Lemma 10.127.1. Let R \to A be a ring map. Consider the category \mathcal{I} of all diagrams of R-algebra maps A' \to A with A' finitely presented over R. Then \mathcal{I} is filtered, and the colimit of the A' over \mathcal{I} is isomorphic to A.
10.127 Colimits and maps of finite presentation
In this section we prove some preliminary lemmas which will eventually help us prove result using absolute Noetherian reduction. In Categories, Section 4.19 we discuss filtered colimits in general. Here is an example of this very general notion.
Proof. The category1 \mathcal{I} is nonempty as R \to R is an object of it. Consider a pair of objects A' \to A, A'' \to A of \mathcal{I}. Then A' \otimes _ R A'' \to A is in \mathcal{I} (use Lemmas 10.6.2 and 10.14.2). The ring maps A' \to A' \otimes _ R A'' and A'' \to A' \otimes _ R A'' define arrows in \mathcal{I} thereby proving the second defining property of a filtered category, see Categories, Definition 4.19.1. Finally, suppose that we have two morphisms \sigma , \tau : A' \to A'' in \mathcal{I}. If x_1, \ldots , x_ r \in A' are generators of A' as an R-algebra, then we can consider A''' = A''/(\sigma (x_ i) - \tau (x_ i)). This is a finitely presented R-algebra and the given R-algebra map A'' \to A factors through the surjection \nu : A'' \to A'''. Thus \nu is a morphism in \mathcal{I} equalizing \sigma and \tau as desired.
The fact that our index category is cofiltered means that we may compute the value of B = \mathop{\mathrm{colim}}\nolimits _{A' \to A} A' in the category of sets (some details omitted; compare with the discussion in Categories, Section 4.19). To see that B \to A is surjective, for every a \in A we can use R[x] \to A, x \mapsto a to see that a is in the image of B \to A. Conversely, if b \in B is mapped to zero in A, then we can find A' \to A in \mathcal{I} and a' \in A' which maps to b. Then A'/(a') \to A is in \mathcal{I} as well and the map A' \to B factors as A' \to A'/(a') \to B which shows that b = 0 as desired. \square
Often it is easier to think about colimits over preordered sets. Let (\Lambda , \geq ) a preordered set. A system of rings over \Lambda is given by a ring R_\lambda for every \lambda \in \Lambda , and a morphism R_\lambda \to R_\mu whenever \lambda \leq \mu . These morphisms have to satisfy the rule that R_\lambda \to R_\mu \to R_\nu is equal to the map R_\lambda \to R_\nu for all \lambda \leq \mu \leq \nu . See Categories, Section 4.21. We will often assume that (I, \leq ) is directed, which means that \Lambda is nonempty and given \lambda , \mu \in \Lambda there exists a \nu \in \Lambda with \lambda \leq \nu and \mu \leq \nu . Recall that the colimit \mathop{\mathrm{colim}}\nolimits _\lambda R_\lambda is sometimes called a “direct limit” in this case (but we will not use this terminology).
Note that Categories, Lemma 4.21.5 tells us that colimits over filtered index categories are the same thing as colimits over directed sets.
Lemma 10.127.2. Let R \to A be a ring map. There exists a directed system A_\lambda of R-algebras of finite presentation such that A = \mathop{\mathrm{colim}}\nolimits _\lambda A_\lambda . If A is of finite type over R we may arrange it so that all the transition maps in the system of A_\lambda are surjective.
Proof. The first proof is that this follows from Lemma 10.127.1 and Categories, Lemma 4.21.5.
Second proof. Compare with the proof of Lemma 10.11.3. Consider any finite subset S \subset A, and any finite collection of polynomial relations E among the elements of S. So each s \in S corresponds to x_ s \in A and each e \in E consists of a polynomial f_ e \in R[X_ s; s\in S] such that f_ e(x_ s) = 0. Let A_{S, E} = R[X_ s; s\in S]/(f_ e; e\in E) which is a finitely presented R-algebra. There are canonical maps A_{S, E} \to A. If S \subset S' and if the elements of E correspond, via the map R[X_ s; s \in S] \to R[X_ s; s\in S'], to a subset of E', then there is an obvious map A_{S, E} \to A_{S', E'} commuting with the maps to A. Thus, setting \Lambda equal the set of pairs (S, E) with ordering by inclusion as above, we get a directed partially ordered set. It is clear that the colimit of this directed system is A.
