Closed

Distinguished coordinates

Lemma 1.11.3. (Distinguished coordinates). For every nonzero g∈Rng\in R_n, n≥1n\geq 1, there is an isometric triangular KK-algebra automorphism ϕ\phi such that ϕ(g)\phi(g), multiplied by a nonzero scalar and a unit of Rn−1R_{n-1}, has the form f=Td+∑i<dbiTi+h,T=Xn,∥bi∥≤1,∥h∥<1,f=T^d+\sum_{i<d}b_iT^i+h,\qquad T=X_n,\quad \|b_i\|\leq1,\quad \|h\|<1, where hh contains only powers of TT greater than dd.

Proof

Proof. Divide gg by a coefficient attaining its nonzero Gauss norm, so that ∥g∥=1\|g\|=1. The set SS of exponents with coefficient norm one is finite and nonempty. Let PP be the sum of these terms and write g=P+h0g=P+h_0. Then ∥h0∥<1\|h_0\|<1: only finitely many coefficients have norm at least 1/21/2, and those among them that are less than one, if any, have a maximum less than one. Taking the maximum with 1/21/2 bounds all remaining coefficients strictly below one.

Choose an integer B≥2B\geq 2 larger than every coordinate of every exponent in SS, and put Ni=BiN_i=B^i for 1≤i<n1\leq i<n. The integers w(ν)=νn+∑i<nNiνiw(\nu)=\nu_n+\sum_{i<n}N_i\nu_i are distinct on SS by uniqueness of base-BB expansions. Make the substitution Xi⟼Xi+TNi(i<n),T⟼T.X_i\longmapsto X_i+T^{N_i}\quad(i<n),\qquad T\longmapsto T. Both this map and the substitution with minus signs are contractive by Proposition 1.2.3; they are inverse on polynomials and therefore on their completions. Hence ϕ\phi is isometric. Regrouping coefficients identifies RnR_n isometrically with Rn−1⟨T⟩R_{n-1}\langle T\rangle: vanishing of the joint coefficient tails is equivalent to vanishing in each row and convergence to zero of the row norms.

Set d=max⁡ν∈Sw(ν)d=\max_{\nu\in S}w(\nu). The coefficient of TdT^d in ϕ(P)\phi(P) is the scalar aa attached to the unique maximizing exponent. Every other term has smaller TT-degree, and all coefficient norms are at most one, since the binomial coefficients have absolute value at most one. The coefficient of TdT^d in ϕ(g)\phi(g) is thus c=a+ec=a+e with ∣a∣=1|a|=1 and ∥e∥<1\|e\|<1. Geometric inversion gives c−1∈Rn−1c^{-1}\in R_{n-1} and ∥c∥=∥c−1∥=1\|c\|=\|c^{-1}\|=1. Multiplying by c−1c^{-1} gives the required lower-coefficient bounds and strictly small higher tail. ◻

Citation

Non-Archimedean Geometry Foundations, Tag NAG-001382, 1.11.3.

Proof in chapter · Fixed-edition record