Explicit truncation bounds for every Dirichlet-series derivative
Proveddirichlet_series_all_derivative_truncation_error_boundanalytic-number-theorydirichlet-seriesformalizationriemann-hypothesis
Let satisfy for all , where and . Let . Write , and choose . For every integer and ,
This supplies a computable approximation error at every derivative order under a signed partial-sum bound. No absolute convergence of the original series and no zero-index normalization are assumed. The positive radius is strictly smaller than the distance to the convergence boundary.
Preamble
import Mathlib.Analysis.Complex.Liouville import Mathlib.Analysis.Calculus.IteratedDeriv.Lemmas import Mathlib.Analysis.Complex.LocallyUniformLimit import Mathlib.Analysis.Calculus.IteratedDeriv.Defs import Mathlib.Tactic.Linarith import Mathlib.Tactic.Ring import Mathlib.NumberTheory.LSeries.Dirichlet import Mathlib.NumberTheory.LSeries.SumCoeff import Mathlib.Analysis.MellinTransform import Mathlib.Tactic.NormNum open MeasureTheory open scoped Topology
Formal statement
theorem dirichlet_series_all_derivative_truncation_error_bound
(f : ℕ → ℂ) {r C : ℝ} (hr : 0 ≤ r) (hC : 0 ≤ C)
(hbound : ∀ N : ℕ, 1 ≤ N →
‖∑ n ∈ Finset.Icc 1 N, f n‖ ≤ C * (N : ℝ) ^ r)
{s : ℂ} {δ : ℝ} (hδ : 0 < δ) (hδgap : δ < s.re - r)
{N : ℕ} (hN : 1 ≤ N) (j : ℕ) :
‖iteratedDeriv j (fun z : ℂ =>
z * mellin (fun t : ℝ => ∑ n ∈ Finset.Icc 1 ⌊t⌋₊, f n) (-z)) s -
(∑ n ∈ Finset.Icc 1 N, f n * (-Complex.log (n : ℂ)) ^ j / (n : ℂ) ^ s)‖ ≤
(j.factorial : ℝ) / δ ^ j * C *
(1 + (‖s‖ + δ) / (s.re - δ - r)) * (N : ℝ) ^ (r - s.re + δ) := by sorrySource
Derived Cauchy-estimate corollary. Pinned Mathlib https://github.com/leanprover-community/mathlib4/blob/0df444a360eaa60ab8c11dca51a86af692955474/Mathlib/Analysis/Complex/Liouville.lean#L44, Complex.norm_iteratedDeriv_le_of_forall_mem_sphere_norm_le, combined with the finite Abel identity https://github.com/leanprover-community/mathlib4/blob/0df444a360eaa60ab8c11dca51a86af692955474/Mathlib/NumberTheory/AbelSummation.lean#L229, sum_mul_eq_sub_integral_mul₀'. The displayed constant is derived here, not quoted as a verbatim source theorem.