§10. Formal Schemes
10.1. Formal affine schemes
(10.1.1) Let be an admissible topological ring (0, 7.1.2); for every ideal of definition of , is identified with the closed subspace of (1.1.11), the set of open prime ideals of ; this topological space does not depend on the ideal of definition considered; let us denote it by . Let be a fundamental system of neighborhoods of in , formed of ideals of definition, and for each , let be the structure sheaf of ; this sheaf is induced on by (which is zero off ). For , the canonical homomorphism therefore defines a homomorphism of sheaves of rings (1.6.1), and is a projective system of sheaves of rings for these homomorphisms. Since the topology of admits a base formed of quasi-compact opens, one may associate to each a sheaf of pseudo-discrete topological rings (0, 3.8.1) which has as its underlying sheaf of rings (without topology), and which we shall again denote by ; and the again form a projective system of sheaves of topological rings (0, 3.8.2). We shall designate by the sheaf of topological rings on , the projective limit of the system ; for every quasi-compact open of , is therefore the topological ring projective limit of the system of discrete rings (0, 3.2.6).
Definition (10.1.2). Given an admissible topological ring , one calls formal spectrum of , and denotes by , the closed subspace of formed of the open prime ideals of . One says that a topologically ringed space is a formal affine scheme if it is isomorphic to a formal spectrum equipped with the sheaf of topological rings , projective limit of the sheaves of pseudo-discrete rings , where runs through the filtered set of ideals of definition of .
When we speak of a formal spectrum as a formal affine scheme, it will always be a question of the topologically ringed space where is defined as above.
One will note that every affine scheme may be considered as a formal affine scheme in one and only one way, by considering as a discrete topological ring: the topological rings are then discrete when is quasi-compact (but not in general when is an arbitrary open of ).
Proposition (10.1.3). If , where is an admissible ring, then is topologically isomorphic to .
Proof. Indeed, since is closed in , it is quasi-compact, and consequently is topologically isomorphic to the projective limit of the discrete rings ; but is isomorphic to (1.3.7); since is separated and complete, it is topologically isomorphic to (0, 7.2.1), whence the proposition.
Proposition (10.1.4). Let be an admissible ring, , and for every , let ; the topologically ringed space is isomorphic to the formal affine spectrum (0, 7.6.15).
Proof. For every ideal of definition of , the discrete ring is identified canonically with (0, 7.6.9); hence (1.2.5 and 1.2.6) the topological space is identified canonically with . Moreover, for every quasi-compact open of contained in , is identified with the module of sections of the structure sheaf of over (1.3.6); hence, if one sets , then is identified with the module of sections , which proves the proposition.
(10.1.5) As a sheaf of rings without topology, the structure sheaf of admits, for every , a stalk which, by virtue of (10.1.4), is identified with the inductive limit for . Consequently (0, 7.6.17 and 7.6.18):
Proposition (10.1.6). For every , the stalk is a local ring whose residue field is isomorphic to . If in addition is adic and Noetherian, is a Noetherian ring.
Since is not reduced to , one concludes from this result that the support of the sheaf of rings is equal to .
10.2. Morphisms of formal affine schemes
(10.2.1) Let , be two admissible rings, and let be a continuous homomorphism. The continuous map (1.2.1) then sends into , for the inverse image under of an open prime ideal of is an open prime ideal of . On the other hand, for every , defines a continuous homomorphism by virtue of (10.1.4), (10.1.3), and (0, 7.6.7); since these homomorphisms satisfy the conditions of compatibility for the restrictions corresponding to the passage from to a multiple of , and since , they define a continuous homomorphism of sheaves of topological rings (0, 3.2.5), which we shall again denote by ; one has thus obtained a morphism of topologically ringed spaces . One will note that, as a homomorphism of sheaves of rings without topology, defines a homomorphism on the stalks, for every .
Proposition (10.2.2). Let , be two admissible topological rings, and let , . For a morphism of topologically ringed spaces to be of the form , where is a continuous ring homomorphism , it is necessary and sufficient that for every , be a local homomorphism .
Proof. The condition is necessary: indeed, let , and let ; if , one has , and it is immediate that the homomorphism deduced from (0, 7.6.7) transforms into a part of ; passing to the inductive limit, one sees therefore (taking account of (10.1.5) and of (0, 7.6.17)) that is a local homomorphism.
Conversely, let be a morphism verifying the condition of the statement; by virtue of (10.1.3), defines a continuous ring homomorphism By virtue of the hypothesis on , for the section of over to have an invertible germ at the point , it is necessary and sufficient that have an invertible germ at the point . But by virtue of (0, 7.6.17), the sections of (resp. ) over (resp. ) whose germ is not invertible at the point (resp. ) are exactly the elements of (resp. ); the preceding remark thus shows that . Finally, for every , the diagram is commutative; by the universal property of the complete rings of fractions (0, 7.6.6), is equal to for every , hence (0, 3.2.5) one has .
We shall say that a morphism of topologically ringed spaces verifying the condition of (10.2.2) is a morphism of formal affine schemes. One may say that the functors in and in define an equivalence between the category of admissible rings and the dual category of the category of formal affine schemes (T, I, 1.2).
(10.2.3) As a particular case of (10.2.2), let us note that, for , the canonical injection of the formal affine scheme induced by on corresponds to the canonical continuous homomorphism . Under the hypotheses of (10.2.2), let be an element of and an element of , a multiple of ; one then has ; the restriction of to , considered as a morphism of into , is the unique morphism making commutative the diagram This morphism corresponds to the unique continuous homomorphism (0, 7.6.7) making commutative the diagram
10.3. Ideals of definition of a formal affine scheme
(10.3.1) Let be an admissible ring, an open ideal of , the formal affine scheme . Let be the set of ideals of definition of contained in ; then is a sheaf of ideals of . Let us designate by the projective limit of the sheaves induced on by , which is identified with a sheaf of ideals of (0, 3.2.6). For every , is the projective limit of , in other words is identified with the open ideal of the ring (0, 7.6.9), and in particular ; one concludes (the forming a base for the topology of ) that one has
(10.3.2) With the notations of (10.3.1), for every , the canonical map of into is surjective and has for kernel (0, 7.6.9); these maps therefore define a continuous surjective homomorphism, said to be canonical, of the sheaf of topological rings onto the sheaf of discrete rings whose kernel is ; this homomorphism is moreover none other than (10.2.1), where is the continuous homomorphism ; the morphism of formal affine schemes (where is moreover the identity homeomorphism of onto itself) is again said to be canonical. One therefore has, by what precedes, a canonical isomorphism
It is clear (by virtue of ) that the map is strictly increasing; by what precedes, for , the sheaf is canonically isomorphic to .
(10.3.3) The hypotheses and notations still being those of (10.3.1), we shall say that a sheaf of ideals of is a sheaf of ideals of definition of (or an Ideal of definition of ) if, for every , there exists an open neighborhood of of the form , where , such that is of the form , where is an ideal of definition of .
Proposition (10.3.4). For every , every Ideal of definition of induces an Ideal of definition of .
Proof. This results from (10.3.1.1).
Proposition (10.3.5). If is an admissible ring, every Ideal of definition of is of the form , where is an ideal of definition of , uniquely determined.
Proof. Indeed, let be an Ideal of definition of ; by hypothesis, and since is quasi-compact, there is a finite number of elements such that the cover and that , where is an ideal of definition of . For each , there exists therefore an open ideal of such that (0, 7.6.9); let be an ideal of definition of contained in all the . The canonical image of in the structure sheaf of (10.3.2) is then such that its restriction to is equal to that of ; one concludes that this canonical image is a quasi-coherent sheaf on , hence of the form , where is an ideal of containing (1.4.1), whence (10.3.2); moreover, since for each there exists an integer such that , one will have, denoting by the largest of the , , and consequently (10.3.2) , whence finally (1.3.13), which proves that is an ideal of definition of (0, 7.1.4).
Proposition (10.3.6). Let be an adic ring, an ideal of definition of such that is an -module of finite type. For every integer , one then has .
Proof. Indeed, for every , one has (since is an open ideal) by virtue of (10.3.1.1) and of (0, 7.6.12). Since is associated to the presheaf (0, 4.1.6), the corollary results from this, since the form a base for the topology of .
(10.3.7) One says that a family of Ideals of definition of is a fundamental system of Ideals of definition if every Ideal of definition of contains one of the ; since , it amounts to the same to say that the form a fundamental system of neighborhoods of in . Let be a family of elements of such that the cover . If is a filtered decreasing family of ideals of such that for every , the family is a fundamental system of Ideals of definition of , then is a fundamental system of Ideals of definition of .
