Validity of the 64-bit evaluator configuration
ProvedArbEcon.OrderK.valid64arda-rh65k-transfernumber-theoryriemann-zeta
Let be nonnegative integers and let . Write for the specified 64-bit fixed-point evaluator configuration, and for the configuration-validity predicate defined with the evaluator. Then
This verifies the configuration used by the certified interval evaluator.
Preamble
import Definitions.Def_ArdaRH65K_Probes_ArbEconomics_Sound
import Definitions.Def_ArdaRS_Arithmetic
import Definitions.Def_ArdaRS_Config64
import Mathlib.Analysis.Analytic.Uniqueness
import Mathlib.Analysis.Complex.Exponential
import Mathlib.Analysis.Real.Pi.Bounds
import Mathlib.Analysis.SpecialFunctions.ImproperIntegrals
import Mathlib.Analysis.SpecialFunctions.Integrals.Basic
import Mathlib.Analysis.SpecialFunctions.Log.Deriv
import Mathlib.Analysis.SpecialFunctions.Pow.Deriv
import Mathlib.Analysis.SpecialFunctions.Pow.Real
import Mathlib.Analysis.SpecialFunctions.Sqrt
import Mathlib.Analysis.SpecialFunctions.Trigonometric.Basic
import Mathlib.Analysis.SpecialFunctions.Trigonometric.Bounds
import Mathlib.MeasureTheory.Function.Floor
import Mathlib.MeasureTheory.Integral.IntegralEqImproper
import Mathlib.MeasureTheory.Integral.IntervalIntegral.IntegrationByParts
import Mathlib.NumberTheory.LSeries.RiemannZeta
import Mathlib.NumberTheory.ZetaValues
import Mathlib.Tactic.IntervalCases
namespace ZetaReflection
end ZetaReflection
namespace ZetaReflection.EMHigh
end ZetaReflection.EMHigh
/- EMZetaHighCheck.lean -- a KERNEL-ONLY assembly for the order-(2K+1) Euler-Maclaurin enclosure
of ζ(1/2 + i t) (lane emhigh, production format).
`EMZetaHighEval.zetaK_re_bounds` turns the evaluator invariants into Re/Im ζ bounds, but an
instance then discharges its rational side conditions (the correction factor `B_K(t, N)`, the
remainder constant, the final inequalities) with `norm_num` on 60-digit rationals: about 3 s of
tactic interpretation and 1.3 MB of proof term per evaluation. At the h280000 scale (445,947
sign evaluations) that is about 400 core-hours of elaboration and 0.6 TB of oleans.
This file replaces all of it by ONE Boolean checker `checkK`, run by `decide +kernel`, over plain
`Nat`/`Int` arithmetic, and ONE soundness theorem `checkK_sound`, proved once:
* `valid64` -- the P = 64 evaluator configuration `EMZetaHighCfg64.cfg64`
(lnbig = sqbig = 256) as a function of the height `t = tn / 2^tq`,
validated ONCE for all heights (no per-instance `norm_num`).
* `betaN`, `Lbeta`, `betaN_spec` -- `B_{i+2}/(i+2)! = betaN i / Lβ`, `Lβ = 2730·12!`, `i ≤ 10`.
* `gpoch`, `pochRI_eq_gpoch` -- the Gaussian-integer rising factorial `(2u)^k (1/2 + i tn/u)_k`.
* `corrData`, `corr_eq` -- the exact correction factor `emCorr K (1/2 + i t) N` as
`(BreN + i BimN) / BD` with integers computed in the kernel.
* `pnK`, `pochNormSq_eq_pnK` -- the remainder numerator `Π_{j<k} ((2j+1)² u² + 4 tn²)`.
* `remEven`, `remOdd` (+ `_sound`) -- the remainder as `EN/ED` in integers, even-saw
(`em_line_remainder_le`) or odd-saw (`em_line_remainder_odd_le`,
π replaced by `3.141592 < π < 3.141593`, about 7 percent fewer terms).
* `checkCore`, `checkCore_sound` -- the four final inequalities for ANY remainder `EN/ED`.
* `checkK`, `checkK_sound` -- the checker (`odd` flag) and its soundness: `checkK … = true`
gives `reLo/D ≤ Re ζ(1/2+it) ≤ reHi/D` and the same for Im.
conjecture1_proved = False. Finite interval arithmetic; nothing here is about RH.
-/
open Complex ZetaReflection ZetaReflection.EMHigh
open scoped Nat Real
section
namespace ArbEcon
end ArbEcon
open ArbEcon
section
namespace ArbEcon.OrderK
end ArbEcon.OrderK
open ArbEcon.OrderK
/-! ## A. The P = 64 configuration (`EMZetaHighCfg64.cfg64`), validated once for every height. -/
Formal statement
theorem ArbEcon.OrderK.valid64 (tn tq : ℕ) (t : ℝ) (ht : t = (tn : ℝ) / 2 ^ tq) : Valid (cfg64 tn tq) t 9 := by sorry
Source