Exposé II. Tangent bundles, Lie algebras
by M. Demazure
1 We propose in this Exposé to construct the analogue, in the theory of schemes, of the tangent bundles and
Lie algebras of the classical theory. It will however be useful not to restrict ourselves to schemes properly speaking,
but also to consider certain functors on the category of schemes which are not necessarily representable (for example
the functors Hom, Norm, etc.). As announced in the preceding Exposé (cf. I 1.1), we shall identify a scheme with the
functor associated to it.
On the other hand, the constructions presented below go beyond the framework of the theory of schemes. They are equally valid, for example, in the theory of analytic spaces with nilpotent elements, modulo a few modifications of detail.
Before beginning this construction, we must lay down a few general definitions that complement those of I 1.7.
1. The functors
Let us resume the notations of I 1.1. We identify the category with a full subcategory of
(in particular we suppress the underlines2 that allowed us
to distinguish graphically an object of Ĉ from an object of ).
Consider the following situation: four objects of Ĉ, denoted S, X, Y, Z, the first being in fact an object of ,
with and above , and above :
X Y
╲ ╱
pX ╲ ╱ pY
╲ ╱
╲ ╱
Z
│
↓
S .
Definition 1.1. We define an object of , denoted , by:
(1) Hom_{Z/S}(X, Y)(S′) = Hom_{Z_{S′}}(X_{S′}, Y_{S′}) = Hom_Z(X ×_S S′, Y),
for every object of . One sees at once that is nothing other than the
subobject of formed by the morphisms compatible with pX and pY, that is to say, the
kernel of the pair of morphisms
Hom_S(X, Y) ⇒ Hom_S(X, Z)
defined: the first by composition with pY; the second as the constant morphism whose "image" is pX.
On the other hand, one sees as in I 1.7 that, for every object of Ĉ above , one has a natural
bijection:3
(2) Hom_S(T, Hom_{Z/S}(X, Y)) ≃ Hom_Z(X ×_S T, Y).
Moreover, by I 1.7.1, if E, F are objects of Ĉ above , one has:
Hom_Z(E, Hom_Z(F, Y)) ≃ Hom_Z(E ×_Z F, Y) ≃ Hom_Z(F, Hom_Z(E, Y)).
Applying this with and , one obtains natural bijections, for every object of :
(3) Hom_S(T, Hom_{Z/S}(X, Y)) ≃ Hom_Z(X ×_S T, Y) ≃ ⎰ Hom_Z(Z ×_S T, Hom_Z(X, Y))
⎱ Hom_Z(X, Hom_Z(Z ×_S T, Y)).
Moreover, these bijections are functorial in , so one obtains isomorphisms of -functors:
(4) Hom_S(T, Hom_{Z/S}(X, Y)) ⥲ Hom_{Z/S}(X, Hom_Z(Z ×_S T, Y))
≃ Hom_{Z/S}(X ×_S T, Y).
Let us point out two particular cases of the definition. If , one has:
Hom_{S/S}(X, Y) = Hom_S(X, Y).
On the other hand, when , one sets
(5) ∏_{Z/S} Y = Hom_{Z/S}(Z, Y),
so that by definition
(∏_{Z/S} Y)(S′) = Hom_Z(Z ×_S S′, Y) ≃ Γ(Y_{S′}/Z_{S′}).
The functor is right adjoint to the base-change functor from to : for every -functor one has
Hom_S(U, ∏_{Z/S} Y) = Hom_Z(U ×_S Z, Y).
(If and if is an -scheme, the functor is called "Weil restriction of scalars".)4 Let us note also that one has an isomorphism:
(6) Hom_{Z/S}(X, Y) ≃ Hom_{X/S}(X, Y ×_Z X) = ∏_{X/S}(Y ×_Z X),
which gives in particular, for , an isomorphism:
(7) Hom_S(X, Y) ≃ ∏_{X/S} Y_X.
Remark 1.2. The functor commutes with products in the following sense: one has a functorial isomorphism
(∗) Hom_{Z/S}(X, Y ×_Z Y′) ≃ Hom_{Z/S}(X, Y) ×_S Hom_{Z/S}(X, Y′).
It follows that if is a -group, resp. a -ring, etc., then is an -group, resp. an -ring, etc.
Remark 1.3.5 Moreover, let be a -functor of O_Y-modules (cf. I, 4.3.3.1). Set
. Then is equipped with a natural structure of
O_H-module; more precisely, for every above , is
equipped with a natural structure of -module.
Indeed, denote by and the morphisms defining the
structures of -(abelian) group and of O_Y-module. Let be an -scheme above
, i.e. one has been given a -morphism , which
therefore makes a -object. Then,
is the set of -morphisms such that , i.e. of -morphisms .
If are two such morphisms, one defines as the -morphism composite
X ×_S H′ ──(φ × ψ)──→ M ×_Y M ──m──→ M
and one verifies that this equips with a structure of abelian group above .6
Likewise, if is an element of , i.e. an -morphism , one defines as the composite , where denotes the -morphism from to with components and ; one verifies that this equips with a structure of -module, functorial in the -object .
2. The schemes
Definition 2.1. Let be a scheme and a quasi-coherent O_S-module. We denote by the
quasi-coherent O_S-algebra (where is considered as an ideal of square zero). We denote by
the -scheme .7
In particular we set , , and we call them respectively the algebra of dual numbers over and the scheme of dual numbers over .
Then is a contravariant functor from the category of quasi-coherent O_S-modules to that of
-schemes. In particular the morphisms and define respectively the structural morphism
and a section of it, which we call the zero section.8
2.1.1.9 Since is a contravariant functor, every defines an -endomorphism of , and one has , , and . Consequently, the -scheme is equipped with a right action of the multiplicative monoid of , which commutes with the -morphisms arising from morphisms ; in particular, the operators preserve the zero section of .
For every and and , we shall write ; then , , and, if , then .10
Remark 2.1.2. The formation of the commutes with base extension: one has canonical isomorphisms
I_S(M)_{S′} ≃ I_{S′}(M ⊗_{O_S} O_{S′}).
To simplify, we shall write ; more generally, if is an -functor (not necessarily representable), we shall write .
2.1.3.11 By what precedes, the multiplicative monoid of acts on the -scheme ,
functorially in , i.e. the -scheme is equipped with a structure of object with monoid of operators
O_S, this structure being functorial in . One thus has a morphism of -schemes
λ : I_S(M) ×_S O_S ⟶ I_S(M),
verifying obvious conditions. For every -functor , one obtains by base change a morphism of -functors:
λ_X : I_X(M) ×_S O_S ⟶ I_X(M)
which makes the -functor into an object with monoid of operators : every element of defines an -endomorphism of ; explicitly, if and , then belongs to and one has:
(m, x) · a = (m · a(x), x).
This operation is functorial in and preserves the zero section , i.e. for every .
Moreover, this operation is "functorial in " in the following sense: if is a morphism of -functors and the corresponding ring morphism (i.e. for every ), then the following diagram is commutative:
a^*
I_X(M) ─────────→ I_X(M)
↑ ↑
π π
u(a)^*
I_Y(M) ─────────→ I_Y(M) .
2.2. Let now and be two quasi-coherent O_S-modules. The commutative diagram
M ⊕ N
↗ ↑ ↘
↗ │ ↘
M │ N
↘ ↓ ↙
↘ │ ↙
0
defines a commutative diagram of -schemes
I_S(M ⊕ N)
↗ ↑ ↖
↗ │ ↖
↗ ε_{M⊕N} ↖
I_S(M) │ I_S(N)
(∗) ↖ │ ↗
↖ │ ↗
ε_M │ ε_N
↘ │ ↙
S .
Proposition 2.2. For every -scheme , the diagram of functors over obtained by applying the functor to diagram (∗) is cartesian:
Hom_S(I_S(M ⊕ N), X)
↙ ↘
Hom_S(I_S(M), X) Hom_S(I_S(N), X)
↘ ↙
Hom_S(S, X) = X .
It must be verified that for every , the diagram of sets obtained by taking the value of the functors on is cartesian. Since the formation of commutes with base extension in the sense explained above, it suffices to do it for , hence to verify that the following diagram of sets is cartesian:
X(I_S(M ⊕ N))
↙ ↘
X(I_S(M)) X(I_S(N))
↘ X(ε_{M⊕N})↙
X(ε_M) ↓ X(ε_N)
X(S) .
Now, if , it follows from SGA 1, III 5.112, that is isomorphic, functorially in , to
where denotes the sheaf of relative differentials of with respect to . Now this latter functor
(in ) evidently transforms a direct sum of O_S-modules into the product of the corresponding sets, whence the
result.
Corollary 2.2.1. Let be an -scheme and a free O_S-module of finite type. The functor
is isomorphic (as functor above ) to a finite product (above ) of copies of
.
Nota 2.2.2. It follows from the proof of the proposition that is isomorphic as -functor to (I 4.6.1), hence representable by the vector fibration13 .
3. The tangent bundle, condition (E)
In this section, unless otherwise stated, the letters , etc., will always denote free O_S-modules of finite
type (that is to say, isomorphic to a finite direct sum of copies of O_S).
We shall systematically use the identifications justified in Exposé I; thus we shall say "functor above " to designate indifferently a functor equipped with a morphism into (= ) or a functor on the category of objects above . Likewise we shall say "group-functor above ", etc.
Definition 3.1. Let be a scheme and a free O_S-module of finite type. Let be a functor above . One
calls tangent bundle to above relative to the O_S-module and denotes by the
-functor
In particular, one calls tangent bundle to above and denotes by the functor14
T_{X/S} = T_{X/S}(O_S) = Hom_S(I_S, X).
Then is a covariant functor from the category of free O_S-modules of finite type to the
category of -functors. In particular the morphisms15 and define respectively an
-morphism
π_M : T_{X/S}(M) ⟶ T_{X/S}(0) ≃ X
and a section of it, called the zero section (or null section).
Remark 3.1.1.16 One should note that the projection is induced by the zero section , while the null section is induced by the structural morphism ; i.e., for every point (resp. ), corresponding to an -morphism (resp. ), one has (resp. ).
It follows from 3.1 that is a covariant functor from the category of free O_S-modules of finite
type to that of functors above . In particular is a monoid of operators of the -functor which
respects "functoriality in ".
Scholie 3.1.2.16 What precedes means, in particular, the following. For every -morphism , set
Σ(X′, M) = Hom_X(X′, T_{X/S}(M)).
One has an action of the multiplicative monoid on , denoted , such that , , and , where is the null section . One has likewise an action of on .17
Moreover, let (resp. ) be the addition (resp. the diagonal map) of , and denote by and the morphisms induced by and . For , denote by (resp. ) multiplication by in (resp. by in ). Since and , one has, for every and :
(†) λ ∗ m(z) = m((λ, λ) ∗ z), m((λ, μ) ∗ δ(x)) = (λ + μ) ∗ x.
Definition 3.2. Let . One calls tangent space to above at the point relative to , and denotes by , the -functor obtained from the -functor by pullback along the morphism :
L^u_{X/S}(M) ──────→ T_{X/S}(M)
│ │
│ │ π
↓ u ↓
S ───────────────→ X .
In particular is denoted . It is the tangent space to above at the point .
Remark 3.2.1.18 It follows from 3.1.1 that, for every , is the set of -morphisms such that , where is the zero section.
Let us note immediately the
Proposition 3.3. If is representable by an -scheme denoted , then and are representable. In particular and are representable by vector fibrations on and on , which are respectively and .
It clearly suffices to prove the proposition for ; the analogous results for follow by pullback. By Corollary 2.2.1, it even suffices to do it for , and in that case, the proposition is nothing other than the remark noted in 2.2.2.
Remark 3.3.1. From this proposition follows a particularly simple description of the vector fibration representing : the image of the section of over is locally closed19, hence defined by a quasi-coherent ideal of a scheme induced on an open subset of . The quotient can be considered as a quasi-coherent module on . It is this module that defines the sought-after vector fibration.
Take for example an algebraic scheme over a field and a -rational point of . Let be the maximal ideal of the local ring and let be the -vector space dual to ; it is the Zariski tangent space of at the point . Then, with the notations of I 4.6.5.1, one has:
Having closed this parenthesis, let us return to the general situation. Let us remark first that is a
covariant functor from the category of free O_S-modules of finite type to that of functors above . In particular
is a set of operators of the -functor which respects functoriality in .20
Proposition 3.4. The formation of and commutes with base extension: for every -scheme , one has isomorphisms functorial in
T_{X_{S′}/S′}(M ⊗ O_{S′}) ⥲ T_{X/S}(M)_{S′},
L^{u′}_{X_{S′}/S′}(M ⊗ O_{S′}) ⥲ L^u_{X/S}(M)_{S′}, where u′ = u_{S′}.