For the last statement, suppose A = R[x_1, \ldots , x_ n]/I. In this case, consider the subset \Lambda ' \subset \Lambda consisting of those systems (S, E) above with S = \{ x_1, \ldots , x_ n\} . It is easy to see that still A = \mathop{\mathrm{colim}}\nolimits _{\lambda ' \in \Lambda '} A_{\lambda '}. Moreover, the transition maps are clearly surjective. \square
It turns out that we can characterize ring maps of finite presentation as follows. This in some sense says that the algebras of finite presentation are the “compact” objects in the category of R-algebras.
Lemma 10.127.3. Let \varphi : R \to S be a ring map. The following are equivalent
\varphi is of finite presentation,
for every directed system A_\lambda of R-algebras the map
\mathop{\mathrm{colim}}\nolimits _\lambda \mathop{\mathrm{Hom}}\nolimits _ R(S, A_\lambda ) \longrightarrow \mathop{\mathrm{Hom}}\nolimits _ R(S, \mathop{\mathrm{colim}}\nolimits _\lambda A_\lambda )is bijective, and
for every directed system A_\lambda of R-algebras the map
\mathop{\mathrm{colim}}\nolimits _\lambda \mathop{\mathrm{Hom}}\nolimits _ R(S, A_\lambda ) \longrightarrow \mathop{\mathrm{Hom}}\nolimits _ R(S, \mathop{\mathrm{colim}}\nolimits _\lambda A_\lambda )is surjective.
Proof. Assume (1) and write S = R[x_1, \ldots , x_ n] / (f_1, \ldots , f_ m). Let A = \mathop{\mathrm{colim}}\nolimits A_\lambda . Observe that an R-algebra homomorphism S \to A or S \to A_\lambda is determined by the images of x_1, \ldots , x_ n. Hence it is clear that \mathop{\mathrm{colim}}\nolimits _\lambda \mathop{\mathrm{Hom}}\nolimits _ R(S, A_\lambda ) \to \mathop{\mathrm{Hom}}\nolimits _ R(S, A) is injective. To see that it is surjective, let \chi : S \to A be an R-algebra homomorphism. Then each x_ i maps to some element in the image of some A_{\lambda _ i}. We may pick \mu \geq \lambda _ i, i = 1, \ldots , n and assume \chi (x_ i) is the image of y_ i \in A_\mu for i = 1, \ldots , n. Consider z_ j = f_ j(y_1, \ldots , y_ n) \in A_\mu . Since \chi is a homomorphism the image of z_ j in A = \mathop{\mathrm{colim}}\nolimits _\lambda A_\lambda is zero. Hence there exists a \mu _ j \geq \mu such that z_ j maps to zero in A_{\mu _ j}. Pick \nu \geq \mu _ j, j = 1, \ldots , m. Then the images of z_1, \ldots , z_ m are zero in A_\nu . This exactly means that the y_ i map to elements y'_ i \in A_\nu which satisfy the relations f_ j(y'_1, \ldots , y'_ n) = 0. Thus we obtain a ring map S \to A_\nu . This shows that (1) implies (2).
It is clear that (2) implies (3). Assume (3). By Lemma 10.127.2 we may write S = \mathop{\mathrm{colim}}\nolimits _\lambda S_\lambda with S_\lambda of finite presentation over R. Then the identity map factors as
for some \lambda . This implies that S is finitely presented over S_\lambda by Lemma 10.6.2 part (4) applied to S \to S_\lambda \to S. Applying part (2) of the same lemma to R \to S_\lambda \to S we conclude that S is of finite presentation over R. \square
Using the basic material above we can give a criterion of when an algebra A is a filtered colimit of given type of algebra as follows.
Lemma 10.127.4. Let R \to \Lambda be a ring map. Let \mathcal{E} be a set of R-algebras such that each A \in \mathcal{E} is of finite presentation over R. Then the following two statements are equivalent
\Lambda is a filtered colimit of elements of \mathcal{E}, and
for any R algebra map A \to \Lambda with A of finite presentation over R we can find a factorization A \to B \to \Lambda with B \in \mathcal{E}.
Proof. Suppose that \mathcal{I} \to \mathcal{E}, i \mapsto A_ i is a filtered diagram such that \Lambda = \mathop{\mathrm{colim}}\nolimits _ i A_ i. Let A \to \Lambda be an R-algebra map with A of finite presentation over R. Then we get a factorization A \to A_ i \to \Lambda by applying Lemma 10.127.3. Thus (1) implies (2).
