Documentation

TauCeti.NumberTheory.Transcendental.GelfondSchneider.SchneiderAuxiliary

Schneider's auxiliary function #

Let K ⊂ ℂ be a number field of degree d, given by an embedding σ : K →+* ℂ, let λ ∈ ℂ and let α, β, γ ∈ K with σ α = exp λ and σ γ = exp (σ β λ). In Schneider's proof of the Gelfond–Schneider theorem the auxiliary function is the exponential polynomial

F(z) = ∑_{i < D₁, j < D₂} a_{ij} z ^ i exp (j λ z),

whose integer coefficients are chosen so that F vanishes at the points u + v σ β of the grid 1 ≤ u, v ≤ L. Since exp (j λ (u + v σ β)) = σ (α ^ (u j) γ ^ (v j)) (RingHom.map_pow_mul_pow_eq_exp), the value F(u + v σ β) is the image under σ of

y_{uv} = ∑_{i, j} a_{ij} (u + v β) ^ i α ^ (u j) γ ^ (v j) ∈ K,

so the vanishing conditions are L² linear equations over K in the D₁ D₂ integer unknowns a_{ij}. When D₁ D₂ ≥ 2 d L², Siegel's lemma with rational integer unknowns (NumberField.exists_ne_zero_int_vec_abs_le_of_house_le) solves them with |a_{ij}| ≤ exp (c (D₁ log L + D₂ L)), where c depends only on K, α, β and γ. To apply it, each equation is multiplied by m ^ (D₁ + 2 L D₂) for a nonzero integer m such that m α, m β and m γ are algebraic integers.

Schneider's proof then evaluates F on the larger grid u, v ≤ c L. The values y_{uv} there have house exp (O (D₁ log L + D₂ L)), which is exp (O (L ^ (3 / 2) log L)) for Schneider's parameters L = 2 d M², D₁ = (2 d)² M³, D₂ = 2 d M, and m ^ (D₁ + 2 c L D₂) y_{uv} is an algebraic integer, where m ^ (D₁ + 2 c L D₂) ≤ exp (O (L ^ (3 / 2))). These are the arithmetic inputs to Liouville's lower bound for a nonzero σ y_{uv}. The analytic upper bound for |F(u + v σ β)|, from the zero estimate for exponential polynomials and Schwarz's lemma with prescribed zeros, is not part of this file.

Main results #

References #

theorem TauCeti.isIntegral_intCast_pow_mul_schneider_term {K : Type u_1} [CommRing K] {m : ℤ} {α β γ : K} (hα : IsIntegral ℤ (↑m * α)) (hβ : IsIntegral ℤ (↑m * β)) (hγ : IsIntegral ℤ (↑m * γ)) {i j u v N : ℕ} (h : i + u * j + v * j ≤ N) :
IsIntegral ℤ (↑m ^ N * ((↑u + ↑v * β) ^ i * (α ^ (u * j) * γ ^ (v * j))))

One term of Schneider's value y_{uv}, multiplied by m ^ N with i + u j + v j ≤ N, is an algebraic integer when m α, m β and m γ are.

theorem TauCeti.isIntegral_intCast_pow_mul_schneider_value {K : Type u_1} [CommRing K] {m : ℤ} {α β γ : K} (hα : IsIntegral ℤ (↑m * α)) (hβ : IsIntegral ℤ (↑m * β)) (hγ : IsIntegral ℤ (↑m * γ)) {D₁ D₂ B : ℕ} (a : Fin D₁ → Fin D₂ → ℤ) {u v : ℕ} (hu : u ≤ B) (hv : v ≤ B) :
IsIntegral ℤ (↑m ^ (D₁ + 2 * B * D₂) * ∑ i : Fin D₁, ∑ j : Fin D₂, ↑(a i j) * (↑u + ↑v * β) ^ ↑i * (α ^ (u * ↑j) * γ ^ (v * ↑j)))

Denominators of Schneider's values. If m α, m β and m γ are algebraic integers, then for u, v ≤ B the multiple m ^ (D₁ + 2 B D₂) ∑_{i, j} a_{ij} (u + v β) ^ i α ^ (u j) γ ^ (v j) is an algebraic integer for all integers a_{ij}, i < D₁, j < D₂.

theorem TauCeti.house_schneider_entry_le {K : Type u_1} [Field K] [NumberField K] {x y z w : K} {L H : ℝ} (hx : NumberField.house x ≤ H) (hy : NumberField.house y ≤ L * H) (hz : NumberField.house z ≤ H) (hw : NumberField.house w ≤ H) (hL : 1 ≤ L) (hH : 0 ≤ H) {i s t n D : ℕ} (hi : i ≤ D) (hn : i + s + t ≤ n) :
NumberField.house (x ^ (n - (i + s + t)) * y ^ i * (z ^ s * w ^ t)) ≤ L ^ D * H ^ n

