Documentation

TauCeti.RingTheory.WittVector.GaussValuation

Gauss valuations on Witt vectors #

Let O be the ring of integers of a valuation v : Valuation K ℝ≥0 on a field K, perfect of characteristic p; for instance the ring of integers 𝒪_F of a perfect nonarchimedean field F of characteristic p, so that 𝕎 O = A_inf. Write a Witt vector as its Teichmüller expansion x = ∑ₙ [xₙ] pⁿ (WittVector.teichmullerCoeff). For ρ < 1 the Gauss valuation

λ_ρ(∑ₙ [xₙ] pⁿ) = supₙ v(xₙ) ρⁿ

is a valuation on 𝕎 O. It takes the value v(a) ρ ^ m on [a] p ^ m, so λ_ρ(p) = ρ and λ_ρ([a]) = v(a), and for 0 < ρ it vanishes only at 0. These are the rank-one valuations from which the Gauss points of Spa(A_inf, A_inf) and the norms of the interval rings of the Fargues–Fontaine curve are built.

The ultrametric inequality rests on a description of the ideals ([c], p ^ (n + 1)) of 𝕎 O: they consist exactly of the Witt vectors whose first n + 1 Teichmüller coordinates have valuation at most v(c) (Valuation.Integers.mem_span_teichmuller_pow_iff), which uses that O is a valuation ring. Multiplicativity follows by comparing the leading Teichmüller terms of a product modulo a suitable power of p.

Main definitions #

Main results #

References #

theorem Valuation.Integers.mem_span_teichmuller_pow_iff {p : ℕ} [Fact (Nat.Prime p)] {R : Type u_1} [CommRing R] [CharP R p] [PerfectRing R p] {K : Type u_2} [Field K] {Γ₀ : Type u_3} [LinearOrderedCommGroupWithZero Γ₀] {v : Valuation K Γ₀} [Algebra R K] (hv : v.Integers R) {c : R} {x : WittVector p R} {n : ℕ} :
x ∈ Ideal.span {(WittVector.teichmuller p) c, ↑p ^ (n + 1)} ↔ ∀ i ≤ n, v ((algebraMap R K) (x.teichmullerCoeff i)) ≤ v ((algebraMap R K) c)

Let R be the ring of integers of a valuation v. Then the ideal ([c], p ^ (n + 1)) of 𝕎 R consists exactly of the Witt vectors whose first n + 1 Teichmüller coordinates have valuation at most v(c).

noncomputable def TauCeti.WittVector.gaussValuation (p : ℕ) [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} (hv : v.Integers O) (ρ : NNReal) (hρ : ρ < 1) :

The Gauss valuation λ_ρ(∑ₙ [xₙ] pⁿ) = supₙ v(xₙ) ρⁿ on the Witt vectors of the ring of integers O of a valuation v : Valuation K ℝ≥0, for O perfect of characteristic p and ρ < 1.

Instances For
    theorem TauCeti.WittVector.gaussValuation_apply {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ : ρ < 1) (x : WittVector p O) :
    (gaussValuation p hv ρ hρ) x = ⨆ (n : ℕ), v ((algebraMap O K) (x.teichmullerCoeff n)) * ρ ^ n

    The Gauss valuation is the supremum of the weighted valuations of the Teichmüller coordinates.

    theorem TauCeti.WittVector.le_gaussValuation {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ : ρ < 1) (x : WittVector p O) (n : ℕ) :
    v ((algebraMap O K) (x.teichmullerCoeff n)) * ρ ^ n ≤ (gaussValuation p hv ρ hρ) x

    Each weighted valuation v(xₙ) ρⁿ of a Teichmüller coordinate is bounded by the Gauss valuation.

    theorem TauCeti.WittVector.gaussValuation_mono {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} (hv : v.Integers O) {ρ₁ ρ₂ : NNReal} (hρ₁ : ρ₁ < 1) (hρ₂ : ρ₂ < 1) (hρ : ρ₁ ≤ ρ₂) (x : WittVector p O) :
    (gaussValuation p hv ρ₁ hρ₁) x ≤ (gaussValuation p hv ρ₂ hρ₂) x

    Increasing the radius increases the Gauss valuation of a Witt vector.