This results immediately from the fact that Hom commutes with base extension.
Corollary 3.4.1. The -functor (resp. the -functor ) is naturally equipped with a
structure of object with operators O_X (resp. O_S), this structure being functorial in .
Let us show this first for . For each above , acts on hence on
; now one verifies that this action is functorial in .
It thus equips , as announced, with a structure of functor with operators O_S.
For it is a little more complicated. For each above , set ; one must equip with a structure of set with monoid of operators in a manner functorial in . For this one constructs the following diagram, where denotes and denotes the section of over defined by :
T_{X_{X′}/X′}(M)
↙ ↘
T_{X/S}(M) ←───── T_{X/S}(M)_{X′}
↓ ↓
X_{X′}
↙ ↘ f′
↙ ↘
X ←──── f X′
↘ ↙
S .
21 This diagram, joined with 3.2.1, shows that is identified with
L^{f′}_{X_{X′}/X′}(M)(X′) = { X′-morphisms ψ : I_{X′}(M) → X_{X′} such that ψ ∘ ε_M = f′ },
on which every acts via its action on , i.e., with the notations of 2.1.1, one has: , i.e. for every and , . One then verifies easily that this construction is functorial in .
The isomorphisms of Proposition 3.4 are then, by construction, isomorphisms for the structures of -objects, resp. -objects.
Remark 3.4.2.22 The action of O_X on can be seen, more simply, as follows. For every
, one has
Hom_X(X′, T_{X/S}(M)) = { φ ∈ Hom_S(I_{X′}(M), X) | φ ∘ ε_M = f },
and one saw in 2.1.3 that , considered as -functor, is equipped with an action of the monoid which preserves the zero section . Consequently, if one denotes by the -endomorphism of defined by , one has: , i.e., for every and ,
(aφ)(m, x′) = φ(m · a(x′), x′)
(note that , whence ).
Likewise, the action of O_S on can be described as follows. For every ,
is the set of -morphisms such that
; for such a and , one has: .
Remark 3.4.3.22 The observations of 3.1.2 hold equally for the action of O_S on and
that of O_X on .
Definition 3.5. Let be a scheme and an -functor. One says that satisfies condition (E) relative to if, for every above and all free -modules and of finite type, the diagram of sets
X(I_{S′}(M ⊕ N))
↙ ↘
X(I_{S′}(M)) X(I_{S′}(N))
↘ ↙
X(S′) ,
obtained by applying to the diagram (∗) defined in 2.1, is cartesian.23
3.5.1.24 It comes to the same thing to say that the functor transforms direct
sums of free O_S-modules of finite type into products of -functors; in this case, the same holds for the functor
, for every .
Proposition 2.2 shows that every representable functor satisfies condition (E).
Abbreviation: instead of saying "X satisfies condition (E) with respect to ", we shall sometimes say " satisfies condition (E)".
If satisfies condition (E), the functor commutes with products and therefore transforms groups into groups. In particular is a commutative -group. For the same reason, is a commutative -group.
Proposition 3.6. If satisfies (E), the abelian group structure on (resp. ) and
the action of O_X (resp. O_S) equip (resp. ) with a structure of O_X-module (resp.
O_S-module).
The action of O_X (resp. O_S) is functorial in ; it therefore respects the abelian group structure that is
deduced by functoriality from that of .25 Indeed, let us resume the notations of 3.1.2. The structure of
(abelian) -group on is defined by the composite:
T_{X/S}(M) ×_X T_{X/S}(M) ≃ T_{X/S}(M ⊕ M) ──m──→ T_{X/S}(M),
and on the other hand the morphism
T_{X/S}(M) ──δ──→ T_{X/S}(M ⊕ M) ≃ T_{X/S}(M) ×_X T_{X/S}(M)
is the diagonal morphism. Taking into account Remark 3.4.3, one deduces from the equalities (†) of 3.1.2 that
λ(x + y) = λx + λy, (λ + μ)x = λx + μx,
for every , and .
Remark 3.6.1. If is representable, in which case it satisfies (E) and and are representable by vector fibrations, the preceding laws are the same as those deduced from the vector fibration structures (cf. I 4.6).26
Proposition 3.4 bis. If satisfies (E), then satisfies (E) and the isomorphisms of 3.4 respect the structures of -modules, resp. of -modules.
Without comment.
Proposition 3.7. The functors and are functorial in , i.e., if is an -morphism, one has commutative diagrams:
T(f) L(f)
T_{X/S}(M) ──────→ T_{X′/S}(M) L^u_{X/S}(M) ──────→ L^{f∘u}_{X′/S}(M)
↓ ↓ ↘ ↙
X ──────────→ X′ , S .
Moreover,27 if is a monomorphism, so are and .
The existence of and , as well as the last assertion, follow immediately from the definitions. The commutativity of the diagrams then follows from the functoriality of these morphisms with respect to and from the fact that .
Remark 3.7.1.28 Suppose and representable and let be the rank of the free O_S-module .
Then, by 2.2.2, is isomorphic to the product over of copies of , and likewise
for . Consequently, the above square is cartesian when is an open immersion, more generally when
, for example if is étale; under these conditions, one has an
isomorphism of -functors
In general, the cartesian square of 3.7 defines a morphism of -functors:
T_{X/S}(M) ────────→ T_{X′/S}(M) ×_{X′} X
↘ ↙
X .
Proposition 3.7 bis. Let be an -morphism; if and satisfy (E) with respect to , then
T(f) L(f)
T_{X/S}(M) ─────────→ T_{X′/S}(M)_X resp. L^u_{X/S}(M) ─────────→ L^{f∘u}_{X′/S}(M)
is a morphism of O_X-modules (resp. of O_S-modules).
This follows from Proposition 3.7 by functoriality in .
Proposition 3.8. Let and be two functors above . One has isomorphisms functorial in :
(3.8.1) T_{X/S}(M) ×_S T_{Y/S}(M) ⥲ T_{(X×_S Y)/S}(M),
(3.8.2) L^u_{X/S}(M) ×_S L^v_{Y/S}(M) ⥲ L^{(u,v)}_{(X×_S Y)/S}(M).
29 The first isomorphism follows from 1.2 (∗); the second is deduced from it by the base change .
Remark 3.8.0. Note that (3.8.1) can also be interpreted as an isomorphism of -functors
T_{X/S}(M) ×_X (X ×_S Y) × T_{Y/S}(M) ×_Y (X ×_S Y) ⥲ T_{(X×_S Y)/S}(M).
(X×_S Y)
Corollary 3.8.1. If is equipped with an algebraic structure defined by finite cartesian products, then is equipped with a structure of the same kind and the projection is a morphism of this kind of structure.
Proposition 3.8 bis. If and satisfy (E), then satisfies (E) and (3.8.1) (resp.
(3.8.2)) is an isomorphism of -modules (resp. O_S-modules).
Proof.30 Suppose that and satisfy (E). Then, by 3.5.1 and (3.8.1), so does . Let us show that (3.8.1) is an isomorphism of -modules.
Let be an -morphism; taking into account 3.4.2, it suffices to see that the map
{ φ ∈ Hom_S(I_Z(M), X) | φ ∘ ε_M = x } × { ψ ∈ Hom_S(I_Z(M), Y) | ψ ∘ ε_M = y }
⟶ { θ ∈ Hom_S(I_Z(M), X ×_S Y) | θ ∘ ε_M = (x, y) }
which to associates is a morphism of -modules. But this is clear, because if then is sent to
(φ ∘ a^*) × (ψ ∘ a^*) = (φ × ψ) ∘ a^* = a · (φ × ψ).
Likewise, using 3.2.1, one shows that (3.8.2) is an isomorphism of O_S-modules.
3.9.0.31 If is an -group and if denotes its unit section, one writes:
i.e., is defined by the cartesian square:
i
Lie(X/S, M) ────────→ T_{X/S}(M)
│ │
│ │ p
↓ e ↓
S ─────────────────→ X .
By Corollary 3.8.1, the projection is a morphism of -groups, and it follows that is equipped with a structure of -group, and is isomorphic via to the kernel of .
If, moreover, satisfies condition (E), we shall see in Proposition 3.9 that the -group structure on , induced by that of , coincides with the abelian group structure induced by functoriality in (cf. 3.5.1). In fact, this result is valid under the weaker hypothesis that be an -functor of monoids or, more generally, an -functor of -sets (cf. the definition below).
Definitions 3.9.0.1. a) Let us introduce the following terminology:32 an -set is a set
equipped with a composition law with two-sided unit, denoted or simply . If is a morphism of
-sets, its kernel Ker f is ; it is a sub--set of .
b) An -object in a category is defined in the usual manner: it is therefore an object of , equipped with a morphism such that there exists a section of (above the final object) possessing the properties of a two-sided unit. Every -monoid, in particular every -group, is therefore a --object. In particular, an -object of the category of functors above the scheme will be called an --functor.
c) If is an --functor (for example, an -group), and if denotes the unit section of , one writes:
Lie(X/S, M) = L^e_{X/S}(M) and Lie(X/S) = Lie(X/S, O_S).
Let us make explicit the following particular case of 3.8.1.33
Corollary 3.9.0.2. If is an --functor (resp. an -group), then and are also --functors (resp. -groups), and one has morphisms of --functors (resp. of -groups):
i p
Lie(X/S, M) ────────→ T_{X/S}(M) ⇄ X ,
s
where is an isomorphism of onto Ker p and is a section of .
Proposition 3.9. Let be an --functor satisfying (E) with respect to . The --functor structure of coming from that of coincides with the -group structure defined in 3.5.1.
It follows from what was said above that is an -object in the category of O_S-modules. The
proposition will then follow from the following lemma:34
Lemma 3.10. Let be a category. Let be an -object in the category of --objects; is therefore a --object (whose composition law we shall denote ) equipped with a morphism of --objects .35 Then and is commutative.
By taking the values of the functors on a variable argument, one reduces, in the usual manner, to verifying the lemma when is the category of sets. One thus has a set and two maps such that
h(f(x, y), f(z, t)) = f(h(x, z), h(y, t)).
One has on the other hand two elements of , namely and , with , . One sees first that
h(f(u, y), f(x, u)) = f(x, y) = h(f(x, u), f(u, y)).
In particular, for , resp. , one obtains, respectively,
x = f(x, e) = h(f(u, e), f(x, u)) = h(u, f(x, u)) = f(x, u),
y = f(e, y) = h(f(e, u), f(u, y)) = h(u, f(u, y)) = f(u, y),
whence, on substituting in the original equality,
h(y, x) = f(x, y) = h(x, y).
This proves the lemma, and thus Proposition 3.9. From 3.9 one then deduces the following corollaries.
Corollary 3.9.1. If is an --functor satisfying (E) with respect to , every element of which projects onto the unit element of is invertible.
Corollary 3.9.2. If is an -monoid satisfying (E) with respect to , an element of is invertible if and only if its image in is.
Corollary 3.9.3. If is an -group satisfying (E) with respect to , the two -group laws on coincide.
Corollary 3.9.4.36 Let be an -group satisfying (E) with respect to . For ,
let be the morphism of -functors defined by . Then the derived morphism
is "multiplication by ", i.e. the map that to every
associates n x.
Let us first remark that is not in general a morphism of groups, but it preserves the unit section , so the derived morphism indeed sends into itself. If one denotes by the inclusion , then is defined by the equality , for every and . Now, by 3.9, the two group laws on (coming from condition (E) and coming from the law of ) coincide, i.e. one has , whence .
Before drawing other consequences of Proposition 3.9, let us prove another functoriality result:
Proposition 3.11. In the situation of Section 1, one has an isomorphism functorial in
Indeed, by definition (cf. 3.1):
T_{Hom_{Z/S}(X,Y)/S}(M) = Hom_S(I_S(M), Hom_{Z/S}(X, Y)).
By the isomorphism (4) of 1.1, applied to , one has:
Hom_S(I_S(M), Hom_{Z/S}(X, Y)) ≃ Hom_{Z/S}(X, Hom_Z(Z ×_S I_S(M), Y)).
Taking into account the isomorphism , this gives
T_{Hom_{Z/S}(X,Y)/S}(M) ≃ Hom_{Z/S}(X, Hom_Z(I_Z(M), Y)) = Hom_{Z/S}(X, T_{Y/Z}(M)).
