Bourbaki

Lemma 2

Tag 06CE, permanent. Cite it as [Bourbaki, Tag 06CE].

Algebra, IV, p. 35, § 4. Formal power series

Let (gi)i(g_i)_i be a family of elements of order ≥2\geq 2 in A[[I]]\mathbf{A}[[\mathbf{I}]]. When I\mathbf{I} is infinite, assume that gig_i tends to 0 along the filter of complements of finite subsets of I\mathbf{I}. There exists one and only one automorphism T of the topological A*-algebra* A[[I]]\mathbf{A}[[\mathbf{I}]] such that T(Xi)=Xi+giT(X_i) = X_i + g_i for all i∈Ii \in \mathbf{I}. *Further,

(23)

ω(T(u)−u)≥ω(u)+1 \omega(T(u) - u) \geq \omega(u) + 1
for each u∈A[[I]]u \in \mathbf{A}[[\mathbf{I}]].

The series fi=Xi+gif_i = X_i + g_i has no constant term and when I\mathbf{I} is infinite, fif_i tends to 0 along the filter of complements of finite subsets of I\mathbf{I}. Therefore (IV, p. 28, Prop. 4) there exists precisely one continuous endomorphism T of the A-algebra A[[I]]\mathbf{A}[[\mathbf{I}]] such that T(Xi)=fiT(X_i) = f_i for all i∈Ii \in \mathbf{I}. For each ν∈N′\nu \in \mathbf{N}' we put

vν=T(Xν)−Xν=∏i∈I(Xi+gi)ν(i)−∏i∈IXiν(i); v_\nu = T(X^\nu) - X^\nu = \prod_{i \in \mathbf{I}} (X_i + g_i)^{\nu(i)} - \prod_{i \in \mathbf{I}} X_i^{\nu(i)} ;

the relations ω(gi)≥2\omega(g_i) \geq 2 imply w(vν)≥∣ν∣+1w(v_\nu) \geq |\nu| + 1, and the relation (23) follows at once from this.

Let us show that TT is injective. Given u∈A[[I]]u \in \mathbf{A}[[\mathbf{I}]] such that T(u)=0T(u) = 0, by (23) we have w(u)≥w(u)+1w(u) \geq w(u) + 1, which is impossible if u≠0u \neq 0 because w(u)w(u) would then be a positive integer.

For every formal series vv in A[[I]]\mathbf{A}[[\mathbf{I}]] we denote by Hn(v)H_n(v) its homogeneous component of degree nn. Let us put S0(v)=H0(v)S_0(v) = H_0(v) and define the continuous mappings S,:A[[I]]→A[[I]]S, : \mathbf{A}[[\mathbf{I}]] \to \mathbf{A}[[\mathbf{I}]] by the recursion equations

(24)

Sn(v)=Hn(v−T(∑k=0n−1Sk(v)))forn≥1 S_n(v) = H_n \left( v - T \left( \sum_{k=0}^{n-1} S_k(v) \right) \right) \quad \text{for} \quad n \geq 1

Put S(v)=∑n≥0Sn(v)S(v) = \sum_{n \geq 0} S_n(v); if ν∈N(I)\nu \in \mathbf{N}^{(\mathbf{I})} and n=∣ν∣n = |\nu|, then the coefficient Sν(v)S^\nu(v) of XνX^\nu in S(v)S(v) is equal to that of XνX^\nu in S,(ν)S,(\nu); since S,S, is a continuous mapping, the mapping Sν:A[[I]]→AS^\nu : \mathbf{A}[[\mathbf{I}]] \to \mathbf{A} is continuous. Hence, by the definition of the product topology on A[[I]]=AN(I)\mathbf{A}[[\mathbf{I}]] = \mathbf{A}^{\mathbf{N}^{(\mathbf{I})}}, the mapping S:A[[I]]→A[[I]]S : \mathbf{A}[[\mathbf{I}]] \to \mathbf{A}[[\mathbf{I}]] is continuous.

We shall prove the relation T(S(v))=vT(S(v)) = v for all v∈A[[I]]v \in \mathbf{A}[[\mathbf{I}]] which will complete the proof of the lemma. Let v∈A[[I]]v \in \mathbf{A}[[\mathbf{I}]], un=Sn(v)u_n = S_n(v) and u=S(v)u = S(v). Let nn be a positive integer such that

(25)n_n

ω(v−T(u))≥n. \omega(v - T(u)) \geq n .