The house bound for an entry of Schneider's linear system: if x, z, w have house at most H and y has house at most L H, then for i ≤ D and i + s + t ≤ n the entry x ^ (n - (i + s + t)) y ^ i z ^ s w ^ t has house at most L ^ D H ^ n.

theorem TauCeti.exists_intCast_pow_mul_schneider_term {K : Type u_1} [Field K] [NumberField K] (α β γ : K) :
∃ (c₀ : ℤ), c₀ ≠ 0 ∧ ∃ (H : ℝ), 1 ≤ H ∧ |↑c₀| ≤ H ∧ ∀ (k u v j N B : ℕ), 1 ≤ B → u ≤ B → v ≤ B → k + u * j + v * j ≤ N → IsIntegral ℤ (↑c₀ ^ N * ((↑u + ↑v * β) ^ k * (α ^ (u * j) * γ ^ (v * j)))) ∧ NumberField.house (↑c₀ ^ N * ((↑u + ↑v * β) ^ k * (α ^ (u * j) * γ ^ (v * j)))) ≤ ↑B ^ k * H ^ N

A common denominator for the terms of Schneider's and Gelfond's values: there are a nonzero integer c₀ and H ≥ 1 with |c₀| ≤ H such that for u, v ≤ B (B ≥ 1) and k + u j + v j ≤ N the multiple c₀ ^ N ((u + v β) ^ k α ^ (u j) γ ^ (v j)) is an algebraic integer of house at most B ^ k H ^ N.

theorem TauCeti.exists_schneider_auxiliary_algebraic {K : Type u_1} [Field K] [NumberField K] (α β γ : K) :
∃ (c : ℝ), 0 < c ∧ ∀ (L D₁ D₂ : ℕ), 3 ≤ L → 2 * Module.finrank ℚ K * L ^ 2 ≤ D₁ * D₂ → ∃ (a : Fin D₁ → Fin D₂ → ℤ), a ≠ 0 ∧ (∀ (i : Fin D₁) (j : Fin D₂), |↑(a i j)| ≤ Real.exp (c * (↑D₁ * Real.log ↑L + ↑D₂ * ↑L))) ∧ ∀ (u v : ℕ), 1 ≤ u → u ≤ L → 1 ≤ v → v ≤ L → ∑ i : Fin D₁, ∑ j : Fin D₂, ↑(a i j) * (↑u + ↑v * β) ^ ↑i * (α ^ (u * ↑j) * γ ^ (v * ↑j)) = 0

Schneider's auxiliary function, algebraic form (Evertse, Lemma 4.22). Let K be a number field of degree d and α, β, γ ∈ K. There is c > 0 such that for all integers L ≥ 3 and D₁, D₂ with D₁ D₂ ≥ 2 d L² there are integers a_{ij}, i < D₁, j < D₂, not all zero, with |a_{ij}| ≤ exp (c (D₁ log L + D₂ L)), such that ∑_{i, j} a_{ij} (u + v β) ^ i α ^ (u j) γ ^ (v j) = 0 for all 1 ≤ u, v ≤ L.

Size and denominators of the values y_{uv} #

Schneider's proof evaluates F at the points u + v σ β of a grid larger than the one on which it was made to vanish. The value there is the image under σ of y_{uv} = ∑_{i, j} a_{ij} (u + v β) ^ i α ^ (u j) γ ^ (v j) ∈ K, and the final contradiction plays an upper bound for |σ y_{uv}| against Liouville's lower bound, which needs the house and a denominator of y_{uv}.

theorem TauCeti.house_schneider_value_le {K : Type u_1} [Field K] [NumberField K] {α β γ : K} {H A : ℝ} (hH : 1 ≤ H) (hα : NumberField.house α ≤ H) (hβ : NumberField.house β ≤ H) (hγ : NumberField.house γ ≤ H) {D₁ D₂ B : ℕ} (hB : 1 ≤ B) {a : Fin D₁ → Fin D₂ → ℤ} (ha : ∀ (i : Fin D₁) (j : Fin D₂), |↑(a i j)| ≤ A) {u v : ℕ} (hu : u ≤ B) (hv : v ≤ B) :
NumberField.house (∑ i : Fin D₁, ∑ j : Fin D₂, ↑(a i j) * (↑u + ↑v * β) ^ ↑i * (α ^ (u * ↑j) * γ ^ (v * ↑j))) ≤ ↑D₁ * ↑D₂ * A * (2 * ↑B) ^ D₁ * H ^ (D₁ + 2 * B * D₂)

