KITP: more multiple populations

In the Good Old Days, Globular Clusters were simple things: spherical, relaxed, coeval and homogenous.
Not so much anymore.
The thorny issue of multiple populations keeps coming up at the workshop.
They're there, something is going on, but what, and how, and Y?

M13 embiggen

Guest author
Natalie Hinkel from ASU put some notes together from the sessions I missed earlier this month, and I post them below, with her permission, with light edits and addition of figures and links by me - any error is mine:

See, here is Good Old M54 (from Chaboyer) - simple.

Multiple Populations: A Summary

Looking at the CMD for any globular cluster, it is usually taken for granted that there is a main sequence (MS), subgiant branch (SGB), red giant branch (RGB), or horizontal branch (HB). However, in a number of clusters (namely NGC 2808, 1851, 6388, 6441, M54, and ω Cen), multiples of these branches have been recently observed. A debate has begun in the astronomy community to determine the causation of these less-than-normal results. And, despite a plethora of scientific papers, computational models, and arguments, there does not seem to be one solution that satisfies the details of this situation. Also, because it is apparent that we do not fully understand what is happening, instead of coining the more common phrase for "multiple populations," it seems more appropriate to say "multiple branches" in these GCs, since that observation can't be denied.

This analysis follows along with, and is a supplement, to Renzini 2008 Renzinie 2008, but remains as unbiased as possible when considering the various models available.

Observational Constraints

Assumption: All observed sources are a result of the same event. Since we are seeing multiple branches, ignoring that sometimes the GC exhibits many MSs, HBs, etc., we take it that the same mechanism is at work. With that in mind, below is a list of the trends noticed in the literature - excluding ω Cen since it is more than just a GC.

1) Only massive GCs (~.5-5*10^6 M_sun) have multiple branches, but not all massive GCs show them.
a) Some exhibit multiple branches in the MS
b) Some in RGB, SGC, or HB
c) Some as a combination of the above.

2) The branches can only be explained in terms of high helium content, compare to that typical of a GC: Y ~ 0.23. Here the abundances are measured indirectly to be Y = [0.28 - 0.40] and possibly as high as Y ~ 0.8 (C. Charbonnel while at KITP).
a) The branches are discrete (Comment: wouldn't the NaO anti-correlation be continuous?).
b) If the branches are thought of as distinct populations, the time required between the formation of each is ~10^7-10^9 yrs (Comment: age is hard to constrain and narrow turnoffs in some of the GCs, particularly NGC 2808, would lend itself away from separate starbursts - see the end of "Reevaluating Assumptions").
c) Again, if considering separate populations, the mass of the 1st generation must be consistent with the enrichment and mass required in the 2nd generation - taking into account that the star formation efficiency will be less than one.
d) Stars with high He content would give rise to a different kind of stellar remnant, which also needs to be explained in this context.

3) Chemical constraints have been determined either directly from observations or indirectly through correlations, including the rise in hydrogen burning products - particularly He.
a) NaO, AlO, AlMg are anticorrelated. CN, CH are bimodal.
b) no general increase in CNO - only an increase in N.
c) These trends are also seen on a smaller global scale in most globular clusters.

Physical Explanations/Models (in rough cosmological order)

1) Pop III stars - This model proposes that some Pop III and Pop II stars were formed at the same time such that the Pop II stars were enriched by the pair-instability SN ejecta from the massive stars. See Choi & Yi '07
Pro: The amount of helium produced in this scenario matches the high He observations.
Con: Dilution of other metals occurs at the same time, which does not match the constraints. This is also a difficult scenario for more than two branches.

Lemma: Gravitationally induced segregation of He to the center of a dark matter halo could pre-enrich Pop III stars, could cause a general rise in He for the entire cluster. However, +2 generations are still a problem. See Chuzhoy '06

2) Nucleated Dwarfs - Possibly for the case of M54 (although there seems to be some recent results contradicting this: Bellazzini et al. '08), dwarf galaxies whose first large (x10 times) generation of stars were tidally torn apart [ed. I don't follow] would leave AGB ejecta trapped in a dark matter potential well. This enables the 2nd generation of stars to be enriched by H-burning products.
Pro: The amount of enriched material for the 2nd generation is not constrained by the total number of 1st generation stars currently observed.
Con: Tidal stripping has to occur at precisely the right time in order to avoid extracting too many 2nd generation stars. After the AGB ejecta has sunk to the center of the cluster, a star burst needs to be triggered - somehow. Iron does not match with observations, which are particularly difficult in this case (M54). Some evidence of the host needs to be traceable nearby.