Proof. Indeed, for every Ideal of definition of , there is a finite covering of by sets such that, for each , is an Ideal of definition of contained in . If is an index such that for all , it results from (10.3.3) that is an Ideal of definition of , evidently contained in , whence our assertion.
10.4. Formal preschemes and morphisms of formal preschemes
(10.4.1) Given a topologically ringed space , one says that an open is a formal affine open (resp. an adic formal affine open, resp. a Noetherian formal affine open) if the topologically ringed space induced by on is a formal affine scheme (resp. such a scheme whose ring is adic, resp. adic and Noetherian).
Definition (10.4.2). One calls formal prescheme a topologically ringed space every point of which admits a formal affine open neighborhood. One says that the formal prescheme is adic (resp. locally Noetherian) if every point of admits an adic (resp. Noetherian) formal affine open neighborhood. One says that is Noetherian if it is locally Noetherian and if its underlying space is quasi-compact (hence Noetherian).
Proposition (10.4.3). If is a formal prescheme (resp. locally Noetherian), the formal affine opens (resp. Noetherian affine opens) form a base for the topology of .
Proof. This results from (10.4.2) and (10.1.4), taking account of the fact that if is an adic Noetherian ring, so is for every (0, 7.6.11).
Corollary (10.4.4). If is a formal prescheme (resp. a locally Noetherian formal prescheme, resp. Noetherian), the topologically ringed space induced on every open of is again a formal prescheme (resp. a locally Noetherian formal prescheme, resp. Noetherian).
Definition (10.4.5). Given two formal preschemes , , one calls morphism (of formal preschemes) of into every morphism of topologically ringed spaces such that, for every , is a local homomorphism .
It is immediate that the composite of two morphisms of formal preschemes is again such a morphism; the formal preschemes therefore form a category, and one will denote by the set of morphisms of a formal prescheme into a formal prescheme .
If is an open part of , the canonical injection into of the formal prescheme induced by on is a morphism of formal preschemes (and even a monomorphism of topologically ringed spaces (0, 4.1.1)).
Proposition (10.4.6). Let be a formal prescheme, a formal affine scheme. There exists a canonical one-to-one correspondence between the morphisms of the formal prescheme into the formal prescheme and the continuous homomorphisms of the ring into the topological ring .
Proof. The demonstration is the same as that of (2.2.4), replacing “homomorphism” by “continuous homomorphism”, “affine open” by “formal affine open”, and using (10.2.2) instead of (1.7.3); we leave the details to the reader.
(10.4.7) Given a formal prescheme , one says that the datum of a formal prescheme and of a morphism defines a formal prescheme over , or a -formal prescheme, being called the structure morphism of the -prescheme . If , where is an admissible ring, one also says that the -formal prescheme is an -formal prescheme or a formal prescheme over . An arbitrary formal prescheme may always be considered as a formal prescheme over (equipped with the discrete topology).
If , are two -formal preschemes, one says that a morphism is a -morphism if the diagram (where the oblique arrows are the structure morphisms) is commutative. With this definition, the -formal preschemes (for fixed ) form a category. One designates by the set of -morphisms of the -formal prescheme into the -formal prescheme . When , one also says -morphism instead of -morphism.
(10.4.8) Since every affine scheme may be considered as a formal affine scheme (10.1.2), every (ordinary) prescheme may be considered as a formal prescheme. It moreover results from (10.4.5) that for ordinary preschemes, the morphisms (resp. -morphisms) of formal preschemes coincide with the morphisms (resp. -morphisms) defined in §2.
10.5. Ideals of definition of formal preschemes
(10.5.1) Let be a formal prescheme; one says that an -Ideal is a sheaf of ideals of definition (or an Ideal of definition) of if every possesses a formal affine open neighborhood such that is an Ideal of definition of the formal affine scheme induced by on (10.3.3); by virtue of (10.3.1.1) and (10.4.3), for every open , is then an Ideal of definition of the formal prescheme induced by on .
One says that a family of Ideals of definition of is a fundamental system of Ideals of definition if there exists a covering of by formal affine opens such that, for every , the family of is a fundamental system of Ideals of definition (10.3.6) of the formal affine scheme induced by on . It results from the final remark of (10.3.7) that when is a formal affine scheme, this definition coincides with the definition given in (10.3.7). For every open of , the restrictions then form a fundamental system of Ideals of definition of the formal prescheme induced on , by virtue of (10.3.1.1). If is a locally Noetherian formal prescheme, and an Ideal of definition of , it results from (10.3.6) that the powers form a fundamental system of Ideals of definition of .
(10.5.2) Let be a formal prescheme, an Ideal of definition of . Then the ringed space is an (ordinary) prescheme, which is affine (resp. locally Noetherian, resp. Noetherian) when is a formal affine scheme (resp. a locally Noetherian formal prescheme, resp. Noetherian); one is indeed at once reduced to the affine case, and then the proposition has already been demonstrated in (10.3.2). Moreover, if is the canonical homomorphism, is a morphism (said to be canonical) of formal preschemes , for here again, this has been seen in the affine case (10.3.2), to which one reduces at once.
Proposition (10.5.3). Let be a formal prescheme, a fundamental system of Ideals of definition of . Then the sheaf of topological rings is the projective limit of the sheaves of pseudo-discrete rings (0, 3.8.1) .
Proof. Since the topology of admits a base of quasi-compact formal affine opens (10.4.3), one is reduced to the affine case, where the proposition is a consequence of (10.3.5), (10.3.2), and of the definition (10.1.1).
It is not certain that every formal prescheme admits Ideals of definition. However:
Proposition (10.5.4). Let be a locally Noetherian formal prescheme. There exists a largest Ideal of definition of ; it is the only Ideal of definition such that the prescheme is reduced. If is an Ideal of definition of , is the inverse image under of the Nilradical of .
Proof. Suppose first that , where is an adic Noetherian ring. The existence and properties of result immediately from (10.3.4) and (5.1.1), taking account of the existence and properties of the largest ideal of definition of (0, 7.1.6 and 7.1.7).
To prove the existence and properties of in the general case, it suffices to show that if are two Noetherian formal affine opens of , the largest Ideal of definition of induces the largest Ideal of definition of ; but since is reduced, this results from what precedes.
One designates by the (ordinary) reduced prescheme .
Corollary (10.5.5). Let be a locally Noetherian formal prescheme, the largest Ideal of definition of ; for every open of , is the largest Ideal of definition of the formal prescheme induced by on .
Proposition (10.5.6). Let , be two formal preschemes, (resp. ) an Ideal of definition of (resp. ), a morphism of formal preschemes.
(i) If , there exists a unique morphism of ordinary preschemes making commutative the diagram
where the vertical arrows are the canonical morphisms.
(ii) Suppose that , are formal affine schemes, , , where (resp. ) is an ideal of definition of (resp. ), and , where is a continuous homomorphism; for , it is necessary and sufficient that , and is then the morphism , where is the homomorphism deduced from by passage to the quotients.
Proof. (i) If , the hypothesis entails that the image under of the sheaf of ideals of is contained in (0, 4.3.5). By passage to the quotients, one therefore deduces from a homomorphism of sheaves of rings moreover, since for every , is a local homomorphism, so is . The morphism of ringed spaces is therefore (2.2.1) the unique morphism of ringed spaces answering the question.
(ii) The canonical functorial correspondence between morphisms of formal affine preschemes and continuous homomorphisms of rings (10.2.2) shows that in the case considered, the relation entails that one has , where is the unique homomorphism making commutative the diagram
The existence of therefore implies that . Conversely, if this condition is verified, then, denoting by the unique homomorphism making commutative the diagram (10.5.6.2) and setting , it is clear that the diagram (10.5.6.1) is commutative; the consideration of the homomorphisms and corresponding to and respectively then shows that this entails the relation .
It is clear that the correspondence defined above is functorial.
10.6. Formal preschemes as inductive limits of preschemes
(10.6.1) Let be a formal prescheme, a fundamental system of Ideals of definition of ; for each , let be the canonical morphism (10.5.2); for , the canonical homomorphism defines a canonical morphism of (ordinary) preschemes such that one has . The preschemes and the morphisms therefore constitute (by virtue of (10.4.8)) an inductive system in the category of formal preschemes.
Proposition (10.6.2). With the notations of (10.6.1), the formal prescheme and the morphisms constitute an inductive limit (T, I, 1.8) of the system in the category of formal preschemes.
