A compact machine lives or dies by its shield: too thin and the magnets cook; too thick and the geometry that makes it cheap is gone. MetroVolt's shield is optimized with real multigroup transport physics — and the early wrong answer is left in the record.
First-pass analytic shielding suggested a tungsten-boride/B₄C recipe at ~0.9 m; the deposited multigroup analyses (S36/S37) corrected it: B₄C-led recipes fail (6× over target at 1.00 m), and the optimized WC/W₂B₅ stack closes the requirement at ≈1.20 m — a minimum that the radial build then absorbs while preserving 0.42 m of spare. The correction is documented, superseding note and all.
Downstream, the shielded conductor fluence ledger totals 3.23×10²² n/m² over life against a 3.0×10²² degradation-onset reference — ~8% above onset, carried openly with the margin posture (onset is not failure) in the magnet chapter.
Shield honesty is magnet honesty: the most expensive components in the plant sit behind that stack. Publishing the failed recipe alongside the working one shows the optimization was real — and the 8%-above-onset flag shows the ledger wasn't tuned to flatter itself.
| Optimized shield | WC/W₂B₅ stack, min ≈1.20 m |
| Rejected recipe | B₄C-led: 6× over at 1.00 m (documented) |
| Spare build after shield | 0.42 m |
| Lifetime conductor fluence | 3.23e22 vs 3.0e22 onset (~8% above; flagged) |
| Method | multigroup transport (S36/S37), deposited |