MAST normalised-beta authority¶
validation/mast_normalised_beta_authority.py is the normalised-beta admission gate
for the canonical disruption channel beta_N. It diagnoses the public
FAIR-MAST EFM_BETAN source without guessing a scale, changing its values, or
promoting a metadata correction into physical authority.
Definition and metadata boundary¶
The pinned IMAS Data Dictionary defines beta_tor_norm as
with unit 1. The pinned FAIR-MAST ingestion mapping gives the same definition
for EFM_BETAN, but targets the older IMAS leaf beta_tor_normal. The live
Level-2 array has source unit T. These facts identify a metadata conflict;
they do not define a physical T-to-1 conversion.
The only non-destructive candidate is therefore to preserve the source values
and repair metadata after authority is complete. The gate sets
numeric_transform to null, forbids range-based scale inference, and never
clamps or discards negative samples.
The source mapping is pinned to the FAIR-MAST ingestion repository. The canonical leaf, definition, and unit are pinned to the IMAS Data Dictionary.
Required authority¶
Canonical binding requires either primary-source authority for the direct array or an independently reproducible formula route. The reproduction route must bind, on one source-attested alignment:
equilibrium.beta_tor, including whether values are fractional or percent;equilibrium.minor_radius, in m;equilibrium.bvac_rgeom, the exact vacuum field used by EFM at the geometric axis;equilibrium.plasma_current_x, including its sign or magnitude convention;equilibrium.time;- an equilibrium validity or reconstruction-quality rule; and
- one-standard-deviation uncertainties plus the covariance policy.
Names, approximate magnitudes, positive majorities, or dimensional cancellation are not substitutes for those declarations.
Real two-shot result¶
The 2026-07-23 bounded run acquired a fresh SourceObjectManifest-v2 snapshot
for MAST shots 30421 and 30424 under the external Samsung datasets tree. It
retains minor_radius without resampling and is pinned to source generations
44d2c93f...a32868 and 1da4ee7c...f6b0. The complete manifest has payload
digest aba5e048...d196a and records 14,699,129 derived bytes.
The direct beta_N evidence contains 107 finite samples per shot. Shot 30421
has 3 negative and 104 positive finite samples, with an observed range of
-13.4791 to 3.9432. Shot 30424 has 4 negative and 103 positive finite samples,
with an observed range of -5.8219 to 3.8770. These values are observations,
not scale or validity authority.
beta_tor, minor_radius, and the equilibrium timebase are present. The exact
bvac_rgeom and fitted plasma_current_x inputs are absent. Both shots remain
blocked for that missing formula lineage, the live source-unit and stale-leaf
conflicts, unresolved sign/scale and negative-value validity, absent
reconstruction-quality authority, and absent one-standard-deviation
uncertainty. Canonical binding, training, scientific validation, facility
prediction, and control admission all remain false.
The self-digested report is retained as non-public operational evidence. Its
payload digest is 6c8e8b4e...c9706a1; its file SHA-256 is
5dc6c759...a425e.
Reproduction¶
Run the gate only against a verified SourceObjectManifest-v2 root:
python -m validation.mast_normalised_beta_authority \
--manifest /external/source-v4/source_object_manifest.json \
--artifact-root /external/source-v4 \
--shot-id 30421 --shot-id 30424 \
--json-out /external/l2f12b-authority.json
The report is deterministic and self-digested. A source-member, metadata,
shape, time-order, formula-lineage, quality, sign/scale, negative-value, or
uncertainty failure keeps canonical_beta_n_binding_admissible false.