Testing a radioactive clock for the Milky Way’s oldest stars

A clock with an uncertain starting point

Dating an ancient star requires a clock whose behaviour and starting conditions are understood. Thorium offers the first ingredient: its long-lived isotope thorium-232 decays with a half-life of about 14 billion years. Europium, another heavy element, is effectively stable over stellar lifetimes. Comparing their abundances can therefore indicate how much time has passed since the elements were produced.

The difficulty is knowing their initial proportion. Different starting amounts of thorium relative to europium would produce different inferred ages, even for identical measurements today. The method, called nucleocosmochronology, depends on both accurate observations and a defensible estimate of that initial ratio.

L. Magrini and colleagues examine these requirements in an arXiv preprint posted on November 5, 2026. The supplied manuscript also records acceptance by Astronomy & Astrophysics on that date. Its findings support using the clock for selected ancient stars, while documenting substantial limitations.

Measuring a faint chemical signature

The researchers measured thorium and europium in eight reference stars from the TITANS sample using archival spectra from UVES at the Very Large Telescope. These stars have carefully determined atmospheric properties, helping reduce uncertainty in the abundance analysis.

Thorium is especially difficult to measure. Its absorption feature near 4019 angstroms overlaps with features from other atoms and carbon-bearing molecules. The team modelled these neighbouring contributions before fitting the thorium signal. Europium measurements drew on up to four spectral lines, with corrections for departures from local thermodynamic equilibrium applied where suitable calculations were available.

Not every measured star was suitable for the subsequent analysis. A carbon-enhanced star was retained for comparison with earlier measurements but excluded from the main analysis. Another star, classified as unusually enriched in actinides—the element family containing thorium—was excluded from the individual age estimates.

Does the clock start consistently?

Combining their measurements with homogeneous literature samples, the authors investigated whether the thorium-to-europium ratio changes systematically with metallicity, here traced by iron abundance relative to the Sun.

Across the explored metal-poor range, approximately [Fe/H] = −3 to −1.3, they found no significant dependence after excluding stars with anomalous actinide abundances. This supports an approximately common initial production ratio for the selected stars within current measurement uncertainties. It does not establish that every heavy-element-producing event delivers the same ratio.

The exclusions matter: about 40% of the combined sample was removed from the production-ratio comparison under the adopted anomaly criteria. The observations also favour stars whose thorium lines are detectable, creating a selection bias toward stronger heavy-element signatures.

The authors additionally corrected observed ratios for radioactive decay using assumed representative ages. Agreement with the theoretical starting ratio is a consistency check, not an independent calibration: the correction already assumes how old the stars are.

Old ages, broad uncertainties

For dating, the study adopted an initial logarithmic thorium-to-europium ratio of −0.24. The six retained stars in the new sample received central age estimates ranging from 8.4 to 15.9 billion years. Their quoted observational uncertainties ranged from 3.2 to 12.8 billion years, illustrating how imprecise individual results remain.

Those error bars exclude uncertainty in the initial production ratio. The authors adopted a representative theoretical uncertainty of 0.10 dex—a logarithmic abundance unit—which adds a systematic uncertainty to the absolute age scale. Assuming one common starting ratio, changing that ratio shifts all inferred ages together rather than changing their relative differences.

A combined subsample restricted to 33 stars with inferred ages above eight billion years had a median age of 14.02 billion years, with quoted observational bounds of −3.22 and +1.87 billion years. The paper explicitly warns that this selected median is not an independent determination of the Universe’s age or the age of the oldest stellar populations.

Sharper spectra and shared birthdays

Simulations suggest that increasing spectral resolving power from roughly 40,000 to 80,000 can better separate thorium from neighbouring features. Typical abundance-ratio uncertainties could fall from about 0.10–0.15 dex to 0.04–0.05 dex, reducing associated observational age uncertainties from roughly five–seven billion years to two–three billion years. These are simulated improvements, not precision already achieved throughout the observed sample.

Globular clusters offer another opportunity: their stars can be treated as approximately coeval. Combining independent measurements could reduce random errors, while testing whether the clock behaves consistently among stars sharing a formation epoch. Shared systematic errors, including uncertainty in the initial ratio, would remain.

Proposed high-resolution multi-object instruments such as HRMOS could enable this approach. The immediate result is a clearer account of when the thorium–europium clock may work and what must improve—not a new precise age for the cosmos.