This paper reports new measurements of thorium, one of the heaviest elements we can observe, in 47 metal-poor stars. The rapid neutron-capture process (r-process) makes thorium and other heavy elements. By studying thorium and its ratio to lanthanide elements like europium, the authors test how similar or variable that production is from one r-process event to another.

The team used high-resolution optical spectra collected on several telescopes between 2016 and 2024. They limited the sample to stars with good data quality (signal-to-noise ratio above about 30 near 4000 Å) and report only thorium detections at or above the 3σ level. The spectra have resolving power R between about 40,000 and 80,000. The 47 stars come from the R-Process Alliance surveys and selected literature sources and include stars in the Milky Way halo, dwarf galaxies, and some clusters.

The key findings are quantitative. The spread (dispersion) in thorium abundance, measured as [Th/H] and [Th/Fe] in logarithmic units (dex), falls from roughly 0.6 dex at the lowest metallicities to about 0.2 dex at higher metallicities. Thorium tracks the lanthanides europium (Eu) and dysprosium (Dy) very closely: on average [Th/Eu] is about 0.0 across stars with metallicities −3.0 ≲ [Fe/H] ≲ −1.5 and across a wide range of europium enrichment. Still, the absolute range of log ε(Th/Eu) in the sample is 1.02 dex, with an observed standard deviation of ±0.20 dex and an estimated intrinsic spread of ±0.11 dex at the lowest metallicities. From this they infer that about 68% of r-process events produce Th/Eu yields that vary only by ±30%, while about 5% of events produce yields that differ by factors larger than ∼3 and up to ∼10.

Why this matters: thorium and other actinides probe the most extreme conditions in r-process sites, because they require very neutron-rich material to form. The small intrinsic scatter for most events implies that many r-process occurrences produce a robust heavy-element mix. At the same time, the existence of rare, large outliers (so-called “actinide-boost” and “actinide-deficient” stars known from earlier work) shows that some events produce very different actinide-to-lanthanide ratios. These results give concrete, numerical constraints that nuclear and astrophysical models must meet when they try to reproduce r-process yields.