 (X). As p is a
smooth epimorphism of stacks, t0 (p) is a smooth and surjective morphism of affine
schemes. It is known (see e.g. [EGAIV]) that there exists a étale covering of affine
schemes X0′ −→ t0 (X) and a commutative diagram
t0 (Y )
x x<
xx
xxx
x 
X0′ / t0 (X).

By Cor. 2.2.2.9, we know that there exists an étale covering X ′ −→ X in k − D− Af f ,

inducing X0′ −→ t0 (X). Taking the homotopy pullback
Y ×hX X ′ −→ X ′
we can replace X by X ′ , and therefore assume that the morphism t0 (p) : t0 (Y ) −→
t0 (X) has a section. We are going to show that this section can be extended to a
section of p, which will be enough to prove what we want.
We use the same trick as in the proof of Cor. 2.2.2.9, and consider once again the
model category k − D− Af f /X≤∗ . Using Thm. 2.2.2.6 (2) we see that the functor
Ho(k − D− Af f /X) −→ Ho(k − D− Af f /X≤∗)
induces an equivalences from the full subcategory Ho(k − D− Af f /X)sm of smooth
morphisms Z → X to the full subcategory Ho(k − D− Af f /X≤∗ )cart,sm consisting of
objects Z∗ such that each Zk −→ X≤k is smooth, and each morphism
Zk−1 −→ Zk ×hZ≤k Z≤k−1
is an isomorphism in Ho(k − D− Af f ). From this we deduce that the space of sections
of p can be described in the following way
M apk−D− Af f /X (X, Y ) ≃ Holimk M apk−D− Af f /X≤k (X≤k , Yk ),
 2.2.3. THE HAG CONTEXT: GEOMETRIC D− -STACKS 153

where Yk ≃ Y≤k ≃ Y ×hX X≤k . But, as there always exists a surjection

π0 (Holimk M apk−D− Af f /X≤k (X≤k , Yk )) −→ Limk π0 (M apk−D− Af f /X≤k (X≤k , Yk )),
we only need to check that Limk π0 (M apk−D− Af f /X≤k (X≤k , Yk )) 6= ∅. For this, it is
enough to prove that for any k ≥ 1, the restriction map
π0 (M apk−D− Af f /X≤k (X≤k , Yk )) −→ π0 (M apk−D− Af f /X≤k−1 (X≤k , Yk−1 ))
is surjective. In other words, we need to prove that a morphism in Ho(k−D− Af f /X≤k )
X≤k−1 / Yk


X≤k

can be filled up to a commutative diagram in Ho(k − D− Af f /X≤k )

X≤k−1 / Yk
xx<
xx
xxx
 xx
X≤k .
Using Prop. 1.4.2.6 and Lem. 2.2.1.1, we see that the obstruction for the existence
of this lift lives in the group
[LBk /A≤k , πk (A)[k + 1]]A≤k −Mods ,
where Yk = Spec Bk and X = Spec A. But, as Yk −→ X≤k is a smooth morphism,
the A≤k -module LBk /A≤k is projective, and therefore
[LBk /A≤k , πk (A)[k + 1]]A≤k −Mods ≃ 0,
by Lem. 2.2.2.2 (5).

It remains to prove that property (5) of Def. 1.4.3.1 is satisfied. Let X −→

Xd [ΩM ] be as in the statement of Def. 1.4.3.1 (4), with X = Spec A and some
connected A-module M ∈ A − M ods , and let U −→ Xd [ΩM ] be an étale morphism.
We know by Lem. 1.4.3.8 (or rather its proof) that this morphism is of the form
Xd′ ′ [ΩM ′ ] −→ Xd [ΩM ]
for some étale morphism X ′ = Spec A′ −→ X = Spec A, and furthermore X ′ −→ X
is equivalent to the homotopy pullback
U ×Xd [ΩM] X −→ X.
Therefore, using Cor. 2.2.2.11, it is enough to prove that
t0 (Xd′ ′ [ΩM ′ ]) −→ t0 (Xd [ΩM ])
is a surjective morphism of affine schemes if and only if
t0 (X ′ ) −→ t0 (X)
is so. But, M being connected, t0 (Xd [ΩM ])red ≃ t0 (X), and in the same way
t0 (Xd′ ′ [ΩM ′ ])red ≃ t0 (X ′ ), this gives the result, since being a surjective morphism
is topologically invariant. 
Corollary 2.2.3.3. (1) Any n-geometric D− -stack has an obstruction the-
ory. In particular, any n-geometric D− -stack has a cotangent complex.
154 2.2. DERIVED ALGEBRAIC GEOMETRY

(2) Any n-representable morphism of D− -stacks f : F −→ G has a relative

obstruction theory. In particular, any n-representable morphism of D− -
stacks has a relative cotangent complex.
Proof. Follows from Prop. 2.2.3.2 and theorem 1.4.3.2. 

We finish by some properties of morphisms, as in Def. 1.3.6.2.

Lemma 2.2.3.4. Let Q be one the following class of morphisms in k − D− Af f .
(1) Flat.
(2) Smooth.
(3) Etale.
(4) Finitely presented.
Then, morphisms in Q are compatible with the étale topology and the class P of
smooth morphisms in the sense of Def. 1.3.6.1.
Proof. Using the explicit description of flat, smooth and étale morphisms given
in Prop. 2.2.2.5 (1), (2), and (3) reduce to the analog well known facts for morphism
between affine schemes. Finally, Cor. 2.2.2.15 implies that the class of finitely pre-
sented morphisms in k − D− Af f is compatible with the étale topology and the class
P. 

Lemma 2.2.3.4 and definition 1.3.6.2 allows us to define the notions of flat, smooth,
étale and locally finitely presented morphisms of D− -stacks, which are all stable by
equivalences, compositions and homotopy pullbacks. Recall that by definition, a flat
(resp. smooth, resp. étale, resp. locally finitely presented) is always n-representable
for some n. Using our general definition Def. 1.3.6.4 we also have notions of quasi-
compact morphisms, finitely presented morphisms, and monomorphisms between D− -
stacks. We also make the following definition.
Definition 2.2.3.5. (1) A morphism of D− -stacks is a Zariski open im-
mersion if it is a locally finitely presented flat monomorphism.
(2) A morphism of D− -stacks F −→ G is a closed immersion if it is repre-
sentable, and if for any A ∈ sk−Alg and any morphism X = RSpec A −→ G
the induced morphism of representable D− -stacks
F ×hG X ≃ RSpec B −→ RSpec A
induces an epimorphism of rings π0 (A) −→ π0 (B).

2.2.4. Truncations
We consider the natural inclusion functor
i : k − Af f −→ k − D− Af f
right adjoint to the functor
π0 : k − D− Af f −→ k − Af f.
The pair (π0 , i) is a Quillen adjunction (for the trivial model structure on k − Af f ),
and as usual we will omit to mention the inclusion functor i, and simply consider
commutative k-algebras as constant simplicial objects. Furthermore, both functors
preserve equivalences and thus induce a Quillen adjunction on the model category of
pre-stacks (using notations of [HAGI, §4.8])
i! : k − Af f ∧ = SP r(k − Af f ) −→ k − D− Af f ∧ k − Af f ∧ ←− k − D− Af f ∧ : i∗ .
 2.2.4. TRUNCATIONS 155

The functor i is furthermore continuous in the sense of [HAGI, §4.8], meaning that
the right derived functor
Ri∗ : Ho(k − D− Af f ∧ ) −→ Ho(k − Af f ∧ )
preserves the sub-categories of stacks. Indeed, by Lem. 1.3.2.3 (2) and adjunction this
follows from the fact that i : k − Af f −→ k − D− Af f preserves co-products, equiva-
lences and étale hypercovers. By the general properties of left Bousfield localizations
we therefore get a Quillen adjunction on the model categories of stacks
i! : k − Af f ∼,ét −→ k − D− Af f ∼,ét k − Af f ∼,ét ←− k − D− Af f ∼,ét : i∗ .
From this we get a derived adjunction on the homotopy categories of stacks
Li! : St(k) −→ D− St(k)
St(k) ←− D− St(k) : Ri∗ .
Lemma 2.2.4.1. The functor Li! is fully faithful.
Proof. We need to show that for any F ∈ St(k) the adjunction morphism
F −→ Ri∗ ◦ Li! (F )
is an isomorphism. The functor Ri∗ commutes with homotopy colimits, as these are
computed in the model category of simplicial presheaves and thus levelwise. Moreover,
as i! is left Quillen the functor Li! also commutes with homotopy colimits. Now, any
stack F ∈ St(k) is a homotopy colimit of representable stacks (i.e. affine schemes), and
therefore we can suppose that F = Spec A, for A ∈ k − Alg. But then Li! (Spec A) ≃
RSpec A. Furthermore, for any B ∈ k − Alg there are natural isomorphisms in
Ho(SSet)
RSpec A(B) ≃ M apsk−Alg (A, B) ≃ Homk−Alg (A, B) ≃ (Spec A)(B).
This shows that the adjunction morphism
Spec A −→ Ri∗ ◦ Li! (Spec A)
is an isomorphism. 
Another useful remark is the following
Lemma 2.2.4.2. The functor i∗ : k − D− Af f ∼,ét −→ k − Af f ∼,ét is right and
left Quillen. In particular it preserves equivalences.
Proof. The functor i∗ has a right adjoint
π0∗ : k − Af f ∼,ét −→ k − D− Af f ∼,ét .
Using lemma 1.3.2.3 (2) we see that π0∗ is a right Quillen functor. Therefore i∗ is left
Quillen. 
Definition 2.2.4.3. (1) The truncation functor is
t0 := i∗ : D− St(k) −→ St(k).
(2) The extension functor is the left adjoint to t0
i := Li! : St(k) −→ D− St(k).
(3) A D− -stack F is truncated if the adjunction morphism
it0 (F ) −→ F
is an isomorphism in D− St(k).
156 2.2. DERIVED ALGEBRAIC GEOMETRY

By lemmas 2.2.4.1, 2.2.4.2 we know that the truncation functor t0 commutes with
homotopy limits and homotopy colimits. The extension functor i itself commutes with
homotopy colimits and is fully faithful. An important remark is that the extension i
does not commute with homotopy limits, as the inclusion functor k −Alg −→ sk −Alg
does not preserve homotopy push-outs.
Concretely, the truncation functor t0 sends a functor
F : sk − Alg −→ SSetV
to
t0 (F ) : k − Alg −→ SSetV
A 7→ F (A).
By adjunction we clearly have
t0 RSpec A ≃ Spec π0 (A)
for any A ∈ sk − Alg, showing that the notation is compatible with the one we did use
before for t0 (Spec A) = Spec π0 (A) as objects in k − D− Af f . The extension functor
i is characterized by
i(Spec A) ≃ RSpec A,
and the fact that it commutes with homotopy colimits.
Proposition 2.2.4.4. (1) The functor t0 preserves epimorphisms of stacks.
(2) The functor t0 sends n-geometric D− -stacks to n-geometric stacks, and flat
(resp. smooth, resp. étale) morphisms between D− -stacks to flat (resp.
smooth, resp. étale) morphisms between stacks.
(3) The functor i preserves homotopy pullbacks of n-geometric stacks along a
flat morphism, sends n-geometric stacks to n-geometric D− -stacks, and flat
(resp. smooth, resp. étale) morphisms between n-geometric stacks to flat
(resp. smooth, resp. étale) morphisms between n−-geometric D− -stacks.
(4) Let F ∈ St(k) be an n-geometric stack, and F ′ −→ i(F ) be a flat morphism
of n-geometric D− -stacks. Then F ′ is truncated (and therefore is the image
by i of an n-geometric stack by (2)).
Proof. (1) By adjunction, this follows from the fact that i : k −Alg −→ sk −Alg
reflects étale covering families (by Prop. 2.2.2.2).

(2) The proof is by induction on n. For n = −1, this simply follows from the
formula
t0 RSpec A ≃ Spec π0 (A),
and Prop. 2.2.2.5 and Thm. 2.2.2.6. Assume that the property is known for any
m < n and let us prove it for n. Let F be an n-geometric D− -stack, which by
Prop. 1.3.4.2 can be written as |X∗ | for some (n − 1)-smooth Segal groupoid X∗ in
k−D− Af f ∼,ét . Using our property at rank n−1 and that t0 commutes with homotopy
limits shows that t0 (X∗ ) is also a (n−1)-smooth Segal groupoid object in k −Af f ∼,ét .
Moreover, as t0 commutes with homotopy colimits we have t0 (F ) ≃ |t0 (X∗ )|, which
by Prop. 1.3.4.2 is an n-geometric stack. This shows that t0 sends n-geometric D− -
stacks to n-geometric stacks and therefore preserves n-representable morphisms. Let
f : F −→ G be a flat (resp. smooth, resp. étale) morphism of D− -stacks which is
n-representable, and let us prove by induction on n that t0 (f ) is flat (resp. smooth,
resp. étale). We let X ∈ St(k) be an affine scheme, and X −→ t0 (G) be a morphism
of stacks. By adjunction between i and t0 we have
X ×ht0 (G) t0 (F ) ≃ t0 (X ×hG F ),
 2.2.4. TRUNCATIONS 157

showing that we can assume that G = X is an affine scheme and thus F to be an

n-geometric D− . By definition of being flat (resp. smooth, resp. étale) there ex-
ists a smooth n-atlas {Ui } of F such that each composite morphism Ui −→ X is
(n − 1)-representable and flat (resp. smooth, resp. étale). By induction and (1)
the family {t0 (Ui )} is a smooth n-atlas for t0 (F ), and by induction each composi-
tion t0 (Ui ) −→ t0 (X) = X is flat (resp. smooth, resp. étale). This implies that
t0 (F ) −→ t0 (X) is flat (resp. smooth, resp. étale).

(3) The proof is by induction on n. For n = −1 this follows from the formula
i(Spec A) ≃ RSpec A, the description of flat, smooth and étale morphisms (Prop.
2.2.2.5 and Thm. 2.2.2.6), and the fact that for any flat morphism of commutative
k-algebras A −→ B and any commutative A-algebra C there is a natural isomorphism
in Ho(A − Algs )
B ⊗A C ≃ B ⊗LA C.
Let us now assume the property is proved for m < n and let us prove it for n.
Let F be an n-geometric stack, and by Prop. 1.3.4.2 let us write it as F ≃ |X∗ |
for some (n − 1)-smooth Segal groupoid object X∗ in k − Af f ∼,ét . By induction,
i(X∗ ) is again a (n − 1)-smooth Segal groupoid objects in k − D− Af f ∼,ét , and as i
commutes with homotopy colimits we have i(F ) ≃ |i(X∗ )|. Another application of
Prop. 1.3.4.2 shows that F is an n-geometric D− -stacks. We thus have seen that i
sends n-geometric stacks to n-geometric D− -stacks.
Now, let F −→ G be a flat morphism between n-geometric stacks, and H −→
G any morphism between n-geometric stacks. We want to show that the natural
morphism
i(F ×hG H) −→ i(F ) ×hi(G) i(H)
is an isomorphism in St(k). For this, we write G ≃ |X∗ | for some (n − 1)-smooth
Segal groupoid object in k − Af f ∼,ét , and we consider the Segal groupoid objects
F∗ := F ×hG X∗ H∗ := H ×hG X∗ ,
where X∗ −→ |X∗ | = G is the natural augmentation in St(k). The Segal groupoid
objects F∗ and H∗ are again (n − 1)-smooth Segal groupoid objects in k − Af f ∼,ét
as G is an n-geometric stack. The natural morphisms of Segal groupoid objects
F∗ −→ X∗ ←− H∗
gives rise to another (n − 1)-smooth Segal groupoid object F∗ ×hX∗ H∗ . Clearly, we
have
F ×hG H ≃ |F∗ ×hX∗ H∗ |.
As i commutes with homotopy colimits, and by induction on n we have
i(F ×hG H) ≃ |i(F∗ ) ×hi(X∗ ) i(H∗ )| ≃ |i(F∗ )| ×h|i(X∗ )| |i(H∗ )| ≃ i(F ) ×hi(G) i(H).
It remains to prove that if f : F −→ G is a flat (resp. smooth, resp. étale) morphism
between n-geometric stacks then i(f ) : i(F ) −→ i(G) is a flat (resp. smooth, resp.
étale) morphism between n-geometric D− -stacks. For this, let {Ui } be a smooth n-
atlas for G. We have seen before that {i(Ui )} is a smooth n-atlas for i(G). As we have
seen that i commutes with homotopy pullbacks along flat morphisms, and because of
the local properties of flat (resp. smooth, resp. étale) morphisms (see Prop. 1.3.6.3
and Lem. 2.2.3.4), we can suppose that G is one of the Ui ’s and thus is an affine
scheme. Now, let {Vi } be a smooth n-atlas for F . The family {i(Vi )} is a smooth
n-atlas for F , and furthermore each morphism i(Vi ) −→ i(G) is the image by i of a
flat (resp. smooth, resp. étale) morphism between affine schemes and therefore is a
flat (resp. smooth, resp. étale morphism) of D− -stacks. By definition this implies
158 2.2. DERIVED ALGEBRAIC GEOMETRY

that i(F ) −→ i(G) is flat (resp. smooth, resp. étale).

(4) The proof is by induction on n. For n = −1 this is simply the description of

flat morphisms of Prop. 2.2.2.5. Let us assume the property is proved for m < n and
let us prove it for n. Let F ′ −→ i(F ) be a flat morphism, with F an n-geometric stack
and F ′ an n-geometric D− -stack. Let {Ui } be a smooth n-atlas for F . Then, {i(Ui )}
is a smooth n-atlas for i(F ). We consider the commutative diagram of D− -stacks
f
it0 (F ′ ) / F′
HH
HH
HH
HH
H$ 
i(F ).
We need to prove that f is an isomorphism in D− St(k). As this is a local property
on i(F ), we can take the homotopy pullback over the atlas {i(Ui )}, and thus suppose
that F is an affine scheme. Let now {Vi } be a smooth n-atlas for F ′ , and we consider
the homotopy nerve of the morphism
a
X0 := Vi −→ F ′ .
i

This is a (n−1)-smooth Segal groupoid objects in D− Af f ∼,ét , which is such that each
morphism Xi −→ i(G) is flat. Therefore, by induction on n, the natural morphism
of Segal groupoid objects
it0 (X∗ ) −→ X∗
is an equivalence. Therefore, as i and t0 commutes with homotopy colimits we find
that the adjunction morphism
it0 (F ′ ) ≃ |it0 (X∗ )| −→ |X∗ | ≃ F ′
is an isomorphism in D− St(k). 
An important corollary of Prop. 2.2.2.5 is the following fact.
Corollary 2.2.4.5. For any Artin n-stack, the D− -stack i(F ) has an obstruction
theory.
Proof. Follows from 2.2.2.5 (3) and Cor. 2.2.3.3. 
One also deduces from Prop. 2.2.2.5 and Lem. 2.1.1.2 the following corollary.
Corollary 2.2.4.6. Let F be an n-geometric D− -stack. Then, for any A ∈
sk − Alg, such that πi (A) = 0 for i > k, the simplicial set RF (A) is (n + k + 1)-
truncated.
Proof. This is by induction on k. For k = 0 this is Lem. 2.1.1.2 and the fact
that t0 preserves n-geometric stacks. To pass from k to k + 1, we consider for any
A ∈ sk − Alg with πi (A) = 0 for i > k + 1, the natural morphisms
RF (A) −→ RF (A≤k ),
whose homotopy fibers can be described using Prop. 1.4.2.5 and Lem. 2.2.1.1. We
find that this homotopy fiber is either empty, or equivalent to
M apA≤k −Mod (LF,x , πk+1 (A)[k + 1]),
which is (k + 1)-truncated. By induction, RF (A≤k ) is (k + 1 + n)-truncated and the
homotopy fibers of
RF (A) −→ RF (A≤k ),
 2.2.5. INFINITESIMAL CRITERIA FOR SMOOTH AND ÉTALE MORPHISMS 159

are (k + 1)-truncated, and therefore RF (A) is (k + n + 2)-truncated. 

Another important property of the truncation functor is the following local de-
scription of the truncation t0 (F ) sitting inside the D− -stack F itself.
Proposition 2.2.4.7. Let F be an n-geometric D− -stack. The adjunction mor-
phism it0 (F ) −→ F is a representable morphism. Moreover, for any A ∈ sk − Alg,
and any flat morphism RSpec A −→ F , the square

it0 (F ) /F
O O

RSpec π0 (A) / RSpec A

is homotopy cartesian. In particular, the morphism it0 (F ) −→ F is a closed immer-

sion in the sense of Def. 2.2.3.5.
Proof. By the local character of representable morphisms it is enough to prove
that for any flat morphism X = RSpec A −→ F , the square

it0 (F ) /F
O O

it0 (X) = RSpec π0 (A) /X

is homotopy cartesian. The morphism

it0 (F ) ×hF X −→ it0 (F )
is flat, and by Prop. 2.2.2.5 (4) this implies that the D− -stack it0 (F ) ×hF X is trun-
cated. In other words the natural morphism
it0 (it0 (F ) ×hF X) ≃ it0 (F ) ×hit0 (F ) it0 (X) ≃ it0 (X) −→ it0 (F ) ×hF X
is an isomorphism. 

Using our embedding

i : St(k) −→ D− St(k)
we will see stacks in St(k), and in particular Artin n-stacks, as D− -stacks. However,
as the functor i does not commute with homotopy pullbacks we will still mention it
in order to avoid confusions.

2.2.5. Infinitesimal criteria for smooth and étale morphisms

Recall that sk − M od1 denotes the full subcategory of sk − M od of connected
simplicial k-modules. It consists of all M ∈ sk − M od for which the adjunction
S(ΩM ) −→ M is an isomorphism in Ho(sk−M od), or equivalently for which π0 (M ) =
0.
Proposition 2.2.5.1. Let f : F −→ G be an n-representable morphism between
D− -stacks. The morphism f is smooth if and only if it satisfies the following two
conditions
(1) The morphism t0 (f ) : t0 (F ) −→ t0 (G) is a locally finitely presented mor-
phism in k − Af f ∼,ét .
160 2.2. DERIVED ALGEBRAIC GEOMETRY

(2) For any A ∈ sk−Alg, any connected M ∈ A−M ods , and any derivation d ∈
π0 (Der(A, M )), the natural projection A ⊕d ΩM −→ A induces a surjective
morphism
 
π0 (RF (A ⊕d ΩM )) −→ π0 RG(A ⊕d ΩM ) ×hRF (A) RG(A) .

