UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Radial distributions of various stellar populations and the evolution of the globular cluster 47 Tucanae Parada Torres, Javiera Fernanda 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata


24-ubc_2016_february_paradatorres_javiera.pdf [ 1.9MB ]
JSON: 24-1.0223163.json
JSON-LD: 24-1.0223163-ld.json
RDF/XML (Pretty): 24-1.0223163-rdf.xml
RDF/JSON: 24-1.0223163-rdf.json
Turtle: 24-1.0223163-turtle.txt
N-Triples: 24-1.0223163-rdf-ntriples.txt
Original Record: 24-1.0223163-source.json
Full Text

Full Text

Radial Distributions of Various Stellar Populations and theEvolution of the Globular Cluster 47 TucanaebyJaviera Fernanda Parada TorresA THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCEinTHE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES(Astronomy)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)January 2016c© Javiera Fernanda Parada Torres, 2016AbstractBlue stragglers (BSS) are stars whose position in the Color-Magnitude Diagram (CMD) placesthem above the main sequence turn-off point in a given cluster. Three possible origins havebeen proposed: stellar collisions, evolution of binary systems, and evolution of hierarchicaltriples. Using data from the core of 47 Tuc in the ultraviolet (UV), we have identified variousstellar populations in the CMD, and used their radial distributions to study the evolution andorigin of BSS. When we separate the BSS in two samples divided by their magnitude, we findthat the bright BSS show a much more centrally concentrated radial distribution and highermass estimates, suggesting an origin involving triple or multiple stellar systems. In contrast,the faint BSS are less concentrated, with a radial distribution similar to the main sequence (MS)binaries pointing to this populations as their progenitors. A sample of evolved BSS was foundon the UV CMD, this put together with available photometric data and MESA evolutionarymodels resulted in time scales and number of observed and expected stars agreeing nicely withthe BSS having a post-MS evolution comparable to that of a normal star of the same massand a MS BSS lifetime of about 200-300 Myr. We also find that the extra population of theasymptotic giant branch (AGB) stars in 47 Tuc is due to evolved BSS, with the bulk of thecontamination being in the red giant branch bump of the BSS that, according to our models,falls in the same magnitude and color range as the observed AGB bump.iiPrefaceThe data reduction, briefly explained in Chapter 2, that led to the photometry files in the ul-traviolet filters used in this thesis, was carried out by Jason Kalirai following the proceduresdescribed in Kalirai et al. (2012). For the ACS data, the photometry file is of public domainand can be obtain at http://www.astro.ufl.edu/ata/publichstgc/databases.html.The same evolutionary models and isochrones presented in this thesis were used in Heylet al. (2015a) of which I am a co-author. The completeness rates have also been previouslypublished in Heyl et al. (2015b) of which I am also a co-author.Everything bother than the above is an original, unpublished, independent work by theauthor, J. Parada.iiiTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiGlossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Colour-Magnitude Diagrams and Stellar Evolution . . . . . . . . . . . . . . . 21.1.1 The Evolution of a Solar Mass Star . . . . . . . . . . . . . . . . . . . 21.1.2 Implications for Globular Clusters . . . . . . . . . . . . . . . . . . . . 51.2 Dynamical Evolution of Globular Clusters . . . . . . . . . . . . . . . . . . . . 51.3 Blue Stragglers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71.3.1 Evolution of Primordial Binaries . . . . . . . . . . . . . . . . . . . . . 81.3.2 Direct Stellar Collisions . . . . . . . . . . . . . . . . . . . . . . . . . 101.3.3 Dynamical Evolution of Hierarchical Triple Systems . . . . . . . . . . 111.3.4 Linking Models to Observations . . . . . . . . . . . . . . . . . . . . . 111.3.5 Blue Stragglers and 47 Tucanae . . . . . . . . . . . . . . . . . . . . . 132 Observations and Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 152.1 Observations and Photometry . . . . . . . . . . . . . . . . . . . . . . . . . . . 152.2 Artificial Star Tests: Correcting for Incompleteness . . . . . . . . . . . . . . . 162.3 The ACS Data Set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18iv3 Stellar Population Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193.1 Main Sequence Binaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193.2 Blue Stragglers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203.3 Reference Population . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213.4 ACS Data Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224 Estimating Masses Outside the MS . . . . . . . . . . . . . . . . . . . . . . . . . . 245 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315.1 Blue Stragglers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315.2 Evolved Blue Stragglers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396.1 Blue Stragglers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396.2 Evolved Blue Stragglers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44vList of TablesTable 4.1 KS-test results between the populations selected on Figure 4.4 . . . . . . . . 28Table 4.2 Estimated mass values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30Table 5.1 KS-test results between the populations selected on Figure 5.1 . . . . . . . . 31Table 5.2 KS-test results between the populations selected on Figure 5.2 . . . . . . . . 35Table 5.3 KS-test results between the populations selected on Figure 5.3 . . . . . . . . 35Table 5.4 Time scales and expected versus observed number of stars for the evolu-tionary stages chosen in the WFC3 CMD . . . . . . . . . . . . . . . . . . . 36Table 5.5 Time scales and expected versus observed number of stars for the evolu-tionary stages chosen in the ACS CMD . . . . . . . . . . . . . . . . . . . . 38viList of FiguresFigure 1.1 Evolution of 1M star . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3Figure 1.2 Differences between CMDs in the visible and ultraviolet range . . . . . . . 6Figure 1.3 BSS formation mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . 9Figure 2.1 47 Tucanae and observed fields . . . . . . . . . . . . . . . . . . . . . . . 16Figure 2.2 Completeness rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17Figure 2.3 The radial distribution of the SMC . . . . . . . . . . . . . . . . . . . . . . 18Figure 3.1 F225W,F225W −F336W CMD locations and radial distributions for MSand MSBn populations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20Figure 3.2 F225W,F225W−F336W CMD locations and radial distributions for faintand bright BSS and RGB populations . . . . . . . . . . . . . . . . . . . . 21Figure 3.3 F606W,F606W −F814W CMD showing the selection of the stellar pop-ulations on the ACS data, and where they fall on the F225W,F225W −F336W CMD as evidence of contamination to the UV HB . . . . . . . . . 23Figure 4.1 Mass segregation along the MS . . . . . . . . . . . . . . . . . . . . . . . 25Figure 4.2 Relationship between log(M) and log(R) . . . . . . . . . . . . . . . . . . 27Figure 4.3 Relationship between log(M) and log(R) reduced to the ACS field radius . 28Figure 4.4 CMD and radial distribution of 5 stellar evolutionary stages from the MSto WDs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29Figure 5.1 F225W,F225W−F336W CMD locations and radial distributions for faintand bright BSS and RGB populations compared to the MSBn . . . . . . . 32Figure 5.2 F225W,F225W −F336W CMD with MESA models and the radial distri-bution for the selected stellar populations . . . . . . . . . . . . . . . . . . 33Figure 5.3 F606W,F606W − F814W and F225W,F225W − F336W CMDs alongwith the radial distributions for the brighter stellar populations . . . . . . . 37viiGlossaryACS Advanced Camera for SurveysAGB Asymptotic Giant Branchb-b Binary-Binary (encounters)bBSS Bright Blue StragglersBSS Blue StragglersCMD Colour-Magnitude DiagramCNO Carbon-Nitrogen-Oxygen (cycle)fBSS Faint Blue StragglersGC Globular ClusterHB Horizontal BranchHST Hubble Space TelescopeKS-test Kolmogorov-Smirnov testMAST Barbara A. Mikulski Archive for Space TelescopesMESA Module for Experiments in Stellar AstrophysicsMS Main SequenceMT Mass Transferpp Proton-Proton (chain)PSF Point-Spread-FunctionRGB Red Giant Branchviiirc Core Radiusrh Half-Light Radiuss-b Single-Binary (encounters)SGB Sub-Giant BranchSMC Small Magellanic Clouds-s Single-Single (encounters)tcross Crossing Timetevol Dynamical Evolution TimeTO Turn Offtrelax Relaxation TimeUV UltravioletWD White DwarfWFC3 Wide Field Camera 3ZAMS Zero Age MSixAcknowledgementsI would like to gratefully acknowledge the help and guidance from my supervisors Dr. HarveyRicher and Dr. Jeremy Heyl throughout this project.I would like to thank to Dr. Jason Kalirai for sharing his data reduction and photometryknowledge, key to this research.I would also like to express my gratitude to Dr. Patricio Rojo, who had no obligation buttook the time to teach me scientific skills that I had no idea were needed for graduate school.Finally and most importantly, I would like to thank my family, especially my parents, fortheir unconditional support (emotional and financial), motivation and love, not only now butthroughout my life. And my friends, life would be boring without them.xChapter 1IntroductionGlobular star clusters (GC) are roughly spherical groups of stars thought to be composed of asimple stellar population, having all the stars born at the same time setting a perfect labora-tory for the study of stellar evolution. With the development of astronomical instrumentation,scientists have been able to study GC in more detail, exposing the presence of different anoma-lous stellar populations. An important example of such stars are blue stragglers (BSS). Firstdiscovered by Sandage (1953) in the GC M3, BSS were described as an extension of the mainsequence (MS) defying normal stellar evolution within a cluster (see section 1.1). How thesestars are formed in GC and where do they go after they leave their MS stage has been a constantdebate.With a large sample of 157 GC in the galaxy (Harris, 1996), these agglomeration of starsare one of the most widely studied systems in astronomy due to their versatility. Not only arethey a good place to study the evolution of stars and dynamics of stellar systems, but they alsogive us information about the structure, chemical composition and dynamical history of theMilky Way. A good example of a well studied system is NGC 104 (47 Tucanae). Visible fromthe southern hemisphere, 47 Tuc is the second largest and brightest GC in the sky. Locatedat ∼ 4.7 kiloparsecs from the Sun (Woodley et al., 2012), 47 Tuc is home to 2 million stars.Although 47 Tuc has been the target of many investigations, this thesis is the first time such abig portion of the core of the cluster has been observed with ultraviolet filters, allowing us togo deeper into the most dense region of this system (more details of the observations procedurewill be given in Chapter 2).One way to study the evolution of GCs and the stars in it, is to use photometric data, whichcan be used to create a colour-magnitude diagram (CMD) using the measured magnitudes andalso radial distributions using the positions for each star. The CMD will allow us to separatethe stars of the cluster into different stellar evolutionary stages while the radial distributionswill tell us about the dynamics of the system.1We will begin with a brief explanation of stellar evolution and how CMDs help us trace thedifferent populations in a GC followed by a section on the dynamics of these systems. The lastsection of the introduction will give a detailed summary of the historic and current results onthe formation and evolution of BSS relevant to this investigation.1.1 Colour-Magnitude Diagrams and Stellar EvolutionA Colour-Magnitude Diagram is a scatter plot showing, as the name suggest, the relationshipbetween apparent magnitude (luminosity) and color (effective temperature) of stars. The ap-parent magnitude of a star depends on the distance of the star with respect to Earth, in this casewe can use such a plot because the stars of GCs are considered to be all at the same distancefrom Earth.To understand the CMD, we need to first understand stellar evolution. To keep thingsconnected, we will trace the evolution of a one solar mass (1M) star in terms of a MESA(Modules for Experiments in Stellar Astrophysics; Paxton et al. 2011, 2013, 2015) modelplaced in CMD space (Figure 1.1) and go through the different stages. We chose this particularmass as it covers all the evolutionary stages seen