Most tokamaks point their plasma cross-section toward the center column. MetroVolt points it the other way — δ = −0.30 — and that one sign flip is the quietest big decision in the design.
Conventional 'positive-triangularity' plasmas earn their confinement in H-mode, a regime whose edge periodically erupts in edge-localized modes (ELMs) — millisecond heat pulses that erode any wall built by humans. Negative-triangularity (NT) plasmas, demonstrated on the DIII-D and TCV tokamaks, hold near-H-mode confinement without that eruptive edge: the magnetic geometry suppresses the instability drive at the boundary while turbulence in the core actually improves.
MetroVolt freezes δ = −0.30 and verifies it the hard way: a free-boundary Grad–Shafranov equilibrium (FreeGS) computes the real shaped plasma at full current, returning a safety factor q95(eq) of 5.24 at the operating 43 MA and 6.37 T (KX-8D; the 6.0 T-era benchmark gave 4.94, which had forced a trim to 42.5 MA) — the q95 ≥ 5.0 floor is met at full current.
An ELM-free edge is not a luxury; it is what lets a first wall survive contact with a commercial duty cycle. Combined with the deliberately low heat load of the D-³He fuel cycle (0.074–0.089 MW/m² at the Mode-D operating ledger), the NT choice converts 'plasma-facing components' from a consumable into plant infrastructure.
Our own deposited δ-scan (S77) even reports the adverse trend honestly: pushing the triangle deeper costs safety-factor margin. The frozen −0.30 sits at the favourable end of the trade — chosen, tested, and published, not asserted.
| Triangularity δ | −0.30 (frozen) |
| Edge regime | ELM-suppressed (NT) |
| q95(eq) at the operating point | 5.24 (KX-8D, FreeGS) |
| Operating floor | q95 ≥ 5.0 — met at the full 43.0 MA |
| Shape verification | free-boundary equilibrium, deposited |