IX-StellaratorForge

Evidence-driven stellarator fusion reactor design and computational co-design program.

IX-StellaratorForge preserves the original IX-Fusion research lineage and builds a reactor-level evidence program above it. SFR-1 Rev A remains the steady-state reference architecture; SFR-2 Rev A remains a separate dynamic-compression assumption breaker; SFR-3 Field Integrity Shell A remains the synthetic magnetic-error-control branch; Dual Boundary AHIS A remains the independently monitored engineering boundary; SFR-4 Integrated Physical-Promotion Campaign A remains the integrated heat/magnetic campaign. Version 0.10.0 adds SFR-5 Reality Gradient and Adaptive Inverse Design A, which converts the SFR-4 magnetic rejection into a fail-closed architecture-search program instead of increasing brute-force sampling of the same family.

Release: 0.10.0: Reality Gradient and Adaptive Inverse Design

Repository verdict: GREEN when all preserved evidence, the SFR-4 integrated campaign, and the SFR-5 adaptive inverse-design autopsy reproduce.

SFR-2 primary verdict: NO_PRIMARY_CASE_CROSSES_OPTIMISTIC_IGNITION_PROXY

Actuation-overlay verdict: NO_DECLARED_BREATHING_CASE_IMPROVES_BOTH_CYCLE_AVERAGE_PROXY_AND_FUSION_POWER

SFR-3 verdict: SYNTHETIC_HARMONIC_CONTROL_DEMONSTRATED__PHYSICAL_CONFINEMENT_UNPROVEN

Dual-boundary verdict: DUAL_BOUNDARY_ARCHITECTURE_SCREEN_PASS__PHYSICAL_SURVIVABILITY_AND_CONFINEMENT_UNPROVEN

SFR-4 verdict: INTEGRATED_REDUCED_CAMPAIGN_COMPLETE__NOMINAL_HEAT_ENVELOPE_SCREEN_PASS__PHYSICAL_COIL_EQUILIBRIUM_CONFINEMENT_AND_FUSION_UNPROVEN

SFR-5 verdict: REALITY_GRADIENT_AUTOPSY_COMPLETE__CURRENT_MAGNETIC_FAMILY_REJECTED__ADAPTIVE_INVERSE_DESIGN_PATH_DEFINED__NO_PHYSICS_PROMOTION

No branch claims demonstrated confinement, ignition, net energy, tritium self-sufficiency, net electricity, buildable hardware, reactor safety, or reactor feasibility.

What v0.10.0 adds

SFR-5 treats the 0 / 80 SFR-4 magnetic result as information about the search representation, not an invitation to sample the same family more densely. Its diagnosis is regenerated from the committed SFR-4 result and thresholds rather than hard-coded into a new narrative.

SFR-5 diagnostic	Executed result
SFR-4 direct-filament candidates	80
Combined topology passes	0
Mean-transform factor required to reach the minimum iota gate	3.8265x
Remaining normalized excursion slack	3.887%
Held-out richer-basis RMS normal-field error / declared limit	12.9825x
Earned fusion-progress credit	0

The two magnetic diagnostics remain separate evidence lanes because they are not the same physical coil representation. SFR-5 rejects continued brute-force search of the fixed helical plus fixed hybrid filament basis as the preferred next move, while explicitly not rejecting stellarators, quasi-isodynamic configurations, nonplanar modular coils or fusion.

The new inverse-design contract allows the plasma boundary, winding surface, nonplanar coil geometry, current groups, REBCO orientation and engineering keep-outs to move. Constraint failures become normalized pressure signals. A backward geometry gradient is only permitted when a real solver supplies analytic, automatic-differentiation or controlled finite-difference sensitivities; the base release deliberately records that step as NOT_RUN.