on the CMD of 47 Tucanae (and some extrastages that due to the short time for which they exist is almost impossible to see on the CMDof a GC). There are though some differences in the evolution of stars with lower and highermasses than 1M and they will be mentioned when the differences become important. Wewill leave aside the formulas that explain the physical processes going on inside the stars andfocus mostly on a description. This section is intended as an introduction to understand whatis going on in each stage that will later be used to trace the evolution of the cluster and not asan introduction to stellar astrophysics.1.1.1 The Evolution of a Solar Mass StarStar formation begins in the interstellar medium when a molecular cloud becomes unstable togravitational collapse and a protostar is formed. The initial stages of stellar formation, andhow we get to a protostar still generate some debate. The end of this rapid contraction phaseis marked by the beginning of the Hayashi track (marked as (1) on Figure 1.1). The star isnow on its way to the MS, the collapse rate slows down and the star becomes fully convective.Contraction continues as the protostar moves down the Hayashi track; luminosity and stellarradius decrease until it develops a radiative core. At this point the protostar enters the Henyeytrack (2), here it becomes hotter until contraction ends and the core is hot enough for theprotostar to begin hydrogen fusion (3) which marks the birth of the star as it reaches the zeroage MS (ZAMS) (Collins, 1989).The MS is the longest evolutionary stage of a star, the lower the mass the more time it2Figure 1.1: MESA evolutionary model for a 1M star. The model was coloured to aCMD using table 1.4 from Sparke & Gallagher (2007). The different colours alongthe model curve represent the different evolutionary stages of the star while thenumbers mark the beginning and/or end of these stages. The inset shows a close upof the red bump along the red giant branch.spends on the MS, and it is characterized by the conversion of hydrogen into helium in the core.There are two burning channels through which a star can accomplish this: the proton-proton(pp) chain and the Carbon-Nitrogen-Oxygen (CNO) cycle. In most stars both mechanisms arepresent but the dominant one is determined by the star’s mass and the temperature it can reachin its core. For a 1M star the principal fusion reaction is the pp-chain, but around 1.2M there3is a transition to the CNO cycle which strongly depends on temperature and thus becomes moreimportant for higher masses (Binney & Merrifield, 1998).Once the star runs out of hydrogen in the core (4) (i.e. reaches the turn off (TO) point)the star becomes a sub giant. We now have an inert helium core and a hydrogen-burningshell surrounding it. In order for the shell to burn hydrogen into helium it must first reach atemperature of∼ 107K (Beccari & Carraro, 2015). Once the shell ignites the hydrogen burning,the density of the regions surrounding the core will decrease, and the core will grow through theaddition of helium from the outside shell. In order for the nuclear energy generation to supportthe whole weight of the star, the core temperature rises slowly which also causes a steadyincrease of the star’s luminosity. To regain equilibrium the star will respond by expanding andcooling the surface while moving towards the region of the CMD dominated by convection(Collins, 1989). At the moment the star is closest to the Hayashi track the sub-giant branch(SGB) ends (5) and the star is now going up the red giant branch (RGB). The time a starspends on the RGB is very mass dependant, the higher the mass the shorter the time on theRGB, becoming non-existent for stars with masses over 15M and very long for stars under2.2M (Beccari & Carraro, 2015). An important feature of the RGB for low-mass stars, isthe red-giant bump, visible on the CMD by the accumulation of stars at a certain magnitude.The accumulation of stars happens as the stars spend more time in this stage of their evolutionthan in any other part of the RGB. Looking at the the inset on Figure 1.1, we can see the star’sevolutionary path going back and forth in luminosity (represented on the CMD as magnitude),these changes in luminosity are believe to be caused by a “jump in mean molecular weightat the dredge-up composition discontinuity, as a consequence of its effect on the hydrostaticstructure of the region immediately above the hydrogen burning shell” (Christensen-Dalsgaard,2015) which leads to a small decrease in luminosity. Once the shell leaves the discontinuitythe star goes back to moving up the RGB.The end of the star on the RGB is marked by the helium flash at the tip of the RGB (6), thecore of the star is now hot enough to burn helium into carbon, and will still have the hydrogenburning shell around it. The flashes will continue until the temperature of the core is highenough to completely remove the degeneracy of the core and the equation of state reverts tothe ideal-gas law (Collins, 1989). The star finally settles at the zero age horizontal branch (HB)where it will continue fusing helium into carbon and oxygen.When the He runs out in the core (7) the star will start going up the asymptotic giant branch(AGB). Here the star has an inert carbon-oxygen core with a helium burning shell around itand a hydrogen burning shell around that. Once again the envelope expands, the temperaturedecreases and the luminosity increases. The star starts losing these shells (8) and it becomes aplanetary nebula. Temperature increases very quickly (with almost a constant luminosity) untilall the gas is dispersed (9). Finally the star starts cooling off and is now a white dwarf (WD).4During the cooling stage there is no nuclear generation of energy and all the energy emittedcomes from the stored thermal energy of the WD.1.1.2 Implications for Globular ClustersAs we mention earlier, stars with higher masses evolve more quickly than those of lowermasses, adding the fact that the stars in a cluster are born at the same time1, this would lead usto expect that any star more massive than the TO should have already evolved off the MS. Fig-ure 1.2 shows two CMDs of 47 Tuc, constructed with data from the Hubble Space Telescope(HST) using, for the left plot, filters in the visible range of the spectrum (F606W and F814W),and the other one using ultraviolet (UV) filters (F225W and F336W, chapter 2 will discussmore about the filters and how the CMDs were obtained). In both CMDs we have highlightedthe different branches discussed previously in order to show where each population is, how theshape of the CMD changes with different filters and how different filters favour different stars(the brightest stars on one CMD are not necessarily bright on the other). Tracing the evolutionof stars on the CMDs seems like an easy task if we stick to the bulk of the population, howeverwe see the presence of stars above the TO as an extension of the MS. These stars are the BSS,and though they do not make up even 1% of the observed sample they are unexpected, andtherefore interesting, members of the cluster.1.2 Dynamical Evolution of Globular ClustersTo describe the dynamical properties and morphology of GCs relevant to this investigation,only a few main parameters are needed. We mention at the beginning that the observations tocomplete this research were performed on the core of 47 Tuc. The core radius, rc, is defined asthe radius where the surface brightness distribution drops by half from its central value. Thereis also the half-light radius, rh, which contains half the total luminosity of the cluster. For 47Tuc these values are 21.6 and 190 arcseconds respectively (Harris, 1996).The dynamics of GCs (or essentially any stellar system) can be described using three differ-ent time scales: the crossing time, tcross, the relaxation time, trelax, and the dynamical evolutiontime, tevol . This last time scale is define by Meylan (2000) as ”the time during which energy-changing mechanisms operate, stars escape, while the size and profile of the system change”.The crossing time, defined as tcross = R/υ (Binney & Tremaine, 2008) (where R is the radiusof the system, and υ , the typical speed of a star), is the time a star takes to cross the system and1There is photometric (Milone et al., 2012a) and dynamical (Richer et al., 2013) evidence that clusters may haveindeed two or three different generations of stars. But these would have only occurred within a period of 1 to 2 Gyr(Ventura et al., 2009), not enough to have stars as massive as the BSS.5Figure 1.2: Comparing CMDs in the visible (left) and UV (right) range. The figure showsthe differences between the CMDs constructed with the same two sets of filtersthat we will use for the analysis. To get the same branches on both CMDs, theF606W,F606W −F814W diagram was completed using data from outside the core(HST cycle 17 GO-11677, PI: Richer) as in those filters its not possible to get tosuch faint stars in the core as the lower MS and WD stars.its related to the second time scale by:trelax ≈ 0.1Nln(N) tcross (1.1)(Binney & Tremaine, 2008) where N is the number of total of stars in the system. The relax-ation time is described as the time it takes the system to have its velocity distribution approach aMaxwellian distribution (Spitzer, 1987). For 47 Tuc, the relaxation time in the core is believedto be about 70 Myr (Harris, 1996; Heyl et al., 2015b).An important process in GC dynamics is mass segregation that happens on a time scaleof trelax. Essentially mass segregation means that more massive stars move towards the centerof the cluster while less massive ones tend to go towards larger radii, completely changing the6mass distribution the cluster began with. This process is the result of two different mechanisms:relaxation and equipartition. The first one comes from the fact that each star wanders awayfrom its initial orbit increasing the entropy of the system leading it to a new configuration witha small, dense core and and a large, low-density halo (Binney & Tremaine, 2008). The secondone comes from kinetic theory which tells us that particle encounters will make those particleswith large kinetic energy lose energy to those with lower energies, leading to a state where themean-square velocity is inversely proportional to mass. In the case of stars, massive ones thatlose energy to less massive stars fall towards the center increasing their velocities and gainingkinetic energy but lose it by falling and continue to fall, while less massive stars rise towardsthe outer parts of the cluster as they slow down (Meylan, 2000; Binney & Tremaine, 2008).One of the first detailed studies of mass segregation fin 47 Tuc was carried out by Anderson(1997). Using images of the core of 47 Tuc, he was able to measure the luminosity functionto which he fitted King-Michie models obtaining the best agreement with those models thatincluded mass segregation. But not only can mass segregation be analysed through luminosityfunctions, if the core of a cluster is indeed relaxed, the radial distribution of different groupsof stars should also exhibit indications of this phenomenon and 47 Tuc should not be the ex-ception. We will show in Chapter 4 how the high quality of this data set, allows us to displayevidence of mass segregation in the core of 47 Tuc by using the mass difference between MSstars of different magnitudes (or mass). This process will also lead us to an estimate of themasses of the stars in the different sequences visible in our CMD.1.3 Blue StragglersIn the last two decades BSS have been found in many GCs as well as open clusters (de Marchiet al., 2006; Ahumada & Lapasset, 2005), in dwarf galaxies (Santana et al., 2012) and in thefield of our galaxy (Santucci et al., 2015). Although there is a large amount of observationaldata revealing important characteristics about BSS, observations alone cannot tell us how orwhen BSS were formed. Determining the possible formation channels and which ones dom-inate in the different environments requires models of formation mechanisms and statisticalanalysis.In order for these stars to look brighter and bluer than the TO, they had to go through somerejuvenating process as there is no evidence of recent star formation episodes in the environ-ments where BSS live. The BSS formation mechanisms can be divided in many different waysbut they all must comply two main conditions: i) there must be at least one MS star involved,and ii) one of the stars involved must gain mass in order to rejuvenate. In fact, the positionsof BSS on the CMD suggest that these stars are in fact more massive than the TO stars. Thefirst attempt to directly measure the mass of a BSS was done by Shara et al. (1997), studyingone of the brightest BSS in the