Consider the category \mathcal{I} of Lemma 10.127.1. By Categories, Lemma 4.19.3 the full subcategory \mathcal{J} consisting of those A \to \Lambda with A \in \mathcal{E} is cofinal in \mathcal{I} and is a filtered category. Then \Lambda is also the colimit over \mathcal{J} by Categories, Lemma 4.17.2. \square
But more is true. Namely, given R = \mathop{\mathrm{colim}}\nolimits _\lambda R_\lambda we see that the category of finitely presented R-modules is equivalent to the limit of the category of finitely presented R_\lambda -modules. Similarly for the categories of finitely presented R-algebras.
Lemma 10.127.5. Let A be a ring and let M, N be A-modules. Suppose that R = \mathop{\mathrm{colim}}\nolimits _{i \in I} R_ i is a directed colimit of A-algebras.
If M is a finite A-module, and u, u' : M \to N are A-module maps such that u \otimes 1 = u' \otimes 1 : M \otimes _ A R \to N \otimes _ A R then for some i we have u \otimes 1 = u' \otimes 1 : M \otimes _ A R_ i \to N \otimes _ A R_ i.
If N is a finite A-module and u : M \to N is an A-module map such that u \otimes 1 : M \otimes _ A R \to N \otimes _ A R is surjective, then for some i the map u \otimes 1 : M \otimes _ A R_ i \to N \otimes _ A R_ i is surjective.
If N is a finitely presented A-module, and v : N \otimes _ A R \to M \otimes _ A R is an R-module map, then there exists an i and an R_ i-module map v_ i : N \otimes _ A R_ i \to M \otimes _ A R_ i such that v = v_ i \otimes 1.
If M is a finite A-module, N is a finitely presented A-module, and u : M \to N is an A-module map such that u \otimes 1 : M \otimes _ A R \to N \otimes _ A R is an isomorphism, then for some i the map u \otimes 1 : M \otimes _ A R_ i \to N \otimes _ A R_ i is an isomorphism.
Proof. To prove (1) assume u is as in (1) and let x_1, \ldots , x_ m \in M be generators. Since N \otimes _ A R = \mathop{\mathrm{colim}}\nolimits _ i N \otimes _ A R_ i we may pick an i \in I such that u(x_ j) \otimes 1 = u'(x_ j) \otimes 1 in M \otimes _ A R_ i, j = 1, \ldots , m. For such an i we have u \otimes 1 = u' \otimes 1 : M \otimes _ A R_ i \to N \otimes _ A R_ i.
To prove (2) assume u \otimes 1 surjective and let y_1, \ldots , y_ m \in N be generators. Since N \otimes _ A R = \mathop{\mathrm{colim}}\nolimits _ i N \otimes _ A R_ i we may pick an i \in I and z_ j \in M \otimes _ A R_ i, j = 1, \ldots , m whose images in N \otimes _ A R equal y_ j \otimes 1. For such an i the map u \otimes 1 : M \otimes _ A R_ i \to N \otimes _ A R_ i is surjective.
To prove (3) let y_1, \ldots , y_ m \in N be generators. Let K = \mathop{\mathrm{Ker}}(A^{\oplus m} \to N) where the map is given by the rule (a_1, \ldots , a_ m) \mapsto \sum a_ j x_ j. Let k_1, \ldots , k_ t be generators for K. Say k_ s = (k_{s1}, \ldots , k_{sm}). Since M \otimes _ A R = \mathop{\mathrm{colim}}\nolimits _ i M \otimes _ A R_ i we may pick an i \in I and z_ j \in M \otimes _ A R_ i, j = 1, \ldots , m whose images in M \otimes _ A R equal v(y_ j \otimes 1). We want to use the z_ j to define the map v_ i : N \otimes _ A R_ i \to M \otimes _ A R_ i. Since K \otimes _ A R_ i \to R_ i^{\oplus m} \to N \otimes _ A R_ i \to 0 is a presentation, it suffices to check that \xi _ s = \sum _ j k_{sj}z_ j is zero in M \otimes _ A R_ i for each s = 1, \ldots , t. This may not be the case, but since the image of \xi _ s in M \otimes _ A R is zero we see that it will be the case after increasing i a bit.