Primary v0.10 artifacts:

• results/sfr5/SFR5_REALITY_GRADIENT_A_RESULT.md
• results/sfr5/sfr5_reality_gradient_a_v0100.json
• configs/reactor/sfr5_reality_gradient_a.json
• docs/reactor/30_SFR5_REALITY_GRADIENT.md
• docs/reactor/31_SFR5_PROMOTION_GATES.md
• external_solvers/sfr5_inverse_design_evidence_contract.json
• provenance/SFR5_REALITY_GRADIENT_TECHNICAL_BASIS_2026.json

What v0.9.0 adds

All seven requested computational workstreams were attempted. The result is deliberately split: the heat architecture passes its declared nominal and steady reduced envelope, while the scanned physical coil family fails.

Workstream	Executed result	Promotion decision
Physical coil field	80 direct-filament Biot-Savart and field-line cases	0 combined topology passes; current family rejected
Finite-beta equilibrium	DESC and VMEC++ adapters explicitly attempted	Dependencies unavailable; no equilibrium result fabricated
Coil/plasma co-design	Expanded geometry/current scan and held-out normal-field reconstruction	Production SIMSOPT co-design not run
Particle confinement	3.5 MeV alpha gyroradius scope, approximately 0.0449 m at 6 T	No guiding-center or transport credit
Burn	Q=20 target requirement with declared alpha deposition and bremsstrahlung	Approximately H_ISS04 2.00 at design iota; not linked to a passing coil
Magnet engineering	Centerline strain, magnetic pressure and stored-energy scopes	Winding-pack FEA, peak conductor field and quench qualification open
Reactor systems	D-T source, breeding coverage, heat and conditional power ledgers	Full 3-D TBR and net-electric prediction open

Heat-exhaust result

The heat problem is separated into first-wall radiation, divertor exhaust and blanket neutron heating. At the declared 1 GW fusion target and Q=20 ledger, the selected requirement sends 60% of 228.98 MW plasma exhaust into controlled radiation and 40% to a 24 m2 effective island-divertor wetted area.

Quantity	v0.9 reduced result
First-wall peak heat flux	0.294 MW/m2
Divertor peak heat flux	5.724 MW/m2
Nominal first-wall tungsten surface	401.2 C
First wall at declared 1 MW/m2 steady upper bound	523.8 C
Divertor tungsten surface	678.3 C
Divertor water mass flow	426.0 kg/s through 960 parallel channels
Mean channel velocity	7.56 m/s
Hydraulic pumping screen	0.122 MW

The retained architecture uses a helium-cooled segmented-tungsten/graded-W-RAFM/ODS-RAFM first wall, an isolated PbLi DCLL blanket, and an independently bounded water-cooled W/OFHC-Cu/CuCrZr divertor. Water and PbLi may not share a boundary, penetration or heat exchanger.

“Heat resolved” is restricted to requirement authority for nominal and declared steady conditions. Stable detachment, the three-dimensional island footprint, critical heat flux, erosion, cyclic fatigue, irradiation, transient events, accidents and hardware qualification remain open.

Magnetic result

The best scoring direct-filament case reaches approximately 0.0653 mean iota against a minimum 0.25 gate. Its radial-excursion screen passes, but the transform screen fails. The held-out richer basis remains approximately 6.49% RMS normal field against a 0.5% screen. Raising current in this classical helical family is rejected as the next design move; a fundamentally different optimized nonplanar modular-coil family is required.

Primary v0.9 artifacts:

• results/sfr4_integrated/SFR4_INTEGRATED_PHYSICAL_PROMOTION_A_RESULT.md
• BOM/SFR4_INTEGRATED_PROMOTION_BOM.md
• docs/reactor/25_SFR4_INTEGRATED_PHYSICAL_CAMPAIGN.md
• docs/reactor/26_SFR4_HEAT_EXHAUST_RESOLUTION.md
• docs/reactor/27_SFR4_SOLVER_ATTEMPTS.md
• docs/reactor/28_SFR4_PROMOTION_GATES.md
• docs/reactor/29_SFR4_DECISION.md

What v0.8.0 adds

The user's "AHIS flipped inside out" idea is implemented as two independent, instrumented engineering boundaries around the unchanged magnetic-confinement concept. The inner lane looks inward from protected positions behind the plasma-facing armor and first wall. The outer lane monitors the double-wall vessel, shielding, cryostat and magnet-support alignment. Neither lane enters the plasma volume, presses on plasma, or receives direct confinement credit.