core of 47 Tuc, they found a mass of 1.7±0.4M, almost twice7the cluster TO mass of ∼ 0.9M (Hesser et al., 1987; Thompson et al., 2010). Later on, dif-ferent studies, including some done on variable BSS, have yielded masses between ∼ 1−2Mfor BSS in different GCs (Gilliland et al., 1998; De Marco et al., 2005). Recent results forpulsating BSS have provided a lower upper limit of ∼ 1.5M (Fiorentino et al., 2014, 2015).In an attempt to recreate the observed BSS populations and their characteristics many sce-narios for the formation of BSS have been proposed, successfully explaining some cases butfailing in others. As we have already mentioned what they have in common, the first differencewe can make then is through which process the mass exchange happens. Following Figure 1.3,we have two mass gaining mechanisms: mass transfer or merger. How we get to this channelsis a much longer story. We will separate the initial scenarios into three different categories fol-lowing the divisions chosen by Perets (2015): i) direct collisions of stars, ii) stellar evolutionof primordial binaries, and iii) dynamical evolution of hierarchical triple systems.At some point, the line between the formation mechanisms becomes hazy, for example, hi-erarchical triple systems can form from binary-binary (b-b) interactions (Antonini et al., 2015),at the same time we a can consider b-b interactions as collisions. As in the latest review (Boffinet al., 2015), we will consider any fast dynamical encounter, involving single stars or binarysystems, as collisions. Considering this issue, we will summarize the characteristics of thedifferent formation mechanisms and their end products, indicating the points at which the di-visions between the formation channels become unclear.1.3.1 Evolution of Primordial BinariesStellar evolution of the individual stars composing a binary system can lead to mass transfer(MT) from one member of the system to the other. This process was first proposed as thepossible origin for BSS by McCrea (1964), around ten years after their discovery. He claimedthat, having enough MT between the members of a close binary system, the secondary star, towhich the mass has been transferred, will end up as an apparently young star. He also predictedthis process could lead to BSS of up to 2.5 magnitudes brighter than the TO, not too far awayfrom the values observed today for BSS in stellar clusters.To predict the outcome of the evolution of a binary system, several characteristics have tobe considered. The most important one of these is the stellar evolutionary state of the donorstar, which also defines the classification system of cases of MT in binaries introduced byKippenhahn & Weigert (1967). These cases are divided in:• Case A: MT during MS.• Case B: MT beyond MS but before helium ignition.• Case C: MT beyond helium ignition8Direct CollisionsHierarchical Triple SystemsPrimordial Binary EvolutionMass TranferThrough Kozai mechanismBSS in a binary systemMerger Single BSSPrimordial or dynamically formedBSS in a triple systemcan lead toFast dynamicalencountersFigure 1.3: Summary of proposed formation mechanisms. The red ovals show the twopossible mass gaining mechanisms while the blue circles have the resulting BSSor BSS system. The different coloured rectangles are the three possibles initialscenarios and following the same coloured arrows one can get to the final productof the process. For the direct collisions mechanism the paths to the final productshas been left out as, depending on the number of stars involved, they can lead to allthe possible end products.as explained by Perets (2015). Other important aspects to consider are the structure of thedonor’s envelope, the mass ratio of the binary, and the type of the accretor, which are not onlykey in determining the end product but are also responsible for the stability of the mass transferprocess (Ivanova, 2015).We will now turn our focus to the conditions that allow a binary system to evolve to forma BSS. Independently of which case of MT the binary follows, the primary star needs to haveenough mass available to transfer to the secondary, to make this last one more massive thanthe TO stars of the cluster (Davies, 2015). Having sufficient mass, the different cases will giveBSS with different characteristics, with the mass of the resulting BSS strongly depending onthe initial binary orbit (Sills, 2010).Case A can either form a single massive BSS, if the MT leads to a merger, or a BSS ina short period binary system (Perets, 2015). This is one of the examples where is hard todifferentiate from one formation mechanism to the other. For a binary system, consisting of9two MS stars, to have a small enough initial separation to exchange mass, it is believe that aprocess outside from the natural evolution of the stars needs to take place (Perets & Fabrycky,2009), for example the perturbation from a third star (Fabrycky & Tremaine, 2007).In order for case B and C to form a BSS, the binary system contains a post MS star thatshould not go through a common envelope stage (Hjellming & Taam, 1991) with its compan-ion. The resulting BSS, would end up in higher period binaries compared to case A, with ahelium WD companion for case B and a carbon-oxygen WD for case C (Perets, 2015). Themass of the BSS is not expected to be very high for these cases, getting very close to that ofthe TO especially for case C.1.3.2 Direct Stellar CollisionsThe first to claim a possible collisional origin for BSS was Hills & Day (1976) during a study ofstellar collisions in GCs. Their research indicated that a MS star in the dense cores of GCs hada 3% chance of colliding with another MS star during the lifetime of the cluster, chances wentup for when one of the stars was in its giant stage. If these these interactions were followed bycoalescence, the product would be a BSS.Following the Hills & Day reasoning, Davies (2015) states two conditions that need to bemet in order for a collision to form a BSS: first, the collision must lead to the merger of theinvolved stars, and second, the end product of the collision must look more massive than theTO stars of the cluster (a rejuvenated star with a mass below that of the TO of a system will notbe observable as a BSS on a CMD). In fact, for low velocity encounters, collisions are believedto be very effective conserving most of the mass of both merged stars (Benz & Hills, 1987).Even though the stars are rejuvenated, they are not reborn, as pointed out by Sills (2010),BSS resulting from collisions are thought to evolve in a similar manner as normal stars of thesame mass, however they are expected to have shorter MS lifetimes. The reason behind thisassumption comes from the fact that the new star is made up of stars that had already beenevolving for some time and thus the initial amount of hydrogen in the core is smaller comparedto a zero age MS star of the same mass. According to Lombardi et al. (1996), for stars of nearlyequal mass, the collision product does not fully mix, instead, the cores of the participating starsend up as the core of the new BSS. In this case, if both stars are close to the TO mass, the BSSwill not have a very long MS lifetime. In contrast, in collisions involving stars with a massratio ≤ 0.5, the hydrogen rich smaller star settles in the core of the merged product, addingnot only the remaining hydrogen in its core but also the hydrogen shell around it, producing alonger MS lifetime compared to the equal mass case.Direct collision of stars do not only happen between single stars (s-s), but binaries can alsobe involved. Single-binary (s-b) and b-b encounters are also possible and more likely than s-scollisions (Sills, 2010). Encounters involving binary systems will be significant when another10star or binary passes within a distance equivalent to the size of the binary (Davies, 2015). Theproduct of these collisions between more than two single stars can leave behind more massiveBSS exceeding twice the TO mass (Fregeau et al., 2004), with a long period (& 103 days)binary companion (Chatterjee et al., 2013).1.3.3 Dynamical Evolution of Hierarchical Triple SystemsThe last option for the origin of BSS is much newer compared to the previous ones. Iben &Tutukov (1999) were the first to describe an scenario where a hierarchical triple system (a thirdstar is orbiting the inner binary) would actually evolve to become a BSS in a binary system. Afew years before Leonard (1993) had a similar idea, but he discarded this possibility claimingthere would need to be a much higher triple system frequency than the one observed at thattime and went back to the physical collisions theory.With the discovery of triple systems harbouring BSS (see van den Berg et al. (2001) foran example), and the disagreement between the observed BSS populations and that obtainedfrom combined N-body and stellar evolution simulations that considered only collisions andprimordial binary evolution, the study of BSS formation through triple systems evolution be-came an independent subject of study. Perets & Fabrycky (2009) claimed that previous BSSformation studies demanded a fraction of the primordial binaries to be short period binarieswhich a previous publication by one of the authors (Fabrycky & Tremaine, 2007) had shownthat such systems actually come from longer period binaries that have been perturbed by a thirdstar via the Lidov-Kozai mechanism (Lidov, 1962; Kozai, 1962). In fact, studies done on shortperiod (Tokovinin, 1997) and contact (Pribulla & Rucinski, 2006) binaries showed that at least40% of these systems have distant companions.Recent studies following the formation channel proposed by Perets & Fabrycky, indicatethat the Lidov-Kozai mechanism has a 21% efficiency when it comes to forming tight binaries(Naoz & Fabrycky, 2014). And, when applied to GC systems, it can contribute up to 10% of thetotal BSS population (Antonini et al., 2015). This population should show some observationaldifferences when comparing them to the BSS from the binary mass transfer scenario. Forinstance their mass could reach much higher values than cases B and C, where part of the massof the system is left in the WD companion. The WD is also another difference as BSS from atriple system are more likely to be left with a MS companion (Perets, 2015).1.3.4 Linking Models to ObservationsAll the above formation mechanisms are able to reproduce the observed properties of singleBSS. But when we observe a system of stars such as a stellar cluster, we are analysing apopulation and not single stars. For a formation mechanism to produce an compelling numberof BSS such that it yields a notable observable population, it must occur at a significant rate,11and the end products must have a lifetime long enough to accumulate (Davies, 2015). Whichmechanism dominates in the different environments where BSS live is still in debate. Althoughnot definite answer has been reach most studies agree that the observed populations today area result of a combination of all the formation channels, with one mechanism prevailing overthe others depending on the system’s properties.When trying to reproduce the observed populations of BSS in clusters including all thefactors becomes an almost impossible task. Besides having simulations producing BSS throughthe different mechanisms, the dynamical evolution of the cluster also needs to be taken intoaccount. One important effect of dynamical interactions is that, over time, these will alter thebinary population of a system. Heggie (1975) showed that dynamical encounters in N-bodysystems will make tight binaries tighter while soft binaries are likely to be destroyed. Thisstatement is supported by the anticorrelation found by Milone et al. (2012b), between binaryfraction and absolute luminosity (mass).Attempts to find the dominating formation mechanism in different GCs, have been basedon finding the strongest correlation between the number of BSS and parameters of the cluster,like total or core mass, binary fraction and collision rate, that are some how related to thedifferent formation channels. Early studies that compared models to observations found noimportant correlations, for example (Piotto et al., 2004), noticed no correlation between thenumber of BSS and any of the cluster parameters except for a very low dependence on thecentral density. As it was very difficult to get good photometric data in the dense cores of GC,it wasnt until 2009 when Knigge et al. found a strong correlation between the number of BSSin the core and the core mass, concluding that most of the BSS come from binary systems, butat the same time these binaries could have been affected by dynamical encounters. With theresults pointing towards a binary origin for BSS, researchers started to look for confirmationof the correlation between BSS frequency and binary fraction already found by Sollima et al.