Topics in Noncommutative Algebraic Geometry,

Homological Algebra and K-theory.

Alexander Rosenberg

Introduction
This text is based on my lectures delivered at the School on Algebraic K-Theory
and Applications which took place at the International Center for Theoretical Physics
(ICTP) in Trieste during the last two weeks of May of 2007. It might be regarded as an
introduction to some basic facts of noncommutative algebraic geometry and the related
chapters of homological algebra and (as a part of it) a non-conventional version of higher
K-theory of noncommutative ’spaces’. Arguments are mostly replaced by sketches of the
main steps, or references to complete proofs, which makes the text an easier reading than
the ample accounts on different topics discussed here indicated in the bibliography.
Lecture 1 is dedicated to the first notions of noncommutative algebraic geometry –
preliminaries on ’spaces’ represented by categories and morphisms of ’spaces’ represented
by (isomorphism classes of) functors. We introduce continuous, affine, and locally affine
morphisms which lead to definitions of noncommutative schemes and more general locally
affine ’spaces’. The notion of a noncommutative scheme is illustrated with important
examples related to quantized enveloping algebras: the quantum base affine spaces and flag
varieties and the associated quantum D-schemes represented by the categories of (twisted)
quantum D-modules introduced in [LR] (see also [T]). Noncommutative projective ’spaces’
introduced in [KR1] and more general Grassmannians and flag varieties studied in [KR3]
are examples of smooth locally affine noncommutative ’spaces’ which are not schemes.
In Lecture 2, we recover some fragments of geometry behind the pseudo-geometric
picture outlined in the first lecture. We start with introducing underlying topological
spaces (spectra) of ’spaces’ represented by abelian categories and describing their main
properties. One of the consequences of these properties is the reconstruction theorem
for commutative schemes [R4] which can be regarded as one of the major tests for the
noncommutative theory. It says, in particular, that any quasi-separated commutative
scheme can be canonically reconstructed uniquely up to isomorphism from its category
of quasi-coherent sheaves. The noncommutative fact behind the reconstruction theorem
is the geometric realization of a noncommutative scheme as a locally affine stack of local
categories on its underlying topological space. The latter is a noncommutative analog of a
locally affine locally ringed topological space, that is a geometric scheme.
Lecture 3 complements this short introduction to the geometry of noncommutative
’spaces’ and schemes with a sketch of the first notions and facts of pseudo-geometry (in
particular, descent) and spectral theory of ’spaces’ represented by triangulated categories.
This is a simple, but quite revelative piece of derived noncommutative algebraic geometry.
Lectures 4, 5, 6 are based on some parts of the manuscript [R8] created out of at-
tempts to find natural frameworks for homological theories which appear in noncommu-
tative algebraic geometry. We start, in Lecture 4, with a version of non-abelian (and

1
often non-additive) homological algebra which is based on presites (– categories endowed
with a Grothendieck pretopology) whose covers consist of one morphism. Although it
does not matter for most of constructions and facts, we assume that the pretopologies are
subcanonical (i.e. representable presheaves are sheaves), or, equivalently, covers are strict
epimorphisms (– cokernels of pairs of arrows). We call such presites right exact categories.
The dual structures, left exact categories, appear naturally and play a crucial role in the
version of higher K-theory sketched in Lecture 5. We develop standard tools of higher
K-theory starting with the long ’exact’ sequence (in Lecture 5) followed by reductions by
resolutions and devissage which are discussed in Lecture 6.

The first version of these lectures was sketched at the Institute des Hautes Études
Scientifiques, Bures-sur-Yvette, in the late Spring of 2007 (in the process of preparation
to actual lectures in Trieste). The present text, which is an extended version of [R9],
was written at the Max Planck Institut für Mathematik in Bonn. I am grateful to both
Institutes for hospitality and excellent working conditions.

References:
Bibliographical references and cross references inside of one lecture are as usual. Ref-
erences to results in another lecture are preceded by the lecture number, e.g. II.3.1.

2
 Lecture 1. Noncommutative locally affine ’spaces’ and schemes.
In Section 1, we review the first notions of noncommutative algebraic geometry –
preliminaries on ’spaces’ represented by categories, morphisms represented by their in-
verse image functors. We recall the notions of continuous, flat and affine morphisms and
illustrate them with a couple of examples. In Section 2, we remind Beck’s theorem char-
acterizing monadic morphisms and apply it to study of affine relative schemes. In Section
3, we introduce the notions of a locally affine morphism and a scheme over a ’space’.
Section 4 is dedicated to flat descent which is one of the basic tools of noncommutative
algebraic geometry. In Section 5, we sketch several examples of noncommutative schemes
and more general locally affine spaces which are among illustrations and/or motivations of
constructions of this work.

1. Noncommutative ’spaces’ represented by categories and morphisms

between them. Continuous, affine and locally affine morphisms.
1.1. Categories and ’spaces’. As usual, Cat, or CatU , denotes the bicategory of
categories which belong to a fixed universum U. We call objects of Catop ’spaces’. For any
’space’ X, the corresponding category CX is regarded as the category of quasi-coherent
sheaves on X. For any U-category A, we denote by |A| the corresponding object of Catop
(the underlying ’space’) defined by C|A| = A.
We denote by |Cat|o the category having same objects as Catop . Morphisms from
f
X to Y are isomorphism classes of functors CY −→ CX . For a morphism X −→ Y , we
denote by f ∗ any functor CY −→ CX representing f and call it an inverse image functor
of the morphism f . We shall write f = [F ] to indicate that f is a morphism having an
f g
inverse image functor F . The composition of morphisms X −→ Y and Y −→ Z is defined
by g ◦ f = [f ∗ ◦ g ∗ ].
1.2. Localizations and conservative morphisms. Let Y be an object of |Cat|o
and Σ a class of arrows of the category CY . We denote by Σ−1 Y the object of |Cat|o such
that the corresponding category coincides with (the standard realization of) the quotient of
the category CY by Σ (cf. [GZ, 1.1]): CΣ−1 Y = Σ−1 CY . The canonical localization functor
p∗ p
Σ
CY −→ Σ−1 CY is regarded as an inverse image functor of a morphism, Σ−1 Y −→
Σ
Y.
f
For any morphism X −→ Y in |Cat|o , we denote by Σf the family of all arrows s
of the category CY such that f ∗ (s) is invertible (notice that Σf does not depend on the
choice of an inverse image functor f ∗ ). Thanks to the universal property of localizations,
f ∗ is represented as the composition of the localization functor p∗f = p∗Σ : CY −→ Σ−1
f CY
f

−1 fc∗
and a uniquely determined functor Σ CY −→ CX . In other words, f = pf ◦ fc for a
fc
uniquely determined morphism X −→ Σ−1
f Y.
f
A morphism X −→ Y is called conservative if Σf consists of isomorphisms, or, equiv-
alently, pf is an isomorphism.
f
A morphism X −→ Y is called a localization if fc is an isomorphism, i.e. the functor
fc∗ is an equivalence of categories.

3
 Thus, f = pf ◦ fc is a unique decomposition of a morphism f into a localization and
a conservative morphism.
1.3. Continuous, flat, and affine morphisms. A morphism is called continuous
if its inverse image functor has a right adjoint (called a direct image functor), and flat
if, in addition, the inverse image functor is left exact (i.e. preserves finite limits). A
continuous morphism is called affine if its direct image functor is conservative (i.e. it
reflects isomorphisms) and has a right adjoint.
1.4. Categoric spectrum of a unital ring. For an associative unital ring R,
we define the categoric spectrum of R as the object Sp(R) of |Cat|o represented by the
category R − mod of left R-modules; i.e. CSp(R) = R − mod.
φ φ̄∗
Let R −→ S be a unital ring morphism and R − mod −→ S − mod the functor S ⊗R −.
The canonical right adjoint to φ̄∗ is the pull-back functor φ̄∗ along the ring morphism φ.
A right adjoint to φ̄∗ is given by

φ̄!
φ! : S − mod −−−→ R − mod, L 7−→ HomR (φ∗ (S), L).

The map    
φ φ̄
R −→ S −7 → Sp(S) −→ Sp(R)

is a functor
Sp
Ringsop −−−→ |Cat|o
which takes values in the subcategory of |Cat|o formed by affine morphisms.
φ̄ φ
The image Sp(R) −→ Sp(T ) of a ring morphism T −→ R is flat (resp. faithful) iff φ
turns R into a flat (resp. faithful) right T -module.
1.4.1. Continuous, flat, and affine morphisms from Sp(S) to Sp(R). Let R
f
and S be associative unital rings. A morphism Sp(S) −→ Sp(R) with an inverse image
functor f ∗ is continuous iff

f ∗ ' M⊗R : L 7−→ M ⊗R L (1)

for an (S, R)-bimodule M defined uniquely up to isomorphism. The functor

f∗ = HomS (M, −) : N 7−→ HomS (M, N ) (2)

is a direct image of f .
By definition, the morphism f is conservative iff M is faithful as a right R-module,
i.e. the functor M ⊗R − is faithful.
The direct image functor (2) is conservative iff M is a generator in the category of left
S-modules, i.e. for any nonzero S-module N , there exists a nonzero S-module morphism
M −→ N .
The morphism f is flat iff M is flat as a right R-module.

4
 The functor (2) has a right adjoint, f ! , iff f∗ is isomorphic to the tensoring (over S)
by a bimodule. This happens iff M is a projective S-module of finite type. The latter is
equivalent to the condition: the natural functor morphism M∗S ⊗S − −→ HomS (M, −) is
an isomorphism. Here M∗S = HomS (M, S). In this case, f ! ' HomR (M∗S , −).
1.5. Example. Let G be a monoid and R a G-graded unital ring. We define the
’space’ SpG (R) by taking as CSpG (R) the category grG R − mod of left G-graded R-modules.
φ∗
There is a natural functor grG R − mod −→ R0 − mod which assigns to each graded R-
module its zero component (’zero’ is the unit element of the monoid G). The functor φ∗
has a left adjoint, φ∗ , which maps every R0 -module M to the graded R-module R ⊗R0 M .
The adjunction arrow IdR0 −mod −→ φ∗ φ∗ is an isomorphism. This means that the functor
φ∗ is fully faithful, or, equivalently, the functor φ∗ is a localization.
The functors φ∗ and φ∗ are regarded as respectively a direct and an inverse image
φ
functor of a morphism SpG (R) −→ Sp(R0 ). It follows from the above that the morphism
φ is affine iff φ is an isomorphism (i.e. φ∗ is an equivalence of categories).
In fact, if φ is affine, the functor φ∗ should be conservative. Since φ∗ is a localization,
this means, precisely, that φ∗ is an equivalence of categories.
1.6. The cone of a non-unital ring. Let R0 be a unital associative ring, and let
R+ be an associative ring, non-unital in general, in the category of R0 -bimodules; i.e. R+
m
is endowed with an R0 -bimodule morphism R+ ⊗R0 R+ −→ R+ satisfying the associativity
condition. Let R = R0 ⊕ R+ denote the augmented ring described by this data. Let TR+
denote the full subcategory of the category R − mod whose objects are all R-modules
annihilated by R+ . Let TR−+ be the Serre subcategory (that is a full subcategory closed
by taking subquotients, extensions, and arbitrary direct sums) of the category R − mod
spanned by TR+ .
We define the ’space’ cone of R+ by taking as CCone(R+ ) the quotient category R −
u∗
mod/TR−+ . The localization functor R − mod −→ R − mod/TR−+ is an inverse image functor
u
of a morphism of ’spaces’ Cone(R+ ) −→ Sp(R). The functor u∗ has a (necessarily fully
faithful) right adjoint, i.e. the morphism u is continuous. If R+ is a unital ring, then u
is an isomorphism (see C3.2.1). The composition of the morphism u with the canonical
affine morphism Sp(R) −→ Sp(R0 ) is a continuous morphism Cone(R+ ) −→ Sp(R0 ). Its
direct image functor is (regarded as) the global sections functor.
1.7. The graded version: ProjG . Let G be a monoid and R = R0 ⊕ R+ a G-graded
ring with zero component R0 . Then we have the category grG R − mod of G-graded R-
modules and its full subcategory grG TR+ = TR+ ∩ grG R − mod whose objects are graded
modules annihilated by the ideal R+ . We define the ’space’ ProjG (R) by setting

CProjG (R) = grG R − mod/grG TR−+ .

Here grG TR−+ is the Serre subcategory of the category grG R−modspanned by grG TR+ . One
can show that grG TR−+ = grG R − mod ∩ TR−+ . Therefore, we have a canonical projection
p
Cone(R+ ) −→ ProjG (R).

5
 The localization functor grG R − mod −→ CProjG (R+ ) is an inverse image functor of a
v v
continuous morphism ProjG (R) −→ SpG (R). The composition ProjG (R) −→ Sp(R0 ) of
φ
the morphism v with the canonical morphism SpG (R) −→ Sp(R0 ) defines ProjG (R) as a
’space’ over Sp(R0 ). Its direct image functor is called the global sections functor.
1.7.1. Example: cone and Proj of a Z+ -graded ring. Let R = ⊕n≥0 Rn be a Z+ -
graded ring, R+ = ⊕n≥1 Rn its ’irrelevant’ ideal. Thus, we have the cone of R+ , Cone(R+ ),
and Proj(R) = ProjZ (R), and a canonical morphism Cone(R+ ) −→ Proj(R).

2. Beck’s Theorem and affine morphisms.

f
2.1. The Beck’s Theorem. Let X −→ Y be a continuous morphism with inverse
image functor f ∗ , direct image functor f∗ , and adjunction morphisms
ηf f
IdCY −→ f∗ f ∗ and f ∗ f∗ −→ IdCX .

Let Ff denote the monad (Ff , µf ) on Y , where Ff = f∗ f ∗ and µf = f∗ f f ∗ .

We denote by Ff − mod, or by (Ff /Y ) − mod the category of Ff -modules. Its objects
are pairs (M, ξ), where M ∈ ObCY and ξ is a morphism Ff (M ) −→ M such that the
diagram
µf (M )
Ff2 (M ) −−−→ Ff (M )
 
Ff (ξ) y yξ
 
ξ
Ff (M ) −−−→ M
commutes and ξ ◦ ηf (M ) = idM . Morphisms from (M, ξ) to M
f, ξ)
e are given by morphisms
g
M −→ Mf of the category CY such that the diagram

Ff (g)
 ) −−−→ Ff(M )
Ff (M f
ξy yξ
  e
g
M −−−→ M
f

commutes. The composition is defined in a standard way.

We denote by Sp(Ff /Y ) the ’space’ represented by the category of Ff -modules and
call it the categoric spectrum of the monad Ff .
There is a commutative diagram

fe∗
CX −−−→ (Ff /Y ) − mod
(3)
f∗ & .f∗
CX

Here fe∗ is the canonical functor

CX −→ (Ff /Y ) − mod, M 7−→ (f∗ (M ), f∗ f (M )),

6
and f∗ is the forgetful functor (Ff /Y ) − mod −→ CY .
The following assertion is one of the versions of Beck’s theorem.
f
2.1.1. Theorem. Let X −→ Y be a continuous morphism.
(a) If the category CY has cokernels of reflexive pairs of arrows, then the functor
f¯∗ has a left adjoint, f¯∗ ; hence f¯∗ is a direct image functor of a continuous morphism
f
X̄ −→ Sp(Ff /Y ).
(b) If, in addition, the functor f∗ preserves cokernels of reflexive pairs, then the ad-
junction arrow f¯∗ f¯∗ −→ IdCX is an isomorphism, i.e. f¯∗ is a localization.
(c) If, in addition to (a) and (b), the functor f∗ is conservative, then f¯∗ is a category
equivalence.
Proof. See [MLM], IV.4.2, or [ML], VI.7.
f
2.1.2. Corollary. Let X −→ Y be an affine morphism (cf. 1.3). If the category
CY has cokernels of reflexive pairs of arrows (e.g. CY is an abelian category), then the
f
canonical morphism X −→ Sp(Ff /Y ) is an isomorphism.
f
2.1.3. Monadic morphisms. A continuous morphism X −→ Y is called monadic
if the functor
fe∗
CX −−−→ Ff − mod, M 7−→ (f∗ (M ), f∗ f (M )),
is an equivalence of categories.
2.2. Continuous monads and affine morphisms. A functor F is called continuous
if it has a right adjoint. A monad F = (F, µ) on a ’space’ Y (i.e. on the category CY ) is
called continuous if the functor F is continuous.
2.2.1. Proposition. A monad F = (F, µ) on Y is continuous iff the canonical
fˆ
morphism Sp(F/Y ) −→ Y is affine.
Proof. A proof in the case of a continuous monad can be found in [KR2, 6.2], or in
[R3, 4.4.1] (see also [R4, 2.2]).
2.2.2. Corollary. Suppose that the category CY has cokernels of reflexive pairs of
f f∗
arrows. A continuous morphism X −→ Y is affine iff its direct image functor CX −→ CY
is the composition of a category equivalence

CX −→ (Ff /Y ) − mod

for a continuous monad Ff on Y and the forgetful functor (Ff /Y ) − mod −→ CY . The
monad Ff is determined by f uniquely up to isomorphism.
Proof. The conditions of the Beck’s theorem are fullfiled if f is affine, hence f∗ is the
composition of an equivalence CX −→ (Ff /Y ) − mod for a monad Ff = (f∗ f ∗ , µf ) in CY
and the forgetful functor (Ff /Y ) − mod −→ CY (see (1)). The functor Ff = f∗ f ∗ has a
right adjoint f∗ f ! , where f ! is a right adjoint to f∗ . The rest follows from 2.2.1.

7
 2.3. The category of affine schemes over a ’space’ and the category of
monads on this ’space’.
2.3.1. Proposition. Let
h
X −−−→ Y
f & .g
S
be a commutative diagram in |Cat|o . Suppose CZ has cokernels of reflexive pairs of arrows.
If f and g are affine, then h is affine.
Let Af fS denote the full subcategory of the category |Cat|o /S of ’spaces’ over S
f
whose objects are pairs (X, X → S), where f is an affine morphism. On the other hand,
we have the category Monc (S) of continuous monads on the ’space’ S (i.e. on the category
CS ) and the functor
Monc (S)op −−−→ Af fS (1)
which assigns to every continuous monad F the object (Sp(F/S, f), where Sp(F/S) is the
’space’ represented by the category F − mod and the morphism f has the forgetful functor
F − mod −→ CS as a direct image functor. It follows from 2.3.1 and 2.2.2 that this functor
is essentially full (that is its image is equivalent to the category Af fS ).
G
For every endofunctor CS −→ CS , let |G| denote the set Hom(IdCS , G) of elements of
G. If F = (F, µ) is a monad, then the set of elements of F has a natural monoid structure;
we denote this monoid by |F|. And we denote by |F|∗ the group of the invertible elements
φ
of the monoid |F|. We say that two monad morphisms F −→ −→ G are conjugate to each
ψ

other of φ = t · ψ · t−1 for some t ∈ |G|∗ .

Let Monrc (S) denote the category whose objects are continuous monads on CS and
morphisms are conjugacy classes of morphisms of monads.
2.3.2. Proposition The functor (1) induces an equivalence between the category
Monrc (S) and the category Af fS of affine schemes over S.
2.3.3. Example. Let S = Sp(R) for an associative ring R. Then the category
Monc (S) of monads on CS = R − mod is naturally equivalent to the category R\Rings
of associative rings over R. The conjugacy classes of monad morphisms correspond to
conjugacy classes of ring morphisms. Let Ass denote the category whose objects are
associative rings and morphisms the conjugacy classes of ring morphisms.
One deduces from 2.3.2 the following assertion:
2.3.3.1. Proposition. The category Af fS of affine schemes over S = Sp(R) is
naturally equivalent to the category (R\Ass)op .

3. Noncommutative schemes and locally affine ’spaces’.

u
i
3.1. Covers. We call a family {Ui −→ X | i ∈ J} of morphisms of ’spaces’ a cover if
– all inverse image functors ui are exact (i.e. the functors u∗i preserve finite limits and
∗

colimits),

8
 – the family {u∗i | i ∈ J} is conservative (i.e. if u∗i (s) is an isomorphism for all i ∈ J,
then s is an isomorphism).
f
3.2. Locally affine morphisms of ’spaces’. We call a morphism X −→ S of
ui
’spaces’ locally affine if there exists a cover {Ui −→ X | i ∈ J} of the ’space’ X such that
all the compositions f ◦ ui are affine.
3.2.1. Semiseparated covers and semiseparated locally affine ’spaces’. A
ui
cover {Ui −→ X | i ∈ J} is called semiseparated if each of the morphisms ui is affine.
A locally affine ’space’ with a semiseparated affine cover is called semiseparated.
3.3. Weak schemes over S. Weak schemes over a ’space’ S are locally affine mor-
ui
phisms X −→ S which have an affine cover {Ui −→ X | i ∈ J} formed by localizatios. The
latter means that each inverse image functor u∗i is the composition of the localization func-
tor CX −→ Σ−1 ∗
u∗ CX , where Σui = {s ∈ HomCX | ui (s) is invertible}, and an equivalence
∗
i
of categories Σ−1
u∗ CX −→ CUi .
i

3.4. Schemes. For a multiplicative system Σ of arrows of a svelte category CX ,

we shall denote by Σ−1 X the ’space’ represented by the category Σ−1 CX and call it the
localization of X at Σ.
u1 u2
For any pair of localizations U1 −→ X ←− U2 , we have a commutative square
u12
U1 ∩ U2 −−−→ U
1
u21 y y u1 (1)
 
u2
U2 −−−→ X

where U1 ∩ U2 is the localization of X at the multiplicative system Σu∗1 ∨ Σu∗2 generated

by Σu∗1 and Σu∗2 and uij are corresponding localizations. The square (1) is cartesian in the
subcategory of ’spaces’ formed by morphisms with exact inverse image functors.
f i u
3.4.1. Definition. A weak scheme X −→ S with an affine cover {Ui −→ X | i ∈ J}
v
is a scheme, if for every i ∈ J and any exact localization V −→ X which is affine over S
(that is f ◦ v is affine), the inverse image functor of the morphism

eui
Ui ∩ V = (Σu∗i ∨ Σv∗ )−1 X −−−→ V = Σ−1
v∗ X

is a localization at a finitely generated multiplicative system.

f u
3.5. Open immersions. Let X −→ S be a morphism and U −→ X an exact
v
localization. We call it a quasi-compact open immersion if for every localization V −→ X
which is affine over S, the inverse image functor of the projection U ∩ V −→ V is a
localization at a finitely generated multiplicative system.
u f
We call a localization U −→ X an open immersion to X −→ S, if there is a cover
ui uui
{Ui −→ U | i ∈ J} such that the composition of arrows Ui −→ X is a quasi-compact
open immersion for each i ∈ J.

9
 f
Thus, X −→ S is a scheme if it has an affine cover consisting of open immersions.

4. Descent: “covers”, comonads, and glueing.

i u
4.1. Comonads associated with “covers”. Let {Ui −→ Xa| i ∈ J} be a family
u
of continuous morphisms and u the corresponding morphism U = Ui −→ X with the
i∈J
inverse image functor

u∗ Y
CX −−−→ CUi = CU , M 7−→ (u∗i (M )|i ∈ J).
i∈J

i u∗
It follows that the family of inverse image functors {CX −→ CUi | i ∈ J} is conserva-
∗
tive iff the functor u is conservative.
Suppose that the category CX has products of |J| objects. Then the morphism
u
a
U= Ui −→ X is continuous: its direct image functor assigns to every object (Li |i ∈ J)
i∈J
Y Y
of the category CU = CUi the product ui ∗ (Li ).
i∈J i∈J
ηu
The adjunction morphism IdCX −→ u∗ u∗ assigns to each object M of CX the mor-
Y ηu i
phism M −→ ui ∗ u∗i (M ) determined by adjunction arrows IdCX −→ ui ∗ u∗i .
i∈J

The adjunction morphism u∗ u∗ −→ u
IdCU assigns to each object L = (Li |i ∈ J) of CU
the morphism (u,i (L)|i ∈ J), where

Y u,i (L)
u∗i ( uj ∗ (Lj )) −−−→ Li
j∈J

is the composition of the image

Y u∗
i (pi )
u∗i ( uj ∗ (Lj )) −−−→ u∗i ui ∗ (Li )
j∈J

ui (Li )
of the image of the projection pi and the adjunction arrow u∗i ui ∗ (Li ) −−−→ Li .
4.2. Beck’s theorem and glueing. Suppose that for each i ∈ J, the category CUi
has kernels of coreflexive pairs of arrows and the functor u∗i preserves them. Then the
inverse and direct image functors of the morphism u satisfy the conditions of Beck’s theo-
rem, hence the category CX is equivalent to the category of comodules over the comonad
Gu = (Gu , δu ) = (u∗ u∗ , u∗ ηu u∗ ) associated with the choice of inverse and direct image
ηu
functors of u together with an adjunction morphism IdCX −→ u∗ u∗ .
Recall that Gu -comodule is a pair (L, ζ), where L is an object of CU and ζ a morphism
L −→ Gu (L) such that u (L) ◦ ζ = idL and Gu (ζ) ◦ ζ = δu (L) ◦ ζ. Beck’s theorem says that

10
if the category CU has kernels of coreflexive pairs of arrows and the functor u∗ preserves
eu∗
and reflects them, then the functor CX −→ (U\Gu ) − comod which assigns to each object
M of CX the Gu -comodule (u∗ (M ), δu (M )) is an equivalence of categories.
ui
In terms of our local data – the “cover” {Ui −→ X | i ∈ J}, a Gu -comodule (L, ζ) is
the data (Li , ζi |i ∈ J), where (Li |i ∈ J) = L and ζi is a morphism
Y
Li −→ u∗i u∗ (L) = u∗i ( uj ∗ (Lj ))
j∈J

which equalizes the pair of arrows

u∗ η u (L)
i u ∗

ui u∗ (L) = ui ( uj ∗ (Lj )) −−−−

−− −−−→ u∗ ( u u∗ ( u (L ))) = u∗ u u∗ u (L)
Y Y Y
∗ ∗
−− −−−→ i m ∗ m j∗ j i ∗ ∗
∗ u (u ∗ ζj )
j i j m j

and such that u,i (L) ◦ ζi = idLi , i ∈ J.

The exactness of the diagram

δu (L)
ζ
L −−−→ Gu (L) −−−→ 2
−−−→ Gu (L)
Gu (ζ)

is equivalent to the exactness of the diagram

u∗ ηu u∗ (L)
i
ζi
uj ∗ (Lj )) −−−−−−−→ u∗ (
Y Y Y
Li −−−→ u∗i ( −−−−−−−→ i um ∗ u∗m ( uj ∗ (Lj ))) (1)
∗
u (u ∗ ζj )
j∈J i j m∈J j∈J

for every i ∈ J. If the functors u∗k preserve products of J objects (or just the products
involved into (1)), then the diagram (1) is isomorphic to the diagram

u∗ ηu u∗ (L)
i
ζi
u∗i uj ∗ (Lj ) −−−−−−−→
Y Y
Li −−−→ −−−−−−−→ u∗i um ∗ u∗m uj ∗ (Lj ) (2)
∗u (u ∗ ζj )
j∈J i j j,m∈J

4.3. Remark. The exactness of the diagram (1) might be viewed as a sort of sheaf
property. This interpretation looks more plausible (or less streched) when the diagram
(1) is isomorphic to the diagram (2), because u∗i uj ∗ (Lj ) can be regarded as the section of
Lj over the ’intersection’ of Ui and Uj and u∗i um ∗ u∗m uj ∗ (Lj ) as the section of Lj over the
intersection of the elements Uj , Um , and Ui of the “cover”.
4.4. The condition of the continuity of the comonad associated with a
ui ∗
CUi −→
“cover”. Suppose that each direct image functorY CX , i ∈ J, has a right adjoint,
! !
ui ; and let u denote the functor CX −→ CU = CUi which maps every object M to
i∈J

11
(u!i (M )|i ∈ J). If the category CX has coproducts of |J| objects, then theafunctor u! has
a left adjoint which maps every object (Li |i ∈ J) of CU to the coproduct ui ∗ (Li ).
i∈J
a Y
Therefore, if the canonical morphism ui ∗ (Li ) −−−→ ui ∗ (Li ) is an isomorphism
i∈J i∈J
for every object (Li |i ∈ J) of the category CU , then (and only then) the functor u! is a
right adjoint to the functor u∗ .
In particular, u! is a right adjoint to u∗ , if the category CX is additive and J is finite.
4.5. Note. If, in addition, the functors ui ∗ are conservative for all i ∈ J, then the
functor u∗ is conservative, and the category CU is equivalent to the category of modules
over the continuous monad Fu = (Fu , µh ), where Fu = u∗ u∗ and µu = u∗ u u∗ for an
u
adjunction morphism u∗ u∗ −→ IdCU .

5. Some motivating examples.

5.1. The base affine ’space’ and the flag variety of a reductive Lie al-
gebra from the point of view of noncommutative algebraic geometry. Let g
be a reductive Lie algebra over C and U (g) the enveloping algebra of g. Let G be the
groupMof integral weights of g and G+ the semigroup of nonnegative integral weights. Let
R= Rλ , where Rλ is the vector space of the (canonical) irreducible finite dimensional
λ∈G+
representation with the highest weight λ. The module R is a G-graded algebra with the
multiplication determined by the projections Rλ ⊗ Rν −→ Rλ+ν , for all λ, ν ∈ G+ . It is
well known that the algebra R is isomorphic to the algebra of regular functions on the base
affine space of g. Recall that G/U , where G is a connected simply connected algebraic
group with the Lie algebra g, and U is its maximal unipotent subgroup.
The category CCone(R) is equivalent to the category of quasi-coherent sheaves on the
base affine space Y of the Lie algebra g. The category CProjG (R) is equivalent to the
category of quasi-coherent sheaves on the flag variety of g.
5.2. The quantized base affine ’space’ and quantized flag variety of a
semisimple Lie algebra. Let now g be a semisimple Lie algebra over a field k of zero
and let Uq (g) be the quantized enveloping algebra of g. Define the G-graded
characteristic,M
algebra R = Rλ the same way as above. This time, however, the algebra R is not
λ∈G+
commutative. Following the classical example (and identifying spaces with categories of
quasi-coherent sheaves on them), we call Cone(R) the quantum base affine ’space’ and
ProjG (R) the quantum flag variety of g.
5.2.1. Canonical affine covers of the base affine ’space’ and the flag variety.
Let W be the Weyl group of the Lie algebra g. Fix a w ∈ W . For any λ ∈ G+ , choose
a nonzero w-extremal vector eλwλ generating the one dimensional vector subspace of Rλ
formed by the vectors of the weight wλ. Set Sw = {k ∗ eλwλ |λ ∈ G+ }. It follows from the
Weyl character formula that eλwλ eµwµ ∈ k ∗ eλ+µ
w(λ+µ) . Hence Sw is a multiplicative set. It

12
was proved by Joseph [Jo] that Sw is a left and right Ore subset in R. The Ore sets
{Sw |w ∈ W } determine a conservative family of affine localizations
−1
Sp(Sw R) −−−→ Cone(R), w ∈ W, (4)

of the quantum base affine ’space’ and a conservative family of affine localizations
−1
SpG (Sw R) −−−→ ProjG (R), w ∈ W, (5)
−1
of the quantum flag variety. We claim that the category grG Sw R − mod of G-graded
−1 −1
Sw R-modules is naturally equivalent to the category (Sw R)0 − mod.
In fact, by 1.5, it suffices to verify that the canonical functor
−1 −1
grG Sw R − mod −→ Sw R)0 − mod
−1
which assigns to every graded Sw R-module its zero component is faithful; i.e. the zero
−1
component of every nonzero G-graded Sw R-module is nonzero. This is, really, the case,
−1
because if z is a nonzero element of the λ-component of a G-graded Sw R-module, then
λ −1
(ewλ ) z is a nonzero element of the zero component of this module.
Thus, we obtain an affine cover
−1
Sp((Sw R)0 ) −−−→ ProjG (R), w ∈ W, (6)

of the quantum flag variety ProjG (R) of the Lie algebra g.

The covers (4) and (6) are scheme structures on respectively quantum base affine
’space’ and quantum flat variety. One can check that all morphisms (4) and (6) are affine,
i.e. the covers (4) and (5) are semiseparated.
5.3. Noncommutative Grassmannians. Fix an associative unital k-algebra R.
Let R\Algk be the category of associative k-algebras over R (i.e. pairs (S, R → S), where
S is a k-algebra and R → S a k-algebra morphism). We call them for convenience R-rings.
We denote by Re the k-algebra R ⊗k Ro . Here Ro is the algebra opposite to R.
5.3.1. The functor GrM,V . Let M, V be left R-modules. Consider the functor,
s
GrM,V : R\Algk −→ Sets, which assigns to any R-ring (S, R → S) the set of isomorphism
classes of epimorphisms s∗ (M ) −→ s∗ (V ) (here s∗ (M ) = S ⊗R M ) and to any R-ring
s φ t
morphism (S, R → S) −→ (T, R → T ) the map GrM,V (S, s) −→ GrM,V (T, t) induced by
φ∗
the inverse image functor S − mod −→ T − mod, N 7−→ T ⊗S N .
5.3.2. The functor GM,V . Denote by GM,V the functor R\Algk −→ Sets which
s v u
assigns to any R-ring (S, R → S) the set of pairs of morphisms s∗ (V ) → s∗ (M ) → s∗ (V )
such that u ◦ v = ids∗ (V ) and acts naturally on morphisms. Since V is a projective module,
the map
π = πM,V : GM,V −→ GrM,V , (v, u) 7−→ [u], (1)
is a (strict) functor epimorphism.

13
 5.3.3. Relations. Denote by RM,V the ”functor of relations” GM,V ×Gr GM,V . By
M,V
s
definition, RM,V is a subfunctor of GM,V × GM,V which assigns to each R-ring, (S, R → S),
the set of all 4-tuples (u1 , v1 ; u2 , v2 ) ∈ GM,V × GM,V such that the epimorphisms u1 , u2
ϕ
are equivalent. The latter means that there exists an isomorphism s∗ (V ) −→ s∗ (V ) such
that u2 = ϕ ◦ u1 , or, equivalently, ϕ−1 ◦ u2 = u1 . Since ui ◦ vi = id, i = 1, 2, these
equalities imply that ϕ = u2 ◦ v1 and ϕ−1 = u1 ◦ v2 . Thus, RM,V (S, s) is a subset of all
(u1 , v1 ; u2 , v2 ) ∈ GM,V (S, s) × GM,V (S, s) satisfying the following relations:

u2 = (u2 ◦ v1 ) ◦ u1 , u1 = (u1 ◦ v2 ) ◦ u2 (2)

in addition to the relations describing GM,V (S, s) × GM,V (S, s):

u1 ◦ v1 = idS⊗R V = u2 ◦ v2 (3)

Denote by p1 , p2 the canonical projections RM,V −→

−→ GM,V . It follows from the sur-
jectivity of GM,V −→ GrM,V that the diagram
p1
−−−→ π
RM,V G
−−p−→ M,V −−−→ GrM,V (4)
2

is exact.
5.3.4. Proposition. If both M and V are projective modules of a finite type, then
the functors GM,V and RM,V are corepresentable.
Proof. See [KR2, 10.4.3].
5.3.5. Quasi-coherent presheaves on presheaves of sets. Consider the category
Aff k of affine k-schemes which we identify with the category of representable functors on
the category Algk of k-algebras, and the fibered category with the base Aff k whose fibers
are categories of left modules over corresponding algebras. Let X be a presheaf of sets on
Aff k . Then we have a fibered category Affg k /X with the base Aff k /X induced by the
forgetful functor Aff k /X −→ Aff k . The category Qcoh(X) of quasi-coherent presheaves
on X is the opposite to the category of cartesian sections of Aff
g k /X.