Item	Selected v0.8 configuration	Reduced-screen result
Plasma-facing stack	Segmented tungsten, graded W-to-RAFM transition, helium-cooled ODS-Eurofer/RAFM first wall	420.7 °C nominal surface; 632.9 °C declared steady-upset upper bound
Breeding and shielding	Enriched PbLi DCLL blanket, electrically insulating SiC/alumina channel inserts, WC/B4C shield, local HfH only after 3-D neutronics	Architecture selected; no TBR, corrosion, MHD-pressure-drop or lifetime credit
Double boundary	Monitored 316LN-class double-wall vessel and interspace, followed by thermal shield, cryostat and monitored support shell	192 paired locations across 24 toroidal sectors and eight poloidal stations
Instrumentation	Two independent inner/outer lanes, hard vacuum/interspace channels and sector-level references	1,736 declared sensing elements; radiation survivability and calibration remain unqualified
Fault campaign	Eleven nominal and fault scenarios	All expected deterministic states reproduced; a sub-sensitivity armor crack is intentionally not detected and requires periodic NDE
Fusion effect	None credited	Earned confinement, fusion, ignition and net-electric improvement remain exactly zero

The 1-D temperature and coefficient-of-thermal-expansion calculations are sizing screens, not thermal FEA, fracture mechanics or lifetime predictions. The reported raw mismatch proxy for the selected stack is approximately 0.212% and is deliberately not converted into stress or pass/fail structural credit. The cooler W/Cu/CuCrZr comparison is deferred because a water-cooled first wall beside PbLi creates a more severe integration and accident problem. The SiCf/SiC comparison remains attractive at high temperature but is deferred by joining, code-qualification and maturity gaps.

The architecture can trigger magnetic trim, power rundown, coolant isolation, safe hold and inspection. It cannot squeeze plasma with wall pressure. Its purpose is to detect damage, preserve geometry and prevent an engineering fault from silently degrading the magnetic cage.

Primary v0.8 artifacts:

• results/sfr3_dual_boundary/SFR3_DUAL_BOUNDARY_AHIS_A_RESULT.md
• BOM/SFR3_DUAL_BOUNDARY_AHIS_BOM.md
• docs/reactor/21_SFR3_DUAL_BOUNDARY_ARCHITECTURE.md
• docs/reactor/22_SFR3_DUAL_BOUNDARY_MATERIAL_SELECTION.md
• docs/reactor/23_SFR3_DUAL_BOUNDARY_FAULT_CAMPAIGN.md
• docs/reactor/24_SFR3_DUAL_BOUNDARY_PROMOTION_GATES.md

What v0.7.0 adds

Seven uploaded repositories—AHIS, PressureX, IX-Vibe, IX-Breath, IX-GCR-SPE, IX-Shield and IX-HfTaZen-Shield—were audited as mechanism donors. No hidden material or liquid confinement mechanism was found. Their defensible patterns were translated into a magnetic field-integrity architecture:

SFR-3 layer	Decision	v0.7 evidence state
Primary confinement	Keep the steady copper-stabilized REBCO field and rigid vessel	Architecture reference only
Active correction	Add 24 individually driven planar trim channels	12-by-24 synthetic response is full row rank; physical response not solved
Passive correction	Study 24 flux-conserving superconducting loops	Transient response only; exactly zero DC-error credit
Observability and FDIR	Add magnetic/current/quench/strain/motion sensing, confidence gates and independent safe hold	Fault-state logic executes deterministically
Structural control	Target measured support modes with warm-side damping and alignment monitoring	Architecture only; no unmeasured damping credit
Magnet protection	Concentrate WC/B4C and solver-dependent HfH shielding around REBCO and streaming paths	No transferred tokamak performance; 3-D OpenMC required
Plasma-facing/edge	Segmented tungsten family; boron or local liquid lithium only as guarded experiments	Zero direct confinement credit