(2008) in low density GCs. Milone et al. (2012b) reaffirmed this correlation for a sample of 59GCs. Leigh et al. (2013) also tried to find a relation between binaries and BSS but their resultsshowed a much stronger correlation with the core mass as found by Knigge et al. (2009),despite the fact that binary fraction in GCs anticorrelates with core mass (Milone et al., 2008).One of the latest studies that included dynamical effects and stellar and binary evolution yielded“a dependence of blue straggler number on cluster mass, a tighter correlation with core mass,a weak dependence on the collisional parameter, and a strong dependence on the number ofbinary stars” (Sills et al., 2013).To compare the observed BSS to the modelled BSS population, many studies use the radialdistribution of these stars, not only to compare them to models but also to other populationsin the same system. The analysis of the observed radial distributions all agree that, for stellarclusters, the BSS are more centrally concentrated that the rest of the populations (Perets, 2015)12which according to mass segregation means that they are more massive than the average star.Other interesting results have been the discovery of bimodal distributions for BSS in GCs likeM3 (Ferraro et al., 1997), 47 Tuc (Ferraro et al., 2004), M55 (Lanzoni et al., 2007) and NGC6229 (Sanna et al., 2012). These distributions, in general, show a peak in the cluster center,decreasing at intermediate distances from the center, to rise again in the outskirts. A goodexplanation for this bimodality, was presented by Mapelli et al. (2006) where they concludethat the BSS in the external regions are almost entirely product of mass transfer in primordialbinaries, in contrast, core BSS are more likely to have originated from collisions. Additionalobservational evidence in favour of the mixed formation mechanism, is the two distinct se-quences of BSS observed in the GC M30 (Ferraro et al., 2009). Here the authors claim thatthe bluer BSS have a collision origin while the redder BSS are the product of the evolution ofclose binaries.1.3.5 Blue Stragglers and 47 TucanaeIn the particular case of 47 Tuc, the study of its population of BSS started with the discoveryof 21 of such stars in one of the first HST observations of the core of this cluster (Paresce et al.,1991). This small sample of BSS already showed signs that the density of BSS is higher inthe central regions of the cluster. Many investigations on the topic have taken place since then,before the discovery of the mentioned bimodal distribution in 47 Tuc, Sills et al. (2000) mod-elled the formation rate of BSS using data outside the core. The results obtained by the authorssuggested that 47 Tuc may have stopped making BSS several billion years ago, undergoing anepoch of enhanced BSS formation around the same time possibly connected to the epoch ofprimordial binary burning.Following the discovery of the bimodal distribution, different attempts to explain the spa-tial layout of BSS in 47 Tuc were made. Mapelli et al. (2004) tried to reproduce the BSS radialdistribution by choosing different formation mechanisms: collisional BSS in the innermostregion and primordial binary evolution outside the core. The best representation of the obser-vational data was obtained when 25% of the BSS come from binaries and 75% from collisionswithin 0.5rc. This result was later refined by Mapelli et al. (2006) obtaining a best fit when46% of the BSS come from mass transfer and 54% from collisions. The models were also ableto predict the minimum in the radial distribution and its surrounding regions named by Mapelliet al. (2004) the “zone of avoidance”, with the condition that external MT BSS productionbegan beyond 30rc.Later on, Monkman et al. (2006), tried to explain the bimodal distribution with a purelycollisional model throughout the cluster. Their results agreed with those found by Mapelli et al.(2004, 2006) for the core of the cluster where the collisional model represents the observationaldata. For their middle region (between 23 and 130 arcseconds) BSS formation would have13needed to stop about half a billion years ago. But for the external regions the collisionalmodels were not able to predict the BSS population, a result that they concluded is likely dueto another formation mechanism dominating the outskirts of 47 Tuc.Around the same time the formation mechanisms debate was taking place, researchersfound evidence that BSS in the core of 47 Tuc have masses larger than twice the MS TO mass.One result that suggested the presence of massive BSS was found by McLaughlin et al. (2006),while studying the proper motion and dynamics of the cluster core, determined that the velocitydispersion of BSS was smaller than that of the cluster giants by a factor of√2 (i.e. twice theirmass). That same year, Knigge et al. (2006) identified a detached binary system consisting ofa 1.5M BSS primary with an active, upper MS companion. These massive BSS can only bethe outcome of a process involving at least three progenitors.Another interesting area of research is the evolution of BSS. One piece of observationalevidence that suggests where evolved BSS might live on the CMD is presented by Beccariet al. (2006). The authors examined the bright end of the CMD and found an overabundanceof massive stars in the AGB of 47 Tuc and concluded that they could be possible related to theevolution of binary systems. This presence of extra stars had already been noticed by Bailynin 1994, who also linked them to the evolution of BSS. We will carry a more detailed analysisof this topic in section 6.2.14Chapter 2Observations and Data AnalysisAlthough obtaining photometry of GCs seems like a direct process, it comes with some com-plications, most of them due to the high stellar density of these systems, especially in or nearto the core. Stars are so close together that, even with the best observing conditions and tech-nology, it is very likely to miss some of them, particularly towards fainter magnitudes. Toavoid any misinterpretation of the data caused by missing sources, we resort to point-spread-function (PSF) photometry (section 2.1) to optimize our object detections, and artificial startests (section 2.2), for incompleteness corrections.2.1 Observations and PhotometryThe data come from observations made with the Hubble Space Telescope (HST) using WideField Camera 3 (WFC3) with two of the most ultraviolet (UV) filters, F225W and F336W,whose central wavelengths are 235.9 nm and 335.9 nm respectively. Ten fields in the core of47 Tuc were obtained between November 2012 and August 2013 during cycle 20 of the HSTprogram GO-12971 (PI: H. Richer). The observations were planned so that each visit includedtwo exposures in each filter, 380s and 700s for F225W and 485s and 720s for F336W. Eachfield was offset from the previous one to map the entire central region of the cluster as shownin Figure 2.1. The combined field of view covers a radius of ∼ 160 arcseconds from the centerof the cluster.The data analysis was performed following the procedure described in Kalirai et al. (2012).First the observations were retrieved from MAST (Barbara A. Mikulski Archive for SpaceTelescopes). All the images were then corrected for geometric distortions using MultiDrizzle(Fruchter & Sosey, 2009). The next step was to register the images onto the same referenceframe, using DAOPHOT II (Stetson, 1987) we selected the brightest stars in each image andobtained their positions to calculate the transformations between each field and the reference.The transformations were then put together in a shift file which MultiDrizzle uses to make one15drizzled image for each filter. Finally, using DAOPHOT II and ALLSTAR (Stetson, 1994) PSFphotometry1 was performed in both the stacked images, and the magnitudes were zero pointedto the VEGAMAG (Bohlin & Gilliland, 2004) photometric system2. The two final photometryfiles are matched into a single catalogue resulting in the UV CMD depicted, for example, inthe figure in the left panel of Figure 2.3. It is important to mention that as we will be analysingradial distributions, the star like shape of the final field has been reduced to a circular areacentred in the center of 47 Tuc as shown by the blue circle on Figure 2.1.320''Figure 2.1: GC 47 Tucanae and the observed field. This image was generated using theAstroView Tool of MAST Data Discovery Portal. The background image is partof the Digitized Sky Survey and was taken by the UK Schmidt Telescope at SidingSpring Observatory in New South Wales, Australia. The orange squares representthe WFC3 fields. The blue circle and dimension indicator are superposed to showthe actual field used for this research and its diameter.2.2 Artificial Star Tests: Correcting for IncompletenessTo estimate the number of stars lost in the photometry process, we ran artificial star tests. Theprocedure, explained in detail in Heyl et al. (2015b), consists in inserting artificial stars into the1DAOPHOT runs aperture photometry on all the stars above a threshold, from which it picks a set of welldefined stars to build a PSF. Finally it measures the positions and magnitudes of all the stars that match the PSF inthe field.2The magnitudes observed are instrumental magnitudes, VEGAMAG is a system to convert these magnitudesinto a common photometric system. VEGAMAG gets its name from the star Vega (a bright AOV star in the con-stellation Lyra with a very smooth spectrum), and is defined such that the magnitude of Vega is 0 at all wavelengths.16images in both F225W and F336W filters and calculating how effectively these are recoveredwhen run through the same photometry process as the real stars. The completeness rate is afunction of both the magnitude of the star and its distance from the center of the cluster, and so,artificial stars were given a range of values covering the observed magnitudes and distances tothe center of 47 Tuc. As can be seen in Figure 2.2, the completeness rate is strongly dependenton both radius and magnitude, with only the brightest stars close to unity.Figure 2.2: Completeness rate as a function of the distance from the center of the clusterand the magnitude of the artificial star.To test our correction for incompleteness, we compared the radial distribution of the SmallMagellanic Cloud (SMC) to that of R2. The SMC is a dwarf galaxy orbiting the Milky Waywhich happens to lie in the background of 47 Tuc. The two objects are completely unrelatedand very far apart (47 Tuc is 4.5 kpc Harris, 1996, away from the Sun, while the SMC is at∼ 60 kpc, Hilditch et al., 2005), but since they share the same sky area, stars from the SMCcontaminate the CMD of 47 Tuc. Figure 2.3 shows where the MS of the SMC lies on ourCMD. Because the SMC is not related to 47 Tuc, the radial distribution of its stars shouldbe proportional to the area of the field of our observations. Looking at the right panel of17Figure 2.3, we can see the comparison between the incompleteness corrected cumulative radialdistributions of the SMC and R2 (as we are only counting the stars within a circular area). Ifour completeness rates were properly obtained then these distributions should be approximatelyequal. In fact the KS-test yields a p-value of 0.60 telling that we cannot reject the hypothesisthat the distributions are in fact the same. The mean completeness fraction of the SMC sampleis less than 70%, so the completeness correction is crucial to obtaining the estimate of theunderlying distribution.Figure 2.3: Left: Selection of SMC stars on the UV CMD. Right: Cumulative radialdistribution of the SMC compared to R2. The legend on the CMD indicates thenumber of stars before correcting for incompleteness, while the legend in the rightplot gives the size of the sample after correcting for incompleteness. The agreementbetween both distributions allow us check the validity of our completeness rates.2.3 The ACS Data SetThe data used to construct the CMD in the visible range (hereafter the ACS data) was obtainedfrom the ACS (Advanced Camera for Surveys) Survey of Galactic Globular Cluster Sarajediniet al. (2007). The survey used the ACS Wide Field Channel to obtain photometric data of65 of the nearest globular clusters and is publicly available at: http://www.astro.ufl.edu/∼ata/public hstgc/databases.html. A description of the data reduction and photometry can be foundin Anderson et al. (2008).18Chapter 3Stellar Population SelectionDue to the high quality of the data, each population is easily identified and can be separatedone from another. Each population is defined to be within a region shown by the differentcolor boxes in figures 2.3 through 5.3. The boundaries of each region were