Bourbaki

Definition 2

Tag 0CUD, permanent. Cite it as [Bourbaki, Tag 0CUD].

Commutative Algebra, chapter IX, § 1. Vecteurs de Witt

Let AA be a ring and let n≥1n \geq 1 be an integer. We denote by Wn(A)W_n(A) the quotient ring W(A)/Vn(A)W(A)/V_n(A).

Given elements a0,...,an−1a_0, ..., a_{n-1} of AA, we denote by [a0,...,an−1][a_0, ..., a_{n-1}] or [ai]0≤i<n[a_i]_{0 \leq i < n} the class modulo Vn(A)V_n(A) of the element (a0,...,an−1,0,0,...)(a_0, ..., a_{n-1}, 0, 0, ...) of W(A)W(A). According to Lemma 4 of No. 6, the mapping (a0,...,an−1)↦[a0,...,an−1](a_0, ..., a_{n-1}) \mapsto [a_0, ..., a_{n-1}] from AnA^n into Wn(A)W_n(A) is a bijection. For this reason, we say that the elements of Wn(A)W_n(A) are the Witt vectors of length nn; by analogy, one sometimes qualifies the elements of W(A)W(A) as Witt vectors of infinite length.

We denote by πn\pi_n the canonical homomorphism from W(A)W(A) into Wn(A)W_n(A). According to Lemma 4 of No. 6, we have

πn(a)=[a0,...,an−1] \pi_n(a) = [a_0, ..., a_{n-1}]
for every a=(an)n∈Na = (a_n)_{n \in \mathbf{N}} in W(A)W(A).

According to the definition of the operations in W(A)W(A), we have the following description of the operations in Wn(A)W_n(A):

[a0,...,an−1]+[b0,...,bn−1]=[Si(a0,...,ai;b0,...,bi)]0≤i<n [a_0, ..., a_{n-1}] + [b_0, ..., b_{n-1}] = [S_i(a_0, ..., a_i; b_0, ..., b_i)]_{0 \leq i < n}
[a0,...,an−1]×[b0,...,bn−1]=[Pi(a0,...,ai;b0,...,bi)]0≤i<n [a_0, ..., a_{n-1}] \times [b_0, ..., b_{n-1}] = [P_i(a_0, ..., a_i; b_0, ..., b_i)]_{0 \leq i < n}
−[a0,...,an−1]=[Ii(a0,...,ai)]0≤i<n. - [a_0, ..., a_{n-1}] = [I_i(a_0, ..., a_i)]_{0 \leq i < n}.

Moreover, the identity element of the addition in Wn(A)W_n(A) is [0,...,0][0, ..., 0] and that of the multiplication is [1,0,...,0][1, 0, ..., 0].

Let ii be an integer such that 0≤i≤n0 \leq i \leq n. By passing to the quotient, the homomorphism Φi\Phi_i from W(A)W(A) into AA defines a homomorphism Φi\Phi_i from Wn(A)W_n(A) into AA. This associates with the Witt vector [a0,...,an−1][a_0, ..., a_{n-1}] the element Φi(a0,...,ai)\Phi_i(a_0, ..., a_i) of AA (also called the ghost component of index ii of [a0,...,an−1][a_0, ..., a_{n-1}]).

Let ρ:B→A\rho : B \to A be a ring homomorphism. By passing to the quotients, the homomorphism W(ρ)W(\rho) from W(B)W(B) into W(A)W(A) defines a homomorphism Wn(ρ)W_n(\rho) from Wn(B)W_n(B) into Wn(A)W_n(A). It is described by the formula

Wn(ρ)[b0,...,bn−1]=[ρ(b0),...,ρ(bn−1)] W_n(\rho)[b_0, ..., b_{n-1}] = [\rho(b_0), ..., \rho(b_{n-1})]
for every [b0,...,bn−1][b_0, ..., b_{n-1}] in Wn(B)W_n(B).

Let mm and nn be two integers such that 1≤n≤m1 \leq n \leq m. We have Vn(A)⊃Vm(A)V_n(A) \supset V_m(A), whence a canonical homomorphism from Wm(A)=W(A)/Vm(A)W_m(A) = W(A)/V_m(A) onto Wn(A)=W(A)/Vn(A)W_n(A) = W(A)/V_n(A); this homomorphism will be denoted by πn,m\pi_{n,m}. Explicitly, we have

πn,m[a0,...,am−1]=[a0,...,an−1] \pi_{n,m}[a_0, ..., a_{m-1}] = [a_0, ..., a_{n-1}]

for [a0,...,am−1][a_0, ..., a_{m-1}] in Wm(A)W_m(A). The family (Wn(A),πn,m)(W_n(A), \pi_{n,m}) is a projective system of rings and the mapping π:a↦(πn(a))n≥1\pi : a \mapsto (\pi_n(a))_{n \geq 1} is a ring homomorphism from W(A)W(A) into lim⁡←Wn(A)\lim_{\leftarrow} W_n(A), called canonical. Since W(A)W(A) is separated and complete for the filtration (Vn(A))n∈Z(V_n(A))_{n \in \mathbf{Z}} (cf. No. 6), the canonical homomorphism π\pi is an isomorphism of topological rings, when Wn(A)W_n(A) is endowed with the discrete topology for every integer n≥1n \geq 1 (III, § 2, No. 6).

Henceforth, the homomorphisms πn\pi_n and πn,m\pi_{n,m} will be called projection homomorphisms from W(A)W(A) into Wn(A)W_n(A), and from Wm(A)W_m(A) into Wn(A)W_n(A) respectively.

Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.

Cited by

Nothing.

Cites

Nothing.

Translated with machine help, not yet reviewed by a person.