The declared synthetic commissioning challenge produces approximately 65.46% nominal RMS reduction, 66.36% with one unavailable trim channel, and 55% passive attenuation of a pure transient challenge. Low sensor confidence disables active correction. Those numbers validate the mathematical control decomposition only; the repo assigns them zero confinement, fusion-power, ignition and net-electric gain.

The 52-row architecture inventory is BOM/SFR3_FIELD_INTEGRITY_SHELL_BOM.csv. The result is results/sfr3_field_integrity/sfr3_field_integrity_shell_a_v070.json. Physical promotion begins by replacing the analytic response matrix with CAD-linked Biot-Savart fields and then testing finite-beta equilibria and islands.

What v0.6.0 adds

The baseline 23 / 26 / 23 / 26 ft ABAB geometry, rigid vessel and steady primary HTS field remain unchanged. The new overlay tests three auxiliary magnetic-field patterns over 720 samples per cycle at depths from 0% through 5%.

Actuation question	v0.6 result
Does synchronous squeeze and expansion sustain an advantage?	No. The 5% case produces an instantaneous proxy peak near 0.9981 during expansion, when uniform fusion power has fallen to about 802.5 MW. Its cycle-average proxy is worse than the unchanged baseline. The peak is not ignition capture.
Does an ABAB-opposed wave improve both principal metrics?	No. No declared case improves both cycle-average optimistic ignition ratio and cycle-average uniform fusion power.
Does the traveling-quadrature wave move closer?	Only nominally in one proxy and not jointly. At 5% depth the cycle-average ratio changes from about 0.961077 to 0.961559 while cycle-average uniform fusion power falls from 1000.000 to about 997.233 MW, before actuator losses.
Does any cycle-average case cross the proxy?	No.
Does the Gemini tri-lobe image supply a fusion mechanism?	No earned gain. Astrophysical accretion and three-body collision claims are rejected. An area-preserving poloidal m=3 harmonic repeated inside all four field periods is retained only as a testable actuator symmetry.
Is the actuator hardware specified?	At architecture level only. The 24-row overlay BOM selects eight normal-conducting triplet stations and 24 independent circuits. Currents, turns, voltage, cooling, forces and placement remain solver or hardware dependent.

The machine-readable result is results/sfr2_actuation/sfr2_actuation_overlay_a_v060.json. The concise interpretation is in results/sfr2_actuation/SFR2_ACTUATION_OVERLAY_A_RESULT.md.

What v0.5.0 adds

v0.5.0 does not rewrite SFR-1. It adds SFR-2 Rev A as a separately gated assumption-breaker study built from the 23 / 26 / 23 / 26 ft concept.

SFR-2 question	v0.5 result
Four machines or one plasma system?	One continuous closed toroidal plasma system. The four dimensions are consecutive ABAB sector arc lengths, not independent plasmas connected by ducts.
Does the 23/26 staggering itself earn confinement credit?	No. v0.5 assigns zero benefit to staggering until a real 3-D equilibrium/transport calculation demonstrates one.
Does “magnetic rifling” help in the empirical screen?	Yes, monotonically inside ISS04. At the highest declared 15 T field with no compression, the target-power-matched ratio rises from ~0.814 at iota=0.6 to ~0.961 at iota=0.9. This is not proof that such a field is realizable.
Does radial compression automatically improve the ignition screen?	No. At the same 1 GW target and H_ISS04=1, 0%, 5%, and 10% squeeze give approximately 0.961, 0.889, and 0.816 respectively at the most favorable declared field/transform point. The smaller minor radius penalizes ISS04 confinement.
Is phase-controlled traveling-wave/RF heating credited?	No. Numerical credit is exactly zero until a self-consistent wave/plasma calculation exists.
Is magnetic-flux compression credited?	No. Axis field is held fixed in the compression screen.
Did any primary SFR-2 point cross the optimistic ignition proxy?	No. The best primary case requires H_ISS04 ≈ 1.0405 in an already optimistic alpha-only balance. That is not “4% away from fusion.”