5.3.6. Quasi-coherent presheaves on GrM,V . Suppose that M and V are projec-

tive modules of a finite type, hence the functors GM,V and RM,V are corepresentable by
R-rings resp. (GM,V , R → GM,V ) and (RM,V , R → RM,V ). Then the category Qcoh(GM,V )
(resp. Qcoh(RM,V )) is equivalent to GM,V − mod (resp. RM,V − mod), and the category
Qcoh(GrM,V ) of quasi-coherent presheaves on GrM,V is equivalent to the kernel of the
diagram
p∗
1

Qcoh(GM,V ) −−−→ Qcoh(R

−−−→ M,V
) (5)
∗ p
2

This means that, after identifying categories of quasi-coherent presheaves in (5) with cor-
responding categories of modules, quasi-coherent presheaves on GrM,V can be realized as

14
pairs (L, φ), where L is a GM,V -module and φ is an isomorphism p∗1 (L) −→
∼
p∗2 (L). Mor-
g
phisms (L, φ) −→ (N, ψ) are given by morphisms L −→ N such that the diagram

p∗
1 (g)
p∗1(L) −−−→ p∗1 (N
 )
φ yo oy ψ
 
p∗
2 (g)
p∗2 (L) −−−→ p∗2 (N )

commutes. The functor

π∗
Qcoh(GrM,V ) −−−→ Qcoh(GM,V ), (L, φ) 7−→ L,

π
is an inverse image functor of the projection GM,V −→ GrM,V (see 5.3.3(4)).
One can show that the functor π ∗ is an inverse image functor of a faithfully flat affine
morphism π̄ from an affine ’space’ Sp(GM,V ) (where GM,V is a ring representing the functor
GM,V ) to the ’space’ GrassM,V represented by the category Qcoh(GrM,V ) of quasi-coherent
sheaves on GrM,V . In our terminology, this means that π̄ is an affine semiseparated cover
of GrassM,V .
5.3.7. Quasi-coherent sheaves of sets. Let X be a presheaf of sets on Aff k Given
a (pre)topology τ on Aff k /X, we define the subcategory Qcoh(X, τ ) of quasi-coherent
sheaves on (X, τ ) [KR4].
5.3.7.1. Theorem ([KR4]). (a) A topology τ on Aff k is subcanonical (i.e. all
representable presheaves are sheaves) iff Qcoh(X) = Qcoh(X, τ ) for every presheaf of
sets X on Aff k (in other words, ’descent’ topologies on Aff k are precisely subcanonical
topologies). In this case, Qcoh(X) = Qcoh(X, τ ) ,→ Qcoh(X τ ) = Qcoh(X τ , τ ), where X τ
is the sheaf associated to X and ,→ is a natural full embedding.
(b) If τ is a topology of effective descent [KR4] (e.g. the fpqc or smooth topology
[KR2]), then the categories Qcoh(X, τ ) and Qcoh(X τ ) are naturally equivalent.
This theorem says, roughly speaking, that the category Qcoh(X) of quasi-coherent
presheaves knows which topologies to choose. A topology that seems to be the most
plausible for Grassmannians, in particular, for N Pnk , is the smooth topology introduced
in [KR2]. It is of effective descent, and the category of quasi-coherent sheaves on N Pnk
defined in [KR1] is naturally equivalent to the category of quasi-coherent sheaves of the
projective space defined via smooth topology on Aff k .

15
 Lecture 2. Underlying topological spaces of noncommutative ’spaces’ and
schemes.
Section 1 contains necessary preliminaries on topologizing, thick and Serre subcate-
gories of an abelian category. In Section 2, we introduce the main notion for the geometric
study of (’spaces’ represented by) abelian categories – the spectrum Spec(−) and define
Zariski topology on the spectrum. In Section 3, we introduce local ’spaces’ and related to
them spectrum Spec− (X). This spectrum is naturally isomorphic to the Gabriel spectrum
in the case when CX is a Grothendieck locally noetherian category. There is a natural
embedding Spec(X) ,→ Spec− (X) whose image, Spec1,1 t (X), consists of all Serre subcat-
t
egories P such that the intersection P of all topologizing subcategories properly containing
P is not equal to P. In Section 4, we study local properties of the spectra with respect to
finite covers. In Section 5, we discuss noncommutative k-’spaces’ and schemes, where k is
an associative unital ring. In Section 6, we introduce the geometric center of a ’space’ X,
which is a locally ringed topological space. The geometric center of a ’space’ X turns to be
a (conventional) scheme if X has a structure of a noncommutative (that is not necessarily
commutative) k-scheme for an associative ring k.
The Reconstruction Theorem for quasi-compact schemes sounds as follows:
If CX is the category of quasi-coherent sheaves on a quasi-compact quasi-separated
commutative scheme, then the geometric center of X is isomorphic to the scheme.
If a scheme (commutative or not) is not quasi-compact, the spectrum Spec(X) should
be replaced by the spectrum Spec0c (X). Its definition and general properties, as well as
the Reconstruction Theorem for non-quasi-compact schemes, are given in Section 7.

1. Topologizing, thick, and Serre subcategories.

1.1. Topologizing subcategories. A full subcategory T of an abelian category CX
is called topologizing if it is closed under finite coproducts and subquotients.
A subcategory S of CX is called coreflective if the inclusion functor S ,→ CX has
a right adjoint; that is every object of CX has a biggest subobject which belongs to S.
Dually, a subcategory T of CX is called reflective if the inclusion functor T ,→ CX has a left
adjoint. We denote by T(X) the preorder with respect to ⊆ of topologizing subcategories
and by Tc (X) (resp. Tc (X)) the preorder of coreflective (resp. reflective) topologizing
subcategories of CX .
1.1.1. The Gabriel product and infinitesimal neighborhoods of topologizing
categories. The Gabriel product, S • T, of the pair of subcategories S, T of CX is the full
subcategory of CX spanned by all objects M such that there exists an exact sequence

0 −→ L −→ M −→ N −→ 0

with L ∈ ObT and N ∈ ObS. It follows that 0 • T = T = T • 0 for any strictly full
subcategory T. The Gabriel product of two topologizing subcategories is a topologizing
subcategory, and its restriction to topologizing categories is associative; i.e. (T(X), •)
is a monoid. Similarly, the Gabriel product of coreflective topologizing subcategories is a

16
coreflective topologizing subcategory, hence Tc (X) is a submonoid of (T(X), •). Dually, the
preorder Tc (X) of reflective topologizing subcategories of CX is a submonoid of (T(X), •).
The nth infinitesimal neighborhood, T(n+1) , of a subcategory T is defined by T(0) = 0
and T(n+1) = T(n) • T for n ≥ 0.
1.2. The preorder  and topologizing subcategories. For any two objects,
M and N , of an abelian category CX , we write M  N if N is a subquotient of a finite
coproduct of copies of M . For any object M of the category CX , we denote by [M ] the
full subcategory of CX whose objects are all L ∈ ObCX such that M  L. It follows
that M  N ⇔ [N ] ⊆ [M ]. In particular, M and N are equivalent with respect
to  (i.e.
M  N  M ) iff [M ] = [N ]. Thus, the preorder {[M ] | M ∈ ObCX }, ⊇ is a canonical
realization of the quotient of (ObCX , ) by the equivalence relation associated with .
1.2.1. Lemma. (a) For any object M of CX , the subcategory [M ] is the smallest
topologizing subcategory containing M .
[The smallest topologizing subcategory spanned by a family of objects S coincides
(b)
with [N ], where SΣ denotes the family of all finite coproducts of objects of S.
N ∈SΣ

Proof. (a) Since  is a transitive relation, the subcategory [M ] is closed with respect
to taking subquotients. If M  Mi , i = 1, 2, then M  M ⊕ M  M1 ⊕ M2 , which
shows that [M ] is closed under finite coproducts, hence it is topologizing. Clearly, any
topologizing subcategory
[ containing M contains the subcategory [M ].
(b) The union [N ] is contained in every topologizing subcategory containing the
N ∈SΣ
family S. It is closed under taking subquotients, because each [N ] has this property. It
is closed under finite coproducts, because if N1 , N2 ∈ SΣ and Ni  Mi , i = 1, 2, then
N1 ⊕ N2  M1 ⊕ M2 .
For any subcategory (or a class of objects) S, we denote by [S] (resp. by [S]c ) the
smallest topologizing resp. coreflective topologizing) subcategory containing S.
1.2.2. Proposition. Suppose that CX is an abelian category with small coproducts.
Then a topologizing subcategory of CX is coreflective iff it is closed under small coproducts.
The smallest
[ coreflective
[ topologizing subcategory spanned by a set of objects S coincides
with [N ] = [N ], where Se is the family of all small coproducts of objects of S.
N ∈Se N ∈S
e
Suppose that CX satisfies (AB4), i.e. it has infinite coproducts and the coproduct of a
set of monomorphisms is a monomorphism. Then, for any object M of CX , the smallest
coreflective topologizing subcategory [M ]c spanned by M is generated by subquotients of
coproducts of sets of copies of M .
Proof. The argument is similar to that of 1.2.1 and left to the reader as an exercise.
1.3. Thick subcategories. A topologizing subcategory T of the category CX is
called thick if T • T = T; in other words, T is thick iff it is closed under extensions.
We denote by Th(X) the preorder of thick subcategories of CX . For a thick subcate-
gory T of CX , we denote by X/T the quotient ’space’ defined by CX/T = CX /T .

17
 1.4. Serre subcategories. We recall the notion of a Serre subcategory of an abelian
category as it is defined in [R, III.2.3.2]. For a subcategory T of CX , let T− denote the full
subcategory of CX generated by all objects L of CX such that any nonzero subquotient of
L has a nonzero subobject which belongs to T.
1.4.1. Proposition. Let T be a subcategory of CX . Then
(a) The subcategory T− is thick.
(b) (T− )− = T− .
(c) T ⊆ T− iff any subquotient of an object of T is isomorphic to an object of T.
Proof. See [R, III.2.3.2.1].
1.4.2. Remark. It follows from 1.4.1 and the definition of T− that, for any sub-
category T of an abelian category CX , the associated Serre subcategory T− is the largest
topologizing (or the largest thick) subcategory of CX such that every its nonzero object
has a nonzero subobject from T.
1.4.3. Definition. A subcategory T of an abelian category CX is called a Serre
subcategory if T− = T. We denote by Se(X) the preorder (with respect to ⊆) of all Serre
subcategories of CX .
The following characterization of Serre subcategories turns to be quite useful.
1.4.4. Proposition. Let T be a subcategory of an abelian category CX closed under
taking subquotients. The following conditions are equivalent:
(a) T is a Serre subcategory.
(b) If S is a subcategory ofTthe category CX which is closed under subquotients and
is not contained in T, then S T⊥ 6= 0.
Proof. (a) ⇒ (b). Let T be a subcategory of CX closed under taking quotients. By the
definition of T− , an object M does not belong to T− iff it has a nonzero subquotient, L,
which does not have a nonzero subobject from T. Since T is closed under taking quotients,
the latter means precisely that Hom(N, L) = 0 for every N ∈ ObT, i.e. L ∈ ObT⊥ . Thus,
M does not belong to T− iff it has a nonzero subquotient which belongs to T⊥ .
(b) ⇒ (a). By the condition (b), if an object M does not belong to T, then it has
a nonzero subquotient which belongs to T⊥ . But, by the observation above, this means
that the object M does not belong to T− . So that T− ⊆ T. The inverse inclusion holds,
because T is closed under taking subquotients (see 1.4.1(c)).
1.4.5. The property (sup). Recall that X (or the corresponding category CX ) has
the property (sup) if for any object M of CX and for any ascending chain, Ω, of subobjects
of M , the supremum of Ω exists, and for any subobject L of M , the natural morphism

sup(N ∩ L | N ∈ Ω) −→ (supΩ) ∩ L

is an isomorphism.
1.4.6. Coreflective thick subcategories and Serre subcategories. Recall that
a full subcategory T of a category CX is called coreflective if the inclusion functor T ,→ CX

18
has a right adjoint. In other words, each object of CX has the largest subobject which
belongs to T .
1.4.6.1. Lemma. Any coreflective thick subcategory is a Serre subcategory. If CX
has the property (sup), then any Serre subcategory of CX is coreflective.
Proof. See [R, III.2.4.4].
1.4.7. Proposition. Let CX have the property (sup). Then for any thick subcategory
T of CX , all objects of T− are supremums of their subobjects contained in T.
Proof. Since CX has the property (sup), the full subcategory Ts of CX whose ob-
jects are supremums of objects from T is thick and coreflective, hence Serre, subcategory
containing T and contained in T− . Therefore it coincides with T− .

2. The spectrum Spec(X). We denote by Spec(X) the family of all nonzero objects
M of the category CX such that L  M for any nonzero subobject L of M .
The spectrum Spec(X) of the ’space’ X is the family of topologizing subcategories
{[M ] | M ∈ Spec(X)} endowed with the specialization preorder ⊇.
Let τ  denote the topology on Spec(X) associated with the specialization preorder:
the closure of W ⊆ Spec(X) consists of all [M ] such that [M ] ⊆ [M 0 ] for some [M 0 ] ∈ W .
2.1. Proposition. (a) Every simple object of the category CX belongs to Spec(X).
The inclusion Simple(X) ,→ Spec(X) induces an embedding of the set of the isomorphism
classes of simple objects of CX into the set of closed points of (Spec(X), τ  ).
(b) If every nonzero object of CX has a simple subquotient, then each closed point of
(Spec(X), τ  ) is of the form [M ] for some simple object M of the category CX .
Proof. (a) If M is a simple object, then Ob[M ] consists of all objects isomorphic to
coproducts of finite number of copies of M . In particular, if M and N are simple objects,
then [M ] ⊆ [N ] iff M ' N .
(b) If L is a subquotient of M , then [L] ⊆ [M ]. If [M ] is a closed point of Spec(X),
this implies the equality [M ] = [L].
Notice that the notion of a simple object of an abelian category is selfdual, i.e.
op
Simple(X) = Simple(X o ), where X o is the dual ’space’ defined by CX o = CX . In
particular, the map M 7−→ [M ] induces an embedding of isomorphism classes of simple
objects of CX into the intersection Spec(X) Spec(X o ).
T

2.1.1. Proposition. If the category CX has enough objects of finite type, then the set

of closed points of (Spec(X), τ ) coincides with Spec(X) Spec(X o ) and is in a natural
T
bijective correspondence with the set Simple(X) of isomorphism classes of simple objects
of the category CX .
Proof. Since every nonzero object of CX has a nonzero subobject of finite type,
Spec(X) consists of [M ] such that M is of finite type and belongs to Spec(X). On the
other hand, if M is of finite type and [M ] belongs to Spec(X o ), then [M ] = [M1 ], where
M1 is a simple quotientTof M . This proves that the set of closed points of Spec(X)
coincides with Spec(X) Spec(X o ). It follows from 2.2(b) that the set closed points of
Spec(X) coincides with the set of isomorphism classes of simple objects of CX .

19
 2.2. Supports of objects. For any object M of the category CX , the support of
M is defined by Supp(M ) = {Q ∈ Spec(X) | Q ⊆ [M ]}. This notion enjoys the usual
properties:
2.2.1. Proposition. (a) If 0 −→ M 0 −→ M −→ M 00 −→ 0, is a short exact
sequence, then [
Supp(M ) = Supp(M 0 ) Supp(M 00 ).

(b) Suppose the category CX has the property (sup). Then

(b1) If M is the supremum of a filtered system {Mi | i ∈ J} of its subobjects, then
[
Supp(M ) = Supp(Mi ).
i∈J

(b2) As a consequence of (a) and (b1), we have

M [
Supp( Mi ) = Supp(Mi ).
i∈J i∈J

Proof. (a) Since [M 0 ] ⊆ [M ] ⊇ [M 00 ], we have the inclusion

[
Supp(M 0 ) Supp(M 00 ) ⊆ Supp(M ).

In order to show the inverse inclusion, notice that for any object L of the subcategory
[M ], there exists an exact sequence 0 −→ L0 −→ L −→ L00 −→ 0 such that L0 is an
object of [M 0 ] and L00 belongs to [M 00 ]. This follows from the fact that L is a subquotient
of a coproduct M ⊕n of n copies of M the related commutative diagram
0 00
0 −−−→ Mx⊕n −−−→ Mx⊕n −−−→ M x⊕n −−−→ 0
cart
  
  
0
0 −−−→ K
 −−−→ K
 −−−→ K00 −−−→ 0 (1)
  
y y y
0 −−−→ L0 −−−→ L −−−→ L00 −−−→ 0

whose rows are exact sequences, the upper vertical arrows are monomorphisms, the lower
ones epimorphisms, and the left upper square is cartesian.
Now if [L] ∈ Spec(X) and the object L0 in the diagram (1) is nonzero, then, by the
definition of the spectrum, [L0 ] = [L], hence [L] ∈ Supp(M 0 ). If L0 = 0, then the arrow
L −→ L00 is an isomorphism, in particular,
[ [L] = [L00 ] ∈ Supp(M 00 ).
(b1) The inclusion Supp(M ) ⊇ Supp(Mi ) is obvious. It follows from the property
i∈J
(sup) that if an object L is a nonzero subquotient of M ⊕n for some n, then it contains a
nonzero subobject, L0 , which is a subquotient of Mi for some i ∈ J. If [L] ∈ Spec(X),
this implies that [L] = [L0 ] ∈ Supp(Mi ).

20
 (b2) If J is finite, the assertion follows from (a). If J is infinite, it is a consequence of
(a) and (b1).
2.3. Topologies on Spec(X). We are interested in topologies compatible with the
specializations: every closed set should contain specializations of all its points.
2.3.1. The topologies τ  and τ . The finest such topology is the topology τ 
introduced at the beginning of the section. The coarsest reasonable topology is the topology
τ having the set of specializations of points of the spectrum as a base of closed sets.
2.3.2. A general construction: supports and topologies. Let Ξ be a class of
\ finite coproducts. For any set E of objects of Xi, let V(E)
objects of CX closed under
denote the intersection Supp(M ). Then, for any family {Ei | i ∈ I} of such sets, we
M ∈E
have, evidently, [ \
V( Ei ) = V(Ei ).
i∈J i∈J
S
It follows from the equality Supp(M ⊕ N ) = Supp(M ) Supp(N ) (see 2.2.1(a)) that
V(E ⊕ E)e = V(E) S V(E).e Here E ⊕ E e def
= {M ⊕ N | M ∈ E, N ∈ E}. e
This shows that Spec(X), ∅, and the subsets V(E) of Spec(X), where E runs through
subsets of Ξ, form the family of all closed sets of a topology, τΞ , on Spec(X).
2.3.3. Examples. Taking as Ξ the class of finite coproducts of the representatives
of the elements of Spec(X), we recover the topology τ . If CX is a category with small
coproducts, then, taking Ξ = ObCX , we recover the topology τ  .
2.4. Zariski topology on the spectrum. Notice that the class Ξf (X) of objects
of finite type is closed under finite coproducts, hence it defines a topology on Spec(X)
which we denote by τf . If the category CX has enough objects of finite type, then we shall
call the topology τf the Zariski topology.
2.4.1. Example. Let R be a commutative unital ring and CX the category R − mod
of R-modules. Then Spec(X) is isomorphic to the prime spectrum Spec(R) of the ring R
and the topology τf corresponds to the Zariski topology on Spec(R).
2.4.2. Note. If the category CX does not have enough objects of finite type, the
topology τf is not a “right” candidate for the Zariski topology, because it should be finer
than the topology τ (introduced in 2.3.1) and it might be not. Zariski topology can be
naturally extended to ’spaces’ X for which the category CX has “locally” enough objects
of finite type. In general, “locally” is an additional data. This data is naturally available
in the case of (not necessarily commutative) schemes.
2.5. Complement: topologizing subcategories and topologies on the spec-
trum. There is another, more universal, way to define topologies on Spec(X).
2.5.1. Lemma. Let W be a family of topologizing subcategories of an abelian category
CX which contains CX , the zero subcategory, is closed under arbitrary intersections and
such that for any pair T1 , T2 of elements of W, there exists T3 ∈ W such that T1 ∪ T2 ⊆

21
T3 ⊆ (T1 ∪ T2 )− . Then τW
o
= {V(T ) = Spec(|T |) | T ∈ W} is a set of closed subsets of a
topology, τW .
Proof. In fact, for any set {Ti | i ∈ J} of topologizing subcategories of the category
CX , we have
\ \
V( Ti ) = V(Ti ). (1)
i∈J i∈J

If T1 , T2 and T3 are topologizing subcategories of CX such that

T1 ∪ T2 ⊆ T3 ⊆ (T1 ∪ T2 )− , (2)

then V(T3 ) = V(T1 ) ∪ V(T2 ). This is a consequence of the inclusions V(T1 ) ∪ V(T2 ) ⊆
V(T3 ) ⊆ V((T1 ∪ T2 )− ) following from (2) and the equality V(T ) = V(T − ) which holds for
any topologizing subcategory T .
2.5.2. A special case. Let Ξ be a set of objects of an abelian category CX closed
fΞ = {[M ] | M ∈ Ξ} and denote by WΞ the set consisting
under finite coproducts. We set W
of CX , 0, and of intersections of arbitrary subfamilies of W
fΞ . The set of topologizing
subcategories WΞ satisfies the conditions of 2.5.1, and the topology it defines coincides
with the topology τΞ introduced in 2.3.
2.5.3. Monoids of topologizing subcategories and associated topologies on
the spectrum. Let W be a set of topologizing subcategories of CX containing CX and
the the zero subcategory 0 and closed under the Gabriel multiplication and arbitrary
intersections. Then W satisfies the conditions of 2.5.1, because for any pair of topologizing
subcategories T1 , T2 , their Gabriel product T1 • T2 contains T1 ∪ T2 and is contained in the
Serre subcategory (T1 ∪ T2 )− generated by T1 ∪ T2 .
Taking as W the monoid T(X) of all topologizing subcategories of CX , we recover the
topology τt0 on Spec(X) associated with the specialization preorder ⊇: the closure of a
subset of the spectrum consists of all specializations of the elements of this subset. This
is the finest among the reasonable topologies on the spectrum.
The map G 7−→ τG is a surjective map from the family of full monoidal subcategories
of (T(X), •) closed under arbitrary intersections onto the set of topologies on Spec0t (X)
which are coarser than the topology τt0 corresponding to T(X).
2.5.4. The coarse Zariski topology. Recall that a full subcategory T of CX is
reflective if the inclusion functor T ,→ CX has a left adjoint. Suppose that the category CX
has supremums of sets of subobjects (for instance, CX has infinite coproducts). Then, by
[R, III.6.2.2], the intersection of any set of reflective topologizing subcategories is a reflective
topologizing subcategory. Taking as W the subcategory Tc (X) of reflective topologizing
subcategories, we obtain the coarse Zariski topology on Spec(X) which we denote by τZ .
2.5.4.1. Proposition. Suppose CX has the property (sup) and a generator of finite
type. Then the topological space (Spec(X), τZ ) is quasi-compact.
Proof. See [R, III.6.5.2.1].

22
 2.5.4.2. Example: the coarse Zariski topology on an affine noncommutative
scheme. Let CX be the category R − mod of left modules over an associative unital ring
R. For every two-sided ideal α in R, let Tα denote the full subcategory of R − mod whose
objects are modules annihilated by the ideal α. By [R, III.6.4.1], the map α 7−→ Tα is an
isomorphism of the preorder (I(R), ⊇) of two-sided ideals of the ring R onto (Tc (X), ⊆).
Moreover, Tα • Tβ = Tαβ for any pair of two-sided ideals α, β. This means that the
map α 7−→ Tα is an isomorphism of monoidal categories (preorders), where the monoidal
structure on I(R) is the multiplication of ideals. Note by passing that it follows from this
description that every reflective topologizing subcategory of CX = R − mod is coreflective.
One of the consequences of 2.5.4.1 is that the topological space (Spec(X), τZ ) is
quasi-compact. This fact is a special case of a more precise assertion: an open subset
U of the space (Spec(X), τZ ) is quasi-compact iff U = U (Tα ) = Spec(X) − V (Tα ) for
a finitely generated two-sided ideal α of the ring R. Two different proofs of this the-
orem can be found in [R]: I.5.6 and III.6.5.3.1. One of its consequences is that quasi-
compact open sets form a base of the Zariski topology on Spec(X). In fact, every two-
sided ideal α is the supremum of [ a set {αi | i ∈ J} of its two-sided subideals, so that
U (Tα ) = U (sup(Tαi | i ∈ J)) = U (Tαi ).
i∈J

2.5.4.3. Digression: the coarse Zariski topology and the Levitski spectrum.
By definition, the Levitski spectrum LSpec(R) of an associative unital ring R consists of
all prime ideals p of R such that the quotient ring R/p has no non-trivial locally nilpotent
ideals. If the ring R is left noetherian, then the Levitski spectrum coincides with the prime
spectrum. We endow the Levitski spectrum LSpec(R) with the Zariski topology induced
from the prime spectrum: its closed sets are V(α) = {p ∈ LSpec(R) | α ⊆ p}, where α
runs through the set of all two-sided ideals of the ring R.
Let CX be the category R − mod of left modules over an associative ring R. The map
which assigns to every R-module M the annihilator Ann(M ) of M reverses the preorder
; that is if [N ] ⊆ [M ], then Ann(N ) ⊆ Ann(M ). This implies that the map

[M ] 7−→ Ann(M )

is well defined and if [M ] is an object of Spec(X), then Ann(M ) = Ann(L) for any nonzero
subobject of M . One can see that the latter implies that Ann(M ) is a prime ideal, if [M ]
is an element of Spec(X). More precise statement is as follows:
Theorem ([R, I.5.3]) The image of the map

Spec(X)−−−→Spec(R), [M ] 7−→ Ann(M ),

coincides with the Levitski spectrum LSpec(R) of the ring R and the map is a quasi-
homeomorphism of (Spec(X), τZ ) onto the Levitski spectrum LSpec(R).
2.5.4.4. Note. Let CX = R − mod for an associative ring R. Then the sets
Supp(R/n) = V([R/n]), where n runs through the set I` (R) of all left ideals of the ring R,
form a base of the Zariski closed sets on Spec(X). By 2.5.4.2, closed sets of the coarse
Zariski topology are precisely the sets Supp(R/α), where α runs through the set I(R) of

23
all two-sided ideals of R. This shows that the coarse Zariski topology is, indeed, coarser
than the Zariski topology on the spectrum, and these topologies coincide if the ring R is
commutative. One can show that they coincide if R is a PI ring.
Unfortunately, the coarse Zariski topology is trivial, or too coarse in many important
examples of noncommutative affine schemes. Thus, the coarse Zariski on the spectrum
of X = Sp(R) is trivial iff R is a simple ring (i.e. it does not have non-trivial two-sided
ideals). In particular, it is trivial if CX is the category of D-modules on the affine space
An , because the algebra An of differential operators on An is simple. The coarse Zariski
topology on Spec(X) is non-trivial, but not sufficiently rich, when CX is the category of
representations of a semisimple Lie algebra over a field of characteristic zero.

3. Local ’spaces’ and Spec− (−).

3.1. Local ’spaces’. A ’space’ X and the representing it abelian category CX are
called local if CX has the smallest nonzero topologizing subcategory, CXt .
One can see that CXt is the only closed point of Spec(X). It follows from definitions
that a ’space’ X is local iff Spec(X) has only one closed point which belongs to support
of any nonzero object of the category CX .
3.1.1. Proposition. Let X be local, and let the category CX have simple objects.
Then all simple objects of CX are isomorphic to each other, and every nonzero object of
CXt is a finite coproduct of copies of a simple object.
Proof. In fact, if M is a simple object in CX , then [M ] is a closed point of Spec(X).
If X is local, this closed point is unique. Therefore, objects of CXt are finite coproducts
of copies of M (see the argument of 2.1).
3.1.2. The residue ’space’ of a local ’space’. Let X be local ’space’ and CXt the
smallest non-trivial topologizing subcategory of the category CX . We regard the inclusion
functor CXt ,→ CX as an inverse image functor of a morphism of ’spaces’ X −→ Xt and
call Xt the residue ’space’ of X.
3.1.3. The residue skew field of a local ’space’. Suppose that X is a local ’space’
such that the category CX has a simple object, M . We denote by kX the ring CX (M, M )o
opposite to the ring of endomorphisms of the object M . Since M is simple, kX is a skew
field which we call the residue skew field of the local ’space’ X. It follows from 3.1.1 that
the residue skew field of X (if any) is defined uniquely up to isomorphism.
It follows that the residue category CXt of the ’space’ X is naturally equivalent to the
category of finite dimensional kX -vector spaces.
3.2. Spec− (X). By definition, Spec− (X) is formed by all Serre subcategories P of
CX such that X/P is a local ’space’. It is endowed with the specialization preorder ⊇.
3.2.1. Support in Spec− (X). We define the support of an object M of CX in
Spec− (X) as the set Supp− (M ) of all P ∈ Spec− (X) which do not contain M , or,
equivalently, the localization of M at P is nonzero. We leave as an exercise proving the
analogue of 2.2.1 for Supp− (−).

24
 3.2.2. Zariski topology. We introduce the topology, τf− , on Spec− (X) the same
way as the topology τf on the spectrum Spec(X): supports of objects of finite type form
a base of its closed sets. If the category CX has enough objects of finite type, then we call
τf− the Zariski topology on Spec− (X).

3.2.3. Remark on topologies on Spec− (X). For any (topologizing) subcategory

T of CX , we set V − (T ) = {P ∈ Spec− (X) | T * P}. If T = [M ] for an object M , then the
set V − (T ) coincides with the support Supp− (M ) of the object M . There is an analogue
of 2.5.1 for the sets V − (T ). In particular, any submonoid M of the monoid (T(X), •) of
topologizing subcategories of CX which is closed under arbitrary intersections determines
a topology τM −
on Spec− (X) whose closed sets are V − (T ), where T runs through M. And
all “reasonable” topologies on Spec− (X) are of this form (see 2.5.3). In particular, taking
the submonoid Tc (X) of reflective topologizing subcategories of CX (and assuming that
CX has supremums of sets of subobjects), we obtain the coarse Zariski topology τZ− on
Spec− (X) (similar to 2.5.4). Details are left to the reader.

3.2.4. Indecomposable injectives and Spec− (−). If CX is a Grothendieck cat-

egory with Gabriel-Krull dimension (say, CX is locally noetherian), then the elements of
Spec− (X) are in bijective correspondence with the set of isomorphism classes of inde-
composable injectives of the category CX . The bijective correspondence is given by the
map which assigns to every indecomposable injective E of CX its left orthogonal – the full
subcategory ⊥ E generated by all objects M of CX such that CX (M, E) = 0.
In other words, Spec− (X) is isomorphic to the Gabriel spectrum of the category CX .
An advantage of the spectrum Spec− (X) is that it makes sense for all abelian cat-
egories, even those which do not have indecomposable injectives at all. For instance, if
CX is the category of coherent sheaves on a noetherian scheme, then its Gabriel spec-
trum is empty, while Spec− (X) coincides with Spec(X) and is homeomorphic to the the
underlying topological space of the scheme.

3.3. The spectra Spec(X), Spec1,1 −

t (X), and Spec (X). Let CX be a svelte
abelian category. For any subcategory P of the category CX , we denote by P t the inter-
section of all topologizing subcategories of CX properly containing P.
The elements of the spectrum Spec1,1 t (X) are all Serre subcategories P of CX such
t 1,1
that P 6= P. We endow Spect (X) with the specialization preorder ⊇.

3.3.1. Proposition. The spectrum Spec1,1 t (X) consists of all topologizing subcate-
def
gories P of the category CX such that Pt = P t P ⊥ is nonzero.
T

Proof. If P ∈ Spec1,1 t (X), i.e. P is a Serre subcategory of CX which is properly

t
contained in P , then it follows from 1.4.4(b) that Pt 6= 0.
Suppose now that P is a topologizing subcategory of CX such that Pt 6= 0. We claim
that then P is a Serre subcategory, i.e. P = P − .
In fact, let S be a topologizing subcategory
T ⊥ of CX which is not contained in P. Then
t
P • S contains P properly and (P • S) P ⊆ S. In particular, Pt ⊆ S. Since Pt 6= 0,
this implies that S is not contained in P − . This (and 1.4.2) shows that P = P − .

25
 For any subcategory Q of the category CX , we denote by Q b the union of all topologiz-
ing subcategories of CX which do not contain Q. It is easy to see, that for a pair Q1 , Q2
topologizing subcategories, Q1 ⊆ Q2 iff Q b1 ⊆ Qb2 .
If Q has one object, L, the subcategory Q is the union of all topologizing subcategories
b
of CX which do not contain L. We shall write hLi instead of Q. b

3.3.2. Proposition. (a) Spec1,1 −

t (X) ⊆ Spec (X).
b is an element of Spec1,1 (X) and the
(b) For any Q ∈ Spec(X), the subcategory Q t
map
Spec(X) −−−→ Spec1,1
t (X), Q 7−→ Q,
b

is an isomorphism of preorders.
Proof. (a) If P ∈ Spec1,1 t
t (X), then P /P is contained in and equivalent to the smallest
nonzero topologizing subcategory of CX /P.
(b1) If Q ∈ Spec(X), then Q b is a Serre subcategory.
In fact, suppose that Q b=6 Qb− , and let M be an object of Q b− which does not belong
to its subcategory Q. b The latter means that Q ⊆ [M ]. Let Q = [L] for some L ∈ Spec(X)
(cf. 2). The inclusion Q ⊆ [M ] means that L is a subquotient of a coproduct of a finite
number, M ⊕n , of copies of M . Since M ⊕n is an object of Q b− , the object L has a nonzero
subobject N which belongs to Q; b i.e. Q * [N ]. But, since L ∈ Spec(X), the subcategories
[N ] and [L] = Q coincide. Contradiction.
(b2) It follows from the definition of Q b that, for any subcategory Q, the subcategory
t b S Q.
Qb coincides with the intersection of all topologizing subcategories of CX containing Q
In particular, Q b belongs to Spec1,1 (X) whenever Q b is a Serre subcategory. Together with
t
(b1), this shows that the assignment Q 7−→ Q b induces a map Spec(X) −→ Spec1,1 (X).
t
(b3) Let P ∈ Spec1,1 t (X). It follows from 3.3.1 that P t 6
= 0. Moreover, by the
argument of 3.3.1, if T is a topologizing subcategory of CX such that T * P, then Pt =
P t P ⊥ ⊆ T.
T
(c) Let P ∈ Spec1,1 t
T ⊥
t (X). Every nonzero object of Pt = P P belongs to Spec(X).
Let L be a nonzero object of Pt and L1 its nonzero subobject of, hence [L1 ] ⊆ [L].
If [L1 ] * [L], then it follows from (b3) above that [L1 ] ⊆ P, or, equivalently, L1 ∈ ObP.
This contradicts to the assumption that the object L is P-torsion free.
(d) Let P ∈ Spec1,1 t
T ⊥
t (X). Then P = hLi for any nonzero object of Pt = P P .
Let L be a nonzero object of Pt . Since L does not belong to the Serre subcategory
hLi, by (b3), we have the inclusion hLi ⊆ P. On the other hand, if hLi $ P, then L ∈ ObP
which is not the case. Therefore P = hLi.
(e) The topologizing subcategory [Pt ] coincides with the subcategory [L] for any
nonzero object L of Pt .
Clearly [L] ⊆ [Pt ] for any L ∈ ObPt . By (b3), if Pt * [L], then [L] ⊆ P, hence L = 0.
Since, by (c), every nonzero object of Pt belongs to Spec(X), this shows that [Pt ] is
an element of Spec(X).
(f) It follows from the argument above that the map

Spec(X) −→ Spec1,1
t (X), Q 7−→ Q,
b

26
is inverse to the map Spec1,1
t (X) −→ Spec(X) which assigns to every P the topologizing
subcategory [Pt ].
3.3.2.1. Note. Let W be a submonoid of the monoid (T(X), •) of topologizing
1,1
subcategories of the category CX which is closed under arbitrary intersections. Let τW
denote the topology on Spec1,1 − −
t (X) induced by the topology τW on Spec (X) (cf. 3.2.2).
Then the map Spec(X) −→ Spec1,1 t (X) of 3.3.2 is an isomorphism from the topological
space (Spec(X), τW ) (defined in 2.5.1) and (Spec1,1 1,1
t (X), τW ).

