Architecture¶

Operational targets¶

Target	Bound
Sensor-to-actuator latency	< 50 nanoseconds
Phase-lock at chamber centre z = 0
Spatial-lock at chamber centre
Plasmoid relative speed at merging	≥ Mach 1 (v_z ≥ 300 km s⁻¹)
Compression peak field	20 tesla

Carrier equations¶

Kinematic FRC merging (PHASE-ORCHESTRATOR + MIF-CORE)¶

For two counter-propagating plasmoids with phases θᵢ, θⱼ, axial velocities v_zi, v_zj, and axial positions zᵢ, zⱼ:

dθᵢ/dt = ωᵢ(t)
       + Kᵢⱼ / (1 + |zᵢ − zⱼ| / L_z) · sin(θⱼ − θᵢ − α)
       + doppler_strength · (v_zi − v_zj) / (0.5 · (|v_zi| + |v_zj|) + ε_v)

The distance-coupling term modulates K, the Doppler term corrects for relative motion using pair-symmetric characteristic-speed normalisation. Both are extracted from the swarmalator family into reusable primitives in scpn-phase-orchestrator (PHA-C.1, PHA-C.2; live versions are reported by the dynamic compatibility matrix, not pinned here). MIF-001 is implemented locally as an upstream-pending Python/Rust/Julia carrier with RK4 phase integration and linear axial positions for the chamber-centre acceptance window.

MIF-002 adds the chamber-fixed moving-frame layer:

dzᵢ/dt = v_zi
reference_error = max_i |zᵢ − z_ref|

It advances [θ, z] with a fixed-step Dormand-Prince RK45 update and exposes reference-window observables while the reusable PHASE-ORCH scpn.upde.moving_frame surface remains upstream-pending.

MIF-003 adds the merge-window monitor:

phase_lock_error = max circular separation(θ)
candidate_lock   = phase_lock_error ≤ 0.01 rad
                   and max_i |zᵢ − z_ref| ≤ 0.002 m
lock_achieved    = candidate_lock for ≥ 3 consecutive samples

It is implemented locally as an upstream-pending Python/Rust monitor until PHASE-ORCH owns the reusable scpn.monitor.merge_window surface.

Pulsed-shot lifecycle (CONTROL + MIF-CORE)¶

MIF-004 models the pulsed-shot lifecycle as the adjacent transition ring:

idle -> ramp_up -> flat_top -> burn -> expansion -> dump -> recharge -> cool_down -> idle

The guard surface consumes plasma telemetry, chamber-centre phase/spatial lock observables, and capacitor-bank telemetry. Python and Rust implementations share the same transition reasons and JSONL audit shape, while Lean proves adjacency determinism and the minimal eight-step return to idle.

Non-adiabatic flux evolution (FUSION-CORE)¶

dψ/dt = −ψ / τ_ψ + R_null · E_θ − η_Spitzer · J_θ

Reference: Ono et al. 1997, Physics of Plasmas 4, 1953, eq. 8. Implemented in scpn-fusion-core as scpn_fusion.core.current_diffusion.solve_flux_evolution_nonadiabatic (FUS-C.3); the live FUSION version is reported by the dynamic compatibility matrix.

Magneto-Rayleigh-Taylor growth (FUSION-CORE)¶

γ(k, a_eff) = √(k · a_eff − k² · B_perp² / (μ₀ ρ))

Reference: Velikovich et al. 2007, Physics of Plasmas 14, 022701, eq. 18. Implemented in scpn-fusion-core as scpn_fusion.core.mrti.mrti_growth_rate (FUS-C.4); the live FUSION version is reported by the dynamic compatibility matrix.

Cross-repository ownership¶

SCPN-FUSION-CORE    canonical physics solvers (Hall-MHD, MRTI, tilt, equilibrium)
SCPN-CONTROL        canonical control facade (Petri-net, NMPC, replay, AER ingest)
SC-NEUROCORE        canonical SNN → SystemVerilog emitter, Q8.8 quantiser, AER HDL
SCPN-PHASE-ORCH     canonical Kuramoto family, distance-coupling, Doppler, monitors
SCPN-QUANTUM-CTRL   canonical QAOA-MPC, pulse shaping, QRNG, PQC trigger signer
SCPN-MIF-CORE       canonical pulsed-FRC kinematic + RTL hot-path lab
                    (sub-50-ns trigger fabric, timing-aware formal, AER bridge,
                     Doppler-Kuramoto, moving-frame UPDE, pulsed-shot lifecycle,
                     capacitor-bank model, Faraday recovery)

Anything that falls under a sibling's canonical scope MUST be upstreamed there, not duplicated here. See the internal scope-and-ownership document for the anti-duplication checklist.

Decision records¶

The reasoning behind the major architectural choices — the ownership boundary, the multi-language dispatch, the upstream-pending and prescribed-input discipline, the public API facade, the delivered-versus-roadmap honesty rule, the formal-verification strategy, and the validation-integrity rules — is recorded as architecture decision records.