The implementation intentionally supersedes earlier conversational percentages. Only repository-generated results count as SFR-2 evidence. See docs/reactor/12_SFR2_DYNAMIC_COMPRESSION.md and results/sfr2/SFR2_REVA_SCREEN_RESULT.md.

What v0.4.0 resolves

v0.4 turns the remaining red-X list into explicit executed calculations or production-solver jobs. It does not rename a surrogate as high fidelity.

Problem	v0.4 result
Failed fixed coil basis	Resolved as rejected. Two new architecture classes are executed: classical helical TF+helix co-design scans and a continuous winding-surface current-potential reconstruction. Neither earns promotion.
G1 equilibrium	Input-generation problem closed; production solve open. Four finite-pressure VMEC-family seed files are generated from the exact analytic SFR-1 screening boundary for DESC/VMEC++ execution and cross-check.
G3/G4 confinement	Requirements/topology quantified; kinetic validation open. Q=10 confinement requirement and vacuum field-line transform/nestedness are executable; alpha-orbit, neoclassical and gyrokinetic production evidence remains required.
G2 HTS magnets	Geometry screen executable; magnet qualification open. Curve length, bend radius and conservative REBCO hard-way strain proxy are calculated; field-performance failure prevents promotion and winding-pack Ic/FEA/quench/manufacturing qualification remains mandatory.
G7 neutronics/TBR	Source and geometric breeding bounds closed; full 3-D transport open. Exact D-T source/tritium ledgers and blanket-coverage constraints are calculated; an OpenMC CSG torus proxy builder and 3-D statepoint analyzer are included.
G8 net electric	Conditional plant equations closed. The current ~704.57 MW uniform screen gives ~204.10 MWe under the declared plant assumptions; the 300 MWe floor requires ~913.04 MW fusion; the 1 GW target ledger gives 340 MWe.
Actual net-electric fusion	Not computationally resolvable. It requires physical hardware and calibrated measurements.

Current numerical snapshot

At the current 8.0 m major radius, 1.7 m minor-radius screening geometry, 6 T axis field, 3% beta and 15 keV uniform D-T screen:

• screened fusion power: 704.57 MW;
• fusion power required to clear the 300 MWe plant floor under the existing plant assumptions: 913.04 MW;
• corresponding fixed-temperature uniform-beta screen for that net floor: ~3.415% beta;
• beta screen for the 1 GW target: ~3.574%;
• current screened net-electric algebra: ~204.10 MWe;
• 1 GW target net-electric algebra: 340 MWe;
• expanded v0.9 best-scoring classical-helical vacuum architecture: 6FP, four alternating helices, 1.0 helical/TF current ratio; it retains the radial-excursion screen but produces only ~0.0653 mean iota, so it fails the transform requirement;
• continuous winding-surface current-potential held-out RMS Bn/B0: ~1.64% (2FP), 2.32% (3FP), 3.05% (4FP), 4.57% (6FP) — all fail the 0.5% screen;
• the geometry-only REBCO strain proxy is below the 0.4% ceiling for the screened TF/helical centerlines, but that is not winding-pack qualification.

These numbers are evidence at their stated authority only. A failed reduced/intermediate screen can reject an architecture; a passing reduced screen cannot prove a reactor.

