# n211 — Construction 1 delivered at Gaussian level: dark states derived, and marginality turns out to be zero-at-rest's protector

**Date:** 2026-07-11 · **Audit gates 7.3/7.4 (the trunk's sharpest).** **Script:** `n211_joint_lindblad.py`, 5/5, fail-closed. **Reproduce:** `python3 n211_joint_lindblad.py`.

## What was built (no longer defined into existence)

An explicit **joint** matter–substrate model: matter = a two-branch superposition (branch label with branch-dependent consumption c(b)); substrate = bosonic modes on the edges of a ring; monitor = one annihilation-type violation operator per site, L_i = √Γ[(Aa)_i + c_i(b)], acting on the joint Hilbert space; substrate Hamiltonian H = Δ·Σa†a with Δ the k-sector's effective gap.

## The three derived results

**1. Dark states are DERIVED (gate 7.3, Gaussian level).** The steady dressing of each branch is an explicit coherent displacement; at Δ = 0 both spatially separated dressed branches are *exact common zero modes of every L_i*, and their superposition is exactly dark — verified two ways: closed-form residual algebra, and exact Lindblad evolution of the coherence block in a truncated Fock space (decay rate 0 to machine).

**2. Zero-at-rest and the sandpile are one fact.** Off the marginal point the steady dressing cannot satisfy the constraint (H rotates it away), the residuals v_b ≠ 0, and the rest decoherence rate is exactly Γ_rest = (Γ/2)|v₁−v₂|², which scales as **(Δ_eff/Γ)²** — measured exponent 1.999, and the Fock-space rate matches the formula to five digits (0.01231 vs 0.01231). So: **zero-at-rest is exact precisely at Δ_eff = 0 — the same marginal point the blind-channel thermostat tunes to.** The theory's two independent pillars (dark states; sandpile marginality) are one mechanism: *the audit protects coherence exactly where it keeps itself critical.* Falsifiable structure gained: any measured rest floor now has a derived meaning — it measures the distance from marginality, Γ_rest/Γ = (Δ_eff/Γ)²·|…|², rather than being a bare parameter.

**3. The relaxation spectrum is DERIVED (gate 7.4).** The poles are (Γ/2)·eig(AA†) — the graph spectrum, **multi-pole** (ring: {1, 1, 2} in units of Γ/2), exactly the "sum of filters" the audit said a real generator would give. The single-pole n48 monitor is the one-mode truncation. And the derived knee *shape* for a locally modulated source on this graph: within 0.1% of a single Lorentzian (the source's overlap weights concentrate on one pole cluster) — compared against n48b's discriminability table (two-pole 8.4%, memory 104%), the derived kernel sits far inside the Lorentzian band. **The single-pole postulate graduates from an assumption to a derived approximation with a computed error.**

## Honest boundary

Gaussian/linear throughout; the matter is still a *parameterized* branch label, not a dynamical matter sector (the full gate-7.3 demand — darkness under a complete matter Hamiltonian — remains); the annihilation-type L_i is a declared monitor choice (Hermitian-quadrature monitors would need the same analysis); the graph is small. What changed: the trunk's flagship property is no longer true-by-definition — it is true-by-computation at a named point, with a derived penalty for leaving it.

**Class: [A-model, exact + Fock-verified]; the marginality identification [B — two standing mechanisms shown to coincide]; nothing about nature claimed.**

## Amendment (n226 sequential audit, 2026-07-11)

Result 3's "within 0.1% of a single Lorentzian" was an artifact of the N = 4 ring's
near-degenerate spectrum. Recomputed sequentially on realistic substrates (diamond
N = 216): max deviation from the best single Lorentzian is **5.8% on the perfect
crystal and 6.1% ± 0.1% on the glass** — inside the two-pole discriminability band
(8.4%, n48b) but far from 0.1%. The pole identity (poles = (Γ/2)·eig(AA†)) stands
graph-generically; the *shape* statement is corrected, and the ~6% multi-pole knee
deviation is registered as a native substrate prediction. Under P3″ the whole
construction reads per-ledger: each sector's zero-at-rest is exact at ITS OWN
marginal point; a measured rest floor measures that sector's distance from its own
marginality.