Corollary 3.11.1. If satisfies (E), then satisfies (E) and the isomorphism of 3.11 respects the -module structures above .37
Proof. Let M, N be two free O_S-modules of finite type. If satisfies (E), then
T_{Y/Z}(M ⊕ N) ≃ T_{Y/Z}(M) ×_Y T_{Y/Z}(N).
The right-hand term is a subfunctor of and via the isomorphism of 1.2 (∗), one obtains an isomorphism
Hom_{Z/S}(X, T_{Y/Z}(M ⊕ N)) ≃
Hom_{Z/S}(X, T_{Y/Z}(M)) ×_{Hom_{Z/S}(X,Y)} Hom_{Z/S}(X, T_{Y/Z}(N)).
Combined with 3.11, this implies:
T_{Hom_{Z/S}(X,Y)/S}(M ⊕ N) ≃
T_{Hom_{Z/S}(X,Y)/S}(M) ×_{Hom_{Z/S}(X,Y)} T_{Hom_{Z/S}(X,Y)/S}(N),
so satisfies (E) with respect to . This proves the first assertion of the corollary.
Let us see the second. Denote by and give ourselves an -morphism , i.e., a -morphism , which therefore makes a -object.
On the one hand, one has a commutative diagram:
Hom_H(H′, Hom_{Z/S}(X, T_{Y/Z}(M))) ───→ Hom_S(H′, Hom_{Z/S}(X, T_{Y/Z}(M)))
Hom_Y(H′ ×_S X, T_{Y/Z}(M)) ──────────→ Hom_Z(H′ ×_S X, T_{Y/Z}(M))
{ ψ ∈ Hom_Z(I_{H′ ×_S X}(M), Y) | ψ ∘ ε_M = δ } ──→ Hom_Z(I_{H′ ×_S X}(M), Y) .
By 1.3, the action of on is given by: for every and ( being then above via ),
(αΨ)(h, x) = α(h, x) Ψ(h, x),
where acts on via the structure of O_Y-module of .
By 3.4.2, this last is given, via the identification
Hom_Y(H′ ×_S X, T_{Y/Z}(M)) = { ψ ∈ Hom_Z(I_{H′ ×_S X}(M), Y) | ψ ∘ ε_M = δ },
by: for every ,
(1) (αψ)(m, h, x) = ψ(m · α(h, x), h, x).
On the other hand, consider the tangent space ; one has a commutative diagram:
Hom_H(H′, T_{H/S}(M)) ──────────→ Hom_S(H′, T_{H/S}(M))
{ Φ ∈ Hom_S(I_{H′}(M), H) | Φ ∘ ε_M = Δ } ────────→ Hom_S(I_{H′}(M), H)
(∗)
{ φ ∈ Hom_Z(I_{H′ ×_S X}(M), Y) | φ ∘ ε_M = δ } ────→ Hom_Z(I_{H′ ×_S X}(M), Y) ,
where the bijection (∗) is given as follows (cf. 1.1 (2) and I 1.7.1): for every and (so that is above via ), one has and
(†) φ(m, h, x) = Φ(m, h) ∘ (x × id_U) ∈ Hom_Z(U, Y).
By 3.4.2 (where one replaces by and by ), the action of on is given by: for every and ,
(aΦ)(m, h) = Φ(m · a(h), h).
Consequently, if (resp. ) is the element of associated with (resp. ), one has, by (†),
(2) (aφ)(m, h, x) = Φ(m · a(h), h) ∘ (x × id_U) = φ(m · a(h), h, x).
Combined with (1), this shows that the isomorphism of 3.11.1 is an isomorphism of -modules; moreover, for every , the structure of -module on extends, in a manner functorial in , to a structure of -module.
In particular, for , one obtains the following corollary.
Corollary 3.11.2. One has an isomorphism functorial in
Moreover, if satisfies (E), then satisfies (E) and the preceding isomorphism respects the -module structures above .
38 Let be an -morphism; one identifies it with the constant morphism such that for every . One sees at once that the fiber product of and of is identified with , where is above via . Consequently, from the definition of and the preceding corollary one deduces:
Corollary 3.11.3. Let be an -morphism. One has an isomorphism functorial in (where in the right-hand term is above via ):
39 It is an isomorphism of O_S-modules if satisfies (E).
In particular, for , is an -functor of monoids, hence a fortiori an --functor; recalling that denotes , where is the unit section (cf. 3.9.0.1), one obtains:
Corollary 3.11.4. One has an isomorphism functorial in
Lie(End_S(X)/S, M) ⥲ ∏_{X/S} T_{X/S}(M);
39 it is an isomorphism of O_S-modules if satisfies (E).
Remark 3.11.5.40 Suppose that satisfies (E). Then is equipped with a structure of -module, i.e. for every ,
Hom_{X/S}(X, T_{X/S}(M))(S′) = { ψ ∈ Hom_X(I_{S′}(M) ×_S X, X) | ψ ∘ (ε_M × id_X) = pr_X }
is equipped with a structure of -module, functorial in . This follows, at one's choice, from 3.6 and the properties of the functor (cf. 1.2), or from the proof of 3.11.1.
We shall now interpret geometrically the definition of the tangent bundle.
41 Let be an -functor; by I 1.7.2, one has isomorphisms functorial in
T_{X/S}(M)(U) = Hom_S(U, Hom_S(I_S(M), X)) ≃ Hom_S(I_S(M), Hom_S(U, X))
≃ Hom_{I_S(M)}(U_{I_S(M)}, X_{I_S(M)}).
In particular, the morphism gives a commutative diagram:
Hom_S(U, T_{X/S}(M)) ⥲ Hom_{I_S(M)}(U_{I_S(M)}, X_{I_S(M)})
│ │
│ π_M │ identity
↓ ↓
Hom_S(U, X) ─────────────────────→ Hom_S(U, X) ,
where the second vertical arrow is obtained by the base change .42
In consequence:
Proposition 3.12. Let be an -morphism. Then is identified with the set of -morphisms from to which restrict to on (viewed as a subobject of via ).
43 In particular, for and , one obtains the
Corollary 3.12.1. The set is identified with the set of -endomorphisms of which induce the identity on , i.e. such that the following diagram be commutative:44
φ
I_X(M) ──────────→ I_X(M)
↑ ↑
ε_M ε_M
id
X ──────────────→ X .
45 On the other hand, by 3.11.2, ; if
moreover satisfies (E), then satisfies (E), so
is an O_S-module by 3.6 (and in fact a -module, by 3.11.5). Applying 3.9, one deduces the
Proposition 3.13. If satisfies (E), the abelian group is identified with the set of -endomorphisms of inducing the identity on . Consequently, every -endomorphism of which induces the identity on is an automorphism.
Corollary 3.13.1. Let be an -isomorphism, satisfying (E). Every -morphism from into which extends is an isomorphism.
Corollary 3.13.2. If satisfies (E), then the monomorphism induces, for every , an isomorphism
Proof.46 One must see that is a bijection, for every . By base change (cf. 3.4), it suffices to do it for . In this case, (resp. ) is the set of -morphisms (resp. automorphisms) which extend , and one applies the preceding corollary.
Corollary 3.13.3. If satisfies (E), the monomorphism induces, for every , an isomorphism . In particular, one has
Lie(Aut_S(X)/S, M) ⥲ Lie(End_S(X)/S, M) ⥲ ∏_{X/S} T_{X/S}(M),
47 so that is equipped with a structure of -module.
3.14. Suppose to conclude that is representable.48 In this case, one saw in 2.2.2 that the -functor is representable by , whence bijections:
Γ(T_{X/S}/X) ≃ Hom_X(Ω¹_{X/S}, O_X) ≃ Dér_{O_S}(O_X).
This is also deduced from what precedes, as follows. By 3.13, is identified with the set of
infinitesimal endomorphisms of (i.e. of I_S-endomorphisms of inducing the identity on ). Now
and have the same underlying topological space, the corresponding rings of sheaves being O_X and
, where is considered as ideal of square zero. Denoting by
the morphism of O_X-algebras that vanishes on , one deduces that to give an
infinitesimal endomorphism of is equivalent to giving a morphism of O_S-algebras such
that , which is equivalent to giving an O_S-derivation of the sheaf of rings O_X.
Moreover, one sees easily that if and if one denotes by the infinitesimal endomorphism corresponding to , then
This shows that the identification
{ infinitesimal endomorphisms of X } ≃ Dér_{O_S}(O_X)
is an isomorphism of abelian groups. Taking into account 3.13 (and 3.11.5), one has therefore constructed an isomorphism of abelian groups (and even of -modules)
which recovers the classical interpretation of tangent vector fields in terms of derivations of the structure sheaf.49 Let us remark in passing that is equal to , where is the dual of .
4. Tangent space to a group — Lie algebras
4.1. Let be a group-functor above . By 3.9.0.2, and are equipped with structures of groups above and one has morphisms of groups
i p
Lie(G/S, M) ────────→ T_{G/S}(M) ⇄ G ;
s
by definition is an isomorphism of onto the kernel of and is a section of . It then follows from I 2.3.7 that this sequence of morphisms allows one to identify with the semi-direct product of by .
Definition 4.1.A.50 The corresponding action of on is denoted
Ad : G ⟶ Aut_{S-gr.}(Lie(G/S, M))
and called the adjoint representation (relative to ) of ; one has therefore by definition, for and :
Ad(x) X = i^{-1}(s(x) i(X) s(x)^{-1}).
If is commutative, then is also commutative and .
Definition 4.1.B.50 If and are two group-functors above and if is a morphism of groups, by functoriality there is deduced a morphism of exact sequences compatible with the canonical sections:
1 ─→ Lie(G/S, M) ─→ T_{G/S}(M) ─→ G ─→ 1
│ L(f) │ T(f) │ f
↓ ↓ ↓
1 ─→ Lie(H/S, M) ─→ T_{H/S}(M) ─→ H ─→ 1 ;
, which we shall also write , is the derived morphism of .51
Remark 4.1.C.52 If and satisfy (E), then respects the O_S-module structures
deduced from "functoriality in " (cf. 3.6).
Proposition 4.1.1. Let . Then is the derived morphism of .
Indeed , which is nothing other than by the very definition of the derived morphism.
Suppose that satisfies (E). Then, by Proposition 3.9, the group structure on defined as above is
nothing other than the structure induced by its O_S-module structure (defined thanks to (E)). One then deduces from
the preceding proposition and the functoriality of the action of O_S (cf. 3.6) the corollary:
Corollary 4.1.1.1. Suppose that satisfies (E). Then Ad sends into the subgroup
Aut_{O_S-mod.}(Lie(G/S, M))
of , i.e., for every , respects the -module
structure of (where ). In other words, Ad is a linear representation (cf.
I, 3.2) of in the O_S-module .
Remarks 4.1.1.2. a) For to satisfy (E), it is necessary and sufficient that for every pair of free
O_S-modules of finite type, the diagram
Lie(G/S, M ⊕ N)
↙ ↘
Lie(G/S, M) Lie(G/S, N)
↘ ↙
Lie(G/S, 0) = S ,
obtained by applying the functor to the diagram (∗) of 2.1, be cartesian.
b) Suppose that satisfies (E). Then the derived morphism of the group law is nothing other than the addition law in . (N.B. is not a morphism of groups, but , so the derived morphism sends into , cf. 3.7 and 3.8.) For every , one shows likewise that if one denotes by the morphism of -functors defined by , then the derived morphism is multiplication by on , cf. 3.9.4.
4.1.2.0.53 Consider now the -functor ; for every , one has (cf. 3.4) and therefore:
Hom_{G/S}(G, T_{G/S}(M))(S′) ≃ Hom_{G_{S′}}(G_{S′}, T_{G_{S′}/S′}(M)) = Γ(T_{G_{S′}/S′}(M)/G_{S′}).
Note first that one has an isomorphism, functorial in ,
(∗) Hom_{S′}(G_{S′}, Lie(G_{S′}/S′, M)) ⥲ Γ(T_{G_{S′}/S′}(M)/G_{S′})
which to every associates the section such that, for every and :
Let be an automorphism of the functor above , not necessarily respecting the group structure. To every section of , one can associate defined by transport of structure: it is, for example, the only section of making commutative the diagram
τ
G_{S′} ─────────→ T_{G_{S′}/S′}(M)
│ │
h T(h)
│ h(τ) │
↓ ↓
G_{S′} ─────────→ T_{G_{S′}/S′}(M) .