chosen with thehelp of MESA evolutionary models, slight modifications on the limits of these regions wouldonly make including stars with higher photometry errors or not real members of the differentbranches more likely. Also, including the stars surrounding the highlighted regions does notchange the number of stars in each box by more than a few percent and tests done includingthese stars show no effect on the shapes of the cumulative radial distributions.3.1 Main Sequence BinariesIn an attempt to identify the population of stars responsible for the formation of BSS, wewill later compare the BSS distribution against the binary stars distribution. We have selecteda sample of main sequence binaries (MSBn) that we expect to be mostly nearly equal massbinaries. Both populations are shown on Figure 3.1. The MSBn selection box starts at amagnitude value of 24 and extends up to brighter stars by about 3 magnitudes with an almostconstant width of 0.4 magnitudes (width reduces at the brighter end of this selection box toavoid contamination by SGB stars) containing a total of 367 stars number that goes up to∼438after correcting for incompleteness.We have also included a selection of stars on the main sequence (MS) to have a referencefor the analysis of this population. To ensure there is no contamination to the MSBn samples,a minimun distance of ∼ 0.2 magnitudes is kept between the binary sequences and its singlestar sequence counterpart.Looking at the right panel of Figure 3.1, we can see that the cumulative radial distributionfor the MSBn is much more centrally concentrated than that of the single MS stars.19Figure 3.1: Left:F225W,F225W −F336W CMD showing the selected stars for the MSand MSBn. Right: Radial distributions for the selected samples. The inset on theCMD has the number of stars before correcting for incompleteness, while the onein the right panel gives the size of the sample after correcting for incompleteness.3.2 Blue StragglersAs can be seen in Figure 3.2 the BSS population is easily spotted on the UV CMD as anextension of the MS of the cluster. Starting a few tenths of magnitudes above the turn-offpoint and extending for almost 4 magnitudes, the total number of BSS on the sample goesup to almost 150 stars. For this study, we have decided to exclude the very faint BSS, andtaken only those that are at least ∼ 0.7 magnitudes brighter than the TO, to avoid any possiblecontamination due to blends. This decision was also based on the fact that when we plotthe BSS sample on the ACS CMD, the fainter BSS on the UV sample are very close to theF606W,F606W −F818W TO, almost blending with the MS and it is important that we haveclean BSS samples on both CMDs.After delimiting the BSS sample we end up with 114 BSS, which we divide into two sub-samples, faint and bright BSS. Each smaller sample started with half the stars of the original,with the bright BSS (bBSS) distribution looking much tighter than the faint BSS (fBSS). Inorder to obtain two distinct BSS populations we have maximized the difference between theirradial distributions ending up with 58 bBSS and 56 fBSS, with the bBSS extending across amuch larger range in magnitude (2.5 magnitudes) compared to the faint sample (0.85 magni-tude). Figure 3.2 shows the final divided BSS samples and their corresponding radial distribu-tions compare to the reference population explained in the next section.20Figure 3.2: Left:F225W,F225W −F336W CMD showing the selected stars for the faintand bright BSS and RGB. Right: Radial distributions for the selected samples. Thelegend on the CMD has the number of stars before correcting for incompleteness,while the legend on the right plot gives the size of the sample after correcting for in-completeness. The division between bright and faint BSS was chosen by maximiz-ing the difference between their radial distributions in order to obtain two distinctBSS populations.3.3 Reference PopulationTo trace the cluster stars we selected the RGB as the reference population. Although previousstudies (Ferraro et al., 2003, 2004) indicate the HB as the most natural reference population, inthe UV CMDs, due to this branch being well separated from other branches, we are concernedwith the contamination of the HB by AGB stars and evolving BSS which will be discussedlater on chapters 5 and 6.In the UV, specifically with the filters chosen for this work, the RGB is well defined andeasy to identify on the CMD. Even though the RGB is not separated from the SGB the shape ofthe CMD makes it easy to get a clean sample. To make sure there is no contamination of SGBstars on our RGB sample, we start our RGB box a few tenths of a magnitudes above and to theright of the end of the SGB. The box pointed out in red on Figure 3.2 shows the final selectionfor the reference population with a total of 2925 stars before completeness correction and asmooth radial distribution corrected for incompleteness coming to a total of ∼ 3050 stars.As mentioned in previous chapters, we will compare our data to the ACS sample. On theACS CMD, the RGB, especially in the fainter part of this branch, is also well defined. Cross21matching the selection of the RGB on the UV CMD to the visible range CMD, also gets usa clean RGB sample of stars starting around ∼ 0.5 magnitudes brighter than the TO, whichtells us that our efforts to exclude SGB stars from our UV sample were successful. Even at thebright end of the RGB on the visible range CMD we can see that this branch is well separatedfrom the horizontal and asymptotic giant branches, making it a suitable reference populationalso in these filters. Because the ACS field is smaller than the WFC3 field, we also expect ourRGB sample to be smaller, coming to a total of ∼ 2200 stars compared to the ∼ 3000 we hadbefore.3.4 ACS Data SelectionIn previous sections we mention the concern about the contamination of the HB by AGB andevolving BSS stars. In an attempt to try to identify the stars polluting the HB we select whatwe think is a cleaner sample of HB stars in the ACS CMD which we call faint HB. In orderto be able to compare the ACS and WFC3 data sets, the data for F225W and F336W werereduced to the same field as the the one covered by the F606W and F814W which is also inthe core but expands to a radius of only 105 arcseconds. In order for a star to be used in thisresearch, it had to be measured in all four filters.As in the WFC3 CMD we used MESA models to choose our regions which are shown inFigure 3.3. The detailed reasoning behind the division of faint and bright HB stars and thedifference between the AGB and the bump on this branch will be explained in chapters 5 and6. The important point for now is confirming the presence of non-HB stars on the HB of theUV CMD that can be pictured on the right panel of the figure. Using the same color code as onthe left panel we can see how bright HB, AGB and the stars on the bump of the AGB pickedon the ACS CMD, fall on the same region as the HB stars on the WFC3 CMD. Although welose some stars as we had to reduce our field to match the ACS field, we can identify wherethe AGB from the evolution of normal stars fall in the UV CMD and obtain a cleaner sampleof HB stars, leading to the proper classification of over 100 stars that we would have otherwiseneeded to ignore.22Figure 3.3: Left: F606W,F606W −F814W CMD with the selection of the stellar pop-ulations on the ACS data. Right: F225W,F225W −F336W CMD showing wherethe stars selected on the ACS data fall on the UV CMD. We can see a clear con-tamination to the UV HB by stars on stellar evolutionary stages different from thenormal HB but that form clear branches on the ACS CMD. The number of stars foreach sample is given in the inset in the left plot.23Chapter 4Estimating Masses Outside the MSAlthough our main focus is not mass segregation, the analysis of the BSS population in thecore of 47 Tuc requires that we have knowledge about the masses of the stars along the manyevolutionary stages and whether or not they have relaxed. As we mentioned in section 1.2,the high quality data and photometry have allowed us to go as faint as 6 magnitudes belowthe TO reaching a significant enough mass difference along the MS to be able to show masssegregation. Figure 4.1 shows the CMD of 47 Tuc in the UV, where we have highlightedthree MS regions with the corresponding masses at the center of each box based on an 11 GyrPARSEC isochrone (Bressan et al., 2012)1 constructed using the metallicity of 47 Tuc and thebolometric corrections of Chen et al. (2014). In order to fit the isochrone to the data, besides thedistance modulus ((m−M)0 = 13.36, Woodley et al. 2012) and reddening (E(B−V ) = 0.04,Salaris et al. 2007) it was necessary to add 0.4 and 0.3 magnitudes of extinction to F225Wand F336W respectively. The isochrone fits the CMD in F606W and F814W without anyadditional corrections. To the right of the CMD we have the radial distributions of the differentMS regions, we can see how the brightest and more massive MS stars are significantly morecentrally concentrated than the faintest sample, with the intermediate mass sample sitting inbetween. The observable difference between the distributions can be confirmed with a KS-testwhich yields p-values of the order of 10−21 or lower.We can now use the radial distributions to predict the masses for different groups of stars in47 Tuc. For each of the three MS regions we take the value of the distance from the center of thecluster where the cumulative distributions reach 20 and 50 percent, we call these distances R20and R50 respectively. Plotting the logarithmic values of each mass against their correspondingR20 and R50, we find a relationship for each R. The logarithmic values of mass (M) and R1Available at http://stev.oapd.inaf.it/cmd24Figure 4.1: Left: UV CMD of the core of 47 Tuc displaying the selection of three MSregions, upper, middle and lower MS, with the green arrow boxes showing the cor-responding masses at the center of each box based on an 11 Gyr PARSEC isochrone.Right: Radial distribution of the regions pointed out on the CMD following the samecolour pattern. The legend on the CMD has the number of stars before correcting forincompleteness, while the legend in the right plot gives the size of the sample aftercorrecting for incompleteness. Having the radial distributions of the more massiveMS stars more centrally concentrated than those with lower masses is evidence ofmass segregation in the core of 47 Tuc.follow a linear relation like the one in equation 4.1:log(Mass) = A× log(R)+B (4.1)As shown in Figure 4.2, we get the following relationships:log(MR20) =−0.83× log(R20)+1.14 (4.2)log(MR20) =−0.99× log(R50)+1.71 (4.3)for R20 and R50 respectively. Equations 4.2 and 4.3 are obtained by fitting a linear function25to the three points retrieved from the MS represented by the blue dots in Figure 4.2. The insetaccompanying the Figure shows the predicted masses for the bBSS, fBSS, MSBn, RGB andbright and faint WD stars at both R20 and R50. We can see the huge difference between themasses for faint (0.97 and 1.04 M) and bright (2.05 and 1.72 M) BSS. Table 4.2 shows thevalues of the masses together with their errors, these are dominated by Poissonian errors. Tocalculate the errors in our predicted masses, we need the error in R (R20 or R50), using R20as the example we calculate the errors using the following equation:error(log(R20)) =±log(r[NR20±√NR20])∓ log(r[NR20]) (4.4)where r[N] is the radius at the Nth star, for example r[NR20] means the radius at the star wherethe cumulative distribution reaches 20%. Then the error in the mass is just:error(log(MR20)) = mR20× error(log(R20)) (4.5)where mR20 is the slope of the fit for R20. Here we have neglected the errors in the determina-tion of the slope and the determination of the masses for the MS as they are very low and hadalmost no effect on the final error values.Looking at the values for the RGB and WDs we can see evidence of mass loss betweenthese two evolutionary stages. To see where this mass loss happens we need the masses of thestages in between the RGB and WD. As we noted before, the HB is contaminated in the UVCMD while the AGB is not identifiable. We then do the same mass prediction exercise limitingthe UV field to 105 arcseconds to match the ACS field and taking data from the ACS CMDwhere possible (normally we would use the matched stars between the two catalogues but asthe MS does not extent to faint magnitudes in the ACS data we do not restrict the selection ofMS stars to stars measured in all four filters but we do restrain it to the same size field. Samefor the WDs). The resulting fits are:log(MR20) =−1.30× log(R20)+1.73 (4.6)log(MR50) =−1.74× log(R50)+2.86 (4.7)The results are shown in Figure 4.3, and again the mass values and errors are found in table4.2.According