Proof. Let be a formal prescheme, and for each index , let be a morphism such that one has for . This last condition and the definition of the entail first that all the are identical to one and the same continuous map of the underlying spaces; moreover, the homomorphisms form a projective system of homomorphisms of sheaves of rings. By passage to the projective limit, one therefore deduces a homomorphism , and it is clear that the morphism of ringed spaces is the only one making commutative the diagrams
It remains therefore to prove that is a morphism of formal preschemes; the question being local on and , one may suppose , , and being admissible rings, with , where is a fundamental system of ideals of definition of (10.3.5); since , the existence of a morphism of formal affine schemes making commutative the diagrams (10.6.2.1) results then from the one-to-one correspondence (10.2.2) between morphisms of formal affine schemes and continuous homomorphisms of rings, and from the definition of the projective limit. But the uniqueness of as a morphism of ringed spaces shows that it coincides with the morphism denoted the same way at the beginning of the demonstration.
The following proposition establishes, under certain supplementary conditions, the existence of the inductive limit of a given inductive system of (ordinary) preschemes in the category of formal preschemes:
Proposition (10.6.3). Let be a topological space, a projective system of sheaves of rings on , having for index set. Let be the kernel of . Suppose that:
a) the ringed space is a prescheme .
b) for every and every , there exists an open neighborhood of in such that the restriction is nilpotent.
c) the homomorphisms are surjective.
Let be the sheaf of topological rings projective limit of the sheaves of pseudo-discrete rings , and let be the canonical homomorphism. Then the topologically ringed space is a formal prescheme; the homomorphisms are surjective; their kernels form a fundamental system of Ideals of definition of , and is the projective limit of the sheaves of ideals .
Proof. Let us first note that on each stalk, is a surjective homomorphism and a fortiori a local homomorphism; hence is a morphism of preschemes () (2.2.1). Suppose first that each is an affine scheme with ring . There exists a ring homomorphism such that (1.7.3); consequently (1.6.3), the sheaf is a quasi-coherent -Module on (for the external law defined by ), associated to considered as an -module by means of . For every , let ; by hypothesis, the opens and are identical in , and the homomorphism of into corresponding to is none other than (1.6.1). But when one considers as an -module, is the -module , hence one also has when is this time considered as a homomorphism of -modules. Then, since is surjective, one concludes that is also (1.3.9), and if is the kernel of , the kernel of is a quasi-coherent -Module equal to . In particular, one has , where is the kernel of . Hypothesis b) entails that is nilpotent: indeed, since is quasi-compact, one may cover by a finite number of opens such that , and taking for the largest of the , one has . One concludes that is nilpotent (1.3.13). Then the ring is admissible (0, 7.2.2), the canonical homomorphism is surjective and its kernel is equal to the projective limit of the for ; the form a fundamental system of neighborhoods of in . The assertions of (10.6.3) result in this case from (10.1.1) and (10.3.2), being none other than .
Still in this same particular case, let us note that if is an element of the projective limit , all the opens (affine open in ) are identified with the open of , the prescheme induced by on being thus identified with the affine scheme .
In the general case, let us remark first that for every quasi-compact open of , each of the is nilpotent, as the reasoning made above shows. We shall see that for every , there is an open neighborhood of in which is an affine open for all the . Indeed, take affine open for , and observe that . Since is nilpotent, by virtue of what precedes, is affine open also for each by virtue of (5.1.9). This being so, for every satisfying the preceding conditions, the study of the affine case made above shows that is a formal prescheme of which the form a fundamental system of ideals of definition and is the projective limit of the ; whence the conclusion.
Corollary (10.6.4). Suppose that for , the kernel of is and that is of finite type over . Then is an adic formal prescheme, and if is the kernel of , one has and is isomorphic to . If in addition is locally Noetherian (resp. Noetherian), is locally Noetherian (resp. Noetherian).
Proof. Since the spaces underlying and are the same, the question is local and one may suppose all the affine; taking account of the relations (with the notations of (10.6.3)), one is at once reduced to the corresponding assertions of (0, 7.2.7 and 7.2.8), noting that is then an -module of finite type (1.3.9).
In particular, every locally Noetherian formal prescheme is the inductive limit of a sequence of (ordinary) locally Noetherian preschemes verifying the conditions of (10.6.3) and (10.6.4): it suffices to consider an Ideal of definition of (10.5.4) and to take ((10.5.1) and (10.6.2)).
Corollary (10.6.5). Let be an admissible ring. For the formal affine scheme to be Noetherian, it is necessary and sufficient that be adic and Noetherian.
Proof. The condition is evidently sufficient. Conversely, suppose that is Noetherian, and let be an ideal of definition of , the corresponding Ideal of definition of . The (ordinary) preschemes are then affine and Noetherian, hence the rings are Noetherian (6.1.3), whence one concludes that is an -module of finite type. Since the form a fundamental system of Ideals of definition of (10.5.1), one has (10.5.3); one concludes (10.1.3) that is topologically isomorphic to , hence is adic and Noetherian (0, 7.2.8).
Remark (10.6.6). With the notations of (10.6.3), let be an -Module, and suppose given, for , a -morphism , so that for . Since the continuous map underlying is the identity, is a homomorphism of sheaves of abelian groups on the space ; moreover, if is the projective limit of the projective system of sheaves of abelian groups, the fact that the are -morphisms permits one to define on a structure of -Module by passage to the projective limit; equipped with this structure, we shall say that is the projective limit (for the ) of the system of -Modules . In the particular case where and where is the identity, we shall say for brevity that is the projective limit of a system such that for (without mentioning the ).
(10.6.7) Let , be two formal preschemes, (resp. ) an Ideal of definition of (resp. ), a morphism such that . One then has, for every integer , ; one may therefore (10.5.6) deduce from a morphism of (ordinary) preschemes , on setting , , and it results at once from the definitions that the diagrams
are commutative for ; in other words, is an inductive system of morphisms.
(10.6.8) Conversely, let (resp. ) be an inductive system of (ordinary) preschemes satisfying the conditions b) and c) of (10.6.3), and let (resp. ) be its inductive limit. By definition of inductive limits, every sequence of morphisms forming an inductive system admits an inductive limit , which is the unique morphism of formal preschemes making commutative the diagrams
Proposition (10.6.9). Let , be two locally Noetherian formal preschemes, (resp. ) an Ideal of definition of (resp. ); the map defined in (10.6.7) is a bijection of the set of morphisms such that onto the set of sequences of morphisms making commutative the diagrams (10.6.7.1).
Proof. If is the inductive limit of such a sequence, it must be shown that . The question being local on and , one may restrict oneself to the case where , are affine, and being adic Noetherian, , , where (resp. ) is an ideal of definition of (resp. ). One then has , , with and , by virtue of (10.3.6) and (10.3.2); , where the homomorphisms form a projective system, hence , where . The commutativity of the diagram (10.6.7.1) for then gives the condition for every , hence, passing to the projective limit, , and this entails (10.5.6, (ii)).
Corollary (10.6.10). Let , be two locally Noetherian formal preschemes, the largest Ideal of definition of (10.5.4).
(i) For every Ideal of definition of and every morphism , one has .
(ii) There is a canonical one-to-one correspondence between and the set of sequences of morphisms making commutative the diagrams (10.6.7.1), where , .
Proof. (ii) results at once from (i) and (10.6.9). To demonstrate (i), one may restrict oneself to the case where , , and being Noetherian, , , where is the largest ideal of definition of and an ideal of definition of . Let , where is a continuous homomorphism; since the elements of are topologically nilpotent (0, 7.1.4, (ii)), so are those of , hence since is the set of topologically nilpotent elements of (0, 7.1.6); whence the conclusion by virtue of (10.5.6, (ii)).
Corollary (10.6.11). Let , , be three locally Noetherian formal preschemes, , morphisms making and into -formal preschemes. Let (resp. , ) be an Ideal of definition of (resp. , ), and suppose that , ; set , , . There is then a canonical one-to-one correspondence between and the set of sequences of -morphisms making commutative the diagrams (10.6.7.1).
Proof. For every -morphism , one has by definition , hence the corollary thus follows from (10.6.9).
One will note that, for , the datum of a morphism determines one and only one morphism making commutative the diagram (10.6.7.1), as one sees at once by reducing to the affine case; one has thus defined a map and the form for the a projective system of sets; (10.6.11) may be stated again by saying that there exists a canonical bijection
10.7. Product of formal preschemes
(10.7.1) Let be a formal prescheme; the -formal preschemes forming a category, one may define the notion of product of -formal preschemes.