3) Accretion Onto MS Seeds - Here the ejecta from the larger 1st generation is thrown onto smaller 1st generation stars who have not yet turned off of the MS, increasing certain aspects of their metallicity.
Pro: The amount of mass needed to create a 2nd generation of enriched stars is not a problem, since they are a fraction of the 1st generation. There is a good match to the multiplicity in He abundances.
Con: A. Sills and F. Rasio are not proponents of this model at all. Accretion would lead to a spread in He, not distinct populations, since this mechanism depends on the mass and velocity of the ISM and the orbit of the seed star. This was said to no longer match observations of all other chemical anomalies.

4) Low Mass RGB Mixing - It was proposed that stars with mass Pro: The CNO anticorrelation and high He abundances are reproduced well.
Con: This process only affects the HB and rules out any MS branches. Post-RGB enrichment and subsequent accumulation would take longer than 1Gyr, which is greater than that observed. The remnant stars from these mixed stars would be mostly carbon. A typical number of low-mass RGB stars ejecting this material is not enough to satisfy the population (~50% M_tot) of an enriched 2nd generation.

5) AGB pollution - Similar to the nucleated dwarf scenario, a 1st massive population of stars is formed. As these stars turn off the MS, AGBs (~3-8 M_sun) are thought to eject a large amount of low velocity material that is high in H-burning products, due to hot bottom burning and second dredge up. At some point, the majority of the 1st generation are torn off via stripping or cluster expansion (90-99% M_tot), leaving the 2nd generation of stars to be formed out of the remnant gas.
Pro: Any mass differential between populations is well explained, so there is little constraint on the amount of gas available to enrich. All chemical signatures (high He and N, the anticorrelations) are well reproduced. Because AGB winds are slow, there is little worry that this matter will escape the potential well of the GC or be stripped off. A similar number of stars in the 1st and 2nd generation is replicated. See D'Ercole et al '08
Con: While this model is clearly a major contender in the current debate and reproduces the chemical trends excellently, I wanted it to be clear that there are still some large issues that need to be resolved. To match the chemistry needed, a peculiar and irregular IMF was used in the initial formation of the 1st generation. Super-AGB stars are required in this scenario to reach Y ~0.40, but He abundances of Y > 0.40 are not reproducable. It was assumed that only the most massive AGBs of the 1st generation were contributing to the ISM (possible mass issues here?), with a large enough population for a 2nd generation. So, then, these stars would have a short enough lifetime to exclude any third dredge up elements (such as s-process, which are not seen in any high abundances here...although others argue that third dredge up must occur more in this situation). However, these models seem to be extraordinarily model dependent such that slightly different prescriptions done by others lead to drastically different results. This, then, begs the question that if different routines describing the same process are enough to clearly change the results of the overall model, then either the model is too sensitive, the adopted prescriptions are wrong, or the "knobs" are too easily manipulated. Also see Bekki & Norris '06 and Karakas et al '06
Caveat: In this scenario, due to overall ISM pollution by the 2nd generation, the 3rd generation (for example in NGC 2808) would actually be the middle branch.

6) Fast Massive Rotating Stars - If stars on the order of 20-120 M_sun are able to reach a critical velocity (where centrifugal acceleration balances gravity) on their surface, that velocity will be maintained for the duration of the star's life - which is short because of the large mass. The fast rotations will then form an equatorial disk around the star, expelling the H-burning products that have been mixed from lower zones up to the surface by the rotation speed. See Decressin et al '07
Pro: The outputted products from these massive stars match the H-burning products, NaO & AlMg anticorrelations, and CNO trends observed in GCs. The heavier winds that would change the chemistry of the 2nd generation, especially those from O-type stars, Wolf-Rayets, and SN ejecta, would be expelled out of the GC potential well by their high velocities.
Con: Again, this is another major candidate theory for the multiple branch observations, however, there are some serious issues yet to be resolved. There are no models for incorporating the ejected gas of the 1st generation into the 2nd generation or if there would be enough mass. This model poorly describes multiple MS branches. The equatorial disk of the massive stars is extremely fragile and would be destroyed by a SN, if not nearby strong winds, thus requiring a short lifetime. This would lead to a quick conversion efficiency and more problems with the mass amount ejected by the rotating stars not consistent with the 2nd generation mass. Rotation at the critical velocity suggested for all massive stars is difficult to maintain in a physically rigorous setting.