The house of Schneider's values. If α, β, γ have house at most H ≥ 1 and the integers a_{ij}, i < D₁, j < D₂, have absolute value at most A, then for u, v ≤ B with B ≥ 1 the house of ∑_{i, j} a_{ij} (u + v β) ^ i α ^ (u j) γ ^ (v j) is at most D₁ D₂ A (2 B) ^ D₁ H ^ (D₁ + 2 B D₂).

theorem TauCeti.exists_house_schneider_value_le_exp {K : Type u_1} [Field K] [NumberField K] (α β γ : K) (c₁ : ℝ) (c : ℕ) :
∃ (C : ℝ), 0 ≤ C ∧ ∀ (L D₁ D₂ : ℕ), 2 ≤ L → ∀ (a : Fin D₁ → Fin D₂ → ℤ), (∀ (i : Fin D₁) (j : Fin D₂), |↑(a i j)| ≤ Real.exp (c₁ * (↑D₁ * Real.log ↑L + ↑D₂ * ↑L))) → ∀ (u v : ℕ), u ≤ c * L → v ≤ c * L → NumberField.house (∑ i : Fin D₁, ∑ j : Fin D₂, ↑(a i j) * (↑u + ↑v * β) ^ ↑i * (α ^ (u * ↑j) * γ ^ (v * ↑j))) ≤ Real.exp (C * (↑D₁ * Real.log ↑L + ↑D₂ * ↑L))

Schneider's values on an extended grid. Let α, β, γ ∈ K, c₁ ∈ ℝ and c ∈ ℕ. There is C ≥ 0 such that for all integers L ≥ 2, D₁ and D₂, all integers a_{ij}, i < D₁, j < D₂, with |a_{ij}| ≤ exp (c₁ (D₁ log L + D₂ L)), and all u, v ≤ c L, the house of ∑_{i, j} a_{ij} (u + v β) ^ i α ^ (u j) γ ^ (v j) is at most exp (C (D₁ log L + D₂ L)).

theorem TauCeti.exists_house_schneider_value_le_exp_rpow {K : Type u_1} [Field K] [NumberField K] (α β γ : K) (c₁ : ℝ) (c : ℕ) :
∃ (c₅ : ℝ), ∀ (M L D₁ D₂ : ℕ), 1 ≤ M → L = 2 * Module.finrank ℚ K * M ^ 2 → D₁ = (2 * Module.finrank ℚ K) ^ 2 * M ^ 3 → D₂ = 2 * Module.finrank ℚ K * M → ∀ (a : Fin D₁ → Fin D₂ → ℤ), (∀ (i : Fin D₁) (j : Fin D₂), |↑(a i j)| ≤ Real.exp (c₁ * (↑D₁ * Real.log ↑L + ↑D₂ * ↑L))) → ∀ (u v : ℕ), u ≤ c * L → v ≤ c * L → NumberField.house (∑ i : Fin D₁, ∑ j : Fin D₂, ↑(a i j) * (↑u + ↑v * β) ^ ↑i * (α ^ (u * ↑j) * γ ^ (v * ↑j))) ≤ Real.exp (c₅ * ↑L ^ (3 / 2) * Real.log ↑L)

The house of Schneider's values (Evertse, Lemma 4.27(ii)). Let K be a number field of degree d, let α, β, γ ∈ K, c₁ ∈ ℝ and c ∈ ℕ. There is c₅ such that, for every integer M ≥ 1 with L = 2 d M², D₁ = (2 d)² M³ and D₂ = 2 d M, all integers a_{ij}, i < D₁, j < D₂, with |a_{ij}| ≤ exp (c₁ (D₁ log L + D₂ L)) (the bound of exists_schneider_auxiliary_algebraic) and all u, v ≤ c L, the value y_{uv} = ∑_{i, j} a_{ij} (u + v β) ^ i α ^ (u j) γ ^ (v j) has house at most exp (c₅ L ^ (3 / 2) log L). In particular ‖σ y_{uv}‖ satisfies the same bound for every embedding σ : K →+* ℂ (NumberField.norm_embedding_le_house). Evertse takes c = 1 + ⌊√(2 d)⌋.