Proposition (10.7.2). Let , be two formal affine schemes over a formal affine scheme . Let , , the -morphisms corresponding (10.2.2) to the canonical (continuous) -homomorphisms and of and into ; then is a product of the -formal affine schemes and .
Proof. By virtue of (10.4.6), everything reduces to verifying that if, to every continuous -homomorphism , where is an admissible ring which is a topological -algebra, one associates the pair , one defines a bijection which is none other than the universal property of the completed tensor product (0, 7.7.6).
Proposition (10.7.3). Given two -formal preschemes , , the product exists.
Proof. The demonstration is identical to that of (3.2.6), replacing therein the affine schemes (resp. the affine opens) by the formal affine schemes (resp. the formal affine opens), and the prop. (3.2.2) by (10.7.2).
All the formal properties of the product of preschemes (3.2.7 and 3.2.8, 3.3.1 to 3.3.12) are valid without any modification for the product of formal preschemes.
(10.7.4) Let , , be three formal preschemes and let , be two morphisms. Suppose that there exist in , , respectively three fundamental systems of Ideals of definition , , respectively, having the same index set , such that and for every . Set , , ; for , , , note that (resp. , ) is a closed subprescheme of (resp. , ) having the same underlying space (10.6.1). Since is a monomorphism of preschemes, one sees first that the products and are identical (3.2.4), then that is identified with a closed subprescheme of having the same underlying space (4.3.1). This being so, the product is the inductive limit of the ordinary preschemes : indeed, one sees as in (10.6.2) that one may reduce to the case where , , and are formal affine schemes. Taking account of (10.5.6, (ii)) and of the hypothesis on the fundamental systems of Ideals of definition of , , and , one sees at once that our assertion results from the definition of the completed tensor product of two algebras (0, 7.7.1).
Moreover, let be an -formal prescheme, a fundamental system of Ideals of definition of having as index set, , two -morphisms such that and . If one sets , and if and are the -morphisms corresponding to and (10.5.6), one verifies at once that is the inductive limit of the -morphisms .
The considerations of this number apply in particular when , , and are locally Noetherian, taking as fundamental systems of Ideals of definition the systems formed of the powers of an Ideal of definition (10.5.1). But one will note that is not necessarily locally Noetherian (see however (10.13.5)).
10.8. Formal completion of a prescheme along a closed part
(10.8.1) Let be a locally Noetherian (ordinary) prescheme, a closed part of the space underlying ; designate by the set of coherent sheaves of ideals in such that the support of is . The set is not empty (5.2.1, 4.1.4, and 6.1.1); we shall order it by the relation .
Lemma (10.8.2). The ordered set is filtered; if is Noetherian, then for every , the set of powers () is cofinal with .
Proof. Indeed, if and belong to , and if one sets , is coherent since is coherent (6.1.1 and 0, 5.3.4), and one has for every , hence for and for , which proves that . On the other hand, if is Noetherian, and if and belong to , there exists an integer such that (9.3.4), which signifies that .
(10.8.3) Let now be a coherent -Module; for every , is a coherent -Module (9.1.1), of support contained in , and which we shall most often identify with its restriction to . When runs through , these sheaves form a projective system of sheaves of abelian groups.
Definition (10.8.4). Given a closed part of a locally Noetherian prescheme and a coherent -Module , one calls completion of along and designates by or by (when no confusion is possible) the restriction to of the sheaf ; one says that its sections over are the formal sections of along .
It is immediate that for every open , one has .
By passage to the projective limit, it is clear that is a sheaf of rings, and that may be considered as an -Module. Moreover, since there exists a base for the topology of formed of quasi-compact opens, one may consider (resp. ) as a sheaf of topological rings (resp. of topological groups) projective limit of the sheaves of pseudo-discrete rings (resp. groups) (resp. ); by passage to the projective limit, then becomes a topological -Module (0, 3.8.1 and 3.8.2); let us recall that for every quasi-compact open , (resp. ) is then the projective limit of the discrete rings (resp. groups) (resp. ).
If now is a homomorphism of -Modules, one deduces from it canonically homomorphisms for every , and these homomorphisms form a projective system. By passage to the projective limit and restriction to , they therefore give a continuous -homomorphism , denoted or and called the completion of the homomorphism along . It is clear that if is a second homomorphism of -Modules, one has , hence is a covariant additive functor in , from the category of coherent -Modules, with values in the category of topological -Modules.
Proposition (10.8.5). The support of is ; the topologically ringed space is a locally Noetherian formal prescheme, and if , is an Ideal of definition of this formal prescheme. If is an affine scheme with Noetherian ring, where is an ideal of , and , then is canonically identified with , where is the separated completion of for the -preadic topology.
Proof. One may evidently restrict oneself to proving the last assertion. One knows (0, 7.3.3) that the separated completion of for the -preadic topology is identified with the ideal of , and that is a -adic Noetherian ring such that (0, 7.2.6). This last relation shows that the open prime ideals of are the ideals , where is a prime ideal of containing , and that one has , whence . Since , the proposition follows at once from the definitions.
One says that the formal prescheme thus defined is the completion of along and one denotes it by or if no confusion is to be feared. When one takes , one may take , and one therefore has .
It is clear that if is a subprescheme induced on an open of , is canonically identified with the formal subprescheme induced by on the open of .
Corollary (10.8.6). The (ordinary) prescheme is the unique reduced subprescheme of having for underlying space (5.2.1). For to be Noetherian, it is necessary and sufficient that be, and it is sufficient that be.
Proof. The determination of being local (10.5.4), one may again suppose that is an affine scheme with Noetherian ring; with the notations of (10.8.5), the ideal of topologically nilpotent elements of is the inverse image under the canonical map of the nilradical of (0, 7.1.3), hence is isomorphic to the quotient of by its nilradical. The first assertion therefore results from (10.5.4) and (5.1.1). If is Noetherian, its underlying space is also, hence the are Noetherian (6.1.2) and so is (10.6.4); the converse is immediate, by virtue of (6.1.2).
(10.8.7) The canonical homomorphisms (for ) form a projective system and therefore give, by passage to the projective limit, a homomorphism of sheaves of rings , denoting by the canonical injection of the underlying spaces. We shall designate by (or ) the morphism (said to be canonical) of ringed spaces.
By tensorization, for every coherent -Module , the canonical homomorphisms give homomorphisms of -Modules which again form a projective system, and therefore give, by passage to the projective limit, a canonical functorial homomorphism of -Modules.
Proposition (10.8.8). (i) The functor (in ) is exact.
(ii) The functorial homomorphism of -Modules is an isomorphism.
Proof. (i) It suffices to prove that if is an exact sequence of coherent -Modules, and an affine open of , with Noetherian ring , the sequence is exact. One then has , , , where , , are three -modules of finite type such that the sequence is exact (1.5.1 and 1.3.11); let and let be an ideal of such that . One then has (1.3.12); hence, by definition of the projective limit, one has the separated completion of for the -preadic topology, and likewise our assertion then results from the fact that when is Noetherian, the functor in is exact on the category of -modules of finite type (0, 7.3.3).
(ii) The question being local, one may suppose that one has an exact sequence (0, 5.3.2); since is functorial, and the functors and are right exact (by (i) and (0, 4.3.1)), one has the commutative diagram whose lines are exact. Moreover, the two functors and commute with finite direct sums (0, 3.2.6 and 4.3.2) and one is therefore reduced to demonstrating our assertion for . One then has (0, 4.3.4), and is a homomorphism of -Modules; it therefore suffices to verify that transforms the unit section of over an open of into itself, which is immediate and therefore shows that in this case is the identity.
Corollary (10.8.9). The morphism of ringed spaces is flat.
Proof. This indeed results from (0, 6.7.3) and from (10.8.8, (i)).
Corollary (10.8.10). If and are coherent -Modules, there exist canonical functorial isomorphisms (in and )
Proof. This results from the canonical identification of and of ; the existence of the first isomorphism is then a result valid for all morphisms of ringed spaces (0, 4.3.3.1) and that of the second a result valid for all flat morphisms (0, 6.7.6), hence follows from (10.8.9).
Proposition (10.8.11). For every coherent -Module , the kernel of the canonical homomorphism deduced from is formed of the sections null in a neighborhood of .
Proof. It results from the definition of that the canonical image of such a section is null. Conversely, if has a null image in , it suffices to see that every admits a neighborhood in in which is null, and one may therefore reduce to the case where is affine, Noetherian, , where is an ideal of , and , where is an -module of finite type. Then is the separated completion of for the -preadic topology, and the homomorphism is the canonical homomorphism . One knows (0, 7.3.7) that the kernel of this homomorphism is the set of annihilated by an element of . One therefore has for an ; for every one deduces , and since is invertible in ( being contained in the maximal ideal of ), one has , which demonstrates the proposition.
