M1 physicist review: benchmark tolerance, pedestal, time step, ASCOT5 and the fallback model
M1 ends in a review by a physicist, the milestone’s human gate. Every M1 task was built and the other three acceptance criteria were met (entries 8, 11 and 12). Open for the review were the tolerance for the published benchmark, which the brief leaves to the physicist (OQ-5), and the choice of fallback transport model (OQ-1). The owner did the review as reviewer of record. The agent set out six items, each with its evidence and a recommendation. The owner accepted all six as written. The full list is on the decisions page.
What was decided
| item | decision |
|---|---|
| benchmark tolerance | against RAPTOR, profiles ≤ 2% (densities ≤ 0.1%), traces ≤ 5%C-061; fusion and Ohmic powers reported, not gated. Result: T_i 1.23% (1.1%), T_e 0.28% (0.71%), n_e 0.03% (0.03%), psi 0.49% (0.41%), q 0.69% (0.70%)C-050, transient 2.4% at dt = 0.05 s, 2.9% at dt = 0.00625 s (1.0% at t = 10 s)C-053; both pass |
| TORAX version | the change from the paper’s TORAX run (T_i 0.13%, T_e 0.50%, n_e 0.01%, psi 0.08%, q 0.16%; Ohmic power density 3.6%C-051) accepted as a code-version effect |
| pedestal | the pinning source removes 87–98%C-038 of the heating; accepted for core profiles, edge power not taken from these runs (K-010) |
| time step | the ITER hybrid example settings check the wrapper only; physics reference runs use a step converged in time to the limit on the decisions page |
| ASCOT5 | orbit tolerance 10-7; 640 alphas give totals (72.2 MW vs 70.5 MWC-059) and the electron/ion split; the radial shape waits for the queued production run; an unexplained excess in birth power (K-011) to be found before beam ions |
| fallback | TORAX’s Bohm/gyro-Bohm model; the critical-gradient model stays in the analytic test loop |
What changed in the code
The tolerance is now part of the benchmark: the limits sit in the benchmark module, a regression test fails if any gated quantity leaves them, and the benchmark command prints each quantity against its limit. Two known limits were added to the ceilings list: the pedestal pinning (K-010) and the alpha birth-power excess (K-011). The claims that the review judged are now marked reviewed in the claims register.
Where this stands: M1 has passed its acceptance criteria and its human gate. The ASCOT5 production run (OQ-13) continues as follow-up work and does not reopen the gate. M2 (training data) starts once the owner provides access to the turbulence code and its licence terms, the operating space to sample, and a compute allocation.
Technical details → Backends: the RAPTOR L-mode benchmark
Technical details → Backends › raptor benchmark
Decisions → OQ-1 · OQ-5