Let A′ be a subring of A containing S. In order that A′ contain Cn, it is necessary and sufficient that one have A′+mA=A.
One has pA⊂mA and Bn consists of the elements of the form a0pn+pa1pn−1+⋯+pnan with a0,...,an in A. Hence one has π(Bn)=kpn, whence π(Cn)=kpn[π(S)]. But since π(S) is a p-basis of k, one has k=kpn[π(S)] (A, V, p. 96), whence π(Cn)=k, that is to say Cn+mA=A.
Let A′ be a subring of A containing S. If A′ contains Cn, we have
A′+mA⊃Cn+mA=A,whenceA′+mA=A. Conversely, suppose that one has
A′+mA=A. Let
a0,...,an be elements of
A; by assumption there exist elements
a0′,...,an′ of
A′ such that
ai≡ai′modmA for
0≤i≤n. By prop. 1 of § 1, No. 1 and the assumption
mAn+1={0}, we therefore have
Φn(a0,...,an)=Φn(a0′,...,an′)∈A′, whence
Bn⊂A′. Since
Cn is the ring generated by
Bn∪S, one has
Cn⊂A′.
In the set S of subrings A′ of A containing S and such that A′+mA=A, there exists by lemma 1 a smallest element C, and one has Cn=C for every integer n≥0 such that mAn+1={0}.
One has C+mA=A by construction and p1C is nilpotent. One obviously has pC⊂C∩mA and lemma 2 below therefore shows that pC is a maximal ideal of C and consequently that C is a Cohen subring of A.
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.
Nothing.
Nothing.