Processing math: 100%

The Stacks project

10.97 Completion for Noetherian rings

In this section we discuss completion with respect to ideals in Noetherian rings.

Lemma 10.97.1. Let I be an ideal of a Noetherian ring R. Denote {}^\wedge completion with respect to I.

  1. If K \to N is an injective map of finite R-modules, then the map on completions K^\wedge \to N^\wedge is injective.

  2. If 0 \to K \to N \to M \to 0 is a short exact sequence of finite R-modules, then 0 \to K^\wedge \to N^\wedge \to M^\wedge \to 0 is a short exact sequence.

  3. If M is a finite R-module, then M^\wedge = M \otimes _ R R^\wedge .

Proof. Setting M = N/K we find that part (1) follows from part (2). Let 0 \to K \to N \to M \to 0 be as in (2). For each n we get the short exact sequence

0 \to K/(I^ nN \cap K) \to N/I^ nN \to M/I^ nM \to 0.

By Lemma 10.87.1 we obtain the exact sequence

0 \to \mathop{\mathrm{lim}}\nolimits K/(I^ nN \cap K) \to N^\wedge \to M^\wedge \to 0.

By the Artin-Rees Lemma 10.51.2 we may choose c such that I^ nK \subset I^ n N \cap K \subset I^{n-c} K for n \geq c. Hence K^\wedge = \mathop{\mathrm{lim}}\nolimits K/I^ nK = \mathop{\mathrm{lim}}\nolimits K/(I^ nN \cap K) and we conclude that (2) is true.

Let M be as in (3) and let 0 \to K \to R^{\oplus t} \to M \to 0 be a presentation of M. We get a commutative diagram

\xymatrix{ & K \otimes _ R R^\wedge \ar[r] \ar[d] & R^{\oplus t} \otimes _ R R^\wedge \ar[r] \ar[d] & M \otimes _ R R^\wedge \ar[r] \ar[d] & 0 \\ 0 \ar[r] & K^\wedge \ar[r] & (R^{\oplus t})^\wedge \ar[r] & M^\wedge \ar[r] & 0 }

The top row is exact, see Section 10.39. The bottom row is exact by part (2). By Lemma 10.96.1 the vertical arrows are surjective. The middle vertical arrow is an isomorphism. We conclude (3) holds by the Snake Lemma 10.4.1. \square

Lemma 10.97.2. Let I be a ideal of a Noetherian ring R. Denote {}^\wedge completion with respect to I.

  1. The ring map R \to R^\wedge is flat.

  2. The functor M \mapsto M^\wedge is exact on the category of finitely generated R-modules.

Proof. Consider J \otimes _ R R^\wedge \to R \otimes _ R R^\wedge = R^\wedge where J is an arbitrary ideal of R. According to Lemma 10.97.1 this is identified with J^\wedge \to R^\wedge and J^\wedge \to R^\wedge is injective. Part (1) follows from Lemma 10.39.5. Part (2) is a reformulation of Lemma 10.97.1 part (2). \square

Lemma 10.97.3. Let I be an ideal of a Noetherian ring R. Denote R^\wedge the completion of R with respect to I. If I is contained in the Jacobson radical of R, then the ring map R \to R^\wedge is faithfully flat. In particular, if (R, \mathfrak m) is a Noetherian local ring, then the completion \mathop{\mathrm{lim}}\nolimits _ n R/\mathfrak m^ n is faithfully flat.

Proof. By Lemma 10.97.2 it is flat. The composition R \to R^\wedge \to R/I where the last map is the projection map R^\wedge \to R/I shows that any maximal ideal of R is in the image of \mathop{\mathrm{Spec}}(R^\wedge ) \to \mathop{\mathrm{Spec}}(R). Hence the map is faithfully flat by Lemma 10.39.15. \square

Lemma 10.97.4. Let R be a Noetherian ring. Let I be an ideal of R. Let M be an R-module. Then the completion M^\wedge of M with respect to I is I-adically complete, I^ n M^\wedge = (I^ nM)^\wedge , and M^\wedge /I^ nM^\wedge = M/I^ nM.

Proof. This is a special case of Lemma 10.96.3 because I is a finitely generated ideal. \square

Lemma 10.97.5. Let I be an ideal of a ring R. Assume

  1. R/I is a Noetherian ring,

  2. I is finitely generated.

Then the completion R^\wedge of R with respect to I is a Noetherian ring complete with respect to IR^\wedge .

Proof. By Lemma 10.96.3 we see that R^\wedge is I-adically complete. Hence it is also IR^\wedge -adically complete. Since R^\wedge /IR^\wedge = R/I is Noetherian we see that after replacing R by R^\wedge we may in addition to assumptions (1) and (2) assume that also R is I-adically complete.

Let f_1, \ldots , f_ t be generators of I. Then there is a surjection of rings R/I[T_1, \ldots , T_ t] \to \bigoplus I^ n/I^{n + 1} mapping T_ i to the element \overline{f}_ i \in I/I^2. Hence \bigoplus I^ n/I^{n + 1} is a Noetherian ring. Let J \subset R be an ideal. Consider the ideal

\bigoplus J \cap I^ n/J \cap I^{n + 1} \subset \bigoplus I^ n/I^{n + 1}.