To prove (4) assume u \otimes 1 is an isomorphism, that M is finite, and that N is finitely presented. Let v : N \otimes _ A R \to M \otimes _ A R be an inverse to u \otimes 1. Apply part (3) to get a map v_ i : N \otimes _ A R_ i \to M \otimes _ A R_ i for some i. Apply part (1) to see that, after increasing i we have v_ i \circ (u \otimes 1) = \text{id}_{M \otimes _ R R_ i} and (u \otimes 1) \circ v_ i = \text{id}_{N \otimes _ R R_ i}. \square
Lemma 10.127.6. Suppose that R = \mathop{\mathrm{colim}}\nolimits _{\lambda \in \Lambda } R_\lambda is a directed colimit of rings. Then the category of finitely presented R-modules is the colimit of the categories of finitely presented R_\lambda -modules. More precisely
Given a finitely presented R-module M there exists a \lambda \in \Lambda and a finitely presented R_\lambda -module M_\lambda such that M \cong M_\lambda \otimes _{R_\lambda } R.
Given a \lambda \in \Lambda , finitely presented R_\lambda -modules M_\lambda , N_\lambda , and an R-module map \varphi : M_\lambda \otimes _{R_\lambda } R \to N_\lambda \otimes _{R_\lambda } R, then there exists a \mu \geq \lambda and an R_\mu -module map \varphi _\mu : M_\lambda \otimes _{R_\lambda } R_\mu \to N_\lambda \otimes _{R_\lambda } R_\mu such that \varphi = \varphi _\mu \otimes 1_ R.
Given a \lambda \in \Lambda , finitely presented R_\lambda -modules M_\lambda , N_\lambda , and R-module maps \varphi _\lambda , \psi _\lambda : M_\lambda \to N_\lambda such that \varphi \otimes 1_ R = \psi \otimes 1_ R, then \varphi \otimes 1_{R_\mu } = \psi \otimes 1_{R_\mu } for some \mu \geq \lambda .
Proof. To prove (1) choose a presentation R^{\oplus m} \to R^{\oplus n} \to M \to 0. Suppose that the first map is given by the matrix A = (a_{ij}). We can choose a \lambda \in \Lambda and a matrix A_\lambda = (a_{\lambda , ij}) with coefficients in R_\lambda which maps to A in R. Then we simply let M_\lambda be the R_\lambda -module with presentation R_\lambda ^{\oplus m} \to R_\lambda ^{\oplus n} \to M_\lambda \to 0 where the first arrow is given by A_\lambda .
Parts (2) and (3) follow from Lemma 10.127.5. \square
Lemma 10.127.7. Let A be a ring and let B, C be A-algebras. Suppose that R = \mathop{\mathrm{colim}}\nolimits _{i \in I} R_ i is a directed colimit of A-algebras.
If B is a finite type A-algebra, and u, u' : B \to C are A-algebra maps such that u \otimes 1 = u' \otimes 1 : B \otimes _ A R \to C \otimes _ A R then for some i we have u \otimes 1 = u' \otimes 1 : B \otimes _ A R_ i \to C \otimes _ A R_ i.
If C is a finite type A-algebra and u : B \to C is an A-algebra map such that u \otimes 1 : B \otimes _ A R \to C \otimes _ A R is surjective, then for some i the map u \otimes 1 : B \otimes _ A R_ i \to C \otimes _ A R_ i is surjective.
If C is of finite presentation over A and v : C \otimes _ A R \to B \otimes _ A R is an R-algebra map, then there exists an i and an R_ i-algebra map v_ i : C \otimes _ A R_ i \to B \otimes _ A R_ i such that v = v_ i \otimes 1.
If B is a finite type A-algebra, C is a finitely presented A-algebra, and u \otimes 1 : B \otimes _ A R \to C \otimes _ A R is an isomorphism, then for some i the map u \otimes 1 : B \otimes _ A R_ i \to C \otimes _ A R_ i is an isomorphism.
Proof. To prove (1) assume u is as in (1) and let x_1, \ldots , x_ m \in B be generators. Since B \otimes _ A R = \mathop{\mathrm{colim}}\nolimits _ i B \otimes _ A R_ i we may pick an i \in I such that u(x_ j) \otimes 1 = u'(x_ j) \otimes 1 in B \otimes _ A R_ i, j = 1, \ldots , m. For such an i we have u \otimes 1 = u' \otimes 1 : B \otimes _ A R_ i \to C \otimes _ A R_ i.