    theorem TauCeti.WittVector.gaussValuation_mul_inv_pow_le_max {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) {ρ₁ ρ₂ : NNReal} (hρ : ρ < 1) (hρ₁ : ρ₁ < 1) (hρ₂ : ρ₂ < 1) (h₀ : 0 < ρ₁) (h₁ : ρ₁ ≤ ρ) (h₂ : ρ ≤ ρ₂) (x : WittVector p O) (m : ℕ) :
    (gaussValuation p hv ρ hρ) x * (ρ ^ m)⁻¹ ≤ max ((gaussValuation p hv ρ₁ hρ₁) x * (ρ₁ ^ m)⁻¹) ((gaussValuation p hv ρ₂ hρ₂) x * (ρ₂ ^ m)⁻¹)

    The Hadamard three circles inequality for Gauss valuations. For radii 0 < ρ₁ ≤ ρ ≤ ρ₂ and m : ℕ, the value λ_ρ(x) / ρ ^ m, which is the Gauss valuation at ρ of x / p ^ m once p is inverted, is at most the larger of its values at ρ₁ and ρ₂: each term v(xₙ) ρ ^ (n - m) is monotone in ρ, increasing if m ≤ n and decreasing otherwise.

    theorem TauCeti.WittVector.gaussValuation_le_one {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ : ρ < 1) (x : WittVector p O) :
    (gaussValuation p hv ρ hρ) x ≤ 1

    The Gauss valuation is at most one.

    @[simp]
    theorem TauCeti.WittVector.gaussValuation_teichmuller_mul_pow {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ : ρ < 1) (a : O) (m : ℕ) :
    (gaussValuation p hv ρ hρ) ((WittVector.teichmuller p) a * ↑p ^ m) = v ((algebraMap O K) a) * ρ ^ m

    The Gauss valuation of [a] p ^ m is v(a) ρ ^ m.

    @[simp]
    theorem TauCeti.WittVector.gaussValuation_teichmuller {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ : ρ < 1) (a : O) :
    (gaussValuation p hv ρ hρ) ((WittVector.teichmuller p) a) = v ((algebraMap O K) a)

    The Gauss valuation of a Teichmüller representative [a] is v(a).

    @[simp]
    theorem TauCeti.WittVector.gaussValuation_p {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ : ρ < 1) :
    (gaussValuation p hv ρ hρ) ↑p = ρ

    The Gauss valuation of p is ρ.

    theorem TauCeti.WittVector.gaussValuation_mul_pow_le_of_eq_sum_add_pow_mul {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ : ρ < 1) {y b : WittVector p O} {m : ℕ} (hb : y = ∑ i ∈ Finset.range m, (WittVector.teichmuller p) (y.teichmullerCoeff i) * ↑p ^ i + ↑p ^ m * b) :
    (gaussValuation p hv ρ hρ) b * ρ ^ m ≤ (gaussValuation p hv ρ hρ) y

    In the splitting y = ∑_{i < m} [yᵢ] pⁱ + pᵐ b of the Teichmüller expansion (WittVector.exists_eq_sum_teichmullerCoeff_add_pow_mul), the tail satisfies λ_ρ(b) ρᵐ ≤ λ_ρ(y).

    theorem TauCeti.WittVector.gaussValuation_le_of_coeff_zero_eq_zero {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ : ρ < 1) {x : WittVector p O} (hx : x.coeff 0 = 0) :
    (gaussValuation p hv ρ hρ) x ≤ ρ

    A Witt vector whose 0-th coordinate vanishes, that is, a multiple of p, has Gauss valuation at most ρ.

    theorem TauCeti.WittVector.gaussValuation_teichmuller_add_teichmuller_sub_le {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ : ρ < 1) (a b : O) :
    (gaussValuation p hv ρ hρ) ((WittVector.teichmuller p) a + (WittVector.teichmuller p) b - (WittVector.teichmuller p) (a + b)) ≤ ρ * max (v ((algebraMap O K) a)) (v ((algebraMap O K) b))

    Teichmüller representatives are additive up to p, quantitatively: λ_ρ([a] + [b] - [a + b]) ≤ ρ max (v(a), v(b)).

