A Dissertation Entitled Stellar Populations in Nearby Merging

A Dissertation entitled

Stellar Populations in Nearby Merging Galaxies

by Alexander J. Mulia

Submitted to the Graduate Faculty as partial fulﬁllment of the requirements for the Doctor of Philosophy Degree in Physics

Dr. Rupali Chandar, Committee Chair

Dr. Jon Bjorkman, Committee Member

Dr. Michael Cushing, Committee Member

Dr. Nikolas Podraza, Committee Member

Dr. Bradley Whitmore, Committee Member

Dr. Patricia R. Komuniecki, Dean College of Graduate Studies

This document is copyrighted material. Under copyright law, no parts of this document may be reproduced without the expressed permission of the author. An Abstract of Stellar Populations in Nearby Merging Galaxies by Alexander J. Mulia

Submitted to the Graduate Faculty as partial fulﬁllment of the requirements for the Doctor of Philosophy Degree in Physics The University of Toledo December 2015

Galaxy mergers were common in the early universe. To better understand this critical step in galaxy evolution, we perform detailed studies of three nearby merging systems. Using images from the Hubble Space Telescope, we identify hundreds of star clusters in these systems, most of which formed as a result of a merger. By studying these clusters, we are able to constrain the properties of their host galaxies.

These properties include: the timescale of the interaction, morphology of the merger’s progenitor galaxies, and the conditions in which stars and clusters formed. We find clusters in all tidal tails of our galaxy sample, even tails that were previously reported to be clusterless. Ages of clusters are similar to ages of their host tidal tails as predicted from simulations. We also find a color gradient across some tails, indicative of a gradient in ages that suggest star formation takes place primarily in the center of the tails, where gas is likely densest. In addition, cluster ages allow us to probe the star formation histories in these systems by predicting past SFRs in various regions of the galaxies using a new method involving the cluster mass function. The mergers also present an interesting environment to study star clusters themselves. We find that the formation and evolution of star clusters in mergers fits the “quasi-universal” picture of clusters seen in many other galaxies.

iii This work is dedicated to two women whose support has been the foundation of my success: my ﬁanc´ee, Ashley Nelson, and my mother, Janiece Mulia. Acknowledgments

There are several people that have helped make this work the best that it could be.

First, I would like to thank my advisor, Rupali Chandar, for not only passing on her

knowledge, but providing me conﬁdence and encouraging growth both inside and out

of academia.

To Brad Whitmore, I am grateful for his detailed and insightful comments on my papers, posters, and everything in between.

Finally, I would like to thank all of my fellow graduate students at the University of

Toledo. Whether talking through fundamental physics, helping with code, or editing my writing, their help was fundamental to both my work and my sanity.

