Closed

Explicit base collars and their whole Gauss fibre

Lemma. (Explicit base collars and their whole Gauss fibre). The domain YεY_\varepsilon is a nonempty generalized affinoid, pure of dimension dd, and η∈Int⁡(Yε/K)\eta\in\operatorname{Int}(Y_\varepsilon/K). For every positive array aa, qaq_a is finite, with qa−1(TP)=M ⁣(RP[T1,…,Tp]/(Pℓ,a(Tℓ)−Sℓ)ℓ=1p),(1)q_a^{-1}(\mathbb T_P) =\mathcal M\!\left( R_P[T_1,\ldots,T_p]/ (P_{\ell,a}(T_\ell)-S_\ell)_{\ell=1}^p\right), \tag{1} where SℓS_\ell are the first pp target coordinates. The displayed algebra is free of rank DaD_a over RPR_P and has its canonical finite-module norm.

Put bℓ=min⁡jaℓjb_\ell=\min_j a_{\ell j}. If PP satisfies ∣uℓ∣<bℓε (1≤ℓ≤p),∣up+b∣<ε (1≤b≤k)for every u∈P,(2)|u_\ell|<b_\ell\varepsilon\ (1\leq\ell\leq p), \qquad |u_{p+b}|<\varepsilon\ (1\leq b\leq k) \quad\hbox{for every }u\in P, \tag{2} then the entire inverse image in (1) lies in the ordinary interior of YεY_\varepsilon in ZanZ^{\rm an}. Consequently it is exactly the saturated collar (Qatg∣Yε)−1(P)(Q_at_g|_{Y_\varepsilon})^{-1}(P).

If ηa\eta_a is the target unit Gauss point, then qa−1(ηa)={η}.(3)q_a^{-1}(\eta_a)=\{\eta\}. \tag{3} For its actual Artinian fibre algebra EaE_a, dim⁡H(ηa)Ea=Da=∑m∈Max⁡Ealength⁡((Ea)m)[κ(m):H(ηa)].(4)\dim_{\mathcal H(\eta_a)}E_a=D_a =\sum_{\mathfrak m\in\operatorname{Max}E_a} \operatorname{length}((E_a)_{\mathfrak m}) [\kappa(\mathfrak m):\mathcal H(\eta_a)] . \tag{4} Equation (3) concerns points; (4) retains all length and field-degree data and does not replace EaE_a by its reduction.

Proof

Proof. Set R=eε>1R=e^\varepsilon>1. A presentation of YεY_\varepsilon uses tℓ,zb,zb−1,(tℓ−cℓj)−1t_\ell,z_b,z_b^{-1},(t_\ell-c_{\ell j})^{-1}, all with radius RR, and their polynomial inverse relations. Indeed ∣tℓ∣≤R|t_\ell|\leq R and ∣cℓj∣≤1<R|c_{\ell j}|\leq1<R give every upper bound ∣tℓ−cℓj∣≤R|t_\ell-c_{\ell j}|\leq R; conversely one such upper bound gives ∣tℓ∣≤R|t_\ell|\leq R. The inverse coordinates give the lower bounds. This is an actual generalized affinoid presentation. All its coordinates have value 1<R1<R at η\eta, so the presentation witnesses absolute interiority there. It is an affinoid domain of the closed polyannulus in (tℓ−cℓ1,zb)(t_\ell-c_{\ell1},z_b), cut out by the additional puncture bounds. That polyannulus is integral and pure of dimension dd by 5.35.4; domain purity 5.35.3 proves the assertion for YεY_\varepsilon.

