# n46c — pre-registration: 3D infrared replication at L=96 (before the run)

Debt from n46b: the 3D dispersion gate fired at its registered L=24
(5.79% vs 5%) and converged in a flagged extension. This file registers
the clean replication IN ADVANCE, as the rules demand. (Declared: the
extension numbers exist in n46b's results section; this registration is
about the gate being pre-committed, not about suspense.)

## Gates (single-q Fourier sums, method of n46b)
- IR2-3D at L=96: |S(q1)/S(q2) - 1| < 5%.
- Stiffness convergence: |S_min(L=96)/S_min(L=48) - 1| < 2%.
- Report: S_3D (the toy's 3D 1/G_N) at L=96.
DEATH: either gate fails -> the 3D quadratic-dispersion claim reverts to
unproven and the n46b "sign owned in 3D" verdict loses its form leg.

## RESULTS (same day; n46c_replication.py)
Both gates PASS: ratio-1 = 0.36% (gate 5%), convergence 0.36% (gate 2%).
S_3D = 493.44. The n46b debt is closed; 3D infrared Coulomb stands on a
pre-registered pass.