In particular, take for the right translation by an element of , i.e., , for every , . Then one has immediately
where denotes the morphism from into defined by
for every , .
It follows that if one makes act by right translations on
Hom_{G/S}(G, T_{G/S}(M)) and Hom_S(G, Lie(G/S, M))
in the following way: for every , , and ,
(σ · x)(g) = σ(g · x^{-1}) · s(x) and (f · x)(g) = f(g · x^{-1}),
for every , , then the isomorphism (∗) above respects the actions of .
In particular, by this isomorphism, the elements of correspond to the constant morphisms from into (i.e. factoring through the projection ) or also to the elements of .
Terminology. The elements of will be called "sections of invariant under right translation".
One then obtains the:
Proposition 4.1.2.54 The map which to associates the section is a bijection of onto the part of formed by the sections invariant under right translation.
Likewise, one makes act on the right on as follows: for every , and ,
(u · x)(g) = u(g · x^{-1}) · x,
for every , . Then the morphism of 3.12.1
respects the actions of and therefore induces, for every , a bijection between and the set of -endomorphisms of which induce the identity on and "commute with right translations", i.e. which verify for every and . One obtains therefore:
Proposition 4.1.3.54 There exists a bijection functorial in between the set and the set of -endomorphisms of inducing the identity on and commuting with the right translations of (in the sense indicated above).
Taking now into account 3.13:
Theorem 4.1.4.54 Let be an -functor of groups; suppose that satisfies (E). Then the group is identified, functorially in , with the group of -automorphisms of inducing the identity on and commuting with the right translations of (in the sense indicated above).
One thus recovers (in the case ) one of the classical definitions of the Lie algebra of a group.
4.2.0.55 Before going further, let us establish new corollaries of 3.11. Let X, Y be above , with
above , as in Section 1. As one saw in 3.11, the isomorphisms 1.1 (4):
(1) Hom_S(I_S(M), Hom_{Z/S}(X, Y)) ⥲ Hom_{Z/S}(X, Hom_Z(I_Z(M), Y))
≃ Hom_{Z/S}(X ×_S I_S(M), Y)
induce the isomorphism below:
(2) T_{Hom_{Z/S}(X,Y)}(M) ⥲ Hom_{Z/S}(X, T_{Y/Z}(M))
≃ Hom_{Z/S}(X ×_S I_S(M), Y) .
By 1.3, if is a -group, so is for every (in particular for ); explicitly, if and , then is defined by , for every .
Suppose now that and are -groups and let us pose the following definition.
Definition 4.2.0.1. Let be the subfunctor of defined by: for every ,
(3) Hom_{(Z/S)-gr.}(X, Y)(S′) = Hom_{Z_{S′}-gr.}(X_{S′}, Y_{S′}).
This definition applies equally when one replaces by the -group .
One then sees easily that corresponds, in the preceding isomorphisms (2), to the -morphisms
φ : X_{S′} ×_{S′} I_{S′}(M) ⟶ Y_{S′}
which are "multiplicative in ", i.e. which verify , and these correspond to the morphisms of -groups . One has thus obtained:
Proposition 4.2.0.2. Let X, Y be -groups, with above . One has isomorphisms of -functors, functorial
in :
In particular, for , one obtains the following corollary. Before stating it, let us remark that if is a commutative -group, so is , then and , and finally .
Corollary 4.2.0.3. Let X, Y be -groups. One has isomorphisms of -functors, functorial in :
If is commutative, these are moreover isomorphisms of abelian -groups.
Definition 4.2.0.4.56 If is an O_S-module, the functor (resp. ) is
equipped with an O_S-module structure deduced from that of . Equipped with this structure, it will be denoted
(resp. ).
Consequently, if X, Y are O_S-modules, then and
, then and ,
are equipped with an O_S-module structure, and one obtains the:
Corollary 4.2.0.5. If X, Y are O_S-modules, one has isomorphisms of O_S-modules, functorial in :
T′_{Hom_{O_S-mod.}(X,Y)/S}(M) ⥲ Hom_{O_S-mod.}(X, T′_{Y/S}(M)).
Definition 4.2.A.57 Let X, L be -groups, with acting on by group automorphisms (cf. I
2.3.5). One defines the subfunctor of as follows: for every
,
Z¹_S(X, L)(S′) = { φ ∈ Hom_{S′}(X_{S′}, L_{S′}) | φ(x_1 x_2) = φ(x_1) (x_1 · φ(x_2)),
for all x_1, x_2 ∈ X(S″), S″ → S′ }.
It is called the "functor of crossed homomorphisms from to ".
Remark 4.2.B.57 a) If is a second -group on which acts by group automorphisms, one has
Z¹_S(X, L ×_S L′) ≃ Z¹_S(X, L) ×_S Z¹_S(X, L′).
b) If is a -O_S-module, coincides with the kernel of the differential
defined in I 5.1; in particular,
is in this case an O_S-module.
4.2. Let be a morphism of -groups; then is described as follows. First, one saw in 3.11.3 that one has an isomorphism of -functors, functorial in :
(†) L^u_{Hom_S(X,Y)/S}(M) ⥲ Hom_{Y/S}(X, T_{Y/S}(M)).
On the other hand, since is an -group, one has
T_{Y/S}(M) = Lie(Y/S, M) · Y = Lie(Y/S, M)_Y;
there follows an isomorphism of -functors, functorial in :
L^u_{Hom_S(X,Y)/S}(M) ⥲ Hom_S(X, Lie(Y/S, M)).
For every , let denote the morphism deduced from by base change. Consider the -functor defined as follows:58 for every ,
Hom_{(Y/S)-gr.}(X, Lie(Y/S, M) · Y)(S′) = Hom_{Y′-gr.}(X′, (Lie(Y/S, M) · Y)_{S′})
= Hom_{Y′-gr.}(X′, Lie(Y′/S′, M) · Y′).
Then one sees easily that the isomorphism (†) induces an isomorphism
(†′) L^u_{Hom_{S-gr.}(X,Y)/S}(M) ⥲ Hom_{(Y/S)-gr.}(X, Lie(Y/S, M) · Y).
On the other hand, the morphism
X ──u──→ Y ──Ad──→ Aut_{S-gr.}(Lie(Y/S, M))
defines an action of on by group automorphisms.
If , then, for every and , one can write in a unique manner
Φ(S′)(x) = φ(S′)(x) · u′(x), where φ(S′)(x) ∈ Lie(Y′/S′, M)(S″);
this determines an element of . Then is a morphism of groups if, and only if, for all one has:
φ(S′)(x_1 x_2) = φ(S′)(x_1) ( u(x_1) φ(S′)(x_2) u(x_1)^{-1} )
= φ(S′)(x_1) (x_1 · φ(S′)(x_2)),
i.e., if and only if . One has thus obtained the:
Proposition 4.2. Let be a morphism of -groups. One has an isomorphism of -functors, functorial in :
L^u_{Hom_{S-gr.}(X,Y)/S}(M) ⥲ Z¹_S(X, Lie(Y/S, M)).
59 Suppose further that satisfies (E). Then it follows from 4.2.0.3, exactly as in the proof of 3.11.1, that satisfies (E). So one has (cf. 3.5.1):
L^u_{Hom_{S-gr.}(X,Y)/S}(M ⊕ N) ≃ L^u_{Hom_{S-gr.}(X,Y)/S}(M) ×_S L^u_{Hom_{S-gr.}(X,Y)/S}(N).
(This also follows from 4.2 and 4.2.B a).) Consequently, is equipped,
like (cf. 4.2.B b)), with an O_S-module structure, deduced from functoriality in . One
deduces that the isomorphism of 4.2 is, in this case, an isomorphism of O_S-modules:
Proposition 4.2 bis.59 Let be a morphism of -groups; suppose that satisfies (E).
One has an isomorphism of O_S-modules, functorial in :
L^u_{Hom_{S-gr.}(X,Y)/S}(M) ⥲ Z¹_S(X, Lie(Y/S, M)).
Moreover, when satisfies (E), one deduces from 3.13.1, exactly as in the proof of 3.13.2, that for every one has an isomorphism functorial in :
One deduces from this the following two corollaries.
Corollary 4.2.1. Let be a morphism of -groups; if satisfies (E), one has an isomorphism of
O_S-modules, functorial in :
L^u_{Isom_{S-gr.}(X,Y)/S}(M) ⥲ Z¹_S(X, Lie(Y/S, M)).
Corollary 4.2.2. Let be an -group; if satisfies (E), one has an isomorphism of O_S-modules,
functorial in :
Lie(Aut_{S-gr.}(X)/S, M) ⥲ Z¹_S(X, Lie(X/S, M)).
60 Moreover, if is commutative, the adjoint action of on is trivial, whence . Therefore:
Corollary 4.2.3. Let be a commutative -group; one has an isomorphism of -functors, functorial in :
L^u_{Hom_{S-gr.}(X,Y)/S}(M) ⥲ Hom_{S-gr.}(X, Lie(Y/S, M)).
4.3. Consider now the case where and are O_S-modules. Recall (cf. 4.2.0.4) that one writes
(resp. ) for the functor (resp. ) equipped with the O_S-module structure
deduced from that of .
When satisfies (E), we shall always write for the functor equipped with the
O_S-module structure defined for every functor satisfying (E). In this case, we know (cf. 3.9) that the abelian-group
structures of and coincide, but the same is not a priori true for those of module (see a
counterexample in §6.3). For every and , we shall write (resp. ) for the
action of on (resp. on ), and similarly for the action of on
and .
One has as O_S-modules; consequently one obtains, exactly as for
Propositions 4.2 and 4.2 bis, the:
Proposition 4.3. Let be a morphism of O_S-modules. One has an isomorphism of -functors,
functorial in :
(∗) L^u_{Hom_{O_S-mod.}(X,Y)/S}(M) ⥲ Hom_{O_S-mod.}(X, Lie′(Y/S, M)).
61 If satisfies (E), then satisfies (E) and (∗) is an
isomorphism of O_S-modules when one equips both sides with the O_S-module structure deduced from functoriality in
.62
Remark 4.3 bis.63 Let be a morphism of O_S-modules; denote by the map which
to every morphism of O_S-modules associates the morphism
u ⊕ φ : X ⟶ T′_{Y/S}(M) = Y ⊕ Lie′(Y/S, M).
Then the isomorphism of 4.3 fits into the following commutative diagram, functorial in :
L^u_{Hom_{O_S-mod.}(X,Y)/S}(M) ⥲ Hom_{O_S-mod.}(X, Lie′(Y/S, M))
│ │
↓ ↓ τ_u
T_{Hom_{O_S-mod.}(X,Y)/S}(M) ⥲ Hom_{O_S-mod.}(X, T′_{Y/S}(M)).
Moreover, when satisfies (E), one deduces from 3.13.1, exactly as in the proof of 3.13.2, that for every , one has
(∗) L^u_{Isom_{O_S-mod.}(X,Y)/S}(M) = L^u_{Hom_{O_S-mod.}(X,Y)/S}(M).
Corollary 4.3.1. Let be an O_S-module satisfying (E) with respect to . One has an isomorphism functorial
in :
Lie(Aut_{O_S-mod.}(X)/S, M) ⥲ Hom_{O_S-mod.}(X, Lie′(X/S, M))
which respects the O_S-module structures deduced from functoriality in .64 In particular,
satisfies (E).
Proof. The first assertion follows from (∗) and 4.3; let us prove the second. As satisfies (E), one has
isomorphisms of O_S-modules Lie′(X/S, M ⊕ N) ≃ Lie′(X/S, M) ×_S Lie′(X/S, N), and thus:
Lie(Aut_{O_S-mod.}(X)/S, M ⊕ N) ≃
Lie(Aut_{O_S-mod.}(X)/S, M) ×_S Lie(Aut_{O_S-mod.}(X)/S, N).
Taking 4.1.1.2 a) into account, this proves that satisfies (E).
4.3.2. Before continuing in this direction, let us examine more closely the relations between , and . Let us remark first that
(1) Lie(O_S/S, M) = Lie′(O_S/S, M) = W(M)
(where is defined in I 4.6) and that one therefore has a canonical isomorphism
Let now be an O_S-module. For every 65, one has a dihomomorphism
(3) ⎰ F(S_1) → F(S_2)
⎱ O(S_1) → O(S_2),
whence a morphism of -modules
F(S_1) ⊗_{O(S_1)} O(S_2) ⟶ F(S_2).