Proof. First of all we can suppose that F and G are fibrant objects in k −
D− Af f ∼,ét .
Suppose first that the morphism f is smooth and let us prove that is satisfies
the two conditions of the proposition. We know by 2.2.4.4 (3) that t0 (f ) is then a
smooth morphism in k − Af f ∼,ét , so condition (1) is satisfied. The proof that (2) is
also satisfied goes by induction on n. Let us start with the case n = −1, and in other
words when f is a smoothand representable morphism. 
We fix a point x in π0 G(A ⊕d ΩM ) ×hF (A) G(A) , and we need to show that the
homotopy fiber taken at x of the morphism
F (A ⊕d ΩM ) −→ G(A ⊕d ΩM ) ×hF (A) G(A)

is non empty. The point x corresponds via Yoneda to a commutative diagram in

Ho(k − D− Af f ∼,ét /G)
X /F

 
Xd [ΩM ] / G,

where X := RSpec A, and Xd [ΩM ] := RSpec (A ⊕d ΩM ). Making a homotopy base

change
X / F ×h Xd [ΩM ]
G

 
Xd [ΩM ] / Xd [ΩM ],

we see that we can replace G be Xd [ΩM ] and f by the projection F ×hG Xd [ΩM ] −→
Xd [ΩM ]. In particular, we can assume that G is a representable D− -stack. The
morphism f can then be written as
f : F ≃ RSpec C −→ RSpec B ≃ G,
and corresponds to a morphism of commutative simplicial k-algebras B −→ C. Then,
using Prop. 1.4.2.6 and Cor. 2.2.3.3 we see that the obstruction for the point x to
lifts to a point in π0 (F (A ⊕d ΩM )) lives in the abelian group [LC/B ⊗LC A, M ]. But,
as B −→ C is assumed to be smooth, the A-module LC/B ⊗LC A is a retract of a free
A-module. This implies that [LC/B ⊗LC A, M ] is a retract of a product of π0 (M ), and
therefore is 0 by hypothesis on M . This implies that condition (2) of the proposition
is satisfied when n = −1.
Let us now assume that condition (2) is satisfied for all smooth m-representable
morphisms for m < n, and let us prove it for a smooth n-representable morphism
f : F −→ G. Using the same trick as above, se see that we can always assume that G
is a representable D− -stack,and therefore that F is ann-geometric D− -stack. Then,
let us chose a point x in π0 G(A ⊕d ΩM ) ×hF (A) G(A) , and we need to show that x
lifts to a point in π0 (F (A ⊕d ΩM )). For this, we use Cor. 2.2.3.3 for F , and consider
 2.2.5. INFINITESIMAL CRITERIA FOR SMOOTH AND ÉTALE MORPHISMS 161

its cotangent complex LF,y ∈ Ho(Sp(A − M ods )), where y ∈ F (A) is the image of x.
There exists a natural functoriality morphism
LG,f (y) −→ LF,y
whose homotopy cofiber is LF/G,y ∈ Ho(Sp(A − M ods )). Then, Prop. 1.4.2.6 tell us
that the obstruction for the existence of this lift lives in [LF/G,y , M ]. It is therefore
enough to show that [LF/G,y , M ] ≃ 0 for any M ∈ A − M ods such that π0 (M ) = 0.
Lemma 2.2.5.2. Let F −→ G be a smooth morphism between n-geometric D− -
stacks with G a representable stack. Let A ∈ sk − Alg and y : Y = RSpec A −→ F be
a point. Then the object
LF/G,y ∈ Ho(Sp(A − M ods ))
is perfect, and its dual L∨ ∈ Ho(Sp(A − M ods )) is 0-connective (i.e. belongs to
F/G,y
the image of Ho(A − M ods ) ֒→ Ho(Sp(A − M ods )).
Proof. Recall first that an object in Ho(Sp(A − M ods )) is perfect if and only if
it is finitely presented, and if and only if it is a retract of a finite cell stable A-module
(see Cor. 1.2.3.8, and also [EKMM, III.2] or [Kr-Ma, Thm. III.5.7]).
The proof is then by induction on n. When F is representable, this is by definition
of smooth morphisms, as then LF/G,y is a projective A-module of finite presentation,
and so is its dual. Let us suppose the lemma proved for all m < n, and lets prove
it for n. First of all, the conditions on LF/G,y we need to prove are local for the
étale topology on A, because of Cor. 1.3.7.8. Therefore, one can assume that the
point y lifts to a point of an n-atlas for F . One can thus suppose that there exists a
representable D− -stack U , a smooth morphism U −→ F , such that y ∈ π0 (RF (A))
is the image of a point z ∈ RU (A). There exists an fibration sequence of stable
A-modules
LF/G,y −→ LU/G,z −→ LU/F,z .
As U −→ G is smooth, LU/G,z is a projective A-module. Furthermore, the stable
A-module LU/F,y can be identified with LU×hF Y /Y,s , where s is the natural section
Y −→ U ×hF Y induced by the point z : Y −→ U . The morphism U ×hF Y −→ Y
being a smooth and (n − 1)-representable morphism, induction tells us that the stable
A-module LU/F,y satisfies the conditions of the lemma. Therefore, LF/G,y is the
homotopy fiber of a morphism between stable A-module satisfying the conditions of
the lemma, and is easily seen to satisfies itself these conditions. 
By the above lemma, we have
[LF/G,y , M ] ≃ π0 (L∨ L ∨
F/G,y ⊗A M ) ≃ π0 (LF/G,y ) ⊗π0 (A) π0 (M ) ≃ 0

for any A-module M such that π0 (M ) = 0. This finishes the proof of the fact that f
satisfies the conditions of Prop. 2.2.5.1 when it is smooth.
Conversely, let us assume that f : F −→ G is a morphism satisfying the lifting
property of 2.2.5.1, and let us show that f is smooth. Clearly, one can suppose that
G is a representable stack, and thus that F is an n-geometric D− -stack. We need to
show that for any representable D− -stack U and any smooth morphism U −→ F the
composite morphism U −→ G is smooth. By what we have seen in the first part of
the proof, we known that U −→ F also satisfies the lifting properties, and thus so
does the composition U −→ G. We are therefore reduced to the case where f is a
morphism between representable D− -stacks, and thus corresponds to a morphism of
commutative simplicial k-algebras A −→ B. By hypothesis on f , π0 (A) −→ π0 (B) is
a finitely presented morphism of commutative rings. Furthermore, Prop. 1.4.2.6 and
162 2.2. DERIVED ALGEBRAIC GEOMETRY

Cor. 2.2.3.3 show that for any B-module M with π0 (M ) = 0, we have [LB/A , M ] = 0.
Let B (I) −→ LB/A be a morphism of B-modules, with B (I) free over some set I,
and such that the induced morphism π0 (B)(I) −→ π0 (LB/A ) is surjective. Let K be
the homotopy fiber of the morphism B (I) −→ LB/A , that, according to our choice,
induces a homotopy fiber sequence of A-modules

B (I) / LB/A / K[1].

The short exact sequence

[LB/A , B (I) ] −→ [LB/A , LB/A ] −→ [LB/A , K[1]] = 0,

shows that LB/A is a retract of B (I) , and thus is a projective B-module. Furthermore,
the homotopy cofiber sequence

LA ⊗LA B / LB / LB/A ,

induces a short exact sequence

[LB , LA ⊗LA B] / [LA ⊗L B, LA ⊗L B] / [LB/A , LA ⊗L B[1]] = 0,

A A A

shows that the morphism LA ⊗LA B −→ LB has a retraction. We conclude that f is

a formally smooth morphism such that π0 (A) −→ π0 (B) is finitely presented, and by
Cor. 2.2.2.11 that f is a smooth morphism. 

From the proof of Prop. 2.2.5.1 we extract the following corollary.

Corollary 2.2.5.3. Let F −→ G be an n-representable morphism of D− -stacks
such that the morphism t0 (F ) −→ t0 (G) is a locally finitely presented morphism of
stacks. The following three conditions are equivalent.
(1) The morphism f is smooth.
(2) For any A ∈ sk − Alg and any morphism of stacks x : X = RSpec A −→ F ,
the object
LF/G,x ∈ Ho(Sp(A − M ods ))
is perfect, and its dual L∨
F/G,y ∈ Ho(Sp(A − M ods )) is 0-connective (i.e.
belongs to the image of Ho(A − M ods ) ֒→ Ho(Sp(A − M ods )).
(3) For any A ∈ sk − Alg, any morphism of stacks x : X = RSpec A −→ F ,
and any A-module M in sk − M od1 , we have
[LF/G,x , M ] = 0.
Proof. That (1) implies (2) follows from lemma 2.2.5.2. Conversely, if LF/G,x
satisfies the conditions of the corollary, then for any A-module M such that π0 (M )
one has [LF/G,x , M ] = 0. Therefore, Prop. 1.4.2.6 shows that the lifting property of
Prop. 2.2.5.1 holds, and thus that (2) implies (1). Furthermore, clearly (2) implies
(3), and conversely (3) together with Prop. 1.4.2.6 implies the lifting property of
Prop. 2.2.5.1. 

Proposition 2.2.5.4. Let f : F −→ G be an n-representable morphism between

D− -stacks. The morphism f is étale if and only if it satisfies the following two
conditions
(1) The morphism t0 (f ) : t0 (F ) −→ t0 (G) is locally finitely presented as a
morphism in k − Af f ∼,τ .
 2.2.5. INFINITESIMAL CRITERIA FOR SMOOTH AND ÉTALE MORPHISMS 163

(2) For any A ∈ sk − Alg, any M ∈ A − M ods whose underlying k-module is in

sk − M od1 , and any derivation d ∈ π0 (Der(A, M )), the natural projection
A ⊕d ΩM −→ A induces an isomorphism in Ho(SSet)
RF (A ⊕d ΩM ) −→ RG(A ⊕d ΩM ) ×hRF (A) RG(A).
Proof. First of all one can suppose that F and G are fibrant objects in k −
D− Af f ∼,ét .
Suppose first that the morphism f is étale and let us prove that is satisfies the
two conditions of the proposition. We know by 2.2.4.4 (3) that t0 (f ) is then a étale
morphism in k − Af f ∼,ét , so condition (1) is satisfied. The proof that (2) is also
satisfied goes by induction on n. Let us start with the case n = −1, and in other
words when f is an étale and representable morphism. In this case the result follows
from Cor. 1.2.8.4 (2). We now assume the result for all m < n and prove it for n.
Let A ∈ sk − Alg, M an A-module with π0 (M ) = 0, and d ∈ π0 (Der(A, M )) be a
derivation. Let x be a point in π0 (RG(A ⊕d ΩM ) ×hRF (A) RG(A)), with image y in
π0 (RF (A)). By Prop. 1.4.2.6 the homotopy fiber of the morphism
RF (A ⊕d ΩM ) −→ RG(A ⊕d ΩM ) ×hRF (A) RG(A)
at the point x is non empty if and only if a certain obstruction in [LF/G,y , M ] vanishes.
Furthermore, if nonempty this homotopy fiber is then equivalent to
M apA−Mod (LF/G,y , Ω(M )).
It is then enough to prove that LF/G,y = 0, and this is contained in the following
lemma.

Lemma 2.2.5.5. Let F −→ G be a étale morphism between n-geometric D− -

stacks. Let A ∈ sk − Alg and y : Y = RSpec A −→ F be a point. Then LF/G,y ≃ 0.
Proof. We easily reduce to the case when G is a representable D− -stack.
The proof is then by induction on n. When F is representable, this is by definition
of étale morphisms. Let us suppose the lemma proved for all m < n, and lets prove
it for n. First of all, the vanishing of LF/G,y is clearly a local condition for the étale
topology on A. Therefore, we can assume that the point y lifts to a point of an n-atlas
for F . We can thus suppose that there exists a representable D− -stack U , a smooth
morphism U −→ F , such that y ∈ π0 (RF (A)) is the image of a point z ∈ RU (A).
There exists an fibration sequence of stable A-modules
LF/G,y −→ LU/G,z −→ LU/F,z .

As U −→ G is étale, LU/G,z ≃ 0. Furthermore, the stable A-module LU/F,y can be

identified with LU×hF Y /Y,s , where s is the natural section Y −→ U ×hF Y induced by
the point z : Y −→ U . The morphism U ×hF Y −→ Y being an étale and (n − 1)-
representable morphism, induction tells us that the stable A-module LU/F,y vanishes.
We conclude that LF/G,y ≃ 0. 

The lemma finishes the proof that if f is étale then it satisfies the two conditions
of the proposition.
Conversely, let us assume that f satisfies the properties (1) and (2) of the proposi-
tion. To prove that f is étale we can suppose that G is a representable D− -stack. We
then need to show that for any representable D− -stack U and any smooth morphism
u : U −→ F , the induced morphism v : U −→ G is étale. But Prop. 1.4.2.6 and
164 2.2. DERIVED ALGEBRAIC GEOMETRY

our assumption (2) easily implies LF/G,v ≃ 0. Furthermore, the obstruction for the
homotopy cofiber sequence
LU/G,v / LF/G,u / LU/F,u ,

to splits lives in [LU/F,u , S(LU/G,v )], which is zero by Cor. 2.2.5.3. Therefore, LU/G,v
is a retract of LF/G,u and thus vanishes. This implies that U −→ G is a formally étale
morphism of representable D− -stacks. Finally, our assumption (1) and Cor. 2.2.2.11
implies that U −→ G is an étale morphism as required. 
Corollary 2.2.5.6. Let F −→ G be an n-representable morphism of D− -stacks
such that the morphism t0 (F ) −→ t0 (G) is a locally finitely presented morphism of
stacks. The following two conditions are equivalent.
(1) The morphism f is étale.
(2) For any A ∈ sk − Alg and any morphism of stacks x : X = RSpec A −→ F ,
one has LF/G,x ≃ 0.
Proof. This follows from Prop. 2.2.5.4 and Prop. 1.4.2.6. 

2.2.6. Some examples of geometric D− -stacks

We present here some basic examples of geometric D− -stacks. Of course we do
not claim to be exhaustive, and many other interesting examples will not be discussed
here and will appear in future works (see e.g. [Go, To-Va1]).

2.2.6.1. Local systems. Recall from Def. 1.3.7.5 the existence of the D− -stack
Vectn , of rank n vector bundles. Recall by 2.2.2.2 that for A ∈ sk−Alg, an A-module
M ∈ A − Algs is a rank n vector bundle, if and only if M is a strong A-module and
π0 (M ) is a projective π0 (A)-module of rank n. Recall also from Lem. 2.2.2.2 that
vector bundles are precisely the locally perfect modules. The conditions of 1.3.7.12
are all satisfied in the present context and therefore we know that Vectn is a smooth
1-geometric D− -stack. As a consequence of Prop. 2.2.4.4 (4) we deduce that the
D− -stack Vectn is truncated in the sense of 2.2.4.3. We also have a stack of rank
n vector bundles Vectn in St(k). Using the same notations for these two different
objects is justified by the following lemma.
Lemma 2.2.6.1. There exists a natural isomorphism in D− St(k)
i(Vectn ) ≃ Vectn .
Proof. As we know that the D− -stack Vectn is truncated, it is equivalent to
show that there exists a natural isomorphism
Vectn ≃ t0 (Vectn )
∼,τ
in Ho(k − Af f ).
We start by defining a morphism of stacks
Vectn −→ t0 (Vectn ).
For this, we construct for a commutative k-algebra A, a natural functor
φA : A − QCohcW −→ i(A) − QCohcW ,
where i(A) ∈ sk − Alg is the constant simplicial commutative k-algebra associated
to A, and where A − QCohcW and i(A) − QCohcW are defined in §1.3.7. Recall that
A − QCohcW is the category whose objects are the data of a B-module MB for any
morphism A −→ B in k − Alg, together with isomorphisms MB ⊗B B ′ ≃ MB ′ for any
 2.2.6. SOME EXAMPLES OF GEOMETRIC D− -STACKS 165

A −→ B −→ B ′ in k − Alg, satisfying the usual cocycle conditions. The morphisms

M → M ′ in A − QCohcW are simply the families of isomorphisms MB ≃ MB′ of
B-modules which commute with the transitions isomorphisms. In the same way, the
objects in i(A) − QCohcW are the data a cofibrant B-module MB ∈ B − M ods for any
morphism i(A) −→ B in sk − Alg, together with equivalences MB ⊗B B ′ −→ MB ′
for any i(A) −→ B −→ B ′ in sk − Alg, satisfying the usual cocycle conditions. The
morphisms M → M ′ in i(A) − QCohcW are simply the equivalences MB −→ MB′
which commutes with the transition equivalences. The functor

φA : A − QCohcW −→ i(A) − QCohcW ,

sends an object M to the object φA (M ), where φA (M )B is the simplicial B-module

defined by
(φA (M )B )n := MBn
(note that B can be seen as a simplicial commutative A-algebra). The functor φA is
clearly functorial in A and thus defines a morphism of simplicial presheaves

QCoh −→ t0 (QCoh).

We check easily that the sub-stack Vectn of QCoh is sent to the sub-stack t0 (Vectn )
of t0 (QCoh), and therefore we get a morphism of stacks

Vectn −→ t0 (Vectn ).

To see that this morphism is an isomorphism of stacks we construct a morphism in the

other direction by sending a simplicial i(A)-module M to the π0 (A)-module π0 (M ).
By 2.2.2.2 we easily see that this defines an inverse of the above morphism. 

For a simplicial set K ∈ SSetU , and an object F ∈ k −D− Af f ∼,ét , we can use the
simplicial structure of the category k − D− Af f ∼,ét in order to define the exponential
F K ∈ k − D− Af f ∼,ét . The model category k − D− Af f ∼,ét being a simplicial model
category the functor

(−)K : k − D− Af f ∼,ét −→ k − D− Af f ∼,ét

is right Quillen, and therefore can be derived on the right. Its right derived functor
will be denoted by
D− St(k) −→ D− St(k)
F 7→ F RK .
Explicitly, we have
F RK ≃ (RF )K
where RF is a fibrant replacement of F in k − D− Af f ∼,ét .

Definition 2.2.6.2. Let K be a U-small simplicial set. The derived moduli stack
of rank n local systems on K is defined to be

RLocn (K) := VectRK

n .

We start by the following easy observation.

Lemma 2.2.6.3. Assume that K is a finite dimensional simplicial set. Then, the
D− -stack RLocn (K) is a finitely presented 1-geometric D− -stack.
166 2.2. DERIVED ALGEBRAIC GEOMETRY

Proof. We consider the following homotopy co-cartesian square of simplicial

sets
SkiO K / Ski+1 K
O

` i+1
` i+1
Hom(∂∆i+1 ,K) ∂∆ Hom(∆i+1 ,K) ∆ ,
/
where Ski K is the skeleton of dimension of i of K. This gives a homotopy pullback
square of D− -stacks
RLocn (Ski+1 K) / RLocn (Ski K)

 
Qh / Qh
Hom(∆i+1 ,K) RLocn (∆i+1 ) Hom(∂∆i+1 ,K) RLocn (∂∆i+1 ).

As finitely presented 1-geometric D− -stacks are stable by homotopy pullbacks, we see

by induction on the skeleton that it only remains to show that RLocn (∂∆i+1 ) is a
finitely presented 1-geometric D− -stack. But there is an isomorphism in Ho(SSet),
L
a
∂∆i+1 ≃ ∗ ∗,
∂∆i

giving rise to an isomorphism of D− -stacks

RLocn (∂∆i+1 ) ≃ RLocn ×hRLocn (∂∆i ) RLocn .
By induction on i, we see that it is enough to show that Vectn is a finitely presented
1-geometric D− -stack which is known from Cor. 1.3.7.12. 
Another easy observation is the description of the truncation t0 RLocn (K). For
this, recall that the simplicial set K has a fundamental groupoid Π1 (K). The usual
Artin stack of rank n local systems on K is the stack in groupoids defined by
Locn (K) : k − Alg −→ {Groupoids}
A 7→ Hom(Π1 (K), Vectn (A)).
In other words, Locn (K)(A) is the groupoid of functors from Π1 (K) to the groupoid
of rank n projective A-modules. As usual, this Artin stack is considered as an object
in St(k).
Lemma 2.2.6.4. There exists a natural isomorphism in St(k)
Locn (K) ≃ t0 RLocn (K).
Proof. The truncation t0 being the right derived functor of a right Quillen
functor commutes with derived exponentials. Therefore, we have
t0 RLocn (K) ≃ (t0 RLocn )RK ≃ (Vectn )RK ∈ St(k).
The stack Vectn is 1-truncated, and therefore we also have natural isomorphisms
(Vectn )RK (A) ≃ M apSSet (K, Vectn (A)) ≃ Hom(Π1 (K), Π1 (Vectn (A))).
This equivalence is natural in A and provides the isomorphism of the lemma. 
Our next step is to give a more geometrical interpretation of the D− -stack
RLocn (K), in terms of certain local systems of objects on the topological realiza-
tion of K.
Let now X be a U-small topological space. Let A ∈ sk − M od, and let us define a
model category A − M ods (X), of A-modules over X. The category A − M ods (X) is
 2.2.6. SOME EXAMPLES OF GEOMETRIC D− -STACKS 167

simply the category of presheaves on X with values in A−M ods . The model structure
on A − M ods (X) is of the same type as the local projective model structure on
simplicial presheaves. We first define an intermediate model structure on A−M ods (X)
for which equivalences (resp. fibrations) are morphism E −→ F in A − M ods (X)
such that for any open subset U ⊂ X the induced morphism E(U ) −→ F(U ) is an
equivalence (resp. a fibration). This model structure exists as A − M ods is a U-
cofibrantly generated model category, and let us call it the strong model structure.
The final model structure on A − M ods (X) is the one for which cofibrations are
the same cofibrations as for the strong model structure, and equivalences are the
morphisms E −→ F such that for any point x ∈ X the induced morphism on the
stalks Ex −→ Fx is an equivalence in A− M ods . The existence of this model structure
is proved the same way as for the case of simplicial presheaves (we can also use the
forgetful functor A−M ods (X) −→ SP r(X) to lift the local projective model structure
on SP r(X) in a standard way).
For a commutative simplicial k-algebra A, we consider A − M ods (X)cW , the sub-
category of A − M ods (X) consisting of cofibrant objects and equivalences between
them. For a morphism of commutative simplicial k-algebras A −→ B, we have a base
change functor
A − M ods (X) −→ B − M ods (X)
E 7→ E ⊗A B
which is a left Quillen functor, and therefore induces a well defined functor
− ⊗A B : A − M ods (X)cW −→ B − M ods (X)cW .
This defines a lax functor A 7→ A − M ods (X)cW , from sk − Alg to Cat, which can be
strictified in the usual way. We will omit to mention explicitly this strictification here
and will do as if A 7→ A − M ods (X)cW does define a genuine functor sk − Alg −→ Cat.
We then define a sub-functor of A 7→ A − M ods (X)cW in the following way. For
A ∈ sk − Alg, let A− Locn (X) be the full subcategory of A− M ods (X)cW consisting of
objects E, such that there exists an open covering {Ui } on X, such that each restriction
E|Ui is isomorphic in Ho(A− M ods (Ui )) to a constant presheaf with fibers a projective
A-module of rank n (i.e. projective A-module E such that π0 (E) is a projective π0 (A)-
module of rank n). This defines a sub-functor of A 7→ A − M ods (X)cW , and thus a
functor from sk − Alg to Cat. Applying the nerve functor we obtain a simplicial
presheaf RLocn (X) ∈ k − D− Af f ∼,ét , defined by
RLocn (X)(A) := N (A − Locn (X)).
Proposition 2.2.6.5. Let K be a simplicial set in U and |K| be its topological
realization. The simplicial presheaf RLocn (|K|) is a D− -stack, and there exists an
isomorphism in D− St(k)
RLocn (|K|) ≃ RLocn (K).
Proof. We first remark that if f : X −→ X ′ is a homotopy equivalence of
topological spaces, then the induced morphism
f ∗ : RLocn (X ′ ) −→ RLocn (X ′ )
is an equivalence of simplicial presheaves. Indeed, a standard argument reduces to
the case where f is the projection X × [0, 1] −→ X, and then to [0, 1] −→ ∗, for which
one can use the same argument as in [To3, Lem. 2.16].
Let us first prove that RLocn (|K|) is a D− -stack, and for this we will prove
that the simplicial presheaf RLocn (|K|) can be written, in SP r(k − D− Af f ), as a
certain homotopy limit of D− -stacks. As D− -stacks in SP r(k − D− Af f ) are stable
by homotopy limits this will prove what we want.
168 2.2. DERIVED ALGEBRAIC GEOMETRY

Let U∗ an open hypercovering of |K| such that each Ui is a coproduct of con-

tractible open subsets in |K| (such a hypercover exists as |K| is a locally contractible
space). It is not hard to show using Cor. B.0.8 and standard cohomological descent,
that for any A ∈ sk − Alg, the natural morphism
N (A − M ods (|K|)cW ) −→ Holimm∈∆ N (A − M ods (Um )cW )
is an isomorphism in Ho(SSet). From this, we easily deduce that the natural mor-
phism
RLocn (|K|) −→ Holimm∈∆ RLocn (Um )
is an isomorphism in Ho(SP r(k − D− Af f )). Therefore, we are reduced to show that
RLocn (Um ) is a D− -stack. But, Um being a coproduct of contractible topological
spaces, RLocn (Um ) is a product of some RLocn (U ) for some contractible space U .
Moreover RLocn (U ) is naturally isomorphic in Ho(SP r(k −D− Af f )) to RLocn (∗) =
Vectn . As we know that Vectn is a D− -stack, this shows that the simplicial presheaf
RLocn (|K|) is a homotopy limit of D− -stacks and thus is itself a D− -stack.
We are left to prove that RLocn (|K|) and RLocn (K) are isomorphic. But, we
have seen that
RLocn (|K|) ≃ Holimm∈∆RLocn (Um ),
for an open hypercover U∗ of |K|, such that each Um is a coproduct of contractible
open subsets. We let K ′ := π0 (U∗ ) be the simplicial set of connected components of
U∗ , and thus
′
Holimm∈∆ RLocn (Um ) ≃ Holimm∈∆RLocn (Km ),
′
where Km is considered as a discrete topological space. We have thus proved that
h
Y ′
RLocn (|K|) ≃ Holimm∈∆ Vectn ≃ (Vectn )RK .
′
Km

But, by [To3, Lem. 2.10], we know that |K| is homotopically equivalent to |K ′ |, and
thus that K is equivalent to K ′ . This implies that
′
RLocn (|K|) ≃ (Vectn )RK ≃ (Vectn )RK ≃ RLocn (|K|).