3.3.3. The difference between Spect1,1 (X) and Spec− (X). Let CX be the
category R − mod of left modules over a commutative associative unital ring R. If the ring
R is noetherian, then the map which assigns to each prime ideal p of R the isomorphism
class of the injective hull of the quotient module R/p is an isomorphism between the
Gabriel spectrum of CX (which is, in this case, naturally isomorphic to Spec− (X)) and
the prime spectrum of the ring R [Gab, Ch.VI]. In this case, Spect1,1 (X) = Spec− (X),
i.e. the map Q 7−→ Q b is an isomorphism between Spec(X) and Spec− (X).
If a commutative ring R is not noetherian, the spectrum Spec− (X) might be much
bigger than (the image of) the prime spectrum Spec(R) of the ring R, while Spec(X)
(hence Spec1,1t (X)) is naturally isomorphic to Spec(R): the isomorphism is given by the
map which assigns to a prime ideal p the topologizing subcategory [R/p]; the inverse map
assigns to every element Q = [M ] of the spectrum Spec(X) the annihilator of the module
M.
3.4. Zariski topology on the spectra of a scheme.
f
3.4.1. Proposition. Let X −→ S be an affine morphism. Suppose that the category
CS has the property (sup) (cf. 1.4.5) and enough objects of finite type. Then the category
CX has enough objects of finite type.
Proof. Let M be an arbitrary nontrivial object. Since the direct image functor f∗
of f is conservative, the object f∗ is nontrivial. By assumption, there is a nonzero mor-
phism L −→ f∗ (M ), where L is an object of finite type. But, then the adjoint morphism
f ∗ (L) −→ M is nonzero. It remains to show that the inverse image functor f ∗ maps
objects of finite type to objects of finite type.
In fact, let L ∈ ObCS be an object of finite type, and let {Ni | i ∈ J} be a filtered
system of subobjects of f ∗ (L) such that supi∈J Ni −→ f ∗ (L) is an isomorphism. Since f∗
is exact and has a right adjoint, it preserves monomorphisms and colimits; in particular,
supi∈J f∗ (Ni ) −→ f∗ f ∗ (L) is an isomorphism. Since the object L is of finite type and
the system of subobjects {f∗ (Ni ) | i ∈ J} of f∗ f ∗ (L) is filtered, and the category CS
has the property (sup), the adjunction morphism L −→ f∗ f ∗ (L) factors through the
morphism f∗ (Ni ) −→ f∗ f ∗ (L) for some i ∈ J. This implies that the identical morphism
f ∗ (L) −→ f ∗ (L) factors through the monomorphism Ni −→ f ∗ (L); hence Ni −→ f ∗ (L) is
an isomorphism.
f
3.4.2. The Zariski topology on Spec− (X). Let X −→ S be a scheme over S (see
I.3.4.1). We assume that CX and CS are abelian categories, inverse image functors of f are
additive, and the category CS has the property (sup) and enough objects of finite type.

27
We call a subset V of the spectrum Spec− (X) Zariski open if for any open immersion
u
U −→ X (cf. I.3.5) which is affine over S (that is f ◦ u is affine), the intersection of (the
image of) Spec− (U ) and the subset V is a Zariski open subset of Spec− (U ).
We use here the fact that, by 3.4.1, the category CU has enough objects of finite type;
so that Zariski topology on Spec− (U ) is well defined (in 2.4).
We denote the Zariski topology on Spec− (X) by τz− .
A standard argument shows that a subset V of Spec− (X) is open iff the intersection
u f
of V with Spec− (Ui ) is open for some affine cover {Ui −→
i
X | i ∈ J} of X −→ S.
3.4.3. Zariski topology on other spectra. Zariski topology on the spectrum
Spec1,1 −
t (X) is induced by the Zariski topology on Spec (X) via the embedding

Spec1,1 −
t (X) ,→ Spec (X).

The Zariski topology on Spec(X) is determined by the requirement that the canonical
bijection Spec(X) −→∼
Spec1,1
t (X) is a homeomorphism with respect to Zariski topologies.

4. ’Locality’ theorems.
Let {Ti | i ∈ J} be a finite set of thick subcategories of an abelian
4.1. Proposition. \
category CX such that Ti = 0; and let u∗i be the localization functor CX −→ CX /Ti .
i∈J
The following conditions on a nonzero coreflective topologizing subcategory Q of CX are
equivalent:
(a) Q ∈ Spec(X),
(b) [u∗i (Q)] ∈ Spec(X/Ti ) for every i ∈ J such that Q * Ti .
Here [u∗i (Q)] denote the topologizing subcategory of CX /Ti spanned by u∗i (Q).
Proof. The assertion follows from [R4, 9.6.1].
4.1.1. Note. The condition (b) of 4.1 can be reformulated as follows:
(b’) For any i ∈ J, either u∗i (Q) = 0, or [u∗i (Q)] ∈ Spec(X/Ti ).
i u
4.2. Proposition. Let CX be an abelian category and U = {Ui −→ X | i ∈ J} a
u∗ i
finite set of continous morphisms such that {CX −→ CUi | i ∈ J} is a conservative family
of exact localizations.
uij
(a) The morphisms Uij = Ui ∩ Uj −→ Ui are continuous for all i, j ∈ J.
(b) Let Li be an object of Spec(Ui ); i.e. [Li ]c ∈ Spec(Ui ) and Li is hLi i-torsion free.
The following conditions are equivalent:
(i) Li ' u∗i (L) for some L ∈ Spec(X);
(ii) for any j ∈ J such that u∗ij (Li ) 6= 0, the object uji ∗ u∗ij (Li ) of CUj has an
associated point; i.e. it has a subobject Lij which belongs to Spec(Uj ).
Proof. The assertion follows from 4.1 (see [R4, 9.7.1]).
4.3. Examples. (a) If CX is the category of quasi-coherent sheaves on a quasi-
compact quasi-separated scheme X and each Ui is the category of quasi-coherent sheaves

28
on an open subscheme of X , then the glueing conditions[
of 4.2 hold for any Li ∈ Spec(Ui );
i.e. the spectrum Spec(X) is naturally identified with Spec(Ui ).
i∈J
(b) Similarly, if CX is the category of holonomic D-modules over a sheaf of twisted
ui
differential operators on a smooth quasi-compact scheme X , and {Ui −→ [ | i ∈ J} is a
X
cover of X corresponding to an open Zariski cover of X , then Spec(X) = Spec(Ui ).
i∈J
This is due to the functoriality of sheaves of holonomic modules with respect to direct
and inverse image functors of open immersions and the fact that holonomic modules are
of finite length (hence they have associated closed points).
i u
4.4. Proposition. Let CX be an abelian category and U = {Ui −→ X | i ∈ J} a
u∗
i
finite set of morphisms of ’spaces’ whose inverse image functors, {CX −→ CUi | i ∈ J},
∗
form a conservative family of exact localizations,
[ and Ker(ui ) is a coreflective subcategory
− −
for every i ∈ J. Then Spec (X) = Spec (Ui ).
i∈J

Proof. The equality is proven in [R4, 9.5].

5. Noncommutative k-schemes.
5.1. k-’Spaces’. Let k be an associative unital ring. A k-’space’ is a continuous
f f∗
morphism X −→ Sp(k), i.e. its inverse image functor, k − mod −→ CX has a right
adjoint, f∗ – the direct image functor of f . The object O = f ∗ (k) plays the role of the
structure sheaf on X. One can show that the object O together with the composition
ϕf ∼
k o −→ CX (O, O) of the isomorphism k o −→ Homk (k, k) (where k o is a ring opposite to k)
and a ring morphism Homk (k, k) −→ CX (f ∗ (k), f ∗ (k)) = CX (O, O) (due to the fact that
f ∗ is an additive functor) determine the functor f ∗ uniquely up to isomorphism; hence the
ϕf f
pair (O, k o −→ CX (O, O) determines the morphism X −→ Sp(k).
f
In particular, every Z-’space’ X −→ Sp(Z) is uniquely determined by the object
O = f ∗ (Z). There is a bijective correspondence between isomorphism classes of objects O
such that ’small’ coproducts of copies of O exist and Z-’space’ structures X −→ Sp(Z).
f
5.2. Affine k-’spaces’. The k-space X −→ Sp(k) is affine if f is affine, i.e. the
direct image functor f∗ is conservative (– reflects isomorphisms) and has a right adjoint,
f ! . This means, precisely, that the object O is a projective generator of finite type.
f
5.3. Noncommutative k-schemes. Consider a k-’space’ (X, X −→ Sp(k)) \ for
which there exists a family {Ti | i ∈ J} of Serre subcategories of CX such that Ti = 0
i∈J
f
and the compositions of X/Ti −→ X and X −→ Sp(k) are affine for all i ∈ J; i.e. (X, f )
is a weak scheme in the sense of I.2.2.
This weak scheme is a scheme, if for every i ∈ J, the set
def
V − (Ti ) = {P ∈ Spec− (X) | Ti * P}

29
is closed in Zariski topology.

6. Zariski geometric center and reconstruction of schemes.

6.1. The center of a category and localizations. Recall that the center, z(CY ),
of a svelte additive category CY is the ring of endomorphisms of its identical functor. If CY
is a category of left modules over a ring R, then the center of CY is naturally isomorphic
to the center of the ring R.
6.1.1. Proposition. Let CX be an abelian category and τ a topology on Spec(X).
The map O eX,τ which assigns to every open subset W of Spec(X) the center of the quotient
\
category CX /SW , where SW = Qb is a presheaf of commutative rings on (Spec(X), τ ).
Q∈W

Proof. This follows from a general (and easily verified) fact that the map which assigns
to a svelte category its center is functorial with respect to localization functors.
6.2. Zariski geometric center. Given a topology τ on Spec(X), we denote by
OX,τ the sheaf associated with the presheaf O eX,τ . The ringed space ((Spec(X), τ ), OX,τ )
is called the geometric center of (X, τ ). If τ is the Zariski topology, then we write simply
(Spec(X), OX ) and call this ringed space the Zariski geometric center of X.
6.2.1. Proposition. Suppose that the category CX has enough objects of finite type.
Then the Zariski geometric center of X is a locally ringed topological space.
Proof. Under the conditions, one can show that the stalk of the sheaf OX at a point
Q of the spectrum is isomorphic to the center of the local category CX /Q.b On the other
hand, the center of any local category (in particular, the center of CX /Q)
b is a local ring
(see [R, Ch. III]).
6.3. Commutative schemes which can be reconstructed from their cate-
gories of quasi-coherent or coherent sheaves. Let X = (X , OX ) be a ringed topo-
j
logical space and U = (U, OU ) −→ (X , OX ) an open immersion. Then the morphism j has
an exact inverse image functor j∗ and a fully faithful direct image functor j∗ . This implies
that Ker(j∗ ) is a Serre subcategory of the category OX − Mod of sheaves of OX -modules
and the unique functor

OX − Mod/Ker(j∗ ) −−−→ OU − Mod

induced by j∗ is an equivalence of categories [Gab, III.5].

Suppose now that X = (X , OX ) is a scheme and QcohX the category of quasi-coherent
sheaves on X. The inverse image functor j∗ of the immersion j maps quasi-coherent sheaves
to quasi-coherent sheaves. Let u∗ denote the functor QcohX −→ QcohU induced by
j∗ . The funtor u∗ , being the composition of the exact full embedding of QcohX into
OX − Mod and the exact functor j∗ , is exact; hence it is represented as the composition of
an exact localization QcohX −−−→ QcohX /Ker(u∗ ) and a uniquely defined exact functor
QcohX /Ker(u∗ ) −−−→ QcohU . If the direct image functor j∗ of the immersion j maps
quasi-coherent sheaves to quasi-coherent sheaves, then it induces a fully faithful functor

30
 u∗
QcohU −−−→ QcohX which is a right adjoint to u∗ . In particular, the canonical functor
QcohX /Ker(u∗ ) −−−→ QcohU is an equivalence of categories.
The reconstruction of a scheme X from the category QcohX of quasi-coherent sheaves
ui
on X is based on the existence of an affine cover {Ui −→ X | i ∈ J} such that the canonical
∗
functors QcohX /Ker(ui ) −−−→ QcohUi , i ∈ J, are category equivalences. It follows from
the discussion above (or from [GZ, I.2.5.2]) that this is garanteed if the inverse image
ui
functor QcohX −→ QcohUi has a fully faithful right adjoint. \ \
−
On the other hand, one can deduce from 4.1 (and the equality Ti− = Ti for
i∈J i∈J
any finite set {Ti | i ∈ J} of topologizing subcategories of CX ; see [R4, 4.1]) that if there
ui
exists an affine cover {Ui −→ X | i ∈ J} such that the canonical functors

QcohX /Ker(u∗i ) −−−→ QcohUi , i ∈ J,

are category equivalences, then Ker(u∗i ) is a Serre subcategory for all i ∈ J, which implies
(in these circumstances) the existence of a right adjoint ui∗ to u∗i for each i ∈ J.
6.4. Proposition. Let X = (X , OX ) be a quasi-compact quasi-separated commutative
scheme. Then the scheme X is isomorphic to the Zariski geometric center (Spec(X), OX )
of the ’space’ X represented by the category QcohX of quasi-coherent sheaves on X.
Proof. The assertion follows from the ’locality’ theorem 4.1. See details in [R4, 9.8].
6.4.1. Note. If the CX is the category of quasi-coherent sheaves on a quasi-
compact quasi-separated scheme, then Zariski topology on Spec(X) coincides with the
coarse Zariski topology (see 2.5.4).
6.5. The reduced geometric center of a ’space’. Fix a ’space’ X and a topology
τ on Spec(X). The map which assigns to each open subset U of Spec(X) the prime
spectrum of the ring OX,τ (U) of global sections of the sheaf OX,τ over U is a functor from
the category of Open(τ ) open sets of the topology τ to the category of topological spaces.
We denote its colimit by Spec(OX,τ ).
6.5.1. Proposition. There is a canonical morphism
pX
(Spec(X), τ ) −−−→ Spec(OX,τ ) (1)

of topological spaces.
Proof. Fix an element Q of Spec(X). For every open subset U of Spec(X) containing
Q, there is a ring morphism from OX,τ (U) to the center z(CX /Q) b of the local category
CX /Q.
b The ring z(CX /Q) b is local [R, Ch.III]. The preimage of its unique maximal ideal
in OX,τ (U) is a prime ideal of OX,τ (U). The image of this prime ideal in Spec(OX,τ ) does
not depend on the choice of U. This defines the map (1).
r
The map (1) is rarely injective. We denote by OX,τ the direct image pX ∗ (OX,τ ) of the
r
sheaf OX,τ and call the ringed topological space (Spec(OX,τ ), OX,τ ) the reduced geometric

31
center of (X, τ ). If τ is the Zariski topology, then we call the reduced geometric center of
(X, τ ) simply the reduced geometric center of X.
6.6. The reduced geometric center of a noncommutative scheme. One can
f
show that the noncommutative scheme structure on X −→ Sp(k) induces a scheme struc-
ture on the reduced geometric center of X. If CX is the category of quasi-coherent sheaves
on a commutative quasi-compact, quasi-separated scheme (X , OX ), then the reduced ge-
ometric center is naturally isomorphic to the geometric center of X. In particular, it is
isomorphic to the scheme.
6.7. Complement: a geometric realization of an abelian category. Let CX
be an abelian category. We have a contravariant pseudo-functor from the category of
the Zariski open sets of the spectrum Spec(X) to Cat which
\ assigns to each open set
U of Spec(X) the quotient category CX /SU , where SU = Q,
b and to each embedding
Q∈U
U ,→ V the corresponding localization functor. To this psedo-functor, there corresponds
(by a standard formalism) a fibered category over the Zariski topology of Spec(X). The
associated stack, FzX , is a stack of local categories: its stalk at each point Q of Spec(X)
is equivalent to the local category CX /Q. b
z
We regard the stack FX as a geometric realization of the abelian category CX .
If X is a (noncommutative) scheme, then the stack FzX is locally affine.
6.7.1. Note. Taking the center of each fiber of the stack FzX , we recover the presheaf
of commutative rings O
eX , hence the geometric center of the ’space’ X.

7. The spectrum Spec0c (X) and “big” schemes.

7.1. The spectrum Spec0c (X). If CX is the category of quasi-coherent sheaves on a
non-quasi-compact scheme, like, for instance, the flag variety of a Kac-Moody Lie algebra,
or a noncommutative scheme which does not have a finite affine cover (say, the quantum
flag variety of a Kac-Moody Lie algebra, or the corresponding quantum D-scheme), then
the spectrum Spec(X) is not sufficient. It should be replaced by the spectrum Spec0c (X)
whose elements are coreflective topologizing subcategories of CX of the form [M ]c (i.e.
generated by the object M ) such that if L is a nonzero subobject of M , then [L]c = [M ]c .
There is a natural map Spec(X) −→ Spec0c (X) which assigns to every Q ∈ Spec(X)
the smallest coreflective subactegory [Q]c containing Q.
If the category CX has enough objects of finite type, this canonical map is a bijection.
7.2. Supports, topologies. The support of an object M of the category CX in the
spectrum Spec0c (X) is the set Suppc (M ) of all Q ∈ Spec0c (X) which are contained in the
coreflective subcategory [M ]c generated by M .
The topologies on Spec0c (X) are defined via the same pattern as the topologies on
Spec(X) – either closed sets obtained as supports of certain family of objects, or as
supports of a family of topologizing subcategories (cf. 2.3 and 2.5).
7.2.1. The Zariski topology and the coarse Zariski topology. In particular,
the coarse Zariski topology on Spec0c (X) is defined similarly to the coarse Zariski topology

32
on Spec(X), using the monoid (under the Gabriel multiplication) of bireflective (that is
reflective and coreflective) topologizing subcategories. The Zariski topology on Spec0c (X)
is defined the same way as the Zariski topology on Spec(X) under condition that X is
a scheme over a ’space’ S and the category CS has property (sup) and enough objects of
finite type (see 3.4).
7.3. The locality theorem. The ’locality’ theorem for the spectrum Spec0c (X) is
as follows:
7.3.1. Proposition. Let {Ti |\i ∈ J} be a set of coreflective thick subcategories of
an abelian category CX such that Ti = 0; and let u∗i denote the localization functor
i∈J
CX −→ CX /Ti . The following conditions on a nonzero coreflective topologizing subcategory
Q of CX are equivalent:
(a) Q ∈ Spec0c (X),
(b) [u∗i (Q)]c ∈ Spec0c (X/Ti ) for every i ∈ J such that Q * Ti .
Proof. See [R4, 10.4.3].
7.4. The reconstruction of commutative schemes. The reconstruction theorem
for non-quasi-compact commutative schemes looks as follows.
7.4.1. Proposition. Let CX be the category of quasi-coherent sheaves on a com-
mutative scheme X = (X , O). Suppose that there is an affine cover {Ui ,→ X | i ∈ J}
of the scheme X such that all immersions Ui ,→ X , i ∈ J, have a direct image functor.
Then the geometric center (Spec0c (X), OX ) corresponding to the (coarse) Zariski topology
is isomorphic to the scheme X.
Proof. See [R4, 10.7.1].
7.4.2. Note. If the X = (X , O) is a quasi-compact and quasi-separated scheme, then
the category CX of quasi-coherent sheaves on X has enough objects of finite type, hence
the spectrum Spec0c (X) coincides with Spec(X). Thus, the reconstruction theorem 6.4 is
a special case of 7.4.1.

33
 Lecture 3. Pseudo-geometry and geometry of ’spaces’ represented by tri-
angulated categories.

This lecture is sketches the beginning of one of the simplest forms of derived noncom-
mutative geometry. Here ’spaces’ are represented by svelte triangulated categories (we call
them t-’spaces’) and morphisms by isomorphism classes of triangle functors. We start with
pseudo-geometry following pattern of Lecture 1, that is we consider continuous morphisms
and look for a triangulated version of Beck’s theorem (which plays a central role for study-
ing ’spaces’ represented by ordinary categories, incorporating both affine schemes and,
in the dual context, descent theory). The triangulated picture, turns to be much easier:
the triangulated version of Beck’s theorem on descent side states that every continuous
morphism is the composition of a comonadic morphism and a continuous localization. In
particular, any faithfully flat (in triangle sense) morphism is comonadic.
1,1
The geometric picture looks even better. There are two spectra, SpecL (X) and
1/2 1,1
SpecL (X) which are triangulated analogs of the spectra respectively Spect (X) and
1/2
Spec(X). There is a natural bijective map SpecL (X) −→ ∼
Spec1,1
L (X). But, unlike the
∼ 1,1
bijection Spec(X) −→ Spect (X) of II.3.3.2, this map does not preserve the specialization
preorder ⊇. Notice that the specialization preorder on Spec1,1 L (X) is what we expect
1,1
from specialization. So that the preorder (SpecL (X), ⊇) is regarded as the ”principal”
1/2
spectrum of the t-’space’ X. On the other hand, the points of the spectrum SpecL (X) are
closed with respect to the topology determined by the specialization preorder, or a natural
1/2
version of Zariski topology on SpecL (X). This gives certain technical advantages (not
used here) and curious interpretations.

1. ’Spaces’ represented by triangulated categories. Recall that a Z-category

is a category endowed with an action of Z, where Z is regarded as a monoidal category:
objects are elements and the tensor product is given by addition. In other words, a Z-
category is a category CX with an auto-equivalence θX and an associativity isomorphism
∼
θX ◦ (θX ◦ θX ) −→ (θX ◦ θX ) ◦ θX satisfying the usual cocycle conditions.
1.1. The category of triangulated categories. Triangulated k-linear categories
are triples (CX , θX ; TrX ), where (CX , θX ) is an additive k-linear Z-category, and TrX a full
subcategory of the category of diagrams of the form

L −→ M −→ N −→ θX (L).

The objects of the subcategory TrX are called triangles. They satisfy to well known
axioms due to Verdier [Ve1]. We denote a triangulated category (CX , θX ; TrX ) by CTX .
A triangle k-linear functor from a triangulated k-linear category CTX = (CX , θX ; TrX )
to a triangulated k-linear category CTY = (CY , θY ; TrY ) is a pair (F, φ), where F is a
∼
k-linear functor CX −→ CY and φ a functor isomorphism θY ◦ F −→ F ◦ θX such that for
any triangle L −→ M −→ N −→ θX (L) of CTX , the diagram

F (L) −→ F (M) −→ F (N ) −→ θY (F (L)),

34
where F (N ) −→ θY (F (L)) is the composition of F (N −→ θX (L)) and the isomorphism
φ(L)
F θX (L) −→ θY (F (L)), is a triangle of the triangulated category CTY .
We denote by TrCatk the category whose objects are svelte triangulated categories
and morphisms are triangle functors between them.
1.2. The category of t-’spaces’. If CTX = (CX , θX ; TrX ) is a svelte Karoubian
(that is the category CX is Karoubian) k-linear triangulated category, we say that it
f
represents a t-’space’ X. A morphism X −→ Y from a t-’space’ X to a t-’space’ Y is an
isomorphism class of triangle functors from CTY to CTX . A representative of a morphism
f will be called an inverse image functor of f and denoted, usually, by f∗ . The composition
f ◦ g is, by definition, the isomorphism class of the composition g ∗ ◦ f ∗ of inverse image
functors of respectively g and f . This defines the category EspTr of t-’spaces’.

2. Triangulated categories and Frobenius Z-categories. We need some facts

about abelianization of triangulated categories which are discussed in more general setting
and in bigger detail in Lecture 4.
For any k-linear category CX , we denote by Mk (X) the abelian category of presheaves
of k-modules on CX and by CXa the full subcategory of Mk (X) generated by all presheaves
of k-modules which a left resolution formed by representable presheaves. Since CXa con-
hX
tains all representable presheaves, the Yoneda functor CX −→ Mk (X) factors through the
embedding CXa −→ Mk (X). We denote the corestriction CX −→ CXa of the Yoneda func-
F a F
tor by HX . Every k-linear functor CX −→ CY induces a right exact functor CXa −→ CYa
F
CX −−−→ CY
hX y y hY
 
(1)
Fa
CXa −−−→ CYa
commutes. The functor Fa is determined uniquely up to isomorphism.
If CX is a Z-category, then the categories Mk (X) and CXa inherit a Z-action such that
F
the functors hX and HX become Z-functors. It follows that for every Z-functor CX −→ CY ,
Fa
the functor CXa −→ CYa is a Z-functor.
2.1. Frobenius abelian Z-categories. An exact k-linear Z-category is called
a Frobenius category if it has enough projectives and injectives and its projectives and
injectives coincide. In this lecture, we are interested only in abelian Frobenius categories.
We denote by FZ Catk the category whose objects are svelte Frobenius k-linear abelian Z-
categories and morphisms are exact k-linear functors which map projectives to projectives.
2.2. Theorem. (a) For any triangulated k-linear category CTX = (CX , θX ; TrX ), the
category CXa is a Frobenius abelian k-linear Z-category. If the category CX is Karoubian,
HX
then the canonical functor CX −→ CXa induces an equivalence between the category CX
and the full subcategory of CXa formed by its projectives.
(b) The correspondence CTX 7−→ CXa extends to a fully faithful functor from the
category TrCatk to the category FZ Catk .

35
 Proof. The assertion follows from Proposition IV.8.7.4. It is equivalent to a part of
Theorem 3.2.1 in [Ve2] (see Remark IV.8.7.5).

3. Localizations, continuous morphisms, and (co)monadic morphisms.

f
3.1. Localizations. Let X −→ Y be a morphism of t-’spaces’. Its inverse image
f∗
functor CY −→ CX is a composition of the localization at the thick subcategory Ker(f∗ )
and a faithful triangle functor. In other words, we have a canonical decomposition f = pf ◦fc ,
where p∗f is the localization functor CY −→ CY /Ker(f∗ ) and f∗c is a faithful triangle functor
determined (uniquely once f∗ is fixed, hence) uniquely up to isomorphism.
f
We call a morphism of t-’spaces’ X −→ Y a localization if fc is an isomorphism, or,
equivalently, if its inverse image functor is a category equivalence.
f
3.2. Continuous morphisms. We call a morphism X −→ Y continuous if its inverse
image functor, f∗ has a right adjoint, f∗ , and this right adjoint is a triangle functor.
3.3. Monads and comonads in triangulated categories. Let TCX = (CX , θX , TrX )
be a triangulated category. A monad on TCX (or a monad on the corresponding t-’space’ X)
µ
is a monad F = (F, µ) on the category CX such that F is a triangle functor and F 2 −→ F is
a morphism of triangle functors. Dually, a comonad on TCX (or X) is a comonad G = (G, δ)
δ
such that G is a triangle functor on TCX and G −→ G2 is a morphism of triangle functors.
The category (F/X) − mod of F-modules has a structure of triangulated category
f∗
induced by the forgetful functor (F/X) − mod −→ CX .
The following assertion is the triangulated version of Beck’s theorem.
3.4. Proposition. Let TCX and TCY be Karoubian triangulated categories and

f∗ f∗
TCY −−−→ TCX −−−→ TCY

a pair of adjoint triangle functors with adjunction morphisms

f ηf
f ∗ f∗ −→ IdTCX and IdTCY −→ f∗ f ∗ .

(a) The canonical functor

ef∗
TCY −−−→ Gf − Comod = (X\Gf ) − Comod, M 7−→ (f ∗ (M ), f ∗ ηf (M )), (1)

is a localization functor. It is a category equivalence iff the functor f ∗ is faithful.

(b) Dually, the canonical functor

ef∗
TCX −−−→ Ff − mod = (Ff /Y) − mod, L 7−→ (f∗ (L), f∗ f (L)), (2)

is a localization. It is a category equivalence iff the functor f∗ is faithful.

36
 Here Gf = (Gf , δf ) = (f ∗ f∗ , f ∗ ηf f∗ ) and Ff = (Ff , µf ) = (f∗ f ∗ , f∗ f f ∗ ) are respec-
tively the comonad and the monad associated with the pair of adjoint functors f ∗ , f∗ .
Proof. It suffices to prove (a), because the two assertions are dual to each other.
Let CXa denote the abelianization of the triangulated category TCX . Any triangle func-
F a F
tor TCX −→ TCY gives rise to an exact Z-functor CXa −→ CYa between the corresponding
abelian Z-categories which maps injective objects to injective objects. In particular, we
have a pair of adjoint exact Z-functors

fa∗ fa∗
CYa −−−→ CXa −−−→ CYa

which map injectives to injectives. Thus, we have the canonical functor

ef∗a
CYa −−−→ (Xa \Gfa ) − Comod.

Since both adjoint functors, fa∗ and fa∗ are exact functors between abelian categories, it
follows from Beck’s theorem that ef∗a is a localization functor. If the functor f ∗ is faithful,
then the functor fa∗ is faithful. This follows from the fact that every object M of the
category CXa is a quotient object of an object N of CX and a subobject of an object
β
L of TCX . Therefore, the composition N −→ L of the epimorphism N −→ M and a
monomorphism M −→ L is nonzero iff M is nonzero. Since the functor f ∗ is faithful,
f ∗ (β) 6= 0 whenever M 6= 0, which, in turn, implies that fa∗ (M ) 6= 0 if M 6= 0.
af∗
Since the functor CYa −→ CXa is exact and the category CYa is abelian, the faithfulness
of fa is equivalent to its conservativeness. The fact that fa∗ is conservative implies that ef∗a
∗

is conservative too. Therefore, being a localization functor, ef∗a is a category equivalence.

f
3.5. Continuous and (co)monadic morphisms of t-spaces. Let X −→ Y be a
continuous morphism. We call it comonadic (resp. monadic) if its inverse (resp. direct)
f
image functor is faithful. By 3.4, a continuous morphism X −→ Y is comonadic iff the
canonical functor

ef∗
TCY −−−→ Gf − Comod = (X\Gf ) − Comod, M 7−→ (f ∗ (M ), f ∗ ηf (M )), (1)

is a category equivalence. Dually, f is a monadic morphism iff

ef∗
TCX −−−→ Ff − mod = (Ff /Y) − mod, L 7−→ (f∗ (L), f∗ f (L)), (2)

is a category equivalence.
f
One can show that if X −→ Y is a continuous morphism, then both localization pf
and the ’faithful’ component fc in the decomposition f = pf ◦ fc (see 3.1) are continuous
morphisms. Thus, every continuous morphism of t-’spaces’ is a unique composition of a
continuous localization and a comonadic morphism.

37
 4. Presite of localizations.
4.1. Proposition. Let T CX be a svelte triangulated category
\
and {Ti \| i ∈ J} a finite
family of thick triangulated subcategories of T CX . Then Ti t S = (Ti t S) for
i∈J i∈J
any thick triangulated subcategory S.
Proof. Let CXa denote the abelianization of the triangulated category T CX . For a
triangulated subcategory T of T CX , let T a denote the smallest thick subcategory of CXa
generated by the image of T in CXa .
(a) If T is a thick triangulated subcategory of T CX , then T = T a CX .
T
In fact, objects of the subcategory T a are arbitrary subquotients of objects of T . Let
M be an object of CX which is a subquotient of an object N of CXa , i.e. there exists a
j e
diagram N −→ K −→ M in which j is a monomorphism and e is an epimorphism. Since
h
M is a projective object, the epimorphism e splits, i.e. there exists a morphism M −→ K
such that e ◦ h = idM . Since M is an injective object of CXa , the monomorphism j ◦ h splits.
If the object N belongs to the subcategory T , then M is also an object of T , because thick
subcategories contain all direct summands of all their objects.
(b) The equality (S t T )a = S a t T a holds for any pair S, T of thick triangulated
subcategories of T CX .
In fact, the squares
CX −−−→ CX/T C
Xa −−−→ CXa/T a
and
   
y y y y
CX /S −−−→ CX /(S t T ) CXa /S a −−−→ CXa /(S a t T a )
are cocartesian and the abelianization functor transforms cocartesian squares into cocarte-
sian squares, which implies that the unique functor

CXa /(S a t T a ) −−−→ CXa /(S t T )a

is a category equivalence.
\ \
a
(c) The equality Tai = Ti holds for any finite family {Ti | i ∈ J} of thick
i∈J i∈J
triangulated subcategories of T CX . \
Replacing T CX by T CX /T and Ti by Ti /T , where T = Ti , we reduce to the case
i∈J
\ \ \
Ti = 0. In this case, the claim is Tai = 0. The equality Ti = 0 means precisely
i∈J i∈J i∈J
that the triangle functor Y
T CX −−−→ T CX /Ti
i∈J

induced by the localization functors {T CX −→ T CX /Ti | i ∈ J} is faithful. But, then its

abelianization, Y
CXa −−−→ CXa /Tia ,
i∈J

38
 \
is a faithful functor, i.e. its kernel, the intersection Tai , equals to zero.
i∈J
(d) It follows from (a), (b) and (c) that
\ \ \ \ \
( Ti ) t S = (( Tia ) t S a ) CX = ( (Tia t S a )) CX =
i∈J i∈J i∈J
\ \ \
((Tia t S a ) CX ) = (Ti t S).
i∈J i∈J

for any finite family {S, Ti | i ∈ J} of thick triangulated subcategories of T CX .

4.2. Presite of exact localizations. Let EspL Tr denote the subcategory of the
category EspTr of t-’spaces’ whose objects are t-’spaces’ and morphisms are localizations
(i.e. their inverse image functors are compositions of localization functors and category
u
equivalences). We call a set {Ui →i X | i ∈ J} of morphisms of EspL Tr a cover of the
t-’space’ X if there is a finite subset J of J such that the family of inverse image functors
iu∗
{TCX −→ TCUi | i ∈ J} is conservative. We denote the set of all such covers of X by Tf (X).
4.2.1. Proposition. The covers defined above form a pretopology, Tf , on EspL
Tr .

