# Matter Harmony Lab 1.5.0 — Frozen hydrogen-isotope holdout

## Result

The finite-range proxy was made predictive rather than refitted. Its range was
frozen at the electron reduced Compton wavelength,
`386.15926796 fm`, and its coupling was fitted only to release 1.4.0's
H--D 1S--2S central residual. The resulting prediction for the separately
measured H--D 2S--8S isotope shift is `-201.996425 Hz`.

The historical transition measurements give

`770859041245.7(6.9) - 770649350012.0(8.6) = 209691233.7 kHz`.

Using the published experimental correlation `0.1225` gives a `10.345613 kHz`
measurement uncertainty. The Pohl et al. bound-state parameterization, finite
size terms, and correlated additive remainders give an ordinary prediction of
`209691226.436128 kHz`. Propagating the radius and remainder covariance raises
the combined uncertainty to `10.367652 kHz`.

| Prediction | Signed error | Signed z |
|---|---:|---:|
| Ordinary reference | -7.263872 kHz | -0.700628 |
| Ordinary + frozen operator | -7.465869 kHz | -0.720112 |

The operator therefore moves the central prediction slightly farther from the
measurement. Its entire holdout effect is only `0.019484` combined standard
deviations; the uncertainty is `51.326` times larger.

## Decision

The frozen operator earns no predictive credit because it does not improve the
absolute holdout residual. It is also not excluded: this target cannot resolve
the two models at anything close to three sigma. Non-rejection is preserved as
ambiguity, not converted into confirmation.

This is a historical published-data holdout, not a blind prospective
experiment. The range is a named physical scale but was not derived from a
fifth-dimensional action. The calculation is written conditional mathematics
plus finite computation using published measured/theory inputs; it is not new
spectroscopy, proof of a physical extra dimension, a universal grid/tick, or a
theological identification.

## Independent verification and next test

An audit that imports neither production isotope module recomputes the Yukawa
matrix elements by direct radial quadrature and independently replays the
covariance. It reproduces the material prediction.

The next useful target must have a predicted operator contrast substantially
larger than its full correlated uncertainty. Rank candidate transitions by
that pre-measurement resolution ratio, freeze one target and ordinary-theory
covariance, then score it once without refitting.