To prove (2) assume u \otimes 1 surjective and let y_1, \ldots , y_ m \in C be generators. Since B \otimes _ A R = \mathop{\mathrm{colim}}\nolimits _ i B \otimes _ A R_ i we may pick an i \in I and z_ j \in B \otimes _ A R_ i, j = 1, \ldots , m whose images in C \otimes _ A R equal y_ j \otimes 1. For such an i the map u \otimes 1 : B \otimes _ A R_ i \to C \otimes _ A R_ i is surjective.
To prove (3) let c_1, \ldots , c_ m \in C be generators. Let K = \mathop{\mathrm{Ker}}(A[x_1, \ldots , x_ m] \to N) where the map is given by the rule x_ j \mapsto \sum c_ j. Let f_1, \ldots , f_ t be generators for K as an ideal in A[x_1, \ldots , x_ m]. We think of f_ j = f_ j(x_1, \ldots , x_ m) as a polynomial. Since B \otimes _ A R = \mathop{\mathrm{colim}}\nolimits _ i B \otimes _ A R_ i we may pick an i \in I and z_ j \in B \otimes _ A R_ i, j = 1, \ldots , m whose images in B \otimes _ A R equal v(c_ j \otimes 1). We want to use the z_ j to define a map v_ i : C \otimes _ A R_ i \to B \otimes _ A R_ i. Since K \otimes _ A R_ i \to R_ i[x_1, \ldots , x_ m] \to C \otimes _ A R_ i \to 0 is a presentation, it suffices to check that \xi _ s = f_ j(z_1, \ldots , z_ m) is zero in B \otimes _ A R_ i for each s = 1, \ldots , t. This may not be the case, but since the image of \xi _ s in B \otimes _ A R is zero we see that it will be the case after increasing i a bit.
To prove (4) assume u \otimes 1 is an isomorphism, that B is a finite type A-algebra, and that C is a finitely presented A-algebra. Let v : B \otimes _ A R \to C \otimes _ A R be an inverse to u \otimes 1. Let v_ i : C \otimes _ A R_ i \to B \otimes _ A R_ i be as in part (3). Apply part (1) to see that, after increasing i we have v_ i \circ (u \otimes 1) = \text{id}_{B \otimes _ R R_ i} and (u \otimes 1) \circ v_ i = \text{id}_{C \otimes _ R R_ i}. \square
Lemma 10.127.8. Suppose that R = \mathop{\mathrm{colim}}\nolimits _{\lambda \in \Lambda } R_\lambda is a directed colimit of rings. Then the category of finitely presented R-algebras is the colimit of the categories of finitely presented R_\lambda -algebras. More precisely
Given a finitely presented R-algebra A there exists a \lambda \in \Lambda and a finitely presented R_\lambda -algebra A_\lambda such that A \cong A_\lambda \otimes _{R_\lambda } R.
Given a \lambda \in \Lambda , finitely presented R_\lambda -algebras A_\lambda , B_\lambda , and an R-algebra map \varphi : A_\lambda \otimes _{R_\lambda } R \to B_\lambda \otimes _{R_\lambda } R, then there exists a \mu \geq \lambda and an R_\mu -algebra map \varphi _\mu : A_\lambda \otimes _{R_\lambda } R_\mu \to B_\lambda \otimes _{R_\lambda } R_\mu such that \varphi = \varphi _\mu \otimes 1_ R.
Given a \lambda \in \Lambda , finitely presented R_\lambda -algebras A_\lambda , B_\lambda , and R_\lambda -algebra maps \varphi _\lambda , \psi _\lambda : A_\lambda \to B_\lambda such that \varphi \otimes 1_ R = \psi \otimes 1_ R, then \varphi \otimes 1_{R_\mu } = \psi \otimes 1_{R_\mu } for some \mu \geq \lambda .
Proof. To prove (1) choose a presentation A = R[x_1, \ldots , x_ n]/(f_1, \ldots , f_ m). We can choose a \lambda \in \Lambda and elements f_{\lambda , j} \in R_\lambda [x_1, \ldots , x_ n] mapping to f_ j \in R[x_1, \ldots , x_ n]. Then we simply let A_\lambda = R_\lambda [x_1, \ldots , x_ n]/(f_{\lambda , 1}, \ldots , f_{\lambda , m}).