Proof. The morphisms of the subcategory EspL Tr are determined, uniquely up to

isomorphism, by the kernel of their inverse image functors. A family of inverse image
iu∗
functors {TCX −→ TCUi | i ∈ J} is conservative iff the intersection of kernels of these
inverse image functors is zero. The assertion follows now from 4.1.
5. The spectra of t-’spaces’.
Fix a svelte triangulated category CTX = (CX , θX , TrX ). Here θX denote its suspension
functor and TrX its category of triangles (otherwise called admissible triangles). We denote
by Tht(X) the preorder (with respect to the inclusion) of all thick triangulated subcategories
of CTX . Recall that a full triangulated subcategory of CTX is called thick if it contains all
direct summands of its objects.
5.1. Spec1L (X) and its decompositions. For any triangulated subcategory T of
CTX , let T ? denote the intersection of all triangulated subcategories of CTX which contain
T properly. And let T? be the intersection of T ? and T ⊥ – the right orthogonal to
T . Recall that T ⊥ is the full subcategory of CTX generated by all objects N such that
CTX (N, M ) = 0 for all M ∈ ObT . It follows that T ⊥ is triangulated (for any subcategory
T which is stable by the translation functor).
We denote by Spec1L (X) the subpreorder of Tht(X) formed by all thick triangulated
subcategories P for which P ? 6= P. We have a decomposition
a
Spec1L (X) = Spec1,1 L (X) Spec1,0
L (X)

of Spec1L (X) into a disjoint union of

Spec1,1
L (X) = {P ∈ Tht(X) | P? 6= 0} and
Spec1,0 1
L (X) = {P ∈ SpecL (X) | P? = 0}.

39
 5.2. L-Local triangulated categories and Spec1,1 L (X). We call a triangulated
category CTY L-local if it has the smallest nonzero thick triangulated subcategory.

5.2.1. Proposition. Let P ∈ Spec1,1 L (X). Then

⊥
(a) P = P∗ .
(b) The triangualted category P ⊥ is L-local and P? is its smallest nonzero thick tri-
angulated subcategory.
Proof. See [R7, 12.7.1].
5.2.2. Proposition. Suppose that infinite coproducts or products exist in CTX . Let P
be a thick triangulated subcategory of CTX . Then the following properties of are equivalent:
(i) P? = P ⊥ ∩ P ? is nonzero, i.e. P ∈ Spec1,1
L (X);
(ii) P belongs to Spec1L (X) and the composition of the inclusion P? ,→ CTX and
Pq∗
the localization functor CTX −→ CTX /P induces an equivalence of triangulated categories
∼ ?
P? −→ P /P.
(iii) P belongs to Spec1L (X) and the inclusion functor P ,→ P ? has a right adjoint.
(iv) P belongs to Spec1L (X) and P ⊥ is nonzero.
Proof. See [R7, 12.7.3].
5.2.3. Corollary. Suppose that infinite coproducts or products exist in CTX . Then
Spec1,0 1 ∗ ⊥
L (X) consists of all P ∈ SpecL (X) such that P 6= P and P = 0.

5.2.4. Remark. Loosely, 5.2.3 says that the elements of Spec1,0

L (X) can be regarded
as ”fat” points – they generate (in a weak sense) the whole category CTX .
5.2.5. Proposition. (a) Let {T\ i | i ∈ J} be a finite set of thick subcategories of a
triangulated category T CX such that Ti = 0. Then
i∈J

[
Spec1L (X) = Spec1L (X/Ti ) (1)
i∈J

⊥
(b) Suppose that (Ti⊥ ) = Ti for all i ∈ J. Then
[
Spec1,1
L (X) = Spec1,1
L (X/Ti ) (2)
i∈J

[
Proof. (a) The inclusion Spec1L (X/Ti ) ⊆ Spec1L (X) follows from the functoriality
i∈J
of Spec1L (−) with respect to localizations. Let P ∈ Spec1L (X). By 4.1,
\
\
P= Ti t P = (Ti t P) (3)
i∈J i∈J

40
which implies that Ti ⊆ P for some i ∈ J. In fact, if Ti * P for all i ∈ J, then Ti t P
\
contains properly Pi for all i ∈ J, hence the intersection (Ti t P) contains properly P,
i∈J
which contradicts to (3).[This proves the inverse inclusion, that is the equality (1).
(b) The inclusion Spec1,1 1,1
L (X/Ti ) ⊆ SpecL (X) follows from the functoriality of
i∈J
Spec1,1
L (−) with respect to localizations at thick subcategories T such that
⊥
(T ⊥ ) = T .
The inverse inclusion follows from (a).
1/2 1/2
5.3. The spectrum SpecL (X). Let SpecL (X) denote the full subpreorder of
Tht(X) whose objects are thick triangulated subcategories Q such that ⊥ Q belongs to
Spec1L (X) and every thick triangulated subcategory of CTX properly containing ⊥ Q con-
tains Q; i.e. ⊥ Q ∨ Q is the smallest thick triangulated subcategory of CTX properly
containing ⊥ Q.
5.3.1. Proposition. (a) The map Q 7−→ ⊥ Q induces a bijective map
1/2 ∼
SpecL (X) −→ Spec1,1
L (X). (1)

1/2
(b) If Q is an object of SpecL (X), then Q is a minimal nonzero thick triangulated
subcategory of CTX .
(c) Suppose that CTX has infinite coproducts or products. Then the following properties
of a thick triangulated subcategory Q are equivalent:
1/2
(i) Q belongs to SpecL (X);
(ii) Q is a minimal nonzero thick triangulated subcategory of CTX such that ⊥ Q
belongs to Spec1L (X).
Proof. See [R7, 12.8.1].
5.4. Flat spectra. Recall the following assertion which is due to Verdier:
5.4.1. Proposition [Ve1, 10–5]. Let T be a thick triangulated subcategory of CTX ,
and let
ι∗ qT∗
T −→T
CTX −→ CTX /T
be the inclusion and localization functors. The following properties are equivalent:
(a) The functor ι∗T has a right adjoint.
(b) The functor qT∗ has a right adjoint.
Let Se(X) denote the family of all thick triangulated subcategories of the triangulated
category CTX which satisfy equivalent conditions of 12.10.3. We define the complete flat
spectrum of X, Spec1fL (X), by setting
\
Spec1fL (X) = Spec1L (X) Se(X). (1)

We define the flat spectrum of X as a full subpreorder, Spec0fL (X), of Tht(X) whose
b ∈ Spec1fL (X)}.
objects are all P such that P

41
 It follows from these definitions that the map P 7−→ P
b defines an injective morphism

Spec0fL (X) −−−→ Spec1fL (X). (2)

1/2 1/2
Let SpecfL (X) denote the full subpreorder of SpecL (X) whose objects are all Q
such that ⊥ Q belongs to Se(X).
5.4.2. Proposition. (a) The map

Tht(X) −→ Tht(X), Q 7−→ ⊥ Q,

induces an isomorphism
1/2
SpecfL (X) −→∼
Spec1fL (X). (3)
1/2
(b) Spec0fL (X) = Spec0L (X) SpecfL (X). The canonical morphism (2) is the com-
T
1/2
position of the inclusion Spec0fL (X) ,→ SpecfL (X) and the isomorphism (3).
Proof. See [R7, 12.10.1].
5.5. Supports and Zariski topology.
5.5.1. Supports. For any object M of the category CX , the support of M in
Spec1L (X) is defined by Supp1 1
L (M ) = {P ∈ SpecL (X) | M 6∈ ObP}. It follows that
Supp1L (L ⊕ M) = Supp1L (L) Supp1L (M).
S

1,1
5.5.2. Topologies on Spec1L (X) and SpecL (X). We follow the pattern of 2.4.
Let Ξ be a class of objects of CX closed under
\ finite coproducts. For any set E of objects
1 1
of Xi, let VL (E) denote the intersection SuppL (M ). Then, for any family {Ei | i ∈ I}
M ∈E
of such sets, we have, evidently,
[ \
V( Ei ) = V(Ei ).
i∈J i∈J

It follows from the equality Supp1L (M ⊕ N ) = Supp1L (M ) Supp1L (N ) (see 2.2.1(a))

S

that VL1 (E ⊕ E)
e = V 1 (E) S V 1 (E).
e Here E ⊕ E e def
= {M ⊕ N | M ∈ E, N ∈ E}.
e
L L
1 1
This shows that the subsets VL (E) of SpecL (X), where E runs through subsets of
Ξ, are all closed sets of a topology, τΞ1 , on the spectrum Spec1L (X).
We denote by τΞ1,1 the induced topology on SpecL 1,1
(X).
5.5.3. Compact topology. The class Ξc (X) of compact objects of the category
CX is closed under finite coproducts, hence it defines a topology on Spec1L (X) which we
denote by τc and call the compact topology.
1,1
Restricting the compact topology to SpecL (X) or to Spec1fL (X), we obtain the com-
pact topology on these spectra.
5.5.4. Zariski topology on Spec1,1 L (X). We define the Zariski topology on the
1,1
spectrum SpecL (X) by taking as a base of closed sets the supports of compact objects
and closures (i.e. the sets of all specializations) of points of Spec1,1
L (X).

42
 If the category CX is generated by compact objects, then the Zariski topology coincides
with the compact topology τc .
1/2
5.5.5. Zariski topology on SpecL (X). It is important to realize that the topolo-
gies we define are determined in the first place by the choice of a preorder on the set of
thick subcategories (or topologizing subcategories in the case of abelian categories). And
so far, the preorder was always the inverse inclusion.
Following these pattern, for any object M of a svelte triangulated category CTX , we
1/2 1/2
define the support of M in SpecL (X) as the set of all Q ∈ SpecL (X) such that the
smallest thick tringulated subcategory [M ]tr containing M contains also Q.
1/2
We define the Zariski topology on SpecL (X) by taking supports of compact objects
1/2
and the finite subsets of SpecL (X) as a base of its closed sets.
It follows from this definition of Zariski topology and 5.2.1(b) that all points of the
1/2 1/2
spectrum SpecL (X) are closed; that is Zariski topology on SpecL (X) is a T1 -topology.
The bijective map
1/2
SpecL (X) −→ ∼
Spec1,1
L (X) (4)
is continuous, but, usually, not a homeomorphism.
5.5.6. Remark. Suppose that CX is the heart of a t-structure on CTX . Then we
have a commutative diagram
1/2
Spec(X)
 −−−→ SpecL (X)
oy yo (5)
 

Spec1,1
t (X) −−−→ 1,1
SpecL (X)

where horizontal arrows are embeddings and vertical arrows are canonical bijections. Thus,
1/2
the Zariski topology on SpecL (X) induces a T1 -topology on the spectrum Spec(X) of
the ’space’ represented by the abelian category CX , which, obviously, differs from Zariski
topology on Spec(X), unless Spec(X) is of zero Krull dimension.
5.6. A geometric realization of a triangulated category. We follow an obvious
modification of the pattern of 6.7. Namely, we assign to a Karoubian triangulated category
TCX having a set of compact generators the contravariant pseudo-functor from the category
1,1
of Zariski open subsets of the spectrum SpecL (X) to the category of svelte triangulated
categories. The associated stack is the stack of local triangulated categories.
5.7. The geometric center. We define the center of a svelte triangulated category
TCY = (CY , θY , TrY ) as the subring OT (Y) of the center z(CX ) of the category CY formed
by θY -invariant endomorphisms of the identical functor of CY . One can show that the ring
OT (Y) is local if the triangulated category TCY is local.
Let TCX be a Karoubian triangulated category with a set of compact generators and
TFZX the corresponding stack of local triangulated categories (cf. 5.6). Assigning to each
fiber of the stack TFZX its center, we obtain a presheaf of commutative rings on the spectrum
Spec1,1 T
L (X) endowed with the Zariski topology. The associated sheaf, OX , is a sheaf of

43
local rings. We call the locally ringed topological space (Spec1,1 T
L (X), OX ) the geometric
centrum of the triangulated category TCX .
5.7.1. Note. Similarly to the abelian case, one can define the reduced geometric
centrum of TCX . Details of this construction are left to the reader.
5.8. On the spectra of a monoidal triangulated category.
5.8.1. A remark on spectral cuisine. There are certain rather simple general
pattern of producing spectra starting from a preorder (they are outlined in [R6]). Here, in
Section 8, these pattern are applied to the preorder Tht(X) of thick triangulated subcate-
gories of the triangulated category CTX .
5.8.2. Application to monoidal triangulated categories. Suppose that a trian-
gulated category TCX has a structure of a monoidal category. Then, replacing the preorder
of thick subcategories with the preorder of those thick subcategories which are ideals of
TCX and mimiking the definitions of Spec1L (X) and Spec1,1 L (X), we obtain the spectra
1 1,1
respectively SpecL,⊗ (X) and SpecL,⊗ (X). If the monoidal category TCX is symmetric,
then Spec1L,⊗ (X) coincides with the spectrum introduced by P. Balmer in different terms,
as a straightforward imitation of the notion of a prime ideal of a commutative ring. This
spectrum has nice properties. Unfortunately, triangulated categories associated with non-
commutative ’spaces’ of interest do not have any symmetric monoidal structure.
It is not clear at the moment what might be the role of the spectra Spec1L,⊗ (X) and
Spec1,1
L,⊗ (X) (if any) in the case of a non-symmetric monoidal category.

44
 Lecture 4. Non-abelian homological algebra.

The preliminaries are dedicated to kernels of arrows of arbitrary categories with initial
objects. They are complemented in Appendix. In the treatment of non-abelian homological
algebra, we adopt here an intermediate level of generality – right or left exact (instead of
fibred or cofibred) categories, which turns the main body of this text into an exercise on
satellites along the lines of [Gr], in which abelian categories are replaced by right (or left)
exact categories with initial (resp. final) objects. An analysis of obtained facts leads to the
notions of stable category of a left exact category and to the notions of quasi-suspended
and quasi-triangulated categories.

1. Preliminaries: kernels and cokernels of morphisms.

f
Let CX be a category with an initial object, x. For a morphism M −→ N we define
the kernel of f as the upper horizontal arrow in a cartesian square

k(f )
 ) −−−→
Ker(f M

f 0y cart yf
 
x −−−→ N

when the latter exists.

Cokernels of morphisms are defined dually, via a cocartesian square

c(f )
Nx −−−→ Cok(f
x )
 0
f  cocart f

M −−−→ y

where y is a final object of CX .

If CX is a pointed category (i.e. its initial objects are final), then the notion of the
f
k(f )
kernel is equivalent to the usual one: the diagram Ker(f ) −−−→ M −−−→
−−−→ N is exact.
0
f
c(f )
Dually, the cokernel of f makes the diagram M −−−→ N −−
−−−→ −→ Cok(f ) exact.
0

1.1. Lemma. Let CX be a category with an initial object x.

f
(a) Let a morphism M −→ N of CX have a kernel. The canonical morphism
k(f ) iN
Ker(f ) −−−→ M is a monomorphism, if the unique arrow x −→ N is a monomorphism.
f M i
(b) If M −→ N is a monomorphism, then x −→ M is the kernel of f .
Proof. The pull-backs of monomorphisms are monomorphisms.
1.2. Corollary. Let CX be a category with an initial object x. The following condi-
tions are equivalent:

45
 f k(f )
(a) If M −→ N has a kernel, then the canonical arrow Ker(f ) −−−→ M is a
monomorphism.
iM
(b) The unique arrow x −→ M is a monomorphism for any M ∈ ObCX .
M i
Proof. (a) ⇒ (b), because, by 1.1(b), the unique morphism x −→ M is the kernel of
the identical morphism M −→ M . The implication (b) ⇒ (a) follows from 1.1(a).
1.3. Note. The converse assertion is not true in general: a morphism might have
a trivial kernel without being a monomorphism. It is easy to produce an example in the
category of pointed sets.
1.4. Examples.
1.4.1. Kernels of morphisms of unital k-algebras. Let CX be the category Algk
of associative unital k-algebras. The category CX has an initial object – the k-algebra k.
ϕ
For any k-algebra morphism A −→ B, we have a commutative square
ϕ
Ax −−−→ B
x
k(ϕ) 
 

(ϕ)
k ⊕ K(ϕ) −−−→ k

where K(ϕ) denote the kernel of the morphism ϕ in the category of non-unital k-algebras
and the morphism k(ϕ) is determined by the inclusion K(ϕ) −→ A and the k-algebra
structure k −→ A. This square is cartesian. In fact, if
ϕ
x −−−→ B
A x
γ
 

ψ
C −−−→ k

is a commutative square of k-algebra morphisms, then C is an augmented algebra: C =

k ⊕ K(ψ). Since the restriction of ϕ ◦ γ to K(ψ) is zero, it factors uniquely through K(ϕ).
β
Therefore, there is a unique k-algebra morphism C = k ⊕ K(ψ) −→ Ker(ϕ) = k ⊕ K(ϕ)
such that γ = k(ϕ) ◦ β and ψ = (ϕ) ◦ β.
ϕ
This shows that each (unital) k-algebra morphism A −→ B has a canonical kernel
Ker(ϕ) equal to the augmented k-algebra corresponding to the ideal K(ϕ).
k(ϕ)
It follows from the description of the kernel Ker(ϕ) −−−→ A that it is a monomor-
phism iff the k-algebra structure k −→ A is a monomorphism.
Notice that cokernels of morphisms are not defined in Algk , because this category
does not have final objects.
1.4.2. Kernels and cokernels of maps of sets. Since the only initial object of the
category Sets is the empty set ∅ and there are no morphisms from a non-empty set to ∅, the

46
 f
kernel of any map X −→ Y is ∅ −→ X. The cokernel of a map X −→ Y is the projection
c(f )
Y −−−→ Y /f (X), where Y /f (X) is the set obtained from Y by the contraction of f (X)
into a point. So that c(f ) is an isomorphism iff either X = ∅, or f (X) is a one-point set.
∧
1.4.3. Presheaves of sets. Let CX be a svelte category and CX the category of
non-trivial presheaves of sets on CX (that is we exclude the trivial presheaf which assigns
∧
to every object of CX the empty set). The category CX has a final object which is the
constant presheaf with values in a one-element set. If CX has a final object, y, then
∧ ∧
yb = CX (−, y) is a final object of the category CX . Since CX has small colimits, it has
cokernels of arbitrary morphisms which are computed object-wise, that is using 1.4.2.
If the category CX has an initial object, x, then the presheaf x b = CX (−, x) is an
∧ ∧
initial object of the category CX . In this case, the category CX has kernels of all its
∧ h ∧
morphisms (because CX has limits) and the Yoneda functor CX −→ CX preserves kernels.
∧
Notice that the initial object of CX is not isomorphic to its final object unless the
category CX is pointed, i.e. initial objects of CX are its final objects.
1.5. Some properties of kernels. See Appendix.

2. Right exact categories and (right) ’exact’ functors.

We define a right exact category as a pair (CX , EX ), where CX is a category and
EX is a pretopology on CX whose covers are strict epimorphisms; that is for any element
M −→ L of E (– a cover), the diagram M ×L M −→ −→ M −→ L is exact. This requirement
means precisely that the pretopology EX is subcanonical; i.e. every representable presheaf
of sets on CX is a sheaf. We call the elements of EX deflations and assume that all
isomorphisms are deflations.
2.1. The coarsest and the finest right exact structures. The coarsest right ex-
act structure on a category CX is the discrete pretopology: the class of deflations coincides
with the class Iso(CX ) of all isomorphisms of the category CX .
Let EsX denote the class of all universally strict epimorphisms of CX ; i.e. elements of
e f
EsX are strict epimorphisms M −→ N such that for any morphism N e −→ N , there exists
a cartesian square
fe
M
f −−−→ M 
ee y cart y e
 
f
N
e −−−→ N
whose left vertical arrow is a strict epimorphism. It follows that EsX is the finest right
exact structure on the category CX . We call this structure canonical.
If CX is an abelian category or a topos, then EsX consists of all epimorphisms.
If CX is a quasi-abelian category, then EsX consists of all strict epimorphisms.
2.2. Right ’exact’ and ’exact’ functors. Let (CX , EX ) and (CY , EY ) be right
F
exact categories. A functor CX −→ CY will be called right ’exact’ (resp. ’exact’) if it
e
maps deflations to deflations and for any deflation M −→ N of EX and any morphism

47
 f
Ne −→ N , the canonical arrow F (Ne ×N M ) −−−→ F (N e ) ×F (N ) F (M ) is a deflation (resp.
an isomorphism).
In other words, the functor F is ’exact’ if it maps deflations to deflations and preserves
pull-backs of deflations.
F
2.3. Weakly right ’exact’ and weakly ’exact’ functors. A functor CX −→ CY
is called weakly right ’exact’ (resp. weakly ’exact’) if it maps deflations to deflations and for
any arrow M −→ N of EX , the canonical morphism F (M ×N M ) −→ F (M ) ×F (N ) F (M )
is a deflation (resp. an isomorphism). In particular, weakly ’exact’ functors are weakly
right ’exact’.
2.4. Note. Of cause, ’exact’ (resp. right ’exact’) functors are weakly ’exact’ (resp.
weakly right ’exact’). In the additive (actually a more general) case, weakly ’exact’ functors
are ’exact’ (see 2.7 and 2.7.2).
2.5. Interpretation: ’spaces’ represented by right exact categories. Weakly
right ’exact’ functors will be interpreted as inverse image functors of morphisms between
’spaces’ represented by right exact categories. We consider the category Espw r whose
objects are pairs (X, EX ), where (CX , EX ) is a svelte right exact category. A morphism
ϕ
from (X, EX ) to (Y, EY ) is a morphism of ’spaces’ X −→ Y whose inverse image functor
ϕ∗
CY −→ CX is a weakly right ’exact’ functor from (CY , EY ) to (CX , EX ). The map which
assigns to every ’space’ X the pair (X, Iso(CX )) is a full embedding of the category |Cat|o
of ’spaces’ into the category Espwr . This full embedding is a right adjoint functor to the
forgetful functor
o
Espw
r −−−→ |Cat| , (X, EX ) 7−→ X.

2.5.1. Proposition. Let (CX , EX ) and (CY , EY ) be additive right exact categories
F
and CX −→ CY an additive functor. Then
(a) The functor F is weakly right ’exact’ iff it maps deflations to deflations and the
sequence
F (e)
F (Ker(e)) −−−→ F (M ) −−−→ F (N ) −−−→ 0
e
is exact for any deflation M −→ N .
(b) The functor F is weakly ’exact’ iff it maps deflations to deflations and the sequence

F (e)
0 −−−→ F (Ker(e)) −−−→ F (M ) −−−→ F (N ) −−−→ 0
e
is ’exact’ for any deflation M −→ N .
Proof. See A.2(b).
2.6. Conflations and fully exact subcategories of a right exact category.
Fix a right exact category (CX , EX ) with an initial object x. We denote by EX the class
k e k
of all sequences of the form K −→ M −→ N , where e ∈ EX and K −→ M is a kernel of e.
Expanding the terminology of exact additive categories, we call such sequences conflations.

48
 2.6.1. Fully exact subcategories of a right exact category. We call a full
subcategory B of CX a fully exact subcategory of the right exact category (CX , EX ), if B
contains the initial object x and is closed under extensions; i.e. if objects K and N in a
k e
conflation K −→ M −→ N belong to B, then M is an object of B.
In particular, fully exact subcategories of (CX , EX ) are strictly full subcategories.
2.6.2. Proposition. Let (CX , EX ) be a right exact category with an initial object x
e
and B its fully exact subcategory. Then the class EX,B of all deflations M −→ N such
that M, N, and Ker(e) are objects of B is a structure of a right exact category on B such
that the inclusion functor B −→ CX is an ’exact’ functor (B, EX,B ) −→ (CX , EX ).
Proof. The argument is an application of facts of Appendix.
2.6.3. Remark. Let (CX , EX ) be a right exact category with an initial object x and
B its strictly full subcategory containing x. Let E be a right exact structure on B such
J
that the inclusion functor B −→ CX maps deflations to deflations and preserves kernels
of deflations. Then E is contained in EX,B . In particular, E is contained in EX,B if the
inclusion functor is an ’exact’ functor from (B, E) to (CX , EX ). This shows that if B is
a fully exact subcategory of (CX , EX ), then EX,B is the finest right exact structure on B
such that the inclusion functor B −→ CX is an exact functor from (B, EX,B ) to (CX , EX ).
2.7. Proposition. Let (CX , EX ) and (CY , EY ) be right exact categories and F a
functor CX −→ CY which maps conflations to conflations. Suppose that the category CY
is additive. Then the functor F is ’exact’.
2.7.1. Corollary. Let (CX , EX ) and (CY , EY ) be additive k-linear right exact cate-
gories and F an additive functor CX −→ CY . Then the functor F is weakly ’exact’ iff it
is ’exact’.
F
Proof. By 2.5.1, a k-linear functor CX −→ CY is a weakly ’exact’ iff it maps conflations
to conflations. The assertion follows now from 2.7.
2.7.2. The property (†). In Proposition 2.7, the assumption that the category CY
is additive is used only at the end of the proof (part (b)). Moreover, additivity appears
there only because it garantees the following property:
(†) if the rows of a commutative diagram

e −−−→ M
L
  −−−→ N
f e

  
y y y
L −−−→ M −−−→ N

are conflations and its right and left vertical arrows are isomorphisms, then the middle
arrow is an isomorphism.
So that the additivity of CY in 2.7 can be replaced by the property (†) for (CY , EY ).
2.7.3. An observation. The following obvious observation helps to establish the
property (†) for many non-additive right exact categories:

49
 F
If (CX , EX ) and (CY , EY ) are right exact categories and CX −→ CY is a conservative
functor which maps conflations to conflations, then the property (†) holds in (CX , EX )
provided it holds in (CY , EY ).
2.7.3.1. Example. Let (CY , EY ) be right exact k-linear category, (CX , EX ) a right
F
exact category, and CX −→ CY is a conservative functor which maps conflations to con-
flations. Then the property (†) holds in (CX , EX ).
For instance, the property (†) holds for the right exact category (Algk , Es ) of associa-
tive unital k-algebras with strict epimorphisms as deflations, because the forgetful functor
f∗
Algk −→ k − mod is conservative, maps deflations to deflations (that is to epimorphisms)
and is left exact. Therefore, it maps conflations to conflations.
2.8. Proposition. (a) Let (CX , EX ) be a svelte right exact category. The Yoneda
j∗
X
embedding induces an ’exact’ fully faithful functor (CX , EX ) −−−→ (CXE , EsXE ), where
CXE is the category of sheaves of sets on the presite (CX , EX ) and EsXE the family of all
universally strict epimorphisms of CXE (– the canonical structure of a right exact category).
ϕ∗
(b) Let (CX , EX ) and (CY , EY ) be right exact categories and (CX , EX ) −−−→ (CY , EY )
e∗
ϕ
a weakly right ’exact’ functor. There exists a functor CXE −−−→ CYE such that the diagram
ϕ∗
CX −−−→ C
Y
j∗X
 ∗
y jY

y (1)
∗
ϕ
e
CXE −−−→ C YE

quasi commutes, i.e. ϕe∗ j∗X ' j∗Y ϕ∗ . The functor ϕ

e∗ is defined uniquely up to isomorphism
and has a right adjoint, ϕ e∗ .
Proof. (a) Since the right exact structure EX of CX is a subcanonical pretopology,
the Yoneda embedding takes values in the category CXE of sheaves on (CX , EX ), hence
it induces a full embedding of CX into CXE which preserves all small limits and maps
deflations to deflations. In particular it is an ’exact’ functor from (CX , EX ) to (CXE , EsXE ).
(b) Every weakly right exact functor (CX , EX ) −→ (CY , EY ) determines a continuous
(i.e. having a right adjoint) functor between the categories of presheaves of sets, which is
compatible with the sheafification functor, hence determines uniquely a continuous functor
between the corresponding categories of sheaves making commute the diagram (1).
2.9. Application: right exact additive categories and exact categories.
2.9.1. Proposition. Let (CX , EX ) be an additive k-linear right exact category.
Then there exists an exact category (CXe , EXe ) and a fully faithful k-linear ’exact’ functor
∗
γX
(CX , EX ) −−−→ (CXe , EXe ) which is universal; that is any ’exact’ k-linear functor from
∗
(CX , EX ) to an exact k-linear category factorizes uniquely through γX .
Proof. We take as CXe the smallest fully exact subcategory of the category CXE of
sheaves of k-modules on (CX , EX ) containing all representable sheaves. Objects of the

50
category CXe are sheaves F such that there exists a finite filtration

0 = F0 −→ F1 −→ . . . −→ Fn = F

such that Fm /Fm−1 is representable for 1 ≤ m ≤ n. The subcategory CXe , being a fully
exact subcategory of an abelian category, is exact. The remaining details are left as an
exercise.

3. Satellites in right exact categories.

3.1. Preliminaries: trivial morphisms, pointed objects, and complexes. Let
CX be a category with initial objects. We call a morphism of CX trivial if it factors through
an initial object. It follows that an object M is initial iff idM is a trivial morphism. If CX
is a pointed category, then the trivial morphisms are usually called zero morphisms.
3.1.1. Trivial compositions and pointed objects. If the composition of arrows
f g
L −→ M −→ N is trivial, i.e. there is a commutative square
f
 −−−→
L M

ξy yg
 
iN
x −−−→ N

where x is an initial object, and the morphism g has a kernel, then f is the composition of
k(g) fg
the canonical arrow Ker(g) −→ M and a morphism L −→ Ker(g) uniquely determined
iN
by f and ξ. If the arrow x −→ N is a monomorphism, then the morphism ξ is uniquely
determined by f and g ; therefore in this case, the arrow fg does not depend on ξ.
3.1.1.1. Pointed objects. In particular, fg does not depend on ξ, if N is a pointed
object. The latter means that therre exists an arrow N −→ x.
3.1.2. Complexes. A sequence of arrows

fn+1 fn fn−1 fn−2

. . . −−−→ Mn+1 −−−→ Mn −−−→ Mn−1 −−−→ . . . (1)

is called a complex if each its arrow has a kernel and the next arrow factors uniquely
through this kernel.
3.1.3. Lemma. Let each arrow in the sequence

f3 f2 f1 f0
. . . −−−→ M3 −−−→ M2 −−−→ M1 −−−→ M0 (2)

of arrows have a kernel and the composition of any two consecutive arrows is trivial. Then

f4 f3 f2
. . . −−−→ M4 −−−→ M3 −−−→ M2 (3)

51
is a complex. If M0 is a pointed object, then (2) is a complex.
k(fi )
Proof. The objects Mi are pointed for i ≥ 2, which implies that (Ker(fi ) −→ Mi+1
are monomorphisms for all i ≥ 2, hence) (3) is a complex (see 3.1.1).
3.1.4. Corollary. A sequence of morphisms
fn+1 fn fn−1 fn−2
. . . −−−→ Mn+1 −−−→ Mn −−−→ Mn−1 −−−→ . . .
unbounded on the right is a complex iff the composition of any pair of its consecutive arrows
is trivial and for every i, there exists a kernel of the morphism fi .
3.1.5. Example. Let CX be the category Algk of unital associative k-algebras. The
algebra k is its initial object, and every morphism of k-algebras has a kernel. Pointed
objects of CX which have a morphism to initial object are precisely augmented k-algebras.
If the composition of pairs of consecutive arrows in the sequence
f3 f2 f1 f0
. . . −−−→ A3 −−−→ A2 −−−→ A1 −−−→ A0
is trivial, then it follows from the argument of 3.1.2 that Ai is an augmented k-algebra for
all i ≥ 2. And any unbounded on the right sequence of algebras with trivial compositions
of pairs of consecutive arrows is formed by augmented algebras.
3.1.6. ’Exact’ complexes. Let (CX , EX ) be a right exact category with an initial
f g
object. We call a sequence of two arrows L −→ M −→ N in CX ’exact’ if the arrow g
k(g) fg
has a kernel, and f is the composition of Ker(g) −→ M and a deflation L −→ Ker(g). A
complex is called ’exact’ if any pair of its consecutive arrows forms an ’exact’ sequence.
3.2. ∂ ∗ -functors. Fix a right exact category (CX , EX ) with an initial object x and
a category CY with an initial object. A ∂ ∗ -functor from (CX , EX ) to CY is a system of
i T
functors CX −→ CY , i ≥ 0, together with a functorial assignment to every conflation
j e di (E)
E = (N −→ M −→ L) and every i ≥ 0 a morphism Ti+1 (L) −−−→ Ti (N ) which depends
functorially on the conflation E and such that the sequence of arrows
T2 (e) d1 (E) T1 (j) T1 (e) d0 (E) T0 (j)
. . . −−−→ T2 (L) −−−→ T1 (N ) −−−→ T1 (M ) −−−→ T1 (L) −−−→ T0 (N ) −−−→ T0 (M )
idTi (x)
is a complex. Taking the trivial conflation x −→ x −→ x, we obtain that Ti (x) −−−→ Ti (x)
is a trivial morphism, or, equivalently, Ti (x) is an initial object, for every i ≥ 1.
Let T = (Ti , di | i ≥ 0) and T 0 = (Ti0 , d0i | i ≥ 0) be a pair of ∂ ∗ -functors from (CX , EX )
fi
to CY . A morphism from T to T 0 is a family f = (Ti −→ Ti0 | i ≥ 0) of functor morphisms
j e
such that for any conflation E = (N −→ M −→ L) of the exact category CX and every
i ≥ 0, the diagram
di (E)
Ti+1(L) −−−→ Ti (N
 )
fi+1 (L) y y fi (N )
 
d0i (E)
0
Ti+1 (L) −−−→ Ti0 (N )

52
commutes. The composition of morphisms is naturally defined. Thus, we have the category
Hom∗ ((CX , EX ), CY ) of ∂ ∗ -functors from (CX , EX ) to CY .
3.2.1. Trivial ∂ ∗ -functors. We call a ∂ ∗ -functor T = (Ti , di | i ≥ 0) trivial if all
Ti are functors with values in initial objects. One can see that trivial ∂ ∗ -functors are
precisely initial objects of the category Hom∗ ((CX , EX ), CY ). Once an initial object y of
the category CY is fixed, we have a canonical trivial functor whose components equal to
the constant functor with value in y – it maps all arrows of CX to idy .
3.2.2. Some natural functorialities. Let (CX , EX ) be a right exact category with
an initial object and CY a category with initial object. If CZ is another category with an
F
initial object and CY −→ CZ a functor which maps initial objects to initial objects, then
for any ∂ ∗ -functor T = (Ti , di | i ≥ 0), the composition F ◦ T = (F ◦ Ti , F di | i ≥ 0) of T
with F is a ∂ ∗ -functor. The map (F, T ) 7−→ F ◦ T is functorial in both variables; i.e. it
extends to a functor

Cat∗ (CY , CZ ) × Hom∗ ((CX , EX ), CY ) −−−→ Hom∗ ((CX , EX ), CZ ). (1)

Here Cat∗ denotes the subcategory of Cat whose objects are categories with initial objects
and morphisms are functors which map initial objects to initial objects.
On the other hand, let (CX , EX ) be another right exact category with an initial object
and Φ a functor CX −→ CX which maps conflations to conflations. In particular, it maps
M
id
initial objects to initial objects (because if x is an initial object of CX , then x −→ M −→ M
id M
is a conflation; and Φ(x −→ M −→ M ) being a conflation implies that Φ(x) is an initial
∗
object). For any ∂ -functor T = (Ti , di | i ≥ 0) from (CX , EX ) to CY , the composition
T ◦ Φ = (Ti ◦ Φ, di Φ| i ≥ 0) is a ∂ ∗ -functor from (CX , EX ) to CY . The map (T, Φ) 7−→ T ◦ Φ
extends to a functor

