The Confinement Scaling Benchmark

Hyperion's energy-confinement estimate is compared to the ITER-basis H-mode scaling, with an explicit note that low aspect ratio pushes the fit beyond much of its source data.

Energy confinement is the crux

The scientific gain of the breeder (Hyperion), Q_sci 3.076, rests on how well the plasma retains energy — the energy confinement time. The community's standard empirical reference is the ITER-basis H-mode scaling, commonly written IPB98(y,2), fitted to a multi-machine database. The breeder analysis compares its predicted confinement time to that scaling as one of the 81 analyses.

# Illustrative form of the ITER-basis H-mode scaling (IPB98(y,2))
# tau_E ~ 0.0562 * Ip^0.93 * B^0.15 * P^-0.69 * n^0.41
#              * M^0.19 * R^1.97 * eps^0.58 * kappa^0.78
# Ip[MA], B[T], P[MW], n[1e19 m^-3], R[m], eps=a/R, kappa=elongation
# The breeder benchmark evaluates this at the design point and
# compares tau_E to the transport-model prediction used for Q_sci.

Where the design sits

At plasma current 9.66 MA and 8 T on-axis, the design point is evaluated against the scaling. The comparison is reported as a ratio of predicted to scaling-implied confinement, with the confidence interval of the scaling propagated through.

The aspect-ratio caveat, stated

IPB98(y,2) was fitted largely at conventional aspect ratio. The breeder runs at A 2.5, low aspect ratio, where spherical tokamaks are known to differ. The record therefore treats the conventional scaling as a cross-check, not as ground truth, and weights the ST-specific database more heavily for confinement inference.

This is where the confinement prediction is most exposed, and the record says so directly. A scaling extrapolated beyond its fitted aspect ratio can be optimistic or pessimistic, and only measurement resolves which. The breeder therefore pre-registers a confinement prediction with an explicit falsifier so the extrapolation is put to a decisive test at FOAK rather than defended by argument.

This page describes a design-and-simulation study, not a built machine. Construction of the breeder (Hyperion) begins Q2 2027; first-of-a-kind first tritium is targeted near 2030. No hardware net-gain claim is made before FOAK.