Parts (2) and (3) follow from Lemma 10.127.7. \square
Lemma 10.127.9. Suppose R \to S is a local homomorphism of local rings. There exists a directed set (\Lambda , \leq ), and a system of local homomorphisms R_\lambda \to S_\lambda of local rings such that
The colimit of the system R_\lambda \to S_\lambda is equal to R \to S.
Each R_\lambda is essentially of finite type over \mathbf{Z}.
Each S_\lambda is essentially of finite type over R_\lambda .
Proof. Denote \varphi : R \to S the ring map. Let \mathfrak m \subset R be the maximal ideal of R and let \mathfrak n \subset S be the maximal ideal of S. Let
As partial ordering we take the inclusion relation. For each \lambda = (A, B) \in \Lambda we let R'_\lambda be the sub \mathbf{Z}-algebra generated by a \in A, and we let S'_\lambda be the sub \mathbf{Z}-algebra generated by b, b \in B. Let R_\lambda be the localization of R'_\lambda at the prime ideal R'_\lambda \cap \mathfrak m and let S_\lambda be the localization of S'_\lambda at the prime ideal S'_\lambda \cap \mathfrak n. In a picture
The transition maps are clear. We leave the proofs of the other assertions to the reader. \square
Lemma 10.127.10. Suppose R \to S is a local homomorphism of local rings. Assume that S is essentially of finite type over R. Then there exists a directed set (\Lambda , \leq ), and a system of local homomorphisms R_\lambda \to S_\lambda of local rings such that
The colimit of the system R_\lambda \to S_\lambda is equal to R \to S.
Each R_\lambda is essentially of finite type over \mathbf{Z}.
Each S_\lambda is essentially of finite type over R_\lambda .
For each \lambda \leq \mu the map S_\lambda \otimes _{R_\lambda } R_\mu \to S_\mu presents S_\mu as the localization of a quotient of S_\lambda \otimes _{R_\lambda } R_\mu .
Proof. Denote \varphi : R \to S the ring map. Let \mathfrak m \subset R be the maximal ideal of R and let \mathfrak n \subset S be the maximal ideal of S. Let x_1, \ldots , x_ n \in S be elements such that S is a localization of the sub R-algebra of S generated by x_1, \ldots , x_ n. In other words, S is a quotient of a localization of the polynomial ring R[x_1, \ldots , x_ n].
Let \Lambda = \{ A \subset R \mid \# A < \infty \} be the set of finite subsets of R. As partial ordering we take the inclusion relation. For each \lambda = A \in \Lambda we let R'_\lambda be the sub \mathbf{Z}-algebra generated by a \in A, and we let S'_\lambda be the sub \mathbf{Z}-algebra generated by \varphi (a), a \in A and the elements x_1, \ldots , x_ n. Let R_\lambda be the localization of R'_\lambda at the prime ideal R'_\lambda \cap \mathfrak m and let S_\lambda be the localization of S'_\lambda at the prime ideal S'_\lambda \cap \mathfrak n. In a picture
It is clear that if A \subset B corresponds to \lambda \leq \mu in \Lambda , then there are canonical maps R_\lambda \to R_\mu , and S_\lambda \to S_\mu and we obtain a system over the directed set \Lambda .
The assertion that R = \mathop{\mathrm{colim}}\nolimits R_\lambda is clear because all the maps R_\lambda \to R are injective and any element of R eventually is in the image. The same argument works for S = \mathop{\mathrm{colim}}\nolimits S_\lambda . Assertions (2), (3) are true by construction. The final assertion holds because clearly the maps S'_\lambda \otimes _{R'_\lambda } R'_\mu \to S'_\mu are surjective. \square
Lemma 10.127.11. Suppose R \to S is a local homomorphism of local rings. Assume that S is essentially of finite presentation over R. Then there exists a directed set (\Lambda , \leq ), and a system of local homomorphism R_\lambda \to S_\lambda of local rings such that
The colimit of the system R_\lambda \to S_\lambda is equal to R \to S.
Each R_\lambda is essentially of finite type over \mathbf{Z}.
Each S_\lambda is essentially of finite type over R_\lambda .
For each \lambda \leq \mu the map S_\lambda \otimes _{R_\lambda } R_\mu \to S_\mu presents S_\mu as the localization of S_\lambda \otimes _{R_\lambda } R_\mu at a prime ideal.
