Stellar collisions in globular clusters: the origin of multiple stellar populations

Two generations of stars, G1 and G2, typically populate Galactic globular clusters (GCs). The origin of G2 stars is unclear. We uncover two empirical dependencies between GC characteristics, which can be explained by the formation of G2 Main-Sequence (MS) stars due to collision\merging of their primordial counterparts (G1). A similar genesis of both G2 stars and peculiar objects like LMXBs and millisecond pulsars is also implied. Indeed, we find a significant (at a confidence level > 99,9%) anti-correlation between the fraction of G1 red giants ($N_{G1}/N_{tot}$) and stellar encounter rates among 51 GCs. Moreover, a Milky Way-like initial mass function (IMF) requires at least ~50% of MS stars located in the mass range [0.1-0.5] $M\odot$. Unlike cluster mass loss, stellar collisions\merging retain these G1 stars by converting them into more massive G2 ones, with mainly M_ms > 0.5 $M\odot$. This process coupled with a decreasing relative mass loss with increasing GC masses implies a smaller ($N_{G1}/N_{tot}$) in more massive GCs with a shallower present day MF. From data for 35 GCs, we find that such an anti-correlation is significant at 98.3% confidence level (Spearman's correlation) for 12 most massive GCs ($M_{GC} > 10^{5.3}$) and it is at a confidence level of 89% for the 12 least massive GCs ($M_{GC} < 10^{5.1} M_\odot$). Other fractions of G1 and G2 stars observed at the bottom of the MS as compared with the red giant branch in a few GCs are consistent with the scenario proposed.