theorem TauCeti.exists_abs_pow_schneider_exponent_le_exp_rpow (d c : ℕ) (m : ℝ) :
∃ (c₆ : ℝ), ∀ (M L D₁ D₂ : ℕ), L = 2 * d * M ^ 2 → D₁ = (2 * d) ^ 2 * M ^ 3 → D₂ = 2 * d * M → |m| ^ (D₁ + 2 * (c * L) * D₂) ≤ Real.exp (c₆ * ↑L ^ (3 / 2))

The size of the denominator of Schneider's values (Evertse, Lemma 4.27). For d, c ∈ ℕ and m ∈ ℝ there is c₆ such that, for L = 2 d M², D₁ = (2 d)² M³ and D₂ = 2 d M, the denominator m ^ (D₁ + 2 c L D₂) of the values y_{uv}, u, v ≤ c L (isIntegral_intCast_pow_mul_schneider_value), satisfies |m| ^ (D₁ + 2 c L D₂) ≤ exp (c₆ L ^ (3 / 2)).

theorem RingHom.map_pow_mul_pow_eq_exp {R : Type u_1} [Semiring R] (σ : R →+* ℂ) {l : ℂ} {α β γ : R} (hα : σ α = Complex.exp l) (hγ : σ γ = Complex.exp (σ β * l)) (u v j : ℕ) :
σ (α ^ (u * j) * γ ^ (v * j)) = Complex.exp (↑j * l * (↑u + ↑v * σ β))

If σ α = exp λ and σ γ = exp (σ β λ), then σ (α ^ (u j) γ ^ (v j)) = exp (j λ (u + v σ β)): the value of the exponential exp (j λ z) at z = u + v σ β lies in the image of σ.

theorem RingHom.map_schneider_value {R : Type u_1} [Ring R] (σ : R →+* ℂ) {l : ℂ} {α β γ : R} (hα : σ α = Complex.exp l) (hγ : σ γ = Complex.exp (σ β * l)) {D₁ D₂ : ℕ} (a : Fin D₁ → Fin D₂ → ℤ) (u v : ℕ) :
σ (∑ i : Fin D₁, ∑ j : Fin D₂, ↑(a i j) * (↑u + ↑v * β) ^ ↑i * (α ^ (u * ↑j) * γ ^ (v * ↑j))) = ∑ i : Fin D₁, ∑ j : Fin D₂, ↑(a i j) * (↑u + ↑v * σ β) ^ ↑i * Complex.exp (↑↑j * l * (↑u + ↑v * σ β))

If σ α = exp λ and σ γ = exp (σ β λ), then σ maps Schneider's value y_{uv} = ∑_{i, j} a_{ij} (u + v β) ^ i α ^ (u j) γ ^ (v j) to the value F(u + v σ β) = ∑_{i, j} a_{ij} (u + v σ β) ^ i exp (j λ (u + v σ β)) of the auxiliary function.

theorem RingHom.exists_schneider_auxiliary {K : Type u_1} [Field K] [NumberField K] (σ : K →+* ℂ) {l : ℂ} {α β γ : K} (hα : σ α = Complex.exp l) (hγ : σ γ = Complex.exp (σ β * l)) :
∃ (c : ℝ), 0 < c ∧ ∀ (L D₁ D₂ : ℕ), 3 ≤ L → 2 * Module.finrank ℚ K * L ^ 2 ≤ D₁ * D₂ → ∃ (a : Fin D₁ → Fin D₂ → ℤ), a ≠ 0 ∧ (∀ (i : Fin D₁) (j : Fin D₂), |↑(a i j)| ≤ Real.exp (c * (↑D₁ * Real.log ↑L + ↑D₂ * ↑L))) ∧ ∀ (u v : ℕ), 1 ≤ u → u ≤ L → 1 ≤ v → v ≤ L → ∑ i : Fin D₁, ∑ j : Fin D₂, ↑(a i j) * (↑u + ↑v * σ β) ^ ↑i * Complex.exp (↑↑j * l * (↑u + ↑v * σ β)) = 0

Schneider's auxiliary function (Evertse, Lemma 4.22). Let K be a number field of degree d with an embedding σ : K →+* ℂ, let λ ∈ ℂ and α, β, γ ∈ K with σ α = exp λ and σ γ = exp (σ β λ). There is c > 0 such that for all integers L ≥ 3 and D₁, D₂ with D₁ D₂ ≥ 2 d L² there are integers a_{ij}, i < D₁, j < D₂, not all zero, with |a_{ij}| ≤ exp (c (D₁ log L + D₂ L)), such that F(z) = ∑_{i, j} a_{ij} z ^ i exp (j λ z) vanishes at z = u + v σ β for all 1 ≤ u, v ≤ L.