We will now describe the cotangent complex of RLocn (K). For this, we fix a
global point
E : ∗ −→ RLocn (K),
which by Lem. 2.2.6.4 corresponds to a functor
E : Π1 (K) −→ Π1 (Vectn (k)),
where Π1 (Vectn (k)) can be identified with the groupoid of rank n projective k-
modules. The object E is thus a local system of rank n projective k-modules on K in
the usual sense. We will compute the cotangent complex LRLocn (K),E ∈ Ho(Sp(sk −
M od)) (recall that Ho(Sp(sk − M od)) can be naturally identified with the unbounded
derived category of k). For this, we let E ⊗k E ∨ be the local system on K of endo-
morphisms of E, and C∗ (K, E ⊗k E ∨ ) will be the complex of homology of K with
coefficients in the local system E⊗k E ∨ . We consider C∗ (K, E⊗k E ∨ ) as an unbounded
complex of k-modules, and therefore as an object in Ho(Sp(sk − M od)).
Proposition 2.2.6.6. There exists an isomorphism in Ho(Sp(sk − M od)) ≃
Ho(C(k))
LRLocn (K),E ≃ C∗ (K, E ⊗k E ∨ )[−1].
 2.2.6. SOME EXAMPLES OF GEOMETRIC D− -STACKS 169

Proof. Let M ∈ sk − M od, and let us consider the simplicial set

DerE (RLocn (K), M ),
of derivations of RLocn (K) at the point E and with coefficients in M . By definition
of RLocn (K) we have
DerE (RLocn (K), M ) ≃ M apSSet/Vectn (k) (K, Vectn (k ⊕ M )),
where K −→ Vectn (k) is given by the object E, and Vectn (k ⊕ M ) −→ Vectn (k)
is the natural projection. At this point we use Prop. A.0.6 in order to describe,
functorially in M , the morphism Vectn (k ⊕ M ) −→ Vectn (k). For this, we let G(k)
to be the groupoid of projective k-modules of rank n. We also define an S-category
G(k ⊕ M ) in the following way. Its objects are projective k-modules of rank n. The
simplicial set of morphisms in G(k ⊕ M ) between two such k-modules E and E ′ is
defined to be
G(k ⊕ M )(E,E ′ ) := HomEq ′ ′
(k⊕M)−Mods (E ⊕ (E ⊗k M ), E ⊕ (E ⊗k M )),

the simplicial set of equivalences from E ⊕ (E ⊗k M ) to E ′ ⊕ (E ′ ⊗k M ), in the model

category (k ⊕ M ) − M ods . It is important to note that E ⊕ (E ⊗k M ) is isomorphic
to E ⊗k (k ⊕ M ), and therefore is a cofibrant object in (k ⊕ M ) − M ods (as the base
change of a cofibrant object E in sk − M od). There exists a natural morphism of
S-categories
G(k ⊕ M ) −→ G(k)
being the identity on the set of objects, and the composition of natural morphisms
HomEq ′ ′ Eq ′
(k⊕M)−Mods (E ⊕ (E ⊗k M ), E ⊕ (E ⊗k M )) −→ Homk−Mods (E, E ) −→

π0 (HomEq ′
k−Mods (E, E )) ≃ G(k)(E,E ′ ) .
on the simplicial sets of morphisms. Clearly, Prop. A.0.6 and its functorial properties,
show that the morphism
Vectn (k ⊕ M ) −→ Vectn (k)
is equivalent to
N (G(k ⊕ M )) −→ N (G(k)),
and in a functorial way in M .
For any E ∈ G(k), we can consider the classifying simplicial set K(E ⊗k E ∨ ⊗k
M, 1) of the simplicial abelian group E ⊗k E ∨ ⊗k M , and for any isomorphism of
projective k-modules of rank n, E ≃ E ′ , the corresponding isomorphism of simplicial
set
K(E ⊗k E ∨ ⊗k M, 1) ≃ K(E ′ ⊗k (E ′ )∨ ⊗k M, 1).
This defines a local system L of simplicial sets on the groupoid G(k), for which the
total space
HocolimG(k) L −→ N (G(k))
is easily seen to be isomorphic to the projection
N (G(k ⊕ M )) −→ N (G(k)).
The conclusion is that the natural projection
Vectn (k ⊕ M ) −→ Vectn (k)
is equivalent, functorially in M , to the total space of the local system E 7→ K(E ⊗k
E ∨ ⊗k M, 1) on Vectn (k). Therefore, one finds a natural equivalence of simplicial
sets
DerE (RLocn (K), M ) ≃ M apSSet/Vectn (k) (K, Vectn (k ⊕ M )) ≃
170 2.2. DERIVED ALGEBRAIC GEOMETRY

C ∗ (K, E ⊗k E ∨ ⊗k M [1]) ≃ M apC(k) (C∗ (K, E ⊗k E ∨ )[−1], M ).

As this equivalence is functorial in M , this shows that
LRLocn (K),E ≃ C∗ (K, E ⊗k E ∨ )[−1],
as required. 
Remark 2.2.6.7. An important consequence of Prop. 2.2.6.6 is that the D− -
stack depends on strictly more than the fundamental groupoid of K. Indeed, the
tangent space of RLocn (K) at a global point corresponding to a local system E on
K is C ∗ (K, E ⊗k E ∨ )[1], which can be non-trivial even when K is simply connected.
In general, there is a closed immersion of D− -stacks (Prop. 2.2.4.7)
it0 (RLocn (K)) −→ RLocn (K),
which on the level of tangent spaces induces the natural morphism
τ≤0 (C ∗ (K, E ⊗k E ∨ )[1]) −→ C ∗ (K, E ⊗k E ∨ )[1]
giving an isomorphism on H 0 and H 1 . This shows that the D− -stack RLocn (K)
contains strictly more information than the usual Artin stack of local systems on K,
and does encode some higher homotopical invariants of K.
2.2.6.2. Algebras over an operad. Recall (e.g. from [Re, Sp]) the notions
of operads and algebras over them, as well as their model structures. Let O be an
operad in k − M od, the category of k-modules. We assume that for any n, the k-
module O(n) is projective. Then, for any A ∈ sk−Alg, we can consider O⊗k A, which
is an operad in the symmetric monoidal category A − M ods of A-modules. We can
therefore consider O ⊗k A − Algs , the category of O-algebras in A − M ods . According
to [Hin1] the category O ⊗k A − Algs can be endowed with a natural structure of
a U-cofibrantly generated model category for which equivalences and fibrations are
defined on the underlying objects in sk − M od. For a morphism A −→ B in sk − Alg,
there is a Quillen adjunction
B ⊗A − : O ⊗k A − Algs −→ O ⊗k B − Algs O ⊗k A − Algs ←− O ⊗k B − Algs : F,
where F is the forgetful functor. The rule A 7→ O ⊗k A − Algs together with base
change functors B ⊗A − is almost a left Quillen presheaf in the sense of Appendix
B, except that the associativity of composition of base change is only valid up to a
natural isomorphism. However, the standard strictification techniques can be applied
in order to replace, up to a natural equivalence, this by a genuine left Quillen presheaf.
We will omit to mention this replacement and will proceed as if A 7→ O ⊗k A − Algs
actually defines a left Quillen presheaf on k − D− Af f .
For any A ∈ sk − Alg, we consider O − Alg(A), the category of cofibrant objects
in O ⊗k A − Algs and equivalences between them. Finally, we let O − Algn (A) be
the full subcategory consisting of objects B ∈ O − Alg(A), such that the underlying
A-module of B is a vector bundle of rank n (i.e. that B is a strong A-module, and
π0 (B) is a projective π0 (A)-module of rank n). The base change functors clearly
preserves the sub-categories O − Algn (A), and we get this way a well defined presheaf
of V-small categories
k − D− Af f −→ CatV
A 7→ O − Algn (A).
Applying the nerve functor we obtain a simplicial presheaf
k − D− Af f −→ SSetV
A 7→ N (O − Algn (A)).
 2.2.6. SOME EXAMPLES OF GEOMETRIC D− -STACKS 171

This simplicial presheaf will be denoted by AlgO

n , and is considered as an object in
k − D− Af f ∼,ét .
Proposition 2.2.6.8. (1) The object AlgO −
n ∈ k − D Af f
∼,ét
is a D− -
stack.
(2) The D− -stack AlgO
n is 1-geometric and quasi-compact.

Proof. (1) The proof relies on the standard argument based on Cor. B.0.8, and
is left to the reader.

(2) We consider the natural morphism of D− -stacks

p : AlgO
n −→ Vectn ,

defined by forgetting the O-algebra structure. Precisely, it sends an object B ∈ O −

Algn (A) to its underlying A-modules, which is a rank n-vector bundle by definition.
We are going to prove that the morphism p is a representable morphism, and this
will imply the result as Vectn is already known to be 1-geometric and quasi-compact.
For this, we consider the natural morphism ∗ −→ Vectn , which is a smooth 1-atlas
for Vectn , and consider the homotopy pullback
^ O O h
Alg n := Algn ×Vectn ∗.

^
It is enough by Prop. 1.3.3.4 to show that Alg O
n is a representable stack. For this,
^
we use [Re, Thm. 1.1.5] in order to show that the D− -stack Alg O
is isomorphic in
n
D− St(k) to M ap(O, End(k n )), defined as follows. For any A ∈ sk − Alg, we consider
End(An ), the operad in A − M ods of endomorphisms of the object An (recall that for
M ∈ A−M ods , the operad End(M ) is defined by End(M )(n) := HomA (M ⊗k n , M )).
We let QO be a cofibrant replacement of the operad O in the model category sk−M od.
Finally, M ap(O, End(k n )) is defined as
sk − Alg −→ SSet
A 7→ M ap(O, End(k n ))(A) := HomOp (QO, End(An )),
where HomOp denotes the simplicial set of morphism in the simplicial category of
^
operads in sk − M od. As we said, [Re, Thm. 1.1.5] implies that Alg O
is isomor-
n
phic to M ap(O, End(k n )). Therefore, it remain to show that M ap(O, End(k n )) is a
representable D− -stack.
For this, we can write O, up to an equivalence, as the homotopy colimit of free
operads
O ≃ Hocolimi Oi .
Then, we have
M ap(O, End(k n )) ≃ Holimi M ap(Oi , End(k n ))
and as representable D− -stacks are stable by homotopy limits, we are reduced to the
case where O is a free operad. This means that there exists an integer m ≥ 0, such
that for any other operad O′ in sk − M od, we have a natural isomorphism of sets
HomOp (O, O′ ) ≃ Hom(k, O′ (m)).
In particular, we find that the D− -stack M ap(O, End(k n )) is isomorphic to the D− -
stack sending A ∈ sk − Alg to the simplicial set HomA−Mods ((An )⊗k m , A), of mor-
phisms from the A-module (An )⊗k m to A. But this last D− -stack is clearly isomorphic
m
to RSpec B, where B is a the free commutative simplicial k-algebra on k n , or in other
words a polynomial algebra over k with nm variables. 
172 2.2. DERIVED ALGEBRAIC GEOMETRY

We still fix an operad O in k − M od (again requiring O(m) to be a projective

k-module for any m), and we let B be an O-algebra in k − M od, such that B is
projective of rank n as a k-module. This defines a well defined morphism of stacks
B : ∗ −→ AlgO
n.

We are going to describe the cotangent complex of AlgO

n at B using the notion of
(derived) derivations for O-algebras.
For this, recall from [Goe-Hop] the notion of O-derivations from B and with
coefficients in a B-module. For any B-module M , one can define the square zero
extension B ⊕ M of B by M , which is another O-algebra together with a natural
projection B ⊕ M −→ M . The k-module of derivations from B to M is defined by
DerkO (B, M ) := HomO−Alg/B (B, B ⊕ M ).
In the same way, for any O-algebra B ′ with a morphism B ′ −→ B we set
DerkO (B ′ , M ) := HomO−Alg/B (B ′ , B ⊕ M ).
The functor B ′ 7→ DerkO (B ′ , M ) can be derived on the left, to give a functor
RDerkO (−, M ) : Ho(O − Algs /B)op −→ Ho(SSetV ).
As shown in [Goe-Hop], the functor M 7→ RDerkO (B, M ) is co-represented by an
object LO
B ∈ Ho(sB − M od), thus there is a natural isomorphism in Ho(SSetV )

RDerkO (B, M ) ≃ M apsB−Mod (LO

B , M ).

The category sB − M od of simplicial B-modules is a closed model category for which

equivalences and fibrations are detected in sk − M od. Furthermore, the category
sB − M od is tensored and co-tensored over sk − M od making it into a sk − M od-
model category in the sense of [Ho1]. Passing to model categories of spectra, we
obtain a model category Sp(sB − M od) which is a Sp(sk − M od)-model category. We
will then set
RDer O O
k (B, M ) := RHomSp(sB−Mod) (LB , M ) ∈ Ho(Sp(sk − M od)),

where RHomSp(sB−Mod) denotes the Ho(Sp(sk − M od))-enriched derived Hom of

Sp(sB − M od). Note that Ho(sB − M od) is equivalent to the unbounded derived
category of B-modules, and as well that Ho(Sp(sk − M od)) is equivalent to the
unbounded derived category of k-modules.
Proposition 2.2.6.9. With the notations as above, the tangent complex of the
D− -stack AlgO
n at the point B is given by

TAlgO
n ,B
≃ RDer O
k (B, B)[1] ∈ Ho(Sp(sk − M od)).

Proof. We have an isomorphism in Ho(Sp(sk − M od))

TAlgO
n ,B
≃ TΩB AlgO
n ,B
[1].
Using Prop. A.0.6, we see that the D− -stack ΩB AlgO
n can be described as

sk − Alg −→ SSet
A 7→ M apeq
O−Algs (B, B ⊗k A)

where M apeq O−Algs denotes the mapping space of equivalences in the category of O-
algebras in sA − M od. In particular, we see that for M ∈ sk − M od, the simplicial set
DerΩB AlgOn
(B, M ) is naturally equivalent to the homotopy fiber, taken at the identity,
of the morphism
M apO−Algs (B, B ⊗k (k ⊕ M )) −→ M apO−Algs (B, B).
 2.2.6. SOME EXAMPLES OF GEOMETRIC D− -STACKS 173

The O-algebra B ⊗k (k ⊕ M ) can be identified with B ⊕ (M ⊗k B), the square

zero extension of B by M ⊗k B, as defined in [Goe-Hop]. Therefore, by definition
of derived derivations this homotopy fiber is naturally equivalent to RDerkO (B, M ).
This shows that
DerΩB AlgO
n
(B, M ) ≃ RDerkO (B, M ⊗k B) ≃ M apSp(sB−Mod) (LO
B , M ⊗k B).

This implies that for any N ∈ sk − M od, we have

M apSp(sk−Mod) (N, TΩB AlgO
n ,B
) ≃ DerΩB AlgO
n
(B, N ∨ ) ≃ M apSp(sB−Mod) (LO ∨
B , N ⊗k B) ≃

M apSp(sk−Mod) (N, RHomSp(sB−Mod) (LO

B , B)) ≃ M apSp(sk−Mod) (N, RDer k (B, B)).
The Yoneda lemma implies the existence of a natural isomorphism in Ho(Sp(sk −
M od))
TΩB AlgOn ,B
≃ RDerk (B, B).
We thus we have
TAlgO
n ,B
≃ RDerk (B, B)[1].

Remark 2.2.6.10. The proof of Prop. 2.2.6.9 actually shows that for any M ∈
sk − M od we have
RHomSp
k (LAlg O
n ,B
, M ) ≃ RDerk (B, B ⊗k M )[1].
2.2.6.3. Mapping D− -stacks. We let X be a stack in St(k), and F be an n-
geometric D− -stack. We are going to investigate the geometricity of the D− -stack of
morphisms from i(X) to F .
Map(X, F ) := Rét Hom(i(X), F ) ∈ D− St(k).
Recall that Rét Hom denotes the internal Hom of the category D− St(k).
Theorem 2.2.6.11. With the notations above, we assume the following three con-
ditions are satisfied.
(1) The stack
t0 (Map(X, F )) ≃ Rét Hom(X, t0 (F )) ∈ St(k)
is n-geometric.
(2) The D− -stack Map(X, F ) has a cotangent complex.
(3) The stack X can be written in St(k) has a homotopy colimit Hocolimi Ui ,
where Ui is a affine scheme, flat over Spec k.
Then the D− -stack Map(X, F ) is n-geometric.
Proof. The only if part is clear. Let us suppose that Map(X, F ) satisfies the
three conditions. To prove that it is an n-geometric D− -stack we are going to lift an
n-atlas of t0 (Map(X, F )) to an n-atlas of Map(X, F ). For this we use the following
special case of J.Lurie’s representability criterion, which can be proved using the
material of this paper (see Appendix C).
Theorem 2.2.6.12. (J. Lurie, see [Lu1] and Appendix C) Let F be a D− -stack.
The following conditions are equivalent.
(1) F is an n-geometric D− -stack.
(2) F satisfies the following three conditions.
(a) The truncation t0 (F ) is an Artin (n + 1)-stack.
(b) F has an obstruction theory.
174 2.2. DERIVED ALGEBRAIC GEOMETRY

(c) For any A ∈ sk − Alg, the natural morphism

RF (A) −→ Holimk RF (A≤k )
is an isomorphism in Ho(SSet).
We need to prove that Map(X, F ) satisfies the conditions (a) − (c) of theorem
2.2.6.12. Condition (a) is clear by assumption. The existence of a cotangent complex
for Map(X, F ) is guaranteed by assumption. The fact that Map(X, F ) is moreover
inf-cartesian follows from the general fact.
Lemma 2.2.6.13. Let F be a D− -stack which is inf-cartesian. Then, for any
D -stack F ′ , the D− -stack Rét Hom(F ′ , F ) is inf-cartesian.
−

Proof. Writing F ′ has a homotopy colimit of representable D− -stacks

F ′ ≃ Hocolimi Ui ,
we find
Rét Hom(F ′ , F ) ≃ Holimi Rét Hom(Ui , F ).
As being inf-cartesian is clearly stable by homotopy limits, we reduce to the case where
F ′ = RSpec B is a representable D− -stack. Let A ∈ sk − Alg, M be an A-module
with π0 (M ) = 0, and d ∈ π0 (Der(A, M )) be a derivation. Then, the commutative
square
Rét Hom(F ′ , F )(A ⊕d ΩM ) / Rét Hom(F ′ , F )(A)

 
Rét Hom(F ′ , F )(A) / Rét Hom(F ′ , F )(A ⊕ M )
is equivalent to the commutative square
RF ((A ⊕d ΩM ) ⊗Lk B) / RF (A ⊗L B)
k

 
RF (A ⊗Lk B) / RF ((A ⊕ M ) ⊗L B).
k

Using the F is inf-cartesian with respect to A ⊗Lk B ∈ sk − Alg, and the derivation
d⊗k B ∈ π0 (Der(A⊗Lk B, M ⊗Lk B)) implies that this last square is homotopy cartesian.
This shows that Rét Hom(F ′ , F ) is inf-cartesian. 
In order to finish the proof of Thm. 2.2.6.11 it only remain to show that Map(X, F )
also satisfies the condition (c) of Thm. 2.2.6.12. For this, we write X as Hocolimi Ui
with Ui affine and flat over k, and therefore Map(X, F ) can be written as the ho-
motopy limit Holimi Map(Ui , F ). In order to check condition (c) we can therefore
assume that X is an affine scheme, flat over k. Let us write X = Spec B, with B a
commutative flat k-algebra. Then, for any A ∈ sk − M od, the morphism
RMap(X, F )(A) −→ Holimk RMap(X, F )(A≤k )
is equivalent to
RF (A ⊗k B) −→ Holimk RF ((A≤k ) ⊗k B).
But, as B is flat over k, we have (A≤k ) ⊗k B ≃ (A ⊗k B)≤k , and therefore the above
morphism is equivalent to
RF (A ⊗k B) −→ Holimk RF ((A ⊗k B)≤k )
and is therefore an equivalence by (1) ⇒ (2) of Thm. 2.2.6.12 applied to F . 
 2.2.6. SOME EXAMPLES OF GEOMETRIC D− -STACKS 175

The following corollaries are direct consequences of Thm. 2.2.6.11. The only non
trivial part consists of proving the existence of a cotangent complex, which we will
assume in the present version of this work.
Corollary 2.2.6.14. Let X be a projective and flat scheme over Spec k, and Y
a projective and smooth scheme over Spec k. Then, the D− -stack Map(i(X), i(Y )) is
a 1-geometric D− -scheme. Furthermore, for any morphism of schemes f : X −→ Y ,
the cotangent complex of Map(i(X), i(Y )) at the point f is given by
LMap(i(X),i(Y )),f ≃ (C ∗ (X, f ∗ (TY )))∨ ,
where TY is the tangent sheaf of Y −→ Spec k.
Let us now suppose that k = C is the field of complex numbers, and let X be a
smooth and projective variety. We will be interested in the sheaf XDR of [S1], defined
by XDR (A) := Ared , for a commutative C-algebra A. Recall that the stack MDR (X)
is defined as the stack of morphisms from XDR to Vectn , and is identified with the
stack of flat bundles on X (see [S1]). It is known that MDR (X) is an Artin 1-stack.
Corollary 2.2.6.15. The D− -stack RMDR (X) := Map(i(XDR ), Vectn ) is 1-
geometric. For a point E : ∗ −→ RMDR (X), corresponding to a flat vector bundle E
on X, the cotangent complex of RMDR (X) at E is given by
∗
LRMDR (X),E ≃ (CDR (X, E ⊗ E ∗ ))∨ [−1],
∗
where CDR (X, E ⊗ E ∗ ) is the algebraic de Rham cohomology of X with coefficients in
the flat bundle E ⊗ E ∗ .
Corollary 2.2.6.15 is only the beginning of the story; in fact we can also produce,
in a similar way, RMDol (X) and RMHod (X), which are derived versions of the
moduli stacks of Higgs bundles and λ-connections of [S1], and this would lead us to a
derived version of non-abelian Hodge theory. We think this is very interesting research
direction because these derived moduli also encode higher homotopical data in their
tangent complexes. We hope to come back to this topic in a future work.
 CHAPTER 2.3

Complicial algebraic geometry

In this chapter we present a second context of application of the general formal-

ism of Part I, in which the base model category is C(k), the category of unbounded
complexes over some ring k of characteristic zero. Contrary to the previously consid-
ered applications, the general notions presented in §1.2 does not produce here notions
which are very close to the usual ones for commutative rings. As a consequence the
geometric intuition is here only a very loose guide.
We will present two different HAG contexts over C(k). The first one is very
weak in the sense that it is very easy for a stack to be geometric in this context (these
geometric stacks will be called weakly geometric). The price to pay for this abundance
of geometric stacks is that this context does not satisfy Artin’s conditions and thus
there is no good infinitesimal theory.
The second HAG context we consider is a bit closer to the geometric intuition,
and satisfies the Artin’s conditions so it behaves well infinitesimally.
Both of these “unbounded” contexts seem interesting as we are able to produce
examples of geometric stacks which cannot be represented by geometric D− -stacks,
i.e. by the kind of geometric stacks studied in the previous chapter.

In this chapter k will be a commutative Q-algebra.