In particular, setting and , one deduces morphisms of -modules, functorial in ,
F(S′) ⊗_{O(S′)} T_{O_S/S}(M)(S′) ⟶ T′_{F/S}(M)(S′);
letting vary, one obtains morphisms of O_S-modules, functorial in ,
(4) F ⊗_{O_S} T_{O_S/S}(M) ⟶ T′_{F/S}(M).
These morphisms are functorial in , hence compatible with the projections of the tangent bundles onto their bases;
they therefore define morphisms of O_S-modules
(5) F ⊗_{O_S} Lie(O_S/S, M) ⟶ Lie′(F/S, M)
such that one has the commutative diagram:
0 ─→ F ⊗_{O_S} Lie(O_S/S, M) ─→ F ⊗_{O_S} T_{O_S/S}(M) ─→ F ─→ 0
↓ ↓ ↓
0 ─────→ Lie′(F/S, M) ──────→ T′_{F/S}(M) ────→ F ─→ 0 .
One may consider the morphisms (5) as morphisms of abelian -groups
(6) F ⊗_{O_S} Lie(O_S/S, M) ⟶ Lie(F/S, M);
tensoring with the isomorphism , one deduces (for ) a morphism of abelian -groups
(7) F ⥲ F ⊗_{O_S} Lie(O_S/S) ⟶ Lie(F/S)
also denoted .
Remark 4.3.3.66 When satisfies (E), the morphisms (6) and (7) are not necessarily morphisms of
O_S-modules, when one equips both sides with the module structures deduced from that of thanks to condition (E).
For example, let be a field of characteristic , , and let be the O_S-module
which to every -scheme associates equipped with the -module structure obtained by
making scalars act through the -th power, i.e., , for , . As -functor
of groups, is isomorphic to . Therefore satisfies (E) and is identified with
. Then, the canonical morphism is, for every , the identity
map : it indeed respects the abelian-group structure but not the O_S-module structure.
Remark 4.3.4.67 One can make the morphisms (4) and (5) explicit as follows. The morphism is defined by: for every , , and ,
Θ(f ⊗ α) = α · (τ_0 ∘ f) = α · (f ∘ ρ),
where is the null section and the structural morphism . Then induces a morphism ; this follows from the "functoriality in " already mentioned after (4), and can be seen explicitly as follows. On the one hand, one has
Lie′(F/S, M)(S′) = { φ ∈ Hom_S(I_{S′}(M), F) | φ ∘ ε_M = e },
where denotes the unit section . On the other hand, is the kernel of the augmentation , and it is therefore a question of seeing that if and , then .
Consider the dihomomorphism (3), in the case where is the -morphism ; one has then a commutative diagram
F(ε_M)
F(I_{S′}(M)) ─────────────→ F(S′)
│ α │ η(α)
↓ F(ε_M) ↓
F(I_{S′}(M)) ─────────────→ F(S′) .
For every , one therefore has , whence if .
In particular, taking , one has and the morphism is given by .
Remark 4.3.5.68 Let be an O_S-module, set and denote by
and the morphisms of O_S-modules given by 4.3.2 (5):
d_F : F ⊗_{O_S} Lie(O_S/S, M) ⟶ Lie′(F/S, M),
d_E : E ⊗_{O_S} Lie(O_S/S, M) ⟶ Lie′(E/S, M).
One deduces from 4.3.4 the following commutative diagram of morphisms of O_S-modules:
d_E
E ⊗_{O_S} Lie(O_S/S, M) ──────────────→ Lie′(E/S, M)
↑ ≃ (∗)
d_F
Hom_{O_S}(F, F ⊗_{O_S} Lie(O_S/S, M)) ─→ Hom_{O_S}(F, Lie′(F/S, M))
where the right vertical arrow is the isomorphism (∗) of 4.3. Therefore (loc. cit.), if satisfies (E), then
satisfies (E) and (∗) is also an isomorphism of O_S-modules when one equips the right-hand terms with the O_S-module
structure deduced from (E).
Remark 4.4.0.69 In 4.3.2, the morphisms (4) are isomorphisms if and only if the morphisms (5) are.
Moreover, if these conditions are verified, then satisfies (E). Indeed, it suffices to verify that
Lie′(F/S, M ⊕ N) ≃ Lie′(F/S, M) ×_S Lie′(F/S, N). Now one has the commutative diagram below, where by hypothesis the
horizontal arrows are isomorphisms:
F ⊗_{O_S} Lie(O_S/S, M ⊕ N) ⥲ Lie′(F/S, M ⊕ N)
│ ≀ │
↓ ↓
F ⊗_{O_S} (Lie(O_S/S, M) ×_S Lie(O_S/S, N)) ⥲ Lie′(F/S, M) ×_S Lie′(F/S, N);
the second vertical arrow is therefore also an isomorphism, i.e. satisfies (E).
Definition 4.4. One says that is a good O_S-module if the morphisms
F ⊗_{O_S} T_{O_S/S}(M) ⟶ T_{F/S}(M)
or, equivalently,
F ⊗_{O_S} Lie(O_S/S, M) ⟶ Lie(F/S, M),
are isomorphisms of abelian -groups (so that satisfies (E)) and if moreover they respect the O_S-module
structures deduced from condition (E).
Corollary 4.4.1.70 Let be an O_S-module. Consider the following conditions:
(i) is a good O_S-module.
(ii) satisfies (E) and is an isomorphism of O_S-modules.
(iii) .
Then one has (i) ⇔ (ii) ⇒ (iii).
Proof. (i) ⇒ (ii) follows from the definition. To prove (ii) ⇒ (i), one must show that the morphisms of abelian -groups (functorial in )
F ⊗_{O_S} Lie(O_S/S, M) ⥲ Lie(F/S, M)
are isomorphisms of O_S-modules. Since satisfies (E), both sides transform finite direct sums of copies of O_S
into finite products of abelian -groups. This reduces us to the case , which follows from the hypothesis.
Finally, (i) ⇒ (iii) follows from the definition and from the fact that the isomorphisms
F ⊗_{O_S} Lie(O_S/S, M) ⥲ Lie′(F/S, M)
of 4.3.2 (5) are morphisms of O_S-modules.
Examples 4.4.2. For every quasi-coherent O_S-Module , the O_S-modules and defined in I 4.6 are
good.
71 Indeed, for every , the morphisms
V(E)(S′) ⊗_{O(S′)} O(I_{S′}(M)) ⟶ T_{V(E)/S}(M)(S′)
W(E)(S′) ⊗_{O(S′)} O(I_{S′}(M)) ⟶ T_{W(E)/S}(M)(S′)
correspond, respectively, to the morphisms:
Hom_{O_{S′}-mod.}(f^*(E), O_{S′}) ⊗_{O(S′)} Γ(S′, D_{O_{S′}}(M))
⟶ Hom_{O_{S′}-mod.}(f^*(E), D_{O_{S′}}(M))
Γ(S′, f^*(E)) ⊗_{O(S′)} Γ(S′, D_{O_{S′}}(M))
⟶ Γ(S′, f^*(E) ⊗_{O_{S′}} D_{O_{S′}}(M));
these are isomorphisms, since is isomorphic, as -module, to a finite direct sum of copies of .
Proposition 4.5. Let be a good O_S-module. Then:
(i) satisfies (E) and one has a functorial isomorphism
Lie(Aut_{O_S-mod.}(F)/S, M) ⥲ Hom_{O_S-mod.}(F, Lie(F/S, M))
which respects the O_S-module structures deduced from condition (E). In particular, one has an isomorphism of
O_S-modules
Lie(Aut_{O_S-mod.}(F)/S) ⥲ End_{O_S-mod.}(F).
(ii)72 Moreover, is a good O_S-module.
Indeed, by 4.4.1, satisfies (E) and
(1) Lie(F/S, M) = Lie′(F/S, M) ≃ F ⊗_{O_S} Lie(O_S/S, M).
Assertion (i) then follows from 4.3.1. Set . By (1) and Remark 4.3.5, one has the following commutative diagram of morphisms of abelian -groups:
d_E
End_{O_S-mod.}(F) ⊗_{O_S} Lie(O_S/S, M) ────→ Lie(End_{O_S-mod.}(F)/S, M)
↑ ≃ (∗)
d_F
Hom_{O_S}(F, F ⊗_{O_S} Lie(O_S/S, M)) ────→ Hom_{O_S}(F, Lie(F/S, M))
where and (∗) are isomorphisms of O_S-modules; consequently, so is . This proves (ii).
Scholie 4.5.1.73 Set (cf. 2.1) (with ), and let be a
good O_S-module. Then, for every , the morphism
F(S′) ⊕ t F(S′) = F(S′) ⊗_{O(S′)} O(I_{S′}) ⟶ F(I_{S′}) = F(S′) ⊕ Lie(F/S)(S′)
(which is the identity on ) induces an isomorphism of -modules . Letting vary, one obtains an isomorphism which one may write .
For every , one has therefore, by 4.5, a commutative diagram:
End_{O_{S′}-mod.}(F_{S′}) ⥲ Hom_{O_{S′}-mod.}(F_{S′}, t F_{S′}) ⥲ Lie(Aut_{O_S-mod.}(F)/S)(S′)
│ │ │
↓ ↓ ↓
Aut_{O_{I_{S′}}-mod.}(F_{I_{S′}}) T_{Aut_{O_S-mod.}(F)/S}(S′)
and one deduces from 4.3 bis that every corresponds to the element of .
Definition 4.6. One says that the -functor of groups is good if satisfies condition (E) and if
is a good O_S-module.
74 Note that if is a good O_S-module, it is a good -group; indeed satisfies (E) and
is a good O_S-module.
Example 4.6.1. If is representable, it is good. Indeed, satisfies (E) and is of the form , hence good, by 4.4.2.
Lemma 4.6.2.75 Let be an -functor of groups such that satisfies (E), and let
. Then satisfies (E) and the morphism of abelian groups respects the O_S-module
structures.
Consequently, is good if and only if is bijective.
Proof. Suppose that satisfies (E). Let M, N be free O_S-modules of finite rank. Write
and the unit section of .
For every , one has and is identified with
{ Φ ∈ Hom_S(I_{S′}(N) ×_{S′} I_{S′}(M), G) |
Φ ∘ (ε_N × id_{I_{S′}(M)}) = e ∈ G(I_{S′}(M)),
Φ ∘ (id_{I_{S′}(N)} × ε_M) = e ∈ G(I_{S′}(N)) }.
This shows that
(1) Lie(F(N)/S, M) ≃ Lie(F(M)/S, N).
As satisfies (E), one deduces:
Lie(F(N)/S, M_1 ⊕ M_2) ≃ Lie(F(M_1 ⊕ M_2)/S, N)
≃ Lie((F(M_1) ×_S F(M_2))/S, N).
By 3.8, the right-hand term is isomorphic to
Lie(F(M_1)/S, N) ×_S Lie(F(M_2)/S, N) ≃ Lie(F(N)/S, M_1) ×_S Lie(F(N)/S, M_2).
It follows that
(2) Lie(F(N)/S, M_1 ⊕ M_2) ≃ Lie(F(N)/S, M_1) ×_S Lie(F(N)/S, M_2),
so satisfies (E), by Remark 4.1.1.2 a).
Let us now show that the morphism of abelian groups respects the O_S-module structures.
Consider the free O_S-module , so that
O(I_{S′}(M)) = O(S′)[t]/(t²) = O(S′) ⊕ t O(S′)
and . We denote by the structural morphism , which corresponds to the injection . Recall (cf. 3.4.2) that, for every -functor , is the set of -morphisms such that , and that the action of is given by , where is the endomorphism of associated with , cf. 2.1.3.
Consequently, by 4.3.4, the morphism is given by: for every and ,
f ↦ f ⊗ t ↦ t · (f ∘ ρ_t) = f ∘ ρ_t ∘ t^*,
where, in the last term, is considered as an -morphism
(such that ). It is a question of seeing that is a
morphism of O_S-modules, i.e., that for every .
Considering as a morphism , one has likewise . On the other hand, by the functoriality in of the action of on (cf. 2.1.3), one has the commutative diagram below:
a^*
I_{S′}(N) ─────────────────→ I_{S′}(N)
↑ ↑
ρ_t ρ_t
u_t(a)^*
I_{I_{S′}(M)}(N) ─────────────→ I_{I_{S′}(M)}(N) .
One has therefore
d(a · f) = f ∘ a^* ∘ ρ_t ∘ t^* = f ∘ ρ_t ∘ u_t(a)^* ∘ t^* = f ∘ ρ_t ∘ t^* ∘ u_t(a)^* = u_t(a) · d(f)
(the second-to-last equality resulting from the fact that is commutative). This completes the proof of Lemma 4.6.2.
