Let b be an element of H, homogeneous of degree >0, and belonging to none of the minimal elements p of Supp(M) such that dim(H/p)=dimH(M). Then dimH(M/bM)=dimH(M)−1.
Put d=dimH(M). By the definition of b, we have dimH(M/bM)<d. By prop. 4, we have
dimH(M/bM)=dimS−1H(S−1M/(b/1).S−1M) and formula (8) of § 3, no. 2 yields the inequality
dimS−1H(S−1M/(b/1).S−1M)≥dimS−1H(S−1M)−1. Finally, we have
dimS−1H(S−1M)=dimH(M)=d by prop. 4. Hence
dimH(M/bM)≥d−1, whence lemma 4.
Let us take up again the proof of Prop. 6, b). We may suppose dimH(M)>0. Let us remark that every minimal element of Supp(M) is graded (apply Lemma 1 of No. 2 to the quotient of H by the annihilator of M). By Prop. 8 of III, § 1, No. 4, there therefore exists a homogeneous element b of H, of degree >0, not belonging to any of the minimal elements p of Supp(M) such that dim(H/p)=dimH(M). By Lemma 4, one has dimH(M/bM)=dimH(M)−1. Suppose further that H is generated by H1 as an H0-algebra and that the residue field k of H0 is infinite. For every minimal element p of Supp(M) such that dim(H/p)=dimH(M), consider the vector subspace Vp=(p∩H1)⊗H0k of the vector space V=H1⊗H0k over k. If one had Vp=V, one would have p∩H1=H1 (II, § 3, No. 2, Prop. 4), whence H1⊂p and dimH(M)=dim(H/p)≤dim(H/H≥1)=0, which is not the case. Since k is supposed infinite, the union of the Vp is distinct from V; if b∈H1 is such that b⊗1 does not belong to any of the Vp, one has dim(M/bM)=dimH(M)−1.
Proceeding by induction on d=dimH(M), one then constructs a sequence (b1,...,bd) of elements of H, with bi homogeneous of degree ni>0 and such that M/∑i=1nbiM is an H-module of finite length. If one supposes that H is generated by H1 as an H0-algebra and that the residue field of H0 is infinite, one may suppose ni=1 for i=1,...,d. By Prop. 4, one has dimS−1H(S−1M)=d and
dimS−1H(S−1M/i=1∑d(bi/1).S−1M)=0. Then
(b1/1,...,bd/1) is a maximal secant sequence for the
S−1H-module
S−1M. Put
ai=bi(n1...nd)/ni for
1≤i≤d; then the
ai are all of the same degree, and
(a1/1,...,ad/1) is a maximal secant sequence for
S−1M (§ 3, No. 2, Remark 3).
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.
Nothing.
Nothing.