2.3.1. Two HA contexts

We let C := C(k), the model category of unbounded complexes of k-modules in
U. The model structure on C(k) is the projective one, for which fibrations are epi-
morphisms and equivalences are quasi-isomorphisms. The model category C(k) is a
symmetric monoidal model category for the tensor product of complexes. Further-
more, it is well known that our assumptions 1.1.0.1, 1.1.0.3, 1.1.0.2 and 1.1.0.4 are
satisfied.
The category Comm(C(k)) is the usual model category of unbounded commu-
tative differential graded algebras over k, for which fibrations are epimorphisms and
equivalences are quasi-isomorphisms. The category Comm(C(k)) will be denoted by
k − cdga, and its objects will simply be called cdga’s. For A ∈ k − cdga, the cat-
egory A − M od is the category of unbounded A-dg-modules, again with its natural
model structure. In order to avoid confusion of notations we will denote the category
A − M od by A − M oddg . Note that for a usual commutative k-algebra k ′ we have
k ′ − M oddg = C(k ′ ), whereas k ′ − M od will denote the usual category of k ′ -modules.
Objects in A − M oddg will be called A-dg-modules.
As for the case of simplicial algebras, we will set for any E ∈ C(k)
πi (E) := H −i (E).
When A is a k-cdga, the Z-graded k-module π∗ (A) has a natural structure of a
commutative graded k-algebra. In the same way for M an A-dg-module, π∗ (M )
becomes a graded π∗ (A)-module. Objects A ∈ k − cdga such that πi (A) = 0 for any
177
178 2.3. COMPLICIAL ALGEBRAIC GEOMETRY

i < 0 will be called (−1)-connected. Any A ∈ k − cdga possesses a (−1)-connected

cover A′ −→ A, which is such that πi (A′ ) ≃ πi (A) for any i ≥ 0.
As in the context of derived algebraic geometry we will denote by M 7→ M [1] the
suspension functor. In the same way, M 7→ M [n] is the n-times iterated suspension
functor (when n is negative this means the n-times iterated loop functor).

The first new feature of complicial algebraic geometry is the existence of two in-
teresting choices for the subcategory C0 , both of them of particular interest depending
on the context. We let C(k)≤0 be the full subcategory of C(k) consisting of complexes
E such that πi (E) = 0 for any i < 0 (or equivalently, such that H i (E) = 0 for all
i > 0, which explains better the notation). We also set k − cdga0 to be the full sub-
category of k − cdga consisting of A ∈ k − cdga such that πi (A) = 0 for any i 6= 0. In
the same way we denote by k − cdga≤0 for the full subcategory of k − cdga consisting
of A such that πi (A) ≡ H −i (A) = 0 for any i < 0.

Lemma 2.3.1.1. (1) The triplet (C(k), C(k), k − cdga) is a HA context.

(2) The triplet (C(k), C(k)≤0 , k − cdga0 ) is a HA context.

Proof. The only non-trivial point is to show that any A ∈ k − cdga0 is C(k)≤0 -
good in the sense of Def. 1.1.0.10. By definition A is equivalent to some usual
commutative k-algebra, so we can replace A by k itself. We are then left to prove
that the natural functor

Ho(C(k))op −→ Ho((C(k)op ∧
≤0 ) )

is fully faithful.
To prove this, we consider the restriction functor

i∗ : Ho((C(k)op )∧ ) −→ Ho((C(k)op ∧
≤0 ) )

induced by the inclusion i : C(k)≤0 ⊂ C(k). We restrict the functor i∗ to the full
sub-categories of corepresentable objects

i∗ : Ho((C(k)op )∧ )corep −→ Ho((C(k)op ∧ corep

≤0 ) ) .

A precision here: we say that a functor F : C(k)≤0 −→ SSet is corepresentable if it

is of the form
D 7→ M apC(k) (E, i(D))

for some object E ∈ C(k) (i.e. it belongs the essential image of Ho(C(k))op −→
Ho((C(k)op ∧ ∗
≤0 ) ). We claim that this restricted functor i is an equivalence of cate-
gories. Indeed, an inverse f can be constructed by sending a functor F : C(k)≤0 −→
SSet to the functor

f (F ) : C(k) −→ SSet
D 7→ Holimn≥0 Ωn F (D(≤ n)[n]),

where D(≤ n) is the naive truncation of D defined by D(≤ n)m = Dm if m ≤ n and

D(≤ n)m = 0 if m > n, and Ωn F (D(≤ n)[n]) is the n-fold loop space of the simplicial
set F (D(≤ n)[n]), based at the natural point ∗ ≃ F (0) → F (D(≤ n)[n]). As any
D ∈ C(k) is functorially equivalent to the homotopy limit Holimn D(≤ n), it is easy
to check that the functor f and i∗ are inverse to each other.
 2.3.1. TWO HA CONTEXTS 179

To finish the proof, it is enough to notice that there exists a commutative diagram
(up to a natural isomorphism)
Rh
Ho(C(k))op / Ho((C(k))∧ )corep
RRR
RRR
RRR i∗
RRR
R( 
Ho((C(k)op
≤0 ) )
∧ corep
.

The Yoneda embedding Rh being fully faithful, this implies that the functor
Ho(C(k))op −→ Ho((C(k)op ∧
≤0 ) )

is also fully faithful. 

Remark 2.3.1.2. There are objects A ∈ k − cdga which are not C(k)≤0 -good.
For example, we can take A with π∗ (A) ≃ k[T, T −1 ] where T is in degree 2. Then,
there is no nontrivial A-dg-module M such that M ∈ C(k)≤0 : this clearly implies
that A can not be C(k)≤0 -good.
Another example is given by A = k[T ] where T is in degree −2, and M be the
A-module A[T −1 ]. As there are no morphisms from M to A-modules in C(k)≤0 , then
M is sent to zero by the functor
Ho(A − M od)op −→ Ho((A − M odop ∧
≤0 ) ).

Definition 2.3.1.3. (1) Let A ∈ k − cdga, and M be an A-dg-module. The

A-dg-module M is strong if the natural morphism
π∗ (A) ⊗π0 (A) π0 (M ) −→ π∗ (M )
is an isomorphism.
(2) A morphism A −→ B in k − cdga is strongly flat (resp. strongly smooth,
resp. strongly étale, resp. a strong Zariski open immersion) if B is strong
as an A-dg-module, and if the morphism of affine schemes
Spec π0 (B) −→ Spec π0 (A)
is flat (resp. smooth, resp. étale, resp. a Zariski open immersion).
One of the main difference between derived algebraic geometry and complicial
algebraic geometry lies in the fact that the strong notions of flat, smooth, étale and
Zariski open immersion are not as easily related to the general notions presented in
§1.2. We have the following partial comparison result.
Proposition 2.3.1.4. Let f : A −→ B be a morphism in k − cdga.
(1) If A and B are (−1)-connected, the morphism f is smooth (resp. i-smooth,
resp. resp. étale, resp. a Zariski open immersion) in the sense of Def.
1.2.6.1, 1.2.7.1, if and only if f is strongly smooth (resp. strongly étale,
resp. a strong Zariski open immersion).
(2) If the morphism f is strongly flat (resp. strongly smooth, resp. strongly
étale, resp. a strong Zariski open immersion), then it is flat (resp. smooth,
resp. étale, resp. a Zariski open immersion) in the sense of Def. 1.2.6.1,
1.2.7.1.
Proof. (1) The proof is precisely the same as for Thm. 2.2.2.6, as the homotopy
theory of (−1)-connected k − cdga is equivalent to the one of commutative simplicial
k-algebras, and as this equivalence preserves cotangent complexes.
180 2.3. COMPLICIAL ALGEBRAIC GEOMETRY

(2) For flat morphisms there is nothing to prove, as all morphisms are flat in the
sense of Def. 1.2.6.1 since the model category C(k) is stable. Let us suppose that
f : A −→ B is strongly smooth (resp. strongly étale, resp. a strong Zariski open
immersion). we can consider the morphism induced on the (−1)-connected covers
f ′ : A′ −→ B ′ ,
where we recall that the (−1)-connected cover A′ −→ A induces isomorphims on πi
for all i ≥ 0, and is such that πi (A′ ) = 0 for all i < 0. As the morphism f is strongly
flat that the square
f′
A′ / B′

 
A /B
f

is homotopy co-cartesian. Therefore, as our notions of smooth, étale and Zariski open
immersion are stable by homotopy push-out, it is enough to show that f ′ is smooth
(resp. étale, resp. a Zariski open immersion). But this follows from (1). 
Example 2.3.1.5. Before going further into complicial algebraic geometry we
would like to present an example illustrating the difference between the strong notions
of flat, smooth, étale and Zariski open immersion and the general notions presented
in §1.2, showing in particular that proposition 2.3.1.4 (2) does not have a converse.
Let A be any commutative k-algebra, X = Spec A the corresponding affine
scheme, and U ⊂ X be a quasi-compact open subscheme. It is easy to see that
there exists a perfect complex of A-modules K, such that U is the open subscheme of
X on which K is acyclic. By Prop. 1.2.10.1 there exists then a morphism A −→ AK
in k − cdga which is a Zariski open immersion. Moreover, the universal property
of A −→ AK shows that if AK is cohomologically concentrated in degree 0 then U
is affine and we have U ≃ Spec π0 (AK ). Therefore, as soon as U is not affine, AK
cannot be concentrated in degree 0. The morphism A −→ AK is thus a Zariski open
immersion, and therefore étale, but is not a strong morphism.
This example also shows that if the scheme U is considered as a scheme over C(k)
(see §2.3.5.1), then U is equivalent to RSpec AK , and thus is affine as a stack over
C(k), even though U is not necessarily an affine subscheme of X over k.

The opposite model category k − cdgaop will be denoted by k − DAf f . We will

endow it with the strong étale model topology.
Definition 2.3.1.6. A family of morphisms {Spec Ai −→ Spec A}i∈I in k −
DAf f is a strong étale covering family (or simply s-ét covering family) if it satisfies
the following two conditions.
(1) Each morphism A −→ Ai is strongly étale.
(2) There exists a finite subset J ⊂ I such that the family {A −→ Ai }i∈J is a
formal covering family in the sense of 1.2.5.1.
Using the definition of strong étale morphisms, we immediately check that a
family of morphisms {Spec Ai −→ Spec A}i∈I in k − DAf f is a s-ét covering family if
and only if there exists a finite sub-set J ⊂ I satisfying the following two conditions.
• For all i ∈ I, the natural morphism
π∗ (A) ⊗π0 (A) π0 (Ai ) −→ π∗ (Ai )
is an isomorphism.
 2.3.2. WEAKLY GEOMETRIC D-STACKS 181

• The morphism of affine schemes

a
Spec π0 (Ai ) −→ Spec π0 (A)
i∈J
is étale and surjective.
Lemma 2.3.1.7. The s-ét covering families define a model topology on k − DAf f ,
which satisfies assumption 1.3.2.2.
Proof. The same as for Lem. 2.2.2.13. 
The model topology s-ét gives rise to a model category of stacks k − DAf f ∼,s-ét .
Definition 2.3.1.8. (1) A D-stack is an object F ∈ k − DAf f ∼,s-ét which
is a stack in the sense of Def. 1.3.2.1.
(2) The model category of D-stacks is k − DAf f ∼,s-ét , and its homotopy cate-
gory is simply denoted by DSt(k).
From Prop. 2.3.1.4 we get the following generalization of Cor. 2.2.2.9.
Corollary 2.3.1.9. Let A ∈ k − cdga and A′ −→ A be its (−1)-connected cover.
Let us consider the natural morphisms
t0 (X) := Spec (π0 A) −→ X ′ = Spec A′ X = Spec A −→ X ′ := SpecA′ .
Then, the homotopy base change functors
Ho(k − DAf f /X ′ ) −→ Ho(k − DAf f /t0 (X))
Ho(k − DAf f /X ′) −→ Ho(k − DAf f /X)
induces equivalences between the full sub-categories of strongly étale morphisms. Fur-
thermore, these equivalences preserve epimorphisms of stacks.
Proof. Using Cor. 2.2.2.9 we see that it is enough to show that the base change
along the connective cover A′ −→ A
Ho(k − DAf f /X ′) −→ Ho(k − DAf f /X)
induces an equivalences from the full sub-categories of strongly étale morphism Y ′ →
X ′ to the full subcategory of étale morphisms Y → X. But an inverse to this functor
is given by sending a strongly étale morphism A −→ B to its connective cover A′ −→
B′. 

2.3.2. Weakly geometric D-stacks

We now let Pw be the class of formally perfect morphisms in k − DAf f , also
called in the present context weakly smooth morphisms.
Lemma 2.3.2.1. The class Pw of fp morphisms and the s-ét model topology satisfy
assumptions 1.3.2.11.
Proof. The only non-trivial thing to show is that the notion of fp morphism
is local for the s-ét topology on the source. As strongly étale morphisms are also
formally étale, this easily reduces to showing that being a perfect module is local for
the s-ét model topology. But this follows from corollary 1.3.7.4. 
We can now state that (C(k), C(k), k − cdga, s-ét, Pw ) is a HAG context in the
sense of Def. 1.3.2.13. From our general definitions we obtain a first notion of geo-
metric D-stacks.
Definition 2.3.2.2. (1) A weakly n-geometric D-stack is a a D-stack F ∈
DSt(k) which is n-geometric for the HAG context (C(k), C(k), k−cdga, s-ét, Pw ).
182 2.3. COMPLICIAL ALGEBRAIC GEOMETRY

(2) A weakly n-representable morphism of D-stacks is an n-representable mor-

phism for the HAG context (C(k), C(k), k − cdga, s-ét, Pw ).
In the context (C(k), C(k), k − cdga, s-ét, Pw ) the i-smooth morphisms are the
formally etale morphisms, and therefore Artin’s conditions of Def. 1.4.3.1 can not be
satisfied. There are actually many interesting weakly geometric D-stacks which do
not have cotangent complexes, as we will see.

2.3.3. Examples of weakly geometric D-stacks

2.3.3.1. Perfect modules. We consider the D-stack Perf , as defined in Def.
1.3.7.5. As we have seen during the proof of Lem. 2.3.2.1 the notion of being perfect
is local for the s-ét topology. In particular, for any A ∈ k − cdga, the simplicial set
Perf (A) is naturally equivalent to the nerve of the category of equivalences between
perfect A-dg-modules.
Proposition 2.3.3.1. (1) The D-stack Perf is categorically locally of finite
presentation in the sense of Def. 1.3.6.4.
(2) The D-stack Perf is weakly 1-geometric. Furthermore its diagonal is (−1)-
representable.
Proof. (1) Let A = Colimi∈I Ai be a filtered colimit of objects in k − cdga. We
need to prove that the morphism
Colimi Perf (Ai ) −→ Perf (A)
is an equivalence. Let j ∈ I, and two perfect Aj -dg-modules P and Q. Then, we have
M apA−Mod (A ⊗LAj P, A ⊗LAj Q) ≃ M apAj −Mod (P, A ⊗LAj Q) ≃

≃ M apAj −Mod (P, Colimi∈j/J Ai ⊗LAj Q).

As P is perfect it is finitely presented in the sense of Def. 1.2.3.1, and thus we get
M apA−Mod (A ⊗LAj P, A ⊗LAj Q) ≃ Colimi∈j/I M apAi −Mod (Ai ⊗LAj P, Ai ⊗LAj Q).
Furthermore, as during the proof of Prop. 1.3.7.14 we can show that the same formula
holds when M ap is replaced by the sub-simplicial set M apeq of equivalences. Invoking
the relations between mapping spaces and loop spaces of nerves of model categories
(see Appendix B), this clearly implies that the morphism
Colimi Perf (Ai ) −→ Perf (A)
induces isomorphisms on πi for i > 0 and an injective morphism on π0 .
It only remains to show that for any perfect A-dg-module P , there exists j ∈ I
and a perfect Aj -dg-module Q such that P ≃ A ⊗LAj Q. For this, we use that perfect
modules are precisely the retract of finite cell modules. By construction, it is clear
that a finite cell A-dg-module is defined over some Aj . We can therefore write P as
a direct factor of A ⊗LAj Q for some j ∈ I and Q a perfect Aj -dg-module. The direct
factor P is then determined by a projector
p ∈ [A ⊗LAj Q, A ⊗LAj Q] ≃ Colimi∈j/I [Ai ⊗LAj Q, Ai ⊗LAj Q].

Thus, p defines a projector pi in [Ai ⊗LAj Q, Ai ⊗LAj Q] for some i, corresponding to a

direct factor Pi of Ai ⊗LAj Q. Clearly, we have

P ≃ A ⊗LAi Pi .
 2.3.3. EXAMPLES OF WEAKLY GEOMETRIC D-STACKS 183

(2) We start by showing that the diagonal of Perf is (−1)-representable. In other

words, we need to prove that for any A ∈ k − cdga, and any two perfect A-dg-modules
M and N , the D-stack
Eq(M, N ) : A − cdga −→ SSetV
(A → B) 7→ M apeq L L
B−Mod (M ⊗A B, N ⊗A B),

(where M apeq denotes the sub-simplicial set of M ap consisting of equivalences) is a

representable D-stack. This is true, and the proof is exactly the same as the proof of
Prop. 1.3.7.14.
To construct a 1-atlas, let us chose a U-small set F of representative for the
isomorphism classes of finitely presented objects in Ho(k − cdga). Then, for any
A ∈ F, let MA be a U-small set of representative for the isomorphism classes of
perfect objects in Ho(A − M oddg ). The fact that these U-small sets F and M exists
follows from the fact finitely presented objects are retracts of finite cell objects (see
Cor. 1.2.3.8).
We consider the natural morphism
a a
p : U := UA,P := RSpec A −→ Perf .
A∈F P ∈MA

By construction the morphism p is an epimorphism of stacks. Furthermore, as the

diagonal of Perf is (−1)-representable each morphism UA,P = RSpec A −→ Perf
is (−1)-representable. Finally, as UA,P = RSpec A and Perf are both categor-
ically locally of finite presentation, we see that for any B ∈ k − cdga and any
Y := RSpec B −→ Perf , the morphism
UA,P ×hPerf Y −→ Y
is a finitely presented morphism between representable D-stacks. In particular, it is
a perfect morphism. We therefore conclude that
a a
p : U := UA,P := RSpec A −→ Perf
A∈F P ∈MA

is a 1-atlas for Perf . This finishes the proof that Perf is weakly 1-geometric. 
2.3.3.2. The D-stacks of dg-algebras and dg-categories. We consider for
any A ∈ k − cdga the model category A − dga, of associative and unital A-algebras
(in U). The model structure on A − dga is the usual one for which equivalences
are quasi-isomorphisms and fibrations are epimorphisms. We consider Ass(A) to be
the subcategory of A − dga consisting of equivalences between objects B ∈ A − dga
satisfying the following two conditions.
• The object B is cofibrant in A − dga.
• The underlying A-dg-module of B is perfect.
For a morphism A −→ A′ in k − cdga, we have a base change functor
A′ ⊗A − : Ass(A) −→ Ass(A′ ),
making A 7→ Ass(A) into a pseudo-functor on k −cdga. Using the usual strictification
procedure, and passing to the nerve we obtain a simplicial presheaf
Ass : k − DAf f −→ SSetV
A 7→ N (Ass(A)).
Proposition 2.3.3.2. (1) The simplicial presheaf Ass is a D-stack.
(2) The natural projection Ass −→ Perf , which forget the algebra structure, is
(−1)-representable.
184 2.3. COMPLICIAL ALGEBRAIC GEOMETRY

Proof. (1) This is exactly the same proof as Thm. 1.3.7.2.

(2) Let A ∈ k − cdga and RSpec A −→ Perf be a point corresponding to a perfect

^
A-dg-module E. We denote by Ass E the homotopy fiber taken at E of the morphism
Ass −→ Perf .
^
Lemma 2.3.3.3. The D-stack Ass E is representable.

Proof. This is the same argument as for Prop. 2.2.6.8. We see using [Re, Thm.
^
1.1.5] that Ass E is the D-stack (over Spec A) M ap(Ass, REnd(E)), of morphisms
from the associative operad to the the (derived) endomorphism operad of E, defined
the same way as in the proof of 2.2.6.8. Again the same argument as for 2.2.6.8,
consisting of writing Ass as a homotopy colimit of free operads, reduces the statement
of the lemma to prove that for a perfect A-dg-module K of, the D-stack
A − cdga −→ SSet
A′ 7→ M ap(C(k)) (K, A′ )
is representable. But this is true as it is equivalent to RSpec B, where B is the
(derived) free A-cdga over K. 
The previous lemma shows that Ass −→ Perf is a (−1)-representable morphism,
and finishes the proof of Prop. 2.3.3.2. 
Corollary 2.3.3.4. The D-stack Ass is weakly 1-geometric.
Proof. Follows from Prop. 2.3.3.1 and Prop. 2.3.3.2. 
We now consider a slight modification of Ass, by considering dg-algebras as dg-
categories with only one objects. For this, let A ∈ k − cdga. Recall that a A-dg-
category is by definition a category enriched over the symmetric monoidal category
A − M oddg , of A-dg-modules. More precisely, a A-dg-category D consists of the
following data
• A set of objects Ob(D).
• For any pair of objects (x, y) in Ob(D) an A-dg-module D(x, y).
• For any triple of objects (x, y, z) in Ob(D) a composition morphism
D(x, y) ⊗A D(y, z) −→ D(x, z),
which satisfies the usual unital and associativity conditions.
The A-dg-categories (in the universe U) form a category A − dgCat, with the
obvious notion of morphisms. For an A − dg-category D, we can form a category
π0 (D), sometimes called the homotopy category of D, whose objects are the same
as D and for which morphisms from x to y is the set π0 (D(x, y)) (with the obvious
induced compositions). The construction D 7→ π0 (D) defines a functor from A−dgCat
to the category of U-small categories. Recall that a morphism f : D −→ E is then
called a quasi-equivalence (or simply an equivalence) if it satisfies the following two
conditions.
• For any pair of objects (x, y) in D the induced morphism
fx,y : D(x, y) −→ E(f (x), f (y))
is an equivalence in A − M oddg .
• The induced functor
π0 (f ) : π0 (D) −→ π0 (E)
is an equivalence of categories.
 2.3.3. EXAMPLES OF WEAKLY GEOMETRIC D-STACKS 185

We also define a notion of fibration, as the morphisms f : D −→ E satisfying the

following two conditions.
• For any pair of objects (x, y) in D the induced morphism
fx,y : D(x, y) −→ E(f (x), f (y))
is a fibration in A − M oddg .
• For any object x in D, and any isomorphism v : f (x) → z in π0 (E), there
exists an isomorphism u : x → y in π0 (D) such that π0 (f )(u) = v.
With these notions of fibrations and equivalences, the category A − dgCat is a
model category. This is proved in [Tab] when A is a commutative ring. The general
case of categories enriched in a well behaved monoidal model category has been worked
out recently by J. Tapia (private communication).
For A ∈ k − cdga, we denote by Cat∗ (A) the category of equivalences between
objects D ∈ A − dgCat satisfying the following two conditions.
• For any two objects x and y in D the A-dg-module D(x, y) is perfect and
cofibrant in A − M oddg .
• The category π0 (D) possesses a unique object up to isomorphism.
For a morphism A −→ A′ in k − cdga, we have a base change functor
− ⊗A A′ : Cat∗ (A) −→ Cat∗ (A′ )
obtained by the formula
Ob(D ⊗A A′ ) := Ob(D) (D ⊗A A′ )(x, y) := D(x, y) ⊗A A′ .
This makes A 7→ Cat∗ (A) into a pseudo-functor on k−cdga. Strictifying and applying
the nerve construction we get a well defined simplicial presheaf
Cat∗ : k − cdga −→ SSetV
A 7→ N (Cat∗ (A)).
It is worth mentioning that Cat∗ is not a stack, since there are non trivial twisted
forms of objects in Cat∗ (k) with respect to the étale topology on k. These twisted
forms can be interpreted as certain stacks in dg-categories having locally a unique
object up to equivalences, but they might have either no global objects or several
nonequivalent global objects. We will not explicitly describe the stack associated to
Cat∗ , as this will be irrelevant for the sequel, and will simply consider Cat∗ as an
object in DSt(k).
There exists a morphism of simplicial presheaves
B : Ass −→ Cat∗ ,
which sends an associative A-algebra C to the A-dg-category BC, having a unique
object ∗ and C as the endomorphism A-algebra of ∗. The morphism B will be
considered as a morphism in DSt(k).
Proposition 2.3.3.5. The morphism B : Ass −→ Cat∗ is weakly 1-representable,
fp and an epimorphism of D-stacks.
Proof. We will prove a more precise result, giving explicit description of the ho-
motopy fibers of B. For this, we start by some model category considerations relating
associative dga to dg-categories. Recall that for any A ∈ k − cdga, we have two model
categories, A − dga and A − dgCat, of associative A-algebras and A-dg-categories.
We consider 1/A − dgCat, the model category of dg-categories together with a dis-
tinguised object. More precisely, 1 is the A-dg-category with a unique object and
A as its endomorphism dg-algebra, and 1/A − dgCat is the comma model category.
The functor B : A − dga −→ 1/A − dgCat is a left Quillen functor. Indeed, its right
186 2.3. COMPLICIAL ALGEBRAIC GEOMETRY

adjoint Ω∗ , sends a pointed A-dg-category D to the A-algebra D(∗, ∗) of endomor-

phisms of the distinguished point ∗. Clearly, the adjunction morphism Ω∗ B ⇒ Id is
an equivalence, and thus the functor
B : Ho(A − dga) −→ Ho(1/A − dgCat)
is fully faithful. As a consequence, we get that for any C, C ′ ∈ A − dga there exists a
natural homotopy fiber sequence of simplicial sets
M apA−dgCat (1, BC ′ ) −→ M apA−dga (B, B ′ ) −→ M apA−dgCat (BC, BC ′ ).
Now, let A ∈ k − cgda and B be an associative A-algebra, corresponding to a
morphism
x : X := RSpec A −→ Ass.
We consider the D-stack
F := Ass ×hCat∗ X.
Using the relations between mapping spaces in A − dga and A − dgCat above we
easily see that the D-stack F is connected (i.e. π0 (F ) ≃ ∗). In particular, in order to
show that F is weakly 1-geometric, it is enough to prove that ΩB F is a representable
D-stack. Using again the homotopy fiber sequence of mapping spaces above we see
that the D-stack ΩB F can be described as
ΩB F : A − cdga −→ SSetV
(A → A′ ) 7→ M ap1/A′ −dgCat (S 1 ⊗L 1, B(B ⊗LA A′ ))

where S 1 ⊗L 1 is computed in the model category A − dgCat. One can easily check
that one has an isomorphism in 1/A − dgCat
S 1 ⊗L 1 ≃ B(A[T, T −1 ]),
where A[T, T −1 ] := A ⊗k k[T, T −1]. Therefore, there is a natural equivalence
M ap1/A′ −dgCat (S 1 ⊗L 1, B(B ⊗LA A′ )) ≃ M apA−dga (A[T, T −1 ], B ⊗LA A′ ),
and thus the D-stack ΩB F can also be described by
ΩB F : A − cdga −→ SSetV
(A → A′ ) 7→ M apA−dga (A[T, T −1 ], B ⊗LA A′ ).