We have w(u−(u0+…+un−1))≥nw(u - (u_0 + \ldots + u_{n-1})) \geq n, whence

ω(T(u)−T(u0+…+un−1)−un)≥n+1 \omega(T(u) - T(u_0 + \ldots + u_{n-1}) - u_n) \geq n + 1

by (23). Now the recursion equation (24) shows that

un=Hn(v−T(u0+⋯+un−1)) u_n = H_n(v - T(u_0 + \cdots + u_{n-1}))

By (26) the formal power series v−T(u)v - T(u) and v−T(u0+…+un−1)−unv - T(u_0 + \ldots + u_{n-1}) - u_n have the same homogeneous component of degree nn, and this component is zero, by (27). We thus have w(v−T(u))≥n+1w(v - T(u)) \geq n + 1, that is (25)n(25)_n implies (25)n+1(25)_{n+1}. Since (25)0(25)_0 clearly holds, we thus have ω(v−T(u))≥n\omega(v - T(u)) \geq n for every integer n≥0n \geq 0, whence v=T(u)=T(S(v))v = T(u) = T(S(v)), as was to be proved.

For the rest of this No. we shall, for any set II, denote by A{I}A\{I\} the set of families (fi)i∈I(f_i)_{i \in I} satisfying the following conditions: (i) for each i∈Ii \in I, fif_i is an element of A[[I]]A[[I]] without constant term; (ii) if II is infinite, fif_i tends to 0 along the filter of complements of finite subsets of II.

The set A{I}A\{I\} is a monoid for the composition law (f,g)↦f∘g(f, g) \mapsto f \circ g, with {Xi}i∈I\{X_i\}_{i \in I} as unit element. The set of invertible elements of A{I}A\{I\} is thus a group.

On the other hand, let EE be the monoid of all continuous unital endomorphisms of the AA-algebra A[[I]]A[[I]] leaving the ideal of all formal power series without constant term invariant. If f∈A{I}f \in A\{I\} and g∈A[[I]]g \in A[[I]], then the element g(f)g(f) is defined. For fixed ff, the mapping g↦g(f)g \mapsto g(f) of A[[I]]A[[I]] into itself is an element WfW_f of EE. If f1,f2∈A{I}f_1, f_2 \in A\{I\} and g∈A[[I]]g \in A[[I]], we have, by formula (5) (IV, p. 30)

Wf1∘f2(g)=g(f1∘f2)=g(f1)∘f2=Wf2(Wf1(g)) W_{f_1 \circ f_2}(g) = g(f_1 \circ f_2) = g(f_1) \circ f_2 = W_{f_2}(W_{f_1}(g))

hence f↦Wff \mapsto W_f is a homomorphism of the monoid opposite to A{I}A\{I\} into EE. By Prop. 4 (IV, p. 28) this homomorphism is bijective.

Let f=(fi)i∈I∈A{I}f = (f_i)_{i \in I} \in A\{I\} and let ∑j∈IαijXj\sum_{j \in I} \alpha_{ij} X_j be the homogeneous component of degree 1 of fif_i. For any fixed jj in II we have αij=0\alpha_{ij} = 0 except for a finite number of suffixes ii, by hypothesis (ii) above. If (λi)∈A(I)(\lambda_i) \in A^{(I)}, we thus have (∑j∈Iαijλj)∈A(I)\left( \sum_{j \in I} \alpha_{ij} \lambda_j \right) \in A^{(I)}.

We denote by TfT_f the AA-linear mapping 1^1

(λi)↦(∑j∈Iαijλj) (\lambda_i) \mapsto \left( \sum_{j \in I} \alpha_{ij} \lambda_j \right)

of A(I)A^{(I)} into A(I)A^{(I)}. If g∈A{I}g \in A\{I\}, it is easily verified that

Tf∘g=Tf∘Tg. T_{f \circ g} = T_f \circ T_g .

1^1 Sometimes TfT_f is called the linear mapping tangent to ff.

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

Cited by

Cites

Transcribed from the printed edition, 2003, Springer. Not translated.