Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Th. 11.21 Step 4: an L×L minor of the interpolation matrix is at most exp(−(Θₙ(L) − L·N₀) + C·L·Ξ), n = dim V, by Lemma 9.2 and Schwarz

Proved
DiazModulus.interp_det_upper

by carlok · Oct 29, 2026 · Mathlib 0df444a (Lean v4.33.1)

diaz-modulus-leannumber-theorytranscendence

Points are p=(p′,p′′)∈Cι0×Cι1p = (p', p'') \in \mathbb{C}^{\iota_0} \times \mathbb{C}^{\iota_1}p=(p′,p′′)∈Cι0​×Cι1​ for finite index sets ι0,ι1\iota_0, \iota_1ι0​,ι1​. For w=(w′,w′′)w = (w', w'')w=(w′,w′′) and a weight vector ttt on ι1\iota_1ι1​, Opwt\mathrm{Op}^t_wOpwt​ is the operator ∑h∈ι0wh′ ∂/∂Xh+∑i∈ι1wi′′ti\sum_{h \in \iota_0} w'_h\,\partial/\partial X_h + \sum_{i \in \iota_1} w''_i t_i∑h∈ι0​​wh′​∂/∂Xh​+∑i∈ι1​​wi′′​ti​ on polynomials in the XhX_hXh​ (h∈ι0h \in \iota_0h∈ι0​), and for vectors wkw_kwk​ (k∈κ0k \in \kappa_0k∈κ0​) and k1,…,kr∈κ0k_1, \dots, k_r \in \kappa_0k1​,…,kr​∈κ0​, Opw,kt=Opwk1t⋯Opwkrt\mathrm{Op}^t_{w,k} = \mathrm{Op}^t_{w_{k_1}} \cdots \mathrm{Op}^t_{w_{k_r}}Opw,kt​=Opwk1​​t​⋯Opwkr​​t​ (the factors commute). For T0,T1≥1T_0, T_1 \geq 1T0​,T1​≥1 the columns are the λ=(τ,t)\lambda = (\tau, t)λ=(τ,t) with τ∈{0,…,T0}ι0\tau \in \{0, \dots, T_0\}^{\iota_0}τ∈{0,…,T0​}ι0​ and t∈{−T1,…,T1}ι1t \in \{-T_1, \dots, T_1\}^{\iota_1}t∈{−T1​,…,T1​}ι1​; there are L=(T0+1)∣ι0∣(2T1+1)∣ι1∣L = (T_0 + 1)^{|\iota_0|}(2T_1 + 1)^{|\iota_1|}L=(T0​+1)∣ι0​∣(2T1​+1)∣ι1​∣ of them. Given uj∈Cι0u_j \in \mathbb{C}^{\iota_0}uj​∈Cι0​ and aji∈Ca_{ji} \in \mathbb{C}aji​∈C (j∈κ1j \in \kappa_1j∈κ1​, i∈ι1i \in \iota_1i∈ι1​), the entry at a row datum (r,k,s)(r, k, s)(r,k,s) (r∈Nr \in \mathbb{N}r∈N, k1,…,kr∈κ0k_1, \dots, k_r \in \kappa_0k1​,…,kr​∈κ0​, s∈Zκ1s \in \mathbb{Z}^{\kappa_1}s∈Zκ1​) and a column (τ,t)(\tau, t)(τ,t) is E(r,k,s),(τ,t)=(Opw,ktXτ)(∑jsjuj)⋅∏i(∏jajisj)tiE_{(r,k,s),(\tau,t)} = (\mathrm{Op}^t_{w,k} X^\tau)(\sum_j s_j u_j) \cdot \prod_i (\prod_j a_{ji}^{s_j})^{t_i}E(r,k,s),(τ,t)​=(Opw,kt​Xτ)(∑j​sj​uj​)⋅∏i​(∏j​ajisj​​)ti​. In the entries, uj=ηj′u_j = \eta'_juj​=ηj′​ and aji=eηj,i′′a_{ji} = e^{\eta''_{j,i}}aji​=eηj,i′′​, for points ηj=(ηj′,ηj′′)\eta_j = (\eta'_j, \eta''_j)ηj​=(ηj′​,ηj′′​). Let V⊆Cι0×Cι1V \subseteq \mathbb{C}^{\iota_0} \times \mathbb{C}^{\iota_1}V⊆Cι0​×Cι1​ be a subspace of dimension nnn containing every wkw_kwk​ (k∈κ0k \in \kappa_0k∈κ0​) and every ηj\eta_jηj​ (j∈κ1j \in \kappa_1j∈κ1​). Then there is C>0C > 0C>0 such that, for all T0,T1≥1T_0, T_1 \geq 1T0​,T1​≥1 and N0,N1∈NN_0, N_1 \in \mathbb{N}N0​,N1​∈N, and for every choice of LLL row data (rλ,kλ,sλ)(r_\lambda, k_\lambda, s_\lambda)(rλ​,kλ​,sλ​), one for each column λ\lambdaλ, with rλ≤N0r_\lambda \leq N_0rλ​≤N0​ and ∣sλ,j∣≤N1|s_{\lambda,j}| \leq N_1∣sλ,j​∣≤N1​, the determinant Δ\DeltaΔ of the L×LL \times LL×L matrix (E(rλ,kλ,sλ),λ′)λ,λ′(E_{(r_\lambda, k_\lambda, s_\lambda), \lambda'})_{\lambda, \lambda'}(E(rλ​,kλ​,sλ​),λ′​)λ,λ′​ satisfies ∣Δ∣≤exp⁡(−(Θn(L)−LN0)+CLΞ)|\Delta| \leq \exp(-(\Theta_n(L) - L N_0) + C L \Xi)∣Δ∣≤exp(−(Θn​(L)−LN0​)+CLΞ), with Ξ=N0log⁡(T0+T1+2)+T0log⁡(N1+2)+T1(N1+1)\Xi = N_0 \log(T_0 + T_1 + 2) + T_0 \log(N_1 + 2) + T_1(N_1 + 1)Ξ=N0​log(T0​+T1​+2)+T0​log(N1​+2)+T1​(N1​+1) and Θn(L)\Theta_n(L)Θn​(L) is the least value of ∑l∣κ′(l)∣\sum_l |\kappa'(l)|∑l​∣κ′(l)∣ over the injective maps κ′:{1,…,L}→Nn\kappa' : \{1, \dots, L\} \to \mathbb{N}^nκ′:{1,…,L}→Nn, with ∣⋅∣|\cdot|∣⋅∣ the total degree of a multi-index (000 if there is no such map).