Let \overline{g}_1, \ldots , \overline{g}_ m be generators of this ideal. We may choose \overline{g}_ j to be a homogeneous element of degree d_ j and we may pick g_ j \in J \cap I^{d_ j} mapping to \overline{g}_ j \in J \cap I^{d_ j}/J \cap I^{d_ j + 1}. We claim that g_1, \ldots , g_ m generate J.

Let x \in J \cap I^ n. There exist a_ j \in I^{\max (0, n - d_ j)} such that x - \sum a_ j g_ j \in J \cap I^{n + 1}. The reason is that J \cap I^ n/J \cap I^{n + 1} is equal to \sum \overline{g}_ j I^{n - d_ j}/I^{n - d_ j + 1} by our choice of g_1, \ldots , g_ m. Hence starting with x \in J we can find a sequence of vectors (a_{1, n}, \ldots , a_{m, n})_{n \geq 0} with a_{j, n} \in I^{\max (0, n - d_ j)} such that

x = \sum \nolimits _{n = 0, \ldots , N} \sum \nolimits _{j = 1, \ldots , m} a_{j, n} g_ j \bmod I^{N + 1}

Setting A_ j = \sum _{n \geq 0} a_{j, n} we see that x = \sum A_ j g_ j as R is complete. Hence J is finitely generated and we win. \square

Lemma 10.97.6. Let R be a Noetherian ring. Let I be an ideal of R. The completion R^\wedge of R with respect to I is Noetherian.

Proof. This is a consequence of Lemma 10.97.5. It can also be seen directly as follows. Choose generators f_1, \ldots , f_ n of I. Consider the map

R[[x_1, \ldots , x_ n]] \longrightarrow R^\wedge , \quad x_ i \longmapsto f_ i.

This is a well defined and surjective ring map (details omitted). Since R[[x_1, \ldots , x_ n]] is Noetherian (see Lemma 10.31.2) we win. \square

Suppose R \to S is a local homomorphism of local rings (R, \mathfrak m) and (S, \mathfrak n). Let S^\wedge be the completion of S with respect to \mathfrak n. In general S^\wedge is not the \mathfrak m-adic completion of S. If \mathfrak n^ t \subset \mathfrak mS for some t \geq 1 then we do have S^\wedge = \mathop{\mathrm{lim}}\nolimits S/\mathfrak m^ nS by Lemma 10.96.9. In some cases this even implies that S^\wedge is finite over R^\wedge .

Lemma 10.97.7. Let R \to S be a local homomorphism of local rings (R, \mathfrak m) and (S, \mathfrak n). Let R^\wedge , resp. S^\wedge be the completion of R, resp. S with respect to \mathfrak m, resp. \mathfrak n. If \mathfrak m and \mathfrak n are finitely generated and \dim _{\kappa (\mathfrak m)} S/\mathfrak mS < \infty , then

  1. S^\wedge is equal to the \mathfrak m-adic completion of S, and

  2. S^\wedge is a finite R^\wedge -module.

Proof. We have \mathfrak mS \subset \mathfrak n because R \to S is a local ring map. The assumption \dim _{\kappa (\mathfrak m)} S/\mathfrak mS < \infty implies that S/\mathfrak mS is an Artinian ring, see Lemma 10.53.2. Hence has dimension 0, see Lemma 10.60.5, hence \mathfrak n = \sqrt{\mathfrak mS}. This and the fact that \mathfrak n is finitely generated implies that \mathfrak n^ t \subset \mathfrak mS for some t \geq 1. By Lemma 10.96.9 we see that S^\wedge can be identified with the \mathfrak m-adic completion of S. As \mathfrak m is finitely generated we see from Lemma 10.96.3 that S^\wedge and R^\wedge are \mathfrak m-adically complete. At this point we may apply Lemma 10.96.12 to S^\wedge as an R^\wedge -module to conclude. \square

Lemma 10.97.8. Let R be a Noetherian ring. Let R \to S be a finite ring map. Let \mathfrak p \subset R be a prime and let \mathfrak q_1, \ldots , \mathfrak q_ m be the primes of S lying over \mathfrak p (Lemma 10.36.21). Then

R_\mathfrak p^\wedge \otimes _ R S = (S_\mathfrak p)^\wedge = S_{\mathfrak q_1}^\wedge \times \ldots \times S_{\mathfrak q_ m}^\wedge

where the (S_\mathfrak p)^\wedge is the completion with respect to \mathfrak p and the local rings R_\mathfrak p and S_{\mathfrak q_ i} are completed with respect to their maximal ideals.

