Argument change on one analytic horizontal piece
ProvedArgHoriz.piece_eqarda-rh65k-transfernumber-theoryriemann-zeta
Let be analytic along the horizontal segment from to , and suppose its values lie in the open half-plane of chart direction . Then the logarithmic derivative is interval-integrable on the segment and
The half-plane condition holds throughout the unordered interval between and .
Preamble
import Definitions.Def_ArdaRH65K_ThetaValue
import Definitions.Def_ArdaRH65K_Turing121_ArctanTaylor
import Definitions.Def_ArdaRH65K_Turing121_ArgHoriz
import Definitions.Def_ArdaRH65K_Turing121_DiffractionCore
import Definitions.Def_ArdaRH65K_Turing121_DlvpTheta
import Definitions.Def_ArdaRH65K_Turing121_DlvpZetaRateEffective
import Definitions.Def_ArdaRH65K_Turing121_RHInBoxAnalytic
import Definitions.Def_ArdaRH65K_Turing121_StripRepr
import Definitions.Def_ArdaRH65K_Turing121_StripReprR1
import Mathlib
import Mathlib.Algebra.Ring.GeomSum
import Mathlib.Analysis.Calculus.Deriv.MeanValue
import Mathlib.Analysis.SpecialFunctions.Complex.Arg
import Mathlib.Analysis.SpecialFunctions.Gamma.Beta
import Mathlib.Analysis.SpecialFunctions.Integrals.Basic
import Mathlib.Analysis.SpecialFunctions.Trigonometric.ArctanDeriv
/- ArgHoriz.lean -- brick K6c of the ANDURIL Arb discharge: piecewise half-plane tracking of the
horizontal argument change (telperion/docs/ANDURIL_ARB_DISCHARGE_2026-09-23.md, inventory row
H2e, brick K6c).
THE BINDERS IT REDUCES. Conjuncts 6 and 7 of the `hArbT` binder of every emitted Turing band
(`TuringBand.BandStatement`, `BandGlue.BandData.EnclHyp`) are
hAHt : DiffractionCore.argChangeHoriz riemannZeta T1 2 (-1) ∈ Set.Icc L2 H2,
hAHb : DiffractionCore.argChangeHoriz riemannZeta T0 2 (-1) ∈ Set.Icc L3 H3,
the argument change of ζ along the horizontal edges `[2, -1] + i T` of the RvM rectangle
(20,758 binders at h280000, 13,235 distinct heights). Unlike K6a/K6b these are NOT generic:
they need values of ζ off the line. This file proves the reduction theorem that turns them
into a finite, kernel-checkable certificate, so that only ENCLOSURES remain.
THE CERTIFICATE (`OctCert`). Breakpoints `2 = p 0, p 1, ..., p m = -1` and one integer label
`k j` per piece `[p j, p (j+1)]`, the octant direction `k j · π/4`. It is valid when
* `StepOK` : consecutive labels differ by at most 3 (a decidable check on integers);
* `NoCross` : on piece `j`, ζ never crosses the line through 0 orthogonal to the octant
direction: `0 < α Re ζ(x + i T) + β Im ζ(x + i T)` for all `x` in the piece,
`(α, β) ∈ {(1,0),(1,1),(0,1),(-1,1),(-1,0),(-1,-1),(0,-1),(1,-1)}` the integer
octant vector (`octX`). One strict sign condition per piece: a range
enclosure of ζ over the piece that avoids the octant boundary discharges it
(`octX_pos_of_box`, bundled per certificate as `OctCert.noCross_of_boxes`).
Then (`argChangeHoriz_eq_of_octCert`, EXACT, no approximation):
argChangeHoriz ζ T 2 (-1) = octAngle (k (m-1)) ζ(-1 + i T) - octAngle (k 0) ζ(2 + i T),
octAngle k w = k π/4 + arctan ((α Im w - β Re w)/(α Re w + β Im w)).
Every interior breakpoint cancels exactly: the chart angles of the two pieces meeting there
agree (`chartAngle_eq`), because both charts contain the point and their directions differ by
less than π. So the only numerics left are TWO POINT ENCLOSURES, of ζ(2 + i T) and ζ(-1 + i T),
turned into an arctan interval by `octAngle_mem_of_box` and the rational arctan brackets
`atanLo ≤ arctan ≤ atanHi` (the emitter may use the sharper `ArctanTaylor.arctan_bracket`).
`hAH_of_octCert_boxes` is the end-to-end binder-form statement.
THE GENERAL THEOREM behind it (`argChangeHoriz_chain`): for any `f` analytic along the segment
and any real chart directions `θ j` with `|θ (j+1) - θ j| < π` and `0 < Re (e^{-i θ j} f)` on
piece `j`, the horizontal argument change is `chartAngle (θ (m-1)) f(end) - chartAngle (θ 0)
f(start)`, `chartAngle θ w = θ + arg (e^{-i θ} w)` (FTC for `log (e^{-i θ} f)` on each piece).
WHAT IS NOT HERE. The off-line ζ evaluator (memo brick B6) that produces the `NoCross` range
enclosures and the two endpoint point enclosures, and the functional-equation fold K6d.
Trust: no hypotheses beyond those stated; axioms [propext, Classical.choice, Quot.sound]
(see AxiomGuardArgChange.lean). No `sorry`.
conjecture1_proved = False. A reduction of one edge of a finite zero count to finitely many
enclosures; nothing here bears on the Riemann Hypothesis. -/
open Complex
section
namespace ArgHoriz
end ArgHoriz
open ArgHoriz
/-! ## 1. Charts: rotated principal arguments -/
/-! ## 2. One piece: FTC for `log (e^{-iθ} f)` -/
Formal statement
theorem ArgHoriz.piece_eq (f : ℂ → ℂ) (T a b θ : ℝ)
(han : ∀ x ∈ Set.uIcc a b, AnalyticAt ℂ f ((x : ℂ) + (T : ℂ) * I))
(hin : ∀ x ∈ Set.uIcc a b, 0 < (rot θ (f ((x : ℂ) + (T : ℂ) * I))).re) :
IntervalIntegrable (fun x : ℝ => logDeriv f ((x : ℂ) + (T : ℂ) * I)) MeasureTheory.volume a b ∧
(∫ x in a..b, logDeriv f ((x : ℂ) + (T : ℂ) * I)).im
= chartAngle θ (f ((b : ℂ) + (T : ℂ) * I)) - chartAngle θ (f ((a : ℂ) + (T : ℂ) * I)) := by sorry
Source