Corollary (10.8.12). The support of is equal to .
Proof. It is clear that is an -Module of finite type (10.8.8, (ii)) and (0, 5.2.4), hence its support is closed (0, 5.2.2) and evidently contained in . To show that it is equal to this last set, one is at once reduced to proving that the relation entails ; now this results from (10.8.11) and from (1.4.1).
Corollary (10.8.13). Let be a homomorphism of coherent -Modules. For to be null, it is necessary and sufficient that be null in a neighborhood of .
Proof. Indeed, by (10.8.8, (ii)), is identified with , hence if one considers as a section over of the sheaf , is the section of over which corresponds to it canonically ((10.8.10.2) and (0, 4.4.6)). It therefore suffices to apply (10.8.11) to the coherent -Module .
Corollary (10.8.14). Let be a homomorphism of coherent -Modules. For to be a monomorphism (resp. an epimorphism), it is necessary and sufficient that be a monomorphism (resp. an epimorphism) in a neighborhood of .
Proof. Let and be the cokernel and the kernel of , so that one has the exact sequence , whence (10.8.8, (i)) the exact sequence If is a monomorphism (resp. an epimorphism), one has (resp. ), hence there is a neighborhood of in which (resp. ) by virtue of (10.8.13).
10.9. Extension of a morphism to the completions
(10.9.1) Let , be two locally Noetherian (ordinary) preschemes, a morphism, (resp. ) a closed part of the underlying space (resp. ), such that . Let (resp. ) be a sheaf of ideals of (resp. ) such that the support of (resp. ) is (resp. ) and that ; one will note that there always exist such sheaves of ideals, for one may for example take for the largest sheaf of ideals of defining a subprescheme of having for underlying space (5.2.1), and the hypothesis then entails (5.2.4). One therefore has, for every integer , (0, 4.3.5); consequently (4.4.6), if one sets , , one deduces from a morphism , and it is immediate that the form an inductive system. We shall designate its inductive limit (10.6.8) by , and we shall say (by abuse of language) that is the extension of to the completions of and along and . It is immediate to verify that this morphism does not depend on the choice of the sheaves of ideals , verifying the conditions above. It suffices indeed to see it when and are affine Noetherian schemes with rings , ; then , , where (resp. ) is an ideal of (resp. ), corresponds to a ring homomorphism such that (4.4.6 and 1.7.4); is then the morphism which corresponds (10.2.2) to the continuous homomorphism , where (resp. ) is the separated completion of (resp. ) for the -preadic (resp. -preadic) topology (10.6.8); and one knows that if one replaces by another sheaf of ideals such that the support of is again , the -preadic and -preadic topologies on are the same (10.8.2).
One will note that, by this definition, the continuous map of the spaces underlying and which corresponds to is none other than the restriction to of .
(10.9.2) It results at once from the preceding definition that the diagram of morphisms of ringed spaces is commutative, the vertical arrows being the canonical morphisms (10.8.7).
(10.9.3) Let be a third prescheme, a morphism, a closed part of such that . If designates the completion along and of the morphism , it results at once from (10.9.1) that one has .
Proposition (10.9.4). Let , be two locally Noetherian -preschemes, being of finite type over . Let , be two -morphisms of into such that , . For , it is necessary and sufficient that and coincide in a neighborhood of .
Proof. The condition is evidently sufficient (without hypothesis of finiteness on ). To see that it is necessary, let us remark first that the hypothesis implies for every . On the other hand, the question being local, one may suppose that and are respective affine open neighborhoods of and of , with Noetherian rings, that is affine, and that is a -algebra of finite type (6.3.3). Then and correspond to two -homomorphisms , of into (1.7.3), and by hypothesis, the extensions by continuity of these homomorphisms to the separated completion of are the same. One concludes from (10.8.11) that for every section , the sections and coincide in a neighborhood of (depending on ); since is an algebra of finite type over , one deduces at once that there exists a neighborhood of such that and coincide in for every section . If is such that is a neighborhood of contained in , one concludes from what precedes and from (1.4.1, d)) that and coincide in .
Proposition (10.9.5). Under the hypotheses of (10.9.1), for every coherent -Module , there exists a canonical functorial isomorphism of -Modules
Proof. If one identifies canonically with and with (10.8.8), the proposition results at once from the commutativity of the diagram of (10.9.2).
(10.9.6) Let now be a coherent -Module, a coherent -Module. If is an -morphism of into , there corresponds to it an -homomorphism , hence by completion a continuous -homomorphism ; and by virtue of (10.9.5) there exists one and only one -morphism such that . If one considers the triples ( being a coherent -Module and a closed part of ) as a category, the morphisms consisting of a morphism of preschemes such that and of an -morphism , one may therefore say that is a functor in , taking its values in the category of pairs formed of a locally Noetherian formal prescheme and of an -Module , the morphisms of this last category consisting of the pairs formed of a morphism of formal preschemes and of a -morphism.
Proposition (10.9.7). Let , , be three locally Noetherian preschemes, , two morphisms, a closed part of , (resp. ) a closed part of (resp. ) such that (resp. ); let ; suppose locally Noetherian, and let , where and are the projections of . Under these conditions, the completion is identified with the product of the -formal preschemes , the structure morphisms being identified with and , and the projections with and .
Proof. It is immediate that the question is local for , , and , and one is therefore reduced to the case where , , , , , , where , , are three ideals such that and , denoting by and the homomorphisms and which correspond to and . Then one knows that and that , where is the ideal . The conclusion results (10.7.2) from the fact that the completed tensor product (where , , are respectively the separated completions of , , for the -, -, and -preadic topologies) is the separated completion of the tensor product for the -preadic topology (0, 7.7.2).
One will note moreover that if is a locally Noetherian -prescheme, , two -morphisms, a closed part of such that , , then the extension to the completions is identified with .
Corollary (10.9.8). Let , be two locally Noetherian -preschemes such that is locally Noetherian; let be a closed part of , (resp. ) a closed part of (resp. ) whose image in is contained in . For every -morphism such that , the graph morphism is identified with the extension of the graph morphism of .
Corollary (10.9.9). Let , be two locally Noetherian preschemes, a morphism, a closed part of , . Then the formal prescheme is identified, by the commutative diagram with the product of formal preschemes.
Proof. It suffices to apply (10.9.7) replacing and by , and by .
Remark (10.9.10). If is the sum (3.1), the union , where is a closed part of (), one sees at once that one has .
10.10. Application to coherent sheaves on formal affine schemes
(10.10.1) Throughout this paragraph, will designate an adic Noetherian ring, an ideal of definition of . Let , , which is identified with the closed part of (10.1.2). Moreover, the definition (10.1.2) and the definition (10.8.4) show that the formal affine scheme is identical to the completion of the affine scheme along the closed part of its underlying space. To every coherent -Module there corresponds therefore an -Module of finite type which is moreover a sheaf of topological modules on the sheaf of topological rings . But every coherent -Module is of the form , where is an -module of finite type (1.5.1); we shall set . Moreover, if is an -homomorphism of -modules of finite type, there corresponds to it a homomorphism , and consequently also a continuous homomorphism , which we shall denote . It is immediate that ; one has thus defined a covariant additive functor from the category of -modules of finite type into that of -Modules of finite type. When is a discrete ring, one has .
Proposition (10.10.2). (i) is an exact functor in , and there exists a canonical functorial isomorphism of -modules .
(ii) If and are two -modules of finite type, there exist canonical functorial isomorphisms
(iii) The map is a functorial isomorphism
Proof. The exactness of results from the exactness of the functors (1.3.5) and (10.8.8). By definition, is the separated completion of the -module for the -preadic topology; but since is complete and of finite type, one knows (0, 7.3.6) that is separated and complete, which finishes proving (i). The isomorphism (10.10.2.1) (resp. (10.10.2.2)) comes from the composition of the isomorphisms (1.3.12, (i)) and (10.8.10.1) (resp. (1.3.12, (ii)) and (10.8.10.2)). Finally, since is an -module of finite type, one may apply (i) to it, which identifies with , and use (10.10.2.2), which proves that the homomorphism (10.10.2.3) is an isomorphism.
One deduces from (10.10.2) a whole series of consequences analogous to those deduced from (1.3.7) and (1.3.12), which we leave to the reader the care of formulating.