to Heyl et al. (2015a), mass loss happens when the star is close to the tip ofthe AGB. As we mention before, the masses for the RGB and BWD show evidence of massloss between these stages of evolution. When we include the masses for the HB and AGBwe see that this drop in mass happens between the AGB and WDs, but we also see an smallincrease in the masses between the RGB and AGB that is consistent with the AGBs having26Figure 4.2: Relationship between log(M) and log(R) for R20 (red) and R50 (blue). Theequations have been obtained through a linear fit using the known masses for thethree regions of the MS (blue dots). The inset shows the predicted masses for thebBSS, fBSS, MSBn, RGB and bright and faint WDs stars at both R20 and R50, theerror values can be found in table 4.2.evolved from slightly more massive stars that ran out of hydrogen earlier than those formingthe current RGB. Two other things might explain this behaviour; one, the errors for the massesat this evolved stage are very large and might account for the extra mass; and two, is possiblethat, even after our efforts to make the HB and AGB as clean as possible, there is still somecontamination by evolved BSS.Another way to test where the mass loss happens is through the cumulative radial distri-butions. If mass loss does not happen until late in the AGB or is very low before this, thenthe radial distribution of the upper MS (UMS), RGB, HB, AGB and BWD should be similar.Looking at Figure 4.4 we can see that the distributions of the aforementioned populations lookvery similar, something that we can confirm through KS-tests which yields p-values over 0.40for any combination of the four regions and > 10−3 for all the samples against the FWD. Thedetailed results of the KS-tests for the regions selected in Figure 4.4 are presented in table 4.1.Also, if there is mass loss that we could notice between the RGB and HB, then the distributionof the HB should be similar to that of stars of lower masses (not necessarily as low as the27Figure 4.3: Relationship between log(M) and log(R). Similar to Figure 4.2 but with thefield reduced to a radius of 105 arcseconds which is the limit for the ACS field. Theerror values for the masses can also be found in table 4.2.FWD). Comparing the radial distribution of the HB against the different MS regions used tobuild the fits for the masses, we confirm that the HB is only related to the UMS, while the dif-ference gets bigger as we go to lower masses with p-values of ∼ 10−4 for MMS (0.74M) and∼ 10−11 for the lower MS (LMS, 0.65M) region. Because the HB lasts for a few relaxationtimes (Heyl et al., 2015b), we can exclude a mass loss of greater than 0.1M on the RGB atnearly the four-sigma level.Table 4.1: KS-test p-value results between the populations selected on Figure 4.4. Thenumbers show that every stellar population could have been drawn from the samesample except for the FWD.UMSRGB 0.91HB 0.40 0.41AGB 0.98 0.89 0.80BWD 0.68 0.60 0.88 0.85FWD ∼ 10−3 ∼ 10−3 ∼ 10−12 ∼ 10−19 ∼ 10−2828Figure 4.4: CMD (left panel) and radial distribution (right panel) including 5 evolution-ary stages: UMS, RGB, HB, AGB and WD, this last one divided into faint andbright WDs. The selection of the stars for the UMS, RGB and WDs are taken di-rectly from the UV CMD, while the samples for the HB and AGB have been donethrough the ACS data. In this case, not all the stars have a counter part on theF606W,F606W −F814W CMD as the WDs were not detected with the filters onthe visible range. Instead the data was reduced to the same field in order to comparetheir radial distribution. The colors for the regions on the CMD are the same as inthe left plot and specified on the legend.29Table 4.2: Results for the mass prediction in both the WFC3 complete 160 arcsecondsfield and the reduced ACS field (105 arcseconds). The masses were calculated usingequations 4.2, 4.3, 4.6 and 4.7, and the errors with equations 4.4 and 4.5.R≤ 160 R≤ 105MR20 Error MR50 Error MR20 Error MR50 ErrorRGB 0.87+0.030.85+0.020.84+0.040.78+0.04-0.02 -0.02 -0.03 -0.03HB N/A N/A 0.90+0.090.83+0.09-0.09 -0.14AGB N/A N/A 1.05+0.170.87+0.27-0.10 -0.31BWD 0.74+0.090.74+0.100.77+0.110.77+0.06-0.08 -0.08 -0.12 -0.10FWD 0.56+0.010.63+0.020.48+0.120.60+0.03-0.01 -0.01 -0.03 -0.02MSBn 1.28+0.061.14+0.24N/A N/A-0.09 -0.13fBSS 0.97+0.121.04+0.12N/A N/A-0.36 -0.27bBSS 2.05+0.091.72+0.31N/A N/A-0.34 -0.60Bump N/A N/A 1.85+0.432.83+1.72-0.40 -0.86BSS 1.78+0.301.39+0.242.55+0.391.73+0.23-0.17 -0.20 -0.45 -0.5030Chapter 5Results5.1 Blue StragglersWe have already pointed out the difference between the masses of faint and bright BSS. Whenlooking at the cumulative radial distribution of the two regions of BSS in Figure 5.1 we can alsosee a significant difference between the two samples. According to the KS-test results, faint andbright BSS are only 1.0% likely to be drawn from the same population and are significantlydifferent from the reference population with p-values of 0.03 for fBSS and ∼ 10−6 for thebBSS.Doing a visual examination of the radial distribution plot we noticed a similarity betweenthe fBSS and the MSBn. This is confirmed by the KS-test which yields a p-value of 0.76suggesting a relation between these two groups of stars that it is not present between the bBSSand MSBn (p-value = 0.01). The KS-test results between the regions highlighted in Figure 5.1are summarized in table Evolved Blue StragglersDifferentiating the various evolutionary stages on the F225W,F225W −F336W CMD afterthe RGB is complicated. Although the HB seems to be clear, the number of stars and theradial distribution of this branch disagree with the models suggesting an over abundance ofTable 5.1: KS-test p-value results between the populations selected on Figure 5.1.bBSSfBSS 0.01MSBn 0.76 0.01RGB ∼ 10−5 0.03 ∼ 10−631Figure 5.1: Left:F225W,F225W −F336W CMD showing the selected stars for the faintand bright BSS, RGB and MSBn. Right: Radial distributions for the selected sam-ples. The legend on the CMD has the number of stars before correcting for in-completeness, while the legend on the right plot gives the size of the sample aftercorrecting for incompleteness.stars. On Figure 5.2, left panel, we can see the upper part of the CMD along with four MESAevolutionary models. The lowest mass model, 0.85M, shows the evolution for a star witha mass approximately equal to the TO mass. According to this model the RGB lasts for ∼4.2×108 years while the HB only∼ 0.7×108 years. This means that we would expect around510 HB stars but instead we get a few more than 700. If we do the same exercise but now fromthe HB to the AGB (15 Myr) we expect an extra 110 stars contaminating the HB, coming to atotal of ∼ 620 expected stars in this region.The time scales for the regions outlined on Figure 5.2, along with the numbers for observedand expected stars are presented in table 5.4. Looking at this table, we can see we have includedthe numbers for the sub-giant branch (SGB). This region, not shown on Figure 5.2 but on theCMD it would extend in the same colour range as the adjacent side of the green region, isincluded to support our theory that the RGB in these filters is not contaminated by evolvedBSS. Doing the calculations to estimate the expected number of stars on the RGB we get a32small difference of only 2.8% with the number of stars observed.Figure 5.2: Left:F225W,F225W −F336W colour-magnitude diagram (CMD), the pinkand green curves are MESA evolutionary models for stars with initial masses of0.85M, 1.1M, 1.4M, and 1.8M from bottom to top. Right: Radial distributionsfor the selected samples on the CMD. As on previous figures, the legend on theCMD has the number of stars before correcting for incompleteness, while the legendin the right plot gives the size of the sample after correcting for incompleteness.Even though we do not see a TO point for the BSS (given that BSS, even of the same mass,can leave the MS at different times depending on the evolutionary stage of the stars that createdit), if we follow the path of the models for the BSS of different masses in Figure 5.2, we noticea region between the SGB, RGB, BSS, and HB highlighted in green, where we would expectmainly stars that have evolved from a BSS (because of the likelihood of blends, especially inthe region right next to the SGB, we have included a mild error cut in the magnitude of thissample). Calculating the time the 1.4M model takes to go from when it leaves the MS tobefore it reaches the region where the normal stars and the evolved BSS share the same CMDspace, and the time it takes the star to evolve from this point up to an AGB star, we get a time of200 Myr for both sections. Considering these two time scales we would expect a contaminationof 100 stars to the HB and surrounding regions.Looking now at the at the radial distributions for the four coloured regions, the BSS distri-33bution looks similar to that of which we are calling evolved BSS. Table 5.2 shows the p-valuesobtained from KS-test performed between the different populations. These results support theidea that both distributions, BSS and evolved BSS, were drawn from the same population witha p-value of 0.98. On the other hand, the HB distribution looks visually similar to that of theRGB but the KS-test rejects the hypothesis of these coming from the same distribution witha p-value of ∼ 10−3. Looking closely we can see that the HB star distribution appears to beslightly more centrally concentrated than the RGB stars. Doing the same for the evolved BSSwe can also reject these stars coming from the same sample as the RGB or HB with p-valuesfor the KS-test of 0.01 and 0.03.To further expand the study and identification of evolved BSS, we now compare our datato photometry obtain from the ACS data. We can see in Figure 5.3 that pointing out the AGBon the ultraviolet CMD is almost impossible, but it becomes much easier on the ACS data,especially at the fainter end of the AGB. By cross-matching the stars from Sarajedini’s CMDto ours, we can identify the AGB stars and obtain a cleaner sample for the radial distribution.Again using this smaller data set we see that the number of stars on the HB and AGB are notconsistent with the models. According to table 5.5 we expect 370 HB stars but we observe∼ 410, for the AGB the numbers would be 80 and ∼ 100 counting the AGB plus the bumphighlighted in blue in Figure 5.3. In summary, taking into account the normal evolution ofstars, we would expect to observe 450 stars from the HB to the observable part of the AGB butwe count 510.Going back to the evolved BSS, using the MESA models on the ACS CMD, we can nowpoint out where the RGB and HB of the BSS fall on the CMD. The first thing we notice isthat the HB for the evolved BSS is brighter than the HB for the normal stars which makes itreasonable to split this population between faint and bright HB. Another interesting result isthe fact that the RGB bump for the evolved BSS falls in the same region where the AGB bumpwas thought to be. This is the reason we have separated this group of stars from the rest of theAGB for further inspection. In this bump alone we count 41 stars. If this is indeed the RGBbump of evolved BSS we would expect to have around 20 stars coming from the evolution ofBSS, which means that at least half the stars in this bump are actually evolved BSS and notAGB stars. We must mention that the numbers of expected stars for the 1.8M model obtainedby starting with a star count of 80 stars on the green region is unrealistic. As we can see onthe UV CMD, there are only 4 stars just above the 1.4M model, if we take this number as theactual observed count of stars we would expect a total of only 1 and 4 stars for the HB andAGB respectively.Before splitting the HB and AGB (HB in faint and bright HB, and AGB in AGB and AGBbump) both radial distributions look more centrally concentrated than the RGB. With the sam-ples separated as explained above, we compare the radial distributions for all the populations34highlighted in Figure 5.3 and summarize the results in table 5.3. The bright HB, AGB bumpand the BSS distributions look very similar and KS-test results show that we cannot reject thepossibility that all three samples are drawn from the same population with p-values of 0.77for BSS against bright HB and 0.58 for BSS versus AGB bump. Is also important to mentionthat using the same statistic we can reject the hypothesis that the AGB and the AGB bump aredrawn from the same distribution with a p-value of 0.04 and the same for the fHB and bHBwith a p-value of 0.08.For completeness, we can confirm that in this reduced sample the KS-test p-value betweenthe radial distributions of the BSS and the evolved BSS still does not allow us reject the ideathat they come from the same population with an even higher result of 0.99. KS-test