High-fidelity execution pack

The repository contains exact fixed-boundary finite-pressure seed inputs:

• external_solvers/inputs/input.SFR1_QA_2FP_REF
• external_solvers/inputs/input.SFR1_QI_3FP
• external_solvers/inputs/input.SFR1_QI_PWO_4FP
• external_solvers/inputs/input.SFR1_C6_QI_6FP

and executable adapters:

• external_solvers/adapters/run_desc_equilibrium.py
• external_solvers/adapters/run_vmecpp_equilibrium.py
• external_solvers/adapters/build_openmc_axisymmetric_proxy.py
• external_solvers/adapters/analyze_openmc_3d_tbr.py
• external_solvers/adapters/validate_production_receipt.py

The adapters stop if the real dependency is unavailable. They never silently replace DESC, VMEC++ or OpenMC with in-repo surrogate physics.

Run the in-repo campaign

python scripts/run_computational_closure.py
python scripts/generate_v040_evidence.py
python scripts/run_sfr2_screen.py
python scripts/run_sfr2_actuation_overlay.py
python scripts/run_sfr3_field_integrity.py
python scripts/run_sfr3_dual_boundary.py
python scripts/run_sfr4_integrated_campaign.py
python scripts/run_sfr5_reality_gradient.py
python check_stellarforge.py

The first command is the expensive deterministic architecture campaign. The release gate validates its tracked evidence rather than repeatedly recomputing the same field scan in several subprocesses.

Reactor envelope

Item	Rev A value	Authority
Major radius	8.0 m	reference-informed target
Minor radius	1.7 m	design target
On-axis field	6.0 T	reference-informed target
Volume-average beta	3.0%	screening/design point
Fusion-power class	1,000 MW	design target
HTS operating temperature	20 K	reference-informed target
Maximum field on conductor	≤20 T	design ceiling
Plasma-to-coil distance	1.35 m target; 1.20 m hard floor	design constraint
Full-3D TBR	≥1.15 target; 1.10 hard floor	production-solver objective
Thermal conversion efficiency	40% target	plant target
Recirculating-power ceiling	120 MW	design target
Net-electric floor	300 MW	design requirement, not prediction

No core is privileged: 2FP QA reference, 3FP QI, 4FP QI/piecewise-omnigenous, 6FP C6 lineage and a direct-J assumption-breaker remain competitors.

BOM / design inventory

BOM/SFR1_FULL_SYSTEM_BOM.csv and .md contain 87 unique system-level rows spanning plasma, magnets, cryogenics, vacuum, PFCs, blanket/shield, tritium/fueling, RF, diagnostics, controls, heat transport, power conversion, electrical, maintenance, safety, facility and solver infrastructure.

BOM/SFR3_FIELD_INTEGRITY_SHELL_BOM.csv contains 52 rows covering direct field sources, trim and passive correction, diagnostics, guarded control, support-mode damping, magnet shielding, plasma-facing and blanket integration, maintenance and explicit rejected/separate concepts. Diamond, glitter, passive dense liquids and flexible-vessel compression are not hidden as active components.

BOM/SFR3_DUAL_BOUNDARY_AHIS_BOM.csv contains 64 rows covering the selected tungsten/RAFM/PbLi stack, two comparison stacks, independent inner and outer sensing, interspace leak monitoring, support/alignment surveillance, protection actions, evidence jobs and explicit rejection/defer decisions. PVDF and unqualified electronics are not placed at the plasma-facing surface.

BOM/SFR4_INTEGRATED_PROMOTION_BOM.csv contains 64 rows covering all seven workstreams, the separated helium/PbLi/water heat architecture, production solver requirements, particle and burn tools, magnet qualification, neutron transport, fuel cycle, remote maintenance and high-heat-flux testing.

“Full” means complete at the system-architecture inventory level. It is not falsely labeled a procurement/fabrication BOM: final quantities, nuclear compositions/enrichment, conductor sizing/current, pressure ratings, safety setpoints, vendors, part numbers and drawings remain solver/hardware dependent.

