Bourbaki

§ 4. Analytic functions (ultrametric case)

In this paragraph, it is assumed that the absolute value of K is ultrametric. We denote by (Ei)1≤i≤n(E_i)_{1 \leq i \leq n} a finite family of normed spaces over K, and by E the product space of the EiE_i, endowed with the norm:

∥x∥=sup⁡∥xi∥if x=(xi). \|x\| = \sup \|x_i\| \quad \text{if } x = (x_i).

We denote by F a separated polynormed space over K.

4.1. Convergent series

4.1.1. Let f=∑afaf = \sum_a f_a be a formal series belonging to P^(E1,…,En;F)\hat{P}(E_1, \ldots, E_n; F), (cf. Appendix). If γ\gamma is a continuous seminorm on F, and R=(Ri)R = (R_i) is a system of nn real numbers >0> 0, we put

∥f∥γ,R=sup⁡a∥fa∥γRα. \|f\|_{\gamma, R} = \sup_a \|f_a\|_\gamma R^\alpha.

The definitions and results of No. 3.1.1 (second paragraph) and No. 3.1.2 apply without change; in particular, we define the spaces

HR(E1,…,En;F)andH(E1,…,En;F). \mathscr{H}_R(E_1, \ldots, E_n; F) \quad \text{and} \quad \mathscr{H}(E_1, \ldots, E_n; F).

4.1.2. The canonical isomorphism jj of P^(E;F)\hat{P}(E; F) onto P^(E1,…,En;F)\hat{P}(E_1, \ldots, E_n; F) gives, by restriction, an isomorphism of topological vector spaces of HR(E;F)\mathscr{H}_R(E; F) onto H(R,…,R)(E1,…,En;F)\mathscr{H}_{(R, \ldots, R)}(E_1, \ldots, E_n; F) for every R∈R+∗R \in \mathbf{R}_+^*; it also gives an isomorphism of H(E;F)\mathscr{H}(E; F) onto H(E1,…,En;F)\mathscr{H}(E_1, \ldots, E_n; F). More precisely, if f=∑mfm∈P^(E;F)f = \sum_m f_m \in \hat{P}(E; F) and if j(f)=∑afaj(f) = \sum_a f_a, one has, for every continuous seminorm γ\gamma on F:

∥fm∥γ=sup⁡∣α∣=m∥fα∥γ \|f_m\|_\gamma = \sup_{|\alpha|=m} \|f_\alpha\|_\gamma
∥f∥γ,R=∥j(f)∥γ,(R,…,R). \|f\|_{\gamma, R} = \|j(f)\|_{\gamma, (R, \ldots, R)}.

4.1.3. Let f=∑afaf = \sum_a f_a be an element of H(E1,…,En;F)\mathscr{H}(E_1, \ldots, E_n; F); let I(f)I(f) be the set of R∈(R+∗)nR \in (\mathbf{R}_+^*)^n such that, for every continuous seminorm γ\gamma on F, the product ∥fα∥γRα\|f_\alpha\|_\gamma R^\alpha tends to zero when ∣α∣|\alpha| tends to infinity. The set I(f)I(f) is nonempty; it is called the strict convergence indicator of ff. The set Ω(f)\Omega(f) of the points

(log⁡R1,…,log⁡Rn)for R∈I(f)(\log R_1, \ldots, \log R_n) \quad \text{for } R \in I(f)

is a convex subset of Rn\mathbf{R}^n.

When n=1n = 1, the set I(f)I(f) is an interval of R\mathbf{R}, open on the left and open or closed on the right; its upper bound (finite or +∞+\infty) is denoted by ρ(f)\rho(f) and is called the strict radius of convergence of ff.

With the notation of 4.1.2, one has R∈I(f)R \in I(f) if and only if

(R,…,R)∈I(j(f)). (R, \ldots, R) \in I(j(f)).
The set of points x=(xi)x = (x_i) such that there exists R=(Ri)∈I(f)R = (R_i) \in I(f) with ∥xi∥≤Ri\|x_i\| \leq R_i for 1≤i≤n1 \leq i \leq n is called the strict domain of convergence of ff and is denoted by C(f)C(f). It is an open subset of EE, union of the polyballs
B(R)={x∈E∣∥xi∥≤Ri for 1≤i≤n}, B(R) = \{ x \in E \mid \|x_i\| \leq R_i \text{ for } 1 \leq i \leq n \},
for R∈I(f)R \in I(f).

4.1.4. The results of 3.1.7 and 3.1.8 remain exact, on replacing everywhere C~(f)\tilde{C}(f) by C(f)C(f) and H~R\tilde{\mathscr{H}}_R by HR\mathscr{H}_R.