Theorem 4.7. If is a good O_S-module, the -group is good.
76 Indeed, by 4.5, satisfies (E) and
Lie(Aut_{O_S-mod.}(F)/S) ≃ End_{O_S-mod.}(F) is a good O_S-module.
4.7.1.77 Let now be an -group and a good O_S-module. Suppose given a linear representation
of in , that is to say (I 4.7.1), a morphism of -groups
If satisfies (E), one deduces from 4.1.C and 4.5 a morphism of O_S-modules, denoted or :
Lie(G/S) ⟶ Lie(Aut_{O_S-mod.}(F)/S) ≃ End_{O_S-mod.}(F).
78 Moreover, setting (with ), one deduces from 4.5.1 that, if and , then one has in the following equality:
(∗) ρ(X) = id + t ρ′(X),
i.e. for every and , one has in the equality .
Definition 4.7.2. Let be a good -group. Then is a good O_S-module, and one has a morphism of
-groups Ad : G → Aut_{O_S-mod.}(Lie(G/S)). One deduces from 4.7.1 a morphism of O_S-modules
ad : Lie(G/S) ⟶ End_{O_S-mod.}(Lie(G/S)),
or, what amounts to the same, an O_S-bilinear morphism:
Lie(G/S) ×_S Lie(G/S) ⟶ Lie(G/S), (x, y) ↦ [x, y] = ad(x) · y
(where and denote two arbitrary elements of ). If is commutative, then .
4.7.3. One can give an equivalent definition of the bracket as follows: let us remark first that it suffices to do
it for . Let us remark next that there is a canonical isomorphism
; to avoid confusions, let us denote by and two copies of I_S and
set , , where . One has then a commutative diagram
I × I′ ──→ I′
│ │
↓ ↓
I ────→ S ,
the two arrows starting from identifying it with the scheme of dual numbers over or over . There follows a commutative diagram of groups (where one writes ):
1 1
│ │
↓ ↓
L(I) ──────→ L(S) ─→ 1
│ │
1 ─→ L(I′) ─→ G(I × I′) ─→ G(I′) ─→ 1
(1) │ │
↓ ↓
1 ─→ L(S) ─→ G(I) ─→ G(S) ─→ 1
│ │
↓ ↓
1 1 .
The ninth piece of the puzzle is nothing other than . If is good, this is and one has therefore the following commutative diagram, where the lines and columns are exact sequences of groups, the five are commutative,79 and where, taking into account the identification (resp. ), the injection (resp. ) is given by (resp. ):
t
z ∈ L(S) ────────→ L(I) ─────→ L(S) ∋ x
│ t′ │
↓ ↓
L(I′) ────→ G(I × I′) ────→ G(I′)
│ │
↓ ↓
y ∈ L(S) ─────→ G(I) ─────→ G(S) .
Now in such a diagram, if one takes two elements and as marked, and one lifts arbitrarily resp. to an element resp. , the commutator in does not depend on the chosen lifts and is the image of an element as marked. The reader will verify that one has .80
Indeed, if one still denotes by the image of by the canonical section (and likewise for ), then and , with , and since and are commutative one has
x̃ y′ x̃^{-1} y′^{-1} = x u y v u^{-1} x^{-1} v^{-1} y^{-1}
= x u y u^{-1} v x^{-1} v^{-1} y^{-1}
= x y x^{-1} y^{-1}.
Moreover, this element is sent to the unit element of and of , hence comes from a (unique) as indicated. Finally, considering (resp. ) as an element of (resp. of ), one has by 4.7.1 (∗):
x y x^{-1} = Ad(x)(y) = (id + t′ ad(x))(y) = y + t′ [x, y],
so the element of is the image of the element of .
On this construction the following two properties appear:
(i) the bracket is "functorial in ": precisely, is a functor from the category of good
-groups to the category of good O_S-modules equipped with an O_S-bilinear composition law.
(ii) One has : indeed the diagram is symmetric with respect to the first diagonal.81
Proposition 4.8. Let be a good O_S-module. Via the identification
Lie(Aut_{O_S-mod.}(F)/S) = End_{O_S-mod.}(F)
one has
Ad(g) · Y = g ∘ Y ∘ g^{-1} and [X, Y] = X ∘ Y − Y ∘ X,
for every , , and
X, Y ∈ Lie(Aut_{O_S-mod.}(F)/S)(S′) = End_{O_{S′}-mod.}(F_{S′}).
Proof.82 By base change, one reduces to , which permits one to lighten the notation. Set
and (with ). Recall (cf. 4.5.1) that the inclusion
i : End_{O_S-mod.}(F) ↪ Aut_{O_I-mod.}(F_I) sends to . Then, by definition of (cf. 4.1.A), one
has:
id + t Ad(g)(Y) = g ∘ (id + t Y) ∘ g^{-1} = id + t (g ∘ Y ∘ g^{-1}),
whence .
Let be a second copy of I_S, set (with ). Apply the results of 4.7.3 to
and L = Lie(G/S) = Aut_{O_S-mod.}(F). One identifies with its image under
the canonical section ; its image in is then , whose inverse is
. Likewise, lifts to , whose inverse is . Then, the commutator
(id + t′ X) ∘ (id + t Y) ∘ (id − t′ X) ∘ (id − t Y) = id + t t′ (X ∘ Y − Y ∘ X)
is the image in of the element of (indeed, is sent to , then to ). By 4.7.3, this shows that .
Corollary 4.8.1. Let be a good -group and . One has:
[x, [y, z]] + [y, [z, x]] + [z, [x, y]] = 0.
83 Indeed, since is good, is a good O_S-module and therefore, by 4.7,
is a good -group. Then, the morphism of -groups
Ad : G ⟶ Aut_{O_S-mod.}(Lie(G/S))
gives, by the functoriality 4.7.3 (i): . Combined with 4.8, this gives:
ad[x, y] = [ad x, ad y] = ad x ∘ ad y − ad y ∘ ad x,
which, applied to an element , gives the Jacobi relation.
Corollary 4.8.2. Let be a good -group acting linearly on a good O_S-module (i.e. let be a
-O_S-module, with and good). Then the linear map ρ′ : Lie(G/S) → End_{O_S-mod.}(F) is a representation,
i.e. one has
ρ′([x, y]) = ρ′(x) ∘ ρ′(y) − ρ′(y) ∘ ρ′(x).
Scholie 4.9. To every good -group (for example representable), one has associated a good O_S-module
functorially equipped with a bilinear map satisfying
[x, y] + [y, x] = 0, [x, [y, z]] + [y, [z, x]] + [z, [x, y]] = 0.
We shall call equipped with this structure the "Lie algebra" of over (the quotation marks being
justified by the fact that one does not know whether is, strictly speaking, an O_S-Lie
algebra84). To every linear representation of in a good O_S-module is associated a
representation of its "Lie algebra". In particular, to the adjoint representation of is associated the adjoint
representation of its "Lie algebra".
Definition 4.10. A group-functor above is said to be very good if it is good and if is an
O_S-Lie algebra (i.e. if one has identically ).
Examples 4.10.1. The following -groups are very good: for every good
O_S-module (cf. 4.7 and 4.8), every representable group (see below), every good -group admitting a monomorphism
into a very good -group, for example every good subgroup-functor of a representable group, or every good -group
admitting a faithful linear representation in a good O_S-module, for example every good -group such that Ad is a
monomorphism . . .
4.11. Suppose now that is a group scheme over . By 4.1.4, is identified with the group of
infinitesimal automorphisms of right-invariant, that is to say, by 3.14, with the group of derivations of O_G
over O_S invariant by right translation. Moreover this identification respects the module structure and is
an85 anti-isomorphism of Lie algebras, as one sees by reasoning as in 4.7.3: set and
and let and . The left translation (resp. ) is
an -automorphism of which induces the identity on (resp. G_I) and which corresponds to an
O_S-automorphism
u = id + t d_x resp. v = id + t′ d_y
of , where are O_S-derivations of O_G invariant by
right translation. Since the correspondence between -automorphisms of and O_S-automorphisms of
is contravariant, corresponds to
. One deduces, by 4.7.3, that the map is
an isomorphism of Lie algebras (for more details, see [DG70], § II.4, 4.4 and 4.6). What precedes is valid for
for every . One thus recovers the classical definition:86
Scholie 4.11.1. Via the isomorphism , is identified with the functor that to every above associates the -Lie algebra of derivations of with respect to invariant by right translation.
Since one knows already, by 4.6.1, that every representable group is good, it follows from what precedes:
Corollary 4.11.2. Every representable group is very good.
Let be the unit section of . Set and recall (cf. 3.3) that is representable by the vector fibration .
Scholie 4.11.3. One has therefore associated functorially to every -group scheme a vector fibration
over , which represents the functor , hence is equipped with a structure
of -scheme of O_S-Lie algebras. Moreover (cf. 3.4 and 3.8), this construction commutes with base extension and
finite products.
Remarks 4.11.4.87 Denote by the morphism .
a) The O_G-module is evidently -equivariant (cf. I, § 6) and therefore, by I
6.8.1, one has . It follows for example that is
locally free (resp. locally free of finite type) if is, which is in particular the case if is the
spectrum of a field (resp. if is the spectrum of a field and locally of finite type over ).
b) Moreover, by I 6.8.2, is equipped with a canonical structure of -O_S-module, which induces
on the adjoint action.88
c) On the other hand (cf. EGA I, 5.3.11 and 5.4.6), is an immersion, and is a closed immersion if is separated over . Therefore is identified with , where is the quasi-coherent ideal defining in an open subset of in which is closed (if is separated, one may take , and if is affine over , is nothing other than the augmentation ideal of , i.e. the kernel of , cf. I 4.2).89
Remark 4.11.5. One can deduce from the isomorphism that the
O_S-module is identified with the sheaf of "differentials of
with respect to invariant on the right", that is to say, with the sheaf whose sections on an open subset of
are the sections of on invariant by right translation (cf. I, 6.8.3, compare also with
VII_A, 2.4).90
Notation 4.11.6. One denotes by the sheaf of sections of the vector fibration ; it is
the O_S-module dual to
(cf. EGA II 1.7.9). It is equipped with a structure of O_S-Lie algebra.
Since this construction does not commute with base extension (in general), the Lie algebra structure on this module
does not allow one to reconstruct the structure of -scheme of O_S-Lie algebras on .91
However one has:
Lemma 4.11.7. Suppose locally free of finite type (which is the case in particular if is smooth over (cf. EGA IV_4, 17.2.4), or if is the spectrum of a field and locally of finite type over ). Then and therefore
Lie(G/S) = V(ω¹_{G/S}) = V(Lie(G/S)^∨) = W(Lie(G/S))
(the last equality following from I 4.6.5.1).
Finally, let be a monomorphism of group-functors. Then is also a monomorphism (cf. 3.7). Since every vector monomorphism of vector fibrations is a closed immersion92 one obtains:
Corollary 4.11.8. Let be a monomorphism of -group schemes.
(i) is a closed immersion and therefore is an epimorphism.
(ii) If is locally free of finite type, then the corresponding morphism is an isomorphism of onto a sub-module of locally direct factor.
5. Computation of some Lie algebras
5.1. Examples of Lie algebras: diagonalizable groups
Let be a diagonalizable group over (I 4.4). Since the formation of commutes with base extension, it suffices to make the construction for . One has then:
G(I_S) = Hom_{gr.}(M, Γ(I_S, O_{I_S})^×) = Hom_{gr.}(M, Γ(S, D_{O_S})^×).
Now one has a split exact sequence
1 ⟶ Γ(S, O_S) ⟶ Γ(S, D_{O_S})^× ⟶ Γ(S, O_S)^× ⟶ 1,
which gives that is identified with equipped with its obvious -module structure. One thus obtains, after base change:
Proposition 5.1. One has isomorphisms
Hom_{S-gr.}(M_S, O_S) ⥲ Lie(D_S(M)/S) and Hom_{gr.}(M̃_S, O_S) ⥲ Lie(D_S(M)/S),
(where, in the second isomorphism, denotes the constant sheaf of groups on defined by , and the sheaf of homomorphisms of sheaves of groups).
Corollary 5.1.1. If is free of finite type (or, as we shall say later, if is a split torus), then (see I, 4.6.5 for the definition of )
W(Lie(D_S(M)/S)) ⥲ Lie(D_S(M)/S),
M^∨ ⊗_ℤ O_S ⥲ Lie(D_S(M)/S),
where denotes the dual of the abelian group . In particular
O_S ⥲ Lie(G_{m,S}/S) and O_S ⥲ Lie(G_{m,S}/S).