resultsalso point towards the evolved BSS being related to the bHB and AGB bump with p-values of0.88 and 0.64 respectively.Separating the HB and AGB has also help us make more sense of the normal evolution ofstars in the cluster. Now the cumulative radial distributions of the RGB, fHB and AGB aremore closely related with p-values of 0.40 for RGB vs. fHB, 0.89 for RGB vs. AGB, and 0.98for fHB vs. AGB.Table 5.2: KS-test p-value results between the populations selected on Figure 5.2BSSeBSS 0.98RGB ∼ 10−4 ∼ 10−5HB ∼ 10−3 0.03 0.01Table 5.3: KS-test p-value results between the populations selected on Figure 5.3BSSeBSS 0.99RGB ∼ 10−4 ∼ 10−5fHB 0.40 ∼ 10−3 ∼ 10−4bHB 0.08 0.04 0.88 0.77AGB 0.17 0.98 0.89 0.06 0.02Bump 0.04 0.36 ∼ 10−4 ∼ 10−3 0.64 0.5835Table 5.4: Times spent in different regions of the CMD according to the MESA modelsfor the regions selected in Figure 5.2, the evolutionary stage named green makesreference to the green region on the UV CMD refer to as Evolved BSS on the WFC3CMD. HB for the BSS models (1.1M,1.4M and 1.8M), corresponds to the timethe stars spend since the end of the green region until the end of the AGB. Theexpected number for HB stars from UV CMD counts the stars expected from the HBand AGB.Models0.85M 1.1M 1.4M 1.8MSGBTime (Myr) 415 171 115 54WFC3Obs 2940 N/AExp N/AGreenTime (Myr) N/A 319 195 54WFC3Obs N/A 100Exp N/ARGB Time (Myr) 420 340 220 84WFC3Obs 3060 N/AExp 2975 ∼ 100HB+AGB Time (Myr) 85 182 180 160WFC3Obs 710 N/AExp 620 ∼ 10036Figure 5.3: Top-left: F606W,F606W − F814W (V,V − I) colour-magnitude diagram(CMD) of the core (with r ≤ 105′′) of 47 Tucanae with the same MESA modelsas Figure 5.2. Top-right: F225W,F225W −F336W (U,U −B) CMD for the sameregion. In both CMDs stars represented by triangles mean they have been selectedon the V,V − I CMD, coloured circles on the U,U −B. Different colours indicatedifferent populations as indicated on the legend of the bottom plot. Bottom: Theradial distributions of the different selected evolutionary stages.37Table 5.5: Times spent in different regions of the CMD according to the MESA modelsfor the regions selected from the ACS CMD as shown in Figure 5.3. The green regionstars were chosen from the WFC3 data. The count of stars observed on the AGB forthe 0.85M model includes the stars from the bump highlighted in blue on Figure 5.3.The ages for the AGB on the BSS models (1.1M,1.4M and 1.8M) are calculatedwithin the same magnitude range as the 0.85M. The numbers for the bump regionmake reference to the RGB bump from the evolution of normal stars for the 0.85Mmodel, while for the BSS models the observed number is the number of stars in theAGB bump. The number of expected stars for the 1.8M models are biased by thenumber of stars in the green region, using the actual number of stars above the 1.4Mmodel on the CMD only 1 star on the HB and 4 on the AGB would be expected.Models0.85M 1.1M 1.4M 1.8MGreenTime (Myr) N/A 319 195 54WFC3Obs N/A 80Exp N/ARGB Time (Myr) 420 340 220 84ACSObs 2200 N/AExp N/A 85 90 120Bump Time (Myr) 35 40 44 20ACSObs 180 40Exp 185 11 18 28HBTime (Myr) 70 68 60 51ACSObs 410 25Exp 370 18 25 72AGB Time (Myr) 15 15 20 30ACSObs 100 ∼ 10Exp 80 ∼ 838Chapter 6Discussion6.1 Blue StragglersThe results reported in section 5.1 indicate there are two distinct BSS sequences present withina radius of 160 arcseconds from the center of 47 Tuc. The p-value of 0.01 obtained for theKS-test between the faint and bright BSS confirms that the two populations have different dis-tributions therefore come from different samples, suggesting different formation mechanisms.Previous studies, including the BSS in the core of 47 Tuc, have also argued in favour of morethan one formation mechanisms (Mapelli et al., 2004, 2006; Monkman et al., 2006) going on inthis GC; primary stellar evolution and direct collisions. Further more, the estimated masses forfaint and bright BSS are also different, with the bright BSS considerably more massive than thefaint ones when using R20. The difference in masses become less obvious when we considerR50 especially taking the large errors into account.The bright BSS are very centrally concentrated (20% of the bright BSS are within 10arcseconds from the center, and the distribution reaches 50% at only ∼ 30 arcseconds) andtheir cumulative radial distribution does not resemble any of the other populations identifiableon the CMD. This prevents us from linking their formation to an specific group of stars. On theother hand, the mass estimate of 2.05M at R20 for the bright BSS, indicates that they mustcome from the interactions of at least three stars, possibly through the evolution of hierarchicaltriple systems or encounters involving more than two stars.In contrast, the faint BSS are less segregated towards the center but still more concentratedthan most of the other populations. Their cumulative radial distribution looks very similar tothat of the MSBn (5.1), confirmed by the 0.76 p-value obtained for the KS-test. The resem-blance of their radial distributions points to a binary origin for the faint BSS. The estimatedmasses of these two populations are also similar, with the faint BSS being a little less massivethat the MSBn as could be expected for a final product of binary evolution.396.2 Evolved Blue StragglersBefore separating the bright and faint HB stars, the combined HB distribution looks morecentrally concentrated than that of the RGB, a difference that is sustained by the KS-test results:∼ 99% probability that the distributions were taken from different samples. According to masssegregation, for a population of stars to be more centrally concentrated would need to be moremassive. This fact contradicts the models and theory about stellar evolution where some massloss is expected between the RGB and HB. Although recent results indicate that the bulk ofthe mass loss happens when the star is closer to the tip of the AGB (Heyl et al. 2015a, andreferences within), there is no evidence of mass gain or any other process that could lead tothe HB stars being significantly more massive than the RGBs. The models superposed on theCMD in Figure 5.2 show that the HB and AGB of the BSS happens to the right of the normalHB, but there are no stars in that part of the CMD. This is due to saturation of the images inthe F336W filter at a magnitude of 15.25, which makes the color of any star with a magnitudeabove the saturation level pushed below the saturation line, in this case into the same CMDposition of the HB for the evolution of normal stars. The contamination to the HB can alsobe noticed by just counting the stars in that region, which gives an observed number of starsmuch higher than expected. This overabundance of stars can be justified if we add the expectednumber of HB stars from normal evolution to the expected number of evolved BSS stars in thatpart of the CMD by using the time scales derived in section 5.2 (see table 5.4).The calculation of the number of BSS contaminating the HB was possible through theidentification of evolved BSS going though their SGB and beginning of the RGB. The ideacame from the fact that there were many stars in a part of the CMD that should not be verypopulated if we consider stars going through the normal evolution, but once we plotted the BSSmodels they appear in the right place. The relation of these stars to the BSS was confirmed bytheir radial distributions with p-value of 0.98 for the KS-test comparing their radial distributionwith those of the BSS. Interestingly, when we plot the stars in this green region on the ACSdata we find they lie very close to, if not on top of, the SGB and RGB of the evolution ofnormal stars, close to the portion of the CMD that Beccari et al. (2006) identified to try to findBSS starting their RGB phase. Selecting these stars on the UV CMD allows us to obtain acleaner sample with a much lower chance of selecting normal RGB stars.The number of BSS compared to that of the evolved BSS in what we call the green region(see Figure 5.2) suggests a short MS lifetime of ∼ 200− 300 Myr. This result disagrees withthose found by Sills et al. (2000) and Chatterjee et al. (2013) (between 1 and 3 Gyr). Thiswould be possible considering the formation mechanism dominating in the core are collisionsand the MS stars in the core are the more massive ones (i.e. close to the TO), which as wementioned in section 1.3.2 would leave a BSS with close to the same amount of hydrogen inthe core as a TO MS star but with more mass around it speeding up the burning process.40The same overabundance of HB stars is observed when analysing the ACS CMD. In thiscase when we superpose the models we noticed that the HB for the BSS is brighter than theHB for the normal evolution of stars. In fact, when we split the HB into faint and bright HB,the numbers of observed and expected stars agree. As noted by Beccari et al. (2006), we alsofind that the distribution of the bright HB and that of the BSS are likely drawn from the samepopulation while the faint HB distribution resembles that of the RGB. Although the result wasexpected, we have used it to obtained a clean HB sample in the UV CMD that has allowed usto confirm the contamination of the UV HB and the predictions from our models.From the ACS data, we can see that the overabundance of stars on the HB also extendsto the AGB. Again we explained this extra population of stars by adding up the number ofexpected stars for the AGB and evolved BSS. According to the numbers reported we expectat least half the stars on the AGB plus AGB bump to be evolved BSS. The fact that we canstatistically state that the AGB and AGB bump do not come from the same distribution butthe BSS and AGB bump most likely do, supports our assumption that this bump is mostlypopulated by evolved BSS going up the RGB for the first time, more specifically our BSSmodels place the RGB bump for the BSS in the same region of the AGB bump. Having theradial distribution of the AGB without the bump agreeing with that of the RGB tells us thispart of the CMD is dominated by stars coming from the evolution of normal stars. This excessof stars in the AGB was studied earlier by Bailyn (1994) and Beccari et al. (2006), who cameto two different conclusions. The first paper suggested that this excess was due to BSS goingthrough their HB stage, but according to our models and as stated in the second paper, the HBof the BSS is much fainter than the AGB bump. Beccari et al., on the other hand, relates thiscontamination to the “high-mass binary by-products currently ascending the RGB for the firsttime”. Our results are in good agreement with the second paper, but we have also been able toconstrain the bulk of the contamination to the AGB bump as due to the RGB bump of the BSS.41Chapter 7ConclusionWe have identified a large sample of over 200 BSS and evolved BSS in the UV data of the coreof 47 Tuc. Expanding our research using available data on the visible range we have studiedthe properties of this population including their masses, possible formation mechanisms, andtheir evolution.When we separate the bright and faint BSS we find that the bright BSS show a much morecentrally concentrated radial distribution and higher mass estimates, properties that suggestan origin involving triple or multiple stellar systems. In contrast, the faint BSS are less con-centrated, with a radial distribution similar to the MSBn pointing to this populations as theirprogenitors.Distinguishing a purely evolved BSS sample from a CMD, had, until now, only been at-tempted on the HB. The evolved BSS selected on the UV CMD along with the MESA modelsand the agreement between the radial distributions of the BSS, evolved BSS, bright HB, andAGB bump, allowed us to construct the story of the evolution of BSS. The time scales andnumber of observed and expected stars agree nicely with the BSS having a post-MS evolutioncomparable to that of a normal star of the same mass. The disagreement between our estimatedMS lifetime and those found by others indicate that a more detailed study of individual BSSproperties is necessary to constrain these values.We have also been able to select clean samples in the different stellar evolutionary stagesfor the normal evolution of stars. Here we find that the cumulative radial distributions for theupper MS, RGB, faint HB and AGB (without the bump), seem to all come from the samesample as expected for stars of the same mass. It is important to mention that in both the AGBand the AGB bump, we might find stars from the evolution of normal stars as well as thosecoming from the evolution from BSS, but the number of stars and their radial distributionshave allowed us to state the dominant population in each sample.Future studies using high quality spectra, will tell us more about the formation and evolu-42tion of