The morphism of k-algebras k[T ] −→ k[T, T −1 ] induces natural morphisms (here B ∨

is the dual of the perfect A-dg-module B)
M apA−dga (A[T, T −1 ], B⊗LA A′ ) −→ M apA−dga (A[T ], B⊗LA A′ ) ≃ M apA−Mod (B ∨ , A′ ).
It is not difficult to check that this gives a morphism of D-stacks
ΩB F −→ RSpec LF (B ∨ ),
where LF (B ∨ ) is the derived free A-cdga over the A-dg-module B ∨ . Furthermore,
applying Prop. 1.2.10.1 this morphism is easily seen to be representable by an open
Zariski immersion, showing that ΩB F is thus a representable D-stack. 

Corollary 2.3.3.6. The D-stack Cat∗ is weakly 2-geometric.

Proof. This follows immediately from Cor. 2.3.3.4, Prop. 2.3.3.5 and the gen-
eral criterion of Cor. 1.3.4.5. 
 2.3.4. GEOMETRIC D-STACKS 187

Let A ∈ k − cdga and B be an associative A-algebra corresponding to a morphism

of D-stacks
B : X := RSpec A −→ Ass.
We define a D-stack B ∗ on A − cdga in the following way
B∗ : A − cdga −→ SSetV
(A → A′ ) 7→ M apA−dga (A[T, T −1 ], B ⊗LA A′ ).
The D-stack B ∗ is called the D-stack of invertible elements in B. The D-stack
B ∗ possesses in fact a natural loop stack structure (i.e. the above functor factors
naturally, up to equivalence, through a functor from A − cdga to the category of
loop spaces), induced by the Hopf algebra structure on A[T, T −1 ]. This loop stack
structure is also the one induced by the natural equivalences
B ∗ (A′ ) ≃ Ω∗ M apA−dgCat (∗, B ⊗LA A′ ).
Delooping gives another D-stack
K(B ∗ , 1) : A − cdga −→ SSetV
(A → A′ ) 7→ K(B ∗ (A′ ), 1).
The following corollary is a reformulation of Prop. 2.3.3.5 and of its proof.
Corollary 2.3.3.7. Let A ∈ k − cdga and B ∈ Ass(A) corresponding to an
associative A-algebra B. Then, there is a natural homotopy cartesian square of D-
stacks
Ass / Cat∗
O O
B

K(B ∗ , 1) / RSpec A.

2.3.4. Geometric D-stacks

We now switch to the HA context (C(k), C(k)≤0 , k − cdga0 ). Recall that C(k)≤0
is the subcategory of C(k) consisting of (−1)-connected object, and k − cdga0 is the
subcategory of k − cdga of objects cohomologically concentrated in degree 0. Within
this HA context we let P to be the class of formally perfect and formally i-smooth
morphisms in k−DAf f . Morphisms in P will simply be called fip-smooth morphisms.
There are no easy description of them, but Prop. 1.2.8.3 implies that a morphism
f : A −→ B be a morphism is fip-smooth if it satisfies the following two conditions.
• The cotangent complex LB/A ∈ Ho(A − M oddg ) is perfect.
• For any R ∈ k − cdga0 , any connected module M ∈ R − M oddg and any
morphism B −→ R, one has
π0 (L∨ L
B/A ⊗B M ) = 0.

The converse is easily shown to be true if A and B are both (−1)-connected.

Lemma 2.3.4.1. The class P of fip-smooth morphisms and the s-ét model topology
satisfy assumptions 1.3.2.11.
Proof. We see that the only non trivial part is to show that the notion of fip-
smooth morphism is local for the s-ét topology. For fp-morphisms this is Cor. 1.3.7.4.
For fi-smooth morphisms this is an easy consequence of the definitions. 
From Lem. 2.3.4.1 we get a HAG context (C(k), C(k)≤0 , k − cdga0 , s − et, P).
Definition 2.3.4.2. (1) An n-geometric D-stack is a D-stack F ∈ DSt(k)
which is n-geometric for the HAG context (C(k), C(k)≤0 , k−cdga0 , s−et, P).
188 2.3. COMPLICIAL ALGEBRAIC GEOMETRY

(2) A strongly n-representable morphism of D-stacks is an n-representable mor-

phism of D-stacks for the HAG context (C(k), C(k)≤0 , k − cdga0 , s − et, P).
Note that P ⊂ Pw and therefore that any n-geometric D-stack is weakly n-
geometric.
Lemma 2.3.4.3. The s-ét topology and the class P of fip-smooth morphisms satisfy
Artin’s conditions of Def. 1.4.3.1.
Proof. This is essentially the same proof as Prop. 2.2.3.2, and is even more
simple as one uses here the strongly étale topology. 

Corollary 2.3.4.4. Any n-geometric D-stack possesses an obstruction theory

(relative to the HA context (C(k), C(k)≤0 , k − cdga0 )).
Proof. This follows from Lem. 2.3.4.3 and Thm. 1.4.3.2. 

2.3.5. Examples of geometric D-stacks

−
2.3.5.1. D -stacks and D-stacks. We consider the normalization functor N :
sk − Alg −→ k − cdga, sending a simplicial commutative k-algebra A to its normal-
ization N (A), with its induced structure of commutative differential graded algebra.
The pullback functor gives a Quillen adjunction
N! : k−D− Af f ∼,et −→ k−DAf f ∼,s−et k−D− Af f ∼,et ←− k−DAf f ∼,s−et : N ∗ .
As the functor N is known to be homotopically fully faithful, the left derived
functor
j := LN! : D− St(k) −→ DSt(k)
is fully faithful. We can actually characterize the essential image of the functor j as
consisting of all D-stacks F for which for any A ∈ k − cdga with (−1)-connected cover
A′ −→ A, the morphism RF (A′ ) −→ RF (A) is an equivalence. In other words, for
any F ∈ k − D− Af f ∼,et , and any A ∈ k − cdga, we have
Rj(F )(A) ≃ RF (D(A′ )),
where D(A′ ) ∈ sk − Alg is a denormalization of A′ , the (−1)-connected cover of A.
Proposition 2.3.5.1. (1) For any A ∈ sk − Alg, we have
j(RSpec A) ≃ RSpec N (A).
(2) The functor j commutes with homotopy limits and homotopy colimits.
(3) The functor j sends n-geometric D− -stacks to n-geometric D-stacks.
Proof. This is clear. 

The previous proposition shows that any n-geometric D− -stack gives rise to an
n-geometric D-stack, and thus provides us with a lot of examples of those.

2.3.5.2. CW-perfect modules. Let A ∈ k − cdga. We define by induction on

n = b − a the notion of a perfect CW-A-dg-module of amplitude contained in [a, b].
Definition 2.3.5.2. (1) A perfect CW-A-dg-module of amplitude contained
in [a, a] is an A-dg-module M isomorphic in Ho(A − M oddg ) to P [−a], with
P a projective and finitely presented A-dg-module (as usual in the sense of
definitions 1.2.3.1 and 1.2.4.1).
 2.3.5. EXAMPLES OF GEOMETRIC D-STACKS 189

(2) Assume that the notion of perfect CW-A-dg-module of amplitude contained

in [a, b] has been defined for any a ≤ b such that b − a = n − 1. A perfect
CW-A-dg-module of amplitude contained in [a, b], with b − a = n, is an
A-dg-module M isomorphic in Ho(A − M oddg ) to the homotopy cofiber of a
morphism
P [−a − 1] −→ N,
where P is projective and finitely presented, and N is a perfect CW-A-dg-
module of amplitude contained in [a + 1, b].
The perfect CW-A-dg-modules satisfy the following stability conditions.
Lemma 2.3.5.3. (1) If M is a perfect CW-A-dg-module of amplitude con-
tained in [a, b], and A −→ A′ is a morphism in k − cdga, then A′ ⊗LA M is
a perfect CW-A-dg-module of amplitude contained in [a, b].
(2) Let A ∈ k − cdga≤0 be a (−1)-connected k − cdga. Then, any perfect A-
dg-module is a perfect CW-A-dg-module of amplitude [a, b] for some integer
a ≤ b.
(3) Let A ∈ k − cdga≤0 be a (−1)-connected k − cdga, and M ∈ A − M oddg .
If there exists a s-et covering A −→ A′ such that A′ ⊗LA M is a perfect
CW-A-dg-module of amplitude contained in [a, b], then so is M .
Proof. Only (2) and (3) requires a proof. Furthermore, (3) clearly follows from
(2) and the local nature of perfect modules (see Cor. 1.3.7.4), and thus it only remains
to prove (2). But this is proved in [EKMM, III.7]. 

We define a sub-D-stack Perf CW [a,b] ⊂ Perf , consisting of all perfect modules locally
equivalent to some CW-dg-modules of amplitude contained in [a, b]. Precisely, for
A ∈ k − cdga, Perf CW[a,b] (A) is the sub-simplicial set of Perf (A) which is the union
of all connected components corresponding to A-dg-modules M such that there is
an s-ét covering A −→ A′ with A′ ⊗LA M a perfect CW-A′ -dg-module of amplitude
contained in [a, b].
Proposition 2.3.5.4. The D-stack Perf CW
[a,b] is 1-geometric.

Proof. We proceed by induction on n = b − a. For n = 0, the D-stack Perf CW [a,a]

is simply the stack of vector bundles Vect, which is 1-geometric by our general result
Cor. 1.3.7.12. Assume that Perf CW [a,b] is known to be 1-geometric for b − a < n.
Shifting if necessary, we can clearly assume that a = 0, and thus that b = n − 1.
That the diagonal of Perf CW [a,b] is (−1)-representable comes from the fact that
CW
Perf [a,b] −→ Perf is a monomorphisms and from Prop. 2.3.3.1. It remains to
construct a 1-atlas for Perf CW
[a,b] .
Let A ∈ k − cdga. We consider the model category M or(A − M oddg ), whose
objects are morphisms in A − M oddg (and with the usual model category structure,
see e.g. [Ho1]). We consider the subcategory M or(A − M oddg )′ consisting cofibrant
objects u : M → N in M or(A−M oddg ) (i.e. u is a cofibration between cofibrant A-dg-
modules) such that M is projective of finite presentation and N is locally a perfect
CW-A-dg-module of amplidtude contained in [0, n − 1]. Morphisms in M or(A −
M oddg )′ are taken to be equivalences in M or(A − M oddg ). The correspondence
A 7→ M or(A − M oddg )′ defines a pseudo-functor on k − cdga, and after strictification
and passing to the nerve we get this way a simplicial presheaf
F : k − cdga −→ SSetV
A 7→ N (M or(A − M oddg )′ ).
190 2.3. COMPLICIAL ALGEBRAIC GEOMETRY

There exists a morphism of D-stacks

F −→ Vect ×h Perf CW
[0,n−1]

that sends an object M → N to (M, N ). This morphism is easily seen to be (−1)-

representable, as its fiber at an A-point (M, N ) is the D-stack of morphisms from M
to N , or equivalently is RSpec B where B is the derived free A-cdga over M ∨ ⊗LA N .
By induction hypothesis we deduce that the D-stack F itself if 1-geometric.
Finally, there exists a natural morphism of D-stacks
p : F −→ Perf CW
[−1,n−1]

sending a morphism u : M → N in M or(A − M oddg )′ to its homotopy cofiber. The

morphism p is an epimorphism of D-stacks by definition of being locally a perfect
CW-dg-module, thus it only remains to show that p is fip-smooth. Indeed, this would
imply the existence of a 1-atlas for Perf CW[−1,n−1] , and thus that it is 1-geometric.
CW
Translating we get that Perf [a,b] is 1-geometric for b − a = n.
In order to prove that p is fip-smooth, let A ∈ k − cdga, and K : RSpec A −→
Perf CW
[−1,n−1] be an A-point, corresponding to a perfect CW-A-dg-module of amplitude
contained in [−1, n − 1]. We consider the homotopy cartesian square
p
FO / Perf CW
[−1,n−1]
O

F′ / RSpec A.
p′

We need to prove that p is a fip-smooth morphism. The D-stack F ′ has natural

projection F ′ −→ Vect given by F ′ −→ F −→ Vect, where the second morphism
sends a morphism M → N to the vector bundle M . We therefore get a morphism of
D-stacks
F ′ −→ Vect ×h RSpec A −→ RSpec A.
As the morphism Vect −→ ∗ is smooth, it is enough to show that the morphism
F ′ −→ Vect ×h RSpec A
is fip-smooth. For this we consider the homotopy cartesian square

FO ′ / Vect ×h RSpec A
O

F0′ / RSpec A,

where the section RSpec A −→ Vect correspond to a trivial rank r vector bundle
Ar . It only remains to show that the morphism F0′ −→ RSpec A is fip-smooth. The
D-stack F0′ over A can then be easily described (using for example our general Cor.
B.0.8) as
F0′ : A − cdga −→ SSetV
(A → A′ ) 7→ M apA−Mod (K[−1], Ar ).
In other words, we can write F0′ ≃ RSpec B, where B is the derived free A − cdga
over (K[−1])r . But, as K is a perfect CW-A-dg-module of amplitude contained in
[−1, n − 1], (K[−1])r is a perfect CW A-dg-module of amplitude contained in [0, n].
The proposition will then follow from the general lemma.
 2.3.5. EXAMPLES OF GEOMETRIC D-STACKS 191

Let A ∈ k − cdga, and recall that the forgetful functor A − cdga −→ A − M oddg
is right Quillen, and that its derived left adjoint
Ho(A − M oddg ) −→ Ho(A − cdga)
sends, by definition, an A-module E to the derived free A-cdga over E.
Lemma 2.3.5.5. Let A ∈ k − cdga, and K be a perfect CW-A-dg-module of am-
plitude contained in [0, n]. Let B be the derived free A − cdga over K, and
p : Y := RSpec B −→ X := RSpec A
be the natural projection. Then p is fip-smooth.
Proof. Let k ′ be any commutative k-algebra, and B −→ k be any morphism
in Ho(k − cdga), corresponding to a point x : Spec k −→ X. Then, we have an
isomorphism in D(k ′ )
LY /X,x ≃ K ⊗LA k ′ .
We thus see that LY /X,x is a perfect complex of k ′ -modules of Tor amplitude concen-
trated in degrees [0, n]. In particular, it is clear that for any complex M of k ′ -modules
such that πi (M ) = 0 for any i > 0, then we have
[LY /X,x , M ]D(k′ ) ≃ π0 (L∨ L
Y /X,x ⊗k′ M ) ≃ 0.
This shows that the A-algebra B is fip-smooth, and thus shows the lemma. 
The previous lemma finishes the proof of the proposition. 
A direct consequence of Prop. 2.3.5.4 and Thm. 1.4.3.2 is the following.
Corollary 2.3.5.6. The D-stack Perf CW [a,b] has an obstruction theory (relative
to the HA context (C(k), C(k)≤0 , k − cdga0 ). For any A ∈ k − cdga0 and any point
E : X := RSpec A −→ Perf CW[a,b] corresponding to a perfect CW-A-dg-module E, there
are natural isomorphisms in Ho(A − M oddg )
LPerf CW ,E ≃ E ∨ ⊗LA E[−1]
[a,b]

TPerf CW ,E ≃ E ∨ ⊗LA E[1].

[a,b]

Proof. The first part of the corollary follows from our Thm. 1.4.3.2 and Prop.
2.3.5.4. Let A and E : X := RSpec A −→ Perf CW [a,b] as in the statement. We have

LPerf CW ,E ≃ LΩE Perf CW ,E [−1].

[a,b] [a,b]

Moreover, ΩE Perf CW
[a,b]≃ RAut(E), where RAut(E) is the D-stack of self-equivalences
of the perfect module E as defined in §1.3.7. By Prop. 1.3.7.14 we know that RAut(E)
is representable, and furthermore that the natural inclusion morphism
RAut(E) −→ REnd(E)
is a formally étale morphism of representable D-stacks. Therefore, we have
LΩE Perf CW ,E ≃ LREnd(E),Id .
[a,b]

Finally, we have
REnd(E) ≃ RSpec B,
where B is the derived free A-cdga over E ∨ ⊗LA E. This implies that
LREnd(E),Id ≃ E ∨ ⊗LA E,
and by what we have seen that
LPerf CW ,E ≃ E ∨ ⊗LA E[−1].
[a,b]
192 2.3. COMPLICIAL ALGEBRAIC GEOMETRY


Remark 2.3.5.7. (1) It is important to note that, if
N ∗ : DSt(k) −→ D− St(k)
denotes the restriction functor, we have
N ∗ (Perf CW
[a,b] ) ≃ Perf [a,b] ,

where Perf [a,b] is the sub-stack of Perf ∈ D− St(k) consisting of perfect

modules of Tor-amplitude contained in [a, b]. However, the two D-stacks
Perf CW
[a,b] and j(Perf [a,b] ) are not the same. Indeed, for any A ∈ k − cdga
with A′ as (−1)-connective cover, we have
j(Perf [a,b] )(A) ≃ Perf [a,b] (A′ ).
In general the natural morphism
− ⊗LA A′ : Perf [a,b] (A′ ) −→ Perf CW
[a,b] (A)
is not an equivalence. For example, let us suppose that there exist a non
zero element x ∈ π−1 (A) ≃ [A, A[1]], then the matrix
 
IdA x
0 IdA[1]
L L
defines an equivalence A A[1] L ≃ A A[1] Lof ′A-dg-modules which is not
induced by an equivalence A′ A′ [1] ≃ A′ A [1] of A′ -dg-modules.
(2) We can also show that the D− -stack Perf [a,b] ∈ D− St(k) is (n+1)-geometric
for n = b − a. The proof is essentially the same as for Prop. 2.3.5.4 and will
not be reproduced here. Of course, the formula for the cotangent complex
remains the same. See [To-Va1] for more details.
2.3.5.3. CW-dg-algebras. Recall the existence of the following diagram of D-
stacks
Ass


Perf CW
[a,b]
/ Perf .

Definition 2.3.5.8. The D-stack of CW-dg-algebras of amplitude contained in

[a, b] is defined by the following homotopy cartesian square
AssCW
[a,b]
/ Ass

 
Perf CW
[a,b]
/ Perf .

Let B be an associative k-dga, and let M be a B-bi-dg-module (i.e. a B⊗Lk B op -dg-

module). We can form the square zero extension B ⊕ M , which is another associative
k-dga together with a natural projection B ⊕M −→ B. The simplicial set of (derived)
derivations from B to M is then defined as
RDerk (B, M ) := M apk−dga/B (B, B ⊕ M ).
The same kind of proof as for Prop. 1.2.1.2 shows that there exists an object LAss
B ∈
Ho(B ⊗Lk B op − M oddg ) an natural isomorphisms in Ho(SSetV )
RDerk (B, M ) ≃ M apB⊗Lk B op −Mod (LAss
B , M ).
 2.3.5. EXAMPLES OF GEOMETRIC D-STACKS 193

We then set
RDerk (B, M ) := RHomB⊗Lk B op −Mod (LAss
B , M ) ∈ Ho(C(k)),

where RHomB⊗Lk B op −Mod denotes the Ho(C(k))-enriched derived Hom’s of the C(k)-
model category B ⊗Lk B op − M oddg .
Corollary 2.3.5.9. The D-stack AssCW [a,b] is 1-geometric. For any global point
CW
B : Spec k −→ Ass[a,b] , corresponding to an associative k-dga B, one has a natural
isomorphism in Ho(C(k))
TAssCW ,B ≃ RDerk (B, B)[1].
[a,b]

Proof. Using Prop. 2.3.3.2 we see that the natural projection

CW
AssCW
[a,b] −→ Perf [a,b]

is (−1)-representable. Therefore, by Prop. 2.3.5.4 we know that AssCW [a,b] is 1-

geometric. In patricular, Thm. 1.4.3.2 implies that it has an obstruction theory
relative to the HA context (C(k), C(k)≤0 , k − cdga0 ). We thus have
TAssCW ,B ≃ TΩB AssCW ,B [1].
[a,b] [a,b]

The identification
TΩB AssCW ,B ≃ RDer k (B, B)
[a,b]

follows from the exact same argument as Prop. 2.2.6.9, using LAss
B instead of LO
B. 

2.3.5.4. The D-stack of negative CW-dg-categories. Recall from §2.3.3.2

the existence of the morphism of D-stacks
B : Ass −→ Cat∗ ,
sending an associative k-algebra C to the dg-category BC having a unique object and
C as endomorphisms of this object.
Definition 2.3.5.10. Let n ≤ 0 be an integer. The D-stack of CW-dg-categories
of amplitude contained in [n, 0] is defined as the full sub-D-stack CatCW
∗,[n,0] of Cat∗
consisting of the essential image of the morphism
AssCW
[n,0] −→ Ass −→ Cat∗ .

More precisely, for A ∈ k − cdga, one sets CatCW

∗,[n,0] (A) to be the sub-simplicial
set of Cat∗ (A) consisting of A-dg-categories D such that for any ∗ ∈ Ob(D) the A-
dg-module D(∗, ∗) is (locally) a perfect CW-A-dg-module of amplitude contained in
[n, 0].
Recall that for any associative k-dga B, one has a model category B ⊗Lk B op −
M oddg of B-bi-dg-modules. This model category is naturally tensored and co-tensored
over the symmetric monoidal model category C(k), making it into a C(k)-model
category in the sense of [Ho1]. The derived Ho(C(k))-enriched Hom’s of B ⊗Lk B op −
M oddg will then be denoted by RHomB⊗Lk B op .
Finally, for any associative k-dga B, one sets
HHk (B, B) := RHomB⊗Lk B op (B, B) ∈ Ho(C(k)),
where B is considered as B-bi-dg-module in the obvious way.
Theorem 2.3.5.11. (1) The morphism
B : AssCW CW
[n,0] −→ Cat∗,[n,0]
is a 1-representable fip-smooth covering of D-stacks.
194 2.3. COMPLICIAL ALGEBRAIC GEOMETRY

(2) The associated D-stack to CatCW ∗,[n,0] is a 2-geometric D-stack.