5.2. Normalizers and centralizers
Let us first prove a few lemmas. Recall (cf. I 3.1.1) that a sequence of O_S-modules is
said to be exact if for every the sequence of -modules is
exact.
Likewise, a sequence of -groups is said to be exact if for every the sequence of groups is exact.
Lemma 5.2.1.93 Let be an exact sequence of -groups.
(i) The sequences and
1 ⟶ Lie(G′/S, M) ⟶ Lie(G/S, M) ⟶ Lie(G″/S, M) ⟶ 1
are then exact.
(ii) Let be a second sequence of groups; it is exact if and only if the sequence below is exact:
1 ⟶ G′ ×_S H′ ⟶ G ×_S H ⟶ G″ ×_S H″ ⟶ 1.
(iii) If two of the -groups satisfy (E), the third also satisfies (E).
(iv) If is an exact sequence of O_S-modules and if two of the modules are
good, the third is also.
(v) If two of the -groups are good, the third is also.
94 The first part of (i) is immediate, and the second part follows. Likewise, (ii) is immediate. Let us prove (iii). To abbreviate, write , , etc. Then one has the following commutative diagram
1 ─→ L′(M ⊕ N) ─→ L(M ⊕ N) ─→ L″(M ⊕ N) ─→ 1
│ │ │
↓ ↓ ↓
1 ─→ L′(M) ×_S L′(N) ─→ L(M) ×_S L(N) ─→ L″(M) ×_S L″(N) ─→ 1
in which the first (resp. the second) row is exact by (i) (resp. (i) and (ii)). Assertion (iii) follows.
Let us prove (iv). One has a commutative diagram
0 ─→ F′ ⊗_{O_S} T_{O_S/S}(M) ─→ F ⊗_{O_S} T_{O_S/S}(M) ─→ F″ ⊗_{O_S} T_{O_S/S}(M) ─→ 0
│ │ │
↓ ↓ ↓
0 ─────→ T_{F′/S}(M) ─────────→ T_{F/S}(M) ─────────→ T_{F″/S}(M) ─────→ 0
with exact rows (the first, because is a free O_S-module hence flat, the second by (i)). It follows
that if two of the modules are good, the third is also. Finally, (v) follows from (iii) and (iv).
Lemma 5.2.2. Let be an -group, E, H two -objects, a -O_S-module.
(i) The canonical homomorphism is an isomorphism.
(ii) If is a good O_S-module, so is .
95 The first assertion is immediate; let us prove the second. For every , one has a commutative diagram
F^G(I_{S′}(M)) ─────────────→ F(I_{S′}(M))
↑ φ ↑ ≃ φ_1
F^G(S′) ⊗_{O(S′)} O(I_{S′}(M)) ─→ F(S′) ⊗_{O(S′)} O(I_{S′}(M))
and one must show that is bijective; now it is evidently injective; let us show that it is surjective. Let be a basis of the free -module and let
u = Σ_{i=0}^n f_i ⊗ t_i
be an element of such that belongs to . Let us show that the belong to .
Let ; one can consider as above via the zero section . Then, for every , one has:
u_{S″} = g · u_{S″} = Σ_{i=0}^n g · (f_i)_{S″} ⊗ (t_i)_{S″}.
Since the form a basis of over , it follows that , whence for every .
Notations.96 If is an -group and a sub--group of , one denotes by the -functor that to every associates the set of classes , . If is a commutative -group then is a commutative -group.
Let now be an -group and a sub--group of ; set and . The adjoint action of on stabilizes , hence induces an action of on the -functor . Then, for every , one has:
(E/F)^K(S′) = { x̄ ∈ E(S′)/F(S′) | for every f : S″ → S′ and k ∈ K(S″),
f^*(x̄^{-1}) Ad(k)(f^*(x̄)) ∈ F(S″) },
where denotes the image of in .
Theorem 5.2.3. Let be an -group, a sub--group of . Write (I 2.3.3)
N = Norm_G(K), Z = Centr_G(K).
Make act on through the adjoint representation of .
(i) If the group law of is commutative97 98, then
Lie(N/S, M) / Lie(K/S, M) = ( Lie(G/S, M) / Lie(K/S, M) )^K.
(ii) If the group law of is commutative97, then
Lie(Z/S, M) = Lie(G/S, M)^K.
(iii) If satisfies (E) (resp. if and satisfy (E)), then satisfies (E) (resp. satisfies (E)).
(iv) Suppose is good; then is good; if moreover is very good, then is very good.
(v) Suppose and are good; then is good; if moreover is very good, then is very good.
To prove (i) and (ii)99 we shall use the following lemma, which follows from the diagram of exact sequences considered in 4.1.B (with and interchanged).
Lemma 5.2.3.0. Let be an -group, a sub--group of , and a free O_S-module of finite type.
Then and are sub--groups of and one has:
T_{H/S}(M) ∩_T Lie(G/S, M) = Lie(H/S, M),
where one has set T_{H/S}(M) ∩_T Lie(G/S, M) :=def T_{H/S}(M) ×_{T_{G/S}(M)} Lie(G/S, M).
Since the functors considered in (i) and (ii) commute with base extension, it suffices to show the equalities of -points.
Set (resp. ) and let . By the preceding lemma and the definition of and (cf. I 2.3.3), one has:
{ X ∈ Lie(G/S, M)(S) ⊂ G(I_S(M)) |
for every f : S′ → I_S(M) and u ∈ K(S′),
f^*(X^α) · u · f^*(X^{-α}) · u^{-1} ∈ K(S′),
resp. f^*(X) · u · f^*(X^{-1}) · u^{-1} = 1 (∗) },
where denotes the image of in .
To simplify the writing, let us write
g = Lie(G/S, M), k = Lie(K/S, M), n = Lie(N/S, M), z = Lie(Z/S, M).
If , the equalities (∗) are valid for every , since factors through (where is the structural morphism and the zero section). One deduces that
n(S)/k(S) ⊂ (g/k)^K(S) and z(S) ⊂ g^K(S).
To prove the reverse inclusions, suppose henceforth commutative; then and are commutative -groups. Let and its image in , suppose that (resp. ) and let us show that (resp. ).
Let ; let us show that the preceding condition (∗) is verified for every . Consider the cartesian square
p
I_{S′}(M) ──────→ I_S(M)
│ ρ′ │ ρ
↓ ρ ∘ f ↓
S′ ──────→ S
and let be the section of defined by . It suffices to show that, for every , one has
(∗∗) p^*(X^α) · v · p^*(X^{-α}) · v^{-1} ∈ K(I_{S′}(M)),
resp. p^*(X) · v · p^*(X^{-1}) · v^{-1} = 1.
Indeed, taking and applying to (∗∗), one obtains (∗), since and .
Let us now show (∗∗); for simplicity, we shall write instead of . Every is written in a unique manner where and . The expression then becomes which, since is commutative, is written . Now this is a priori in ; taking into account Lemma 5.2.3.0, the condition (∗∗) for therefore becomes: for every ,
⎰ Ad(k)(X) = X, if H = Z;
⎱ X^α Ad(k)(X^{-α}) ∈ Lie(K/S, M)(S′), if H = N.
When , this condition is indeed a consequence of the hypothesis . When , the condition is also written:
(∗′) Ad(k)(X) = X and Ad(k)(X^{-1}) = X^{-1},
now the second condition of (∗′) is a consequence of the first, since the action of on respects the group structure of this latter. Therefore, when , the condition (∗∗) for is indeed a consequence of the hypothesis . This proves (i) and (ii).
To prove (iii)–(v), let us write and define similarly , and . If satisfies (E), then and therefore, by 5.2.2 (i),
g(M ⊕ N)^K ≃ g(M)^K ×_S g(N)^K,
whence , so satisfies (E). If moreover satisfies (E), one obtains successively isomorphisms:
(g/k)(M ⊕ N) ≃ (g/k)(M) ×_S (g/k)(N),
(g/k)^K(M ⊕ N) ≃ (g/k)(M)^K ×_S (g/k)(N)^K,
then a commutative diagram:
0 ─→ k(M ⊕ N) ─→ n(M ⊕ N) ─→ (n/k)(M ⊕ N) ─→ 0
≀ │ │ ≀ │
↓ ↓ ↓
0 ─→ k(M) ×_S k(N) ─→ n(M) ×_S n(N) ─→ (n/k)(M) ×_S (n/k)(N) ─→ 0
from which it follows that , so satisfies (E).
Henceforth, write and define similarly k, z, n. If is good, is a good O_S-module
so, by 5.2.2 (ii), is also, so (which satisfies (E) by (iii)) is good. If moreover is good,
then is a good O_S-module so, by 5.2.1 (iv) and 5.2.2 (ii), so are and . Taking into account the
exact sequence
one obtains, by 5.2.1 (iv) again, that is good. Finally, if in addition to the preceding conditions, is very good, i.e., if one has identically for every , it is clear that and are very good. This proves (iii), (iv) and (v).
Corollary 5.2.3.1. One has Lie(Centr(G)/S) = Lie(G/S)^G if the group law of is commutative.
Corollary 5.2.3.2. If the group law of is commutative and if is a normal subgroup of , then
( Lie(G/S) / Lie(K/S) )^K = Lie(G/S) / Lie(K/S).
5.3. Linear representations
Let be a good -group acting linearly on a good O_S-module via
One has defined (4.7.1 and 4.8.2) a corresponding linear representation
ρ′ : Lie(G/S) ⟶ End_{O_S-mod.}(F).
The sub--groups and are defined for every part of , for example for every
sub-O_S-module of .
Definition 5.3.0. One will pose analogously: for every ,
Norm_{Lie(G/S)}(E)(S′) = { X ∈ Lie(G/S)(S′) | ρ′(X) E_{S′} ⊂ E_{S′} };
Centr_{Lie(G/S)}(E)(S′) = { X ∈ Lie(G/S)(S′) | ρ′(X) E_{S′} = 0 }.
(Note that this construction can be made for any linear representation of an O_S-"Lie algebra" (in the sense of 4.9)
and that the two subobjects constructed are sub-O_S-modules stable under the bracket.)
Theorem 5.3.1. Let be a good -group acting linearly on a good O_S-module and let be a
sub-O_S-module of .
(i) One has Lie(Centr_G(E)/S) = Centr_{Lie(G/S)}(E) and is a good -group; it is very good if
is.
(ii) Suppose that is a good O_S-module. Then100 Lie(Norm_G(E)/S) = Norm_{Lie(G/S)}(E) and
is a good -group; it is very good if is.
The proof is left to the reader.
5.3.2. Let be a good -group; what precedes applies in particular to the case where one takes for the adjoint representation of . Let be a good101 sub-module of ; to it one therefore associates two subgroups of , its centralizer and its normalizer. By 5.3.1, their Lie algebras are respectively the centralizer and the normalizer of in computed as usual by means of the bracket:
Centr_{Lie(G/S)}(E)(S′) = { X ∈ Lie(G/S)(S′) | [X, E_{S′}] = 0 },
Norm_{Lie(G/S)}(E)(S′) = { X ∈ Lie(G/S)(S′) | [X, E_{S′}] ⊂ E_{S′} }.
5.3.3. Let be a sub--group of , then is a sub-O_S-module of ; suppose that
is a good O_S-module101 (which is the case if is a good -group). One evidently has
Centr_G(K) ⊂ Centr_G(Lie(K/S)),
Norm_G(K) ⊂ Norm_G(Lie(K/S)),
whence, by 5.3.1,
Lie(Centr_G(K)/S) ⊂ Centr_{Lie(G/S)}(Lie(K/S)),
Lie(Norm_G(K)/S) ⊂ Norm_{Lie(G/S)}(Lie(K/S)),
but none of these four inclusions is a priori an equality; we shall see by the sequel many examples.
It follows in particular from these inclusions that if is a normal subgroup of , then is an ideal of .
6. Miscellaneous remarks
6.1. One can define the bracket of two infinitesimal automorphisms for an -functor which is not necessarily a
group. It suffices to apply the results of this Exposé to the group . To arrive at an
agreeable formalism, one is led to suppose good, i.e. to suppose that the O_X-module is good (if is
an -group, this definition evidently coincides with Definition 4.6).