v Contents

Abstract iii

Acknowledgments v

Contents vi

List of Tables x

List of Figures xi

List of Abbreviations xiii

1 Introduction 1

1.1 The Role of Mergers in Galaxy Evolution ...... 1

1.1.1 ModelingtheMergers ...... 4

1.1.2 Diﬃculties and Limitations of Simulations ...... 5 1.2 StarClusters ...... 6

1.2.1 StarClusters:TheBasics ...... 7

1.2.2 Star Cluster Evolution: A Theorist’s Perspective ...... 9

1.2.3 Star Cluster Evolution: An Observer’s Perspective ...... 11

2 Star Clusters in the Tidal Tails of Merging Galaxies 15

2.1 Introduction...... 15

2.2 Observations, Data Reduction, and Cluster Selection ...... 17

2.2.1 ObservationsandDataReduction ...... 17

vi 2.2.2 Cluster Selection ...... 21

2.3 Results...... 29

2.3.1 ClusterDensity ...... 29

2.3.2 Luminosity Functions (LFs) ...... 30

2.3.3 Colors and Ages of Clusters in Tidal Tails ...... 32 2.3.3.1 The Impact of Internal Reddening on Cluster Colors 35

2.3.3.2 Ages...... 36

2.3.4 Diﬀuse Light in Tidal Tails ...... 37

2.4 Discussion...... 42

2.4.1 Comparing Cluster Ages with Tail Ages ...... 42

2.4.1.1 NGC520 ...... 44 2.4.1.2 NGC2623...... 45

2.4.1.3 NGC3256...... 46

2.4.1.4 The Timescale of Cluster Formation ...... 46

2.4.2 Comparing Cluster Ages Across Galaxies ...... 47

2.4.3 How are the Tidal Tail Age and Luminosity Function Related? 48

2.5 Conclusions ...... 49

3 The Cluster Population of NGC 3256 51

3.1 Introduction...... 51

3.2 Observations, Data Reduction, and Cluster Selection ...... 52 3.2.1 ObservationsandDataReduction ...... 52

3.2.2 Cluster Selection ...... 54

3.3 ClusterAgeandMassDetermination ...... 56

3.3.1 SEDFitting...... 58

3.3.2 Calibrating the Hα Filter ...... 59

3.4 Results...... 62

vii 3.4.1 Luminosity Function ...... 62

3.4.2 Age Distribution ...... 63

3.4.3 MassFunction...... 65

3.4.3.1 Uncertainties on β ...... 67 3.4.3.2 IsThereanUpperMassCutoﬀ? ...... 68

3.5 How Does Distance Aﬀect the Observed Cluster Distributions? . . .. 69

3.6 HowEﬃcientlyDoesNGC3256FormClusters? ...... 72

3.6.1 CMF/SFRStatistic...... 74

3.6.2 TheΓStatistic ...... 76

3.7 SummaryandConclusions ...... 78

4 The Cluster Populations of NGC 520 and NGC 2623 81

4.1 Introduction...... 81

4.2 ObservationsandClusterSelection ...... 82 4.3 Results...... 85

4.3.1 Luminosity Functions ...... 85

4.3.2 Color Distributions ...... 87

4.3.2.1 Ages...... 89

4.4 Constraining Star Formation Rates Using the CMF/SFR Statistic . . 89

4.4.1 Approximating the Cluster Mass Function ...... 90 4.4.2 SFRs in the Main Bodies of Mergers ...... 91

4.4.3 SFRs in Tidal Tails of Mergers ...... 92

4.4.4 Uncertainties ...... 93

4.5 Discussion...... 95

4.5.1 LFsAcrosstheGalaxySample ...... 95

4.5.2 Do NGC 520 and NGC 2623 Lack Recent Star Formation? . . 96 4.5.2.1 GasDynamics ...... 96

viii 4.5.2.2 Other Considerations ...... 97

4.5.3 Timescales of Star Formation in NGC 520 ...... 99

4.5.4 Conditions of Star Formation in NGC 520 ...... 99

4.6 Conclusions ...... 100

5 Conclusions and Future Work 102

5.1 Cluster Evolution in Merging Galaxies ...... 102

5.2 UsingClustersasaTool ...... 103

5.3 FutureWork...... 104

References 106

ix List of Tables

2.1 Galaxyproperties...... 19

2.2 Selectioncuts ...... 23

2.3 Propertiesofstarclusters ...... 28

2.4 Diﬀuselightmeasurements...... 41

3.1 Estimates of cluster formation eﬃciency in NGC 3256 ...... 79

x List of Figures

1-1 HubbleExtremeDeepField ...... 2

1-2 ImageoftheAntennae ...... 3

1-3 SimulatedevolutionofNGC2623 ...... 5

1-4 HST imageofthemainbodyoftheAntennae ...... 7

1-5 Examplesofaglobularandopencluster ...... 8 1-6 Sample mass-age diagram for a synthetic population of clusters ..... 12

1-7 Examples of YMC age distributions and mass functions ...... 13

2-1 Fieldofviewforgalaxysample ...... 17

2-2 ACS/WFC F 814W imagesofthegalaxysample...... 18

2-3 Thumbnailsofstarclusters ...... 24

2-4 NGC520starclustercandidates ...... 25 2-5 NGC2623starclustercandidates ...... 26

2-6 Piewedgestarclustercandidates ...... 26

2-7 NGC3256starclustercandidates ...... 27

2-8 Luminosity functions of tidal tail star clusters ...... 31

2-9 Color-color diagrams of tidal tail star clusters ...... 33

2-10 Spatial distribution of young and old NGC 520 clusters ...... 34 2-11 Spatial distribution of young and old NGC 3256 clusters ...... 34

2-12 LocationofhaloclustersinNGC520andNGC3256 ...... 37

2-13 B − I colormapofgalaxysample...... 38

2-14 Diﬀuse light color-color diagram for galaxy sample ...... 39

xi 2-15 Color-color and density diagrams for tidal tail clusters ...... 43

3-1 BVIHα imageofNGC3256...... 53

3-2 Concentration index and mV distributionsforNGC3256 ...... 55 3-3 Color-colordiagramofNGC3256clusters ...... 57

3-4 Hα line emission map of NGC 3256 with cluster locations and ages . . . 61 3-5 LuminosityfunctionsofNGC3256clusters...... 62