(3) If C is an associative k-dg-algebra, corresponding to a point C : Spec k −→
AssCW
[n,0] , then one has a natural isomorphism in Ho(C(k))

TCatCW ,BC ≃ HHk (C, C)[2].

∗,[n,0]

Proof. (1) Using Cor. 2.3.3.7, it is enough to show that for any A ∈ k − cdga,
and any associative A-algebra C, which is a perfect CW-A-dg-module of amplitude
contained in [n, 0], the morphism of D-stack K(C ∗ , 1) −→ RSpec A is 1-representable
and fip-smooth. For this it is clearly enough to show that the D-stack C ∗ is rep-
resentable and that the morphism C ∗ −→ RSpec A is fip-smooth. We already have
seen during the proof of Prop. 2.3.3.5 that C ∗ is representable and a Zariski open
sub-D-stack of RSpec F , where F is the free A-cdga over C ∨ . It is therefore enough
to see that A −→ F is fip-smooth, which follows from Lem. 2.3.5.5 as by assumption
C ∨ is a perfect CW-A-dg-module of amplitude contained in [0, n].

(2) Follows from Cor. 2.3.5.9 and Cor. 1.3.4.5.

(3) We consider a point C : Spec k −→ AssCW [n, 0], and the homotopy cartesian
square
AssCW / CatCW
O [n,0] ∗,[n,0]
O
C

K(C ∗ , 1) / Spec k.
By (2), Cor. 2.3.5.9 and Thm. 1.4.3.2 we know that all the stacks in the previous
square have an obstruction theory, and thus a cotangent complex (relative to the
HA context (C(k), C(k)≤0 , k − cdga0 )). Therefore, one finds a homotopy fibration
sequence of complexes of k-modules
TK(C ∗ ,1),∗ / TAssCW ,C / TCatCW ,BC
[n,0] ∗,[n,0]

that can also be rewritten as

C[1] / RDerk (C, C)[1] / TCatCW ,BC .
∗,[n,0]

The morphism C −→ RDer k (C, C) can be described in the following way. The C-
bi-dg-module LAss
C can be easily identified with the homotopy fiber (in the model
category of C ⊗Lk C op -dg-modules) of the multiplication morphism
C ⊗Lk C op −→ C.
The natural morphism LAss
C −→ C ⊗Lk C op then induces our morphism on the level of
derivations
C ≃ RHomC⊗Lk C op −Mod (C⊗Lk C op , C) −→ RHomC⊗Lk C op −Mod (LAss
C , B) ≃ RDer k (C, C).

In particular, we see that there exists a natural homotopy fiber sequence in C(k)
HHk (C, C) = RHomC⊗Lk C op −Mod (C, C) /C / RDer (C, C).
k

We deduce that there exists a natural isomorphism in Ho(C(k))

TCatCW ,BC [−2] ≃ HHk (C, C).
∗,[n,0]


 2.3.5. EXAMPLES OF GEOMETRIC D-STACKS 195

An important corollary of Thm. 2.3.5.11 is given by the following fact. It appears

in many places in the literature but we know of no references including a proof of
it. For this, we recall that for any commutative k-algebra k ′ we denote by Cat∗ (k ′ )
the category of equivalences between k ′ -dg-categories satisfying the following two
conditions.
• For any two objects x and y in D the complex of k ′ -module D(x, y) is perfect
(and cofibrant in C(k ′ )).
• The category π0 (D) possesses a unique object up to isomorphism.
[n,0]
We finally let Cat∗ (k ′ ) be the full subcategory of Cat∗ (k ′ ) consisting of objects D
such that for any two objects x and y the perfect complex D(x, y) has Tor amplitude
contained in [n, 0] for some n ≤ 0.
[n,0]
Corollary 2.3.5.12. Let D ∈ Cat∗ (k ′ ). Then, the homotopy fiber DefD ,
taken at the point D, of the morphism of simplicial sets
[n,0] [n,0]
N (Cat∗ (k[ǫ])) −→ N (Cat∗ (k))
is given by
DefD ≃ M apC(k) (k, HHk (D, D)[2]).
In particular, we have
πi (DefD ) ≃ HH2−i
k (D, D).

In the above corollary we have used HHk (D, D), the Hochschild complex of a
dg-category D. It is defined the same way as for associative dg-algebras, and when D
is equivalent to BC for an associative dg-alegbra C we have
HHk (D, D) ≃ HHk (C, C).
Finally, we would like to mention that restricting to negatively graded dg-categories
seems difficult to avoid if we want to keep the existence of a cotangent complex.
Corollary 2.3.5.13. Assume that k is a field. Let C ∈ AssCW [a,b] (k) be a k-point
corresponding to an associative dg-algebra C. If we have H i (C) 6= 0 for some i > 0
then the D-stack Cat∗ does not have a cotangent complex at the point BC.
Proof. Suppose that Cat∗ does have a cotangent complex at the point BC.
Then, as so does the D-stack AssCW[a,b] (by Cor. 2.3.5.9), we see that the homotopy
fiber, taken at the point BC, of the morphism B : Ass −→ Cat∗ has a cotangent
complex at C. This homotopy fiber is K(C ∗ , 1) (see Cor. 2.3.3.7), and thus we would
have
LK(C ∗ ,1),C ≃ C ∨ [−1].
Let k[ǫi−1 ] be the square zero extension of k by k[i − 1] (i.e. ǫi−1 is in degree −i + 1),
for some i > 0 as in the statement. We now consider the homotopy fiber sequence
M apC(k) (C ∨ [−1], k[i − 1]) / K(C ∗ , 1)(k[ǫi−1 ]) / K(C ∗ , 1)(k).

Considering the long exact sequence in homotopy we find

π1 (K(C ∗ , 1)(k[ǫi−1 ])) = H 0 (C)∗ ⊕ H i−1 (C) / π1 (K(C ∗ , 1)(k)) = H 0 (C)∗ /

π0 (M apC(k) (C ∨ [−1], k[i − 1])) = H i (C) / π0 (K(C ∗ , 1)(k[ǫi−1 ])) = ∗.

This shows that the last morphism must be injective, which can not be the case as
soon as H i (C) 6= 0. 
196 2.3. COMPLICIAL ALGEBRAIC GEOMETRY

Remark 2.3.5.14. (1) Some of the results in Thm. 2.3.5.11 were announced
in [To-Ve2, Thm. 5.6]. We need to warn the reader that [To-Ve2, Thm.
^O briefly given before that
5.6] is not correct for the description of RCat
theorem (the same mistake appears in [To2, Thm. 4.4]). Indeed, RCat ^O
would correspond to isotrivial deformations of dg-categories, for which the
underlying complexes of morphisms stays locally constant. Therefore, the
^O can not be the full Hochschild complex as stated
tangent complex of RCat
in [To-Ve2, Thm. 5.6]. Our theorem 2.3.5.11 corrects this mistake.
(2) We like to consider our Thm. 2.3.5.11 (3) and Cor. 2.3.5.12 as a possible
explanation of the following sentence in [Ko-So, p. 266]:
“In some sense, the full Hochschild complex controls deformations of the
A∞ -category with one object, such that its endomorphism space is equal to
A.”

Furthermore, our homotopy fibration sequence

K(C ∗ , 1) / AssCW
[n,0]
/ CatCW
∗,[n,0]

is the geometric global counter-part of the well known exact triangles of

complexes (see e.g. [Ko, p. 59])
+1
C[1] / RDerk (C, C)[2] / HHk (C, C)[2] /

as we pass from the former to the latter by taking tangent complexes at the
point C.
(3) We saw in Cor. 2.3.5.13 that the full D-stack Cat∗ can not have a reasonable
infinitesimal theory. We think it is important to mention that even Cor.
2.3.5.12 cannot reasonably be true if we remove the assumption that D(x, y)
is of Tor-amplitude contained in [n, 0] for some n ≤ 0. Indeed, for any
commutative k-algebra k ′ , the morphism
B : Ass(k ′ ) −→ Cat∗ (k ′ )
is easily seen to induce an isomorphism on π0 and a surjection on π1 . From
this and the fact that TAss,C = RDer k (C, C)[1], we easily deduce that the
natural morphism
π0 (RDer k (C, C)[1]) −→ π0 (DefBC )
is surjective. Therefore, if Cor. 2.3.5.12 were true for D = BC, we would
have that the morphism
H 1 (RDerk (C, C)) −→ HH2 (C, C)
is surjective, or equivalently that the natural morphism
H 2 (C) −→ H 2 (RDerk (C, C))
is injective. But this is not the case in general, as the morphism C −→
RDer k (C, C) can be zero (for example when C is commutative). It is there-
fore not strictly correct to state that the Hochschild cohomology of an as-
sociative dg-algebra controls its deformation as a dg-category, contrary to
what appears in several references (including some of the authors !).
(4) For n = 0, we can easily show that the restriction N ∗ (CatCW −
∗,[0,0] ) ∈ D St(k)
is a 2-geometric D− -stack in the sense of §2.2. Furthermore, its truncation
t0 N ∗ (CatCW
∗,[0,0] ) ∈ Ho(k − Af f
∼,et
) is naturally equivalent to the Artin
2-stack of k-linear categories with one object. The tangent complex of
 2.3.5. EXAMPLES OF GEOMETRIC D-STACKS 197

N ∗ (CatCW∗,[0,0] ) at a point C corresponding to a k-alegbra projective of fi-

nite type over k is then the usual Hochschild complex HH(C, C) computing
the Hochschild cohomology of C.
Of course CatCW ∗,[n,0] is only a rough approximation to what should be
the stack of dg-categories, and in particular we think that CatCW ∗,[n,0] is not
suited for dealing with dg-categories coming from algebraic geometry. Let
for example X be a smooth and projective variety over a field k; it is known
that the the derived category Dqcoh (X) is of the form Ho(B − M oddg ) for
some associative dg-algebra B which is perfect as a complex of k-modules. It
is however very unlikely that B can be chosen to be concentrated in degrees
[n, 0] for some n ≤ 0 (by construction H i (B) = Exti (E, E) for some compact
generator E ∈ Dqcoh (X)). So the dg-algebra B when considered as a dg-
category will not define a point in CatCW ∗,[n,0] . Another, more serious problem
comes from the fact that the dg-algebra B is not uniquely determined, but is
only unique up to Morita equivalence. As a consequence, the variety X can
deform and B might not follow this deformation (though another, Morita
equivalent, dg-algebra will follow the deformation), and thus X 7→ B will
not be a morphism of stacks (even locally around X). Therefore, it seems
very important to consider the stack Cat∗ modulo Morita equivalences. We
also think that passing to Morita equivalences will solve the problem of the
non geometricity of Cat∗ mentioned above. This direction is currently being
investigated by M. Anel.
 CHAPTER 2.4

Brave new algebraic geometry

In this final chapter we briefly present brave new algebraic geometry1, i.e. alge-
braic geometry over ring spectra. We will emphasize the main differences with derived
algebraic geometry, and the subject will be studied in more details in future works.
As in the case of complicial algebraic geometry, we will present two distinct HAG
contexts (see Cor. 2.4.1.11) which essentially differs in the choice of the class P.
The first one, where P is chosen to be the class of strongly étale morphisms (see
Def. 2.4.1.3), is suited for defining brave new Deligne-Mumford stacks, and, as all
the contexts based on “strong” morphisms, it is geometrically very close to usual
algebraic geometry. The second HAG context, where P is chosen to be the class of
fip-smooth morphisms (i.e. formally perfect and formally i-smooth morphisms), is
weaker and is similar to the corresponding weak context already presented in com-
plicial algebraic geometry: it allows to define brave new geometric stacks which are
not Deligne-Mumford. Since the notion of fip-smooth morphism for brave new rings
behaves differently from commutative rings (see Prop. 2.4.1.5), the geometric intu-
ition in this context is once again a bit far from standard algebraic geometry (e.g.
smooth morphisms are not necessary flat). Nonetheless we think that this context
is very interesting, as it is not only geometrically reasonable (e.g. it satisfies Artin’s
conditions), but it is also able to “see” some of the interesting new phenomena arising
in the theory of structured ring spectra.

2.4.1. Two HAG contexts

We let C := SpΣ , the category of symmetric spectra in U. The model structure
we are going to use on SpΣ is the so called positive stable model structure described
in [Shi]. This model structure is Quillen equivalent to the usual model structure,
but is much better behaved with respect to homotopy theory of monoids and mod-
ules objects. The model category SpΣ is a symmetric monoidal model category for
the smash product of symmetric spectra. Furthermore, all our assumptions 1.1.0.1,
1.1.0.3, 1.1.0.2 and 1.1.0.4 are satisfied thanks to [MMSS, Theorem 14.5], [Shi, Thm.
3.1, Thm. 3.2], and [Shi, Cor. 4.3] in conjunction with Lemma [HSS, 5.4.4].
The category Comm(SpΣ ) is the category of commutative symmetric ring spectra,
together with the positive stable model structure. The category Comm(SpΣ ) will be
denoted by S − Alg, and its objects will simply be called commutative S-algebras or
also bn rings (where bn stands for brave new ). For any E ∈ SpΣ , we will set
πi (E) := πistab (RE),
where RE is a fibrant replacement of E in SpΣ , and πistab (RE) are the naive stable
homotopy groups of the Ω-spectrum RE. Note that if E ∈ SpΣ is fibrant, then there
is a natural isomorphism π∗stab E ≃ π∗ E, and that a map f : E ′ → E ′′ is a weak
equivalence in SpΣ if and only if π∗ f is an isomorphism.
1
The term “brave new rings” was invented by F. Waldhausen to describe structured ring spectra;
we have only adapted it to our situation.

199
200 2.4. BRAVE NEW ALGEBRAIC GEOMETRY

When A is a commutative S-algebra, the Z-graded abelian group π∗ (A) has a

natural structure of a commutative graded algebra. In the same way, when M an
A-module, π∗ (M ) becomes a graded π∗ (A)-module. An object E will be called con-
nective, or (−1)-connected, if πi (E) = 0 for all i < 0. We let C0 be SpΣ c , be full
subcategory of connective objects in SpΣ . We let A be S − Alg0, the full subcategory
of Comm(SpΣ ) consisting of commutative S-algebras A with πi (A) = 0 for any i 6= 0.
If we denote by H : CommRings −→ Comm(C) the Eilenberg-MacLane functor, then
S − Alg0 is the subcategory of S − Alg formed by all the commutative S-algebras
equivalent to some Hk for some commutative ring k.
Lemma 2.4.1.1. The triplet (SpΣ , SpΣ
c , S − Alg0 ) is a HA context.

Proof. The only thing to check is that any object A ∈ S − Alg0 is SpΣ
c -good in
the sense of Def. 1.1.0.10. Thanks to the equivalence between the homotopy theory
of Hk-modules and of complexes of k-modules (see [EKMM, Thm. IV.2.4]), this has
already been proved during the proof of Lem. 2.3.1.1 (2) . 
The following example lists some classes of formally étale maps in brave new
algebraic geometry, according to our general definitions in Chapter 1.1.
Example 2.4.1.2.
(1) If A and B are connective S-algebras, a morphism A → B is formally thh-
étale if and only if it is formally étale ([Min, Cor. 2.8]).
(2) A morphism of (discrete) commutative rings R → R′ is formally étale if and
only if the associated morphism HR → HR′ of bn rings is formally étale
if and only if the associated morphism HR → HR′ of bn rings is formally
thh-étale ([HAGI, §5.2]).
(3) the complexification map KO → KU is thh-formally étale (by [Ro, p. 3])
hence formally étale. More generally, the same argument shows that any
Galois extension of bn rings, according to J. Rognes [Ro], is formally thh-
étale, hence formally étale.
(4) There exist examples of formally étale morphisms of bn-rings which are not
thh-étale (see [Min] or [HAGI, §5.2]).
As in the case of complicial algebraic geometry (Ch. 2.3) we find it useful to
introduce also strong versions for properties of morphisms between commutative S-
algebras.
Definition 2.4.1.3. (1) Let A ∈ S − Alg, and M be an A-module. The
A-module M is strong if the natural morphism
π∗ (A) ⊗π0 (A) π0 (M ) −→ π∗ (M )
is an isomorphism.
(2) A morphism A −→ B in S − Alg is strongly flat (resp. strongly (formally)
smooth, resp. strongly (formally) étale, resp. a strong Zariski open immer-
sion) if B is strong as an A-module, and if the morphism of affine schemes
Spec π0 (B) −→ Spec π0 (A)
is flat (resp. (formally) smooth, resp. (formally) étale, resp. a Zariski open
immersion).
One of the main difference between derived algebraic geometry and unbounded
derived algebraic geometry was that the strong notions of flat, smooth, étale and
Zariski open immersion are not as easily related to the corresponding general notions
presented in §1.2. In the present situation, the comparison is even more loose as
 2.4.1. TWO HAG CONTEXTS 201

typical phenomena arising from the existence of Steenrod operations in characteristic

p, make the notion of smooth morphisms of S-algebras rather subtle, and definitely
different from the above notion of strongly smooth morphisms. We do not think this
is a problem of the theory, but rather we think of this as an interesting new feature of
brave new algebraic geometry, as compared to derived algebraic geometry, and well
worthy of investigation.
Proposition 2.4.1.4. Let f : A −→ B be a morphism in S − Alg.
(1) If A and B are connective, the morphism f is étale (resp. a Zariski open
immersion) in the sense of Def. 1.2.6.1, if and only if f is strongly étale
(resp. a strong Zariski open immersion).
(2) If the morphism f is strongly flat ( resp. strongly étale, resp. a strong
Zariski open immersion), then it is flat (resp. étale, resp. a Zariski open
immersion) in the sense of Def. 1.2.6.1.
Proof. (1) The proof is the same as for Thm. 2.2.2.6.

(2) The proof is the same as for Prop. 2.3.1.4. 

The reader will notice that the proof of Thm. 2.2.2.6 (2) does not apply to the
present context as a smooth morphism of commutative rings is in general not smooth
when considered as a morphism of commutative S-algebras. The typical example of
this phenomenon is the following.
Proposition 2.4.1.5. • The canonical map HQ → H(Q[T ]) is smooth,
i-smooth and perfect.
• The canonical map HFp → H(Fp [T ]) is strongly smooth but not formally
smooth, nor formally i-smooth (Def. 1.2.7.1).
Proof. For any discrete commutative ring k, we have a canonical map ak :
Hk[T ] := FHk (Hk) → H(k[T ]) of commutative Hk-algebras, corresponding to the
map H(k → k[T ]) pointing the element T . Now
M
π∗ (Hk[T ]) ≃ H∗ (Σr , k)
r≥0

where in the group homology H∗ (Σr , k), k is a trivial Σr -module. Since Hn (Σr , Q) = 0,
for n 6= 0, and H0 (Σr , Q) ≃ Q, for any r ≥ 0 we see that aQ is a stable homo-
topy equivalence, and therefore a weak equivalence ([HSS, Thm. 3.1.11]). In other
words H(Q[T ]) “is” the free commutative HQ-algebra on one generator; therefore it
is finitely presented over HQ and, since for any (discrete) commutative ring k we have
LHk[T ]/Hk ≃ Hk[T ],
the cotangent complex LH(Q[T ])/HQ is free of rank one over H(Q[T ]), hence projective
and perfect. So HQ → H(Q[T ]) is smooth and perfect.
Let’s move to the char p > 0 case. It is clear that HFp → H(Fp [T ]) is strongly
smooth; let’s suppose that it is formally smooth. In particular LH(Fp [T ])/HFp is a
projective H(Fp [T ])-module. Therefore π∗ LH(Fp [T ])/HFp injects into
a Y
π∗ H(Fp [T ]) ≃ Fp [T ]
E E

(concentrated in degree 0). But, by [Ba-McC, Thm. 4.2] and [Ri-Rob, Thm. 4.1],
π∗ LH(Fp [T ])/HFp ≃ (H(Fp [T ]))∗ (HZ)
202 2.4. BRAVE NEW ALGEBRAIC GEOMETRY

and the last ring has (HFp )∗ (HZ) as a direct summand (using the augmentation
H(Fp [T ]) → HFp ). Now, it is known that (HFp )∗ (HZ) is not concentrated in degree
0 (it is a polynomial Fp -algebra in positive degrees generators, for p = 2, and the
tensor product of such an algebra with an exterior Fp -algebra for odd p). Therefore
HFp → H(Fp [T ]) cannot be formally smooth. 
We remark again that, as now made clear by the proof above, the conceptual
reason for the non-smoothness of HFp → H(Fp [T ]) is essentially the existence of
(non-trivial) Steenrod operations in characteristic p > 0. Since formal smoothness
is stable under base-change, we also conclude that HZ → H(Z[T ]) is not formally
smooth. The same argument also shows that this morphism is not formally i-smooth.

The following example shows that the converse of Prop. 2.4.1.4 (2) is false in
general.
Example 2.4.1.6. The complexification map m : KO → KU is formally étale
but not strongly formally étale. In fact, we have
π∗ m : π∗ (KO) = Z[η, β, λ±1 ]/(η 3 , 2η, ηβ, β 2 − 4λ) −→ π∗ (KO) = Z[ν ±1 ],
with deg(η) = 1, deg(β) = 4, deg(λ) = 8, deg(ν) = 2, π∗ m(η) = 0, π∗ m(β) = 2ν 2 and
π∗ m(λ) = ν 4 . In particular π0 m is an isomorphism (hence étale) but m is not strong.
We address the reader to [HAGI, Rmk. 5.2.9] for an example, due to M. Mandell,
of a non connective formally étale extension of HFp , which is therefore not strongly
formally étale. There also exist examples of Zariski open immersion HR −→ A, here
R is a smooth commutative k-algebra, such that A possesses non trivial negative
homotopy groups (see [HAGI, §5.2] for more details).
The opposite model category S − Alg will be denoted by SAf f . We will endow
it with the following strong étale model topology.
Definition 2.4.1.7. A family of morphisms {Spec Ai −→ Spec A}i∈I in SAf f
is a strong étale covering family (or simply s-ét covering family) if it satisfies the
following two conditions.
(1) Each morphism A −→ Ai is strongly étale.
(2) There exists a finite sub-set J ⊂ I such that the family {A −→ Ai }i∈J is a
formal covering family in the sense of 1.2.5.1.
Using the definition of strong étale morphisms, we immediately check that a
family of morphisms {Spec Ai −→ Spec A}i∈I in SAf f is a s-ét covering family if
and only if there exists a finite sub-set J ⊂ I satisfying the following two conditions.
• For all i ∈ I, the natural morphism
π∗ (A) ⊗π0 (A) π0 (Ai ) −→ π∗ (Ai )
is an isomorphism.
• The morphism of affine schemes
a
Spec π0 (Ai ) −→ Spec π0 (A)
i∈J
is étale and surjective.
Lemma 2.4.1.8. The s-ét covering families define a model topology on SAf f , that
satisfies assumption 1.3.2.2.
Proof. The same as for Lem. 2.2.2.13. 
The model topology s-ét gives rise to a model category of stacks SAf f ∼,s-ét .
 2.4.2. ELLIPTIC COHOMOLOGY AS A DELIGNE-MUMFORD S-STACK 203

Definition 2.4.1.9. (1) An S-stack is an object F ∈ SAf f ∼,s-ét which is

a stack in the sense of Def. 1.3.2.1.
(2) The model category of S-stacks is SAf f ∼,s-ét , and its homotopy category
will be simply denoted by St(S).
We now set P to be the class of fip-smooth morphisms (i.e. formally perfect
and formally i-smooth morphisms) in SAlg, and Ps-ét be the class of strongly étale
morphisms.
Lemma 2.4.1.10. (1) The class Ps-ét of strongly étale morphisms and the
s-ét model topology satisfy assumptions 1.3.2.11.
(2) The class P of fip-smooth morphisms and the s-ét model topology satisfy
assumptions 1.3.2.11.
Proof. It is essentially the same as for Lem. 2.3.2.1. 
Corollary 2.4.1.11. (1) The 5-tuple (SpΣ , SpΣ , S − Alg, s-ét, Ps-ét ) is a
HAG context.
(2) The 5-tuple (SpΣ , SpΣ
c , S − Alg0 , s-ét, P) is a HAG context.