Evidence gates after v0.10

1. G1 equilibrium — production execution open. Inputs are complete; real finite-beta DESC and VMEC++ convergence/cross-code evidence has not been executed in this build runtime.
2. G2 coils — new architectures executed, no promoted set. Low/intermediate tests reject current fixed boundary/coil combinations; true plasma/coil co-optimization, REBCO Ic/strain over winding packs, loads/support FEA, tolerances and quench remain open.
3. G3/G4 confinement — requirements and vacuum topology quantified, kinetic transport open.
4. G5 edge/divertor — open.
5. G6 RF — resonance-scale screen retained; 3-D deposition/wall-plug closure open.
6. G7 neutronics — source/coverage bounds complete; full 3-D OpenMC/DAGMC transport open.
7. G8 system — conditional algebra closed; coupled prediction waits on G1–G7.
8. G9 hardware — open by definition.

SFR-2 has its own independent gates. Only SFR2_G0_SPEC is PASS_SPEC_ONLY; dynamic equilibrium, coils/field, particle and alpha orbits, transport/MHD, transient edge heat flux, RF/phase control, neutronics/TBR, integrated burn/plant, and hardware are all NOT_RUN. SFR-1 evidence cannot silently promote SFR-2.

The actuation overlay is independently gated. Only SFR2A_G0_OVERLAY_SPEC is PASS_SPEC_ONLY. Time-dependent equilibrium, coils/electromagnetics, magnetic-pumping kinetics, particle and alpha orbits, transport/MHD, wall loads, integrated power and hardware all remain NOT_RUN.

SFR-3 is independently gated. SFR3_G0_ARCHITECTURE_SPEC is PASS_SPEC_ONLY and SFR3_G1_SYNTHETIC_CONTROLLABILITY is PASS_LOW_AUTHORITY_SYNTHETIC_ONLY. Physical Biot-Savart response, free-boundary equilibrium/islands, particle and alpha orbits, transport/MHD, magnet engineering, 3-D neutronics/TBR, integrated burn/plant and hardware are all NOT_RUN. The synthetic pass cannot promote any of them.

Dual Boundary AHIS A is also independently gated. Only its configuration and reduced thermal/fault-logic screens pass. Conjugate thermal FEA, disruption/EM structural FEA, W/RAFM joint qualification, coolant and PbLi MHD/corrosion loops, sensor irradiation/calibration, 3-D neutronics, remote maintenance and integrated hardware tests remain NOT_RUN. No passing reduced screen can promote those gates.

SFR-4 is independently gated. Its specification passes; its scanned coil family fails; its nominal and declared steady reduced heat screen passes. CAD-linked nonplanar coil optimization, DESC/VMEC++ equilibrium, islands/stability, alpha and thermal transport, 3-D edge/divertor physics, CFD/FEA/fatigue/irradiation, OpenMC TBR, integrated burn/plant analysis and hardware remain NOT_RUN.

SFR-5 is independently gated. Only SFR5_G0_SPEC_AND_AUTOPSY is PASS_REDUCED: the v0.9 magnetic evidence is reinterpreted as an architecture-family rejection and a declared inverse-design search direction. Movable-geometry sensitivities, winding-surface feasibility, discrete nonplanar-coil realization, single-stage finite-beta plasma/coil co-design, transport, 3-D heat/magnet/neutronics closure, integrated plant analysis and hardware remain NOT_RUN. Constraint pressure is diagnostic until a real movable-geometry evaluator supplies sensitivities; SFR-5 assigns zero unearned physics or fusion credit.

Quality gate

python check_stellarforge.py

IX-STELLARATORFORGE: GREEN means release integrity, the complete deterministic test suite, preserved SFR-1/SFR-2/SFR-3 evidence, the dual-boundary reduced screen, the SFR-4 integrated campaign, the SFR-5 Reality Gradient autopsy, BOM contracts, license, tracked computations and solver contracts reproduce. It never means fusion was achieved.

