How Uncertainty Is Quantified

Breeder results are reported with uncertainties, not as point values; this page describes how input uncertainty is propagated to every headline number.

A number without an uncertainty is incomplete

The breeder (Hyperion) headline figures — Q_sci 3.076, 85.0 MW, and the product yields — are the central estimates of distributions, not exact truths. Each rests on inputs that carry uncertainty: confinement scaling coefficients, profile assumptions, cross-sections, and geometry. Uncertainty quantification (UQ) propagates those input uncertainties through the analyses to bracket the outputs.

# Monte-Carlo style propagation (schematic)
import numpy as np
def propagate(model, priors, N=10000):
    samples = [model(**draw(priors)) for _ in range(N)]
    return np.mean(samples), np.percentile(samples,[5,95])
# Reported: central value + a stated confidence interval,
# not a bare point estimate.

What is propagated

• Scaling-law coefficient uncertainty from the empirical fits.
• Profile and boundary assumptions varied across plausible ranges.
• Nuclear-data uncertainty in the neutronics and breeding analyses.
• Geometry tolerances where they affect coverage.

Honest intervals

Where an interval is wide, the record says so rather than quoting only the optimistic edge. The TBR lever at 1.1, 1.5, and 1.8 is itself a UQ posture: the record spans the range instead of committing to a single breeding number, because the underlying uncertainty is real. Sensitivity — which inputs drive the spread — is covered in the sensitivity analysis.

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.