Closure of a Deuterium–Helium-3 Tandem Mirror with Direct Conversion: A 35-Simulation Validation Package

Supplementary data and software for a deuterium–helium-3 tandem-mirror generator with direct energy conversion. The engineering gain and neutron fraction are length-independent; net electric power scales with central-cell length. The same machine is deployed in two housings — MetroVolt (data-centre generator) and Aegis (fixed-site defence generator); the difference is deployment, not a separate physics case. Scope — physics only. All commercial analysis (cost, LCOE, $/kg, capital, market, price) is excluded. The validation suite operates at the design point. run_all.py (H41–H80) evaluates the closing configuration; the H41 gate reproduces Q_E 1.3091 and the suite runs 33 OK / 0 FAIL / 2 SKIP. The design point is a near-thermal point (Te/Ti 0.9975) that closes; H47 verifies the near-thermal Te, and H50/H75/H80 report closure conditional on the end-plug requirement (H53). Also reproducible standalone via tools/reproduce_design_point.py. The design point (M-45, reproducible): one locked config — n_p/n_c 16, x(³He) 0.30, Ti 90 keV (Te 89.78, near-thermal), ne 2.6e20, B_m 17 T, β_c 0.55, a_c 0.86 m, ions-only → Q_E 1.31, f_n 5.44% (length-independent); l_c 55/440/1400 m give P_fus 0.54/4.3/13.7 GW and net +104/+850/+2832 MWe. Low-neutron, not aneutronic.Mandatory caveat (M-47): closure is REQUIREMENT-CLASS, not demonstrated — contingent on end-plug density n_p/n_c 16 → 4.16e21 m⁻³ = 347× GDT-measured / 26× the best published mirror (H53); at n_p/n_c 10 it does not close (Q_E 0.63, net −160 MWe). Fuel-cleanliness trajectory (M-46): closes across x(³He) ∈ [0.20, ~0.43]; f_n falls 9.53% → 2.77%; design x=0.30 (f_n 5.44%), free clean-shift to x=0.35 (f_n 4.18%, still net-positive). Near-aneutronic x ≥ 0.45 does not close. Withdrawn 2026-08-02 (non-reproducible): Q_E 1.002/1.191/1.825 and net +2.78/+128.5/+363. Two-tier honesty: across the 45-row validation manifest (the 35-simulation H-suite plus the B- and M-series freeze entries): 24 PASS, 14 REQUIREMENT, 7 INFORMATIVE. Named extrapolations: the thermal barrier, electron/thermal direct conversion (unmeasured in-band; the design point doesn't use it), the end-plug winding at the required throat field (26.49 T / 39.74 T, both below an existing 32.35 T record — the winding structure is the extrapolation, not the field), alpha-channelling, holdoff voltage (1.67 MV required vs 76 kV ever held), the plug density ratio, and the ³He supply requirement. Novel physics result: synchrotron loss is not a fixed fraction of fusion power — the effective harmonic cutoff scales as the ⅓ power of minor radius. Framing results: pure D-D does not close in a mirror (Q_E 0.43); the failure is the charged-power budget, not confinement. The Aegis shipboard variant closes negative — Aegis is a fixed-site generator. Contents: the 35 physics/validation simulations plus B/M freeze entries; the shared reduced-order evaluator; the design card, reference card and recorded ledgers; figures; and a machine-readable validation manifest. Reproduce with python3 run_all.py on the pinned environment. Project page (physics validation & simulation): https://www.kronosfusionenergy.com/physics_Validation_Simulation License CC BY 4.0. DOI 10.5281/zenodo.21746479. Companion (Hyperion breeder) 10.5281/zenodo.21746157.