4.1.5. Let F1,…,FmF_1, \ldots, F_m be complete normed spaces and suppose that FF is quasi-complete. Let f=(fi)1≤i≤mf = (f_i)_{1 \leq i \leq m}, with fi∈H(E1,…,En;Fi)f_i \in \mathscr{H}(E_1, \ldots, E_n; F_i) and let g∈H(F1,…,Fm;F)g \in \mathscr{H}(F_1, \ldots, F_m; F), such that the point (fi(0))1≤i≤m(f_i(0))_{1 \leq i \leq m} of EE belong to the strict domain of convergence of gg. Then, for every α∈Nm\alpha \in \mathbf{N}^m, the formal series gα∘fg_\alpha \circ f belong to H(E1,…,En;F)\mathscr{H}(E_1, \ldots, E_n; F) and the family of the gα∘fg_\alpha \circ f is summable in H(E1,…,En;F)\mathscr{H}(E_1, \ldots, E_n; F) (therefore a fortiori in P^(E1,…,En;F)\hat{P}(E_1, \ldots, E_n; F)). Its sum will be denoted by g∘fg \circ f.

More precisely, there exist R∈⋂iI(fi)R \in \bigcap_i I(f_i) and R′∈I(g)R' \in I(g) such that

sup⁡∣α∣>0∥fi,α∥Rα<Ri′(for 1≤i≤m). \sup_{|\alpha| > 0} \|f_{i,\alpha}\| R^\alpha < R'_i \quad (\text{for } 1 \leq i \leq m).
Under these conditions the formal series gα∘fg_\alpha \circ f belong to HR(E1,…,En;F)\mathscr{H}_R(E_1, \ldots, E_n; F) and the family of the gα∘fg_\alpha \circ f is summable in HR(E1,…,En;F)\mathscr{H}_R(E_1, \ldots, E_n; F). Finally, if x∈B(R)x \in B(R), then f(x)=(fi(x))f(x) = (f_i(x)) belong to C(g)C(g) and one has:
g(f(x))=(g∘f)(x). g(f(x)) = (g \circ f)(x).

Suppose moreover that, for each ii, there exists a family (eji)(e_j^i) of elements of EiE_i such that every element xx of EiE_i is the sum of a summable family (λjeji)(\lambda_j e_j^i) (with λj∈K\lambda_j \in K) such that ∥x∥=sup⁡j∣λj∣\|x\| = \sup_j |\lambda_j|. One can then in the preceding paragraph replace the condition sup⁡∣α∣>0∥fi,α∥Rα<Ri′\sup_{|\alpha| > 0} \|f_{i,\alpha}\| R^\alpha < R'_i by the condition ∥fi∥R≤Ri′\|f_i\|_R \leq R'_i.

4.1.6. Suppose that Ei=KE_i = K for 1≤i≤n1 \leq i \leq n. The space P^(Kn;F)\hat{P}(K^n; F) then identifies with the space of formal series in nn indeterminates X1,…,XnX_1, \ldots, X_n with coefficients in FF and an element ff of P^(Kn;F)\hat{P}(K^n; F) is written:

f=∑aXacawith ca∈F. f = \sum_a X^a c_a \quad \text{with } c_a \in F.
If R∈(R+∗)nR \in (\mathbf{R}_+^*)^n and if γ\gamma is a continuous seminorm on FF, one has:
∥f∥γ,R=sup⁡∥ca∥γ⋅Rα. \|f\|_{\gamma, R} = \sup \|c_a\|_\gamma \cdot R^\alpha.

The last paragraph of 3.1.10 remains valid, as does 3.1.11.

4.1.7. Suppose that K is a closed subfield of a complete valued field (necessarily ultrametric) L. For y∈Ei⊗KLy \in E_i \otimes_K L, put:

∥y∥=inf⁡k(sup⁡∣ak∣⋅∥xk∥) \|y\| = \inf_k (\sup |a_k| \cdot \|x_k\|)

the greatest lower bound being extended to all finite families of pairs (xk,ak)∈Ei×L(x_k, a_k) \in E_i \times L such that y=∑kxk⊗aky = \sum_k x_k \otimes a_k. One thus obtains a norm on the L-vector space Ei⊗KLE_i \otimes_K L, which induces on EiE_i the given norm. The completion of Ei⊗KLE_i \otimes_K L for this norm is denoted by EiLE_i^L.

Let now F be a separated complete polynormed L-vector space. For every K-continuous polynomial fαf_\alpha, homogeneous of multidegree α\alpha, on E1×⋯×EnE_1 \times \cdots \times E_n, with values in the underlying polynormed K-vector space FKF_K of F, there exists one and only one L-continuous polynomial f~α\tilde{f}_\alpha, homogeneous of the same multidegree, on E1L×⋯×EnLE_1^L \times \cdots \times E_n^L, with values in F, which extends fαf_\alpha. For every continuous seminorm on the L-vector space F, one has

∥f~α∥i=∥fα∥i. \|\tilde{f}_\alpha\|_i = \|f_\alpha\|_i.