Hom∗ ((CX , EX ), CY ) × Ex∗ ((CX , EX ), (CX , EX )) −−−→ Hom∗ ((CX , EX ), CY ), (2)

where Ex∗ ((CX , EX ), (CX , EX )) denotes the full subcategory of Hom(CX , CX ) whose ob-
jects are preserving conflations functors CX −→ CX .
3.3. Universal ∂ ∗ -functors. Fix a right exact category (CX , EX ) with an initial
object x and a category CY with an initial object y.
A ∂ ∗ -functor T = (Ti , di | i ≥ 0) from (CX , EX ) to CY is called universal if for every ∂ ∗ -
g
functor T 0 = (Ti0 , d0i | i ≥ 0) from (CX , EX ) to CY and every functor morphism T00 −→ T0 ,
fi
there exists a unique morphism f = (Ti0 −→ Ti | i ≥ 0) from T 0 to T such that f0 = g.
3.3.1. Interpretation. Consider the functor

Ψ∗
Hom∗ ((CX , EX ), CY ) −−−→ Hom(CX , CY ) (3)

which assigns to every ∂ ∗ -functor (resp. every morphism of ∂ ∗ -functors) its zero compo-
F
nent. For any functor CX −→ CY , we have a presheaf of sets Hom(Ψ∗ (−), F ) on the

53
category Hom∗ ((CX , EX ), CY ). Suppose that this presheaf is representable by an object
(i.e. a ∂ ∗ -functor) Ψ∗ (F ). Then Ψ∗ (F ) is a universal ∂ ∗ -functor.
Conversely, if T = (Ti , di | i ≥ 0) is a universal ∂ ∗ -functor, then T ' Ψ∗ (T0 ).
3.3.2. Proposition. Let (CX , EX ) be a right exact category with an initial object x;
and let CY be a category with initial objects, kernels of morphisms, and limits of filtered
F
systems. Then, for any functor CX −→ CY , there exists a unique up to isomorphism
universal ∂ ∗ -functor T = (Ti , di | i ≥ 0) such that T0 = F .
In other words, the functor
Ψ∗
Hom∗ ((CX , EX ), CY ) −−−→ Hom(CX , CY ) (3)

which assigns to each morphism of ∂ ∗ -functors its zero component has a right adjoint, Ψ∗ .
F
Proof. For an arbitrary functor CX −→ CY , we set

S− (F )(L) = lim Ker(F (k(e))),

e
where the limit is taken by the (filtered) system of all deflations M −→ L. Since deflations
S− (F )
form a pretopology, the map L 7−→ S− (F )(L) extends naturally to a functor CX −−−→ CY .
j e
By the definition of S− (F ), for any conflation E = (N −→ M −→ L), there exists a unique
ed0F (E)
morphism S− (F )(L) −−−→ Ker(F (j)). We denote by d0F (E) the composition of e d0F (E) and
the canonical morphism Ker(F (j)) −→ F (N ).
Notice that the correspondence F 7−→ (S− (F ), d0F ) is functorial. Applying the iter-
ations of the functor S− to F , we obtain a ∂ ∗ -functor S− •
(F ) = (S−i
(F )|i ≥ 0). This
∗
∂ -functor is universal.
3.3.3. Remark. Let the assumptions of 3.3.2 hold. Then we have a pair of adjoint
functors
Ψ∗ Ψ∗
Hom∗ ((CX , EX ), CY ) −−−→ Hom(CX , CY ) −−−→ Hom∗ ((CX , EX ), CY )

By 3.3.2, the adjunction morphism Ψ∗ Ψ∗ −→ Id is an isomorphism which means that Ψ∗

is a fully faithful functor and Ψ∗ is a localization functor at a left multiplicative system.
3.3.4. Proposition. Let (CX , EX ) be a right exact category with an initial object and
T = (Ti , di | i ≥ 0) a ∂ ∗ -functor from (CX , EX ) to CY . Let CZ be another category with
an initial object and F a functor from CY to CZ which preserves initial objects, kernels of
morphisms and limits of filtered systems. Then
(a) If T is a universal ∂ ∗ -functor, then F ◦ T = (F ◦ Ti , F di | i ≥ 0) is universal.
(b) If, in addition, the functor F is fully faithful, then the ∂ ∗ -functor F ◦T is universal
iff T is universal.
Proof. (a) Since the functor F preserves kernels of morphisms and filtered limits (that
is all types of limits which appear in the construction of S− (G)(L)), the natural morphism

F ◦ S− (G)(L) −→ S− (F ◦ G)(L)

54
 G
is an isomorphism for any functor CX −→ CY such that S− (G)(L) = lim Ker(G(k(e)))
◦G
exists. Moreover, dF0 is naturally isomorphic to F dG
0 . Here naturally isomorphic means
j e
that for any conflation E = (N −→ M −→ L), there is a commutative diagram

F dG
0 (E)
F ◦ S−(G)(L) −−−→ F ◦G(N )
oy y id
 

dF
0
G
(E)
S− (F ◦ G)(L) −−−→ F ◦ G(N )

ϕi
i i
commutes. Therefore, the natural morphisms F ◦ S− (T0 ) −−−→ S− (F ◦ T0 ) are are
∗
isomorphisms for all i ≥ 0 and ϕ = (ϕi | i ≥ 0) is an isomorphism of ∂ -functors

∼
◦T0
i
(F ◦ S− (T0 ), F dTi 0 | i ≥ 0) −−−→ (S−
i
(F ◦ T0 ), dF
i | i ≥ 0).

∼
(b) By (a), we have a functor isomorphism F ◦ Ti+1 −→ F ◦ S− (Ti ) for all i ≥ 0. Since
the functor F is fully faithful, this isomorphism is the image of a uniquely determined
∼
isomorphism Ti+1 −→ S− (Ti ). The assertion follows now from (the argument of) 3.3.2.
Details are left as an exercise.

3.3.5. An application. Let (CX , EX ) be a right exact category and CY a category.

We assume that both categories, CX and CY have initial objects. Consider the Yoneda
embedding
hY
CY −−−→ CY∧ , M 7−→ M
c = CY (−, M ).

of the category CY into the category CY∧ of presheaves of sets on CY . The functor hY is
fully faithful and preserves all limits. In particular, it satisfies the conditions of 3.3.4(b).
Therefore, a ∂ ∗ -functor T = (Ti , di | i ≥ 0) from (CX , EX ) to CY is universal iff the
def
∂ ∗ -functor Tb = hY ◦ T = (Tbi , b
di | i ≥ 0) from (CX , EX ) to CY∧ is universal.
Since the category CY∧ has all limits (and colimits), it follows from 3.3.2 that, for
G
any functor CX −→ CY∧ , there exists a unique up to isomorphism universal ∂ ∗ -functor
T = (Ti , di | i ≥ 0) = Ψ∗ (G) whose zero component coincides with G. In particular, for
F
every functor CX −→ CY , there exists a unique up to isomorphism universal ∂ ∗ -functor
T = (Ti , di | i ≥ 0) from (CX , EX ) such that T0 = hY ◦ F = Fe. It follows from 3.3.4(b)
that a universal ∂ ∗ -functor whose zero component coincides with F exists if and only if
for all L ∈ ObCX and all i ≥ 1, the presheaves of sets Ti (L) are representable.

3.3.6. Remark. Let (CX , EX ) be a svelte right exact category with an initial object
x and CY a category with an initial object y and limits. Then, by the argument of 3.3.2,
we have an endofunctor S− of the category Hom(CX , CY ) of functors from CX to CY ,
λ
together with a cone S− −→ y, where y is the constant functor with the values in the
j e
initial object y of the category CY . For any conflation E = (N −→ M −→ L) of (CX , EX )

55
 F
and any functor CX −→ CY , we have a commutative diagram
λ(L)
S− F(L) −−−→ y

d0 (E) y
 
y
Fj Fe
F (N ) −−−→ F (M ) −−−→ F (L)

3.4. The dual picture: ∂-functors and universal ∂-functors. Let (CX , IX ) be
op op
a left exact category, which means by definition that (CX , IX ) is a right exact category.
∗
A ∂-functor on (CX , IX ) is the data which becomes a ∂ -functor in the dual right exact
category. A ∂-functor on (CX , IX ) is universal if its dualization is a universal ∂ ∗ -functor.
We leave to the reader the reformulation in the context of ∂-functors of all notions and
facts about ∂ ∗ -functors.
3.5. Universal ∂ ∗ -functors and ’exactness’.
3.5.1. The properties (CE5) and (CE5∗ ). Let (CX , EX ) be a right exact category.
We say that it satisfies (CE5∗ ) (resp. (CE5)) if the limit of a filtered system (resp. the
colimit of a cofiltered system) of conflations in (CY , EY ) exists and is a conflation.
In particular, if (CX , EX ) satisfies (CE5∗ ) (resp. (CE5)), then the limit of any filtered
system (resp. the colimit of any cofiltered system) of deflations is a deflation.
The properties (CE5) and (CE5∗ ) make sense for left exact categories as well. Notice
that a right exact category satisfies (CE5∗ ) (resp. (CE5)) iff the dual left exact category
satisfies (CE5) (resp. (CE5∗ )).
3.5.2. Note. If (CX , EX ) is an abelian category with the canonical exact structure,
then the property (CE5) for (CX , EX ) is equivalent to the Grothendieck’s property (AB5)
and, therefore, the property (CE5∗ ) is equivalent to (AB5∗ ) (see [Gr, 1.5]).
The property (CE5) holds for Grothendieck toposes.
In what follows, we use (CE5∗ ) for right exact categories and the dual property (CE5)
for left exact categories.
3.5.3. Proposition. Let (CX , EX ), (CY , EY ) be right exact categories, and (CY , EY )
satisfy (CE5∗ ). Let F be a weakly right ’exact’ functor (CX , EX ) −→ (CY , EY ) such that
j e
S− (F ) exists. Then for any conflation E = (N −→ M −→ L) in (CX , EX ), the sequence
S− (F )(j) S− (F )(e) d0 (E) F (j)
S− (F )(N ) −−−→ S− (F )(M ) −−−→ S− (F )(L) −−−→ F (N ) −−−→ F (M ) (1)

is ’exact’. The functor S− (F ) is a weakly right ’exact’ functor from (CX , EX ) to (CY , EY ).
3.5.4. ’Exact’ ∂ ∗ -functors and universal ∂ ∗ -functors. Fix right exact categories
(CX , EX ) and (CY , EY ), both with initial objects. A ∂ ∗ -functor T = (Ti , di | i ≥ 0) from
j e
(CX , EX ) to CY is called ’exact’ if for every conflation E = (N −→ M −→ L) in (CX , EX ),
the complex
T2 (e) d1 (E) T1 (j) T1 (e) d0 (E) T0 (j)
. . . −−−→ T2 (L) −−−→ T1 (N ) −−−→ Ti (M ) −−−→ T1 (L) −−−→ T0 (N ) −−−→ T0 (M )

56
is ’exact’.
3.5.4.1. Proposition. Let (CX , EX ), (CY , EY ) be right exact categories. Suppose
that (CY , EY ) satisfies (CE5∗ ). Let T = (Ti | i ≥ 0) be a universal ∂ ∗ -functor from
(CX , EX ) to (CY , EY ). If the functor T0 is right ’exact’, then the universal ∂ ∗ -functor T
is ’exact’.
Proof. If T0 is right ’exact’, then, by 3.5.3, the functor T1 ' S− (T0 ) is right ’exact’
j e
and for any conflation E = (N −→ M −→ L), the sequence

T1 (j) T1 (e) d0 (E) T0 (j)

T1 (N ) −−−→ T1 (M ) −−−→ T1 (L) −−−→ T0 (N ) −−−→ T0 (M )

is ’exact’. Since Tn+1 = S− (Tn ), the assertion follows from 3.5.3 by induction.
3.5.4.2. Corollary. Let (CX , EX ) be a right exact category. For each object L of
CX , the ∂-functor Ext•X (L) = (ExtiX (L) | i ≥ 0) is ’exact’.
Suppose that the category CX is k-linear. Then for each L ∈ ObCX , the ∂-functor
Ext•X (L) = (ExtiX (L) | i ≥ 0) is ’exact’.
Proof. In fact, the ∂-functor Ext•X (L) is universal by definition (see 3.4.1), and the
functor Ext0X (L) = CX (−, L) is left exact. In particular, it is left ’exact’.
If CX is a k-linear category, then the universal ∂-functors Ext•X (L), L ∈ ObCX , with
the values in the category of k-modules (defined in 3.4.2) are ’exact’ by a similar reason.

4. Coeffaceable functors, universal ∂ ∗ -functors, and pointed projectives.

4.1. Projectives and projective deflations. Fix a right exact category (CX , EX ).
We call an object P of CX projective if every deflation M −→ P splits. Equivalently, any
f e
morphism P −→ N factors through any deflation M −→ N .
We denote by PEX the full subcategory of CX generated by projective objects.
4.1.1. Example. Let (CX , EX ) be a right exact category whose deflations split.
Then every object of CX is a projective object of (CX , EX ).
A deflation M −→ L is called projective if it factors through any deflation of L.
Any deflation P −→ L with P projective is a projective deflation. On the other hand,
an object P is projective iff the identical morphism P −→ P is a projective deflation.
4.2. Coeffaceble functors and projectives. Let (CX , EX ) be a right exact cate-
F
gory and CY a category with an initial object. We call a functor CX −→ CY coeffaceable,
t
or EX -coeffaceable, if for any object L of CX , there exists a deflation M −→ L such that
F (t) is a trivial morphism.
F
It follows that if a functor CX −→ CY is EX -coeffaceable, then it maps all projectives
to initial objects and all projective deflations to trivial arrows.
So that if the right exact category (CX , EX ) has enough projective deflations (resp.
F
enough projectives), then a functor CX −→ CY is EX -coeffaceable iff F (e) is trivial for
any projective deflation e (resp. F (M ) is an initial object for every projective object M ).

57
 4.3. Proposition. Let (CX , EX ) be a right exact category with initial objects and
T = (Ti , di | i ≥ 0) a universal ∂ ∗ -functor from (CX , EX ) to CY . Then Ti (P ) is an initial
object for any pointed projective object P and for all i ≥ 1.
4.3.1. Corollary. Let (CX , EX ) be a right exact category with initial objects and
T = (Ti , di | i ≥ 0) a universal ∂ ∗ -functor from (CX , EX ) to CY . Suppose that (CX , EX )
has enough projectives and projectives of (CX , EX ) are pointed objects. Then the functors
Ti are coeffaceable for all i ≥ 1.
Proof. The assertion follows from 4.3 and 4.2.
4.4. Proposition. Let (CX , EX ) and (CY , EY ) be right exact categories with initial
objects; and let T = (Ti , di | i ≥ 0) be an ’exact’ ∂ ∗ -functor from (CX , EX ) to (CY , EY ).
If the functors Ti are EX -coeffaceable for i ≥ 1, then T is a universal ∂ ∗ -functor.
Proof. The argument is similar to the proof in [Gr] of the corresponding assertion for
abelian categories.
4.5. Proposition. Let (CX , EX ), (CY , EY ), and (CZ , EZ ) be right exact categories.
Suppose that (CX , EX ) has enough projectives and CY has kernels of all morphisms. If
T = (Ti | i ≥ 0) is a universal, ’exact’ ∂ ∗ -functor from (CX , EX ) to (CY , EY ) and F a
functor from (CY , EY ) to (CZ , EZ ) which respects conflations, then the composition F ◦T =
(F ◦ Ti | i ≥ 0) is a universal ’exact’ ∂ ∗ -functor.
Proof. Under the conditions of the proposition, the composition F ◦ T is an ’exact’
functor such that the functors F ◦ Ti , i ≥ 1, map pointed projectives of (CX , EX ) to trivial
objects (because Ti map pointed projectives to trivial objects by 4.3 and F maps trivial
objects to trivial objects). Since there are enough pointed projectives (hence all projectives
are pointed), this implies that the functors F ◦ Ti are coeffaceable for i ≥ 1. Therefore, by
4.4, F ◦ T is a universal ∂ ∗ -functor.
4.6. Sufficient conditions for having enough pointed projectives.
4.6.1. Proposition. Let (CX , EX ) and (CZ , EZ ) be right exact categories and
f∗
CZ −→ CX a functor having a right adjoint f∗ . Suppose that f ∗ maps deflations of
(M )
the form N −→ f∗ (M ) to deflations and the adjunction arrow f ∗ f∗ (M ) −−−→ M is a
deflation for all M (which is the case if any morphism t of CX such that f∗ (t) is a split
epimorphism belongs to EX ). Let (CZ , EZ ) have enough projectives, and all projectives are
pointed objects. Then each projective of (CX , EX ) is a pointed object.
If, in addition, f∗ maps deflations to deflations, then (CX , EX ) has enough projectives.
4.6.2. Note. The conditions of 4.6.1 can be replaced by the requirement that if
N −→ f∗ (M ) is a deflation, then the corresponding morphism f ∗ (N ) −→ M is a deflation.
This requirement follows from the conditions of 4.6.1, because the morphism f ∗ (N ) −→ M
t
corresponding to N −→ f∗ (M ) is the composition of f ∗ (t) and the adjunction arrow
(M )
f ∗ f∗ (M ) −−−→ M .
4.6.3. Example. Let (CX , EX ) be the category Algk of associative k-algebras en-
dowed with the canonical (that is the finest) right exact structure. This means that class

58
EX of deflations coincides with the class of all are strict epimorphisms of k-algebras. Let
(CY , EY ) be the category of k-modules with the canonical exact structure, and f∗ the
forgetful functor Algk −→ k − mod. Its left adjoint, f ∗ preserves strict epimorphisms, and
the functor f∗ preserves and reflects deflations; i.e. a k-algebra morphism t is a strict epi-
morphism iff f∗ (t) is an epimorphism. In particular, the adjunction arrow f ∗ f∗ (A) −→ A
is a strict epimorphism for all A. By 4.6.1, (CX , EX ) has enough projectives and each
projective has a morphism to the initial object k; that is each projective has a structure
of an augmented k-algebra.
4.7. Acyclic objects and the universality of ∂ ∗ -functors. Given a ∂ ∗ -functor
T = (Ti | i ≥ 0) from a right exact category (CX , EX ) to a category CY , we call an object
M of CX T -acyclic if Ti (M ) is a trivial object for all i ≥ 1.
4.7.1. Proposition. Let (CA , EA ) and (CX , EX ) be right exact categories with initial
G
objects and CA −→ CX a functor which preserves conflations. Let T = (Ti | i ≥ 0) be an
’exact’ ∂ ∗ -functor from (CX , EX ) to a category CZ with initial objects. If there are enough
objects M of CX such that G(M ) is a T -acyclic object, then T ◦G is a universal ∂ ∗ -functor.
Proof. Since the functor G maps conflation to conflations, and the ∂ ∗ -functor T is
’exact’, its composition T ◦ G is an ’exact’ ∂ ∗ -functor. Since there are enough objects in
CA which the functor G maps to acyclic objects (i.e. for each object L of CA , there is a
deflation M −→ L such that G(M ) is T -acyclic), the functor Ti ◦ G is effaceable for all
i ≥ 1. Therefore, by 4.6, the composition T ◦ G is a universal ∂ ∗ -functor.

5. Universal problems for universal ∂ ∗ - and ∂-functors.

5.1. The categories of universal ∂ ∗ - and ∂-functors. Fix a right exact svelte
category (CX , EX ) with an initial object. Let ∂ ∗ Un(X, EX ) denote the category whose
objects are universal ∂ ∗ -functors from (CX , EX ) to categories CY (with initial objects).
Let T be a universal ∂ ∗ - functor from (CX , EX ) to CY and Te a universal ∂ ∗ - functor from
(CX , EX ) to CZ . A morphism from T to T 0 is a pair (F, φ), where F is a functor from CY to
CZ and φ is a ∂ ∗ -functor isomorphism F ◦T −→ ∼
T 0 . If (F 0 , φ0 ) is a morphism from T 0 to T 00 ,
then the composition of (F, φ) and (F 0 , φ0 ) is defined by (F 0 , φ0 ) ◦ (F, φ) = (F 0 ◦ F, φ0 ◦ F 0 φ).
Dually, for a left exact category (CX , IX ) with a final object, we denote by ∂Un(X, IX )
the category whose objects are universal ∂-functors from (CX , IX ) to categories with final
object. Given two universal ∂-functors T and T 0 from (CX , IX ) to respectively CY and
CZ , a morphism from T to T 0 is a pair (F, ψ), where F is a functor from CY to CZ and ψ
is a functor isomorphism T 0 −→ ∼
F ◦ T . The composition is defined by (F 0 , ψ 0 ) ◦ (F, ψ) =
(F 0 ◦ F, F 0 ψ ◦ ψ 0 ).
5.2. Universal problems for universal ∂-functors with values in complete
categories and ∂ ∗ -functors with values in cocomplete categories.
Let (CX , EX ) be a svelte right exact category. We denote by ∂ ∗ Unc (X, EX ) the subcat-
egory of ∂ ∗ Un(X, EX ) whose objects are universal ∂ ∗ -functors from (CX , EX ) to complete
(i.e. having limits of small diagrams) categories CY and morphisms between these universal
∂ ∗ -functors are pairs (F, φ), where the functor F preserves limits.

59
 For a svelte left exact category (CX , IX ), we denote by ∂Unc (X, IX ) the subcategory
of ∂Un(X, IX ) whose objects are ∂-functors with values in cocomplete categories and mor-
phisms are pairs (F, ψ) such that the functor F preserves small colimits.
5.2.1. Proposition. Let (CX , EX ) be a svelte right exact category with initial objects
and (CX , IX ) a svelte left exact category with final objects. The categories ∂ ∗ Unc (X, EX )
and ∂Unc (X, IX ) have initial objects.
Proof. It suffices to prove the assertion about ∂Unc (X, IX ), because the assertion
about ∂ ∗ -functors is obtained via dualization.
Consider the Yoneda embedding

hX
∧
CX −−−→ CX , M 7−→ CX (−, M ).

We denote by Ext•X,IX the universal ∂-functor from (CX , IX ) to CX ∧

such that Ext0X,IX =
•
hX . The claim is that ExtX,IX is an initial object of the category ∂Unc (X, IX ).
In fact, let CY be a cocomplete category. By [GZ, II.1.3], the composition with the
h
X ∧ ∧
Yoneda embedding CX −→ CX is an equivalence between the category Homc (CX , CY ) of
continuous (that is having a right adjoint, or, equivalently, preserving colimits) functors
∧ F
CX −→ CY and the category Hom(CX , CY ) of functors from CX to CY . Let CX −→ CY be
∧ c F
an arbitrary functor and CX −→ CY the corresponding continuous functor. By definition,
j
S+ F (L) = colim(Cok(F (M −→ Cok(j)), where L −→ M runs through inflations of L.
Since Fc preserves colimits, it follows from (the dual version of) 3.3.4(a) that Fc ◦ Ext•X,IX
is a universal ∂-functor whose zero component is Fc ◦ Ext0X,IX = Fc ◦ hX = F. Therefore,
by (the dual version of the argument of) 3.3.2, the universal ∂-functor Fc ◦ Ext•X,IX is
•
isomorphic to the right satellite S+ F of the functor F . This shows that Ext•X,IX is an
initial object of the category ∂Unc (X, IX ).
5.3. The universal problem for arbitrary universal ∂- and ∂ ∗ -functors. Let
(CX , IX ) be a svelte left exact category with final objects. Let CXs denote the smallest
∧
strictly full subcategory of the category CX containing all presheaves ExtnX,IX (L), L ∈
ObCX , n ≥ 0. Let Ext•X,IX denote the corestriction of the ∂-functor Ext•X,IX to the
subcategory CXs . Thus, Ext•X,IX is the composition of the ∂-functor Ext•X,IX and the
JX
∧
inclusion functor CXs −−−→ CX . It follows that Ext•X,IX is a universal ∂-functor.
5.3.1. Proposition. Let (CX , IX ) a svelte left exact category with final objects.
For any universal ∂-functor T = (Ti , di | i ≥ 0) from (CX , IX ) to a category CY (with
Ts
final objects), there exists a unique (up to isomorphism) functor CXs −→ CY such that
T = T s ◦ Ext•X,IX and the diagram

T0∗ op
∧
CxX −−−→ CxY∨
 o
JX   hY

Ts
CXs −−−→ CY

60
 op
commutes. Here CY∨ denote the category of presheaves of sets on CYop (i.e. functors
op
CY −→ Sets) and hoY the (dual) Yoneda functor CY −→ CY∨ , L 7−→ CY (L, −).; and T0∗
is a unique continuous (i.e. having a right adjoint) functor such that T0∗ ◦ hX = hoY ◦ T0 .
op
Proof. The category CY∨ is cocomplete (and complete) and the functor hoY preserves
colimits. Therefore, by 3.3.4, the composition hoY ◦T is a universal ∂-functor from (CX , IX )
op
to CY∨ . By 5.2.1, the ∂-functor hoY ◦ T is the composition of the universal ∂-functor
T∗ op
Ext•X,IX from (CX , IX ) to CX ∧
and the unique continuous functor CX ∧
−→0
CY∨ such that
∗ o o
T0 ◦ hX = hY ◦ T0 . Since the functor hY is fully faithful, this implies that the universal ∂-
functor T = (Ti , di | i ≥ 0) is isomorphic to the composition of the corestriction of Ext•X,IX
Ts
to the subcategory CXs and a unique functor CXs −−−→ CY such that the composition
hoY ◦ T s coincides with the restriction of the functor T0∗ to the subcategory CXs .
5.3.2. Note. The formulation of the dual assertion about the universal ∂ ∗ -functors
is left to the reader.
5.4. The k-linear version. Fix a right exact svelte k-linear additive category
(CX , EX ). Let ∂k∗ Un(X, EX ) denote the category whose objects are universal k-linear ∂ ∗ -
functors from (CX , EX ) to k-linear additive categories CY . Let T be a universal k-linear
∂ ∗ -functor from (CX , EX ) to CY and Te a universal k-linear ∂ ∗ - functor from (CX , EX ) to
CZ . A morphism from T to T 0 is a pair (F, φ), where F is a k-linear functor from CY to CZ
and φ is a ∂ ∗ -functor isomorphism F ◦ T −→ ∼
T 0 . If (F 0 , φ0 ) is a morphism from T 0 to T 00 ,
then the composition of (F, φ) and (F 0 , φ0 ) is defined by (F 0 , φ0 ) ◦ (F, φ) = (F 0 ◦ F, φ0 ◦ F 0 φ).
We denote by ∂k∗ Unc (X, EX ) the subcategory of ∂k∗ Un(X, EX ) whose objects are k-
linear ∂ ∗ -functors with values in complete categories and morphisms are pairs (F, φ) such
that the functor F preserves small limits.
Dually, for a left exact svelte k-linear additive category (CX , IX ), we denote by
∂k Un(X, IX ) the category whose objects are universal k-linear ∂-functors from (CX , IX )
to additive k-linear categories. Given two universal ∂-functors T and T 0 from (CX , IX ) to
respectively CY and CZ , a morphism from T to T 0 is a pair (F, ψ), where F is a k-linear
functor from CY to CZ and ψ a functor isomorphism T 0 −→ ∼
F ◦ T . The composition is
0 0 0 0 0
defined by (F , ψ ) ◦ (F, ψ) = (F ◦ F, F ψ ◦ ψ ).
We denote by ∂k Unc (X, IX ) the subcategory of ∂k Un(X, IX ) whose objects are k-linear
∂-functors with values in cocomplete categories and morphisms are pairs (F, ψ) such that
the functor F preserves small colimits.
5.4.1. Proposition. Let (CX , EX ) (resp. (CX , IX )) be a svelte right (resp. left) exact
additive k-linear category. The categories ∂k∗ Unc (X, EX ) and ∂k Unc (X, IX ) have initial
objects.
Proof. By duality, it suffices to prove that the category ∂k Unc (X, IX ) has an initial
∧
object. The argument is similar to the argument of 5.2.1, except for the category CX of
presheaves of sets on the category CX is replaced by the category Mk (X) of presheaves of
k-modules on CX . The initial object of the category ∂k Unc (X, IX ) is the universal k-linear
∂-functor Ext•(X,IX ) from (CX , IX ) to the category Mk (X) whose zero component is the
(k-linear) Yoneda embedding CX −→ Mk (X), L 7−→ CX (−, L).

61
 Let (CX , IX ) be a svelte k-linear additive left exact category. Let Msk (X denote the
smallest additive strictly full subcategory of the category Mk (X) containing all presheaves
Extn(X,IX ) (L), L ∈ ObCX , n ≥ 0. Let Ext•(X,IX ) denote the corestriction of the ∂-functor
Ext•(X,IX ) to the subcategory Msk (X. Thus, Ext•(X,IX ) is the composition of the k-linear
∂-functor Ext•X,IX and the inclusion functor

JX
Msk (X) −−−→ Mk (X).

It follows that Ext•X,IX is a universal ∂-functor.

5.4.2. Proposition. Let (CX , IX ) be a svelte left exact category with final objects.
For any universal ∂-functor T = (Ti , di | i ≥ 0) from (CX , IX ) to a category CY (with
Ts
final objects), there exists a unique (up to isomorphism) functor CXs −→ CY such that
T = T s ◦ Ext•X,IX and the diagram

T0∗ op
∧
CxX −−−→ CxY∨
 o
JX   hY

Ts
CXs −−−→ CY
op
commutes. Here CY∨ denote the category of presheaves of sets on CYop (i.e. functors
op
CY −→ Sets) and hoY the (dual) Yoneda functor CY −→ CY∨ , L 7−→ CY (L, −).; and T0∗
is a unique continuous (i.e. having a right adjoint) functor such that T0∗ ◦ hX = hoY ◦ T0 .
Proof. The argument is similar to that of 5.3.1.

6. The stable category of a left exact category.

6.1. Reformulations. Fix a svelte left exact category (CX , IX ). Let Θ b ∗ denote
X
∧ ∧
the continuous (that is having a right adjoint) functor CX −→ CX determined (uniquely
j
b ∗ ◦ hX . To any conflation N −→ e
up to isomorphism) by the equality Ext1X = Θ X M −→ L,
corresponds the diagram

bj be
N
b −−−→ M
 −−−→
c L
b

y d0 (E)
 
y (1)
λ(N
b)
x −−−→ Θb ∗ (N
b)
b X

∧
where L
b = hX (L) and xb is the final object of the category CX of presheaves on CX .
•
Due to the universality of ExtX , all the information about universal ∂-functors from
j e
the left exact category (CX , IX ), is encoded in the diagrams (1), where N −→ M −→ L
runs through the class of conflations of (CX , IX ).
In fact, it follows from the (argument of) 3.3.4(a) that the universal ∂-functor Ext•X is
isomorphic to the ∂-functor of the form (Θ b ∗n ◦ hX , Θ
b ∗n (d0 ) | n ≥ 0); and for any functor
X X

62
F from CX to a category CY with colimits and final objects, the universal ∂-functor
(Ti , di | i ≥ 0) from (CX , IX ) to CY with T0 = F is isomorphic to

F ∗ ◦ Ext•X = (F ∗ Θ
b ∗n ∗ b ∗n
X ◦ hX , F ΘX (d0 ) | n ≥ 0). (2)

6.2. Note. If CX is a pointed category, then the presheaf x b = CX (−, x) is both a

λ(N
b)
∧ b ∗ (N
final and an initial object of the category CX . In particular, the morphism x b −−−→ Θ X
b)
in (1) is uniquely defined, hence is not a part of the data. In this case, the data consists
of the diagrams
bj be d0 (E)
Nb −−−→ M c −−−→ L b −−−→ Θ b ∗X (N
b ),
j e
where E = (N −→ M −→ L) runs through conflations of (CX , IX ).
6.3. Stable category of (CX , IX ). Consider the full subcategory CXs of the cat-
∧ b ∗n (M), where M runs through representable presheaves
egory CX whose objects are Θ X
and n through nonnegative integers. We denote by θXs the endofunctor CXs −→ CXs
induced by Θb ∗ . It follows that CX is the smallest Θ
b ∗ -stable strictly full subcategory of
X s X
∧ c = CX (−, M ), M ∈ ObCX .
the category CX containing all presheaves M
6.3.1. Triangles. We call the diagram

bj be d0 (E)
b −−−→ M
N c −−−→ L
b −−−→ Θ b ∗X (N
b ), (1)

j e
where E = (N −→ M −→ L) is a conflation in (CX , IX ), a standard triangle.
A triangle is any diagram in CXs of the form

j e d
N −→ M −→ L −→ θXs (N ), (2)

which is isomorphic to a standard triangle. It follows that for every triangle, the diagram
e
M
 −−−→ L

yd
 
y
λ(N )
x b ∗ (N )
−−−→ Θ
b X

commutes. Triangles form a category TrXs : morphisms from

j e d
N −−−→ M −−−→ L −−−→ θX (N )

to
j0 e0 d0
N 0 −−−→ M0 −−−→ L0 −−−→ θX (N 0 )

63
are given by commutative diagrams
j e d
N
 −−−→ M
 −−−→ 
L −−−→ θX
(N )
f y yg yh y θX (f )
   
j0 e0 d0
N0 −−−→ M0 −−−→ L0 −−−→ θX (N 0 )

The composition is obvious.

6.3.2. The prestable category of a left exact category. Thus, we have obtained
a data (CXs , (θXs , λ), TrXs ). We call this data the prestable category of the left exact
category (CX , IX ).
6.3.3. The stable category of a left exact category with final objects. Let
(CX , IX ) be a left exact category with a final object x and (CXs , θXs , λ; TrXs ) the associ-
ated with (CX , IX ) presuspended category. Let Σ = ΣθXs be the class of all arrows t of
CXs such that θXs (t) is an isomorphism.
We call the quotient category CXs = Σ−1 CXs the stable category of the left exact
category (CX , IX ). The endofunctor θXs determines a conservative endofunctor θXs of
q∗
Σ
the stable category CXs . The localization functor CXs −→ CXs maps final objects to final
λ
objects. Let λs denote the image x e = q∗Σ (b
x) −→ θXs of the cone x b −→ θXs .
Finally, we denote by TrXs the strictly full subcategory of the category of diagrams
of the form N −→ M −→ L −→ θXs (N ) generated by q∗Σ (TrXs ).
The data (CXs , θXs , λs ; TrXs ) will be called the stable category of (CX , IX ).
6.4. Dual notions. If (CX , EX ) is a right exact category with an initial object,
one obtains, dualizing the definitions of 6.3.2 and 6.3.3, the notions of the precostable and
costable category of (CX , EX ).
6.5. The k-linear version. Let (CX , JX ) be a k-linear additive svelte left exact
∧
category. Replacing the category of CX of presheaves of sets by the category Mk (X)
of presheaves of k-modules on CX and the functor Ext1(X,IX ) by its k-linear version,
Ext1(X,IX ) , we obtain the k-linear versions of prestable and stable categories of the left
exact category (CX , JX ).
6.5.1. Note. If (CX , JX ) is a k-linear exact category (that is JX happen to be
the class of inflations of a k-linear exact category) with enough injectives, than its stable
category defined above is equivalent to the conventional stable category of (CX , JX ). Recall
that the latter has the same objects as CX and its morphisms are homotopy classes of
f

morphisms of CX : two morphisms M −→ −→ N are homotopy equivalent to each other if

g
their difference f − g factors through an injective object.
Notice that our construction of stable category of (C, IX ) does not require any addi-
tional hypothesis. In particular, it extends the notion of the stable category to arbitrary
exact categories.