BSS. Each formation mechanism leaves BSS with different chemical properties and pos-sible companions, both of which could be identified and characterised through spectroscopy.43BibliographyAhumada, J. A., & Lapasset, E. 2005, Boletin de la Asociacion Argentina de Astronomia LaPlata Argentina, 48, 177 → pages 7Anderson, A. J. 1997, PhD thesis, University of California, Berkeley → pages 7Anderson, J., Sarajedini, A., Bedin, L. R., King, I. R., Piotto, G., Reid, I. N., Siegel, M.,Majewski, S. R., Paust, N. E. Q., Aparicio, A., Milone, A. P., Chaboyer, B., & Rosenberg,A. 2008, The Astronomical Journal, 135, 2055 → pages 18Antonini, F., Chatterjee, S., Rodriguez, C. L., Morscher, M., Pattabiraman, B., Kalogera, V.,& Rasio, F. A. 2015, ArXiv e-prints, arXiv:1509.05080 → pages 8, 11Bailyn, C. D. 1994, The Astronomical Journal, 107, 1073 → pages 14, 41Beccari, G., & Carraro, G. 2015, in Astrophysics and Space Science Library, Volume 413,Ecology of Blue Straggler Stars, ed. H. M. J. Boffin, G. Carraro, & G. Beccari (Springer),1, doi:10.1007/978-3-662-44434-4 1 → pages 4Beccari, G., Ferraro, F. R., Lanzoni, B., & Bellazzini, M. 2006, The Astrophysical JournalLetters, 652, L121 → pages 14, 40, 41Benz, W., & Hills, J. G. 1987, The Astrophysical Journal, 323, 614 → pages 10Binney, J., & Merrifield, M. 1998, Galactic Astronomy (Princeton University Press) → pages4Binney, J., & Tremaine, S. 2008, Galactic Dynamics: Second Edition (Princeton UniversityPress) → pages 5, 6, 7Boffin, H. M. J., Carraro, G., & Beccari, G. 2015, Ecology of Blue Straggler Stars (Springer),doi:10.1007/978-3-662-44434-4 → pages 8Bohlin, R. C., & Gilliland, R. L. 2004, The Astronomical Journal, 127, 3508 → pages 16Bressan, A., Marigo, P., Girardi, L., Salasnich, B., Dal Cero, C., Rubele, S., & Nanni, A.2012, Monthly Notices of the Royal Astronomical Society, 427, 127 → pages 24Chatterjee, S., Rasio, F. A., Sills, A., & Glebbeek, E. 2013, The Astrophysical Journal, 777,106 → pages 11, 4044Chen, Y., Girardi, L., Bressan, A., Marigo, P., Barbieri, M., & Kong, X. 2014, MonthlyNotices of the Royal Astronomical Society, 444, 2525 → pages 24Christensen-Dalsgaard, J. 2015, Monthly Notices of the Royal Astronomical Society, 453,666 → pages 4Collins, G. W. I. 1989, The fundamentals of stellar astrophysics (New York: W. H. Freemanand Co.) → pages 2, 4Davies, M. B. 2015, in Astrophysics and Space Science Library, Volume 413, Ecology ofBlue Straggler Stars, ed. H. M. J. Boffin, G. Carraro, & G. Beccari (Springer), 203,doi:10.1007/978-3-662-44434-4 9 → pages 9, 10, 11, 12de Marchi, F., de Angeli, F., Piotto, G., Carraro, G., & Davies, M. B. 2006, Astronomy &Astrophysics, 459, 489 → pages 7De Marco, O., Shara, M. M., Zurek, D., Ouellette, J. A., Lanz, T., Saffer, R. A., & Sepinsky,J. F. 2005, The Astrophysical Journal, 632, 894 → pages 8Fabrycky, D., & Tremaine, S. 2007, The Astrophysical Journal, 669, 1298 → pages 10, 11Ferraro, F. R., Beccari, G., Rood, R. T., Bellazzini, M., Sills, A., & Sabbi, E. 2004, TheAstrophysical Journal, 603, 127 → pages 13, 21Ferraro, F. R., Sills, A., Rood, R. T., Paltrinieri, B., & Buonanno, R. 2003, The AstrophysicalJournal, 588, 464 → pages 21Ferraro, F. R., Paltrinieri, B., Fusi Pecci, F., Cacciari, C., Dorman, B., Rood, R. T., Buonanno,R., Corsi, C. E., Burgarella, D., & Laget, M. 1997, Astronomy & Astrophysics, 324, 915→ pages 13Ferraro, F. R., Beccari, G., Dalessandro, E., Lanzoni, B., Sills, A., Rood, R. T., Pecci, F. F.,Karakas, A. I., Miocchi, P., & Bovinelli, S. 2009, Nature, 462, 1028 → pages 13Fiorentino, G., Lanzoni, B., Dalessandro, E., Ferraro, F. R., Bono, G., & Marconi, M. 2014,The Astrophysical Journal, 783, 34 → pages 8Fiorentino, G., Marconi, M., Bono, G., Dalessandro, E., Ferraro, F. R., Lanzoni, B., Lovisi,L., & Mucciarelli, A. 2015, The Astrophysical Journal, 810, 15 → pages 8Fregeau, J. M., Cheung, P., Portegies Zwart, S. F., & Rasio, F. A. 2004, Monthly Notices ofthe Royal Astronomical Society, 352, 1 → pages 11Fruchter, A., & Sosey, M. 2009, HST MultiDrizzle Handbook (Space Telescope ScienceInstitute) → pages 15Gilliland, R. L., Bono, G., Edmonds, P. D., Caputo, F., Cassisi, S., Petro, L. D., Saha, A., &Shara, M. M. 1998, The Astrophysical Journal, 507, 818 → pages 8Harris, W. E. 1996, The Astronomical Journal, 112, 1487, 2010 edition → pages 1, 5, 6, 1745Heggie, D. C. 1975, Monthly Notices of the Royal Astronomical Society, 173, 729 → pages12Hesser, J. E., Harris, W. E., & Vandenberg, D. A. 1987, Publications of the AstronomicalSociety of the Pacific, 99, 1148 → pages 8Heyl, J., Kalirai, J., Richer, H. B., Marigo, P., Antolini, E., Goldsbury, R., & Parada, J. 2015a,The Astrophysical Journal, 810, 127 → pages iii, 26, 40Heyl, J., Richer, H. B., Antolini, E., Goldsbury, R., Kalirai, J., Parada, J., & Tremblay, P.-E.2015b, The Astrophysical Journal, 804, 53 → pages iii, 6, 16, 28Hilditch, R. W., Howarth, I. D., & Harries, T. J. 2005, Monthly Notices of the RoyalAstronomical Society, 357, 304 → pages 17Hills, J. G., & Day, C. A. 1976, Astrophysical Letters, 17, 87 → pages 10Hjellming, M. S., & Taam, R. E. 1991, The Astrophysical Journal, 370, 709 → pages 10Iben, Jr., I., & Tutukov, A. V. 1999, in Astronomical Society of the Pacific Conference Series,Vol. 169, 11th European Workshop on White Dwarfs, ed. S.-E. Solheim & E. G. Meistas,432 → pages 11Ivanova, N. 2015, in Astrophysics and Space Science Library, Volume 413, Ecology of BlueStraggler Stars, ed. H. M. J. Boffin, G. Carraro, & G. Beccari (Springer), 179,doi:10.1007/978-3-662-44434-4 8 → pages 9Kalirai, J. S., Richer, H. B., Anderson, J., Dotter, A., Fahlman, G. G., Hansen, B. M. S.,Hurley, J., King, I. R., Reitzel, D., Rich, R. M., Shara, M. M., Stetson, P. B., & Woodley,K. A. 2012, The Astronomical Journal, 143, 11 → pages iii, 15Kippenhahn, R., & Weigert, A. 1967, Zeitschrift fur Astrophysik, 65, 251 → pages 8Knigge, C., Gilliland, R. L., Dieball, A., Zurek, D. R., Shara, M. M., & Long, K. S. 2006, TheAstrophysical Journal, 641, 281 → pages 14Knigge, C., Leigh, N., & Sills, A. 2009, Nature, 457, 288 → pages 12Kozai, Y. 1962, The Astronomical Journal, 67, 591 → pages 11Lanzoni, B., Dalessandro, E., Perina, S., Ferraro, F. R., Rood, R. T., & Sollima, A. 2007, TheAstrophysical Journal, 670, 1065 → pages 13Leigh, N., Knigge, C., Sills, A., Perets, H. B., Sarajedini, A., & Glebbeek, E. 2013, MonthlyNotices of the Royal Astronomical Society, 428, 897 → pages 12Leonard, P. J. T. 1993, in Bulletin of the American Astronomical Society, Vol. 25, AmericanAstronomical Society Meeting Abstracts, 1329 → pages 11Lidov, M. L. 1962, Planetary and Space Science, 9, 719 → pages 1146Lombardi, Jr., J. C., Rasio, F. A., & Shapiro, S. L. 1996, The Astrophysical Journal, 468, 797→ pages 10Mapelli, M., Sigurdsson, S., Colpi, M., Ferraro, F. R., Possenti, A., Rood, R. T., Sills, A., &Beccari, G. 2004, The Astrophysical Journal Letters, 605, L29 → pages 13, 39Mapelli, M., Sigurdsson, S., Ferraro, F. R., Colpi, M., Possenti, A., & Lanzoni, B. 2006,Monthly Notices of the Royal Astronomical Society, 373, 361 → pages 13, 39McCrea, W. H. 1964, Monthly Notices of the Royal Astronomical Society, 128, 147 → pages8McLaughlin, D. E., Anderson, J., Meylan, G., Gebhardt, K., Pryor, C., Minniti, D., &Phinney, S. 2006, The Astrophysical Journal Supplement Series, 166, 249 → pages 14Meylan, G. 2000, in Astronomical Society of the Pacific Conference Series, Vol. 211, MassiveStellar Clusters, ed. A. Lanc¸on & C. M. Boily, 215 → pages 5, 7Milone, A., Piotto, G., Bedin, L. R., King, I. R., Anderson, J., Marino, A. F., Bellini, A.,Gratton, R., Renzini, A., Stetson, P. B., Cassisi, S., Aparicio, A., Bragaglia, A., Carretta, E.,D’Antona, F., Di Criscienzo, M., Lucatello, S., Monelli, M., & Pietrinferni, A. 2012a, TheAstrophysical Journal, 744, 58 → pages 5Milone, A. P., Piotto, G., Bedin, L. R., & Sarajedini, A. 2008, Memorie della SocietAstronomica Italiana, 79, 623 → pages 12Milone, A. P., Piotto, G., Bedin, L. R., Aparicio, A., Anderson, J., Sarajedini, A., Marino,A. F., Moretti, A., Davies, M. B., Chaboyer, B., Dotter, A., Hempel, M., Marı´n-Franch, A.,Majewski, S., Paust, N. E. Q., Reid, I. N., Rosenberg, A., & Siegel, M. 2012b, Astronomy& Astrophysics, 540, A16 → pages 12Monkman, E., Sills, A., Howell, J., Guhathakurta, P., de Angeli, F., & Beccari, G. 2006, TheAstrophysical Journal, 650, 195 → pages 13, 39Naoz, S., & Fabrycky, D. C. 2014, The Astrophysical Journal, 793, 137 → pages 11Paresce, F., Meylan, G., Shara, M., Baxter, D., & Greenfield, P. 1991, Nature, 352, 297 →pages 13Paxton, B., Bildsten, L., Dotter, A., Herwig, F., Lesaffre, P., & Timmes, F. 2011, TheAstrophysical Journal Supplement Series, 192, 3 → pages 2Paxton, B., Cantiello, M., Arras, P., Bildsten, L., Brown, E. F., Dotter, A., Mankovich, C.,Montgomery, M. H., Stello, D., Timmes, F. X., & Townsend, R. 2013, The AstrophysicalJournal Supplement Series, 208, 4 → pages 2Paxton, B., Marchant, P., Schwab, J., Bauer, E. B., Bildsten, L., Cantiello, M., Dessart, L.,Farmer, R., Hu, H., Langer, N., Townsend, R. H. D., Townsley, D. M., & Timmes, F. X.2015, The Astrophysical Journal Supplement Series, 220, 15 → pages 247Perets, H. B. 2015, in Astrophysics and Space Science Library, Volume 413, Ecology of BlueStraggler Stars, ed. H. M. J. Boffin, G. Carraro, & G. Beccari (Springer), 251,doi:10.1007/978-3-662-44434-4 11 → pages 8, 9, 10, 11, 12Perets, H. B., & Fabrycky, D. C. 2009, The Astrophysical Journal, 697, 1048 → pages 10, 11Piotto, G., De Angeli, F., King, I. R., Djorgovski, S. G., Bono, G., Cassisi, S., Meylan, G.,Recio-Blanco, A., Rich, R. M., & Davies, M. B. 2004, The Astrophysical Journal Letters,604, L109 → pages 12Pribulla, T., & Rucinski, S. M. 2006, The Astronomical Journal, 131, 2986 → pages 11Richer, H. B., Heyl, J., Anderson, J., Kalirai, J. S., Shara, M. M., Dotter, A., Fahlman, G. G.,& Rich, R. M. 2013, The Astrophysical Journal Letters, 771, L15 → pages 5Salaris, M., Held, E. V., Ortolani, S., Gullieuszik, M., & Momany, Y. 2007, Astronomy &Astrophysics, 476, 243 → pages 24Sandage, A. R. 1953, The Astronomical Journal, 58, 61 → pages 1Sanna, N., Dalessandro, E., Lanzoni, B., Ferraro, F. R., Beccari, G., & Rood, R. T. 2012,Monthly Notices of the Royal Astronomical Society, 422, 1171 → pages 13Santana, F. A., Mun˜oz, R. R., Geha, M., Cote´, P., Stetson, P., Simon, J., & Djorgovski, S. G.2012, in Astronomical Society of the Pacific Conference Series, Vol. 458, GalacticArchaeology: Near-Field Cosmology and the Formation of the Milky Way, ed. W. Aoki,M. Ishigaki, T. Suda, T. Tsujimoto, & N. Arimoto, 339 → pages 7Santucci, R. M., Placco, V. M., Rossi, S., Beers, T. C., Reggiani, H. M., Lee, Y. S., Xue,X.-X., & Carollo, D. 2015, The Astrophysical Journal, 801, 116 → pages 7Sarajedini, A., Bedin, L. R., Chaboyer, B., Dotter, A., Siegel, M., Anderson, J., Aparicio, A.,King, I., Majewski, S., Marı´n-Franch, A., Piotto, G., Reid, I. N., & Rosenberg, A. 2007,The Astronomical Journal, 133, 1658 → pages 18Shara, M. M., Saffer, R. A., & Livio, M. 1997, The Astrophysical Journal Letters, 489, L59→ pages 7Sills, A. 2010, in American Institute of Physics Conference Series, Vol. 1314, AmericanInstitute of Physics Conference Series, ed. V. Kalogera & M. van der Sluys, 105–112 →pages 9, 10Sills, A., Bailyn, C. D., Edmonds, P. D., & Gilliland, R. L. 2000, The Astrophysical Journal,535, 298 → pages 13, 40Sills, A., Glebbeek, E., Chatterjee, S., & Rasio, F. A. 2013, The Astrophysical Journal, 777,105 → pages 12Sollima, A., Lanzoni, B., Beccari, G., Ferraro, F. R., & Fusi Pecci, F. 2008, The AstrophysicalJournal, 481, 701 → pages 1248Sparke, L. S., & Gallagher, III, J. S. 2007, Galaxies in the Universe: An Introduction(Cambridge University Press) → pages 3Spitzer, L. 1987, Dynamical evolution of globular clusters (Princeton University Press) →pages 6Stetson, P. 1994, Publications of the Astronomical Society of the Pacific, 106, 250 → pages16Stetson, P. B. 1987, Publications of the Astronomical Society of the Pacific, 99, 191 → pages15Thompson, I. B., Kaluzny, J., Rucinski, S. M., Krzeminski, W., Pych, W., Dotter, A., &Burley, G. S. 2010, The Astronomical Journal, 139, 329 → pages 8Tokovinin, A. A. 1997, Astronomy Letters, 23, 727 → pages 11van den Berg, M., Orosz, J., Verbunt, F., & Stassun, K. 2001, Astronomy & Astrophysics,375, 375 → pages 11Ventura, P., Caloi, V., D’Antona, F., Ferguson, J., Milone, A., & Piotto, G. P. 2009, MonthlyNotices of the Royal Astronomical Society, 399, 934 → pages 5Woodley, K. A., Goldsbury, R., Kalirai, J. S., Richer, H. B., Tremblay, P.-E., Anderson, J.,Bergeron, P., Dotter, A., Esteves, L., Fahlman, G. G., Hansen, B. M. S., Heyl, J., Hurley, J.,Rich, R. M., Shara, M. M., & Stetson, P. B. 2012, The Astronomical Journal, 143, 50 →pages 1, 2449


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items