Let us note that the property of exactness of , applied to the exact sequence , shows that the sheaf of ideals of designated here by coincides with the one which had been denoted in the same way in (10.3.1), by virtue of (10.3.2).
Proposition (10.10.3). Under the hypotheses of (10.10.1), is a coherent sheaf of rings.
Proof. If , one knows that is an adic Noetherian ring (0, 7.6.11) and since the question is local, one is reduced (10.1.4) to proving that the kernel of a homomorphism is an -Module of finite type. One then has , where is an -homomorphism (10.10.2); since is Noetherian, the kernel of is of finite type, in other words one has a homomorphism such that the sequence is exact. One concludes (10.10.2) that the sequence is exact, which proves that the kernel of is of finite type.
(10.10.4) With the preceding notations, set , and let be the affine scheme , being the sheaf of ideals of definition of corresponding to the ideal . Let be the morphism of preschemes corresponding to the canonical homomorphism for ; the formal scheme is the inductive limit of the for the (10.6.3).
Proposition (10.10.5). Under the hypotheses of (10.10.1), let be an -Module. The following conditions are equivalent:
a) is a coherent -Module.
b) is isomorphic to the projective limit (10.6.6) of a sequence of coherent -Modules such that .
c) There exists an -module of finite type (determined up to canonical isomorphism by (10.10.2, (i))) such that is isomorphic to .
Proof. Let us first show that b) implies c). One has , where is an -module of finite type, and the hypothesis entails that for (1.6.5); the therefore form a projective system for the canonical -homomorphisms (), and it results at once from the definition of the that this projective system verifies the conditions of (0, 7.2.9); its projective limit is consequently an -module of finite type such that for every . One deduces that is induced on by , hence by definition (10.8.4).
Conversely, c) entails b); indeed, if is the immersion morphism , is induced on by , and by definition (10.8.4); since for , the verify the conditions of b), whence our assertion.
Let us now show that c) implies a): indeed, one has by definition ; being the cokernel of a homomorphism , it results from (10.10.2) that is the cokernel of a homomorphism , and since the sheaf of rings is coherent (10.10.3), so is (0, 5.3.4).
Finally, a) entails b). Considered as an -Module, one has ; is a coherent -Module (0, 5.3.5), and since it is also an -Module and is coherent, one concludes that is a coherent -Module (0, 5.3.10), and it is immediate that for (recalling that the continuous map of the underlying spaces is the identity of ). The sheaf is therefore a coherent -Module, since one has seen that b) entails a). The canonical homomorphisms form a projective system, which by passage to the limit gives a canonical homomorphism , and everything reduces to demonstrating that is bijective. The question now being local, one may restrict oneself to the case where is the cokernel of a homomorphism ; this homomorphism being of the form , where is a homomorphism (10.10.2), is isomorphic to , where (10.10.2). One then has, by virtue of (10.10.2), , and since the -adic topology on is discrete, one has (as -Module); one has seen above that and is therefore indeed in this case the identity. Q.E.D.
Corollary (10.10.6). If verifies condition b) of (10.10.5), the projective system is isomorphic to the system of the .
(10.10.7) Let now , be two adic Noetherian rings, a continuous homomorphism; one will designate by (resp. ) an ideal of definition of (resp. ), such that , and one will set , , , . Let be the morphism of preschemes corresponding to (1.6.1), its extension to the completions (10.9.1), which is also the morphism of formal preschemes corresponding to (10.2.2).
Proposition (10.10.8). For every -module of finite type, there exists a canonical functorial isomorphism of -Modules
Proof. Indeed, denoting by and the canonical morphisms, one has (10.8.8), up to canonical functorial isomorphisms, and (1.6.5); the proposition therefore results from the commutativity of the diagram (10.9.2).
Corollary (10.10.9). For every ideal of , one has .
Proof. Indeed, let be the canonical injection , to which corresponds the canonical injection of sheaves of -Modules; by definition, is the image of the homomorphism ; but this homomorphism is identified with by (10.10.8). Since the image of is the ideal of , the image of is therefore by virtue of (10.10.2), whence the conclusion.
10.11. Coherent sheaves on formal preschemes
Proposition (10.11.1). If is a locally Noetherian formal prescheme, the sheaf of rings is coherent and every sheaf of ideals of definition of is coherent.
Proof. The question being local, one is reduced to the case of a Noetherian formal affine scheme, and the proposition therefore results from (10.10.3) and (10.10.5).
(10.11.2) Let be a locally Noetherian formal prescheme, a sheaf of ideals of definition of , the (ordinary) locally Noetherian prescheme , so that is the inductive limit of the sequence for the canonical morphisms (10.6.3). With these notations:
Theorem (10.11.3). For an -Module to be coherent, it is necessary and sufficient that it be isomorphic to a projective limit of a sequence , where is a coherent -Module such that for (10.6.6). The projective system is then isomorphic to the system of the , being the canonical morphism .
Proof. The question being local, one is reduced to the case where is a Noetherian formal affine scheme, and the theorem is then a consequence of (10.10.5) and (10.10.6).
One may therefore say that the datum of a coherent -Module is equivalent to that of a projective system of coherent -Modules such that for .
Corollary (10.11.4). If and are two coherent -Modules, one may (with the notations of (10.11.3)) define a canonical functorial isomorphism
Proof. The projective limit of the second member is to be understood for the maps () of into . The homomorphism (10.11.4.1) makes correspond to an element the sequence ; one sees at once that one defines a homomorphism inverse to the preceding by making correspond to the projective system its projective limit in , taking account of (10.11.3).
Corollary (10.11.5). For a homomorphism to be surjective, it is necessary and sufficient that the corresponding homomorphism be.
Proof. The question being local, one is reduced to the case where , being adic Noetherian, , and , where and are -modules of finite type and a homomorphism ; one then has moreover , where is the homomorphism ; the conclusion results from the fact that and (resp. and ) are simultaneously surjective (1.3.9 and 10.10.2) and from the fact that and are simultaneously surjective (0, 7.1.14).
(10.11.6) The th. (10.11.3) shows that one may consider every coherent -Module as a topological -Module, by considering it as the projective limit of the sheaves of pseudo-discrete groups (0, 3.8.1). It then results from (10.11.4) that every homomorphism of coherent -Modules is automatically continuous (0, 3.8.2). Moreover, if is a coherent sub--Module of a coherent -Module , then for every open , is a closed subgroup of the topological group , for the functor being left exact, is the kernel of the homomorphism , which is continuous by what precedes, since is coherent (0, 5.3.4); our assertion results from the fact that is a separated topological group.
Proposition (10.11.7). Let and be two coherent -Modules. One may define (with the notations of (10.11.3)) canonical functorial isomorphisms of topological -Modules (10.11.6)
Proof. The existence of the isomorphism (10.11.7.1) results from the formula and from (10.11.3). The isomorphism (10.11.7.2), where the two members are considered as sheaves of modules without topology, results from the definition of the sections of and and from the existence of the isomorphism (10.11.4.1), applied to the prescheme induced on an arbitrary Noetherian formal affine open of . It remains to prove that the isomorphism (10.11.7.2) is bicontinuous over a quasi-compact set, and one is therefore reduced to the case where , being adic Noetherian, whence (10.10.5) , , , being -modules of finite type; taking account of (10.10.2.1), (10.10.2.3), and (1.3.12, (ii)), one is reduced to showing that the canonical isomorphism (with , ) is continuous, which was proved in (0, 7.8.2).
(10.11.8) Since is the group of sections of the sheaf of topological groups , it is equipped with a group topology. If is Noetherian, it results from (10.11.7.2) that a fundamental system of neighborhoods of in this group is obtained by taking the subgroups ( arbitrary).
Proposition (10.11.9). Let be a Noetherian formal prescheme, and two coherent -Modules. In the topological group the surjective (resp. injective, bijective) homomorphisms form an open part.
Proof. By virtue of (10.11.5), the set of surjective homomorphisms in is the inverse image, under the continuous map , of a part of the discrete group , whence the first assertion. To demonstrate the second, let us cover by a finite number of Noetherian formal affine opens . For to be injective, it is necessary and sufficient that all its images under the (continuous) restriction maps be; one is therefore reduced to the affine case, and then this has already been proved in (0, 7.8.3).
10.12. Adic morphisms of formal preschemes
(10.12.1) Let , be two locally Noetherian formal preschemes; we shall say that a morphism is adic if there exists an Ideal of definition of such that is an Ideal of definition of ; one also says then that is an adic -prescheme (for ). When this is so, for every Ideal of definition of , is an Ideal of definition of . Indeed, the question being local, one may suppose and affine Noetherian; there exists therefore an integer such that and (10.3.6 and 0, 7.1.4), whence and . The first of these relations proves that , where is an open ideal of , and the second proves that is an ideal of definition of (0, 7.1.4), whence our assertion.
