Roy's strong six exponentials theorem, from the elementary multiplicity estimate: for Q̄-free x ∈ ℂ², y ∈ ℂ³, some xiyj lies outside ℒ̃
Proved
DiazModulus.strong_six_exponentials
by carlok · Oct 29, 2026 · Mathlib 0df444a (Lean v4.33.1)
diaz-modulus-leannumber-theorytranscendence
Here Q is the field of algebraic numbers, L is the set of logarithms of algebraic numbers (the z∈C with ez algebraic), and L is the Q-vector space spanned by 1 and L.
Points are p=(p′,p′′)∈Cι0×Cι1 for finite index sets ι0,ι1. For w=(w′,w′′) and a weight vector t on ι1, Opwt is the operator ∑h∈ι0wh′∂/∂Xh+∑i∈ι1wi′′ti on polynomials in the Xh (h∈ι0), and for vectors wk (k∈κ0) and k1,…,kr∈κ0, Opw,kt=Opwk1t⋯Opwkrt (the factors commute). For T0,T1≥1 the columns are the λ=(τ,t) with τ∈{0,…,T0}ι0 and t∈{−T1,…,T1}ι1; there are L=(T0+1)∣ι0∣(2T1+1)∣ι1∣ of them. Given uj∈Cι0 and aji∈C (j∈κ1, i∈ι1), the entry at a row datum (r,k,s) (r∈N, k1,…,kr∈κ0, s∈Zκ1) and a column (τ,t) is E(r,k,s),(τ,t)=(Opw,ktXτ)(∑jsjuj)⋅∏i(∏jajisj)ti.
The hypothesis hME : MultiplicityEstimateElem is the published definition MultiplicityEstimateElem (DiazModulus.MultiplicityEstimateElem, module Definitions.Def_MultiplicityEstimateElem), a proposition that is assumed here, not proved. It is an elementary form of Philippon's multiplicity estimate: the case G−={e}, G+=G of DALAG Th. 8.1 (§8.1) for G=GaA0×GmA1, written in coordinates. It says: let A0,A1,B0,B1 be finite, d=∣A0∣+∣A1∣, wb∈CA0×CA1 (b∈B0), uj∈CA0 and non-zero aji∈C (j∈B1, i∈A1), D0,D1≥1, M0,M1∈N, and c=0 a vector on the box of the (τ,t) with τh≤D0 and ∣ti∣≤D1. If ∑λcλE(r,k,s),λ=0 for every r≤dM0, every k1,…,kr∈B0 and every s with ∣sj∣≤dM1 (the entries E are those defined above, with A0,A1,B0,B1 for ι0,ι1,κ0,κ1 and D0,D1 for T0,T1), then there are a subspace V⊆CA0, integer vectors φ1,…,φq∈ZA1 with common complex kernel L1, and ℓ with ℓ+dim(W∩(V×L1))=dimW, W the span of the wb, such that dimV+dimL1<d and every set R of vectors s with ∣sj∣≤M1, pairwise incongruent modulo (V,φ), has (ℓM0+ℓ)∣R∣D0dimVD1dimL1≤KdD0∣A0∣D1∣A1∣, with Kd=2d(d+1)/2d!2∣A1∣∣A0∣∣A0∣; two vectors s,s′∈ZB1 are congruent modulo (V,φ) when ∑j(sj−sj′)uj∈V and ∏i(∏jajisj−sj′)φk,i=1 for every k.
Assume hME. Let x∈C2 and y∈C3 each be linearly independent over Q. Then xiyj∈/L for some i,j.
This is Roy's strong six exponentials theorem, the case d=2, ℓ=3 of DALAG Cor. 11.16 (§11.6.1), conditional on hME. hME is carried as an explicit hypothesis and is not proved here. It is intended as the case G−={e} of DALAG Th. 8.1 (Philippon's multiplicity estimate); that derivation is not proved formally here. A separate check of the statement against Th. 8.1, including exact computation on small cases, found no counterexample.
Proof. Suppose xiyj∈L for all i,j. DiazModulus.exists_sse_lifts gives m≥1, a Q-free algebraic ω and lifts ηa,j∈Q2×Lm×2 with Υ(ηa,j)=ωayjx, where Υ(u,v)i=ui+∑kωkvk,i. Apply DiazModulus.linear_subgroup_transcendence (DALAG Th. 11.21) in G=Ga2×Gm2m with V=Υ−1(Cx), of dimension 2m+1 by DiazModulus.sse_setup (so d−n=1), the algebraic spanning family (−ωkei,ek,i) of kerΥ (DiazModulus.sse_setup) and the lifts η. It returns (V′,φ,L1,ℓ) with ℓ<d′ and 3m−ρ+d1′≤2m(d′−ℓ). By DiazModulus.sse_tangent_count, d′−ℓ=δ=2−dimT for T=Υ(V′×L1), so δ≥1. If T⊆Cx, DiazModulus.sse_relRank_rational gives L1=0 and ρ=0, so 5m≤2mδ≤4m. Otherwise dimT=1 and T∩Cx=0, and DiazModulus.sse_relRank_generic gives ρ≤2m−dimL1, so 3m≤2m. Both contradict m≥1.
Novelty. Classical: Roy's strong six exponentials theorem (Roy 1992, Cor. 2; DALAG Cor. 11.16, §11.6.1). The proof follows DALAG: Roy's map Υ of §11.6.2 and the linear subgroup theorem, Th. 11.21; the case analysis of the obstruction (DiazModulus.sse_relRank_generic, DiazModulus.sse_relRank_rational) does not use that it is defined over Q and does not use Baker's theorem; that variation of the route is routine. The multiplicity estimate (DALAG Th. 8.1, after Philippon) enters only as the hypothesis hME, which is not proved here.
open Complex ComplexConjugate
namespace DiazModulus
theorem strong_six_exponentials (hME : MultiplicityEstimateElem) :
∀ x : Fin 2 → ℂ, ∀ y : Fin 3 → ℂ, LinearIndependent (↥Qbar) x → LinearIndependent (↥Qbar) y →
¬ ∀ i j, x i * y j ∈ LogAlgTilde := by
sorry
end DiazModulus
Source
Roy's strong six exponentials theorem: D. Roy, Matrices whose coefficients are linear forms in logarithms, J. Number Theory 41 (1992), 22–47, Cor. 2; M. Waldschmidt, Diophantine Approximation on Linear Algebraic Groups, Grundlehren der mathematischen Wissenschaften 326, Springer, 2000, Cor. 11.16 (§11.6.1; this node is its case d = 2, ℓ = 3), through Th. 11.21 (§11.7.1). The multiplicity estimate (M. Waldschmidt, Diophantine Approximation on Linear Algebraic Groups, Grundlehren der mathematischen Wissenschaften 326, Springer, 2000, Th. 8.1, §8.1; P. Philippon, Lemmes de zéros dans les groupes algébriques commutatifs, Bull. Soc. Math. France 114 (1986), 355–383) is carried as the explicit hypothesis hME, the definition MultiplicityEstimateElem, and is not proved in this round. Uses DiazModulus.exists_sse_lifts, DiazModulus.sse_setup, DiazModulus.sse_tangent_count, DiazModulus.sse_relRank_generic, DiazModulus.sse_relRank_rational and DiazModulus.linear_subgroup_transcendence. C1.4 of the Diaz modulus mission. Formal proof: Diaz modulus mission, 29 October 2026 (C. Perassi).