Proof. The first equality follows from Lemma 10.97.1. We may replace R by the localization R_\mathfrak p and S by S_\mathfrak p = S \otimes _ R R_\mathfrak p. Hence we may assume that R is a local Noetherian ring and that \mathfrak p = \mathfrak m is its maximal ideal. The \mathfrak q_ iS_{\mathfrak q_ i}-adic completion S_{\mathfrak q_ i}^\wedge is equal to the \mathfrak m-adic completion by Lemma 10.97.7. For every n \geq 1 prime ideals of S/\mathfrak m^ nS are in 1-to-1 correspondence with the maximal ideals \mathfrak q_1, \ldots , \mathfrak q_ m of S (by going up for S over R, see Lemma 10.36.22). Hence S/\mathfrak m^ nS = \prod S_{\mathfrak q_ i}/\mathfrak m^ nS_{\mathfrak q_ i} by Lemma 10.53.6 (using for example Proposition 10.60.7 to see that S/\mathfrak m^ nS is Artinian). Hence the \mathfrak m-adic completion S^\wedge of S is equal to \prod S_{\mathfrak q_ i}^\wedge . Finally, we have R^\wedge \otimes _ R S = S^\wedge by Lemma 10.97.1. \square

Lemma 10.97.9. Let R be a ring. Let I \subset R be an ideal. Let 0 \to K \to P \to M \to 0 be a short exact sequence of R-modules. If M is flat over R and M/IM is a projective R/I-module, then the sequence of I-adic completions

0 \to K^\wedge \to P^\wedge \to M^\wedge \to 0

is a split exact sequence.

Proof. As M is flat, each of the sequences

0 \to K/I^ nK \to P/I^ nP \to M/I^ nM \to 0

is short exact, see Lemma 10.39.12 and the sequence 0 \to K^\wedge \to P^\wedge \to M^\wedge \to 0 is a short exact sequence, see Lemma 10.96.1. It suffices to show that we can find splittings s_ n : M/I^ nM \to P/I^ nP such that s_{n + 1} \bmod I^ n = s_ n. We will construct these s_ n by induction on n. Pick any splitting s_1, which exists as M/IM is a projective R/I-module. Assume given s_ n for some n > 0. Set P_{n + 1} = \{ x \in P \mid x \bmod I^ nP \in \mathop{\mathrm{Im}}(s_ n)\} . The map \pi : P_{n + 1}/I^{n + 1}P_{n + 1} \to M/I^{n + 1}M is surjective (details omitted). As M/I^{n + 1}M is projective as a R/I^{n + 1}-module by Lemma 10.77.7 we may choose a section t : M/I^{n + 1}M \to P_{n + 1}/I^{n + 1}P_{n + 1} of \pi . Setting s_{n + 1} equal to the composition of t with the canonical map P_{n + 1}/I^{n + 1}P_{n + 1} \to P/I^{n + 1}P works. \square

Lemma 10.97.10. Let A be a Noetherian ring. Let I, J \subset A be ideals. If A is I-adically complete and A/I is J-adically complete, then A is J-adically complete.

Proof. Let B be the (I + J)-adic completion of A. By Lemma 10.97.2 B/IB is the J-adic completion of A/I hence isomorphic to A/I by assumption. Moreover B is I-adically complete by Lemma 10.96.8. Hence B is a finite A-module by Lemma 10.96.12. By Nakayama's lemma (Lemma 10.20.1 using I is in the Jacobson radical of A by Lemma 10.96.6) we find that A \to B is surjective. The map A \to B is flat by Lemma 10.97.2. The image of \mathop{\mathrm{Spec}}(B) \to \mathop{\mathrm{Spec}}(A) contains V(I) and as I is contained in the Jacobson radical of A we find A \to B is faithfully flat (Lemma 10.39.16). Thus A \to B is injective. Thus A is complete with respect to I + J, hence a fortiori complete with respect to J. \square


Comments (5)

Comment #3389 by Vignesh on

A different presentation of the proof of Lemma 10.96.1:

Consider a presentation of module M (which is f.g. by assumption) We get that is exact. (Here is viewed as a submodule of .) One sees this by observing that maps surjectively onto .

By Lemma 10.86.1, we get that is exact. By Artin-Rees lemma, Therefore we get is exact as required.

A correction in the proof of Lemma 10.96.2:

Here I think you mean an arbitrary ideal in R (and not specifically the ideal w.r.t. which is completed): .

Comment #3390 by Vignesh on

Sorry about the typo in my comment, in the second paragraph, I wanted to say:

By Lemma 10.86.1, we get that is exact. By Artin-Rees lemma, . Therefore we get is exact as required.

Comment #3459 by on

OK, I tried to improve the exposition a little bit based on your comments. Thanks! See here for the corresponding latex changes.

Comment #4200 by Qijun Yan on

Just a remark on the last part of the proof of Lemma 0DYC: As is a quotient of and flat over , then it seems that we must have by the definition of flatness.

Comment #4202 by on

@#4200 Well, it may be that we can shorten the proof at the end there, but flat surjections of rings aren't necessarily isomorphisms.


Your email address will not be published. Required fields are marked.

In your comment you can use Markdown and LaTeX style mathematics (enclose it like $\pi$). A preview option is available if you wish to see how it works out (just click on the eye in the toolbar).

Unfortunately JavaScript is disabled in your browser, so the comment preview function will not work.

All contributions are licensed under the GNU Free Documentation License.