64
 7. Complement: presuspended and quasi-suspended categories.
It is tempting to follow Keller’s example [Ke1], [KeV] and turn essential properties
of prestable and stable categories of a left exact category into axioms. We call the corre-
sponding notions respectively presuspended and quasi-suspended categories.
7.1. Presuspended and quasi-suspended categories. Fix a category CX with a
θX
e
final object x and a functor CX −−−→ x\CX , or, what is the same, a pair (θX , λ), where θX
is an endofunctor CX −→ CX and λ is a cone x −→ θX . We denote by Tr fX the category
whose objects are all diagrams of the form

j e d
N −−−→ M −−−→ L −−−→ θX (N )

such that the square

e
M
 −−−→ L

yd
 
y
λ(N )
x −−−→ θX (N )
commutes. Morphisms from

j e d
N −−−→ M −−−→ L −−−→ θX (N )

to
j0 e0 d0
N 0 −−−→ M0 −−−→ L0 −−−→ θX (N 0 )
f g h
are triples (N −→ N 0 , M −→ M0 , L −→ L0 ) such that the diagram
j e d
N
 −−−→ M
 −−−→ 
L −−−→ θX
(N )
f y yg yh y θX (f )
   
j0 e0 d0
N0 −−−→ M0 −−−→ L0 −−−→ θX (N 0 )

commutes. The composition of morphisms is natural.

7.1.1. Presuspended categories. A presuspended category is a triple (CX , θeX , TrX ),
where CX and θeX = (θX , λ) are as above and TrX is a strictly full subcategory of the
category Tr
fX whose objects are called triangles, which satisfies the following conditions:
(PS1) Let CX0 denote the full subcategory of CX generated by objects N such that
j e d
there exists a triangle N −→ M −→ L −→ θX (N ). For every N ∈ ObCX0 , the diagram

idN λ(N )
N −−−→ N −−−→ x −−−→ θX (N )

is a triangle.

65
 j e d f
(PS2) For any triangle N −→ M −→ L −→ θX (N ) and any morphism N −→ N 0
j0 e0 d0
with N 0 ∈ ObCX0 , there is a triangle N 0 −→ M0 −→ L −→ θX (N 0 ) such that f extends
to a morphism of triangles

j e d (f,g,idL ) j0 e0 d0
(N −→ M −→ L −→ θX (N )) −−−−−−−→ (N 0 −→ M0 −→ L −→ θX (N 0 )).

(PS3) For any pair of triangles

j e d j0 e0 d0
N −→ M −→ L −→ θX (N ) and N 0 −→ M0 −→ L0 −→ θX (N 0 )

and any commutative square

j
N
 −−−→ M

f y yg
 
j0
N0 −−−→ M0
h
there exists a morphism L −→ L0 such that (f, g, h) is a morphism of triangles, i.e. the
diagram
j e d
N
 −−−→ M
 −−−→ 
L −−−→ θX
(N )
f y yg yh y θX (f )
   
j0 e0 d0
N0 −−−→ M0 −−−→ L0 −−−→ θX (N 0 )
commutes.
(PS4) For any pair of triangles
u v w x s r
N −→ M −→ L −→ θX (N ) and M −→ M0 −→ M
f −→ θX (M ),

there exists a commutative diagram

u v w
N −−−→ M
 −−−→ L
 −−−→ θX
(N )
id y xy yy y id
   
u0 v0 w0
0 0
N −−−→ M
 −−−→ L
 −−−→ θX
(N )
sy yt y θX (u)
  
(2)
id r
M
f
 −−−→ M
f
 −−−→ θX (M)
 0
ry yr

θX (v)
θX (M) −−−→ θX (L)

whose two upper rows and two central columns are triangles.
j e d
(PS5) For every triangle N −→ M −→ L −→ θX (N ), the sequence

. . . −−−→ CX (θX (N ), −) −−−→ CX (L, −) −−−→ CX (M, −) −−−→ CX (N , −)

66
is exact.

7.1.2. Remarks. (a) If CX is an additive category, then three of the axioms above
coincide with the corresponding Verdier’s axioms of triangulated category (under condition
that CX0 = CX ). Namely, (PS1) coincides with the first half of the axiom (TRI), the axiom
(PS3) coincides with the axiom (TRIII), and (PS4) with (TRIV) (see [Ve2, Ch.II]).
(b) It follows from (PS4) that if N −→ M −→ L −→ θX (N ) is a triangle, then all
three objects, N , M, and L, belong to the subcategory CX0 .
(c) The axiom (PS2) can be regarded as a base change property, and axiom (PS4)
expresses the stability of triangles under composition. So that the axioms (PS1), (PS2)
and (PS4) say that triangles form a ’pretopology’ on the subcategory CX0 . The axiom
(PS5) says that this pretopology is subcanonical: the representable presheaves are sheaves.
These interpretations (as well as the axioms themselves) come from the main examples:
prestable and stable categories of a left exact category.

7.1.3. Quasi-suspended categories. A presuspended category (CX , θX , λ; TrX )

will be called quasi-suspended if the functor θX is conservative. We denote by SCat the
full subcategory of the category PCat of presuspended categories whose objects are con-
servative presuspended svelte categories.

7.2. Examples.

7.2.1. The presuspended category of presheaves of sets on a left exact

b ∗ be a continuous endofunctor of
category. Fix a left exact category (CX , IX ). Let Θ X
∧ b ∗ ◦ hX . We
CX = CX∧ determined uniquely up to isomorphism by the equality Ext1X,I = Θ X
call a standard triangle all diagrams of the form

bj be d0 (E)
b −−−→ M
N c −−−→ L
b −−−→ Θ b ∗ (N
b ), (1)
X

j e
where E = (N −→ M −→ L) is any conflation in (CX , IX ). Triangle is an object
of the category Tr fX∧ which is isomorphic to a standard triangle. We denote by Tr cX
the full subcategory of the category TrX∧ whose objects are triangles. One can see that
f
TCX∧ = (CX∧ , Θ b ∗ , λX ; Tr
cX ) is a presuspended category.
X
∧
In fact, CX0 is the full subcategory of CX generated by all representable functors. The
N id
property (PS1) holds, because N −→ N −→ x is a conflation for any object N of CX .
j e
The property (PS2) holds, because for any conflation E = (N −→ M −→ L) and any
f
morphism N −→ N 0 , we have a commutative diagram

j e
N −−−→ M
 −−−→ L

f y cocart yf y idL
  e 
j0 e0
N0 −−−→ M0 −−−→ L

67
whose rows are conflations and left square is cocartesian. The property (PS3) holds,
because for any commutative diagram
j e
N −−−→ 
M −−−→ L
f y yg
 
j0 e0
N0 −−−→ M 0 −−−→ L0
h
whose rows are conflations, there exists a unique arrow L −→ L0 which makes the diagram
j e
N −−−→ M
 −−−→ L

f y yg yh
  
j0 e0
N0 −−−→ M 0 −−−→ L0

commute, i.e. (f, g, h) is a morphism of conflations. Since Ext•X is a ∂-functor, this implies
the commutativity of the diagram
j e d0 (E)
N −−−→ M −−−→ L b ∗ (N )
−−−→ Θ
   X
 b∗
f y yg yh y ΘX (f )
  

j0 e0 d0 (E 0 )
0 0 0 b ∗ (N 0 )
N −−−→ M −−−→ L −−−→ Θ X

j0 e0
where E 0 denotes the conflation N 0 −→ M 0 −→ L0 .
j e
For any conflation E = (N −→ M −→ L), the sequence

bj be d0 (E) b∗ (j)
Θ
b −−−→ M
N c −−−→ L
b −−−→ Θ b ) −−X−→ . . .
b ∗X (N

is exact, because Ext•X is an ’exact’ ∂-functor. This implies the property (PS5), that is
the exactness of
b ∗ (N ), −) −−−→ CX (L, −) −−−→ CX (M, −) −−−→ CX (N, −).
. . . −−−→ CX (Θ X

7.2.2. The associated quasi-suspended category. It is obtained via localization

b ∗ , λX ; Tr
of the presuspended category TCX∧ = (CX∧ , Θ cX ) (see 7.2.1) at the class of arrows
X
b∗ b∗
ΣΘb∗ = {s ∈ HomCX∧ | ΘX (s) is an isomorphism}. Since ΘX is a continuous functor, the
X
localization q∗X at ΣΘ
b∗X is a continuous (that is having a right adjoint) functor too. In
particular, the functor q∗X preserves colimits of small diagrams. The fact that q∗X is right
exact implies that the category TCXc obtained by applying the localization functor q∗X to
TCX∧ inherits all the properties of TCX∧ , including the exactness of the sequence

bj be d0 (E) b∗ (j)
Θ
q∗X N b ) −−X−→ . . . .
b ∗ (N


b −−−→ M
c −−−→ L
b −−−→ Θ
X

68
By construction, the suspension functor θXc induced by Θ b ∗ on the quotient category CX =
X c

Σ−1∗ CX∧ (it is uniquely determined by the equality θXc ◦ q∗X = q∗X ◦ Θ b ∗ ) is conservative;
X
Θ
bX
i.e. TCXc = (CXc , θXc , λXc ; TrXc ) is a quasi-suspended category.
Notice that the category CXc is cocomplete and complete. This follows from the
∧
corresponding properties of the category CX∧ = CX of presheaves of sets on CX and the
∗
fact that the localization functor qX has a right adjoint [GZ, I.1].
7.2.3. Reduced presuspended categories. Let TCX = (CX , θX , λX ; T rX ) be a
presuspended category and CX0 the full subcategory of CX generated by objects N such
j e d
that there exists a triangle N −→ M −→ L −→ θX (N ). Let CX1 be the smallest θX -stable
strictly full subcategory of CX containing the subcategory CX0 and θX1 the endofunctor
of CX1 induced by θX . One can see that TCX1 = (CX1 , θX1 , λX1 ; T rX ) is a presuspended
category. It is quasi-suspended if TCX is quasi-suspended.
We call TCX1 the reduced presuspended category associated with TCX . In particular,
we call the presuspended category TCX reduced if it coincides with TCX1 .
7.2.4. Prestable and stable categories of a left exact category. The reduced
presuspended category associated with the presuspended category TCX∧ of presheaves of
sets on a left exact category (CX , IX ) (see 7.2.1) coincides with the prestable category of
(CX , IX ) defined in 6.3.2.
The reduced presuspended (actually, quasi-suspended) category associated with the
quasi-suspended category associated with TCX∧ (see 7.2.2) is naturally equivalent to the
stable category of (CX , IX ) introduced in 6.3.3.
7.3. The category of presuspended categories. Let TCX = (CX , θX , λX ; T rX )
and TCY = (CY , θY , λY ; T rY ) be presuspended categories. A triangle functor from TCX
to TCY is a pair (F, φ), where F is a functor CX −→ CY which maps initial object to an
initial object and φ is a functor isomorphism F ◦ θX −→ θY ◦ F such that for every triangle
u v w
N −→ M −→ L −→ θX (N ) of TCX , the sequence

F (u) F (v) φ(N )F (w)

F (N ) −−−→ F (M) −−−→ F (L) −−−→ θY (F (N ))

is a triangle of TCY . The composition of triangle functors is defined naturally:

(G, ψ) ◦ (F, φ) = (G ◦ F, ψF ◦ Gφ).

Let (F, φ) and (F 0 , φ0 ) be triangle functors from T− CX to T− CY . A morphism from

λ
(F, φ) to (F 0 , φ0 ) is given by a functor morphism F −→ F 0 such that the diagram
φ
◦F
θY  −−−→ F 
◦ θX
θY λ y y λθX
 
φ0
θY ◦ F 0 −−−→ F 0 ◦ θX

commutes. The composition is the compsition of the functor morphisms.

69
 Altogether gives the definition of a bicategory PCat formed by svelte presuspended
categories, triangle functors as 1-morphisms and morphisms between them as 2-morphisms.
As usual, the term “category PCat” means that we forget about 2-morphisms.
Dualizing (i.e. inverting all arrows in the constructions above), we obtain the bicat-
egory Po Cat formed by precosuspended svelte categories as objects, triangular functors as
1-morphisms, and morphisms between them as 2-morphisms.
7.3.1. The subcategory of quasi-suspended categories. We denote by SCat
the full subcategory of the category PCat of presuspended svelte categories whose objects
are quasi-suspended categories.
7.3.2. From presuspended categories to quasi-suspended categories. Let
(CX , θX , λ; TrX ) be a presuspended category and Σ = ΣθX the class of all arrows s of the
category CX such that θX (s) is an isomorphism. Let ΘX denote the endofunctor of the
quotient category Σ−1 CX uniquely determined by the equality ΘX ◦ q∗Σ = q∗Σ ◦ θX , where
q∗Σ denotes the localization functor CX −→ Σ−1 CX . Notice that the functor q∗Σ maps final
e denote the morphism q∗ (x) −→ ΘX induced by x −→ λ
objects to final objects. Let λ Σ θX
q∗
Σ (λ)
(that is by q∗Σ (x) −−−→ q∗Σ ◦ θX = ΘX ◦ q∗Σ ) and Tr
fX the essential image of TrX . Then the
data (Σ−1 CX , ΘX , λ; fX ) is a quasi-suspended category.
e Tr
The constructed above map (CX , θX , λ; TrX ) 7−→ (Σ−1 CX , ΘX , λ; fX ) extends to a
e Tr
J∗ ∗ J
functor PCat −→ SCat which is a left adjoint to the inclusion functor SCat −→ PCat.
q∗
Σ
The natural triangle (localization) functors (CX , θX , λ; TrX ) −−−→ (Σ−1 CX , ΘX , λ; fX )
e Tr
form an adjunction arrow IdPCat −→ J∗ J∗ . The other adjunction arrow is identical.
7.4. Quasi-triangulated categories. Let (CX , θX , λ; TrX ) be a presuspended cate-
gory. We call it quasi-triangulated, if the endofunctor θX is an auto-equivalence.
In particular, every quasi-triangulated category is quasi-suspended. Let QTr denote
the full subcategory of PCat (or SCat) whose objects are quasi-triangulated subcategories.
We call a quasi-triangulated category strict if θX is an isomorphism of categories.
7.4.1. Proposition. The inclusion functor QTr −→ PCat has a left adjoint. More
precisely, for each prestable category, TCX = (CX , θX , λ; TrX ), there is a triangle functor
from TCX to a strict quasi-triangulated category such that any triangle functor to a quasi-
triangulated category factors uniquely through this functor.
Proof. The argument is a standard procedure of inverting a functor, which was orig-
inated, probably, in Grothendieck’s work on derivators. One can mimik the argument of
the similar theorem (from suspended to strict triangulated categories) from [KeV].
7.5. The k-linear version. It is obtained by restricting to k-linear additive cate-
gories and k-linear functors. Otherwise all axioms and facts look similarly. Details are left
to the reader.
7.5.1. Remark. Notice that the notion of a quasi-suspended k-linear category pre-
sented here differs from the notion of suspended category proposed by Keller and Vossieck
[KeV1]. In particular, the notion of a quasi-triangulated k-linear category is different from
the notion of a triangulated k-linear category.

70
 7.6. Dual notions. Dualizing the notion of a presuspended category, we obtain the
notion of a precosuspended category. The corresponding, dual, axioms will be denoted by
(PS1∗ ), ... , (PS5∗ ). A precosuspended category TCX = (CX , θX , λ; TrX ) will be called
quasi-cosuspended if the functor θX is conservative and quasi-cotriangulated if θX is an
auto-equivalence.

8. Complement: cohomological and homological functors.

8.1. Cohomological functors.
Fix a svelte presuspended category TCX = (CX , θX , λ; TrX ). Let (CY , IY ) be a left
F
exact category. We say that a functor CX −→ CY is a cohomological functor from TCX to
u v w
(CY , IY ), if for any triangle N −→ M −→ L −→ θX (N ) of TCX , the sequence

F (u) F (v) F (w) F θX (u)

F (N ) −−−→ F (M) −−−→ F (L) −−−→ F θX (N ) −−−→ . . .

is ’exact’. We denote by CF(X) the category whose objects are cohomological functors
from the presuspended category TCX to svelte left exact categories. Morphisms from a
H G
cohomological functor TCX −→ (CY , IY ) to a cohomological functor TCX −→ (CZ , IZ ) is
∼
a pair (F, φ), where F is a functor CY −→ CZ and φ a functor isomorphism F ◦ H −→ G.
The composition is defined in a standard way.
8.1.1. Note. The axiom (PS5) says that the (dual) Yoneda functor

ho
X
∨ op
CX −−−→ (CX ) , M 7−→ CX (M, −),

is cohomological. Equivalently, all representable functors, CX (−, V), are cohomological

functors from TCX to Setsop , or homological functors from the precosuspended category
op
TCX = TCXo to Sets.
8.2. Universal ’exact’ ∂-functors and cohomological functors. Fix a svelte
left exact category (CX , IX ) and consider the category ∂Anc (X, I) of ∂-functors from
(CX , IX ) to cocomplete categories (cf. 5.2 and 5.1). By 5.2.1, every universal ∂-functor
T = (Ti , di | i ≥ 0) from (CX , IX ) to a cocomplete category CY is the composition of the
universal ∂-functor Ext•X,I from (CX , IX ) to CX ∧
= CX∧ and a continuous (i.e. having a
T0∗
right adjoint) functor CX∧ −−−→ CY which is determined uniquely up to isomorphism
by the equality T0 = T0∗ ◦ hX , where hX is the Yoneda embedding CX −→ CX∧ = CX ∧
.
Suppose now that T is an ’exact’ ∂-functor from (CX , IX ) to (CY , IY ) for some left
exact structure IY on the category CY . Then the functor T0∗ maps the exact sequence

bj be d0 (E) b∗ (j)
Θ
b −−−→ M
N c −−−→ L
b −−−→ Θ b ) −−X−→ . . .
b ∗X (N

to an ’exact’ sequence, i.e. T0∗ is a cohomological functor from the presuspended category
TCX∧ = (CX∧ , Θ b ∗ , λX ; Tr
cX ) to the left exact category (CY , IY ).
X

71
 def
Set T + = (Ti , di | i ≥ 1). It follows that T + is the composition of

Ext+ b ∗i b ∗i b∗ • b∗ b ∗i b ∗i
X,I = (ΘX , ΘX d0 | i ≥ 1) ◦ hX = ΘX ◦ ExtX,I = ΘX ◦ (ΘX , ΘX d0 | i ≥ 0) ◦ hX

and the continuous functor T0∗ . Thus,

∗ ∗
Ext+ i i e i i e
X,I ' (θXc , θXc d0 | i ≥ 1) ◦ (qX ◦ hX ) = θXc ◦ (θXc , θXc d0 | i ≥ 0) ◦ (qX ◦ hX ),

where θXc is the suspension endofunctor of the category CXc = Σ−1∗ CX∧ (cf. 7.2.2) which is
Θ
bX
uniquely determined by the equality θXc ◦q∗X = q∗X ◦Θ b ∗ and e
X d0 is uniquely determined by the
∗ ∗ +
equality ed0 qX = qX d0 . This shows that T determines uniquely up to isomorphism (and
is determined by) a continuous cohomological functor T0∗ ◦ θXc from the quasi-suspended
category TCXc = (CXc , θXc , λXc ; TrXc ) associated with the presuspended category TCX∧ of
presheaves of sets on the left exact category (CX , IX ) (see 7.2.2).
It follows from 7.2.3 and 7.2.4 that T + determines a cohomological functor HT from
the stable category CXs of the left exact category of (CX , IX ). The functor HT is the
restriction of the functor T0∗ ◦ θXc to the stable category CXs , which is a subcategory of the
category CXc .
8.3. Homological functors. Homological functors from a precosuspended category
TCX are defined dually. We denote by HF(X) the category of homological functors from
TCX to svelte right exact categories with initial objects.
8.4. Homological functors to cocomplete right exact categories.
8.4.1. Cocomplete right exact categories. We call a right exact category
(CY , EY ) cocomplete if CY has colimits and initial objects, and EY consists of all strict
epimorphisms (in particular, EY is the finest right exact structure on CY ).
8.4.2. Examples. (a) Any Grothendieck topos endowed with the canonical pre-
topology is a cocomplete right exact category.
(b) If (CY , EY ) is an abelian (or a quasi-abelian) category with the canonical right
exact structure, then it is cocomplete iff the category CY has small coproducts.
8.4.3. The category HFc (X). Let TCX = (CX , θX , λ; TrX ) be a svelte precosuspended
category. We denote by HFc (X) the subcategory of the category HF(X) of homological
functors from TCX (cf. 8.3) whose objects are homological functors from TCX to svelte
cocomplete right exact categories and morphisms are morphisms (F, φ) of HF(X) such
that F is a continuous (i.e. having a right adjoint) functor.
8.4.4. Proposition. The category HFc (X) has an initial object.
Proof. By the axiom (PS5∗ ) (dual to (PS5)), the Yoneda functor
hX
∧
CX −−−→ CX , M 7−→ CX (−, M),

is a homological functor from TCX to the category CX∧ endowed with the canonical pre-
hX ∧
topology. The claim is that the cohomological functor TCX −→ CX is universal; i.e. it is
an initial object of the category HFc (X).

72
 F
In fact, for any functor CX −→ CY , where CY is a cocomplete category, there is a
∗
∧ F
continuous functor CX −→ CY which is determined uniquely up to isomorphism by the
equality F ∗ ◦hoX = F . By [GZ, II.1.3], the map F 7−→ F∗ extends to an equivalence between
∧
the category Hom(CX , CY ) of functors from CX to CY and the category Homc (CX , CY ) of
∧
continuous functors from CX to CY .
8.5. Homological functors to fully right exact categories.
8.5.1. Fully right exact and fully left exact categories. We call a right exact
category (CY , EY ) fully right exact if the Yoneda embedding of CY into the category CYE
of sheaves of sets on (CY , EY ) establishes an equivalence between (CY , EY ) and a fully
exact subcategory of the category CYE .
A fully left exact category is defined dually.
8.5.2. Note. The additive version of these notions coincides with the notion of an
exact category, because, in additive case, any fully right (or left) exact category is a fully
exact subcategory of an abelian category.
8.5.3. Proposition. (a) Any Grothendieck topos is a fully right exact category.
(b) Any fully exact (that is full and closed under extensions) subcategory of a fully
right exact category is a fully right exact category.
Proof. (a) A right exact category (CX , EX ) is a Grothendieck topos iff the canonical
j∗
X
functor (CX , EX ) −→ (CXE , EXE ) is an equivalence of right exact categories.
(b) The argument is left to the reader.
8.5.4. The category of homological functors to fully right exact cate-
gories. Let HFr (X) denote the subcategory of the category HF(X) of homological functors
whose objects are homological functors with values in fully right exact svelte categories
H
and morphisms from a homological functor TCX −→ (CY , EY ) to a homological functor
G F
TCX −→ (CZ , EZ ) is a pair (F, φ) such that the functor CY −→ CZ maps deflations to
deflations and conflations to ’exact’ sequences.
8.5.5. Proposition. The category HFr (X) has an initial object.
∧
Proof. Let CXe denote the smallest θX -stable fully exact subcategory of the category CX
(endowed with the canonical right exact structure) containing all representable presheaves.
hX
The Yoneda embedding induces a homological functor CX −→ CXe . The claim is that this
homological functor is an initial object of the category HFr (X).
In fact, let (CY , EY ) be a right exact category and G a homological functor from TCX
j∗
to (CY , EY ). The composition of G with the Yoneda embedding (CY , EY ) −→ Y
(CYE , EsYE )
(cf. 2.8) is a homological functor. Since the category of sheaves CYE is cocomplete, there
is a quasi-commutative diagram
G
CX −−−→ CY
 ∗
hX y y jY

(1)
G∗
∧
CX −−−→ CYE

73
where G ∗ is a continuous functor determined uniquely up to isomorphism by the quasi-
commutativity of the diagram (1). Since the functor G ∗ is right exact, the preimage
−1
G ∗ (CZ ) of any fully exact subcategory CZ of CYE , is a fully exact subcategory of CX ∧
.
∧
In particular, the preimage of CY is a fully exact subcategory of CX . By hypothesis, CY is
a fully exact category, i.e. the full subcategory C eY of CY generated by all representable
E
−1
functors is a fully exact subcategory of CX . Therefore, G ∗ (C
∧ eY ) is a fully exact subcate-
∧
gory of CX containing all representable functors; so that it contains the subcategory CXe ;
i.e. the restriction of the functor G ∗ to the subcategory CXe takes values in the subcategory
CeY . Therefore, G ∗ the composition of G ∗ and the inclusion functor CX ,→ C ∧ is isomorphic
e X
G∗
to the composition j∗Y ◦ Ge∗ , where the functor CXe −→
e
CY is determined uniquely up to
isomorphism.
8.6. Universal cohomological functors. For the reader’s convenience, we sketch
some details of the dual picture. Fix a svelte presuspended category TCX = (CX , θX , λ; TrX ).
We denote by CHc (X) the category whose objects are cohomological functors from TCX
H
to svelte complete categories. Morphisms from a cohomological functor CX −→ CY to
G
a cohomological functor CX −→ CZ is a pair (F, φ), where F is a cocontinuous (that is
∼
having a left adjoint) functor CY −→ CZ and φ a functor isomorphism F ◦ H −→ G. The
composition is defined in a standard way.
8.6.1. Proposition. The category CHc (X) has an initial object.
Proof. By the axiom (PS5), the dual Yoneda functor

ho
X
∨ op
CX −−−→ (CX ) , M 7−→ CX (M, −),

is cohomological. The claim is that the cohomological functor hoX is universal; i.e. it is an
initial object of the category CHc (X).
F
In fact, for any functor CX −→ CY , where CY is a complete category, there is a
∨ op F∗
cocontinuous functor (CX ) −→ CY which is determined uniquely up to isomorphism by
the equality F∗ ◦ hoX = F . By (the dual version of) [GZ, II.1.3], the map F 7−→ F∗ extends
to an equivalence between the category Hom(CX , CY ) of functors from CX to CY and the
∨ op ∨ op
category Homc ((CX ) , CY ) of cocontinuous functors from (CX ) to CY .
8.6.2. Cohomological functors to fully left exact categories. Let CHF` (X)
denote the subcategory of the category CHF(X) of cohomological functors whose objects are
cohomological functors with values in fully left exact svelte categories and morphisms from
H G
a cohomological functor TCX −→ (CY , EY ) to a cohomological functor TCX −→ (CZ , EZ )
F
is a morphism (F, φ) of CHF(X) such that the functor CY −→ CZ maps inflations to
inflations and conflations to ’exact’ sequences.
8.6.3. Proposition. The category CHF` (X) has an initial object.
Proof. The assertion follows from 8.5.5 by duality.

74
 8.7. Universal homological and cohomological functors in k-linear case. Let
TCX = (CX , θX , λ; TrX ) be a k-linear precosuspended category. We denote by HFk (X) the
category whose objects are k-linear homological functors from TCX to svelte exact k-linear
categories and morphisms are morphisms (F, φ) of homological functors (that is morphisms
of the category HFr (X)) such that the (right ’exact’) functor F is k-linear.
8.7.1. Proposition. The category HFk (X) has an initial object.
Proof. (a) Let CXk denote the smallest θX -stable fully exact subcategory of the cate-
gory Mk (X) of presheaves of k-modules on CX containing all representable functors. The
hX
k
Yoneda embedding induces a k-linear homological functor TCX −→ (CXk , EXk ), where EXk
is the (right) exact structure induced by the canonical right exact structure on Mk (X).
The claim is that the homological functor hXk is an initial object of the category HFk (X).
The argument follows is similar to the argument of 8.5.5.
8.7.2. The k-linear additive categories and exact categories with enough
projectives. For any k-linear additive category CY , let CYa denote the full subcategory of
the category Mk (Y ) of presheaves of k-modules on CY whose objects are those presheaves
of k-modules which have a left resolution formed by representable presheaves. One can
deduce from [Ba, I.6.7] that CYa is a fully exact subcategory of the abelian k-linear category
Mk (Y ). Since every representable functor is a projective object of the abelian category
Mk (Y ) and every deflation is a strict epimorphism, it follows that representable functors
are projectives of the exact category CYa . It follows from the definition of CYa that it has
enough projectives with respect to the exact structure induced from Mk (Y ).
8.7.2.1. Proposition. The correspondence CY 7−→ CYa is a functor from the cate-
gory Addk of svelte k-linear additive categories and k-linear functors to the category Pr Exk
of svelte exact categories with enough projectives and right exact functors which map pro-
jectives to projectives.
ϕ
Proof. In fact, any k-linear functor CY −→ CZ extends uniquely up to isomorphism
∗
ϕ
to a continuous k-linear functor Mk (Y ) −→ Mk (Z) such that the diagram
ϕ
CY −−−→ CZ
hY y y hZ
 
ϕ∗
Mk (Y ) −−−→ Mk (Z)

commutes. Since the functor ϕ∗ is right exact and maps representable functors to repre-
ϕa
sentable functors, it induces a functor CYa −→ CZa which is right ’exact’.
8.7.2.2. Remarks. (a) Let CY be a k-linear additive svelte category. Each projective
of the associated exact category CYa is a direct summand of a representable functor.
a hY
Therefore, if the category CY is Karoubian, then the canonical embedding CY −→ CYa
induces an equivalence of the category CY and the full subcategory of CYa generated by
all projectives of (CYa , EYa ).

75
 (b) Suppose that CY is a k-linear additive category endowed with an action

Φ
e
M × CY −−−→ CY

of a monoidal category M f = (M, , I). Then the category CY is endowed with a natural
a

action of M by right exact endofunctors which preserve projectives.

f
In fact, the action M×CY −→ CY extends uniquely up to isomorphism to a continuous
f on the category Mk (Y ) of sheaves
action (i.e. an action by continuous endofunctors) of M
of k-modules on CY which is compatible with the Yoneda embedding. This action induces
an action Φe a of M
f on CY such that the diagram
a

IdM ×hYa
M× CY −−−−−−−→ M ×
 C Ya
Φ
e  y Φa
 e
y
hYa
CY −−−−−−−→ C Ya

quasi-commutes. Notice that the action Φ e a preserves projectives.

(b1) It follows from (a) above that if the category CY is Karoubian, then the functor
hYa
CY −→ CYa induces an equivalence of M-category
f (CY , Φ)
e and the full M-subcategory
f of
the exact M-category (CYa , Φa ) generated by all its projectives.
f e
(c) Let TCX = (CX , θX , λ; TrX ) be a svelte additive k-linear precosuspended category.
It follows from (b) that CXa is a svelte exact k-linear Z+ -category, which has enough pro-
hXa
jectives. It follows that the canonical embedding CX −−−→ CXa is a k-linear homological
functor from TCX to the exact k-linear category CXa .
By (b1), if the category CX is Karoubian, then it is equivalent to the full Z+ -subcategory
of CXa generated by all projectives of the right exact category (CXa , EXa ).
8.7.3. The subcategory HFek (X) of homological functors. For a k-linear preco-
suspended category TCX , we denote by HFek (X) the subcategory of the category HFk (X)
G
whose objects are k-linear homological functors TCX −→ (CZ , EZ ) such that for any arrow
f of the category CX , there exists a cokernel of G(f ). Morphisms of HFek (X) are morphisms
(F, φ) such that the k-linear functor F is ’exact’.
8.7.4. Proposition. (a) Let TCX = (CX , θX , λ; TrX ) be a precosuspended cate-
f
gory having the following property: any morphism M −→ L of CX extends to a triangle
d g f hXa
θX (L) −→ N −→ M −→ L. Then the homological functor CX −−−→ CXa (see 8.7.2.2(c))
is an initial object of the category HFek (X).
(b) Suppose that TCX is a quasi-triangulated k-linear category satisfying the condition
of (a). Then the exact category CXa is abelian.
Proof. (a) Suppose that TCX = (CX , θX , λ; TrX ) is a k-linear precosuspended category
a hX
satisfying the condition of (a). Then the corestriction CX −→ CXa of the Yoneda embedding
to the subcategory CXa is an initial object of the category HFek (X).

76
 The proof of this fact is the same as the argument of [R8, C4.3.4].
(b) If the translation functor θX is an auto-equivalence, then the category CXz is
abelian. The proof of this fact follows the arguments of [R8, C4.4].
8.7.5. Remark. If TCX is a triangulated category, then, by 8.7.4(b), the category
hX a
CXa is abelian. In this case, the universal homological functor TCX −→ CXa is equivalent
to the ’abelianization’ functor of Verdier [Ve2, II.3]. The latter follows from the fact that
the Verdier’s abelianization functor is universal among the homological functors to abelian
categories. More precisely, it is an initial object of the full subcategory of the category
HFek (X) whose objects are homological functors from the triangulated category TCX to
abelian categories.
8.7.6. Cohomological functors. The formulation of the corresponding facts about
k-linear cohomological functors is left to the reader.

Appendix: some properties of kernels.

f
A.1. Proposition. Let M −→ N be a morphism of CX which has a kernel pair,
p1
M ×N M −→ −→ M. Then the morphism f has a kernel iff p1 has a kernel, and these two
p2

kernels are naturally isomorphic to each other.

Proof. Suppose that f has a kernel, i.e. there is a cartesian square
k(f )
 ) −−−→
Ker(f M

f 0y yf (1)
 
iN
x −−−→ N
Then we have the commutative diagram
γ p2
 ) −−−→ M ×
Ker(f N M −−−→ M

f0 y p1 y yf
  
(2)
iM f
x −−−→ M −−−→ N
N i
which is due to the commutativity of (1) and the fact that the unique morphism x −→ N
f
factors through the morphism M −→ N . The morphism γ is uniquely determined by
the equality p2 ◦ γ = k(f ). The fact that the square (1) is cartesian and the equalities
p2 ◦ γ = k(f ) and iN = f ◦ iM imply that the left square of the diagram (2) is cartesian,
γ
i.e. Ker(f ) −−−→ M ×N M is the kernel of the morphism p1 .
Conversely, if p1 has a kernel, then we have a diagram
k(p1 ) p2
 1 ) −−−→ M ×
Ker(p N M −−−→ M
0 
p1 y cart p1 y cart y f
 
iM f
x −−−→ M −−−→ N

77
which consists of two cartesian squares. Therefore the square
k(f )
 1 ) −−−→
Ker(p M

0
p1 y cart yf

iN
x −−−→ N

with k(f ) = p2 ◦ k(p1 ) is cartesian.

