Zero-free slabs confine ball zeros to the band interior
ProvedEdgeClearGlue.hins_of_slabCleararda-rh65k-transfernumber-theoryriemann-zeta
Let and let . Suppose , , and . If the slabs and are zero-free in the critical strip, then every zeta zero in the ball's finite zero support satisfies
The conclusion places the support strictly inside the counting rectangle.
Preamble
import Definitions.Def_ArdaRH65K_Turing121_DiffractionCore
import Definitions.Def_ArdaRH65K_Turing121_DlvpTheta
import Definitions.Def_ArdaRH65K_Turing121_DlvpZetaRateEffective
import Definitions.Def_ArdaRH65K_Turing121_EdgeClearGlue
import Definitions.Def_ArdaRH65K_Turing121_RHInBoxAnalytic
import Definitions.Def_ArdaRH65K_Turing121_StripRepr
import Definitions.Def_ArdaRH65K_Turing121_StripReprR1
import Definitions.Def_ArdaRH65K_Turing121_TuringBand
import Definitions.Def_ArdaRH65K_XiLineZeros
import Mathlib
import Theorems.Thm_EdgeClearGlue_mem_ball_and_zero_of_mem_zeroFinset
import Theorems.Thm_ZetaZeroConfinement_zeta_zero_iff_completed_zero_of_im_ne
/- EdgeClearGlue.lean -- brick K1 of the ANDURIL Arb discharge
(telperion/docs/ANDURIL_ARB_DISCHARGE_2026-09-23.md, section 1.3: `hArbT_nonvanishing_of_edge_clear`).
Every emitted Turing band `RHInBoxT_*.rh_in_box_*` states the canonical
`TuringBand.BandStatement` and takes an Arb binder `hArbT` with nine conjuncts. The first four
are non-vanishing and confinement facts about the RvM rectangle `[-1, 2] x [T0, T1]` and the
Blaschke ball `ball c R` that covers it:
hnzb : ∀ x ∈ [-1, 2], ζ(x + T0 i) ≠ 0 (H2a, bottom edge)
hnzt : ∀ x ∈ [-1, 2], ζ(x + T1 i) ≠ 0 (H2a, top edge)
hnzl : ∀ y ∈ [T0, T1], ζ(-1 + y i) ≠ 0 (H2b, left edge)
hins : every zero of ζ in `ball c R` lies in (-1, 2) x (T0, T1) (H2c, confinement)
This file derives all four from
* `ZetaZeroConfinement.zeta_zero_re_mem_strip`: an off-real zero of ζ has 0 < Re < 1;
* the ball geometry: three rational/sqrt inequalities about c, R, T0, T1 (decidable per band);
* two EDGE-CLEARANCE facts, the memo's E(T0, δ0) and E(T1, δ1), as zero-free SLABS on the
OUTER side of each edge: `SlabClear (T0 - δ0) T0` and `SlabClear T1 (T1 + δ1)`, where
`SlabClear a b` says ζ has no zero with 0 < Re < 1 and a ≤ Im ≤ b.
H2b needs no input at all: `hnzl_discharged` proves the conjunct for EVERY `T0`, `T1`, with no
hypotheses (off the real axis by the strip theorem; at y = 0 by ζ(-1) = -1/12). That removes
one binder from each of the 10,379 bands of the h280000 ladder.
What remains numerical after this file: two one-sided slab facts per band, about the open strip
0 < Re < 1 only. WHY ONE-SIDED, AND HOW TALL (measured on the h280000 ladder with the float
zero list at grid 0.01, the tight cases refined with mpmath; untrusted numerics, design only):
* ONE-SIDED: at about 2,831 band edges a zero lies on the INNER side of the edge closer than
the ball cap (as close as the grid resolves), so even a two-sided window of half-height
equal to the cap would be false there. Inner-side zeros are counted by the band; only
the outer side must be clear, and that is all `hins` needs.
* TALL ENOUGH, NOT TOO TALL: the slab must reach past the ball cap (δ ≥ cap; caps are 0.0574
to 0.0922) and must stop before the nearest zero beyond the edge. The margin between the
two is as small as 4.3e-5 (top edge 154974: cap 0.092160, zero at 154974.092203). So δ
must be the band's own cap rounded up finely (emit_band_glue.py uses 1e-6); a round
δ = 1/10 would make the slab hypothesis false at about 385 of the 10,379 bands.
Trust: no hypotheses beyond those named in each statement; axioms
[propext, Classical.choice, Quot.sound] (see AxiomGuardBandGlue.lean). No `sorry`.
conjecture1_proved = False. This is glue for a finite verification up to a fixed height; it
says nothing about the Riemann Hypothesis. -/
open Complex
section
namespace EdgeClearGlue
end EdgeClearGlue
open EdgeClearGlue
/-! ## Zeta off the critical strip -/
/-! ## Edge clearance: zero-free slabs -/
/-! ## Confinement of the Blaschke-ball zeros -/
Formal statement
theorem EdgeClearGlue.hins_of_slabClear {T0 T1 δ0 δ1 : ℝ} {c : ℂ} {R : ℝ} (hs1 : (1 : ℂ) ∉ Metric.ball c R)
(hpos : 0 ≤ c.im - R) (hlo : T0 - δ0 ≤ c.im - R) (hhi : c.im + R ≤ T1 + δ1)
(hE0 : SlabClear (T0 - δ0) T0) (hE1 : SlabClear T1 (T1 + δ1)) :
∀ ρ ∈ RHInBoxAnalytic.zeroFinset c R hs1,
(-1 : ℝ) < ρ.re ∧ ρ.re < 2 ∧ T0 < ρ.im ∧ ρ.im < T1 := by sorry
Source