Globular clusters were long considered simple stellar populations: all stars born at the same time from a single gas cloud, so their chemical composition should be identical. But observations over the past decades have shown that almost every cluster contains at least two groups of stars that differ in light elements such as sodium, oxygen, aluminum, and magnesium. These groups are called the first and second generations. Meanwhile, in elements heavier than iron, such as barium or europium, clusters are usually homogeneous. Only a few metal-poor clusters, including M15 and M92, show a spread in the abundances of r-process elements, produced by rapid neutron capture. This work is part of a series of studies looking for a link between stellar generations and this spread: in M92 and NGC 2298, it was already found that the first generation is more heterogeneous in heavy elements than the second. The present paper tests this scenario on M15 with a larger sample of stars than before.
The authors used archival spectra of 89 stars in M15 obtained with the DEIMOS spectrograph on the Keck II telescope. The medium-resolution spectra cover the wavelength range from 3900 to 9100 angstroms. For each star, they constructed a synthetic spectrum using the ATLAS-9/SYNTHE package and fitted it to the observations in three stages. First, they determined temperature, surface gravity, overall metallicity, and radial velocity. Then they refined the carbon and oxygen abundances to account for their effect on the continuum. Finally, using a Markov chain Monte Carlo method, they measured the abundances of 18 elements, including magnesium, yttrium, barium, lanthanum, and europium. To separate stars into generations, they used k-means clustering on magnesium abundance: a threshold of [Mg/Fe] = 0.28 separates the first generation (high magnesium) from the second (low magnesium).
The spread in yttrium abundance does not differ statistically between the first and second generations. However, for barium, lanthanum, and europium, the variance in the first generation is significantly larger than in the second, with the difference reaching the 2σ level under different measurement error thresholds. For example, for barium, the standard deviation of [Ba/Fe] is 0.26 dex in the first generation versus 0.13 dex in the second (for errors < 0.1 dex). For europium, the difference is visible if one excludes a star with anomalously low Eu abundance, which may actually belong to the first generation. The derived variances are consistent with a picture where the gas that gave birth to the first generation was enriched by the r-process unevenly, and by the time the second generation formed, it had become well mixed.
The result adds M15 to the list of clusters where the first generation is more heterogeneous in r-process elements than the second, following M92 and NGC 2298. This strengthens the hypothesis that the generations formed at different times: enough time passed between the birth of the first and second generations for the gas to mix. Such a picture supports models with multiple episodes of star formation, rather than simultaneous formation followed by surface pollution of stars. Unlike neighboring studies that used photometry or small samples, this work applies spectroscopic analysis of 89 stars with detailed accounting for systematic effects from carbon and oxygen, making the conclusion about the difference in variances more robust.
The next step is to check whether the same pattern is seen in other metal-poor globular clusters with known r-process element spreads. If the pattern holds, it would become an important argument for the multiple-generation scenario. Also needed are models that explain why the r-process source managed to enrich only part of the first-generation gas and what type of event (magnetorotational supernova, collapsar, or something else) has the appropriate delay.
The result affects models of globular cluster formation and our understanding of r-process sources in the early Universe.
Conduct a similar analysis for other clusters with heavy-element dispersion and increase the sample of stars in M15 to refine the variances.
For globular cluster astrophysics, the central unresolved problem is the origin of multiple populations: why stars with different light-element abundances coexist in the same cluster. This work connects this problem to the chemical evolution of heavy elements and shows that the spread in r-process elements correlates with generations, which constrains possible formation scenarios.
This section is addressed to the author of the paper and their colleagues. Neighbouring works were found by meaning-based search across our archive, not by word overlap. We do not review: we show the field around the work.
What the work does
The authors measured the abundances of Mg, Y, Ba, La, and Eu in 89 stars of M15 using archival DEIMOS spectra and split the sample into first and second generations based on [Mg/Fe]. They found that the dispersion of Ba, La, and Eu abundances in the first generation is significantly higher than in the second, indicating incomplete gas mixing at early times and homogenization by the time the second generation formed.
Where its strength is
A strong point of this work is the direct link between stellar generations, identified by light elements, and the scatter of heavy r-process elements, which provides a timescale for gas mixing in the protocluster.
Where it could go next
Worth trying
Why this may matter
Understanding how quickly gas mixes in protoclusters is important for models of globular cluster formation and the origin of elements in the early Universe.
Nearby works in our archive
The archive is sparse around this work: the closest review is only 0.60, against a usual 0.64. It looks like few have drilled nearby.
The difference in barium dispersion between M15 generations reaches a factor of two: 0.26 versus 0.13 dex, which corresponds to a spread of 1.8 times in absolute abundance.