According to our general theory, the notions of morphisms in P and Ps-ét gives
two notions of geometric stacks in SAf f ∼,s-ét .
Definition 2.4.1.12. (1) A n-geometric Deligne-Mumford S-stack is an n-
geometric S-stack with respect to the class Ps-ét of strongly étale morphisms.
(2) An n-geometric S-stack is an n-geometric S-stack with respect to the class
P of fip-smooth morphisms.
Of course, as Ps-ét is included in P any strong n-geometric Deligne-Mumford
S-stack is an n-geometric S-stack.

Finally, the reader can easily check the following proposition.

Proposition 2.4.1.13. (1) The topology s-ét and the class Ps-ét satisfy Artin
condition (for the HA context (SpΣ , SpΣ , S − Alg).
(2) The topology s-ét and the class P satisfy Artin’s condition (for the HA con-
text (SpΣ , SpΣ
c , S − Alg0 ).

In particular, we obtain as a corollary of Thm. 1.4.3.2 that any n-geometric S-

stack has an obstruction theory relative to the HA context (SpΣ , SpΣ c , S − Alg0 ). In
the same way, any strong n-geometric Deligne-Mumford S-stack has an obstruction
theory relative to the context (SpΣ , SpΣ , S − Alg).
Without going into details, we mention that all the examples of geometric D-
stacks given in the previous chapter can be generalized to examples of geometric
S-stacks. One fundamental example is Perf CW [a,b] of perfect CW-modules of amplitude
contained in [a, b], which by a similar argument as for Prop. 2.3.5.4 is a 1-geometric
S-stack.

2.4.2. Elliptic cohomology as a Deligne-Mumford S-stack

In this final section we present the construction of a 1-geometric Deligne-Mumford
S-stack using the sheaf of spectra of topological modular forms.

The Eilenberg-MacLane spectrum construction (see [HSS, Ex. 1.2.5]) gives rise
to a fully faithful functor
LH! : St(Z) −→ St(S),
204 2.4. BRAVE NEW ALGEBRAIC GEOMETRY

which starts from the homotopy category of stacks on the usual étale site of affine
schemes, i.e.
St(Z) := Ho(Z − Af f ∼,ét).
This functor has a right adjoint, called the truncation functor
h0 := H ∗ : St(S) −→ St(Z),
simply given by composing a simplicial presheaf F : SAf f op −→ SSet with the func-
tor H : Af f −→ S − Af f .

Let us denote by E the moduli stack of generalized elliptic curves with integral
geometric fibers, which is the standard compactification of the moduli stack of elliptic
curves by adding the nodal curves at infinity (see e.g. [Del-Rap, IV], where it is
denoted by M(1) ); recall that E is a Deligne-Mumford stack, proper and smooth over
Spec Z ([Del-Rap, Prop. 2.2]).
As shown by recent works of M. Hopkins, H. Miller, P. Goerss, N. Strickland,
C. Rezk and M. Ando, there exists a natural presheaf of commutative S-algebras on
the small étale site E ét of E. We will denote this presheaf by tmf. Recall that by
construction, if U = Spec A −→ E is an étale morphism, corresponding to an elliptic
curve E over the ring A, then tmf(U ) is the (connective) elliptic cohomology theory
associated to the formal group of E (in particular, one has π0 (tmf(U )) = A). Recall
also that the (derived) global sections RΓ(E, tmf), form a commutative S-algebra, well
defined in Ho(S − Alg), called the spectrum of topological modular forms, and denoted
by tmf 2.
Let U −→ E be a surjective étale morphism with U an affine scheme, and let us
consider its nerve
U∗ : ∆op −→ Af f
[n] 7→ Un := U ×E U ×E · · · ×E U .
| {z }
n times

This is a simplicial object in E ét , and by applying tmf we obtain a co-simplicial

object in S − Alg
tmf(U∗ ) : ∆ −→ S − Alg
[n] 7→ tmf(Un ).
Taking RSpec of this diagram we obtain a simplicial object in the model category
S − Af f ∼,s-ét
RSpec (tmf(U∗ )) : ∆op −→ S − Af f ∼,s-ét
[n] 7→ RSpec (tmf(Un )).
The homotopy colimit of this diagram will be denoted by
E S := hocolimn∈∆op RSpec (tmf(U∗ )) ∈ St(S).
The following result is technically just a remark as there is essentially nothing to
prove; however, we prefer to state it as a theorem to emphasize its importance.
Theorem 2.4.2.1. The stack E S defined above is a strong Deligne-Mumford 1-
geometric S-stack. Furthermore E S is a “brave new derivation” of the moduli stack E
of elliptic curves, i.e. there exists a natural isomorphism in St(Z)
h0 (E S ) ≃ E.

2The notation here is a bit nonstandard: what is usually called the spectrum of topological
modular forms is actually the connective cover of the spectrum we have denoted by tmf
 2.4.2. ELLIPTIC COHOMOLOGY AS A DELIGNE-MUMFORD S-STACK 205

Proof. To prove that E S is geometric, it is enough to check that the simplicial

object RSpec (tmf(U∗ )) is a strongly étale Segal groupoid. For this, recall that for
any morphism U = Spec B → V = Spec A in E ét , the natural morphism
π∗ (tmf(V )) ⊗π0 (tmf(V )) π0 (tmf(U )) ≃ π∗ (tmf(V )) ⊗A B −→ π∗ (tmf(U ))
is an isomorphism. This shows that the functor
RSpec (tmf(−)) : E ét −→ S − Af f ∼,s-ét
preserves homotopy fiber products and therefore sends Segal groupoid objects to
Segal groupoid objects. In particular, RSpec (tmf(U∗ )) is a Segal groupoid object.
The same fact also shows that for any morphism U = Spec B → V = Spec A in E ét ,
the induced map tmf(V ) −→ tmf(U ) is a strong étale morphism. This implies that
RSpec (tmf(U∗ )) is a strongly étale Segal groupoid object in representable S-stacks,
and thus shows that E S is indeed a strong Deligne-Mumford 1-geometric S-stack.
The truncation functor h0 clearly commutes with homotopy colimits, and there-
fore
h0 (E S ) ≃ hocolimn∈∆op h0 (RSpec (tmf(Un ))) ∈ St(Z).
Furthermore, for any connective representable S-stack, RSpec A, one has a natural
isomorphism h0 (RSpec A) ≃ Spec π0 (A). Therefore, one sees immediately that there
is a natural isomorphism of simplicial objects in Z − Af f ∼,ét
h0 (RSpec (tmf(U∗ ))) ≃ U∗ .
Therefore, we get
h0 (E S ) ≃ hocolimn∈∆op h0 (RSpec (tmf(Un ))) ≃ hocolimn∈∆op Un ≃ E,
as U∗ is the nerve of an étale covering of E. 
Theorem 2.4.2.1 tells us that the presheaf of topological modular forms tmf pro-
vides a natural geometric S-stack E S whose truncation is the usual stack of elliptic
curves E. Furthermore, as one can show that the small strong étale topoi of E S and
E coincide (this is a general fact about strong étale model topologies), we see that
tmf := RΓ(E, tmf) ≃ RΓ(E S , O),
and therefore that topological modular forms can be simply interpreted as functions
on the geometric S-stack E S . Of course, our construction of E S has essentially been
rigged to make this true, so this is not a surprise. However, we have gained a bit from
the conceptual point of view: since after all E is a moduli stack, now that we know
the existence of the geometric S-stack E S we can ask for a modular interpretation of
it, or in other words for a direct geometric description of the corresponding simplicial
presheaf on S − Af f . An answer to this question not only would provide a direct
construction of tmf, but would also give a conceptual interpretation of it in a geometric
language closer the usual notion of modular forms.
Question 2.4.2.2. Find a modular interpretation of the S-stack E S .
Essentially, we are asking for the brave new “objects” that the S-stack E S clas-
sifies. We could also consider the non-connective version of the S-stack E S (defined
through the non-connective version of tmf) for which a modular interpretation seems
much more accessible.
Very recent work by J. Lurie (see [Lu2] for a detailed announcement of his results)
answers in fact to Question 2.4.2.2; he shows that such a variant of E S classifies brave
new versions of [AHS]’s elliptic spectra plus additional data (called orientations).
This moduli-theoretic point of view makes use of some very interesting notions of
206 2.4. BRAVE NEW ALGEBRAIC GEOMETRY

brave new abelian varieties, brave new formal groups and their geometry. The com-
plete picture (possibly extended to higher chromatic levels) does not only give an
alternative construction of the spectrum tmf (and a better functoriality) but it could
be the starting point of a rather new3 and deep interaction between stable homo-
topy theory and homotopical algebraic geometry, involving many new questions and
objects, and probably also new insights on classical objects of algebraic topology.

3J. Lurie’s approach has as a byproduct also a natural construction of G-equivariant versions
of elliptic cohomology, for any compact G.
 APPENDIX A

Classifying spaces of model categories

The classifying space of a model category M is defined to be N (MW ), the nerve of

its subcategory of equivalences. More generally, if C ⊂ MW is a full subcategory of the
category of equivalences in M , which is closed by equivalences in M , the classifying
space of C is N (C), the nerve of C.
We fix a V-small model category M and a full subcategory C ⊂ MW closed by
equivalences. We consider the model category (C, C)∧ , defined in [HAGI, §2.3.2].
op
Recall that the underlying category of (C, C)∧ is the category SSetC V , of V-small
simplicial presheaves on C. The model structure of (C, C)∧ is defined as the left
op
Bousfield localization of the levelwise projective model structure on SSetC V , by in-
verting all the morphisms in C. The important fact is that local objects in (C, C)∧
are functors F : C op −→ SSetV sending all morphisms in C to equivalences.
We define an adjunction
N : (C, C)∧ −→ SSet/N (C) (C, C)∧ ←− SSet/N (C) : S
in the following way. A functor F : C op −→ SSet is sent to the simplicial set N (F ),
for which the set of n-simplices is the set of parirs
N (F )n := {(c0 → c1 → · · · → cn , α)}
where (c0 → c1 → · · · → cn ) is an n-simplex in N (C) and α ∈ F (cn )n is an n-simplex
in F (cn ). Put in an other way, N (F ) is the diagonal of the bi-simplicial set
(n, m) 7→ N (C/Fn )m
where C/Fn is the category of objects of the presheaf of n-simplices in F . The functor
N has a right adjoint
S : SSet/N (C) −→ (C, C)∧
sending X −→ N (C) to the simplicial presheaf
S(X) : C op −→ SSet
x 7→ HomSSet/N (C) (N (hx ), X),
where hx is the presheaf of sets represented by x ∈ C (note that N (hx ) → N (C) is
isomorphic to N (C/x) → N (C)).
Proposition A.0.3. The adjunction (N, S) is a Quillen equivalence.
Proof. First of all we need to check that (N, S) is a Quillen adjunction. For
this we use the standard properties of left Bousfield localizations, and we see that it
is enough to check that
N : SP r(C) −→ SSet/N (C) SP r(C) ←− SSet/N (C) : S
is a Quillen adjunction (where SP r(C) is the projective model structure of simplicial
presheaves on C), and that S preserves fibrant objects. These two facts are clear by
definition of S and the description of fibrant objects in N (C, C)∧ .
207
208 A. CLASSIFYING SPACES OF MODEL CATEGORIES

For any x ∈ C, the morphism N (hx ) = N (C/x) −→ N (C) is isomorphic in

Ho(SSet/N (C)) to x −→ N (C). Therefore, for X ∈ Ho(SSet/N (C)) and x ∈ C, the
simplicial set
RS(X)(x) ≃ RHomSSet/N (C) (x, X)
is naturally isomorphic in Ho(SSet) to the homotopy fiber of X −→ N (C) taken at
x. This clearly implies that the right derived functor
RS : Ho(SSet/N (C)) −→ Ho((C, C)∧ )
is conservative. In particular, it only remains to show that the adjunction morphism
Id −→ RSLN is an isomorphism. But this last assumption follows from the definition
the functor N and from a standard lemma (se for example [Q3]), which shows that
the homotopy fiber at x ∈ C of N (F ) −→ N (C) is naturally equivalent to F (x) when
F is fibrant in (C, C)∧ . 
Recall from [HAGI, Lem. 4.2.2] that for any object x ∈ C, one can construct
a local model for hx as hR(x) , sending y ∈ C to HomC (Γ∗ (y), R(x)), where Γ∗ is a
co-simplicial replacement functor in M . The natural morphism hx −→ hR(x) being an
equivalence in (C, C)∧ , one finds using Prop. A.0.3 natural equivalences of simplicial
sets
Hom(hy , hR(x) ) ≃ M apeq h
M (y, x) ≃ RHomSSet/N (C) (N (hy ), N (hx )) ≃ y ×N (C) x,

where M apeqM (y, x) is the sub-simplicial set of the mapping space M apM (y, x) con-
sisting of equivalences. As a corollary of this we find the important result due to
Dwyer and Kan. For this, we recall that the simplicial monoid of self equivalences of
an object x ∈ M can be defined as
Aut(x) := Hom(C,C)∧ (hR(x) , hR(x) ).
Corollary A.0.4. Let C ⊂ MW be a full subcategory of equivalences in a model
category M , which is stable by equivalences. Then, one has a natural isomorphism in
Ho(SSet)
a
N (C) ≃ BAut(x)
x∈π0 (N (C)))

where Aut(x) is the simplicial monoid of self equivalences of x in M .

Another important consequence of Prop. A.0.3 is the following interpretation
of mapping spaces in term of homotopy fibers between classifying spaces of certain
model categories.
Corollary A.0.5. Let M be a model category and x, y ∈ M be two fibrant
and cofibrant objects in M . Then, there exists a natural homotopy fiber sequence of
simplicial sets
M apM (x, y) −→ N ((x/M )W ) −→ N (MW ),
where the homotopy fiber is taken at y ∈ M .
Proof. This follows easily from Cor. A.0.4. 
We will need a slightly more functorial interpretation of Cor. A.0.4 in the par-
ticular case where the model category M is simplicial. We assume now that M is a
V-small simplicial U-model category (i.e. the model category M is a SSetU -model
category in the sense of [Ho1]). We still let C ⊂ MW be a full subcategory of the
category of equivalences in M , and still assume that C ⊂ MW is stable by equiva-
lences.
 A. CLASSIFYING SPACES OF MODEL CATEGORIES 209

We define an S-category G(C) is the following way (recall that an S-category is

a simplicially enriched category, see for example [HAGI, §2.1] for more details and
notations). The objects of G(C) are the objects of C which are furthermore fibrant
and cofibrant in M . For two objects x and y, the simplicial set of morphisms is defined
to be
G(C)(x,y) := Homeq M (x, y),
where by definition Homeq M (x, y)n is the set of equivalences in M from ∆n ⊗ x to y
eq
(i.e. HomM (x, y) is the sub-simplicial set of HomM (x, y) consisting of equivalences).
Clearly, the S-category G(C) is groupoid like, in the sense that its category of con-
nected components π0 (G(C)) (also denoted by Ho(G(C)) is a groupoid (or equivalently
every morphism in G(C) has an inverse up to homotopy). Let C c,f be the full subcat-
egory of C consisting of fibrant and cofibrant objects in C. There exist two natural
morphisms of S-categories
C ←− C c,f −→ G(C),
where a category is considered as an S-category with discrete simplicial sets of mor-
phisms. Passing to the nerves, one gets a diagram of simplicial sets
N (C) ←− N (C c,f ) −→ N (G(C)),
where the nerve functor is extended diagonally to S-categories (see e.g. [D-K1]).
Another interpretation of corollary A.0.4 is the following result.
Proposition A.0.6. With the above notations, the two morphisms
N (C) ←− N (C c,f ) −→ N (G(C)),
are equivalences of simplicial sets.
Proof. It is well known that the left arrow is an equivalence as a fibrant-cofibrant
replacement functor gives an inverse up to homotopy. For the right arrow, we let
N (G(C))n be the category of n-simplicies in N (G(C)), defined by having the same
objects and with
(N (G(C))n )x,y := (N (G(C))x,y )n .
By definition of the nerve, one has a natural equivalence
N (G(C)) ≃ Hocolimn∆op (N (G(C))n ).
Furthermore, it is clear that each functor
C c,f = G(C)0 −→ G(C)n
induces an equivalence of the nerves, as the 0-simplex [0] → [n] clearly induces a
functor
N (G(C)n ) −→ N (C c,f )
which is a homotopy inverse. 
Proposition A.0.6 is another interpretation of Prop. A.0.3, as the S-category
G(C) is groupoid-like, the delooping theorem of G. Segal implies that there exists a
natural equivalence of simplicial sets
x ×hN (C) y ←− x ×hN (C c,f ) y −→ x ×hN (G(C)) y ←− G(C)(x,y) = Homeq
M (x, y).

The advantage of Prop. A.0.6 over the more general proposition A.0.3 is that it is
more easy to state a functorial property of the equivalences in the following particular
context (the equivalence in Prop. A.0.3 can also be made functorial, but it requires
some additional work, using for example simplicial localization techniques).
We assume that G : M −→ N is a simplicial, left Quillen functor between V-
small simplicial U-model categories. We let C ⊂ MW and D ⊂ NW be two full
210 A. CLASSIFYING SPACES OF MODEL CATEGORIES

sub-categories stable by equivalences, and we suppose that all objects in M and N

are fibrant. Finally, we assume that the functor G restricted to cofibrant objects sends
C c := M c ∩ C to Dc := N c ∩ D. In this situation, we define an S-functor
G : G(C) −→ G(D)
simply by using the simplicial enrichment of G. Then, one has a commutative diagram
of S-categories
Co Cc / G(C)

G G
 
Do Dc / G(D),
and thus a commutative diagram of simplicial sets
N (C) o N (C c ) / N (G(C))

G G
 
N (D) o N (Dc ) / N (G(D)).

The important fact here is that this construction is associative with respect to com-
position of the simplicial left Quillen functor G. In other words, if one has a diagram
of simplicial model categories Mi and simplicial left Quillen functors, together with
sub-categories Ci ⊂ (Mi )W satisfying the required properties, then one obtain a com-
mutative diagram of diagrams of simplicial sets
N (Ci ) o N (Cic ) / N (G(Ci ))

Gi Gi
 
N (Di ) o N (Dic ) / N (G(Di )).
 APPENDIX B

Strictification

Let I be a U-small category. For any i ∈ I we let Mi be a U-cofibrantly generated

model category, and for any u : i → j morphism in I we let
u∗ : Mj −→ Mi Mj ←− Mi : u∗
be a Quillen adjunction. We suppose furthermore that for any composition
u /j v /k
i
one has an equality of functors
u∗ ◦ v ∗ = (v ◦ u)∗
Such a data ({Mi }i , {u∗ }u ) will be called, according to [H-S], a (U-)cofibrantly gen-
erated left Quillen presheaf over I, and will be denoted simply by the letter M .
For any cofibrantly generated left Quillen presheaf M on I, we consider the cat-
egory M I , of I-diagrams in M in the following way. Objects in M I are given by the
data of objects xi ∈ Mi for any i ∈ I, together with morphisms φu : u∗ (xj ) −→ xi
for any u : i → j in I, making the following diagram commutative
u∗ (φv )
u∗ v ∗ (xk ) / u∗ (xj )

Id φu
 
(v ◦ u)∗ (xk ) / xi
φv◦u

u /j v / k in I. Morphisms in M I are simply given by families mor-

for any i
phisms fi : xi −→ yi , such that fi ◦ φu = φu ◦ fj for any i → j in I.
The category M I is endowed with a model structure for which the fibrations or
equivalences are the morphisms f such that for any i ∈ I the induced morphism fi
is a fibration or an equivalence in Mi . As all model categories Mi are U-cofibrantly
generated, it is not hard to adapt the general argument of [Hi, 11.6] in order to prove
that M I is also a U-cofibrantly generated model category.
We define an object x ∈ M I to be homotopy cartesian if for any u : i → j in I
the induced morphism
Lu∗ (xi ) −→ xj
is an isomorphism in Ho(Mj ). The full subcategory of cartesian objects in M I will
I
be denoted by Mcart .

There exists a presheaf of categories (−/I)op over I, having the opposite comma
category (i/I)op as value over the object i, and the natural functor (i/I)op −→ (j/I)op
for any morphism j −→ i in I. For any U-cofibrantly left Quillen presheaf M over I,
we define a morphism of presheaves of categories over I
M I × (−/I)op −→ M,
211
212 B. STRICTIFICATION

where M I is seen as a constant presheaf of categories, in the following way. For an

object i ∈ I, the functor
M I × (i/I)op −→ Mi
sends an object (x, u : i → j) to u∗ (xj ) ∈ Mi , and a morphism (x, u : i → j) −→
(y, v : i → k), given by a morphism x → y in M I and a commutative diagram in I

i >>
>> u
v >>
>
 w >
k / j,

is sent to the morphism in Mi

u∗ (xj ) ≃ v ∗ (w∗ (xj )) / v ∗ (xk ).

For a diagram u : i → j in I, the following diagram

M I × (j/I)op / Mj

u∗
 
M I × (i/I)op / Mi

clearly commutes, showing that the above definition actually defines a morphism
M I × (−/I)op −→ M.
I
We let (Mcart )cof I
W be the subcategory of M consisting of homotopy cartesian
I
and cofibrant objects in M and equivalences between them. In the same way we
c
consider the sub-presheaf MW whose value at i ∈ I is the subcategory of Mi consisting
of cofibrant objects in Mi and equivalences between them. Note that the functors
u∗ : Mj −→ Mi being left Quillen for any u : i −→ j, preserves the sub-categories of
equivalences between cofibrant objects.
We have thus defined a morphism of presheaves of categories
I
(Mcart )cW × (−/I)op −→ MW
c
,
and we now consider the corresponding morphism of simplicial presheaves obtained
by applying the nerve functor
I
N ((Mcart )cW ) × N ((−/I)op ) −→ N (MW
c
),
that is considered as a morphism in the homotopy category Ho(SP r(I)), of simplicial
presheaves over I. As for any i the category (i/I)op has a final object, its nerve
N ((i/I)op ) is contractible, and therefore the natural projection
I
N ((Mcart )cW ) × N ((−/I)op ) −→ N ((Mcart
I
)cW )
is an isomorphism in Ho(SP r(I)). We therefore have constructed a well defined
I
morphism i Ho(SP r(I)), from the constant simplicial presheaf N ((Mcart )cW ) to the
c
simplicial presheaf N (MW ). By adjunction this gives a well defined morphism in
Ho(SSet)
I
N ((Mcart )cW ) −→ Holimi∈I N (MW
c
).
The strictification theorem asserts that this last morphism is an isomorphism in
Ho(SSet). As this seems to be a folklore result we will not include a proof.
 B. STRICTIFICATION 213

Theorem B.0.7. For any U-small category I and any U-cofibrantly generated left
Quillen presheaf M on I, the natural morphism
I
N ((Mcart c
)cW ) −→ Holimi∈I op N (MW )
is an isomorphism in Ho(SSet).
Proof. When M is the constant Quillen presheaf of simplicial sets this is proved
in [D-K3]. The general case can treated in a similar way. See also [H-S, Thm. 18.6]
for a stronger result. 