License and permission contact

This repository is source-available for research/evaluation only under LicenseRef-IX-StellaratorForge-Eval-Only-1.1; it is not OSI open source. Physical construction, manufacturing, deployment, reactor/lab-plasma operation, commercial use, power generation, or use outside the grant requires a separate written license from the project owner.

Licensing/permission inquiries: https://www.linkedin.com/in/brycewdesign/

A LinkedIn connection, message, discussion, download, citation or repository access does not itself grant additional rights. See LICENSE, LICENSING.md, and NOTICE.

Start here

• FINAL_STATUS.md
• PROOF_OF_CONCEPT.md
• docs/reactor/30_SFR5_REALITY_GRADIENT.md
• docs/reactor/31_SFR5_PROMOTION_GATES.md
• results/sfr5/SFR5_REALITY_GRADIENT_A_RESULT.md
• results/sfr5/sfr5_reality_gradient_a_v0100.json
• docs/reactor/25_SFR4_INTEGRATED_PHYSICAL_CAMPAIGN.md
• docs/reactor/26_SFR4_HEAT_EXHAUST_RESOLUTION.md
• docs/reactor/27_SFR4_SOLVER_ATTEMPTS.md
• docs/reactor/28_SFR4_PROMOTION_GATES.md
• docs/reactor/29_SFR4_DECISION.md
• results/sfr4_integrated/SFR4_INTEGRATED_PHYSICAL_PROMOTION_A_RESULT.md
• BOM/SFR4_INTEGRATED_PROMOTION_BOM.md
• docs/reactor/21_SFR3_DUAL_BOUNDARY_ARCHITECTURE.md
• docs/reactor/22_SFR3_DUAL_BOUNDARY_MATERIAL_SELECTION.md
• docs/reactor/23_SFR3_DUAL_BOUNDARY_FAULT_CAMPAIGN.md
• docs/reactor/24_SFR3_DUAL_BOUNDARY_PROMOTION_GATES.md
• results/sfr3_dual_boundary/SFR3_DUAL_BOUNDARY_AHIS_A_RESULT.md
• BOM/SFR3_DUAL_BOUNDARY_AHIS_BOM.md
• docs/reactor/17_SFR3_DONOR_TRANSLATION_LEDGER.md
• docs/reactor/18_SFR3_FIELD_INTEGRITY_SHELL.md
• docs/reactor/19_SFR3_MATERIAL_AND_LIQUID_BRANCHES.md
• docs/reactor/20_SFR3_PROMOTION_GATES.md
• results/sfr3_field_integrity/SFR3_FIELD_INTEGRITY_SHELL_A_RESULT.md
• BOM/SFR3_FIELD_INTEGRITY_SHELL_BOM.md
• docs/reactor/12_SFR2_DYNAMIC_COMPRESSION.md
• docs/reactor/13_SFR2_PROMOTION_GATES.md
• results/sfr2/SFR2_REVA_SCREEN_RESULT.md
• docs/reactor/14_SFR2_PHASE_PROGRAMMED_BREATHING.md
• docs/reactor/15_TRILOBE_CONCEPT_TRANSLATION.md
• docs/reactor/16_SFR2_ACTUATION_AND_TRILOBE_GATES.md
• results/sfr2_actuation/SFR2_ACTUATION_OVERLAY_A_RESULT.md
• BOM/SFR2_ACTUATION_OVERLAY_BOM.md
• docs/closure/06_MAXIMUM_COMPUTATIONAL_CLOSURE.md
• results/computational_closure/SFR1_V040_RESULT.md
• BOM/SFR1_FULL_SYSTEM_BOM.md
• external_solvers/README.md

The inherited IX-Fusion documentation/results remain in place as foundation evidence and are not rewritten into a success claim.