Proof. By assumption we may choose an isomorphism \Phi : (R[x_1, \ldots , x_ n]/I)_{\mathfrak q} \to S where I \subset R[x_1, \ldots , x_ n] is a finitely generated ideal, and \mathfrak q \subset R[x_1, \ldots , x_ n]/I is a prime. (Note that R \cap \mathfrak q is equal to the maximal ideal \mathfrak m of R.) We also choose generators f_1, \ldots , f_ m \in I for the ideal I. Write R in any way as a colimit R = \mathop{\mathrm{colim}}\nolimits R_\lambda over a directed set (\Lambda , \leq ), with each R_\lambda local and essentially of finite type over \mathbf{Z}. There exists some \lambda _0 \in \Lambda such that f_ j is the image of some f_{j, \lambda _0} \in R_{\lambda _0}[x_1, \ldots , x_ n]. For all \lambda \geq \lambda _0 denote f_{j, \lambda } \in R_{\lambda }[x_1, \ldots , x_ n] the image of f_{j, \lambda _0}. Thus we obtain a system of ring maps
Set \mathfrak q_\lambda the inverse image of \mathfrak q. Set S_\lambda = (R_\lambda [x_1, \ldots , x_ n]/ (f_{1, \lambda }, \ldots , f_{m, \lambda }))_{\mathfrak q_\lambda }. We leave it to the reader to see that this works. \square
Remark 10.127.12. Suppose that R \to S is a local homomorphism of local rings, which is essentially of finite presentation. Take any system (\Lambda , \leq ), R_\lambda \to S_\lambda with the properties listed in Lemma 10.127.10. What may happen is that this is the “wrong” system, namely, it may happen that property (4) of Lemma 10.127.11 is not satisfied. Here is an example. Let k be a field. Consider the ring
Set S = R/zR. As system take \Lambda = \mathbf{N} and R_ n = k[[z, y_1, \ldots , y_ n]]/(\{ y_ i^2 - zy_{i + 1}\} _{i \leq n-1}) and S_ n = R_ n/(z, y_ n^2). All the maps S_ n \otimes _{R_ n} R_{n + 1} \to S_{n + 1} are not localizations (i.e., isomorphisms in this case) since 1 \otimes y_{n + 1}^2 maps to zero. If we take instead S_ n' = R_ n/zR_ n then the maps S'_ n \otimes _{R_ n} R_{n + 1} \to S'_{n + 1} are isomorphisms. The moral of this remark is that we do have to be a little careful in choosing the systems.
Lemma 10.127.13. Suppose R \to S is a local homomorphism of local rings. Assume that S is essentially of finite presentation over R. Let M be a finitely presented S-module. Then there exists a directed set (\Lambda , \leq ), and a system of local homomorphisms R_\lambda \to S_\lambda of local rings together with S_\lambda -modules M_\lambda , such that
The colimit of the system R_\lambda \to S_\lambda is equal to R \to S. The colimit of the system M_\lambda is M.
Each R_\lambda is essentially of finite type over \mathbf{Z}.
Each S_\lambda is essentially of finite type over R_\lambda .
Each M_\lambda is finite over S_\lambda .
For each \lambda \leq \mu the map S_\lambda \otimes _{R_\lambda } R_\mu \to S_\mu presents S_\mu as the localization of S_\lambda \otimes _{R_\lambda } R_\mu at a prime ideal.
For each \lambda \leq \mu the map M_\lambda \otimes _{S_\lambda } S_\mu \to M_\mu is an isomorphism.
Proof. As in the proof of Lemma 10.127.11 we may first write R = \mathop{\mathrm{colim}}\nolimits R_\lambda as a directed colimit of local \mathbf{Z}-algebras which are essentially of finite type. Next, we may assume that for some \lambda _1 \in \Lambda there exist f_{j, \lambda _1} \in R_{\lambda _1}[x_1, \ldots , x_ n] such that
Choose a presentation
of M over S. Let A \in \text{Mat}(t \times s, S) be the matrix of the presentation. For some \lambda _2 \in \Lambda , \lambda _2 \geq \lambda _1 we can find a matrix A_{\lambda _2} \in \text{Mat}(t \times s, S_{\lambda _2}) which maps to A. For all \lambda \geq \lambda _2 we let M_\lambda = \mathop{\mathrm{Coker}}(S_\lambda ^{\oplus s} \xrightarrow {A_\lambda } S_\lambda ^{\oplus t}). We leave it to the reader to see that this works. \square
Lemma 10.127.14. Suppose R \to S is a ring map. Then there exists a directed set (\Lambda , \leq ), and a system of ring maps R_\lambda \to S_\lambda such that
The colimit of the system R_\lambda \to S_\lambda is equal to R \to S.
