Documentation

TauCeti.RingTheory.MvPowerSeries.TateAlgebra.Groebner

Restricted power series over a Euclidean normed ring are noetherian #

Let R be a complete ring whose norm is multiplicative and ultrametric, and let deg : R → ℕ be a Euclidean degree compatible with the norm:

Then for every finite set of variables σ and all positive radii c : σ → ℝ, the ring R{T/c} of restricted power series with the Gauss norm at the radii c is left noetherian (TauCeti.MvPowerSeries.isNoetherianRing_isRestricted_subring_of_division). This is Kedlaya's analogue of the Hilbert basis theorem, stated for an abstract Euclidean normed ring. Kedlaya proves it for the completion A^r of W(𝒪_L)[[x] : x ∈ L] for the Gauss norm λ_r, which is Euclidean in this sense for the degree read off from its Newton polygon (his Remark 2.7 and Proposition 2.9), and deduces that the rings B^I of the Fargues–Fontaine curve are strongly noetherian. A complete nonarchimedean division ring satisfies the hypotheses with deg = 0, so its Tate algebras at arbitrary radii are left noetherian (TauCeti.MvPowerSeries.isNoetherianRing_isRestricted_subring).

The argument gives more than finite generation. Every left ideal H has finitely many nonzero elements x such that each y ∈ H is a combination ∑ a x * x with ‖a x‖ * ‖x‖ ≤ ‖y‖ (TauCeti.MvPowerSeries.exists_finset_eq_sum_mul_isRestricted_subring_of_division): no cancellation between the terms is needed to represent y. For Tate algebras at unit polyradius the generators can be normalised to Gauss norm one, and the coefficients are then bounded by ‖y‖ (Ideal.exists_finset_norm_eq_one_isRestricted_subring). In particular, the map (a x)ₓ ↦ ∑ a x * x from the finite free module with the maximum norm onto H induces on H its subspace topology, making H complete and hence closed (TauCeti.MvPowerSeries.isClosed_ideal_isRestricted_subring_of_division).

Running the same Gröbner reduction on an arbitrary series y whenever y is strictly farther from 0 than from H — using an element of H strictly closer to y than ‖y‖ to supply the divisor at the leading index of y — shows that every left ideal is strictly closed: for every y, the coset y + H has an element of minimal Gauss norm (TauCeti.MvPowerSeries.exists_norm_sub_le_isRestricted_subring_of_division, Ideal.exists_norm_sub_le_isRestricted_subring), even when the coefficient ring is not spherically complete.

Proof #

The argument is a Gröbner-basis argument. Fix a monomial order on the exponents. The leading index of a nonzero restricted series is the largest exponent, for the monomial order, at which its weighted coefficient norm attains the Gauss norm. For an ideal H, the pairs (leading index, degree of the leading coefficient) of the nonzero elements of H lie in the well-quasi-ordered set (σ →₀ ℕ) × ℕ (Dickson's lemma). So finitely many elements of H realise all the minimal pairs, and every nonzero element of H has a pair above one of theirs.

Fix ε < 1 bounding, relative to the norm, the weighted coefficients of these finitely many elements beyond their leading indices. Let y have ‖y‖ ≤ M and ‖y‖ > θ ≥ ε M, and suppose some h ∈ H satisfies ‖y - h‖ < ‖y‖ (for instance h = y when y ∈ H). Then h shares the leading index and leading-coefficient degree of y. A reduction step divides the leading coefficient of y by the leading coefficient of a chosen x, shifted to the leading index of y. This replaces that coefficient by a remainder of smaller degree or of weighted norm at most θ. It changes coefficients at larger indices only by at most ε M ≤ θ, which by the compatibility of deg with the norm keeps the degrees of the large ones. Recording, at each index, one more than the degree of each coefficient of weighted norm above θ, reduction strictly decreases in the colexicographic order. That order is well founded:

This follows Kedlaya's proof, with two reorganisations: Dickson's lemma is applied directly to the pairs above instead of to Kedlaya's minimal leading degrees d_I, and his argument by contradiction on an infinite reduction sequence is replaced by the well-founded potential.

