Owner limits the MAST-U certification scope
M4’s closed-loop check runs the coupled simulation once with the transport surrogate and once with TGLF, the turbulence code it was trained on, and compares the temperature profiles. The limit is 15%. ITER passes (served-face T_e 7.42%, T_i 8.18%, both passing the 15% limit; whole profile 7.72/9.23%; 34 of 51 final-state faces servedC-178) and MAST-U fails (served-face T_e 35.33%, T_i 49.15%, both failing 15%; whole profile 41.51/62.39%; cells with both adjacent faces served give 11.86/23.24%, still failing T_i; 13 of 26 final-state faces servedC-179). The MAST-U remediation retrained the surrogate. It then stopped at the step that needs a TGLF reference trajectory to compare the retrained candidate with. At some MAST-U states TGLF returns negative heat diffusivities, which the guard rejects, so those faces of the reference use the fallback transport model instead. Refining TGLF’s numerical settings did not converge those values (entries 59–62, last entry 62).
The projected reference
For the reference only, the owner chose TGLF at its standard settings with every negative heat diffusivity set to zero, each projection logged, and a stop if any projected heat flux reaches 1 gB (the gyro-Bohm flux unit) (OQ-38). The production surrogate and guard were unchanged. The single approved run stopped on that limit and saved no trajectory.
To see whether a rerun could pass, we applied the same rule, read-only, to every face of the 40 saved states of the earlier unprojected reference run, which has identical profiles:
| quantity | result |
|---|---|
| projections, and those reaching the stop | 150 negative raw heat diffusivities at 19 faces; 46 would reach the 1 gB stop, at 31 of 40 states, the first at state 1C-240 |
| of those, on faces the reference would serve | 26 would be served, the first at state 4 (faces 11 and 12, ion, −8.0 and −33.8 gB); near the axis the projected magnitude reaches 6020 gBC-241 |
| candidate against the guarded reference where both serve TGLF | T_e 32.0%, T_i 41.5% on the 14 faces where both serve TGLF, against the 15% limitC-242 |
The stop applies to every projection, served or not, and the first one to reach it comes at state 1 (first table row). So the rule cannot complete on this scenario, and no rerun can produce the reference. Even counting only served faces, it would stop early. The screen is not a replay: its fluxes are inferred by inverting TORAX’s conversion, and a projected run would leave the saved trajectory after its first served projection. The last row compares against the guarded TGLF/fallback hybrid, so it describes the distance and is not a gate result.
Decision
The owner chose the recommended option of OQ-42, limited MAST-U scope: - the MAST-U closed-loop comparison is recorded as not evaluable; - MAST-U is certified at the M4 review with that limitation stated, with ceilings K-030 and K-032; - ITER’s measured pass stands; - the model, the guard, the temperature and flux acceptance limits and the frozen data splits are unchanged; - no further MAST-U reference compute is planned.
The other two options were to redefine the comparison against the guarded hybrid, which the last row suggests would still fail the limit, or to raise the stop threshold, rejected because projected magnitudes reach thousands of gB.
Where this stands: M4 moves on to the formal proofs, calibration, the EC closed-loop comparison, certificate-bound guards and the per-surrogate reports, then the certification review.
Technical details → verification. Decisions → OQ-38, OQ-42.
Technical details → Verification › mastu scope
Decisions → OQ-38 · OQ-42