Each R_\lambda is of finite type over \mathbf{Z}.
Each S_\lambda is of finite type over R_\lambda .
Proof. This is the non-local version of Lemma 10.127.9. Proof is similar and left to the reader. \square
Lemma 10.127.15. Suppose R \to S is a ring map. Assume that S is integral over R. Then there exists a directed set (\Lambda , \leq ), and a system of ring maps R_\lambda \to S_\lambda such that
The colimit of the system R_\lambda \to S_\lambda is equal to R \to S.
Each R_\lambda is of finite type over \mathbf{Z}.
Each S_\lambda is finite over R_\lambda .
Proof. Consider the set \Lambda of pairs (E, F) where E \subset R is a finite subset, F \subset S is a finite subset, and every element f \in F is the root of a monic P(X) \in R[X] whose coefficients are in E. Say (E, F) \leq (E', F') if E \subset E' and F \subset F'. Given \lambda = (E, F) \in \Lambda set R_\lambda \subset R equal to the \mathbf{Z}-subalgebra of R generated by E and S_\lambda \subset S equal to the \mathbf{Z}-subalgebra generated by F and the image of E in S. It is clear that R = \mathop{\mathrm{colim}}\nolimits R_\lambda . We have S = \mathop{\mathrm{colim}}\nolimits S_\lambda as every element of S is integral over S. The ring maps R_\lambda \to S_\lambda are finite by Lemma 10.36.5 and the fact that S_\lambda is generated over R_\lambda by the elements of F which are integral over R_\lambda by our condition on the pairs (E, F). The lemma follows. \square
Lemma 10.127.16. Suppose R \to S is a ring map. Assume that S is of finite type over R. Then there exists a directed set (\Lambda , \leq ), and a system of ring maps R_\lambda \to S_\lambda such that
The colimit of the system R_\lambda \to S_\lambda is equal to R \to S.
Each R_\lambda is of finite type over \mathbf{Z}.
Each S_\lambda is of finite type over R_\lambda .
For each \lambda \leq \mu the map S_\lambda \otimes _{R_\lambda } R_\mu \to S_\mu presents S_\mu as a quotient of S_\lambda \otimes _{R_\lambda } R_\mu .
Proof. This is the non-local version of Lemma 10.127.10. Proof is similar and left to the reader. \square
Lemma 10.127.17. Suppose R \to S is a ring map. Assume that S is of finite presentation over R. Then there exists a directed set (\Lambda , \leq ), and a system of ring maps R_\lambda \to S_\lambda such that
The colimit of the system R_\lambda \to S_\lambda is equal to R \to S.
Each R_\lambda is of finite type over \mathbf{Z}.
Each S_\lambda is of finite type over R_\lambda .
For each \lambda \leq \mu the map S_\lambda \otimes _{R_\lambda } R_\mu \to S_\mu is an isomorphism.
Proof. This is the non-local version of Lemma 10.127.11. Proof is similar and left to the reader. \square
Lemma 10.127.18. Suppose R \to S is a ring map. Assume that S is of finite presentation over R. Let M be a finitely presented S-module. Then there exists a directed set (\Lambda , \leq ), and a system of ring maps R_\lambda \to S_\lambda together with S_\lambda -modules M_\lambda , such that
The colimit of the system R_\lambda \to S_\lambda is equal to R \to S. The colimit of the system M_\lambda is M.
Each R_\lambda is of finite type over \mathbf{Z}.
Each S_\lambda is of finite type over R_\lambda .
Each M_\lambda is finite over S_\lambda .
For each \lambda \leq \mu the map S_\lambda \otimes _{R_\lambda } R_\mu \to S_\mu is an isomorphism.
For each \lambda \leq \mu the map M_\lambda \otimes _{S_\lambda } S_\mu \to M_\mu is an isomorphism.
In particular, for every \lambda \in \Lambda we have
Proof. This is the non-local version of Lemma 10.127.13. Proof is similar and left to the reader. \square
Comments (0)