A.2. Remarks. (a) Needless to say that the picture obtained in (the argument of)
τf0
A.1 is symmetric, i.e. there is an isomorphism Ker(p1 ) −→ Ker(p2 ) which is an arrow in
the commutative diagram
k(p1 ) p1
 1 ) −−−→ M ×N M
Ker(p −−−→ M

0 
τf yo τf yo y idM
 
k(p2 ) p2
Ker(p2 ) −−−→ M ×N M −−−→ M
p1
f
(b) Let a morphism M −→ N have a kernel pair, M ×N M −→
−→ M, and a kernel. Then,
p2
k(p1 )
by A.1, there exists a kernel of p1 , so that we have a morphism Ker(p1 ) −−−→ M ×N M
∆M
and the diagonal morphism M −−−→ M ×N M . Since the left square of the commutative
diagram
p01
x
 −−−→ Ker(p  1) −−−→ x

cart c(p1 )y
  
y y
∆M p1
M −−−→ M ×N M −−−→ M
is cartesian and compositions of the horizontal arrows are identical morphisms, it follows
that its left square is cartesian too. Loosely, one can say that the intersection
` of Ker(p1 )
with the diagonal of M ×N M is zero. If there exists a coproduct Ker(p1 ) M , then the
k(p1 ) ∆M
pair of morphisms Ker(p1 ) −−−→ M ×N M ←−−− M determine a morphism
a
Ker(p1 ) M −−−→ M ×N M.

If the category
` CX is additive, then this morphism is an isomorphism, or, what is the
same, Ker(f ) M ' M ×N M . In general, it is rarely the case, as the reader can find
out looking at the examples of 1.4.
A.3. Proposition. Let
fe
f −−−→
M
 N
e
ge y cart yg (3)
 
f
M −−−→ N

78
be a cartesian square. Then Ker(f ) exists iff Ker(fe) exists, and they are naturally iso-
morphic to each other.
A.4. The kernel of a composition and related facts. Fix a category CX with
an initial object x.
f g
A.4.1. The kernel of a composition. Let L −→ M and M −→ N be morphisms
such that there exist kernels of g and g ◦ f . Then the argument similar to that of A.3
shows that we have a commutative diagram
fe g0
Ker(gf) −−−→ Ker(g)
 −−−→ x

k(gf ) y

cart y k(g) cart

y iN
 (1)
f g
L −−−→ M −−−→ N
whose both squares are cartesian and all morphisms are uniquely determined by f, g and
the (unique up to isomorphism) choice of the objects Ker(g) and Ker(gf ).
Conversely, if there is a commutative diagram
u g0
K −−−→ Ker(g)
 −−−→ x

t y cart y k(g) y iN
  
f g
L −−−→ M −−−→ N
t
whose left square is cartesian, then its left vertical arrow, K −→ L, is the kernel of the
g◦f
composition L −−−→ N .
f
A.4.2. Remarks. (a) It follows from A.3 that the kernel of L −→ M exists iff
fe
the kernel of Ker(gf ) −−−→ Ker(g) exists and they are isomorphic to each other. More
precisely, we have a commutative diagram
k(fe) fe g0
Ker(
 f)
e −−−→ Ker(gf) −−−→ Ker(g)
 −−−→ x

oy k(gf ) y cart y k(g) cart y iN
   
k(f ) f g
Ker(f ) −−−→ L −−−→ M −−−→ N
whose left vertical arrow is an isomorphism.
(b) Suppose that (CX , EX ) is a right exact category (with an initial object x). If the
morphism f above is a deflation, then it follows from this observation that the canonical
fe
morphism Ker(gf ) −−−→ Ker(g) is a deflation too. In this case, Ker(f ) exists, and we
have a commutative diagram
k(fe) fe
Ker( e) −−−→ Ker(gf )
f −−−→ Ker(g)
  
oy k(gf ) y cart y k(g)
  
k(f ) f
Ker(f ) −−−→ L −−−→ M

79
whose rows are conflations.
The following observations is useful (and are used) for analysing diagrams.
g
A.4.3. Proposition.(a) Let M −→ N be a morphism with a trivial kernel. Then
f
a morphism L −→ M has a kernel iff the composition g ◦ f has a kernel, and these two
kernels are naturally isomorphic one to another.
(b) Let
f
L
 −−−→ M

γ y yg
 
φ
f −−−→
M N
be a commutative square such that the kernels of the arrows f and φ exist and the kernel
of g is trivial. Then the kernel of the composition φ ◦ γ is isomorphic to the kernel of the
morphism f , and the left square of the commutative diagram

∼ k(f ) f
Ker(f ) −−−→ Ker(φγ)
 −−−→ L
 −−−→ M

γ cart γ y yg
  
ey
k(φ) φ
Ker(φ) −−−→ f −−−→
M N

is cartesian.
Proof. (a) Since the kernel of g is trivial, the diagram A.4.1(1) specializes to the
diagram
fe idx
Ker(gf ) −−−→ x −− −→  x
k(gf ) y cart y k(g) y iN
  
f g
L −−−→ M −−−→ N
with cartesian squares. The left cartesian square of this diagram is the definition of Ker(f ).
The assertion follows from A.4.1.
(b) Since the kernel of g is trivial, it follows from (a) that Ker(f ) is naturally isomor-
phic to the kernel of g ◦ f = φ ◦ γ. The assertion follows now from A.4.1.
f
A.4.4. Corollary. Let CX be a category with an initial object x. Let L −→ M be a
g k(g)
strict epimorphism and M −→ N a morphism such that Ker(g) −−−→ M exists and is a
monomorphism. Then the composition g ◦ f is a trivial morphism iff g is trivial.
A.4.4.1. Note. The following example shows that the requirement ”Ker(g) −→ M
is a monomorphism” in A.4.4 cannot be omitted.
Let CX be the category Algk of associative unital k-algebras, and let m be an ideal
of the ring k such that the epimorphism k −→ k/m does not split. Then the identical
morphism k/m −→ k/m is non-trivial, while its composition with the projection k −→ k/m
is a trivial morphism.

80
 f
A.5. The coimage of a morphism. Let M −→ N be an arrow which has a kernel,
i.e. we have a cartesian square

k(f )
 ) −−−→
Ker(f M

f 0y cart yf
 
iN
x −−−→ N

k(f )
with which one can associate a pair of arrows Ker(f ) −→
−→ M, where 0f is the composition
0 f

0 iM
of the projection f and the unique morphism x −→ M . Since iN = f ◦ iM , the morphism
k(f )
f equalizes the pair Ker(f ) −→
−→ M. If the cokernel of this pair of arrows exists, it will
0 f

be called the coimage of f and denoted by Coim(f ), or. loosely, M/Ker(f ).

f
Let M −→ N be a morphism such that Ker(f ) and Coim(f ) exist. Then f is the
pf
composition of the canonical strict epimorphism M −−−→ Coim(f ) and a uniquely defined
jf
morphism Coim(f ) −−−→ N .
f
A.5.1. Lemma. Let M −→ N be a morphism such that Ker(f ) and Coim(f ) exist.
∼
There is a natural isomorphism Ker(f ) −→ Ker(pf ).
Proof. The outer square of the commutative diagram

f0
Ker(f) −−−→ x
 −−−→ x

k(f )y cart (1)
  
y y
pf jf
M −−−→ Coim(f ) −−−→ L

is cartesian. Therefore, its left square is cartesian which implies, by A.3, that Ker(pf ) is
isomorphic to Ker(f 0 ). But, Ker(f 0 ) ' Ker(f ).
A.5.2. Note. By A.4.1, all squares of the commutative diagram

f0
Ker(f
 ) −−−→ x

id y cart
 
y
pf
e
Ker(jf pf ) −−−→ Ker(j −−−→ (2)
 f) x

k(f )y cart cart
  
y y
pf jf
M −−−→ Coim(f ) −−−→ L

are cartesian.

81
 f
If CX is an additive category and M −→ L is an arrow of CX having a kernel and
jf
a coimage, then the canonical morphism Coim(f ) −−−→ L is a monomorphism. Quite a
few non-additive categories have this property.
A.5.3. Example. Let CX be the category Algk of associative unital k-algebras.
Since cokernels of pairs of arrows exist in Algk , any algebra morphism has a coimage. It
ϕ
follows from 1.4.1 that the coimage of an algebra morphism A −→ B is A/K(ϕ), where
K(ϕ) is the kernel of φ in the usual sense (i.e. in the category of non-unital algebras).
The canonical decomposition ϕ = jϕ ◦ pϕ coincides with the standard presentation of ϕ as
the composition of the projection A −→ A/K(ϕ) and the monomorphism A/K(ϕ) −→ B.
In particular, ϕ is strict epimorphism of k-algebras iff it is isomorphic to the associated
pϕ
coimage map A −−−→ Coim(ϕ) = A/K(ϕ).

82
 Lecture 5. Universal K-functors.

1. Preliminaries: left exact categories of right exact ’spaces’.

We start with left exact stuctures formed by localizations of ’spaces’ represented by
svelte categories. Then the obtained facts are used to define natural left exact structures
on the category of ’spaces’ represented by right exact categories.
The following proposition is a refinement of [R3, 1.4.1].
f q
1.1. Proposition. Let Z ←− X −→ Y be morphisms of ’spaces’ such that q (i.e. its
q∗
inverse image functor CY −→ CX ) is a localization. Then
q a
(a) The canonical morphism Z −→ Z Y is a localization.
e
f,q
(b) If q is a continuous localization, then qe is a continuous localization.
(c) If Σq∗ = {s ∈ HomCY | q ∗ (s) is invertible} is a left (resp. right) multiplicative
system, then Σe q∗
has the same property.
f q
1.2. Corollary. Let Z ←− X −→ Y be morphisms of ’spaces’ such that q is a
q a
localization, and let Z −→ Z Y be a canonical morphism. Suppose that the category CY
e
f,q
has finite limits (resp. finite colimits). Then qe∗ is a left (resp. right) exact localization, if
the localization q ∗ is left (resp. right) exact.
Proof. By 1.1(a), qe∗ is a localization functor.
q∗
Suppose that the category CY has finite limits and the localization functor CY −→ CX
is left exact. Then it follows from [GZ, I.3.4] that Σq∗ = {s ∈ HomCY | q ∗ (s) is invertible}
is a right multiplicative system. The latter implies, by 1.1(c), that Σe q∗
is a right multi-
∗
plicative system. Therefore, by [GZ, I.3.1], the localization functor qe is left exact.
The following assertion is a refinement of [R3, 1.4.2].
p q
1.3. Proposition. Let X ←− Z −→ Y be morphisms of ’spaces’ such that p∗ and q ∗
are localization functors. Then the square
q
Z
 −−−→ Y

py y p1
 
q1 a
X −−−→ X Y
p,q

is cartesian.
1.4. Left exact structures on the category of ’spaces’. Let L denote the
class of all localizations of ’spaces’ (i.e. morphisms whose inverse image functors are
q
localizations). We denote by L` (resp. Lr ) the class of localizations X −→ Y of ’spaces’
such that Σq∗ = {s ∈ HomCY | q ∗ (s) is invertible} is a left (resp. right) multiplicative

83
system. We denote by Le the intersection of L` and Lr (i.e. the class of localizations q
such that Σq∗ is a multiplicative system) and by Lc the class of continuous (i.e. having a
direct image functor) localizations of ’spaces’. Finally, we set Lce = Lc ∩ Le ; i.e. Lce is the
q
class of continuous localizations X −→ Y such that Σq∗ is a multiplicative system.
1.4.1. Proposition. Each of the classes of morphisms L, L` , Lr , Le , Lc , and Lce
are structures of a left exact category on the category |Cat|o of ’spaces’.
Proof. It is immediate that each of these classes is closed under composition and
contains all isomorphisms of the category |Cat|o . It follows from 1.1 that each of the
classes is stable under cobase change. In other words, the arrows of each class can be
regarded as cocovers of a copretopology. It remains to show that these copretopologies are
subcanonical. Since L is the finest copretopology, it suffices to show that L is subcanonical.
The copretopology L being subcanonical means precisely that for any localization
q
X −→ Y , the square
q
 −−−→
X Y

qy y q1
 
q2 a
Y −−−→ Y Y
q,q

is cartesian. But, this follows from 1.3.

1.5. Observation. Each object of the left exact category (|Cat|o , Lc ) is injective.
q
In fact, a ’space’ X is an injective object of (|Cat|o , Lc ) iff each inflation X −→ Y
t
is split; i.e. there is a morphism Y −→ X such that t ◦ q = idX . Since the morphism
q is continuous, it has a direct image functor, q∗ , which is fully faithful, because q ∗ is a
localization functor. The latter means precisely that the adjunction arrow q ∗ q∗ −→ IdCX
t
is an isomorphism; i.e. the morphism Y −→ X whose inverse image functor coincides with
q∗ satisfies the equality t ◦ q = idX .
1.6. Left exact structures on the category of right (or left) exact ’spaces’.
A right exact ’space’ is a pair (X, EX ), where X is a ’space’ and EX is a right exact
structure on the category CX . We denote by Espr the category whose objects are right
exact ’spaces’ (X, EX ) and morphisms from (X, EX ) to (Y, EY ) are given by morphisms
f
X −→ Y of ’spaces’ whose inverse image functor, f ∗ , is ’exact’; i.e. f ∗ maps deflations to
deflations and preserves pull-backs of deflations.
Dually, a left exact ’space’ is a pair (Y, IY ), where (CY , IY ) is a left exact category. We
denote by Esp` the category whose objects are left exact ’spaces’ (Y, IY ) and morphisms
(Y, IY ) −→ (Z, IZ ) are given by morphisms Y −→ Z whose inverse image functors are
’coexact’, which means that they preserve inflations and their push-forwards.
1.6.1. Note. The categories Espr and Esp` are naturally isomorphic to each other:
the isomorphism is given by the dualization functor (X, EX ) 7−→ (X o , Eop X ). Therefore,
every assertion about the category Espr of right exact ’spaces’ translates into an assertion
about the category Esp` of left exact ’spaces’ and vice versa.
1.6.2. Proposition. The category Espr has fibered coproducts.

84
 1.6.3. Canonical left exact structures on the category Espr . Let Les denote
q
the class of all morphisms (X, EX ) −→ (Y, EY ) of right exact ’spaces’ such that q ∗ is a
localization functor and each arrow of EX is isomorphic to an arrow q ∗ (e) for some e ∈ EY .
If Σq∗ is a left or right multiplicative system, then this condition means that EX is
the smallest right exact structure containing q ∗ (EY ).
1.6.3.1. Proposition. The class Les is a left exact structure on the category Espr
of right exact ’spaces’.
1.6.3.2. Corollary. Each of the classes of morphisms of ’spaces’ L` , Lr , Le , Lc ,
and Lce (cf. 1.4, 1.4.1) induces a structure of a left exact category on the category Espr of
right exact ’spaces’.
Proof. The class L` induces the class Les ` of morphisms of the category Espr formed
q q
by all arrows (X, EX ) −→ (Y, EY ) from Les such that the morphism of ’spaces’ X −→ Y
e,c
belongs to L` . Similarly, we define the classes Les es c
` , Lr , Les , and Les .

1.6.3.3. The left exact structure Les sq . For a right exact ’space’ (X, EX ), let
Sq(X, EX ) denote the class of all cartesian squares in the category CX some of the arrows
of which (at least two) belong to EX .
q
The class Lessq consists of all morphisms (X, EX ) −→ (Y, EY ) of right exact ’spaces’
such that its inverse image functor, q∗ , is equivalent to a localization functor and each
square of Sq(X, EX ) is isomorphic to some square of q∗ (Sq(Y, EY )).
1.6.3.4. Proposition. The class Les sq is a left exact structure on the category Espr
of right exact ’spaces’ which is coarser than Les and finer than Lesr .

Proof. The argument is left to the reader.

1.7. Relative right exact ’spaces’. The category Espr of right exact ’spaces’ has
initial objects and no final object. Final objects appear if we fix a right exact ’space’
S = (S, ES ) and consider the category Espr /S instead of Espr . The category Espr /S has
a natural final object and cokernels of all morphisms. It also inherits left exact structures
from Espr , in particular those defined above (see 1.6.3.2). Therefore, our theory of derived
functors (satellites) can be applied to functors from Espr /S.
1.8. The category of right exact k-’spaces’. For a commutative unital ring k,
we denote by Esprk the category whose objects are right exact ’spaces’ (X, EX ) such that
CX is a k-linear additive category and morphisms are morphisms of right exact ’spaces’
whose inverse image functors are k-linear.
e,c
Each of the left exact structures Les , Les es es c
` , Lr , Lr , Les , and Les induces a left exact
structure on the category Esprk of right exact k-’spaces’. We denote them by respectively
e,c
Les (k), Les es es c
` (k), Lr (k), Lr (k), Les (k), and Les (k).

2. The group K0 of a right (or left) exact ’space’.

2.1. The group Z0 |CX |. For a svelte category CX , we denote by |CX | the set of
isomorphism classes of objects of CX , by Z|CX | the free abelian group generated by |CX |,

85
and by Z0 (CX ) the subgroup of Z|CX | generated by differences [M ] − [N ] for all arrows
M −→ N of the category CX . Here [M ] denotes the isomorphism class of an object M .
2.2. Proposition. (a) The maps X 7−→ Z|CX | and X 7−→ Z0 (CX ) extend natu-
rally to presheaves of Z-modules on the category of ’spaces’ |Cat|o (i.e. to functors from
(|Cat|o )op to Z − mod).
(b) If the category CX has an initial (resp. final) object x, then Z0 (CX ) is the sub-
group of Z|CX | generated by differences [M ] − [x], where [M ] runs through the set |CX | of
isomorphism classes of objects of CX .
Proof. The argument is left to the reader.
op op
2.3. Note. Evidently, Z|CX | ' Z|CX | and Z0 (CX ) ' Z0 (CX ).
2.4. The group K0 of a right exact ’space’. Let (X, EX ) be a right exact ’space’.
We denote by K0 (X, EX ) the quotient of the group Z0 |CX | by the subgroup generated by
the expressions [M 0 ] − [L0 ] + [L] − [M ] for all cartesian squares

fe
0
M −−−→ M

e0 y cart ye
 
f
L0 −−−→ L

whose vertical arrows are deflations.

We call K0 (X, EX ) the group K0 of the right exact ’space’ (X, EX ).
2.4.1. Example: the group K0 of a ’space’. Any ’space’ X is identified with the
trivial right exact ’space’ (X, Iso(CX )). We set K0 (X) = K0 (X, Iso(CX )). That is K0 (X)
coincides with the group Z0 (CX ).
2.4.2. Proposition. Let (X, EX ) be a right exact ’space’ such that the category CX
has initial objects. Then K0 (X, EX ) is isomorphic to the quotient of the group Z0 (X) by the
subgroup generated by the expressions [M ] − [L] − [N ] for all conflations N −→ M −→ L.
2.5. Proposition. (a) The map (X, EX ) 7−→ K0 (X, EX ) extends to a contravariant
functor, K0 , from the category Espr of right exact ’spaces’ to the category Z − mod.
f
(b) Let (X, EX ) −→ (Y, EY ) be a morphism of Espr such that every object of the
category CX is isomorphic to the inverse image of an object of CY . Then the map
K0 (f )
K0 (Y, EY ) −−−→ K0 (X, EX ) is a group epimorphism.
In particular, the functor K0 maps ’exact’ localizations to epimorphisms.

3. Higher K-groups of right exact ’spaces’.

3.1. The relative functors K0 and their derived functors. Fix a right exact
K0
’space’ Y = (Y, EY ). The functor (Espr )op −−−→ Z − mod induces a functor

K0Y
(Espr /Y)op −−−→ Z − mod

86
defined by
ξ K0 (ξ)
K0Y (X , ξ) = K0Y (X , X → Y) = Cok(K0 (Y) −−−→ K0 (X ))
and acting correspondingly on morphisms.
The main advantage of the functor K0Y is that its domain, the category Espr /Y has
a final object, cokernels of morphisms, and natural left exact structures induced by left
exact structures on Espr . Fix a left exact structure I on Espr (say, one of those defined in
6.8.3.2) and denote by IY the left exact structure on Espr /Y induced by I. Notice that,
since the category Z − mod is complete (and cocomplete), there is a well defined satellite
endofunctor of Hom((Espr /Y)op , Z − mod), F 7−→ SIY F. So that for every functor F
from (Espr /Y)op to Z − mod, there is a unique up to isomorphism universal ∂ ∗ -functor
(SIi Y F, di | i ≥ 0).
In particular, there is a universal contravariant ∂ ∗ -functor K•Y,I = (KiY,I , di | i ≥ 0)
from the right exact category (Espr /Y, IY ) of right exact ’spaces’ over Y to the category
Z − mod of abelian groups; that is KiY,I = SIi Y K0Y,I for all i ≥ 0.
We call the groups KiY,I (X , ξ) universal K-groups of the right exact ’space’ (X , ξ)
over Y with respect to the left exact structure I.
3.2. ’Exactness’ properties. In general, the ∂ ∗ -functor K•Y,I is not ’exact’. The
purpose of this section is to find some natural left exact structures I on the category Espr /Y
of right exact ’spaces’ over Y and some of its subcategories for which the ∂ ∗ -functor K•Y,I
is ’exact’.
q
3.2.1. Proposition. Let (X, ξ) −→ (X 0 , ξ 0 ) be a morphism of the category Espr /Y
q
such that X −→ X 0 belongs to Les (cf. 6.8.3) and has the following property:
s
(#) if M −→ L is a morphism of CX 0 such that q ∗ (s) is invertible, then the element
K0 (cq )
[M ] − [L] of the group K0 (X 0 ) belongs to the image of the map K0 (X 00 ) −−−→ K0 (X 0 ),
cq q
where (X 0 , ξ 0 ) −→ (X 00 , ξ 00 ) is the cokernel of the morphism (X, ξ) −→ (X 0 , ξ 0 ).
Suppose, in addition, that one of the following two conditions holds:
(i) the category CX 0 has an initial object;
f s
(ii) for any pair of arrows N −→ L ←− M, of the category CX 0 such that q∗ (s) is
invertible, there exists a commutative square

fe
N
e
 −−−→ M

ty ys
 
f
N −−−→ L

such that q∗ (t) is invertible.

q cq
Then for every conflation (X, ξ) −→ (X 0 , ξ 0 ) −→ (X 00 , ξ 00 ) of the left exact category
(Espr /Y, IY ) the sequence

K0Y (cq ) K0Y (q)

K0Y (X 00 , ξ 00 ) −−−→ K0Y (X 0 , ξ 0 ) −−−→ K0Y (X, ξ) −−−→ 0

87
of morphisms of abelian groups is exact.
q
3.2.2. Proposition. The class LY 0 0
es of all morphisms (X, ξ) −→ (X , ξ ) of Espr /Y
q
such that X −→ X 0 belongs to Les and satisfies the condition (#) of 3.2.1, is a left exact
structure on the category Espr /Y.
q
3.2.2.1. Proposition. The class LY 0 0 Y
es,r of all morphisms (X, ξ) −→ (X , ξ ) og Les
q∗
such that the functor CX 0 −→ CX satisfies the condition (ii) of 3.2.1, is a left exact
structure on the category Espr /Y.
3.2.3. Proposition. Let Y = (Y, EY ) be a right exact ’space’, and let I be a left exact
structure on the category Espr /Y which is coarser than LYes,r (cf. 3.2.2). Then the universal
∗ Y Y
∂ -functor K• = (Ki , di | i ≥ 0) from the left exact category (Espr /Y, IY ) to the category
q cq
Z−mod of abelian groups is ’exact’; i.e. for any conflation (X, ξ) −→ (X 0 , ξ 0 ) −→ (X 00 , ξ 00 ),
the associated long sequence

K1Y (q) d0 K0Y (cq ) K0Y (q)

. . . −−−→ K1Y (X, ξ) −−−→ K0Y (X 00 , ξ 00 ) −−−→ K0Y (X 0 , ξ 0 ) −−−→ K0Y (X, ξ) −−−→ 0

is exact.
Proof. Since the left exact structure IY is coarser than LY
es , it satisfies the condition
q cq
(#) of 3.2.1. Therefore, by 3.2.1, for any conflation (X, ξ) −→ (X 0 , ξ 0 ) −→ (X 00 , ξ 00 ) of the
left exact category (Esp∗r /Y, IY ), the sequence

K0Y (cq ) K0Y (q)

K0Y (X 00 , ξ 00 ) −−−→ K0Y (X 0 , ξ 0 ) −−−→ K0Y (X, ξ) −−−→ 0

of Z-modules is exact. Therefore, by [Lecture III, 3.5.4.1], the universal ∂ ∗ -functor K•Y =
(KiY , di | i ≥ 0) from (Esp∗r /Y, IY )op to Z − mod is ’exact’.
The following proposition can be regarded as a machine for producing universal ’exact’
K-functors.
3.2.4. Proposition. Let Y = (Y, EY ) be a right exact ’space’, (CS , IS ) a left
exact category with final objects, and F a functor CS −→ Espr /Y which maps conflations
of (CS , IS ) to conflations of the left exact category (Espr /Y, LY
es,r ). Then there exists a
∗ S,F S,F
(unique up to isomorphism) universal ∂ -functor K• = (Ki , di | i ≥ 0) from the right
exact category (CS , IS )op to Z − mod whose zero component, K0S,F , is the composition of
op Fop
the functor CS −−−→ Espr /Y op and the functor K0Y .
The ∂ ∗ -functor K•S,F is ’exact’.
Proof. The existence of the ∂ ∗ -functor K•S,F follows, by [Lecture III, 3.3.2], from the
completeness (– existence of limits of small diagrams) of the category Z − mod of abelian
groups. The main thrust of the proposition is the ’exactness’ of K•S,F .
By hypothesis, the functor F maps conflations to conflations. Therefore, it follows
from 3.2.1 that for any conflation X −→ X0 −→ X00 of the left exact category (CS , IS ),

88
the sequence of abelian groups K0S,F (X00 ) −→ K0S,F (X0 ) −→ K0S,F (X) −→ 0 is exact. By
[Lecture III, 3.5.4.1], this implies the ’exactness’ of the ∂ ∗ -functor K•Y .
3.3. The ’absolute’ case. Let |Cat∗ |o denote the subcategory of the category
|Cat|o of ’spaces’ whose objects are ’spaces’ represented by categories with initial objects
and morphisms are those morphisms of ’spaces’ whose inverse image functors map initial
objects to initial objects. The category |Cat∗ |o is pointed: it has a canonical zero (that is
both initial and final) object, x, which is represented by the category with one (identical)
morphism. Thus, the initial objects of the category |Cat|o of all ’spaces’ are zero objects
of the subcategory |Cat∗ |o .
f cf
Each morphism X −→ Y of the category |Cat∗ |o has a cokernel, Y −→ C(f), where
the category CC(f) representing the ’space’ C(f) is the kernel Ker(f∗ ) of the functor f∗ .
By definition, Ker(f∗ ) is the full subcategory of the category CY generated by all objects
of CY which the functor f∗ maps to initial objects. The inverse image functor c∗f of the
canonical morphism cf is the natural embedding Ker(f∗ ) −→ CY .
Let Esp∗r denote the category formed by right exact ’spaces’ with initial objects and
those morphisms of right exact ’spaces’ whose inverse image functor is ’exact’ and maps
initial objects to initial objects. The category Esp∗r is pointed and the forgetfull functor

J∗
Esp∗r −−−→ |Cat∗ |o , (X, EX ) 7−→ X,

J
∗
is a left adjoint to the canonical full embedding |Cat∗ |o −→ Esp∗r which assigns to every
’space’ X the right exact category (X, Iso(CX )). Both functors, J∗ and J∗ , map zero
objects to zero objects.
Let x be a zero object of the category Esp∗r . Then Esp∗r /x is naturally isomorphic to
Esp∗r and the relative K0 -functor K0x coincides with the functor K0 .
3.3.1. The left exact structure L∗es . We denote by L∗es the canonical left exact
structure Lxes ; it does not depend on the choice of the zero object x. It follows from
q
the definitions above that L∗es consists of all morphisms (X, EX ) −→ (Y, EY ) having the
following properties:
q∗
(a) CY −→ CX is a localization functor (which is ’exact’), and every arrow of EX is
isomorphic to an arrow of q∗ (EY ).
s
(b) If M −→ M 0 is an arrow of CY such that q∗ (s) is an isomorphism, then [M ] − [M 0 ]
is an element of KerK0 (q)|.
F
3.3.2. Proposition. Let (CS , IS ) be a left exact category, and CS −→ Esp∗r
a functor which maps conflations of (CS , IS ) to conflations of the left exact category
(Esp∗r , L∗es ). Then there exists a (unique up to isomorphism) universal ∂ ∗ -functor K•S,F =
(KiS,F , e
di | i ≥ 0) from (CS , IS )op to Z − mod whose zero component, K0S,F , is the com-
op Fop
position of the functor CS −−−→ (Esp∗r )op and the functor K0 .
The ∂ ∗ -functor K•S,F is ’exact’. In particular, the ∂ ∗ -functor K• = (Ki , di | i ≥ 0)
from (Esp∗r , L∗es ) to Z − mod is ’exact’.

89
 Proof. The assertion is a special case of 3.2.4.
3.4. Universal K-theory of abelian categories. Let Espak denote the category
whose objects are ’spaces’ X represented by k-linear abelian categories and morphisms
f
X −→ Y are represented by k-linear exact functors.
There is a natural functor
F
Espak −−−→ Esp∗r (1)
which assigns to each object X of the category Espak the right exact (actually, exact) ’space’
(X, Est st
X ), where EX is the standard (i.e. the finest) right exact structure on the category
f f
CX , and maps each morphism X −→ Y to the morphism (X, Est st
X ) −→ (Y, EY ) of right
exact ’spaces’. One can see that the functor F maps the zero object of the category Espak
(represented by the zero category) to a zero object of the category Esp∗r .
3.4.1. Proposition. Let CX and CY be k-linear abelian categories endowed with the
q∗
standard exact structure. Any exact localization functor CY −→ CX satisfies the conditions
(a) and (b) of 3.3.1.
eh
Proof. In fact, each morphism q∗ (M ) −→ q∗ (N ) is of the form q∗ (h)q∗ (s)−1 for some
h s
morphisms M 0 −→ N and M 0 −→ M such that q∗ (s) is invertible. The morphism h is a
(unique) composition j ◦ e, where j is a monomorphism and e is an epimorphism. Since the
functor q∗ is exact, q∗ (j) is a monomorphism and q∗ (e) is an epimorphism. Therefore, e
h is
∗
an epimorphism iff q (j) is an isomorphism. This shows that the condition (a) holds.
s
Let M −→ M 0 be a morphism and
s
0 −→ Ker(s) −→ M −→ M 0 −→ Cok(s) −→ 0

the associated with s exact sequence. Representing s as the composition, j◦e, of a monomor-
phism j and an epimorphism e, we obtain two short exact sequences,
e j
0 −→ Ker(s) −→ M −→ N −→ 0 and 0 −→ N −→ M 0 −→ Cok(s) −→ 0,

hence [M ] = [Ker(s)] + [N ] and [M 0 ] = [N ] + [Cok(s)], or [M 0 ] = [M ] + [Ker(s)] − [Cok(s)]

in K0 (Y ). It follows from the exactness of the functor q∗ that the morphism q∗ (s) is an
isomorphism iff Ker(s) and Cok(s) are objects of the category Ker(q∗ ). Therefore, in this
case, it follows that [M 0 ] = [M ] modulo Z|Ker(q∗ )| in K0 (Y ).
q
3.4.2. Proposition. (a) The class La of all morphisms X −→ Y of the category
q∗
Espak such that CY −→ CX is a localization functor, is a left exact structure on Espak .
F
(b) The functor Espak −−−→ Esp∗r is an ’exact’ functor from the left exact category
(Espk , La ) to the left exact category (Esp∗r , L∗es ). Moreover, La = F−1 (L∗es ), that is the left
a

exact structure La is induced by the left exact structure L∗es via the functor F.
3.4.3. The universal Grothendieck K-functor. The composition K0a of the
functor
Fop
(Espak )op −−−→ (Esp∗r )op

90
 K0∗
∗ op
and the functor (Espr ) −−−→ Z − mod assigns to each object Xof the category Espak
the abelian group K0∗ (X, Est K0∗ (X, Est
X ). It follows from 2.4.2 that the group X ) coincides
with the Grothendieck group of the abelian category CX . Therefore, we call K0a the
Grothendieck K0 -functor.
3.4.4. Proposition. There exists a universal ∂ ∗ -functor K•a = (Kia , dai | i ≥ 0) from
the right exact category (Espak , La )op to the category Z − mod whose zero component is the
Grothendieck functor K0 . The universal ∂ ∗ -functor K•a is ’exact’; that is for any exact
q
localization X −→ X 0 , the canonical long sequence
K1a (q) da
0 (q) K0a (cq ) K0a (q)
. . . −−−→ K1a (X) −−−→ K0a (X 00 ) −−−→ K0a (X 0 ) −−−→ K0a (X) −−−→ 0 (3)
is exact.
F
Proof. By 3.4.2(b), the functor Espak −−−→ Esp∗r is an ’exact’ functor from the left
exact category (Espak , La ) to the left exact category (Esp∗r , L∗es ) which maps the zero object
of the category Espak (– the ’space’ represented by the zero category) to a zero object of
the category Esp∗r . Therefore, F maps conflations to conflations.
The assertion follows now from 3.3.2.1 applied to the functor F.
3.4.5. The universal ∂ ∗ -functor K•a and the Quillen’s K-theory. For a ’space’
X represented by a svelte k-linear abelian category CX , we denote by KiQ (X) the i-th
Quillen’s K-group of the category CX . For each i ≥ 0, the map X 7−→ KiQ (X) extends
naturally to a functor
KiQ
(Espak )op −−−→ Z − mod
It follows from the Quillen’s localization theorem [Q, 5.5] that for any exact localization
q dQ
i (q)
0
X −→ X and each i ≥ 0, there exists a connecting morphism Q
Ki+1 (X) −−−→ K0Q (X 00 ),
where CX 00 = Ker(q∗ ), such that the sequence
K1Q (q) dQ
0 (q) K0Q (cq ) K0Q (q)
. . . −−−→ K1Q (X) −−−→ K0Q (X 00 ) −−−→ K0Q (X 0 ) −−−→ K0Q (X) −−−→ 0 (4)
is exact. It follows (from the proof of the Quillen’s localization theorem) that the con-
necting morphisms dQ i (q), i ≥ 0, depend functorially on the localization morphism q. In
∗
other words, K•Q = (KiQ , dQ i | i ≥ 0) is an ’exact’ ∂ -functor from the left exact category
(Espak , La )op to the category Z − mod of abelian groups.
Naturally, we call the ∂ ∗ -functor K•Q the Quillen’s K-functor.
Since K•a = (Kia , dai | i ≥ 0) is a universal ∂ ∗ -functor from (Espak , La )op to Z − mod,
the identical isomorphism K0Q −→ K0a extends uniquely to a ∂ ∗ -functor morphism
ϕQ
•
K•Q −−−→ K•a . (5)

4. The universal K-theory of exact categories. Let Espek denote the subcate-
gory of the category Esp∗r whose objects are ’spaces’ represented by svelte exact k-linear
categories and inverse image of morphisms are k-linear functors.