It results at once from what precedes that if and are two adic -preschemes, every -morphism is adic: indeed, if , are the structure morphisms, and an Ideal of definition of , one has , hence is an Ideal of definition of , and by hypothesis is an Ideal of definition of .
(10.12.2) In what follows, we shall suppose fixed a locally Noetherian formal prescheme and an Ideal of definition of ; we shall set . The adic (locally Noetherian) -preschemes evidently form a category. We shall say that an inductive system of (ordinary) locally Noetherian -preschemes is an adic -inductive system if the structure morphisms are such that, for , the diagrams are commutative and identify with the product . The adic inductive systems form a category: it suffices indeed to define a morphism of such systems as an inductive system of -morphisms such that is identified with for . This being so:
Theorem (10.12.3). There is a canonical equivalence between the category of adic -preschemes and the category of adic -inductive systems.
Proof. The equivalence in question is obtained in the following way: if is an adic -prescheme, the structure morphism, is an Ideal of definition of and one makes correspond to the inductive system of the , the structure morphism corresponding to (10.5.6). Let us first show that is an adic inductive system: if , one has , hence for every , and (by the exactness of the functor ) for ; our conclusion therefore results from (4.4.5). It is immediate moreover to verify that to a -morphism of adic -preschemes there corresponds (with evident notations) an inductive system of -morphisms such that is identified with for .
The fact that one has indeed thus defined an equivalence will result from the following more precise proposition:
Proposition (10.12.3.1). Let be an inductive system of -preschemes; suppose that the structure morphisms are such that the diagrams (10.12.2.1) are commutative and identify with for . Then the inductive system verifies the conditions b) and c) of (10.6.3); let be its inductive limit, the morphism inductive limit of the inductive system . Then, if is locally Noetherian, is locally Noetherian and is an adic morphism.
Proof. Since the sheaf of ideals of which defines the subprescheme of is nilpotent, so is, by virtue of (4.4.5), the sheaf of ideals of defining the subprescheme of , hence the conditions of (10.6.3) are indeed verified. The question is consequently local on and and one may suppose that , , being a -adic Noetherian ring, and ; if , the hypothesis entails that is Noetherian and that if one sets , . The kernel of is therefore and the kernel of is for ; moreover, since is Noetherian, is of finite type over , hence is of finite type over , and a fortiori over ; the fact that is Noetherian then results from (10.6.4); if one has , and if is the kernel of , . If is the homomorphism corresponding to , one therefore has ; as the homomorphism corresponding to is equal to , the ideal of is dense in , and since every ideal of is closed (0, 7.3.5), one has . If , the relation then results from (10.10.9) and finishes the demonstration.
(10.12.3.2) The preceding equivalence furnishes, for two adic -preschemes , , a canonical bijection the projective limit being relative to the maps for .
10.13. Morphisms of finite type
Proposition (10.13.1). Let be a locally Noetherian formal prescheme, an Ideal of definition of , a morphism of formal preschemes. The following conditions are equivalent:
a) is locally Noetherian, is an adic morphism (10.12.1) and if one sets , the morphism deduced from is of finite type.
b) is locally Noetherian, and is the inductive limit of an adic -inductive system such that the morphism is of finite type.
c) Every point of possesses a Noetherian formal affine open neighborhood having the following property:
(Q) is the union of a finite family of Noetherian formal affine opens such that the adic Noetherian ring is topologically isomorphic to the quotient of an algebra of restricted formal power series (0, 7.5.1) over , by an ideal (necessarily closed).
Proof. It is immediate that a) entails b) by virtue of (10.12.3). To show that b) entails c), one may, since the question is local on , suppose that , where is adic Noetherian; let , being an ideal of definition of . Since by hypothesis is of finite type over , is the finite union of affine opens such that the ring of the affine scheme induced by on is an algebra of finite type over the ring of (6.3.2). By virtue of (5.1.9), is also an affine open in each of the Noetherian preschemes , and if is the ring of the affine scheme induced by on , hypothesis b) entails that for , is isomorphic to . Consequently, the formal prescheme induced on by is isomorphic to , where (10.6.4); is a -adic ring, and , isomorphic to , is an algebra of finite type over . One concludes (0, 7.5.5) that is topologically isomorphic to a quotient of an algebra of restricted formal power series over (by a necessarily closed ideal, since such an algebra is Noetherian (0, 7.5.4)).
To demonstrate that c) entails a), one may limit oneself to the case where is also affine, being an adic Noetherian ring, isomorphic to the quotient of an algebra of restricted formal power series over by a closed ideal. Then (0, 7.5.5), is an algebra of finite type over , and is an ideal of definition of , hence, by virtue of (10.10.9), the conditions of a) are satisfied.
One will note that if the conditions of prop. (10.13.1) are fulfilled, property a) is valid for every Ideal of definition of (by virtue of c)), and consequently, in property b), all the are morphisms of finite type.
Corollary (10.13.2). If the conditions of (10.13.1) are verified, every Noetherian formal affine open of possesses property (Q), and if is Noetherian, so is .
Proof. This results at once from (10.13.1) and from (6.3.2).
Definition (10.13.3). When the equivalent properties a), b), c) of (10.13.1) are verified, one says that the morphism is of finite type, or that is a -formal prescheme of finite type, or a formal prescheme of finite type over .
Corollary (10.13.4). Let , be two Noetherian formal affine schemes; for to be of finite type over , it is necessary and sufficient that the adic Noetherian ring be isomorphic to the quotient of an algebra of restricted formal power series over by a closed ideal.
Proof. Indeed, with the notations of (10.13.1), if is of finite type over , is then a -algebra of finite type by virtue of (6.3.3) and is an ideal of definition of (10.10.9). One therefore concludes by (0, 7.5.5).
Proposition (10.13.5). (i) The composite of two morphisms of formal preschemes which are of finite type is of finite type.
(ii) Let , , be three locally Noetherian (resp. Noetherian) formal preschemes, , two morphisms. If is of finite type, is locally Noetherian (resp. Noetherian) and is of finite type over .
(iii) Let be a locally Noetherian formal prescheme, , two locally Noetherian -formal preschemes such that is locally Noetherian. If , are locally Noetherian -formal preschemes, , two -morphisms of finite type, is locally Noetherian and is a -morphism of finite type.
Proof. (iii) is deduced from (i) and (ii) by the formal reasoning of (3.5.1) and it therefore suffices to prove (i) and (ii).
Let , , be three locally Noetherian formal preschemes, , two morphisms of finite type. If is an Ideal of definition of , is one for and is one for . Set , , and let , be the morphisms corresponding to and . Since by hypothesis and are of finite type, so is (6.3.4) which corresponds to ; hence is of finite type by (10.13.1).
Under the conditions of (ii), (resp. , ) is the inductive limit of a sequence (resp. , ) of locally Noetherian preschemes and one may suppose (10.13.1) that for . The formal prescheme is then the inductive limit of the preschemes (10.7.4), and one has Moreover, is locally Noetherian since is of finite type over (6.3.8). One concludes first (10.12.3.1) that is locally Noetherian; moreover, since is of finite type over (6.3.8), it results from (10.12.3.1) and from (10.13.1) that is of finite type over , which finishes proving (ii) (the assertion relative to the Noetherian preschemes being an immediate consequence of (6.3.8)).
Corollary (10.13.6). Under the hypotheses of (10.9.9), if is a morphism of finite type, so is its extension to the completions.
10.14. Closed subpreschemes of formal preschemes
Proposition (10.14.1). Let be a locally Noetherian formal prescheme, a coherent sheaf of ideals of . If is the (closed) support of , the topologically ringed space is a locally Noetherian formal prescheme, which is Noetherian if is.
Proof. Let us note that is coherent by virtue of (10.10.3) and (0, 5.3.4), hence its support is closed (0, 5.2.2). Let be an Ideal of definition of , and let ; the sheaf of rings is the projective limit of the sheaves (10.11.3), which all have for support. The sheaf is a coherent -Module, since is coherent, hence is also a coherent -Module (0, 5.3.10); if is the closed subprescheme of defined by this sheaf of ideals, it is immediate that is the formal prescheme inductive limit of the , and since the conditions of (10.6.4) are satisfied, this proves that this formal prescheme is locally Noetherian, and Noetherian if is (since then is by virtue of (6.1.4)).