Main results #

References #

theorem TauCeti.MvPowerSeries.exists_finset_eq_sum_mul_isRestricted_subring_of_division {σ : Type u_1} {R : Type u_2} [NormedRing R] [IsUltrametricDist R] {c : σ → ℝ} [∀ (i : σ), Fact (0 < c i)] [NormMulClass R] [CompleteSpace R] [Finite σ] (deg : R → ℕ) (hdeg : ∀ (x y : R), ‖x - y‖ < ‖x‖ → deg y = deg x) (hdiv : ∀ (x y : R), x ≠ 0 → ∃ (z : R) (w : R), y = z * x + w ∧ ‖w‖ ≤ ‖y‖ ∧ (w = 0 ∨ deg w < deg x)) (H : Ideal ↥(MvPowerSeries.IsRestricted.subring c)) :
∃ (F : Finset ↥(MvPowerSeries.IsRestricted.subring c)), (∀ x ∈ F, x ∈ H ∧ x ≠ 0) ∧ ∀ y ∈ H, ∃ (a : ↥(MvPowerSeries.IsRestricted.subring c) → ↥(MvPowerSeries.IsRestricted.subring c)), y = ∑ x ∈ F, a x * x ∧ ∀ x ∈ F, ‖a x‖ * ‖x‖ ≤ ‖y‖

Strict generators of ideals of restricted power series, from the argument of Kedlaya's Theorem 3.2 for an abstract Euclidean normed ring. Let R be a complete ring with a multiplicative ultrametric norm, and deg : R → ℕ a degree such that deg y = deg x whenever ‖x - y‖ < ‖x‖, and such that for x ≠ 0 every y can be divided as y = z * x + w with ‖w‖ ≤ ‖y‖ and either w = 0 or deg w < deg x. Then for finitely many variables and all positive radii c, every left ideal H of the ring of restricted power series has a finite set F of nonzero elements such that every y ∈ H is a combination ∑ x ∈ F, a x * x with ‖a x‖ * ‖x‖ ≤ ‖y‖. In particular F generates H, and F is empty when H = ⊥.

theorem TauCeti.MvPowerSeries.isClosed_ideal_isRestricted_subring_of_division {σ : Type u_1} {R : Type u_2} [NormedRing R] [IsUltrametricDist R] {c : σ → ℝ} [∀ (i : σ), Fact (0 < c i)] [NormMulClass R] [CompleteSpace R] [Finite σ] (deg : R → ℕ) (hdeg : ∀ (x y : R), ‖x - y‖ < ‖x‖ → deg y = deg x) (hdiv : ∀ (x y : R), x ≠ 0 → ∃ (z : R) (w : R), y = z * x + w ∧ ‖w‖ ≤ ‖y‖ ∧ (w = 0 ∨ deg w < deg x)) (H : Ideal ↥(MvPowerSeries.IsRestricted.subring c)) :

Left ideals of restricted power series over a Euclidean normed ring are closed. Under the hypotheses of exists_finset_eq_sum_mul_isRestricted_subring_of_division, every left ideal H of IsRestricted.subring c has a finite strict generating system and is therefore complete and closed for the Gauss norm.

theorem TauCeti.MvPowerSeries.exists_norm_sub_le_isRestricted_subring_of_division {σ : Type u_1} {R : Type u_2} [NormedRing R] [IsUltrametricDist R] {c : σ → ℝ} [∀ (i : σ), Fact (0 < c i)] [NormMulClass R] [CompleteSpace R] [Finite σ] (deg : R → ℕ) (hdeg : ∀ (x y : R), ‖x - y‖ < ‖x‖ → deg y = deg x) (hdiv : ∀ (x y : R), x ≠ 0 → ∃ (z : R) (w : R), y = z * x + w ∧ ‖w‖ ≤ ‖y‖ ∧ (w = 0 ∨ deg w < deg x)) (H : Ideal ↥(MvPowerSeries.IsRestricted.subring c)) (y : ↥(MvPowerSeries.IsRestricted.subring c)) :
∃ h ∈ H, ∀ h' ∈ H, ‖y - h‖ ≤ ‖y - h'‖

Least-norm coset representatives for ideals of restricted power series. Under the hypotheses of exists_finset_eq_sum_mul_isRestricted_subring_of_division, every left ideal H is strictly closed: for every restricted series y, there is h ∈ H minimising ‖y - h‖ over H.

theorem TauCeti.MvPowerSeries.isNoetherianRing_isRestricted_subring_of_division {σ : Type u_1} {R : Type u_2} [NormedRing R] [IsUltrametricDist R] {c : σ → ℝ} [∀ (i : σ), Fact (0 < c i)] [NormMulClass R] [CompleteSpace R] [Finite σ] (deg : R → ℕ) (hdeg : ∀ (x y : R), ‖x - y‖ < ‖x‖ → deg y = deg x) (hdiv : ∀ (x y : R), x ≠ 0 → ∃ (z : R) (w : R), y = z * x + w ∧ ‖w‖ ≤ ‖y‖ ∧ (w = 0 ∨ deg w < deg x)) :