The algebraic coordinate ring of ZZ is K[S1±1,…,Sp±1,Z1±1,…,Zk±1][T1,…,Tp]/(Pℓ,a(Tℓ)−Sℓ)ℓ.K[S_1^{\pm1},\ldots,S_p^{\pm1},Z_1^{\pm1},\ldots,Z_k^{\pm1}] [T_1,\ldots,T_p]/(P_{\ell,a}(T_\ell)-S_\ell)_\ell . To check the inverse coordinates in this identity, use (Tℓ−cℓj)−1=Sℓ−1(Tℓ−cℓj)aℓj−1∏h≠j(Tℓ−cℓh)aℓh.(T_\ell-c_{\ell j})^{-1} =S_\ell^{-1}(T_\ell-c_{\ell j})^{a_{\ell j}-1} \prod_{h\ne j}(T_\ell-c_{\ell h})^{a_{\ell h}}. Successive division by the separate monic polynomials gives the free basis ∏ℓTℓiℓ\prod_\ell T_\ell^{i_\ell}, 0≤iℓ<Aℓ0\leq i_\ell<A_\ell. The frozen analytification theorem 2.33.26 identifies the whole inverse image of any target affinoid with its ordinary tensor algebra and canonical finite-module norm. It gives (1).

Here is the estimate which supplies saturation. At any analytic point, fix a block ℓ\ell. If ∣tℓ∣>1|t_\ell|>1, all its puncture distances are equal to ∣tℓ∣|t_\ell|. If ∣tℓ∣≤1|t_\ell|\leq1, all those distances are at most 11, and at most one is less than 11, because the difference of two punctures has norm 11. Thus its weight vector belongs to Fℓ=⋃j=1mℓR≥0eℓj ∪ R≥0(−eℓ1−⋯−eℓmℓ).(5)F_\ell=\bigcup_{j=1}^{m_\ell}\mathbb R_{\geq0}e_{\ell j} \ \cup\ \mathbb R_{\geq0}(-e_{\ell1}-\cdots-e_{\ell m_\ell}). \tag{5} On the positive ray jj, uℓ=aℓjvℓju_\ell=a_{\ell j}v_{\ell j}; on the last ray, uℓ=−Aℓsu_\ell=-A_\ell s with all vℓj=−sv_{\ell j}=-s. In either case ∣vℓj∣≤∣uℓ∣/bℓ.(6)|v_{\ell j}|\leq |u_\ell|/b_\ell. \tag{6} The final kk weights are unchanged. Bounds (2) therefore put every inverse-image point in the strict moment inequalities. This proves the collar assertion without selecting any branch.

We prove (3) over the original field. The unit Gauss point maps to ηa\eta_a: after reduction, the separate nonconstant monic polynomials P‾ℓ,a(t‾ℓ)\overline P_{\ell,a}(\overline t_\ell), together with the z‾b\overline z_b, are algebraically independent over K~\widetilde K. For example, their distinct leading multidegrees prove the injectivity of polynomial substitution. Scaling a largest coefficient of any Laurent polynomial then proves the unit Gauss norm identity.

Conversely let yy map to ηa\eta_a. Equation (6), with u=0u=0, gives ∣tℓ−cℓj∣=∣zb∣=1|t_\ell-c_{\ell j}|=|z_b|=1 at yy, hence ∣tℓ∣≤1|t_\ell|\leq1. In the residue field of H(y)\mathcal H(y), the elements P‾ℓ,a(t‾ℓ),z‾b\overline P_{\ell,a}(\overline t_\ell),\overline z_b are algebraically independent: any relation would contradict the target Gauss seminorm after lifting its finitely many coefficients. Each t‾ℓ\overline t_\ell is integral over the field generated by these elements. Consequently the field generated by t‾1,…,t‾p,z‾1,…,z‾k\overline t_1,\ldots,\overline t_p,\overline z_1, \ldots,\overline z_k has transcendence degree dd. These dd elements are therefore algebraically independent. For any nonzero polynomial in t,zt,z, divide by a largest coefficient. Its reduction is nonzero at yy, so its value is the maximum coefficient norm. The same holds at η\eta. Inverting the puncture coordinates and the zbz_b proves equality on the entire algebraic coordinate ring, hence equality of the analytic points y=ηy=\eta. No separability or defectlessness assertion enters this argument.

Finally the finite tensor comparison 1.18.7 identifies EaE_a with the ordinary base change of the free module in (1). Its dimension is DaD_a. An Artinian composition series gives the second equality in (4), with full residue-field degrees. ◻

Citation

Non-Archimedean Geometry Foundations, Tag NAG-002686.

Proof in chapter · Fixed-edition record