Proposition 4.2.2.2(2) as printed fails: the solution of is not supported in
ProvedMasulloAudit.periodization_kernel_support_counterexampleContext. For a real number , Masullo's manuscript studies the periodization operator on functions on , and for a compact the space of smooth functions supported in . Proposition 4.2.2.2(2) (p. 171) states:
Assume for real numbers , [with mapping onto , e.g. ]. Let . Let be a real number such that , and define and . There exists such that .
Statement (refutes the printed proposition). Let , with (so that , i.e. , satisfies the surjectivity assumption), and let be any real number with (so ; the choice is the one made in the manuscript's proof). Put . Then there is a smooth function with
- , and
- for every (the family is summable; in fact only finitely many terms are nonzero),
such that no function whatsoever (no smoothness or continuity is required) that vanishes outside satisfies for all .
For instance, for one has and , so the printed statement would force every to vanish, which is false.
Relation to the source. This is a counterexample to the exact printed statement of Proposition 4.2.2.2(2), for , every admissible , and every (for the printed support claim is correct, since then ). The error in the printed proof is the two-sided estimate (4.2.2.6): dividing the Mellin transform by improves the exponential type only as , not as . The compactly supported solution is unique, , and its support lies in , which is not contained in in the stated range. The statement says nothing about the Riemann Hypothesis; it concerns the support bookkeeping used in Proposition 4.2.2.5 (p. 173).
Formalization notes. Functions on are represented as functions on supported in ; ContDiff ℝ ∞ φ is -smoothness, Function.support φ ⊆ Set.Icc a⁻¹ a is the support condition (the interval is closed, so this is the same as a condition on the closed support), and the kernel condition is stated with HasSum over .
import Mathlib set_option autoImplicit false open scoped ContDiff
namespace MasulloAudit
/-- Counterexample to Masullo, Proposition 4.2.2.2(2) (p. 171). For `γ > 1`, `K = [1/a, a]` with
`a > 1`, `a² > γ`, and any `A₀` with `a ≤ A₀ < aγ` (so `K ⊆ [1/A₀, A₀]`), put
`A = |log A₀ - log γ|`. There is a smooth `φ` supported in `K` whose `γ^ℤ`-periodization
`∑_k φ(γ^k x)` vanishes on `(0, ∞)`, but no function `ψ` at all (no regularity required) that
vanishes outside `K_A = [e^{-A}, e^{A}]` solves `ψ(x) - ψ(γx) = φ(x)` on `(0, ∞)`. -/
theorem periodization_kernel_support_counterexample (γ a A₀ : ℝ) (hγ : 1 < γ) (ha : 1 < a)
(haγ : γ < a ^ 2) (hA₀ : a ≤ A₀) (hA₀' : A₀ < a * γ) :
∃ φ : ℝ → ℂ, ContDiff ℝ ∞ φ ∧ Function.support φ ⊆ Set.Icc a⁻¹ a ∧
(∀ x : ℝ, 0 < x → HasSum (fun k : ℤ => φ (γ ^ k * x)) 0) ∧
¬ ∃ ψ : ℝ → ℂ,
Function.support ψ ⊆ Set.Icc (Real.exp (-|Real.log A₀ - Real.log γ|))
(Real.exp |Real.log A₀ - Real.log γ|) ∧
∀ x : ℝ, 0 < x → ψ x - ψ (γ * x) = φ x := by sorry
end MasulloAudit