Vaughan's Lemma 2.2, weight form
ProvedCircleMethod.aux_sum_min_leThe central counting estimate of the circle method, Vaughan's Lemma 2.2, in an abstracted weight form.
Classically the statement is that for and ,
The mechanism is equidistribution: as runs over a block of consecutive integers and , the residues run over a complete residue system, so the values are spread out at spacing roughly . Within one block at most a bounded number of can have in any given dyadic range, and summing over produces the logarithm. There are blocks.
Rather than the , the statement here quantifies over an arbitrary weight subject to three hypotheses: , , and . Any family of exponential sums satisfies these, whatever their lengths, so the abstraction is what makes the lemma applicable after Weyl differencing — where the inner sums have lengths depending on the shift and a fixed-length statement cannot be used. Taking recovers the classical form.
Formalization Note. The multiplicative hypothesis is used in place of deliberately. In Lean , so the divided form would be silently weaker than intended exactly when — the case that matters most. The multiplicative form is equivalent away from that degeneracy and correct at it.
import Definitions.Def_CircleMethod_char import Mathlib.Analysis.SpecialFunctions.Pow.Real import Mathlib.Analysis.SpecialFunctions.Log.Basic import Mathlib.NumberTheory.Divisors open Finset
namespace CircleMethod
theorem aux_sum_min_le {ε : ℝ} (hε : 0 < ε) :
∃ C : ℝ, 0 < C ∧ ∀ (H N q : ℕ) (a : ℤ) (α : ℝ) (g : ℕ → ℝ), 0 < H → 0 < N → 0 < q →
Nat.Coprime a.natAbs q → |α - (a : ℝ) / (q : ℝ)| ≤ 1 / (q : ℝ) ^ 2 →
(∀ k : ℕ, 0 ≤ g k) → (∀ k : ℕ, g k ≤ (N : ℝ)) →
(∀ k : ℕ, 2 * nrm ((k : ℝ) * α) * g k ≤ 1) →
(∑ k ∈ Finset.Icc 1 H, g k)
≤ C * ((H : ℝ) / (q : ℝ) + 1) * ((N : ℝ) + (q : ℝ)) * (1 + Real.log (1 + (q : ℝ))) := by sorry
end CircleMethod