    @[simp]
    theorem TauCeti.WittVector.gaussValuation_eq_zero_iff {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ₀ : 0 < ρ) (hρ : ρ < 1) {x : WittVector p O} :
    (gaussValuation p hv ρ hρ) x = 0 ↔ x = 0

    For 0 < ρ, the Gauss valuation vanishes only at zero.

    theorem TauCeti.WittVector.gaussValuation_pos {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ₀ : 0 < ρ) (hρ : ρ < 1) {x : WittVector p O} (hx : x ≠ 0) :
    0 < (gaussValuation p hv ρ hρ) x

    For 0 < ρ, the Gauss valuation of a nonzero Witt vector is positive.

    theorem TauCeti.WittVector.isContinuous_gaussValuation {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} [TopologicalSpace (WittVector p O)] {ϖ : O} (hI : IsAdic (Ideal.span {↑p, (WittVector.teichmuller p) ϖ})) (hv : v.Integers O) (hρ : ρ < 1) (hϖ : v ((algebraMap O K) ϖ) < 1) :

    The Gauss valuations are continuous for the (p, [ϖ])-adic topology on 𝕎 O, when v(ϖ) < 1: λ_ρ is at most 1 everywhere, and at most max ρ v(ϖ) < 1 at p and [ϖ].

    The degree #

    For 0 < ρ < 1 and x ≠ 0, the weighted valuations v(xₙ) ρⁿ attain their supremum λ_ρ(x) at finitely many indices. Kedlaya calls the largest of them the degree of x. Viewing x as a power series in p with Teichmüller coefficients, it is the degree of the leading part of x for λ_ρ, and it behaves like the degree of a polynomial: it is additive on products and only depends on x up to terms of smaller Gauss valuation.

    Both facts come from the radii slightly larger than ρ. There λ_ρ'(x) = v(x_d) ρ' ^ d for the degree d, because enlarging the radius favours the larger indices among those attaining λ_ρ(x). Since every λ_ρ' is a valuation, the degree is the exponent of this power law, and it is computed from the valuations λ_ρ' alone, with no reference to Witt vector arithmetic.

    noncomputable def TauCeti.WittVector.gaussDegree (p : ℕ) [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} (hv : v.Integers O) (ρ : NNReal) (hρ : ρ < 1) (x : WittVector p O) :

    The degree of a Witt vector for the Gauss valuation λ_ρ: the largest index d with v(x_d) ρ ^ d = λ_ρ(x), for x ≠ 0 and 0 < ρ < 1.

    Instances For
      @[simp]
      theorem TauCeti.WittVector.gaussDegree_zero {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ : ρ < 1) :
      gaussDegree p hv ρ hρ 0 = 0

      The degree of 0 is 0: every index attains its Gauss valuation 0.

      theorem TauCeti.WittVector.gaussValuation_eq_mul_pow_gaussDegree {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ : ρ < 1) (x : WittVector p O) :
      (gaussValuation p hv ρ hρ) x = v ((algebraMap O K) (x.teichmullerCoeff (gaussDegree p hv ρ hρ x))) * ρ ^ gaussDegree p hv ρ hρ x

      Every Witt vector attains its Gauss valuation at its degree, including at radius zero: λ_ρ(x) = v(x_d) ρ ^ d for d = gaussDegree p hv ρ hρ x.

      theorem TauCeti.WittVector.le_gaussDegree {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ₀ : 0 < ρ) (hρ : ρ < 1) {x : WittVector p O} (hx : x ≠ 0) {n : ℕ} (hn : v ((algebraMap O K) (x.teichmullerCoeff n)) * ρ ^ n = (gaussValuation p hv ρ hρ) x) :
      n ≤ gaussDegree p hv ρ hρ x

      Every index at which a nonzero Witt vector attains its Gauss valuation is at most its degree.

      theorem TauCeti.WittVector.mul_pow_lt_gaussValuation_of_gaussDegree_lt {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ₀ : 0 < ρ) (hρ : ρ < 1) {x : WittVector p O} (hx : x ≠ 0) {n : ℕ} (hn : gaussDegree p hv ρ hρ x < n) :
      v ((algebraMap O K) (x.teichmullerCoeff n)) * ρ ^ n < (gaussValuation p hv ρ hρ) x

      Beyond its degree, the weighted valuations of a nonzero Witt vector are smaller than its Gauss valuation.