7) External Pollution - This is the newest theory presented to explain the multiple branches phenomena. The model looks specifically at a "peculiar" region in the galactic halo, before the formation of the GC, that has been pre-enriched from an early SNe II and contains ~4-7 M_sun AGB stars. As the AGBs begin polluting, one or a few nearby SN Ia explodes. The enriched material is swept up in the SN expansion and is condensed into one region as the gas collapses. When the GC forms the stars are born out of this unique chemistry and the chemically anomalous stars are born first. With pollution from low metallicity SNe IIs and general chemical evolution, the metallicity of the population moves to that of more "normal" stars. This reverses the order of the stellar generations seen in the above models. See Marcolini et al '09, especially Figure 1.
Pro: The 2nd generation of stars is not mass limited by the 1st generation, nor is there an implementation of a peaked IMF or required 90-99% M_tot loss. Current prescriptions are not highly affected by the lack of or significant presence of third dredge up AGB stars. Multiple branches in the MS, SGB, RGB, and HB are all reconstructed with similar population fractions. The NaO, AlMg, and CNO trends are recreated.
Con: While some of their chemistry has been matched to observations, it has been artificially manipulated to correlate. For example, AGBs were assumed to produce more Al and four times less carbon than typically theorized - making it such that the 12C/13C ratios weren't reproduced. The abundances seen in helium were also not produced. Additionally, they specified that the GCs were ten times more metal rich than coeval halo stars. It is also not certain if this technique will yield distinct CMD branches, or merely a spread.

Reevaluating Assumptions
Initially some assumptions were made regarding this analysis. However, those premises, too, need to be questioned.

• Are all of these multi-branch populations as a result of the same event? Or should they be split into sub-categories according to which branch is observed in multiples, since clearly different stellar processes are affected at different times in the life of a star?
• Helium cannot be directly measured, so indirect correlations lead to the observed abundances. Could other factors be affecting these correlations - making it appear that He is the main divergence between branches?
• If there are truly multiple stellar populations, with a finite time between the birth of each, how are narrow TOs defined? While at KITP, Alison Sills produced a plot showing how the stellar evolutionary track of stars with varying masses and metallicities appear on a CMD. The main result of this plot was to show that the TOs of these stars, even with extremely varying He abundances, are similar. This is particularly relevant to NGC 2808. But is this true in all the GC observed?

click to embiggen

It's a standard HR diagram (log L/Lsun vs log Teff). The solid line is
for a normal 0.8 solar mass star with Z=0.001 ('typical' globular
cluster metallicity) and Y=0.23. It has a turnoff age of 13 Gyr. The
dotted line is for the same mass and metallicity, but Y=0.4. It has a
turnoff age of about 4 Gyr. The dashed line is a 0.6 solar mass star,
same metallicity but Y=0.4; its turnoff age is 12.7 Gyr.

GC Specifics
Below are small annotated lists describing the various GCs that have been observed and the characteristics that have been observed or determined in the literature.

• NGC 2808
  3 HBs, 3 MSs
  Y = 0.25, 0.30, 0.37
  delta t = 10^8 - 10^9 yr between branch formation
  very little iron differences between branches, extremely narrow turn off
  Piotto et al. '07, '08
• NGC 1851
  2 SGB
  Y = 0.25 - 0.253 (fit to isochrones)
  delta t = 10^7-18^8 yr
  one branch has normal alpha-enhancement, the other branch has higher CNO abundances, CNO-Na anticorrelation, and is s-process enriched.
  Zocallini et al. 2009, Renzini '08
• M54
  +3 TOs, 2 RGBs
  Y delta t ~ 2 Gyr
  [Fe/H] = -[0.4, 0.6]
  Siegel et al. '07, Bellazzini et al. '08
• NGC 6388
  extended HB hook, multiple SGBs
  Y ~ 0.35, .38-.4
  delta t = 0
  very strong NaO anticorrelation, twin GC to NGC 6441, lots of contamination in the observations
  Caloi & D'Antona '07, Busso et al. '07
• NGC 6441
  extended HB hook, multiple SGBs
  0.27 delta t = 0
  Busso et al. '07

N. Hinkel (natalie.hinkelATgmail.youknowhat), ASU