Globular Clusters: Chemical Memory of the First Molecular Clouds
Using cosmological radiation-hydrodynamic simulations with a chemical enrichment model, a population of giant molecular clouds was identified whose internal chemical patterns reproduce the key features of globular clusters: a wide spread in light element abundances and an anti-correlation of nitrogen and oxygen with nearly constant iron abundance. These clouds form when star formation reignites after a previous burst, as previously ejected oxygen-rich gas collides with nitrogen-bearing galactic gas. In such regions, dense star clusters are born. Thus, the chemical inhomogeneity of globular clusters does not require a long, stepwise self-enrichment inside the cluster—it is directly inherited from the structured interstellar medium shaped by the baryon cycle. Globular clusters turn out to be a sort of chronicle of enrichment processes and gas flows in galaxies at high redshifts.
Context
Globular clusters are true fossil remnants of the Big Bang: their age often exceeds 10 billion years. Since the work of Edwin Hubble, they have served as important indicators of galaxy evolution. Thanks to spectroscopy, pioneered by Cecilia Payne-Gaposchkin, we know that stars are mostly made of hydrogen. However, the chemical composition of globular clusters has long puzzled scientists. The main mystery is the large spreads and anti-correlations of light element abundances, such as nitrogen and oxygen, with nearly constant iron abundance. Traditional self-pollution models required that first-generation polluting stars somehow selectively enriched the gas for the second generation, but faced a catastrophic mass shortage.
Methods
The authors used spectroscopic data from the THESAN-ZOOM simulations—a complex cosmological model with radiation transfer, tracking enrichment by nine elements. These numerical experiments, like a virtual time machine, allow us to peer into the early Universe. In them, scientists identified giant molecular clouds whose chemical composition resembles globular clusters. Selection criteria included a large dispersion in the N/O ratio (over 0.5 dex) and an [N/Fe]–[O/Fe] anti-correlation. Using machine learning methods, a key factor was highlighted: the most important was not the local properties of the cloud, but the star formation history of the galaxy—especially a burst 20–50 million years ago. Remarkably, the calculations explicitly included constraints imposed by the speed of light, which is particularly important for correctly modeling ionizing radiation.
Results
The simulations revealed a whole population of clouds satisfying the conditions. Their masses are systematically higher than those of ordinary molecular clouds, and metallicities lower—just like real globular clusters. The formation mechanism turned out to be unexpectedly elegant. In a typical dwarf galaxy, after a starburst, supernova explosions eject oxygen-rich gas into the halo. Then, when the galaxy temporarily 'dies', old stars on the asymptotic giant branch enrich the remaining gas with nitrogen. Later, the oxygen-rich gas cools and falls back, colliding with the nitrogen-rich medium. This collision creates a compression zone—almost like accretion in black holes, but on the scale of an entire cloud. As a result, the future cluster inherits ready-made anti-correlations without waiting millions of years.
Implications
The scenario solves the long-standing mass deficit problem: there is no need for the first generation of stars to be tens of times more massive than the second. Instead, all stars form simultaneously from already mixed gas. Moreover, the model does not require preferential loss of stars with normal chemical composition, which agrees better with observations. These conclusions would not have been possible without long-term observations by Hubble, which studied globular clusters in our Galaxy for decades.
Future development
In the coming years, modeling will become even more precise: future simulations will be able to track the evolution of dense star clusters over cosmological timescales. It is important to include sodium production in the calculations to directly verify the Na–O anti-correlation. Perhaps gravitational waves will also contribute: next-generation detectors will be able to register mergers of compact objects in cluster cores, providing independent information about their dynamics.
Impact
The work affects the interpretation of data from JWST, revealing the connection between young clusters and nitrogen pollution in galaxies. It also emphasizes the importance of matter recycling in the early Universe.
Next steps
The researchers plan to test key predictions: very short cluster formation times (less than 3–5 million years) and the presence of nitrogen anomalies even in unformed protoclusters. This will require observations of gravitationally lensed systems and ultraviolet spectroscopy.
Key open problems
This result links the chemistry of globular clusters with the nuances of the baryon cycle and even with the distribution of dark matter, which was first discovered by Vera Rubin through galaxy rotation curves. Moreover, it may help solve the missing satellites problem and the overcooling problem, important for modern cosmology.
🎯 Globular clusters are so old that they contain virtually no elements heavier than helium born in previous generations of stars—this makes them true 'time capsules' from the reionization epoch.
Key numbers
- median mass difference: ~0.5 dex
- age of oldest clusters: >10 billion years
- maximum simulation resolution: 142 M⊙
- allowed metallicity dispersion: σ([Fe/H]) <0.1 dex
- threshold surface density: ~700 M⊙/pc^2