Proof. Fix a basis of VVV and let ψ:Cn→V\psi : \mathbb{C}^n \to Vψ:Cn→V be the coordinate map. Each column gives the entire function fλ∘ψf_\lambda \circ \psifλ​∘ψ, fλ(z)=z′τe⟨t,z′′⟩f_\lambda(z) = z'^\tau e^{\langle t, z'' \rangle}fλ​(z)=z′τe⟨t,z′′⟩, and each row the point ψ−1(∑jsjηj)\psi^{-1}(\sum_j s_j \eta_j)ψ−1(∑j​sj​ηj​) and the directions ψ−1(wk)\psi^{-1}(w_k)ψ−1(wk​). By DiazModulus.iteratedFDeriv_expMono_eq_op the entries of F(z)=det⁡(Dr(f∘ψ)(zζ)(v))F(z) = \det(D^{r}(f \circ \psi)(z\zeta)(v))F(z)=det(Dr(f∘ψ)(zζ)(v)) are the polynomial parts at zp′zp'zp′ times ez⟨t,p′′⟩e^{z\langle t, p'' \rangle}ez⟨t,p′′⟩, so F(1)=ΔF(1) = \DeltaF(1)=Δ up to transposition. DiazModulus.interpolation_det_order (DALAG Lemma 9.2) gives ord0F≥Θn(L)−LN0\mathrm{ord}_0 F \geq \Theta_n(L) - L N_0ord0​F≥Θn​(L)−LN0​, and DiazModulus.schwarz_order_at_zero (Schwarz) with R=eR = eR=e gives ∣Δ∣≤e−ordsup⁡∣z∣=e∣F∣|\Delta| \leq e^{-\mathrm{ord}} \sup_{|z| = e} |F|∣Δ∣≤e−ordsup∣z∣=e​∣F∣. On ∣z∣=e|z| = e∣z∣=e each entry is at most eβe^\betaeβ, by DiazModulus.op_monomial_eval_norm_le for the polynomial part and ∣eu∣≤e∣u∣|e^u| \leq e^{|u|}∣eu∣≤e∣u∣, so ∣F∣≤L! eLβ|F| \leq L!\, e^{L\beta}∣F∣≤L!eLβ; finally log⁡L+β≤KΞ\log L + \beta \leq K\XilogL+β≤KΞ is elementary.

