2025MNRAS.544.3601B

Globular clusters host complex stellar populations whose chemical signatures suggest early ($3\times 10^6$-$10^9\, \mathrm{yr}$) retention and reprocessing of stellar ejecta, yet direct evidence for intracluster gas is lacking. Here we present a unified theoretical framework for the evolution of gas in young globular clusters, and its implications for the production of multiple stellar populations. We show that low-velocity AGB winds are gravitationally retained in clusters more massive than a few $\times 10^5\, \text{M}_\odot$. In addition, AGB winds in such clusters collide with each other and the previously retained winds, triggering a rapid 'switch' to efficient gas retention. Expected gas retention fractions agree well with the observed second population fractions in Galactic globular clusters. Furthermore, the accumulated gas cannot form new stars because protostellar cores are disrupted by encounters with pre-existing stars. Instead, the gas is accreted onto pre-existing main-sequence stars and compact objects. Time-dependent core-halo models indicate that compact objects can grow and collapse within $\sim 10^8$-$10^9\, \mathrm{yr}$, while lower-mass main-sequence stars can be 'rejuvenated' into the 4-$6\, \text{M}_\odot$ range required to reproduce key abundance patterns. Therefore, in our model, the multiple populations will be found in sufficiently massive clusters, with the second-population stars being formed from the inner subset of first-population stars that accreted large fractions of their mass from the AGB-processed retained gas. Finally, we argue that a combination of feedback processes, including accretion luminosity onto compact objects, novae, pulsar winds, and binary supernovae, will clear the gas by $\lesssim 10^9\, \mathrm{yr}$, thus reproducing the gas-poor conditions observed for present-day clusters.