3-6 Mass-agerelationforNGC3256clusters ...... 64

3-7 AgedistributionsforNGC3256clusters ...... 65

3-8 MassfunctionsforNGC3256clusters...... 66

3-9 Schechter functions ﬁt to NGC 3256 mass functions ...... 69

3-10 Source detections in the Antennae from original and degradedimages . . 70 3-11 Luminosity functions for Antennae cluster candidates ...... 71

3-12 Color-color diagrams for Antennae cluster candidates ...... 73

3-13CMF/SFRstatisticforNGC3256...... 76

4-1 HRCcoverageofNGC520andNGC2623 ...... 83

4-2 Concentration index and mV distributionforNGC520 ...... 84 4-3 Cluster candidate locations in NGC 520 ...... 85 4-4 Cluster candidate locations in NGC 2623 ...... 86

4-5 Luminosity function of NGC 520 and NGC 2623 cluster candidates ... 86

4-6 Color-color diagrams of NGC 520 and NGC 2623 cluster candidates... 88

4-7 Color-color and density diagram for clusters in the secondary nucleus of

NGC520 ...... 91

4-8 CMF/SFR statistic for clusters in the secondary nucleus of NGC 520 .. 92 4-9 CMF/SFRstatisticforpiewedgeclusters ...... 93

4-10 CMF/SFR statistic for clusters in the southern tidal tail of NGC520 .. 94

4-11 CMF/SFR statistic for clusters in the eastern tidal tail of NGC 3256 .. 94

xii List of Abbreviations

ACS ...... Advanced Camera for Surveys AGN ...... active galactic nucleus CFR ...... cluster formation rate CMF ...... cluster mass function FWHM ...... full width at half maximum HLA ...... Hubble Legacy Archive HRC ...... High Resolution Camera HST ...... Hubble Space Telescope IMF ...... initial mass function IR ...... infrared LF ...... luminosity function LIRG ...... luminous infrared galaxy NED ...... NASA/IPAC Extragalactic Database NIR ...... near-infrared PN ...... primary nucleus SED ...... spectral energy distribution SFH ...... star formation history SFR ...... star formation rate S/N ...... signal to noise ratio SN ...... secondary nucleus SPH ...... smoothed particle hydrodynamics SSP ...... simple stellar population WFC ...... Wide Field Camera WFC3 ...... Wide Field Camera 3 WFPC2 ...... Wide Field Planetary Camera 2 YMC ...... young massive cluster

xiii Chapter 1

Introduction

1.1 The Role of Mergers in Galaxy Evolution

In 1929, Edwin Hubble discovered that the universe is expanding, and conversely that galaxies in the early universe were much closer together. Less space between galaxies made mergers far more frequent in the early universe. Some of the best evidence for this was gathered in 1995, when the Hubble Space Telescope’s (HST )

Wide Field Planetary Camera 2 (WFPC2) gathered optical light from one of the darkest patches on the sky for ≈ 140 hours, providing the world with a glimpse of the universe approximately 12 billion years ago, now known as the Hubble Deep

Field. HST has since taken similar observations using newer instruments. The most recent and deepest image was taken with HST ’s Wide Field Camera 3 (WFC3) and is known as the Hubble Extreme Deep Field, which is shown in Figure 1-1. Close examination of this image reveals many objects that appear irregular and distorted, and this distortion is caused by the interaction of two or more galaxies, often resulting in stripping of part of the galaxy. Cameron et al. (2011) used one of the Deep Field images to show that the fraction of irregular, distorted galaxies decreased substantially starting at ≈ 3 billion years after the big bang, meaning mergers have become much less frequent over time. Because HST performed deep ﬁeld observations on three

1 Figure 1-1: Hubble Extreme Deep Field.

diﬀerent areas of the sky and found galaxy interactions to be common in all of them,

we conclude that galaxy mergers and interactions were frequent in the early universe.

Therefore, to understand how galaxies form and evolve, we need to learn more about

the important role of galaxy interactions and mergers. Because detailed studies of these distant systems are diﬃcult, we turn to nearby interacting systems. While more rare, there are a number of interacting galaxies in the local universe. Figure 1-2 shows the nearest major merging system, the Antennae.

In addition to the central heart shaped region, two bright, elongated “tails” can be seen streaming away from the main body of the galaxy. These tails are the result of tidal interactions between the galaxies, which consist of stripped stars, gas, and

2 Figure1-2: The Antennae (NGC 4038/39), the nearest pair of merging galaxies. Copyright Rolf Wahl Olsen, www.rolfolsenastrophotography.com. dust from the disks of the progenitor galaxies, and they are a common product of interacting galaxies. Focusing on the main body of Figure 1-2, the abundance of bright pink spots that indicate Hα emission, which is radiation resulting from Hydrogen electrons transitioning from the second to ﬁrst excited states, showcase the strong star formation occurring in the Antennae. We ﬁnish by concluding that if galaxy interactions were common in the early universe, and interactions appear to result in increased star formation, much of the star formation in the early universe might be attributed to merging galaxies.