Definition (10.14.2). One calls closed subprescheme of a formal prescheme every formal prescheme where is a coherent -Module; one says that this prescheme is the closed subprescheme defined by .
It is clear that the correspondence thus defined between coherent -Modules and closed subpreschemes of is one-to-one.
The morphism of topologically ringed spaces , where is the injection and the canonical homomorphism , is evidently (10.4.5) a morphism of formal preschemes, which one calls the canonical injection of into . One will note that if , where is adic Noetherian, one has , where is an ideal of (10.10.5), and it results at once from what precedes that one then has up to an isomorphism, and that corresponds (10.2.2) to the canonical homomorphism .
One says that a morphism of locally Noetherian formal preschemes is a closed immersion if it factors as , where is an isomorphism of onto a closed subprescheme of and the canonical injection. Since is a monomorphism of ringed spaces, and are necessarily unique.
Proposition (10.14.3). A closed immersion is a morphism of finite type.
Proof. One reduces at once to the case where is a formal affine scheme and ; the proposition results from (10.13.1, c)).
Lemma (10.14.4). Let be a morphism of locally Noetherian formal preschemes, and let be a covering of by Noetherian formal affine opens of , such that the are Noetherian formal affine opens of . For to be a closed immersion, it is necessary and sufficient that be a closed part of and that, for every , the restriction of to correspond (10.4.6) to a surjective homomorphism .
Proof. The conditions are evidently necessary. Conversely, if they are fulfilled, and if one designates by the kernel of , one defines a coherent sheaf of ideals of by taking , and by taking null in the complement of the union of the . Indeed, since is closed and the support of is , everything reduces to verifying that and induce the same sheaf on a Noetherian formal affine open . Now, the restriction of to being a closed immersion of this formal prescheme into , is a Noetherian formal affine open in and the restriction of to is a closed immersion; if is the kernel of the surjective homomorphism corresponding to this restriction, it is immediate (10.10.2) that induces on . The sheaf of ideals being thus defined, it is then clear that , where is the canonical injection of the closed subprescheme of defined by , and an isomorphism of onto .
Proposition (10.14.5). (i) If , are closed immersions of locally Noetherian formal preschemes, is a closed immersion.
(ii) Let , , be three locally Noetherian formal preschemes, a closed immersion, a morphism. Then the morphism is a closed immersion.
(iii) Let be a locally Noetherian formal prescheme, , two -formal preschemes locally Noetherian such that is locally Noetherian. If , are locally Noetherian -formal preschemes, , two -morphisms which are closed immersions, then is a closed immersion.
Proof. By virtue of (3.5.1), it again suffices to prove (i) and (ii).
To demonstrate (i), one may suppose that (resp. ) is a closed subprescheme of (resp. ) defined by a coherent sheaf (resp. ) of ideals of (resp. ); if is the injection of the underlying spaces, is a coherent sheaf of ideals of (0, 5.3.12), hence also a coherent -Module (0, 5.3.10); the kernel of is therefore a coherent sheaf of ideals of (0, 5.3.4), and is isomorphic to , which proves that is isomorphic to a closed subprescheme of .
To demonstrate (ii), it is immediate that one may limit oneself to the case where , , , being a -adic Noetherian ring, , where is an ideal of , an adic and Noetherian topological -algebra. Everything reduces to proving that the homomorphism is surjective: now, is an -module of finite type, and its topology is the -adic topology; it then results from (0, 7.7.8) that is identified with , whence our assertion.
Corollary (10.14.6). Under the hypotheses of (10.14.5, (ii)), let , be the projections, so that the diagram is commutative. For every coherent -Module , one then has a canonical isomorphism of -Modules
Proof. To define a homomorphism , one knows that it amounts to the same to define a homomorphism (0, 4.4.3): we shall take , where is the canonical homomorphism (0, 4.4.3). To see that is an isomorphism, one reduces at once to the case where , , are formal spectra of adic Noetherian rings , , , with the conditions seen above in (10.14.5, (ii)); one then has , where is an -module of finite type (10.10.5), and the two members of (10.14.6.1) are identified respectively, by virtue of (10.10.8), with and , whence the corollary, since is canonically identified with .
Corollary (10.14.7). Let be a locally Noetherian ordinary prescheme, a closed subprescheme of , the canonical injection , a closed part of and ; then is a closed immersion, and for every coherent -Module , one has
Proof. Since , it suffices to use (10.9.9) and to apply (10.14.5) and (10.14.6).
10.15. Separated formal preschemes
Definition (10.15.1). Let be a formal prescheme, a -formal prescheme, the structure morphism. One calls diagonal morphism (also denoted ) the morphism . One says that is separated over , or a -formal scheme, or that is a separated morphism, if the image under of the underlying space is a closed part of the space underlying . One says that a formal prescheme is separated, or is a formal scheme, if it is separated over .
Proposition (10.15.2). Suppose that the formal preschemes , are respectively inductive limits of sequences , of ordinary preschemes, and that the morphism is the inductive limit of a sequence of morphisms . For to be separated, it is necessary and sufficient that the morphism be.
Proof. Indeed, is then the inductive limit of the sequence of morphisms (10.7.4), and the image under of the underlying space (resp. the underlying space ) is identical to the image under of the underlying space (resp. to the underlying space ); whence the conclusion.
Proposition (10.15.3). Suppose in what follows that all the formal preschemes (resp. morphisms of formal preschemes) considered are inductive limits of sequences of ordinary preschemes (resp. of morphisms of ordinary preschemes).
(i) The composite of two separated morphisms is separated.
(ii) If , are two separated -morphisms, is separated.
(iii) If is a separated -morphism, the -morphism is separated for every extension of the formal base prescheme.
(iv) If the composite of two morphisms is separated, is separated.
(One understands in this statement that if one and the same formal prescheme intervenes several times in one and the same proposition, one considers it as the inductive limit of the same sequence of ordinary preschemes everywhere it figures, and the morphisms of into a formal prescheme (resp. of a formal prescheme into ) as inductive limits of morphisms of the into ordinary preschemes (resp. of ordinary preschemes into the ).)
Proof. With the notations of (10.15.2), one has indeed , and , and the assertions of (10.15.3) are then immediate consequences of (10.15.2) and of the corresponding assertions of (5.5.1) for ordinary preschemes.
We leave to the reader the care of stating, for the same kind of formal preschemes and of morphisms as in (10.15.3), the propositions corresponding to (5.5.5), (5.5.9), and (5.5.10) (replacing therein “affine open” by “formal affine open verifying condition b) of (10.6.3)”).
An analogous reasoning also shows that every Noetherian formal affine scheme is separated, which justifies the terminology.
Proposition (10.15.4). Let be a locally Noetherian formal prescheme, , two locally Noetherian -formal preschemes, such that or is of finite type over (so that is locally Noetherian) and that is separated over . Let be a -morphism; then the graph morphism is a closed immersion.
Proof. One may suppose that is the inductive limit of a sequence of locally Noetherian preschemes, (resp. ) the inductive limit of a sequence (resp. ) of -preschemes, the inductive limit of a sequence of -morphisms ; then is the inductive limit of the sequence and of the sequence (10.7.4); by hypothesis, is separated over (10.15.2), hence the space is a closed subspace of ; since the underlying spaces of (resp. ) and (resp. ) are the same, one sees already that is a closed subspace of . Let us now remark that when runs through the set of pairs formed of a Noetherian formal affine open (resp. ) of (resp. ) such that , the opens form a covering of in , and if is the restriction of to , is the restriction of to . If we show that is a closed immersion, it will be so also for (10.14.4); in other words, one is reduced to the case where , , are affine (, , adic Noetherian), corresponding to a continuous -homomorphism ; then corresponds to the unique continuous homomorphism which, composed with the canonical homomorphisms and , gives respectively the identity and . Now, it is clear that is surjective, whence our assertion.
Corollary (10.15.5). Let be a locally Noetherian formal prescheme, a -prescheme of finite type; for to be separated over , it is necessary and sufficient that the diagonal morphism be a closed immersion.
Proposition (10.15.6). A closed immersion of locally Noetherian formal preschemes is a separated morphism.
Proof. With the notations of (10.14.2), is a closed immersion, hence a separated morphism, and it suffices to apply (10.15.2).
Proposition (10.15.7). Let be a locally Noetherian (ordinary) prescheme, a closed part of and . For to be separated, it is necessary and sufficient that be, and it is sufficient that be.
Proof. Indeed, with the notations of (10.8.5), for to be separated, it is necessary and sufficient that be (10.15.2), and since , it is equivalent to say that is (5.5.1, (vi)).