Let I and M be as in the statement of Thm. B.0.7, and let M0 be a U-cofibrantly

model category. We consider M0 as a constant left Quillen presheaf on I, for which
all values are equal to M0 , and all transition functors are identities. We assume that
there exists a left Quillen natural transformation φ : M0 −→ M . By this, we mean
the data of left Quillen functors φi : M0 −→ Mi for any i ∈ I, such that for any
u : i → j one has u∗ ◦ φj = φi . In this case, we define a functor
φ : M0 −→ M I ,
by the obvious formula φ(x)i := φi (x) for x ∈ M0 , and the transition morphisms of
φ(x) all being identities. The functor φ is not a left Quillen functor, but preserves
equivalences between cofibrant objects, and thus possesses a left derived functor
Lφ : Ho(M0 ) −→ Ho(M I ).
One can even show that this functor possesses a right adjoint, sending an object
x ∈ M I to the homotopy limit of the diagram in M0 , i 7→ Rψi (xi ), where ψi is the
right adjoint to φi .
One also has a natural transformation of presheaves of categories
(M0 )cW −→ MW
c

inducing a natural morphism of simplicial presheaves on I

N ((M0 )cW ) −→ N (MW
c
),
and thus a natural morphism in Ho(SSet)
c
N ((M0 )cW ) −→ Holimi∈I op N (MW ).
Corollary B.0.8. Let I, M and M0 be as above, and assume that the functor
Lφ : Ho(M0 ) −→ Ho(M I )
is fully faithful and that its image consists of all homotopy cartesian objects in Ho(M I ).
Then the induced morphism
c
N ((M0 )cW ) −→ Holimi∈I op N (MW )
is an isomorphism in Ho(SSet).
Proof. Indeed, we consider the functor
G : (M0 )cW I
−→ (Mcart )cW
x 7→ φ(Qx),
where Qx is a functorial cofibrant replacement of x. By hypothesis, the induced
morphism on the nerves
N ((M0 )cW ) −→ N (Mcart
I
)cW
214 B. STRICTIFICATION

is an isomorphism in Ho(SSet). We now consider the commutative diagram in

Ho(SSet)
N ((M0 )cW ) / N (M I )c
cart W
RRR
RRR
RRR
RRR
R( 
c
Holimi∈I op N (MW ).
The right vertical arrow being an isomorphism by Thm. B.0.7 we deduce the corollary.

 APPENDIX C

Representability criterion (after J. Lurie)

The purpose of this appendix is to give a sketch of a proof of the following special
case of J. Lurie’s representability theorem. Lurie’s theorem is much deeper and out
of the range of this work. We will not need it in its full generality and will content
ourselves with this special case, largely enough for our applications.
Theorem C.0.9. (J. Lurie, see [Lu1]) Let F be a D− -stack. The following
conditions are equivalent.
(1) F is an n-geometric D− -stack.
(2) F satisfies the following three conditions.
(a) The truncation t0 (F ) is an Artin (n + 1)-stack.
(b) F has an obstruction theory relative to sk − M od1 .
(c) For any A ∈ sk − Alg, the natural morphism
RF (A) −→ Holimk RF (A≤k )
is an isomorphism in Ho(SSet).
Sketch of proof. The only if part is the easy part. (a) is true by Prop. 2.2.4.4
and (b) by Cor. 2.2.3.3. For (c) one proves the following more general lemma.
Lemma C.0.10. Let f : F −→ G be an n-representable morphism. Then for any
A ∈ sk − Alg, the natural square
RF (A) / Holimk RF (A≤k )

 
RG(A) / Holimk RG(A≤k )

is homotopy cartesian.
Proof. We prove this by induction on n. For n = −1, one reduces easily to
the case of a morphism between representable D− -stacks, for which the result simply
follows from the fact that A ≃ Holimk A≤k . Let us now assume that n ≥ 0 and the
result prove for m < n. Let A ∈ sk − Alg and
x ∈ π0 (Holimk RG(A) ×hRG(A≤k ) RF (A≤k ))
with projections
xk ∈ π0 (RG(A) ×hRG(A≤k ) RF (A≤k )).
We need to prove that the homotopy fiber H of
RF (A) −→ Holimk RG(A) ×hRG(A≤k ) RF (A≤k )
at x is contractible. Replacing F by F ×hG X where X := RSpec A, and G by X, one
can assume that G is a representable stack and F is an n-geometric stack. As G is
representable, the morphism
RG(A) −→ Holimk RG(A≤k )
215
216 C. REPRESENTABILITY CRITERION (AFTER J. LURIE)

is an equivalence. Therefore, we are reduced to the case where G = ∗. The point x is

then a point in π0 (Holimk RF (A≤k )), and we need to prove that the homotopy fiber
H, taken at x, of the morphism
RF (A) −→ Holimk RF (A≤k )
is contractible. Using Cor. 2.2.2.9 one sees easily that this last statement is local on
the small étale site of A. By a localization argument we can therefore assume that
each projection xk ∈ π0 (RF (A≤k )) of x is the image of a point yk ∈ π0 (RU (A≤k )),
for some representable D− -stack U and a smooth morphism U −→ F .
Using Prop. 1.4.2.6 and Lem. 2.2.1.1 we see that the homotopy fiber of the
morphism
RU (A≤k+1 ) −→ RU (A≤k ) ×hRF (A≤k ) RF (A≤k+1 )
taken at yk is equivalent to M apA≤k −Mod (LU/F,yk , πk+1 (A)[k + 1]). Cor. 2.2.5.3 then
implies that when k is big enough, the homotopy fibers of the morphisms
RU (A≤k+1 ) −→ RU (A≤k ) ×hRF (A≤k ) RF (A≤k+1 )
are simply connected, and thus this morphism is surjective on connected components.
This easily implies that the points yk can be thought as a point y ∈ π0 (Holimk RU (A≤k ))
whose image in π0 (Holimk RD(A≤k )) is equal to x.
We then consider the diagram
RU (A) / Holimk RU (A≤k )

 
RF (A) / Holimk RF (A≤k ).

By induction on n we see that this diagram is homotopy cartesian, and that the
top horizontal morphism is an equivalence. There, the morphism induced on the
homotopy fibers of the horizontal morphisms is an equivalence, showing that H is
contractible as required. 

Conversely, let F be a D− -stack satisfying conditions (a)−(c) of C.0.9. The proof

goes by induction on n. Let us first n = −1. We start by a lifting lemma.
Lemma C.0.11. Let F be a D− -stack satisfying the conditions (a) − (c) of Thm.
C.0.9. Then, for any affine scheme U0 , and any étale morphism U0 −→ t0 (F ), there
exists a representable D− -stack U , a morphism u : U −→ F , with LF,u ≃ 0, and a
homotopy cartesian square in k − D− Af f ∼,ét
U0 / t0 (F )

 
U / F.

Proof. We are going to construct by induction a sequence of representable D− -

stacks
U0 / U1 . . . / Uk / Uk+1 . . . /F

in k − D− Af f ∼,ét /F satisfying the following properties.

• One has Uk = RSpec Ak with πi (Ak ) = 0 for all i > k.
• The corresponding morphism Ak+1 −→ Ak induces isomorphisms on πi for
all i ≤ k.
 C. REPRESENTABILITY CRITERION (AFTER J. LURIE) 217

• The morphism uk : Uk −→ F are such that

πi (LUk /F,uk ) = 0 ∀ i ≤ k + 1.
Assume for the moment that this sequence is constructed, and let A := HolimAk ,
and U := RSpec A. The points uk , defines a well defined point in π0 (Holimk RF (Ak )),
which by condition (d) induces a well defined morphism of stacks u : U −→ F . Clearly,
one has a homotopy cartesian square
U0 / t0 (F )

 
U / F.
Let M be any A-module. Again, using condition (d), one sees that
DerF (U, M ) ≃ Holimk DerF (Uk , M≤k ) ≃ Holimk M apAk −Mods (LUk /F,uk , Mk ) ≃ 0.
This implies that LU/F,u = 0.
It remains to explain how to construct the sequence of Uk . This is done by
induction. For k = 0, the only thing to check is that
πi (LU0 /F,u0 ) = 0 ∀ i ≤ 1.
This follows easily from Prop. 1.4.2.6 and the fact that u0 : U0 −→ t0 (F ) is étale.
Assume now that all the Ui for i ≤ k have been constructed. We consider uk :
Uk −→ F , and the natural morphism
LUk −→ LUk /F,uk −→ (LUk /F,uk )≤k+2 = Nk+1 [k + 2],
where Nk+1 := πk+2 (LUk /F,uk ). The morphism
LUk −→ Nk+1 [k + 2]
defines a square zero extension of Ak by Nk+1 [k + 1], Ak+1 −→ Ak . By construction,
and using Prop. 1.4.2.5 there exists a well defined point in
RHomUk /k−D− Af f ∼,ét /F (Uk+1 , F ),
corresponding to a morphism in Ho(k − D− Af f ∼,ét /F )
Uk E / Uk+1
EE
EE
EE
E" 
F.
It is not hard to check that the corresponding morphism RSpec Ak+1 = Uk+1 −→ F
has the required properties. 
Let us come back to the case n = −1. Lemma C.0.11 implies that there exists
a representable D− -stack U , a morphism U −→ F such that LF,u ≃ 0, and inducing
an isomorphism t0 (U ) ≃ t0 (F ). Using Prop. 1.4.2.6 and Lem. 2.2.1.1, it is not hard
to show by induction on k, that for any A ∈ sk − Alg with πi (A) = 0 for i > k, the
induced morphism
RU (A) −→ RF (A)
is an isomorphism. Using condition (c) for F and U one sees that this is also true for
any A ∈ sk − Alg. Therefore, U ≃ F , and F is a representable D− -stack.

We now finish the proof of Thm. C.0.9 by induction on n. Let us suppose we

know Thm. C.0.9 for m < n and let F be a D− -stack satisfying conditions (a) − (c)
for rank n. By induction we see that the diagonal of F is (n−1)-representable and the
218 C. REPRESENTABILITY CRITERION (AFTER J. LURIE)

hard point is to prove that F has an n-atlas. For this we lift an n-atlas of it0 (F ) in
the following way. Starting from a smooth morphism U0 −→ it0 (F ), we to construct
by induction a sequence of representable D− -stacks
U0 / U1 . . . / Uk / Uk+1 . . . /F

in k − D− Af f ∼,ét /F satisfying the following properties.

• One has Uk = RSpec Ak with πi (Ak ) = 0 for all i > k.
• The corresponding morphism Ak+1 −→ Ak induces isomorphisms on πi for
all i ≤ k.
• The morphism uk : Uk −→ F is such that for any M ∈ Ak − M ods with
πi (M ) = 0 for all i > k + 1 and π0 (M ) = 0, one has
[LUk /F,uk , M ]Sp(Ak −Mods ) = 0.
Such a sequence can be constructed by induction on k using obstruction theory as
in Lem. C.0.11. We then let A = Holimk Ak , and U = RSpec A. Then, condition
(c) implies the existence of a well defined morphism U −→ F , which is seen to be
smooth using Cor. 2.2.5.3. Using this lifting of smooth morphisms from it0 to F , one
produces an n-atlas for F by lifting an n-atlas of it0 (F ). 
 Bibliography

[EGAI] A. Grothendieck, J. Dieudonné, Eléments de Géométrie Algébrique, I, Springer-Verlag, New

York 1971.
[AHS] M. Ando, M. J. Hopkins, N. P. Strickland, Elliptic spectra, the Witten genus, and the theorem
of the cube, Inv. Math. 146, (2001), 595-687.
[Ar] M. Artin, Algebraization of formal moduli I, in “Global Analysis” (papers in honor of K. Ko-
daira), University of Tokyo Press, 1969, p. 21-71.
[SGA4-I] M. Artin, A. Grothendieck, J. L. Verdier, Théorie des topos et cohomologie étale des
schémas- Tome 1, Lecture Notes in Math 269, Springer Verlag, Berlin, 1972.
[SGA4-II] M. Artin, A. Grothendieck, J. L. Verdier, Théorie des topos et cohomologie étale des
schémas- Tome 2, Lecture Notes in Math 270, Springer Verlag, Berlin, 1972.
[Ba] M. Basterra, André-Quillen cohomology of commutative S-algebras, J. Pure Appl. Algebra 144,
1999, 111-143.
[Ba-McC] M. Basterra, R. McCarthy, Gamma Homology, Topological André-Quillen Homology and
Stabilization, Topology and its Applications 121/3, 2002, 551-566.
[Be1] K. Behrend, Differential Graded Schemes I: Perfect Resolving Algebras, Preprint
math.AG/0212225.
[Be2] K. Behrend, Differential Graded Schemes II: The 2-category of Differential Graded Schemes,
Preprint math.AG/0212226.
[Bl] B. Blander, Local projective model structure on simpicial presheaves, K-theory 24 (2001) No.
3, 283-301.
[Ci-Ka1] I. Ciocan-Fontanine, M. Kapranov, Derived Quot schemes, Ann. Sci. Ecole Norm. Sup. (4)
34 (2001), 403-440.
[Ci-Ka2] I. Ciocan-Fontanine, M. Kapranov, Derived Hilbert Schemes, preprint available at
math.AG/0005155.
[Del1] P. Deligne, Catégories Tannakiennes, in Grothendieck Festschrift Vol. II, Progress in Math.
87, Birkhauser, Boston 1990.
[Del2] P. Deligne, Le groupe fondamental de la droite projective moins trois points, in Galois groups
over Q, Math. Sci. Res. Inst. Publ., 16, Springer Verlag, New York, 1989.
[Del-Rap] P. Deligne, M. Rapoport, Les schémas de modules de courbes elliptiques, 143-317 in Mod-
ular functions of one variable II, LNM 349, Springer, Berlin 1973.
[Dem-Gab] M. Demazure, P. Gabriel, Groupes algébriques, Tome I, Masson & Cie. Paris, North-
Holland publishing company, 1970.
[Du] D. Dugger Universal homotopy theories, Adv. Math. 164 (2001), 144-176.
[Du2] D. Dugger Combinatorial model categories have presentations, Adv. in Math. 164 (2001),
177-201.
[D-K1] W. Dwyer, D. Kan, Simplicial localization of categories, J. Pure and Appl. Algebra 17 (1980),
267-284.
[D-K2] W. Dwyer, D. Kan, Equivalences between homotopy theories of diagrams, in Algebraic topol-
ogy and algebraic K-theory, Annals of Math. Studies 113, Princeton University Press, Prince-
ton, 1987, 180-205.
[D-K3] W. Dwyer, D. Kan, Homotopy commutative diagrams and their realizations, J. Pure Appl.
Algebra 57 (1989) No. 1, 5-24.
[DHK] W. Dwyer, P. Hirschhorn, D. Kan, Model categories and more general abstract homotopy
theory, Book in preparation, available at http://www-math.mit.edu/∼ psh.
[DS] W. Dwyer, J. Spalinski, Homotopy theories and model categories, Handbook of Algebraic Topol-
ogy, edited by I. M. James, Elsevier, 1995, 73-126.
[EKMM] A.D. Elmendorf, I. Kriz, M.A. Mandell, J.P. May, Rings, modules, and algebras in stable
homotopy theory, Mathematical Surveys and Monographs, vol. 47, American Mathematical
Society, Providence, RI, 1997.

219
220 BIBLIOGRAPHY

[Goe-Ja] P. Goerss, J.F. Jardine, Simplicial homotopy theory, Progress in Mathematics, Vol. 174,
Birkhauser Verlag 1999.
[Goe-Hop] P. Goerss, M. Hopkins, André-Quillen (co)homology for simplicial algebras over sim-
plicial operads, Une Dégustation Topologique [Topological Morsels]: Homotopy Theory in the
Swiss Alps (D. Arlettaz and K. Hess, eds.), Contemp. Math. 265, Amer. Math. Soc., Providence,
RI, 2000, pp. 41-85.
[Go] J. Gorski, Representability of the derived Quot functor, in preparation.
[Gr] A.Grothendieck, Catégories cofibrées additives et complexe cotangent relatif, Lecture Note in
Mathematics 79, Springer-Verlag, Berlin, 1968.
[EGAI] A. Grothendieck, J. Dieudonné, Eléments de Géométrie Algébrique I, Springer Verlag,
Berlin, 1971.
[EGAIV] A. Grothendieck, Eléments de Géométrie Algébrique IV . Etude locale des schémas et des
morphismes de schémas, Publ. Math. I.H.E.S., 20, 24, 28, 32 (1967).
[Ha] M. Hakim, Topos annelés et schémas relatifs, Ergebnisse der Mathematik und ihrer Grenzge-
biete, Band 64. Springer-Verlag Berlin-New York, 1972.
[Hin1] V. Hinich, Homological algebra of homotopical algebras, Comm. in Algebra 25 (1997), 3291-
3323.
[Hin2] V. Hinich, Formal stacks as dg-coalgebras, J. Pure Appl. Algebra 162 (2001), No. 2-3, 209-
250.
[Hi] P. S. Hirschhorn, Model Categories and Their Localizations, Math. Surveys and Monographs
Series 99, AMS, Providence, 2003.
[H-S] A. Hirschowitz, C. Simpson, Descente pour les n-champs, preprint available at
math.AG/9807049.
[Hol] S. Hollander, A homotopy theory for stacks, preprint available at math.AT/0110247.
[Ho1] M. Hovey, Model categories, Mathematical surveys and monographs, Vol. 63, Amer. Math.
Soc., Providence 1998.
[Ho2] M. Hovey, Spectra and symmetric spectra in general model categories, J. Pure Appl. Alg. 165
(2001), 63-127.
[HSS] M. Hovey, B.E. Shipley, J. Smith, Symmetric spectra, J. Amer. Math. Soc. 13 (2000), no. 1,
149-208.
[Ill] L. Illusie, Complexe cotangent et déformations I, Lectur Notes in Mathematics 239, Springer
Verlag, Berlin, 1971.
[Ja1] J. F. Jardine, Simplicial presheaves, J. Pure and Appl. Algebra 47 (1987), 35-87.
[Ja2] J. F. Jardine, Stacks and the homotopy theory of simplicial sheaves, in Equivariant stable
homotopy theory and related areas (Stanford, CA, 2000). Homology Homotopy Appl. 3 (2001),
No. 2, 361-384.
[Jo1] A. Joyal, Letter to Grothendieck.
[Jo2] A. Joyal, unpublished manuscript.
[Jo-Ti] A. Joyal, M. Tierney, Strong stacks and classifying spaces, in Category theory (Como, 1990),
Lecture Notes in Mathematics 1488, Springer-Verlag New York, 1991, 213-236.
[Ka2] M. Kapranov, Injective resolutions of BG and derived moduli spaces of local systems, J. Pure
Appl. Algebra 155 (2001), No. 2-3, 167-179.
[K-P-S] L. Katzarkov, T. Pantev, C. Simspon, Non-abelian mixed Hodge structures, preprint
math.AG/0006213.
[Ko] M. Kontsevich, Operads and motives in deformation quantization, Moshé Flato (1937-1998),
Lett. Math. Phys. 48, (1999), No. 1, 35-72.
[Ko-So] M. Kontsevich, Y. Soibelman, Deformations of algebras over operads and the Deligne con-
jecture, Conférence Moshé Flato 1999, Vol. 1 (Dijon), 255-307, Math. Phys. Stud. 21, Kluwer
Acad. Publ, Dordrecht, 2000.
[Kr-Ma] I. Kriz, J. P. May, Operads, algebras, modules and motives, Astérisque 233, 1995.
[La-Mo] G. Laumon and L. Moret-Bailly, Champs algébriques, A Series of Modern Surveys in Math-
ematics vol. 39, Springer-Verlag 2000.
[Lu1] J.Lurie, PhD Thesis, MIT, Boston, 2004.
[Lu2] J. Lurie, A survey of elliptic cohomology, Preprint December 2005 (available at
http://www.math.harvard.edu/∼lurie/papers/survey.pdf).
[MMSS] M. Mandell, J. P. May, S. Schwede, B. Shipley Model categories of diagram spectra, Proc.
London Math. Soc. 82 (2001), 441–512.
[May] J.P. May, Pairings of categories and spectra, JPAA 19 (1980), 299-346.
[May2] J.P. May, Picard groups, Grothendieck rings, and Burnside rings of categories, Adv. in Math.
163, 2001, 1-16.
[Mil] J. S. Milne, Étale cohomology, Princeton University Press, 1980.
 BIBLIOGRAPHY 221

[Min] V. Minasian, André-Quillen spectral sequence for THH, Topology and Its Applications, 129,
(2003) 273-280.
[MCM] R. Mc Carthy, V. Minasian, HKR theorem for smooth S-algebras, Journal of Pure and
Applied Algebra, Vol 185, 2003, 239-258, 2003.
[Q1] D. Quillen, Homotopical algebra, Lecture Notes in Mathematics 43, Springer Verlag, Berlin,
1967.
[Q2] D. Quillen, On the (co-)homology of commutative rings, Applications of Categorical Algebra
(Proc. Sympos. Pure Math., Vol XVII, New York, 1964), 65-87. Amer. Math. Soc., Providence,
P.I.
[Q3] D. Quillen, Higher algebraic K-theory I, in Algebraic K-theory I-Higher K-theories, Lecture
Notes in Mathematics 341, Springer Verlag, Berlin.
[Re] C. Rezk, Spaces of algebra structures and cohomology of operads, Thesis 1996, available at
http://www.math.uiuc.edu/ rezk.
[Ri-Rob] B. Richter, A. Robinson, Gamma-homology of group algebras and of polynomial algebras,
To appear in the “Proceedings of the Northwestern conference” 2002.
[Ro] J. Rognes, Galois extensions of structured ring spectra, Preprint math.AT/0502183.
[Schw-Shi] S. Schwede, B. Shipley, Stable model categories are categories of modules, Topology 42
(2003), 103 − −153.
[Shi] B. Shipley, A convenient model category for commutative ring spectra, Preprint 2002.
[S1] C. Simpson, Homotopy over the complex numbers and generalized cohomology theory, in Moduli
of vector bundles (Taniguchi Symposium, December 1994), M. Maruyama ed., Dekker Publ.
(1996), 229-263.
[S2] C. Simpson, A Giraud-type characterization of the simplicial categories associated to closed
model categories as ∞-pretopoi, Preprint math.AT/9903167.
[S3] C. Simpson, Algebraic (geometric) n-stacks, Preprint math.AG/9609014.
[S4] C. Simpson, The Hodge filtration on non-abelian cohomology, Preprint math.AG/9604005.
[Sm] J. Smith, Combinatorial model categories, unpublished.
[Sp] M. Spitzweck, Operads, algebras and modules in model categories and motives, Ph.D.
Thesis, Mathematisches Institüt, Friedrich-Wilhelms-Universität Bonn (2001), available at
http://www.uni-math.gwdg.de/spitz/.
[Tab] G. Tabuada, Une structure de catégorie de modèles de Quillen sur la catégorie des dg-
catégories, Preprint math.KT/0407338.
[To1] B. Toën, Champs affines, Preprint math.AG/0012219.
[To2] B. Toën, Homotopical and higher categorical structures in algebraic geometry, Hablitation
Thesis available at math.AG/0312262
[To3] B. Toën, Vers une interprétation Galoisienne de la théorie de l’homotopie, Cahiers de topolo-
gie et geometrie differentielle categoriques, Volume XLIII (2002), 257-312.
[To-Va1] B. Toën, M. Vaquié, Moduli of objects in dg-categories, Preprint math.AG/0503269.
[To-Va2] B. Toën, M. Vaquié, Au-dessous de Spec Z, Preprint math.AG/0509684.
[HAGI] B. Toën, G. Vezzosi, Homotopical algebraic geometry I: Topos theory, Advances in Mathe-
matics, 193, Issue 2 (2005), p. 257-372.
[To-Ve1] B. Toën, G. Vezzosi, Segal topoi and stacks over Segal categories, December 25, 2002,
to appear in Proceedings of the Program “Stacks, Intersection theory and Non-abelian Hodge
Theory”, MSRI, Berkeley, January-May 2002 (also available as Preprint math.AG/0212330).
[To-Ve2] B. Toën, G. Vezzosi, From HAG to DAG: derived moduli spaces, p. 175-218, in “Axiomatic,
Enriched and Motivic Homotopy Theory”, Proceedings of the NATO Advanced Study Institute,
Cambridge, UK, (9-20 September 2002), Ed. J.P.C. Greenlees, NATO Science Series II, Volume
131 Kluwer, 2004.
[To-Ve3] B. Toën, G. Vezzosi, “Brave New” algebraic geometry and global derived moduli spaces of
ring spectra, to appear in Proceedings of the Euroworkshop “Elliptic Cohomology and Higher
Chromatic Phenomena” (9 - 20 December 2002), Isaac Newton Institute for Mathematical Sci-
ences (Cambridge, UK), H. Miller, D. Ravenel eds. (also available as Preprint math.AT/0309145).
[To-Ve4] B. Toën, G. Vezzosi, Algebraic geometry over model categories. A general approach to
Derived Algebraic Geometry, Preprint math.AG/0110109.
[We] C. Weibel, An introduction to homological algebra, Cambridge Univ. Press, Cambridge, 1995.