3 1.1.1 Modeling the Mergers

We are unable to observe galaxies interacting in real time, because the timescale for two galaxies to merge is of the order of a few billion years. However, there are enough ongoing mergers in the local universe that we can study interacting systems at diﬀerent stages of evolution. This was ﬁrst attempted by Toomre (1977), who compiled 11 likely candidates for interacting galaxies into a sequence of evolution using simulations from Toomre & Toomre (1972); this pioneering work is now referred to as the Toomre Sequence. Simulations have since evolved using N-body codes, and they are the best way to constrain the properties of interacting galaxies such as the number and timing of previous close passages.

N-body simulations start with a number of gravitationally interacting, collisionless particles with some set of initial conditions, and allow gravity alone to drive the evolution of the system. This is a good approximation for stars and dark matter, because their kinematics are dominated by gravitational forces. The initial conditions that diﬀerentiate one simulation from another can be simpliﬁed into four independent parameters, such as disk orientation, pericenter separation, orbital eccentricity, and mass ratio (Privon et al. 2013).

Once the simulation is completed, it is compared to observations such as optical stellar light and HI maps. This comparison is typically done by examining the tidal tails, because they evolve relatively quickly over time, as well as retain information regarding the initial interaction. The kinematics of ejected stars behave ballistically, i.e. they are often not majorly eﬀected by self gravity (Barnes & Hibbard 2009). By comparing the orientation, length, and overall morphology of the simulated tidal tails to observations, one can estimate key properties of a merging system. Unfortunately, matching three-dimensional simulations to observations, which are projections on the sky, introduces additional “viewing” parameters: time since galactocentric pericenter,

4 Figure 1-3: Figure 8 in Privon et al. 2013. N-body simulation of the evolution of NGC 2623, a major merger. Times given are relative to the present day. viewing angles, length and velocity scaling, and system center of mass.

Once the best matching simulation has been found, it can be used to study the history and future evolution of that merging system. For example, Figure 1-3 shows the history of the merging system NGC 2623, according to the best matching simulation.

The times given are relative to the present day.

1.1.2 Diﬃculties and Limitations of Simulations

There are several challenges of the modeling approach. The large parameter space discussed in the previous section introduces degeneracies. We refer the reader to Privon et al. (2013) for a discussion on the eﬀect of degeneracies, and we note that usually a uniquely matching simulation can be found.

5 Another challenge is that while N-body particles are collisionless, the HI gas in observations is not collisionless. Because the real gas is collisional, it does not always follow the ballistic motion of the stellar tail (Barnes & Hibbard 2009). The gas can dissipate and fall back on to the main body, or encounter shocks that render the collisionless simpliﬁcation inaccurate. Even gas in the tidal tails does not completely act ballistically and can contract to form stars and star clusters within the tail.

More complicated simulations use smoothed particle hydrodynamics (SPH) to take gas into account and model star formation. SPH + N-body simulations have the beneﬁt of remedying the collisionless simpliﬁcation of N-body simulations, and provide an estimate of the star formation history of the interacting system. However, these simulations require more assumptions about the system, i.e. initial gas mass, gas temperature, prescription of star formation used, etc. (for further discussion, see

Mihos & Hernquist 1994). Simulations can be further complicated by the inclusion of feedback from stellar evolution, which will directly impact their gaseous compo- nents. In addition, many known interacting systems harbor an active galactic nucleus

(AGN), so feedback from these AGN must also be accounted for.

By studying star formation empirically, we can constrain these models to better represent physical systems. In this thesis, we study the stars and star clusters of three merging systems. We test that the star formation histories and spatial distributions of clusters obtained from observations are consistent with simulations.

1.2 Star Clusters

The majority of star formation occurs in stellar clusters, making clusters an eﬀec- tive way to study the recent star formation of external galaxies (Lada & Lada 2003).

For example, in the Antennae, Fall, Chandar, & Whitmore (2005) estimated that at least 20% and up to 100% of massive stars form in clusters. This system is famous

6 Figure 1-4: Optical and Hα HST image of the main body of the Antennae. Many thousands of newly formed clusters can be seen throughout the galaxy. for the thousands of young clusters found in the main body, which can be seen in

Figure 1-4 as red and bluish knots, many of which are also associated with Hα emission. However, in order to use clusters to understand the star formation history of a galaxy, it is critical to have an understanding of the complex evolution of clusters.

