Closed

The Gauss section

Theorem 1.19.1. (The Gauss section). Let KK be a complete real-valued non-Archimedean field, possibly trivially valued, and let AA be a commutative unital Banach KK-algebra with a homogeneous submultiplicative norm. Let r=(r1,…,rs)r=(r_1,\ldots,r_s) be finite and free over KK, allowing the empty tuple. Put Ar=A⊗^KKrA_r=A\widehat\otimes_K K_r, using separated completed projective tensor products. In the coefficient description of Theorem 1.16.2, the formula σr(x)(∑ν∈ZsaνTν)=max⁡ν∈Zs∣aν(x)∣rν\sigma_r(x)\left(\sum_{\nu\in\mathbf Z^s}a_\nu T^\nu\right) =\max_{\nu\in\mathbf Z^s}|a_\nu(x)|r^\nu defines a bounded multiplicative seminorm on ArA_r extending the specified absolute value of KrK_r, for every x∈MK(A)x\in\mathcal M_K(A); the value is zero when all coefficients have value zero. It gives a continuous section σr:MK(A)⟶MKr(Ar)ofπr:MKr(Ar)⟶MK(A).\sigma_r:\mathcal M_K(A)\longrightarrow\mathcal M_{K_r}(A_r) \quad\text{of}\quad \pi_r:\mathcal M_{K_r}(A_r)\longrightarrow\mathcal M_K(A). Thus πr\pi_r is surjective. The sections commute with bounded unital KK-algebra homomorphisms. No affinoid, reducedness, spherical-completeness or residue-field freeness assumption is imposed. For the zero algebra both spectra are empty.

Proof

Proof. Step 1: convergence and multiplication. A bounded multiplicative seminorm xx on any such Banach algebra is contractive. Indeed, if its bound is CC, then ∣a(x)∣n≤C∥an∥≤C∥a∥n|a(x)|^n\leq C\|a^n\|\leq C\|a\|^n; taking nnth roots proves ∣a(x)∣≤∥a∥|a(x)|\leq\|a\|. This also gives ∣∣a(x)∣−∣b(x)∣∣≤∥a−b∥\bigl||a(x)|-|b(x)|\bigr|\leq\|a-b\|. Consequently the coefficient weights in the formula tend to zero. A positive supremum is attained, and only finitely many terms attain it. The formula is ultrametric, bounded by the coefficient norm, and agrees with xx on constants.

For finite tensor representations z=∑iai⊗uiz=\sum_i a_i\otimes u_i and w=∑jbj⊗vjw=\sum_j b_j\otimes v_j, multiplication gives ∥zw∥π≤max⁡i,j∥ai∥∥bj∥∣ui∣∣vj∣.\|zw\|_\pi\leq\max_{i,j}\|a_i\|\|b_j\||u_i||v_j|. Taking both infima proves submultiplicativity of the projective seminorm. Its zero-seminorm subspace is an ideal, so multiplication passes to separation and completion. Finite Laurent families are dense by Theorem 1.16.2. Their products therefore show that multiplication in ArA_r is convergent coefficient convolution. In particular, if F=∑aαTαF=\sum a_\alpha T^\alpha and G=∑bβTβG=\sum b_\beta T^\beta, then (FG)κ=∑α+β=κaαbβ.(FG)_\kappa=\sum_{\alpha+\beta=\kappa}a_\alpha b_\beta. For fixed κ\kappa, the terms converge to zero in AA: their norms, multiplied by rκr^\kappa, are bounded by (∥aα∥rα)(∥bκ−α∥rκ−α)(\|a_\alpha\|r^\alpha)(\|b_{\kappa-\alpha}\|r^{\kappa-\alpha}), the product of a vanishing family and a bounded family. Completeness makes this sum converge independently of its enumeration. Continuous coefficient projection identifies it with the limit of finite products.

Step 2: multiplicativity with tied maximal terms. Write M=σr(x)(F)M=\sigma_r(x)(F) and N=σr(x)(G)N=\sigma_r(x)(G). Every term of every product coefficient has weighted xx-value at most MNMN. The preceding convergence and continuity of xx give σr(x)(FG)≤MN\sigma_r(x)(FG)\leq MN, including the case MN=0MN=0. Suppose M,N>0M,N>0, and let D,ED,E be the finite sets where the respective maxima are attained. There are 0≤λF,λG<10\leq\lambda_F,\lambda_G<1 bounding all other weights by λFM\lambda_FM, respectively λGN\lambda_GN: among the finitely many weights at least half the maximum, the nonmaximal ones have a strictly smaller maximum, and all remaining weights are at most half the maximum.

Choose the lexicographically greatest indices μ∈D\mu\in D and ν∈E\nu\in E. The pair (μ,ν)(\mu,\nu) is the only pair in D×ED\times E summing to κ=μ+ν\kappa=\mu+\nu: if α<μ\alpha<\mu, translation invariance forces κ−α>ν\kappa-\alpha>\nu. Every other summand of (FG)κ(FG)_\kappa therefore has weighted xx-value at most λMN\lambda MN, where λ=max⁡(λF,λG)<1\lambda=\max(\lambda_F,\lambda_G)<1. The entire convergent remainder satisfies this same bound. The distinguished term has weighted value MNMN, so the ultrametric inequality in both directions gives ∣(FG)κ(x)∣rκ=MN.|(FG)_\kappa(x)|r^\kappa=MN. This proves the reverse inequality and hence multiplicativity. Only an order on a finite set of maximal terms was used; no freeness over the field attached to xx was needed. For the empty tuple the only index is zero.

Step 3: scalars and topology. On the scalar copy of KrK_r, the coefficients lie in KK, and the formula is its specified absolute value from Lemma 1.16.1. Thus it is a KrK_r-spectral point. For a finite Laurent polynomial PP, the function x↦σr(x)(P)x\mapsto\sigma_r(x)(P) is a finite maximum of continuous evaluation functions. For every F∈ArF\in A_r and every such PP, uniformly in xx, ∣σr(x)(F)−σr(x)(P)∣≤∥F−P∥r.\bigl|\sigma_r(x)(F)-\sigma_r(x)(P)\bigr|\leq\|F-P\|_r. Finite truncations approximate FF in norm. Its evaluation is therefore a uniform limit of continuous functions, hence continuous. This is the required continuity for the evaluation topology. Constants give πrσr=id\pi_r\sigma_r=\mathrm{id}. A bounded homomorphism acts coefficientwise after extension; applying a point to each coefficient proves the asserted naturality. If A=0A=0, the coefficient space is zero and the statement is vacuous. ◻

Selected scopes (1)

Citation

Non-Archimedean Geometry Foundations, Tag NAG-001569, 1.19.1.

Proof in chapter · Fixed-edition record