6.2. There exist functors possessing infinitesimal endomorphisms that are not automorphisms, hence a fortiori not satisfying condition (E).102 For every pointed set , let be the free abelian monoid generated by and let be the abelian monoid obtained by quotienting by the equivalence relation generated by the relation . Then is the left adjoint of the forgetful functor from the category of abelian monoids to that of pointed sets; one will say that is the "free abelian monoid on the pointed set ".
Take then for the functor that to every scheme associates the free abelian monoid on the set , pointed by the zero element. Every morphism corresponds to an element of square zero of , and hence defines an endomorphism of by (sum in ). One thus obtains an endomorphism of , defined as follows. For every and ,
φ((x, f)) = (x + u_f, f).
If is the composite of the structural morphism and the zero section of , the corresponding element is , and so , since in . This shows that induces the identity on ; it is therefore an infinitesimal endomorphism of which is evidently not an automorphism.
6.3. There exist modules that are not good. On the one hand, one can modify the preceding counterexample slightly:103 take for the free -module with basis the elements of ; then does not satisfy condition (E) with respect to ; moreover, let be the -group defined by , where and is the -algebra of the abelian group ; then the -group is not commutative.
104 On the other hand, one can give the following counterexamples. Let , with a field of characteristic .
a) Let be the O_S-module which to every -scheme associates equipped with the
-module structure obtained by making scalars act through the -th power, that is to say, ,
for , . As -functor of groups, is isomorphic to . Therefore satisfies (E)
and is identified with . Then, the canonical morphism is,
for every , the identity map : it respects the abelian-group structure but not the module structure. Therefore is not good (cf. 4.4.1).
b) Let be the -functor of groups which to every -scheme associates
α_{p,k}(T) = { x ∈ O(T) | x^p = 0 };
it is represented by so it is a very good -group; it is also equipped with an
O_S-module structure, but it is not a good O_S-module, because the canonical morphism
is not bijective.
6.4. Let be a group-functor on . One has by definition the following implications
(G/S satisfies (E)) ⇐ (G is good) ⇐ (G is very good).
105 It would be interesting to prove or counter-exemplify the implications in the reverse direction.
6.5.106 Let Nil be the -functor of groups defined as follows: for every scheme ,
is the ideal of formed by the nilpotent elements, i.e.
Nil(S) = { x ∈ O(S) | there exists n ∈ ℕ such that x^n = 0 }.
(Nil is very good but is not representable.) Let , and be the -functors of
groups that to every scheme associate, respectively, the ideal and
O_{réd}(S) = O(S) / Nil(S), F(S) = O(S) / Nil(S)².
One sees easily that , so the -module is not good
(although it is a good -group). If M, N are -modules free of finite rank, one has
Nil²(I_S(M ⊕ N)) ≃ Nil²(S) ⊕ Nil(S) ⊗_ℤ M ⊕ Nil(S) ⊗_ℤ N
and therefore
F(I_S(M ⊕ N)) ≃ F(S) ⊕ O_{réd}(S) ⊗_ℤ M ⊕ O_{réd}(S) ⊗_ℤ N.
One deduces, on the one hand, that the -functor of groups satisfies condition (E) and, on the other hand, that ; since this latter is not a good -module, this shows that is an example of a -group that satisfies (E) but is not good.
6.5.1. Let us also give the counterexamples mentioned in 5.3.1–5.3.3. Let be a scheme, the good O_S-module
equipped with the natural action of the good -group , and the sub-O_S-module
of formed by the pairs such that is nilpotent. Set . Then
while, for every , one has
Norm_{Lie(G/S)}(E)(S′) = { ⎡a b⎤ | a, b, c, x ∈ O(S′), x nilpotent },
⎣x c⎦
so Lie(Norm_G(E)/S) ≠ Norm_{Lie(G/S)}(E).
By considering the semi-direct product , one obtains an analogous counterexample where is a
sub-O_S-module of ; moreover, with the notations introduced above, where is the
subgroup of (i.e., for every , is formed by the pairs such
that ).
Bibliography
[DG70] M. Demazure, P. Gabriel, Groupes Algébriques, Masson & North-Holland, 1970.
[HM69] G. Hochschild, G. D. Mostow, Automorphisms of affine algebraic groups, J. Algebra 13 (1969), 535–543.
[MO67] H. Matsumura, F. Oort, Representability of group functors and automorphisms of algebraic schemes, Invent. math. 4 (1967), 1–25.
Footnotes
Version of 14/10/2024.
Rendered in this edition by boldface characters F, G, V, W, cf. Exposé I. However, for functors such as
Norm and Centr (cf. 5.2) the underlines have been kept in the original, and they have been added for the functor
Lie, this to distinguish the functor from the -Lie algebra ,
used for example in Exposé VII_A.
One has detailed the points (2), (3), (4); in particular, (4) will be used in 3.11.
One has added the two preceding sentences.
One has added this remark, useful for Corollary 3.11.1.
Of course, if is a -group (not necessarily abelian) and if one sets , this makes a group above .
Note that has the same underlying space as .
Compare with 3.1.1.
One has detailed the original in what follows, and one has added the numbering 2.1.1 to 2.1.3.
In the sequel, one will be principally interested in the case where , acting by homotheties
on . For example, if is a free O_S-module of rank , then, for every ,
is identified with the set of -tuples
such that for all i, j, and one
has for every
.
One has added this paragraph, which will be useful in 3.4.2 and 4.6.2.
See also the addition 0.1.8 in Exp. III.
Cf. N.D.E. (33) of Exp. I.
When and is a -scheme, one has ; one thus recovers one of the usual definitions of the tangent bundle.
One has corrected " and " to: " and ".
One has added the paragraphs 3.1.1 and 3.1.2.
In fact, one is interested only in the action of (resp. ) on (resp. ), cf. below and the proof of 3.6.
One has added this remark.
Cf. EGA I, 5.3.11.
Cf. 3.1.2.
One has detailed the original in what follows.
One has added Remarks 3.4.2 and 3.4.3.
For examples of -functors and -groups not satisfying (E), see 6.2 and 6.3.
One has added what follows.
One has detailed what follows.
I.e., for every -morphism , the action of on corresponds, via the identification , to the natural action of on the right-hand term; this follows from the proof of SGA 1, III 5.1 (see also the addition 0.1.8 in Exp. III).
One has added the sentence that follows.
One has added the hypothesis that and be representable, and one has detailed what follows.
One has detailed what follows.
One has added this proof.
One has added this paragraph, in order to explain the introduction of the notion of --functor.
This is inspired by the notion of -space in topology.
One has added Corollary 3.9.0.2, which will be useful in 4.1.
Inspired by the standard proof showing that the fundamental group of an -space is abelian.
is equipped with the law , .
One has added this corollary, which amplifies Remark 4.1.1.2 below.
Setting , this implies in particular that is equipped with a natural structure of module over . This result has seemed somewhat surprising to the editors; for this reason one has detailed the proof.
One has added the sentences that follow.
One has added the sentence that follows.
One has added this remark.
One has simplified the original in what follows.
Via the isomorphism , this is equivalent to Remark 3.1.1. Likewise, 3.12 is equivalent to the first part of Remark 3.4.2.
One has detailed the original in what follows.
When , these are called "infinitesimal endomorphisms" of ; see 6.2 at the end of this Exposé.
One has detailed the original in what follows.
One has added the proof that follows.
One has added what follows.
One has added the reminder of 2.2.2, and detailed the sequel.
Moreover, this isomorphism is functorial in : if , setting , one has .
One has added the numbering 4.1.A and 4.1.B to highlight these definitions.
If is a monomorphism, so are and , cf. 3.7.
One has added this remark.
One has detailed the original in what follows, in order to show the action of on the functors and .
The statements 4.1.2, 4.1.3 and 4.1.4 are obtained more simply by remarking that the automorphisms of invariant by right translations are the left translations.108
One has added paragraph 4.2.0, whose results are used implicitly in 4.2 and 4.7 of the original.
One has inserted here this definition, which in the original appeared in 4.3.
One has detailed the original in what follows; in particular, one has added Definition 4.2.A and Remark 4.2.B.
This is Definition 4.2.0.1 applied to and to the -groups and .
One has added the sentences that follow and Proposition 4.2 bis, implicit in the original.
One has added the sentence that follows.
One has added the sentence that follows.
Note that on the right-hand term, this is the structure defined by , where denotes the action of on . This differs from the action , but coincides with it if .
One has added this remark, which will be useful in 4.5.1.
Cf. N.D.E. (61). Moreover, one has added the sentence that follows, as well as its proof.
One has changed to , because one must let S_2 and S_1 vary (cf.
below). Moreover, one has detailed the original in what follows.
One found in the original the assertion that when satisfies (E), the morphisms (6) (and hence (7))
are morphisms of O_S-modules, assertion contradicted by a counterexample given in 6.3. One has deleted this
assertion, inserted here the example just mentioned, and modified Definition 4.4 accordingly. This is without
consequence for the sequel.
One has added this remark, which will be useful in 4.6.2.
One has added this remark, which will be useful in 4.5 and 4.7.
Taking into account the modifications in Definition 4.4 (cf. N.D.E. (65)), one has inserted here this remark, which in the original figured in the proof of 4.4.1.
The original showed that (i) implies (ii) and (iii); one has added the implication (ii) ⇒ (i).
One has added what follows.
One has added the point (ii), immediate consequence of what precedes.
One has added this scholie, implicit in the original, which will be useful in 4.7.1. (Here and in the sequel, one writes , etc. for variables of square zero, the letter being reserved for the unit section of groups.)
One has added the sentence that follows.
One has added this lemma.
One has simplified the original, using the additions made in 4.3.1 and 4.5.
One has added the numbering 4.7.1, 4.7.2, 4.7.3.
One has added the sentence that follows.
One has detailed the original in what follows.
The original indicated in a footnote: "(∗) The author acknowledges, at Gabriel's request, that the exercise is not immediate; this is moreover the reason why it is not in the text". One has given the details, taking into account the addition made in 4.7.1; see also [DG70], § II.4, 4.2.
I.e. and play symmetric roles: the image in of [y, x] equals the commutator
, which is the inverse of . On
the other hand (cf. 4.10 below), one does not know whether one necessarily has : if one considers as
an element (resp. ) it is not a priori clear that and commute. .
.
One has detailed and simplified the original in what follows.
One has detailed the original in what follows.
Because one does not know whether , see 4.10 which follows.
One has corrected a sign error in the original, cf. [DG70], § II.4, 4.4 and 4.6.
One has detailed the original in what follows; in particular, one has added, for later references, the numbering 4.11.1 to 4.11.8.
One has detailed what follows, taking into account the additions made in Exp. I, § 6.8.
For this reason, the linear action of on could be called the "pre-adjoint
action"; in fact, by abuse of language, one will still speak of "the adjoint action" on . Let us
point out here a slightly more general construction, which will be used in Exp. III (cf. in particular III 0.8).
Suppose given, for every , an O_Y-module , and for every -morphism , a morphism
of O_Z-modules , which is functorial in (this is the case, for example,
if is a quasi-coherent ideal of O_S and if one sets , cf. Exp. III); then the action of on
induces a (generalized) "adjoint action" of on the -functor of abelian groups that to
every associates .
One has added the preceding description of , which will be useful later (for example, in VII_A, 5.5).
This description of in terms of invariant differentials will not be used in the sequel.
Because does not necessarily determine . For example, if ( a field) and if is the -group whose fiber at is and the other fibers are the unit group, then but .
Let be a morphism of O_S-modules and . If is a
monomorphism, the surjective morphism factors through O_S, so .
One has detailed the statement of the lemma (and its proof); this will be useful in 5.3.1.
One has added the proof of points (iii) and (v).
One has detailed the original in what follows.
One has detailed these notations.
Condition automatically verified if satisfies (E) (cf. 3.9), for example if is representable.
For an example where the -group is not commutative, see 6.3.
One has detailed the proof, adding in particular Lemma 5.2.3.0, implicit in the original.
One has corrected the original, which stated the equality that follows without hypotheses on ; see 6.5.1 for counterexamples.
One has added this hypothesis; cf. 6.5.1.
One has detailed the original in what follows. In particular, the correct definition of the "free abelian monoid on a pointed set" was pointed out to us by O. Gabber.
One has added what follows.
In what follows, one has detailed example a) and added example b).
And is very good if it is representable (4.11). For representability criteria, see for example [MO67]. On the other hand, for the automorphisms of an affine algebraic group (over an algebraically closed field of characteristic 0), let us cite [HM69].
One has added this paragraph.
Additional references cited in this Exposé.
See for example the proof of Propositions 4.6 and 2.5 of [DG70], § II.4.