      theorem TauCeti.WittVector.eventually_gaussValuation_eq_mul_pow_gaussDegree {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ₀ : 0 < ρ) (hρ : ρ < 1) (x : WittVector p O) :
      ∀ᶠ (ρ' : NNReal) in nhdsWithin ρ (Set.Ici ρ), ∀ (hρ' : ρ' < 1), (gaussValuation p hv ρ' hρ') x = v ((algebraMap O K) (x.teichmullerCoeff (gaussDegree p hv ρ hρ x))) * ρ' ^ gaussDegree p hv ρ hρ x

      The Gauss valuations near ρ follow a power law of the degree. For ρ' slightly larger than ρ, the Gauss valuation λ_ρ'(x) of a Witt vector is attained at its degree d for λ_ρ: λ_ρ'(x) = v(x_d) ρ' ^ d.

      theorem TauCeti.WittVector.gaussDegree_eq_of_eventually_eq_mul_pow {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ₀ : 0 < ρ) (hρ : ρ < 1) {x : WittVector p O} (hx : x ≠ 0) {c : NNReal} {n : ℕ} (h : ∀ᶠ (ρ' : NNReal) in nhdsWithin ρ (Set.Ici ρ), ∀ (hρ' : ρ' < 1), (gaussValuation p hv ρ' hρ') x = c * ρ' ^ n) :
      gaussDegree p hv ρ hρ x = n

      The degree is the exponent of the power law. If, for all ρ' slightly larger than ρ, the Gauss valuation λ_ρ'(x) of a nonzero Witt vector is c ρ' ^ n, then n is the degree of x for λ_ρ.

      theorem TauCeti.WittVector.gaussDegree_eq_of_gaussValuation_sub_lt {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ₀ : 0 < ρ) (hρ : ρ < 1) {x y : WittVector p O} (h : (gaussValuation p hv ρ hρ) (x - y) < (gaussValuation p hv ρ hρ) x) :
      gaussDegree p hv ρ hρ y = gaussDegree p hv ρ hρ x

      The degree only depends on the leading part (Kedlaya, Remark 2.7): if λ_ρ(x - y) < λ_ρ(x), then x and y have the same degree.

      theorem TauCeti.WittVector.gaussDegree_mul {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ₀ : 0 < ρ) (hρ : ρ < 1) {x y : WittVector p O} (hx : x ≠ 0) (hy : y ≠ 0) :
      gaussDegree p hv ρ hρ (x * y) = gaussDegree p hv ρ hρ x + gaussDegree p hv ρ hρ y

      The degree is additive on products (Kedlaya, Lemma 2.6).

      theorem TauCeti.WittVector.gaussDegree_teichmuller_mul_pow {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ₀ : 0 < ρ) (hρ : ρ < 1) {a : O} (ha : a ≠ 0) (m : ℕ) :
      gaussDegree p hv ρ hρ ((WittVector.teichmuller p) a * ↑p ^ m) = m

      The degree of [a] p ^ m is m, for a ≠ 0.

      @[simp]
      theorem TauCeti.WittVector.gaussDegree_teichmuller_mul {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ : ρ < 1) {c : O} (hc : c ≠ 0) (x : WittVector p O) :
      gaussDegree p hv ρ hρ ((WittVector.teichmuller p) c * x) = gaussDegree p hv ρ hρ x

      Multiplying by a Teichmüller representative [c] with c ≠ 0 does not change the degree, including at radius zero.

      theorem TauCeti.WittVector.gaussDegree_sum_range_teichmuller_mul_pow_lt {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} {O : Type u_2} [Field K] [CommRing O] [Algebra O K] [CharP O p] [PerfectRing O p] {v : Valuation K NNReal} {ρ : NNReal} (hv : v.Integers O) (hρ₀ : 0 < ρ) (hρ : ρ < 1) {a : ℕ → O} {n : ℕ} (h : ∑ j ∈ Finset.range n, (WittVector.teichmuller p) (a j) * ↑p ^ j ≠ 0) :
      gaussDegree p hv ρ hρ (∑ j ∈ Finset.range n, (WittVector.teichmuller p) (a j) * ↑p ^ j) < n

      A nonzero finite Teichmüller sum ∑_{j < n} [aⱼ] pʲ has degree less than n.