Logbook — 2026-04-13 — WP-E v0.3 decisions

Context. WP revised from v0.2 → v0.3 incorporating Guardian v0.2 and external v0.2 reviews. This entry records the decisions taken on Q0–Q8, the preflight protocol specification, and the Hamiltonian-code line that fixes the φ_α sign convention.

1. Decisions on open questions Q0–Q8

Q	Decision	Status	Section in v0.3
Q0	WP-E — Forward Map and Observability (new node)	Resolved	§1, §8
Q1	Observable = complex C + σ_z	Resolved	§2, §8
Q2	2D slices S1/S2/S3 (Option B); 3D cube deferred	Resolved	§4, §8
Q3	Reference models R1, R2, R12; residuals Δη, Δt, cross	Resolved; LD closed-form TBD	§3.1, §4, §8
Q4	δ₀ ±6 MHz;	α	∈ {0,1,3,5}; φ_α ≥ 48, 64 preferred
Q5	Extend `stroboscopic_sweep.py` (default); new file only if invasive	Conditional — decide after preflight	§8
Q6	φ_α anchored to engine Hamiltonian (see §2 of this entry)	Resolved	§2.2, §8
Q7	Pure-state unitary; reopens if §4a D2 trips	Resolved with gate	§2, §4a, §8
Q8	Fix N = 22;	C	normalised to reference state (

Q0 was resolved in favour of the external reviewer's recommendation (WP-E, a new observable-forward-map node) rather than folding into WP-B or WP-A.1. Rationale: the forward map is the object every downstream injectivity/backaction claim depends on; giving it its own node makes that dependency auditable.

The π/2 inconsistency (N · θ_pulse = 1.53 rad vs. π/2 = 1.571 rad) was resolved by acknowledgement rather than parameter adjustment (A2, E3). Adjusting θ_pulse to 1.571/22 = 0.0714 rad would require retuning Ω_eff or δt, which misrepresents the nominal device parameters. Defining |C| relative to the reference state (§2.1) removes the ambiguity without touching the parameters.

2. Hamiltonian-code line fixing the φ_α sign convention

Reference: ../../scripts/stroboscopic_sweep.py.

Primary anchor — coherent-state preparation (line 103):

alpha = alpha_abs * np.exp(1j * theta)   # with theta = np.radians(alpha_phase_deg)

This is the unambiguous definition of φ_α in the engine: the complex phase of the coherent-state amplitude α passed to the Fock-basis expansion.

Supporting anchor — displacement generator (line 138):

gen = alpha * adag - np.conj(alpha) * a   # D(α) = exp(α a† − α* a)

Standard Wigner-convention displacement operator. Combined with the position quadrature definition X = a + adag (line 88), this fixes:

φ_α = 0 → coherent state on +X̂ axis in phase space (⟨X̂⟩ > 0, ⟨P̂⟩ = 0);
φ_α = π/2 → coherent state on +P̂ axis.

Coupling Hamiltonian — line 198–200:

H[:nmax, nmax:] = (omega_r / 2) * C        # ↓↑ block (σ₋), C = exp(iη X)
H[nmax:, :nmax] = (omega_r / 2) * Cdag     # ↑↓ block (σ₊)

with C = exp(iη(a + a†)) constructed at line 96 via expm(1j * eta * X). This confirms that the stroboscopic analysis pulse couples through the exponential-in-X operator and that all sign conventions for ∂C/∂φ_α follow from the chain:

alpha (line 103)  →  D(α) (line 138)  →  X̂ = a + a† (line 88)  →  H_int = (Ω/2)(C σ₋ + C† σ₊) (line 199–200)

Consequence for the WP. Every residual plot's φ_α sign is interpreted in this convention. Any future engine (new phase_contrast_map.py or a rewrite) must reproduce the φ_α → (⟨X̂⟩, ⟨P̂⟩) mapping bit-for-bit at preflight anchor.

3. Preflight protocol (§4a of v0.3)

Anchor point. (δ₀ = 0, |α| = 0, φ_α = 0).

Engines under test. 1. HDF5 adaptive-learner (legacy — produced 0.61/0.71/0.84/0.75 contrast series). 2. JSON-uniform engine (legacy — produced ≈ 0.56 contrast). 3. Candidate engine (extended stroboscopic_sweep.py or new phase_contrast_map.py).

Observables recorded per engine. - Per-pulse Bloch vector (⟨σ_x⟩, ⟨σ_y⟩, ⟨σ_z⟩) at pulse indices 1 … 22. - Accumulated phase arg C(0, 0, 0). - Final contrast |C(0, 0, 0)| (raw, not yet normalised).

Gate outcomes. - Agree within integrator tolerance → proceed to §4 deliverables; legacy engines used for spot-check cross-validation. - Disagree with cause identified among (i) φ_α sampling, (ii) hidden ensemble averaging, (iii) Ω vs Ω_eff calibration, (iv) integrator tolerance → publish logbook entry, patch/document/quarantine, re-run. - Disagree with no identified cause → main sweep paused; mismatch becomes a WP-E sub-deliverable.

D2 clause. If the legacy engines' noise models are not documented to be identical to the candidate engine's pure-unitary configuration, Q7 is reopened before the main sweep — not after.

4. Outstanding actions

[ ] Code-reading pass on ../../scripts/stroboscopic_sweep.py to confirm: (a) whether ⟨σ_x⟩, ⟨σ_y⟩ are exposed or require backend patch (Q1 + Q5); (b) whether a closed-form LD linear prediction exists for R1 (Q3); (c) runtime estimate for 64 × 121 × 5 = 38,720 runs (Q4).
[ ] Separate task: update cross-references in ../../ARCHITECTURE.md and ../../ideal-limit-principles.md with the WP-E node. Not modified in v0.3 per the non-goals clause of the revision task.
[ ] Preflight run (§4a). First logbook entry after this one should report preflight results.

5. Files touched in this revision

../README.md: v0.2 → v0.3 full rewrite incorporating Blocks A–H of the Guardian task specification.
[this file]: new logbook entry.

No code changes to ../../scripts/stroboscopic_sweep.py or to ../../ideal-limit-principles.md were made, per the non-goals clause.