Bourbaki

Lemma 3

Tag 000I, permanent. Cite it as [Bourbaki, Tag 000I].

Algebra, VIII, p. 6, § 1. Artinian Modules and Noetherian Modules

Let A be a ring and nn a natural number. An Artinian A-module M that is the sum of a family of submodules of length ⩽n\leqslant n has finite length.

We use induction on nn. First, suppose n=1n= 1. If M were not of finite length, then we could construct a sequence (Mm)m∈N(M_m)_{m\in\mathbf{N}} of submodules of M of length 1 with Mm⊄(∑i<mMi)M_m\not\subset (\sum_{i<m}M_i) for every m∈Nm\in \mathbf{N}. We would then have Mm∩∑i<mMi=0M_m\cap \sum_{i<m}M_i= 0 for every mm, and the sum of the family (Mm)m∈N(M_m)_{m\in\mathbf{N}} would be direct. However, this contradicts the fact that the A-module M is Artinian (VIII, p. 2, Example 2).

Now, suppose n⩾2n\geqslant 2. Let (Mi)i∈I(M_i)_{i\in I} be a family of submodules of M of length ⩽n\leqslant n with sum M. For every i∈Ii\in I, choose a submodule Mi′M'_i of MiM_i of length ⩽n−1\leqslant n-1 such that Mi/Mi′M_i/M'_i has length ⩽1\leqslant 1. Set M′=∑Mi′M'=\sum M'_i, and denote the image of MiM_i in M′′=M/M′M''= M/M' by Mi′′M''_i. The modules Mi′′M''_i are of length ⩽1\leqslant 1, and their sum is M′′M''. The modules M′M' and M′′M'' are Artinian (VIII, p. 3, Proposition 3); by the induction hypothesis, they are of finite length. Hence, M has finite length (II, §1, No. 10, p. 212, Proposition 16).

Let us now prove Theorem 1. Let S\mathscr{S} denote the set of left ideals a\mathfrak{a} of A such that the module As/aA_s/\mathfrak{a} has finite length. Let (ai)i∈I(\mathfrak{a}_i)_{i\in I} be a finite family of elements of S\mathscr{S}. By Proposition 1 of VIII, p. 2, the A-module As/aiA_s/\mathfrak{a}_i is Artinian and Noetherian for every i∈Ii\in I. Consequently, As/⋂i∈IaiA_s/\bigcap_{i\in I}\mathfrak{a}_i is Artinian and Noetherian (VIII, p. 4, Corollary of Proposition 3), hence of finite length (VIII, p. 2, Proposition 1). This proves that S\mathscr{S} is a left directed set for the inclusion. Since the ring A is left Artinian, the set S\mathscr{S} has a least element b\mathfrak{b}. We denote the length of the A-module As/bA_s/\mathfrak{b} by nn.

Let xx be an element of AsA_s and a\mathfrak{a} its annihilator (II, §1, No. 12, p. 219). The A-module AxAx is isomorphic to As/aA_s/\mathfrak{a}. If AxAx has finite length, then a\mathfrak{a} belongs to S\mathscr{S}, so a\mathfrak{a} contains b\mathfrak{b} and AxAx has length ⩽n\leqslant n. Thus, every monogenous left ideal of A of finite length has length ⩽n\leqslant n. Let c\mathfrak{c} be the sum of these ideals; this is a left ideal of A, of finite length by Lemma 3. Every left ideal of A of finite length is a sum of monogenous left ideals of finite length and is therefore contained in c\mathfrak{c}. Hence, c\mathfrak{c} is the largest left ideal of A of finite length.

If c\mathfrak{c} were distinct from A, then the set of left ideals of A containing c\mathfrak{c} and distinct from c\mathfrak{c} would have a minimal element c′\mathfrak{c}'. The A-module c′/c\mathfrak{c}'/\mathfrak{c} would then have length 1, and c′\mathfrak{c}' would have finite length, which contradicts the fact that c\mathfrak{c} is maximal. We therefore have c=A\mathfrak{c}= A; the A-module AsA_s has finite length.

Bourbaki sets the proof directly after the statement, and the corpus does not separate them, so what is above is the statement and its proof together.

Cited by

Cites

Transcribed from the printed edition, 2023, Springer Nature. Not translated.