Kedlaya's Hilbert basis theorem for restricted power series, the argument of Kedlaya's Theorem 3.2 for an abstract Euclidean normed ring. Let R be a complete ring with a multiplicative ultrametric norm, and deg : R → ℕ a degree such that deg y = deg x whenever ‖x - y‖ < ‖x‖, and such that for x ≠ 0 every y can be divided as y = z * x + w with ‖w‖ ≤ ‖y‖ and either w = 0 or deg w < deg x. Then for finitely many variables and all positive radii c, the ring of restricted power series with its Gauss norm at the radii c is left noetherian.

Tate algebras are left noetherian. Over a complete nonarchimedean division ring K, the ring of restricted power series in finitely many variables at arbitrary positive radii c is left noetherian: K satisfies the hypotheses of isNoetherianRing_isRestricted_subring_of_division with the zero degree.

theorem Ideal.exists_finset_eq_sum_mul_isRestricted_subring {σ : Type u_1} {c : σ → ℝ} [∀ (i : σ), Fact (0 < c i)] {K : Type u_3} [NormedDivisionRing K] [IsUltrametricDist K] [CompleteSpace K] [Finite σ] (H : Ideal ↥(MvPowerSeries.IsRestricted.subring c)) :
∃ (F : Finset ↥(MvPowerSeries.IsRestricted.subring c)), (∀ x ∈ F, x ∈ H ∧ x ≠ 0) ∧ ∀ y ∈ H, ∃ (a : ↥(MvPowerSeries.IsRestricted.subring c) → ↥(MvPowerSeries.IsRestricted.subring c)), y = ∑ x ∈ F, a x * x ∧ ∀ x ∈ F, ‖a x‖ * ‖x‖ ≤ ‖y‖

Strict generators of ideals of Tate algebras. Over a complete nonarchimedean division ring K, in finitely many variables and at arbitrary positive radii c, every left ideal H of the Tate algebra has a finite set F of nonzero elements such that every y ∈ H is a combination ∑ x ∈ F, a x * x with ‖a x‖ * ‖x‖ ≤ ‖y‖.

theorem Ideal.exists_finset_norm_eq_one_isRestricted_subring {σ : Type u_1} {K : Type u_3} [NormedDivisionRing K] [IsUltrametricDist K] [CompleteSpace K] [Finite σ] (H : Ideal ↥(MvPowerSeries.IsRestricted.subring fun (x : σ) => 1)) :
∃ (F : Finset ↥(MvPowerSeries.IsRestricted.subring fun (x : σ) => 1)), (∀ x ∈ F, x ∈ H ∧ ‖x‖ = 1) ∧ ∀ y ∈ H, ∃ (a : ↥(MvPowerSeries.IsRestricted.subring fun (x : σ) => 1) → ↥(MvPowerSeries.IsRestricted.subring fun (x : σ) => 1)), y = ∑ x ∈ F, a x * x ∧ ∀ x ∈ F, ‖a x‖ ≤ ‖y‖

Strict generators of norm one. Over a complete nonarchimedean division ring K, every left ideal H of the Tate algebra K⟨X⟩ at unit polyradius in finitely many variables has a finite set F of generators of Gauss norm one such that every y ∈ H is a combination ∑ x ∈ F, a x * x with ‖a x‖ ≤ ‖y‖. The set F is empty when H = ⊥.

theorem Ideal.exists_norm_sub_le_isRestricted_subring {σ : Type u_1} {c : σ → ℝ} [∀ (i : σ), Fact (0 < c i)] {K : Type u_3} [NormedDivisionRing K] [IsUltrametricDist K] [CompleteSpace K] [Finite σ] (H : Ideal ↥(MvPowerSeries.IsRestricted.subring c)) (y : ↥(MvPowerSeries.IsRestricted.subring c)) :
∃ h ∈ H, ∀ h' ∈ H, ‖y - h‖ ≤ ‖y - h'‖

Strict closedness of ideals of Tate algebras ([Bosch, Güntzer, Remmert, Non-Archimedean Analysis][bosch-guntzer-remmert], §5.2.7, Proposition 2). Over a complete nonarchimedean division ring K, in finitely many variables and at arbitrary positive radii c, every left ideal H is strictly closed: for every y, there is h ∈ H minimising ‖y - h‖ over H.