Loading web-font TeX/Math/Italic

The Stacks project

10.98 Taking limits of modules

In this section we discuss what happens when we take a limit of modules.

Lemma 10.98.1. Let I \subset A be a finitely generated ideal of a ring. Let (M_ n) be an inverse system of A-modules with I^ n M_ n = 0. Then M = \mathop{\mathrm{lim}}\nolimits M_ n is I-adically complete.

Proof. We have M \to M/I^ nM \to M_ n. Taking the limit we get M \to M^\wedge \to M. Hence M is a direct summand of M^\wedge . Since M^\wedge is I-adically complete by Lemma 10.96.3, so is M. \square

Lemma 10.98.2. Let I \subset A be a finitely generated ideal of a ring. Let (M_ n) be an inverse system of A-modules with M_ n = M_{n + 1}/I^ nM_{n + 1}. Set M = \mathop{\mathrm{lim}}\nolimits M_ n. Then M/I^ nM = M_ n and M is I-adically complete.

Proof. By Lemma 10.98.1 we see that M is I-adically complete. Since the transition maps are surjective, the maps M \to M_ n are surjective. Consider the inverse system of short exact sequences

0 \to N_ n \to M \to M_ n \to 0

defining N_ n. Since M_ n = M_{n + 1}/I^ nM_{n + 1} the map N_{n + 1} + I^ nM \to N_ n is surjective. Hence N_{n + 1}/(N_{n + 1} \cap I^{n + 1}M) \to N_ n/(N_ n \cap I^ nM) is surjective. Taking the inverse limit of the short exact sequences

0 \to N_ n/(N_ n \cap I^ nM) \to M/I^ nM \to M_ n \to 0

we obtain an exact sequence

0 \to \mathop{\mathrm{lim}}\nolimits N_ n/(N_ n \cap I^ nM) \to M^\wedge \to M

Since M is I-adically complete we conclude that \mathop{\mathrm{lim}}\nolimits N_ n/(N_ n \cap I^ nM) = 0 and hence by the surjectivity of the transition maps we get N_ n/(N_ n \cap I^ nM) = 0 for all n. Thus M_ n = M/I^ nM as desired. \square

Lemma 10.98.3. Let A be a Noetherian graded ring. Let I \subset A_+ be a homogeneous ideal. Let (N_ n) be an inverse system of finite graded A-modules with N_ n = N_{n + 1}/I^ n N_{n + 1}. Then there is a finite graded A-module N such that N_ n = N/I^ nN as graded modules for all n.

Proof. Pick r and homogeneous elements x_{1, 1}, \ldots , x_{1, r} \in N_1 of degrees d_1, \ldots , d_ r generating N_1. Since the transition maps are surjective, we can pick a compatible system of homogeneous elements x_{n, i} \in N_ n lifting x_{1, i}. By the graded Nakayama lemma (Lemma 10.56.1) we see that N_ n is generated by the elements x_{n, 1}, \ldots , x_{n, r} sitting in degrees d_1, \ldots , d_ r. Thus for m \leq n we see that N_ n \to N_ n/I^ m N_ n is an isomorphism in degrees < \min (d_ i) + m (as I^ mN_ n is zero in those degrees). Thus the inverse system of degree d parts

\ldots = N_{2 + d - \min (d_ i), d} = N_{1 + d - \min (d_ i), d} = N_{d - \min (d_ i), d} \to N_{-1 + d - \min (d_ i), d} \to \ldots

stabilizes as indicated. Let N be the graded A-module whose dth graded part is this stabilization. In particular, we have the elements x_ i = \mathop{\mathrm{lim}}\nolimits x_{n, i} in N. We claim the x_ i generate N: any x \in N_ d is a linear combination of x_1, \ldots , x_ r because we can check this in N_{d - \min (d_ i), d} where it holds as x_{d - \min (d_ i), i} generate N_{d - \min (d_ i)}. Finally, the reader checks that the surjective map N/I^ nN \to N_ n is an isomorphism by checking to see what happens in each degree as before. Details omitted. \square

Lemma 10.98.4. Let A be a graded ring. Let I \subset A_+ be a homogeneous ideal. Denote A' = \mathop{\mathrm{lim}}\nolimits A/I^ n. Let (G_ n) be an inverse system of graded A-modules with G_ n annihilated by I^ n. Let M be a graded A-module and let \varphi _ n : M \to G_ n be a compatible system of graded A-module maps. If the induced map

\varphi : M \otimes _ A A' \longrightarrow \mathop{\mathrm{lim}}\nolimits G_ n

is an isomorphism, then M_ d \to \mathop{\mathrm{lim}}\nolimits G_{n, d} is an isomorphism for all d \in \mathbf{Z}.

Proof. By convention graded rings are in degrees \geq 0 and graded modules may have nonzero parts of any degree, see Section 10.56. The map \varphi exists because \mathop{\mathrm{lim}}\nolimits G_ n is a module over A' as G_ n is annihilated by I^ n. Another useful thing to keep in mind is that we have

\bigoplus \nolimits _{d \in \mathbf{Z}} \mathop{\mathrm{lim}}\nolimits G_{n, d} \subset \mathop{\mathrm{lim}}\nolimits G_ n \subset \prod \nolimits _{d \in \mathbf{Z}} \mathop{\mathrm{lim}}\nolimits G_{n, d}

where a subscript {\ }_ d indicates the dth graded part.

Injective. Let x \in M_ d. If x \mapsto 0 in \mathop{\mathrm{lim}}\nolimits G_{n, d} then x \otimes 1 = 0 in M \otimes _ A A'. Then we can find a finitely generated submodule M' \subset M with x \in M' such that x \otimes 1 is zero in M' \otimes _ A A'. Say M' is generated by homogeneous elements sitting in degrees d_1, \ldots , d_ r. Let n = d - \min (d_ i) + 1. Since A' has a map to A/I^ n and since A \to A/I^ n is an isomorphism in degrees \leq n - 1 we see that M' \to M' \otimes _ A A' is injective in degrees \leq n - 1. Thus x = 0 as desired.

Surjective. Let y \in \mathop{\mathrm{lim}}\nolimits G_{n, d}. Choose a finite sum \sum x_ i \otimes f'_ i in M \otimes _ A A' mapping to y. We may assume x_ i is homogeneous, say of degree d_ i. Observe that although A' is not a graded ring, it is a limit of the graded rings A/I^ nA and moreover, in any given degree the transition maps eventually become isomorphisms (see above). This gives

A = \bigoplus \nolimits _{d \geq 0} A_ d \subset A' \subset \prod \nolimits _{d \geq 0} A_ d

Thus we can write

f'_ i = \sum \nolimits _{j = 0, \ldots , d - d_ i - 1} f_{i, j} + f_ i + g'_ i

with f_{i, j} \in A_ j, f_ i \in A_{d - d_ i}, and g'_ i \in A' mapping to zero in \prod _{j \leq d - d_ i} A_ j. Now if we compute \varphi _ n(\sum _{i, j} f_{i, j}x_ i) \in G_ n, then we get a sum of homogeneous elements of degree < d. Hence \varphi (\sum x_ i \otimes f_{i, j}) maps to zero in \mathop{\mathrm{lim}}\nolimits G_{n, d}. Similarly, a computation shows the element \varphi (\sum x_ i \otimes g'_ i) maps to zero in \prod _{d' \leq d} \mathop{\mathrm{lim}}\nolimits G_{n, d'}. Since we know that \varphi (\sum x_ i \otimes f'_ i) is y, we conclude that \sum f_ ix_ i \in M_ d maps to y as desired. \square


Comments (0)


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.