91
 There is a natural functor
Fr
Espek −−−→ Esp∗r (1)
which maps objects and morphisms of the category Espek to the corresponding objects and
morphisms of the category Esp∗r .
Fr
4.1. Proposition. The functor Espek −−−→ Esp∗r preserves cocartesian squares and
maps the zero object of the category Espek to the zero object of the category Esp∗r .
Proof. The argument is similar to that of 7.5.2(b). Details are left to the reader.
4.2. Corollary. The class of morphisms Lek = F−1 ∗
r (Les ) is a left exact structure on
the category Espek and Fr is an ’exact’ functor from the left exact category (Espek , Lek ) to
the left exact category (Esp∗r , L∗es ).
The composition K0e of the inclusion functor

Fop
(Espak )op −−−→ (Esp∗r )op

K0∗
∗ op
and the functor (Espr ) −−−→ Z − mod assigns to each object X of the category Espak
∗
the abelian group K0 (X, Est
X ) which coincides with the Quillen’s group K0 of the exact
category (CX , EX ).
4.3. Proposition. There exists a universal ∂ ∗ -functor K•e = (Kie , dei | i ≥ 0) from
the right exact category (Espek , Le )op to the category Z − mod whose zero component is
the functor K0e . The universal ∂ ∗ -functor K•e is ’exact’; that is for any exact localization
q
(X, EX ) −→ (X 0 , EX 0 ) which belongs to Le , the canonical long sequence

K1e (q) K1e (cq ) de1 (q)

K1e (X.EX ) ←−−− K1e (X 0 , EX 0 ) ←−−− K0e (X 00 , EX 00 ) ←−−− ...
de0 (q) y (4)

K0e (cq ) K0e (q)
K0e (X 00 , EX 00 ) −−−→ K0e (X 0 , EX 0 ) −−−→ K0e (X, EX ) −−−→ 0

is exact.
F
Proof. The functor Espek −−−→ Esp∗r is an ’exact’ functor from the left exact category
(Espek , Le ) to the left exact category (Esp∗r , L∗es ) which maps the zero object of the category
Espek (– the ’space’ represented by the zero category) to a zero object of the category Esp∗r .
Therefore, F maps conflations to conflations. It remains to apply 3.3.2.

92
 Lecture 6. Comparison theorems for higher K-theory: reduction by reso-
lution, additivity, devissage. Towards some applications.

In the first four sections, we fix a left exact subcategory of the left exact category
(Esp∗r , L∗es ) of right exact ’spaces’, or a left exact subcategory of the left
exact category
Esprk of k-linear right exact ’spaces’, endowed with the induced left exact structure. The
higher K-functors are computed as satellites of the restriction of the functor K0 to this left
exact subcategory.

1. Reduction by resolution.
1.1. Proposition. Let (CX , EX ) be a right exact category with initial objects and
CY its fully exact subcategory such that
(a) If M 0 −→ M −→ M 00 is a conflation with M ∈ ObCY , then M 0 ∈ ObCY .
(b) For any M 00 ∈ ObCX , there exists a deflation M −→ M 00 with M ∈ ObCY .
Then the morphism K• (Y, EY ) −→ K• (X, EX ) is an isomorphism.
Proof. The first part of the argument of 1.1 shows that if CY is a fully exact subcat-
egory of a right exact category (CX , EX ) satisfying the condition (b) and F0 is a functor
from Espop
r to a category with filtered limits such that F0 (Y, EY ) −→ F0 (X, EX ) is an
n n
isomorphism, then S− F0 (Y, EY ) −→ S− F0 (X, EX ) is an isomorphism for all n ≥ 0.
The condition (a) is used only in the proof that K0 (Y, EY ) −→ K0 (X, EX ) is an
isomorphism.
1.2. Proposition. Let (CX , EX ) and (CZ , EZ ) be right exact categories with initial
objects and T = (Ti , di | i ≥ 0) an ’exact’ ∂ ∗ -functor from (CX , EX ) to (CZ , EZ ). Let CY
be the full subcategory of CX generated by T -acyclic objects (that is objects V such that
Ti (V ) is an initial object of CZ for i ≥ 1). Assume that for every M ∈ ObCX , there is
a deflation P −→ M with P ∈ ObCY , and that Tn (M ) is an initial object of CZ for n
sufficiently large. Then the natural map K• (Y, EY ) −→ K• (X, EX ) is an isomorphism.
Proof. The assertion is deduced from 1.1 in the usual way (see [Q]).
1.3. Proposition. Let (CX , EX ) be a right exact category with initial objects; and
let
β10 α01
0 00
Ker(f
 ) −−−→ Ker(f
 ) −−−→ Ker(f
 )
0  00
ky ky yk

β1 α1
Ker(α −−−→ −−−→ A001 (3)
 1) A
1
0  00
fy f y yf

β2 α2
Ker(α2 ) −−−→ A2 −−−→ A002
be a commutative diagram (determined by its lower right square) such that Ker(k00 ) and
Ker(β2 ) are trivial. Then
(a) The upper row of (3) is ’exact’, and the morphism β10 is the kernel of α10 .
(b) Suppose, in addition, that the arrows f 0 , α1 and α2 in (3) are deflations and
(CX , EX ) has the following property:

93
 e p
(#) If M −→ N is a deflation and M −→ M an idempotent morphism (i.e. p2 = p)
which has a kernel and such that the composition e ◦ p is a trivial morphism, then the
k(p) e
composition of the canonical morphism Ker(p) −−−→ M and M −→ N is a deflation.
Then the upper row of (3) is a conflation.
1.4. Proposition. Let (CX , EX ) be a right exact category with initial objects having
the property (#) of 1.3. Let CY be a fully exact subcategory of a right exact category
(CX , EX ) which has the following properties:
(a) If N −→ M −→ L is a conflation in (CX , EX ) and N, M are objects of CY , then
L belongs to CY too.
(b) For any deflation M −→ L with L ∈ ObCY , there exist a deflation M −→ L with
M ∈ ObCY and a morphism M −→ M such that the diagram
M
. &
M −−−→ L
commutes.
` (c) If P, M are objects of CY and P −→ x is`a morphism to initial object, then
P M exists (in CX ) and the sequence P −→ P M −→ M (where the left arrow
id
is the canonical coprojection and the right arrow corresponds to the M −→ M and the
composition of P −→ x −→ M) is a conflation.
Let CYn be the full subcategory
S of CX generated by all objects L having a CY -resolution
of the length ≤ n, and CY∞ = n≥0 CYn . Then CYn is a fully exact subcategory of (CX , EX )
for all n ≤ ∞ and the natural morphisms
∼ ∼ ∼ ∼
K• (Y, EY ) −→ K• (Y1 , EY1 ) −→ . . . −→ K• (Yn , EYn ) −→ K• (Y∞ , EY∞ )
are isomorphisms for all n ≥ 0.
1.5. Proposition. Let (CX , EX ) be a right exact category with initial objects having
the property (#) of 1.3. Let CY be a fully exact subcategory of a right exact category
(CX , EX ) satisfying the conditions (a) and (c) of 1.4. Let M 0 −→ M −→ M 00 be a
conflation in (CX , EX ), and let P 0 −→ M 0 , P 00 −→ M 00 be CY -resolutions of the length
n ≥ 1. Suppose that resolution P 00 −→ M 0 is projective. Then there exists a CY -resolution
0
` 00
P −→ M of the length n such that Pi = Pi Pi for all i ≥ 1 and the splitting ’exact’
sequence P 0 −→ P −→ P 00 is an ’exact’ sequence of complexes.

2. Additivity of ’characteristic’ filtrations.

2.1. Characteristic ’exact’ filtrations and sequences.
2.1.1. The right exact ’spaces’ (Xn , EXn ). For a right exact exact ’space’ (X, EX ),
let CXn be the category whose objects are sequences Mn −→ Mn−1 −→ . . . −→ M0 of n
morphisms of EX , n ≥ 1, and morphisms between sequences are commutative diagrams
Mn −−−→ M n−1 −−−→ ... −−−→ 
M0
fn y fn−1 y ... y f0
  

Mn0 −−−→ 0
Mn−1 −−−→ ... −−−→ M00

94
 Notice that if x is an initial object of the category CX , then x −→ . . . −→ x is an
initial object of CXn .
We denote by EXn the class of all morphisms (fi ) of the category CXn such that
fi ∈ EX for all 0 ≤ i ≤ n.
2.1.1.1. Proposition. (a) The pair (CXn , EXn ) is a right exact category.
(b) The map which assigns to each right exact ’space’ (X, EX ) the right exact ’space’
(Xn , EXn ) extends naturally to an ’exact’ endofunctor of the left exact category (Espr , Les )
of right ’exact’ ’spaces’ which induces an ’exact’ endofunctor Pn of its exact subcategory
(Esp∗r , L∗es ).
Proof. The argument is left to the reader.
2.1.2. Proposition. (Additivity of ’characteristic’ filtrations) Let (CX , EX ) and
t tn−1 t
(CY , EY ) be right exact categories with initial objects and fn∗ −→
n ∗
fn−1 −→ . . . −→ 1
f0∗ a
sequence of deflations of ’exact’ functors from (CX , EX ) to (CY , EY ) such
Xthat the functors
∗ ∗
ki = Ker(ti ) are ’exact’ for all 1 ≤ i ≤ n. Then K• (fn ) = K• (f0 ) + K• (ki ).
1≤i≤n

Proof. The argument uses facts on kernels (see Appendix A to Lecture 4).
2.1.3. Corollary. Let (CX , EX ) and (CY , EY ) be right exact categories with initial
objects and g ∗ −→ f ∗ −→ h∗ a conflation of ’exact’ functors from (CX , EX ) to (CY , EY ).
Then K• (f ) = K• (g) + K• (h).
2.1.4. Corollary. (Additivity for ’characteristic’ ’exact’ sequences) Let

f∗n −→ f∗n−1 −→ . . . −→ f∗1 −→ f∗0

be an ’exact’ sequence of ’exact’ functors from (CX , EX ) to (CY , EY ) which map initial
objects to initial objects. Suppose that f∗1 −→ f∗0 is a deflation and f∗n −→ f∗n−1 is the kernel
X
of f∗n−1 −→ f∗n−2 . Then the morphism (−1)i K• (fi ) from K• (X, EX ) to K• (Y, EY ) is
0≤i≤n
equal to zero.
Proof. The assertion follows from 2.1.3 by induction.
A more conceptual proof goes along the lines of the argument of 2.1.2. Namely, we
assign to each right exact category (CY , EY ) the right exact category (CYne , EYne ) whose
objects are ’exact’ sequences L = (Ln −→ Ln−1 −→ . . . −→ L1 −→ L0 ), where L1 −→ L0
is a deflation and Ln −→ Ln−1 is the kernel of Ln−1 −→ Ln−2 . This assignment defines
an endofunctor Pen of the category Esp∗r of right exact ’spaces’ with initial objects, and
maps L 7−→ Li determine morphisms Pen −→ IdEsp∗r . The rest of the argument is left to
the reader.

3. Infinitesimal ’spaces’. Devissage.

3.1. The Gabriel multiplication in right exact categories. Fix a right exact
category (CX , EX ) with initial objects. Let T and S be subcategories of the category CX .

95
The Gabriel product S • T is the full subcategory of CX whose objects M fit into a
g h
conflation L −→ M −→ N such that L ∈ ObS and N ∈ ObT.
3.1.1. Proposition. Let (CX , EX ) be a right exact category with initial objects. For
any subcategories A, B, and D of the category CX , there is the inclusion

A • (B • D) ⊆ (A • B) • D.

Proof. An exercise on kernels and cartesian squares.

3.1.2. Corollary. Let (CX , EX ) be an exact category. Then the Gabriel multiplica-
tion is associative.
Proof. Let A, B, and D be subcategories of CX . By 3.1.1, we have the inclusion
A • (B • D) ⊆ (A • B) • D. The opposite inclusion holds by duality, because (A • B)op =
B op • Aop .
3.2. The infinitesimal neighborhoods of a subcategory. Let (CX , EX ) be a
right exact category with initial objects. We denote by OX the full subcategory of CX
generated by all initial objects of CX . For any subcategory B of CX , we define subcategories
B (n) and B(n) , 0 ≤ n ≤ ∞, by setting B (0) = OX = B(0) , B (1) = B = B(1) , and
[
B (n) = B (n−1) • B for 2 ≤ n < ∞; and B (∞) = B (n) ;
n≥1
[
B(n) = B • B(n−1) for 2 ≤ n < ∞; and B(∞) = B(n)
n≥1

It follows that B (n) = B(n) for n ≤ 2 and, by 3.1.1, B(n) ⊆ B (n) for 3 ≤ n ≤ ∞.
We call the subcategory B (n+1) the upper nth infinitesimal neighborhood of B and the
subcategory B(n+1) the lower nth infinitesimal neighborhood of B. It follows that B (n+1)
is the strictly full subcategory of CX generated by all M ∈ ObCX such that there exists a
sequence of arrows
j1 j2 jn
M0 −−−→ M1 −−−→ . . . −−−→ Mn = M
i e
with the property: M0 ∈ ObB, and for each n ≥ i ≥ 1, there exists a deflation Mi −→ Ni
ji ei
with Ni ∈ ObB such that Mi−1 −→ Mi −→ Ni is a conflation.
Similarly, B(n+1) is a strictly full subcategory of CX generated by all M ∈ ObCX such
that there exists a sequence of deflations
en e2 e1
M = Mn −−−→ . . . −−−→ M1 −−−→ M0

such that M0 and Ker(ei ) are objects of B for 1 ≤ i ≤ n.

3.2.1. Note. It follows that B (n) ⊆ B (n+1) for all n ≥ 0, if B contains an initial
object of the category CX .

96
 3.3. Fully exact subcategories of a right exact category. Fix a right exact
category (CX , EX ). A subcategory A of CX is a fully exact subcategory of (CX , EX ) if
A • A = A.
3.3.1. Proposition. Let (CX , EX ) be a right exact category with initial objects. For
any subcategory B of CX , the subcategory B (∞) is the smallest fully exact subcategory of
(CX , EX ) containing B.
Proof. Let A be a fully exact subcategory of the right exact category (CX , EX ), i.e.
A = A • A. Then B (∞) ⊆ A, iff B is a subcategory of A.
On the other hand, it follows from 3.1.1 and the definition of the subcategories B (n)
(see 3.2) that B (n) • B (m) ⊆ B (m+n) for any nonnegative integers n and m. In particular,
B (∞) = B (∞) • B (∞) , that is B (∞) is a fully exact subcategory of (CX , EX ) containing B.
3.4. Cofiltrations. Fix a right exact category (CX , EX ) with initial objects. A
cofiltration of the length n+1 of an object M is a sequence of deflations
en e2 e1
M = Mn −−−→ . . . −−−→ M1 −−−→ M0 . (1)

The cofiltration (1) is said to be equivalent to a cofiltration

een ee2 ee1

fm −−−→
M =M . . . −−−→ Mf1 −−−→ M
f0

if m = n and there exists a permutation σ of {0, . . . , n} such that Ker(ei ) ' Ker(eeσ(i) )
for 1 ≤ i ≤ n and M0 ' M f0 .
The following assertion is a version (and a generalization) of Zassenhouse’s lemma.
3.4.1. Proposition. Let (CX , EX ) have the following property:
t1 t2
(‡) for any pair of deflations M1 ←− M −→ M2 , there is a commutative square
t1
M
 −−−→ M
1
t2 y y p2
 
p1
M2 −−−→ M3
of deflations such that the unique morphism M −→ M1 ×M3 M2 is a deflation.
Then any two cofiltrations of an object M have equivalent refinements.
3.5. Devissage.
3.5.1. Proposition. (Devissage for K0 .) Let ((X, EX ), Y ) be an infinitesimal
’space’ such that (X, EX ) has the following property (which appeared in 3.4.1):
t1 t2
(‡) for any pair of deflations M1 ←− M −→ M2 , there is a commutative square
t1
M
 −−−→ M
1
t2 y y p2
 
p1
M2 −−−→ M3

97
of deflations such that the unique morphism M −→ M1 ×M3 M2 is a deflation.
Then the natural morphism

K0 (Y, EY ) −−−→ K0 (X, EX ) (1)

is an isomorphism.
3.5.2. The ∂ ∗ -functor K•sq . Let Les
o denote the left exact structure on the category
Espo of Espr (cf. 3.9.4) induced by the (defined in 6.8.3.3) left exact structure Les sq on
sq
the category Espr of right exact ’spaces’. Let Ki (X, EX ) denote the i-th satellite of the
functor K0 with respect to the left exact structure Les
o.

3.5.3. Proposition. Let ((X, EX ), Y ) be an infinitesimal ’space’ such that the right
exact ’space’ (X, EX ) has the property (‡) of 3.4.1, the category CX has final objects, and
all morphisms to final objects are deflations. Then the natural morphism

Kisq (Y, EY ) −−−→ Kisq (X, EX ) (8)

is an isomorphism for all i ≥ 0.

Proof. The assertion follows from a general devissage theorem for universal ∂ ∗ -functors
whose zero component satisfy devissage property (like K0 , by 3.5.1).

4. An application: K-groups of ’spaces’ with Gabriel-Krull dimension.

4.1. Gabriel-Krull filtration. We recall the notion of the Gabriel filtration of
an abelian category as it is presented in [R, 6.6]. Let CX be an abelian category. The
Gabriel filtration of X assigns to every cardinal α a Serre subcategory CXα of CX which
is constructed as folows:
Set CX0 = O.
If α is not a limit cardinal, then CXα is the smallest Serre subcategory of CX containing
∗
all objects M such that the localization qα−1 (M ) of M at CXα−1 has a finite length.
If β is a limit cardinal, then CXβ is the smallest Serre subcategory containing all
subcategories CXα for α < β.
Let CXω denote the smallest Serre subcategory containing all the subcategories CXα .
Clearly the quotient category CX /CXω has no simple objects.
An object M is said to have the Gabriel-Krull dimension β, if β is the smallest cardinal
such that M belongs to CXβ .
The ’space’ X has a Gabriel-Krull dimension if X = Xω .
Every locally noetherian abelian category (e.g. the category of quasi-coherent sheaves
on a noetherian scheme, or the category of left modules over a left noetherian associative
algebra) has a Gabriel-Krull dimension. [
It follows that for any limit ordinal β, we have K• (Xβ ) = K• (Xα ). Therefore,
α<β
[
K• (Xω ) = K• (Xα ), where Orn denotes the set of non-limit ordinals.
α∈Orn

98
 4.2. Reduction via localization. If α is a non-limit ordinal, we have the exact
∗
qα−1
localization CXα −−−→ CXα /CXα−1 = CXαq , hence the corresponding long exact sequence

dα
. . . −→ Kn+1 (Xαq ) −→
n
Kn (Xα−1 ) −→ Kn (Xα ) −→ Kn (Xαq ) −→ . . . −→ K0 (Xαq ) (1)

of K-groups.
4.3. Reduction by devissage. Suppose that the category CX is noetherian, i.e.
all objects of CX are noetherian. Then the quotient category CXαq = CXα /CXα−1 is
noetherian. Notice that the Krull dimension of Xαq equals to zero; hence all objects of the
category CXαq have a finite length. Let CXα,s
q denote the full subcategory of CXαq generated
q
by semisimple objects. By devissage, the natural morphism K• (Xα,s ) −→ K• (Xαq ) is an
isomorphism. If CY is a a svelte abelian category whose objects are semisimple of finite
length, then K• (Y ) = K• (Sp(DQ )), where DQ is the residue skew field of the
Q∈Spec(Y )
point Q of the spectrum of Y , which is the skew field CY (M, M )o of the endomorphisms
of the simple object M such that Q = [M ]. In particular,
a
K• (Xαq ) = K• (Sp(DQ ))
q
Q∈Spec(Xα )

for every non-limit ordinal α.

5. First definitions of K-theory and G-theory of noncommutative schemes.

The purpose of this section is to sketch the first notions which allow extension of K-
theory and G-theory to noncommutative schemes and more general locally affine ’spaces’.
We consider here only the class of so-called semiseparated locally affine ’spaces’ and schemes
which includes the main examples of noncommutative schemes and locally affine ’spaces’,
starting from quantum flag varieties and noncommutative Grassmannians. Commutative
∆X
semiseparated schemes are schemes X whose diagonal moprhism X −→ X × X is affine.
In particular, every separated scheme is semiseparated.
Semiseparated noncommutative (in particular, commutative) schemes and locally affine
’spaces’ over an affine scheme are particularly convenient, because the category of quasi-
coherent sheaves on them is described by a linear algebra data provided by flat descent.
5.1. Semiseparated schemes. Flat descent. We shall consider semiseparated
schemes and more general locally affine ’spaces’ over an affine scheme, S = Sp(R). These
are pairs (X, f ), where X is a ’space’ and f a continuous morphism X −→ S for which
ui
there exists a finite affine cover {Ui −→ X | i ∈ J} such that every morphism ui is flat
and affine. In this case, the corresponding morphism

u
a
UJ = Ui −→ X
i∈J

99
is fflat (=faithfully flat) and affine. By the (dual version of) Beck’s theorem (Lecture 1,
2.3.1), there is a commutative diagram

CX −−−→ Gu − Comod
u∗ & .bu∗
Rf u − mod

where the horizontal arrow is a category equivalence. Here we identify the category CUJ
with the category of Rf u -modules for a ring Rf u over R corresponding (by Beck’s theorem)
fu
to the affine morphism UJ −→ Sp(R) – the monad Ff u on R − mod is isomorphic to
the monad Rf u ⊗R − (see Lecture 1). Since the morphism u is affine, the associated
comonad Gu = (Gu , δu ), that is the functor Gu = u∗ u∗ , is continuous: the composition
u! u∗ is its right adjoint. Therefore, Gu is isomorphic to the tensoring Mu ⊗Rf u − by an
Rf u -bimodule Mu determined uniquely up to isomorphism. The comonad structure δu
δu
e
induces a map M −−−→ Mu ⊗Rf u Mu which turnes Mu into a coalgebra in the monoidal
category of Rf u -bimodules. Thus, the category CX is naturally equivalent to the category
ζ
(Mu , δeu ) − Comod of (Mu , δeu ). Its objects are pairs (V, V −→ Mu ⊗Rf u V ), where V is
a left Rf u -module, which satisfy the usual comodule conditions. The structure morphism
f
X −→ Sp(R) is encoded in the structure object O = f ∗ (R), or, what is the same, a
ζf u
comodule structure Rf u −→ Mu ⊗Rf u Rf u on the left module Rf u , which we can replace,
ζf u
thanks to an isomorphism Mu ⊗Rf u Rf u ' Mu , by a morphism Rf u −→ Mu satisfying
εu
the natural associativity condition and whose composition with counit M −→ Rf u of the
coalgebra (Mu , δeu ) is the identical morphism.
Thus, Beck’s theorem provides a description of the category of quasi-coherent sheaves
on a semiseparated noncommutative (that is not necessarily commutative) scheme in terms
of linear algebra.

5.2. The category of vector bundles. Fix a locally affine ’space’ (X, f ). We call
an object M of the category CX a vector bundle if its inverse image, u∗J (M) is a projective
ΓUJ -module of finite type, or, equivalently, u∗i (M) is a projective ΓUi -module of finite
type for each i ∈ J. We denote by P(X) the full subcategory of the category CX whose
objects are vector bundles on X.

5.3. The category of coherent objects. We call an object M of the category CX

coherent if u∗i (M) is coherent for each i ∈ J. We denote by Coh(X) the full subcategory
of CX generated by coherent objects.

5.3.1. Proposition. (a) The notions of a projective and coherent objects are well
defined.
(b) Coh(X) is a thick subcategory of CX . In particular, it is an abelian category.
(c) P(X) an fully exact (i.e. closed under extensions) subcategory of CX . In particu-
lar, P(X) is an exact category.

100
 uJ uI e
Proof. (a) Semiseparated finite covers form a filtered system: if UJ −→ X ←− UI are
fflat and affine, then all arrows in the cartesian square

UJ ×X UeI −−−→ Ue

I
cart
 
y y
UJ −−−→ X
are fflat and affine. This follows from the categorical description of the cartesian product
corresponding to direct image functors of UJ −→ X and Ue −→ X.
(b) & (c). An exercise for the reader.
5.4. The category of locally affine semiseparated ’spaces’. Let LaffS denote
the subdiagram of the category |Cat|oS of S-’spaces’ whose objects are locally affine quasi-
f
compact semiseparated S-’spaces’ and morphisms are those morphisms X −→ Y of S-
’spaces’ which can be lifted to a morphism of semiseparated covers. More precisely, for
f πY
any morphism X −→ Y of LaffS and any affine cover UY −→ Y , there is a commutative
diagram
fe
UX −−−→ U Y
πX y y πY
 
f
X −−−→ Y
where the left vertical arrow is an affine cover of X.
One can see that LaffS is a subcategory of |Cat|oS .
For each object (X, f ) of LaffS , let XP denote the ’space’ defined by CXP = P(X)
and XC the ’space’ defined by CXC = Coh(X, f ).
5.5. Proposition. The map (X, f ) 7−→ XP is a functor from LaffS to the category
Espx whose objects are ’spaces’ represented by exact categories and whose morphisms have
’exact’ inverse image functors.
Proof. In fact, restricted to the affine schemes, the functor takes values in the category
Ex, because an inverse image of (automatically affine) morphism between affine S-’spaces’
maps conflations to conflations. The general case follows from the affine case via affine
covers, because the inverse image functors of the covers are fflat.
5.6. The functor K• . We define the K-theory functor K• as the universal ∂-functor
from the category LaffS semiseparated locally affine ’spaces’ endowed with left exact struc-
ture induced by the functor from LaffS to the category of right exact ’spaces’ which assigns
to every locally affine semiseparated ’space’ the right exact ’space’ represented by the cat-
egory of vector bundles.
5.7. The category Lafffl S . We denote this way the subcategory of the category LaffS
of locally affine ’spaces’ formed by flat morphisms.
5.7.1. Proposition. The map (X, f ) 7−→ XP is a functor from LaffflS to the category
Espa whose objects are ’spaces’ represented by abelian categories and whose morphisms
have exact inverse image functors.

101
 Proof. An exercise for the reader.
5.8. The functor G• . We endow the category Lafffl S with the left exact structure IS )
e
induced by the standard left exact structure on Espa (inverse image functors of inflations
are exact localizations) via the functor of 5.7.1. We define the ∂-functor G• as the universal
∂-functor from the left exact category (Lafffl S , IS ), whose zero component assigns to every
e
locally affine semiseparated ’space’ (X, f ) the K0 -group of the ’space’ represented by the
category of coherent sheaves on (X, f ).
5.9. Proposition. Let i 7−→ (Xi , fi ) be a filtered projective system of locally affine
S-’spaces’ such that the transition morphisms (Xi , fi ) −→ (Xj , fj ) are affine, and let
(X, f ) = lim(Xi , fi ). Then

K• (X, f ) ' colim(K• (Xi , fi )). (2)

If in addition the transition morphisms are flat, then

G• (X, f ) ' colim(G• (Xi , fi )). (20 )

Proof. It follows from the assumptions that a filtered projective system of locally affine
S-’spaces’ and affine morphisms induces a filtered inductive system of the exact categories
P(Xi , fi ) of vector-bundles. Its colimit, P(X, f ) is an exact category whose conflations are
images of conflations of P(Xi , fi ). Whence the isomorphism (2).
If, in addition, the transition morphisms are flat, then the inverse image functors of
the transition functors induce exact functors between categories of coherent objects. This
implies the isomorphism (2’).
5.10. Regular locally affine ’spaces’. For a locally affine S-’space’ (X, f ), we
denote by H(X, f ) the full subcategory of the category Coh(X, f ) which have a P(X)-
resolution.
5.10.1. Proposition. (a) H(X, f ) is a fully exact subcategory of the category
Coh(X, f ). In particular, it is an exact category.
(b) Set H(X, f ) = CXH . The embedding of categories P(X, f ) ,→ H(X, f ) induces an
def ∼
isomorphism K• (X, f ) = K• (XP ) −→ K• (XH ).
Proof. (a) By a standard argument.
(b) The fact is a consequence of the Resolution Theorem.
5.10.2. Definition. A locally affine ’space’ is called regular if H(X, f ) = Coh(X, f ).
Thus, if (X, f ) is a regular locally affine ’space’, then K• (X, f ) = G• (X, f ).
5.10.3. Remark. If (X, f ) is an affine S-’space’, then the regularity coincides with
the usual notion of regularity of rings (S is assumed to be affine). Similarly, if (X, f ) is an
S-’space’ corresponding to a commutative scheme.
The notion of H(X, f ) is local in the following sense:

102
 5.10.4. Proposition. Let (X, f ) be a locally affine S-’space’. The following condi-
tions on an object M of CX are equivalent:
(a) M belongs to H(X, f );
ui
(b) u∗i (M) belongs to H(Ui , f ui ) for some finite cover {Ui −→ X | i ∈ J} of (X, f )
and for all i ∈ J;
ui
(c) u∗i (M) belongs to H(Ui , f ui ) for some finite cover {Ui −→ X | i ∈ J} of (X, f )
and for all i ∈ J.
Proof. Obviously, (c) ⇒ (b). In the rest of the argument, one can assume that the
covers consist of one fflat affine morphism. The assertion follows from the fact that such
covers form a filtered system. Details are left as an exercise.
5.10.5. Examples. The quantum flag varieties and the corresponding twisted quan-
tum D-schemes [LR] are examples of regular schemes. Noncommutative Grassmannians
[KR1], [KR3] are examples of regular locally affine ’spaces’ which are not schemes.

6. Remarks on K-theory and quantized enveloping algebras.

In a sense, the standard K-theory based on the category of vector bundles, or G-theory
based on the category of all coherent sheaves, do not give much valuable information from
the point of view of representation theory. For instance, if g is a finite-dimensional Lie
algebra over a field k, then K• (U (g)) ' K• (k) and, similarly, K• (An (k)) ' K• (k), where
An (k) is the n-th Weyl algebra over k. This indicates that one should study K-theory
of other subcategories of the category of U (g)-modules. The subcategory which received
most of attention in seventies and the beginning of eighties was the category O = O(g) of
representations of a semi-simple (or reductive) Lie algebra g introduced by I.M. Gelfand
and his collaborators. The highlight of its study was Kazhdan-Lusztig conjecture and,
the most important, its prove, which led to the reformulation of the representation theory
of reductive algebraic groups in terms of D-modules and D-schemes making it a part of
noncommutative algebraic geometry, even before this branch of mathematics emerged.
The main basic fact which allowed to reduce the problems of representation theory
to the study of D-modules on flag varieties is the Beilinson-Bernstein localization theorem
which says that the global section functor induces an equivalence between the category of
D-modules on the flag variety of a reductive Lie algebra g over a field of zero characteristic
and the category of U (g)-modules with trivial central character (and its twisted version).
Harish-Chandra modules and their different generalizations turned out to be holonomic
D-modules. As a result, holonomic modules on flag varieties became the main object of
study in representation theory of reductive algebraic groups.
On the other hand, the notions of quantum flag variety and the appropriate categories
of twisted D-modules were introduced in [LR]. And it was established a quantum version
of Beilinson-Bernstein localization theorem [LR], [T], which reduces the study of represen-
tations of the quantized enveloping algebra Uq (g) to the study of twisted D-modules on
quantum flag variety, like in the classical case. The notion of a holonomic D-module is
extended to the setting of noncommutative algebraic geometry [R5]. In particular, there
exists a notion of a quantum holonomic D-module.
All initial ingredients are present and the area of research is wide open.

103
 References
[Ba] H. Bass, Algebraic K-theory, Benjamin 1968
[Gr] A. Grothendieck, Sur quelques points d’algèbre homologique. Tohuku Math. J., Ser.
II, 9, 120–221 (1957)
[GZ] P. Gabriel and M. Zisman, Calculus of fractions and homotopy theory, Springer
Verlag, Berlin-Heidelberg-New York, 1967
[Ke1] B. Keller, Derived categories and their uses, preprint, 1994.
[KeV] B. Keller, D. Vossieck, Sous les catégories dérivées, C. R. Acad. Sci. Paris 305
(1987), 225–228.
[KR1] M. Kontsevich, A.L. Rosenberg, Noncommutative smooth spaces, in “The Gelfand
Mathematical Seminar 1996–1999” (2000), 87–109.
[KR2] M. Kontsevich, A.L. Rosenberg, Noncommutative spaces and flat descent, preprint
MPI, 2004(36), 108 pp.
[KR3] M. Kontsevich, A.L. Rosenberg, Noncommutative spaces, preprint MPI, 2004(35),
79 pp.
[KR4] M. Kontsevich, A. Rosenberg, Noncommutative stacks,
preprint MPI, 2004(37), 55 pp.
[KR5] M. Kontsevich, A. Rosenberg, Noncommutative Grassmannians and related con-
structions, to appear.
[KR6] M. Kontsevich, A. Rosenberg, Orthogonals, in preparation.
[KR7] M. Kontsevich, A. Rosenberg, Noncommutative ’spaces’ and stacks, a monograph
in preparation.
[LR] V. Lunts, A.L. Rosenberg, Localization for quantum groups, Selecta Mathematica,
New Series, 5 (1999), 123–159.
[ML] S. Mac-Lane, Categories for the working mathematicians, Springer - Verlag; New
York - Heidelberg - Berlin (1971)
[MLM] S. Mac-Lane, L. Moerdijk, Sheaves in Geometry and Logic, Springer - Verlag; New
York - Heidelberg - Berlin (1992)
[Q] D. Quillen, Higher algebraic K-theory, LNM 341, 85–147 (1973)
[R] A.L. Rosenberg, Noncommutative algebraic geometry and representations of quantized
algebras, Kluwer Academic Publishers, Mathematics and Its Applications, v.330 (1995),
328 pages.
[R1] A.L. Rosenberg, Noncommutative local algebra, Geometric and Functional Analysis
(GAFA), v.4, no.5 (1994), 545-585
[R2] A.L. Rosenberg, Noncommutative schemes, Compositio Mathematica 112 (1998), 93-
125
[R3] A.L. Rosenberg, Noncommutative spaces and schemes, preprint MPIM, 1999
[R4] A. L. Rosenberg, Spectra of ’spaces’ represented by abelian categories, preprint MPIM,
2004(115), 73 pp.
[R5] A. L. Rosenberg, Spectra, associated points, and representations, preprint MPIM,
2007(10), 71 pp.
[R6] A.L. Rosenberg, Spectra of noncommutative spaces, preprint MPIM, 2003(110)
[R7] A.L. Rosenberg, Spectra related with localizations, preprint MPIM, 2003(112), 77 pp
[R8] A.L. Rosenberg, Homological algebra of noncommutative ’spaces’ I, to appear.

104
[R9] A.L. Rosenberg, Trieste Lectures, to appear in Lecture Notes of the School on Alge-
braic K-Theory and Applications, ICTP, Italy, 92 pp.
[R10] A.L. Rosenberg, Geometry of right exact ’spaces’, to appear.
[R11] A.L. Rosenberg, Geometry of noncommutative ’spaces’ and schemes, to appear.
[S] J.-P. Serre, Faisceaux algébriques cohérents, Annals of Math.62, 1955
[T] Tanisaki, T., The Beilinson-Bernstein correspondence for quantized enveloping alge-
bras, arXiv. Math. QA/0309349, v1(2003).
[Ve1] J.–L. Verdier, Catégories dérivées, Séminaire de Géométrie Algébrique 4 1/2, Coho-
mologie Étale, LNM 569, pp. 262–311, Springer, 1977
[Ve2] J.–L. Verdier, Des catégories dérivées des catégories abéliennes, Astérisque, v.239,
1996

105