Novelty. Classical; routine formalisation. Step 4 of the proof of DALAG Th. 11.21: the analytic upper bound.

Preamble
import Definitions.Def_DiazModulus
Formal statement
open Complex ComplexConjugate

namespace DiazModulus

theorem interp_det_upper {ι₀ ι₁ κ₀ κ₁ : Type} [Fintype ι₀] [Fintype ι₁] [Fintype κ₀] [Fintype κ₁]
    [DecidableEq ι₀] [DecidableEq ι₁]
    (V : Submodule ℂ ((ι₀ → ℂ) × (ι₁ → ℂ))) (w : κ₀ → (ι₀ → ℂ) × (ι₁ → ℂ)) (hwV : ∀ k, w k ∈ V)
    (η : κ₁ → (ι₀ → ℂ) × (ι₁ → ℂ)) (hηV : ∀ j, η j ∈ V) :
    ∃ C : ℝ, 0 < C ∧ ∀ T₀ T₁ N₀ N₁ : ℕ, 1 ≤ T₀ → 1 ≤ T₁ →
    ∀ (ro : (ι₀ → Fin (T₀ + 1)) × (ι₁ → Fin (2 * T₁ + 1)) → ℕ) (ka : (l : (ι₀ → Fin (T₀ + 1)) × (ι₁ → Fin (2 * T₁ + 1))) → Fin (ro l) → κ₀)
      (sv : (ι₀ → Fin (T₀ + 1)) × (ι₁ → Fin (2 * T₁ + 1)) → κ₁ → ℤ), (∀ l, ro l ≤ N₀) → (∀ l j, |sv l j| ≤ (N₁ : ℤ)) →
      ‖(Matrix.of fun l l' : (ι₀ → Fin (T₀ + 1)) × (ι₁ → Fin (2 * T₁ + 1)) =>
        MvPolynomial.eval (∑ j, (((sv l) j : ℤ) : ℂ) • (fun j => (η j).1) j)
          ((List.ofFn fun k : Fin (ro l) =>
            ((∑ h, (w ((ka l) k)).1 h • (MvPolynomial.pderiv h).toLinearMap
              + (∑ i, (w ((ka l) k)).2 i * (((l'.2 i : ℕ) : ℂ) - T₁)) • LinearMap.id :
                Module.End ℂ (MvPolynomial ι₀ ℂ)))).prod
            (∏ h, MvPolynomial.X h ^ (l'.1 h : ℕ)))
        * ∏ i, (∏ j, (fun j i => Complex.exp ((η j).2 i)) j i ^ (sv l) j) ^ (((l'.2 i : ℕ) : ℤ) - T₁)).det‖
        ≤ Real.exp (-(((sInf {S | ∃ κ' : Fin ((T₀ + 1) ^ Fintype.card ι₀ * (2 * T₁ + 1) ^ Fintype.card ι₁) → (Fin (Module.finrank ℂ ↥V) → ℕ), Function.Injective κ' ∧ S = ∑ l, ∑ i, κ' l i} : ℕ) : ℝ)
              - (((T₀ + 1) ^ Fintype.card ι₀ * (2 * T₁ + 1) ^ Fintype.card ι₁ : ℕ) : ℝ) * N₀)
            + C * (((T₀ + 1) ^ Fintype.card ι₀ * (2 * T₁ + 1) ^ Fintype.card ι₁ : ℕ) : ℝ) * ((N₀ : ℝ) * Real.log ((T₀ : ℝ) + T₁ + 2) + (T₀ : ℝ) * Real.log ((N₁ : ℝ) + 2)
          + (T₁ : ℝ) * ((N₁ : ℝ) + 1))) := by
  sorry

end DiazModulus
Source
Step 4 (upper bound) of the proof of M. Waldschmidt, Diophantine Approximation on Linear Algebraic Groups, Grundlehren der mathematischen Wissenschaften 326, Springer, 2000, Th. 11.21 (the linear subgroup theorem; proof in §11.7.1); compare Lemma 9.2; Schwarz's lemma is Lemma 2.4. Uses DiazModulus.interpolation_det_order, DiazModulus.schwarz_order_at_zero, DiazModulus.iteratedFDeriv_expMono_eq_op and DiazModulus.op_monomial_eval_norm_le. C1.4 of the Diaz modulus mission. Formal proof: Diaz modulus mission, 29 October 2026 (C. Perassi).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me