If f=∑fα∈H(E1,…,En;FK)f = \sum f_\alpha \in \mathscr{H}(E_1, \ldots, E_n; F_K), then f~=∑f~α∈H(E1L,…,EnL;F)\tilde{f} = \sum \tilde{f}_\alpha \in \mathscr{H}(E_1^L, \ldots, E_n^L; F).

The series ff and f~\tilde{f} have the same strict convergence indicator (and the same strict radius of convergence when n=1n = 1).

Conversely, let L be a closed non-discrete subfield of K and let Ei0E_i^0 and F0F^0 be the spaces over L obtained by restriction of scalars from the EiE_i and F. If f=∑fα∈H(E1,…,En;F)f = \sum f_\alpha \in \mathscr{H}(E_1, \ldots, E_n; F), then fα∈Pα(E10,…,En0;F0)f_\alpha \in P_\alpha(E_1^0, \ldots, E_n^0; F^0); if one sets f0=∑f~α∈P^(E10,…,En0;F0)f^0 = \sum \tilde{f}_\alpha \in \hat{P}(E_1^0, \ldots, E_n^0; F^0), then f0∈H(E10,…,En0;F0)f^0 \in \mathscr{H}(E_1^0, \ldots, E_n^0; F^0).

One has C(f)⊂C(f0)C(f) \subset C(f^0) and f(x)=f0(x)f(x) = f^0(x) for every x∈C(f)x \in C(f).

4.2. Analytic functions

4.2.1. The definitions and results of 3.2.1 and 3.2.2 remain valid without changes.

4.2.2. With the notation of 3.2.2, the strict convergence indicator of the power series expansion of Δaf\Delta^a f at a point aa of U contains that of the power series expansion of ff at aa.

4.2.3. The results of 3.2.4, 3.2.5, 3.2.7, 3.2.8 and 3.2.11 remain exact. That of 3.2.6 also, provided in addition that K is assumed to be of characteristic zero.

4.2.4. Suppose F quasi-complete and let f∈H(E1,…,En;F)f \in \mathscr{H}(E_1, \ldots, E_n; F). The mapping x↦f(x)x \mapsto f(x) is analytic in C(f)C(f). For every a∈C(f)a \in C(f), the convergence indicator of the power series expansion of ff at aa is equal to that of ff.

4.3. Some inequalities

4.3.1. It is assumed that K satisfies at least one of the following conditions: (a) the residue field of K is infinite ; (b) the image of K by the mapping a↦∣a∣a \mapsto |a| is dense in R+\mathbf{R}_+. (In other words, it is assumed that K is not locally compact). Let f=∑αfα∈H(E1,…,En;F)f = \sum_\alpha f_\alpha \in \mathscr{H}(E_1, \ldots, E_n; F) and let R∈I(f)R \in I(f). One has:

sup⁡x∈B(R)∥f(x)∥y=sup⁡αsup⁡x∈B(R)∥fα(x)∥y \sup_{x \in B(R)} \|f(x)\|_y = \sup_\alpha \sup_{x \in B(R)} \|f_\alpha(x)\|_y
for every continuous seminorm γ\gamma on F (“Cauchy inequalities”).

4.3.2. There exists a constant a>0a > 0 such that for every continuous homogeneous polynomial fα∈Pα(E1,…,En;F)f_\alpha \in P_\alpha(E_1, \ldots, E_n; F) and every R∈(R+∗)nR \in (\mathbf{R}_+^*)^n, one has:

a∣α∣Rα∣α∣!∥fα∥y≤sup⁡x∈B(R)∥fα(x)∥y≤∥fα∥yRα a^{|\alpha|} R^{\alpha} |\alpha|! \|f_\alpha\|_y \leq \sup_{x \in B(R)} \|f_\alpha(x)\|_y \leq \|f_\alpha\|_y R^\alpha
for every continuous seminorm γ\gamma on F. If K satisfies condition (b) of 4.3.1 or if the image of EiE_i by the mapping x↦∥x∥x \mapsto \|x\| is contained in the image of K by the mapping a↦∣a∣a \mapsto |a| and contains RiR_i (for 1≤i≤n1 \leq i \leq n), one can take a=1a = 1.

4.3.3. If K is of characteristic zero, the formal series f=∑αfαf = \sum_\alpha f_\alpha belongs to H(E1,…,En;F)\mathscr{H}(E_1, \ldots, E_n; F) if and only if there exists R∈(R+∗)nR \in (\mathbf{R}_+^*)^n such that

sup⁡αsup⁡x∈B(R)∥fα(x)∥y<+∞ \sup_\alpha \sup_{x \in B(R)} \|f_\alpha(x)\|_y < +\infty
for every continuous seminorm γ\gamma on F.

Machine translation of the English, by gpt-5-6, against glossary version 34. Not checked by a person.