teal-sea / zeta-labstate of record · compiled 14 Aug 2026 · revision 9ebdea0 · source

Library · hunts/jensen_clock/MISSION.md

Hunt #6 — the Jensen clock (`jensen_clock/`)

1,362 words · 147 lines · source

Question. zeta/li.py's real-rootedness lane asks about Jensen polynomials J^{d,n}; hunt #4 (flow_repair/) measured when the de Bruijn–Newman flow lands the Davenport–Heilbronn off-line pairs. This hunt connects them: at what finite Jensen degree does hyperbolicity actually witness the counterexample's off-line pair, and is the finite degree itself a heat clock?

The observation driving the design (found in a scratch prototype before this mission was written — recorded honestly as such): with E(x) = Σ γ(n) xⁿ/n! and γ(n) = n!/(2n)! · M₂ₙ the moment sequence of the DH weight Φ_DH, the degree-d Jensen polynomial satisfies exactly

J^{d,0}(x/d) = Σ_j γ(j)/j! · Π_{i<j}(1 − i/d) · x^j,

and the damping Π(1 − i/d) ≈ e^{−j²/2d}, read at the cosh-series saddle j ≈ zu/2, acts like the flow multiplier e^{t u²} with an effective time

t_eff(d) = |x₀| / (8 d), x₀ = (β − 1/2 + iγ)² the pair's image.

A prototype trajectory ladder saw the DH height-85.7 pair's image in J^{d,0} drift to the real axis and land between d = 10⁴ and 3·10⁴, with the drift rate agreeing with hunt #4's measured pair dynamics through the dictionary above at the percent level. A naive grid scan (d ≤ 32, all shifts n ≤ 250) sees nothing, which the dictionary explains: those degrees carry t_eff far beyond the pair's landing time, so the degree itself has smoothed the violation away before the polynomial is even inspected.

Pre-registered predictions

Targets below use hunt #4's measured values, read from hunts/flow_repair/results.json: pair 1 (β ≈ 0.8085, γ ≈ 85.699, t\* = 0.044126344551623946, isolated-pair naive value y₀²/2 = 0.047591…), pair 2 (β ≈ 0.6508, γ ≈ 114.163, t\* ≈ 0.0111296).

Phase 2 — the shift direction (registered 2026-08-11, after phase 1 closed)

Phase 1 fixed the shift n = 0 and measured the degree axis. The shifted Jensen polynomial is exactly the damped n-th derivative, J^{d,n}(x/d) = Σ_j γ(n+j)/j! · Π_{i<j}(1−i/d) · x^j = damped E⁽ⁿ⁾(x), so the shift axis asks: what does differentiation in x do to the off-line pair's image, on the same clock? Differentiation of real entire functions is believed to act on zeros as a smoothing flow (Gauss–Lucas pulls complex pairs toward the real hull; there is a modern PDE literature on repeated differentiation as an erosion/heat-like flow on the zero distribution — a literature pass is part of this phase). Pre-registered questions and predictions, written before any phase-2 measurement:

Phase 2 may also touch only hunts/jensen_clock/.

Phase 3 — the falsifier and the trichotomy (registered 2026-08-11, after

phase 2 closed)

Scope

May touch: hunts/jensen_clock/ only. Reads (never writes) hunts/flow_repair/results.json, zeta.epstein, zeta.heatflow, zeta.li. No changes to zeta/, ontology/, harness/, no cache invalidation, no ledger entry unless something survives the checklist in hunts/README.md.

Everything here is the accurate regime — mpmath floats with measured cross-route defects; the strongest words used are measured and observed. The truncated-series detector reports its own validity margin (tail bound vs minimum modulus on the contour) rather than assuming it.

Prior-art hooks, recorded so nothing is overclaimed: coefficient multiplier sequences of Gaussian type are classical (Pólya–Schur; de Bruijn's e^{−λD²} operators; Turán), and Griffin–Ono–Rolen–Zagier relate large-shift Jensen polynomials to Hermite polynomials through exactly this kind of heat limit. The degree-damping-as-heat reading may well be implicit in that literature; this hunt claims measurements on the counterexample, not novelty.