1.2.1 Star Clusters: The Basics

A star cluster is a gravitationally bound system of stars. Star clusters outside the Local Group are too distant to be resolved into individual stars, and the clusters presented in this work appear as point-like sources but ones that are broader than the point spread function. Because we cannot test whether or not these clusters are gravitationally bound, we use the term “star cluster” for any concentrated aggregate of stars with a density much higher than that of the surrounding stellar ﬁeld, whether or not it also contains gas and whether or not it is gravitationally bound. This is the standard deﬁnition used by the star formation community (Lada & Lada 2003;

7 Figure 1-5: Examples of a globular and open cluster. Left: The globular cluster M3. Image Credit: Thomas Allen, Michelle Deady, Kevin Hardegree-Ullman, and Alexander Mulia. Right: The open cluster the Pleiades.

McKee & Ostriker 2007).

In our own Milky Way, star clusters are typically classiﬁed into two diﬀerent categories. Globular clusters are spherically shaped, massive (most are a few ×104 −

6 10 M⊙), and generally old (∼ 12 Gyr), and they are typically found in the galactic halo and bulge. Open clusters, on the other hand, are irregularly shaped, much

1 3 younger (. a few 100 Myr), less massive (10 −10 M⊙), and they reside in the galactic disk (Portegies Zwart, McMillan, & Gieles 2010). Figure 1-5 shows an example of both types of clusters.

More recently, bright, massive blue clusters have been found in a number of star forming galaxies, including mergers (e.g., Whitmore & Schweizer 1995; Zepf et al. 1999), and they do not ﬁt into either category of globular or open cluster. Rather, these young massive clusters (YMCs) seem to bridge the gap in age and mass between the two traditional types of clusters. In fact, some YMCs in the Antennae and other

7 merging galaxies have masses upwards of 10 M⊙ (Fall, Chandar, & Whitmore 2009).

Given these high masses, it is natural to question if YMCs are the younger versions of globular clusters. One of the more widely accepted theories of galaxy formation is that many galaxies formed through multiple merging events (e.g., Dekel et al. 2009;

8 Naab et al. 2009; Zolotov et al. 2009; Oser et al. 2010; Font et al. 2011; Lackner et

al. 2012; Navarra-Gonz´alez et al. 2013), and it is possible that globular clusters were

born during these mergers, much as YMCs are formed in mergers today.

1.2.2 Star Cluster Evolution: A Theorist’s Perspective

Since a large fraction of stars form in YMCs, we can learn about the star formation history (SFH) of a galaxy by studying its YMC population. However, clusters are dynamic and evolving systems, and clusters as we observe them today do not tell us the full story of their formation. We therefore need an understanding of how YMCs evolve with time.

In general, clusters will become fainter over time, and this has two independent causes. One is that the brightest, most massive stars will die off quickly, a concept known as “fading.” The other is that clusters lose mass over time, and the physical dominant mechanism responsible for this mass loss changes over time. To understand the drivers of this mass loss, we simplify the dynamical evolution of a stellar system into three major stages. We caution that this is indeed a simplified description, and external factors such as the tidal field of the host galaxy and passing giant molecular clouds also play a role. The first stage occurs when the clusters are still embedded within their natal gas, which typically lasts for the first ∼ 10 Myr. This gas is expelled from the system by supernovae and stellar winds from the most massive stars of the system. This rapid gas removal can gravitationally unbind many clusters completely, while only partially destroying others, a process referred to as “infant mortality” (Fall, Chandar,

& Whitmore 2005). The remaining clusters will be either unbound or only weakly bound, making them vulnerable to dissolution from other processes.

Mass loss due to stellar evolution drives the cluster dynamics on ≈ 10 − 100 Myr

timescales. In this stage, supernovae and stellar winds will reduce the net cluster

9 mass by 10 − 50% (Fall & Chandar 2012). Further, more massive stars sink toward

the center of the cluster in a process known as mass segregation. Mass segregation is a result of dynamical friction, the concept that a massive object moving through

a distribution of less massive objects will exchange energy via gravity, causing the

massive object to slow down. As the massive stars slow down, they sink in toward

the cluster center. Once these massive stars undergo their ﬁnal stages of evolution

and shed their mass, they cause the cluster core to expand, further weakening the

gravitational binding. On longer timescales (& 100 Myr), two-body interactions drive cluster relaxation,

eventually leading to evaporation. Two-body interactions produce a Maxwellian ve-

locity distribution, leaving a fraction of stars with velocities greater than the escape

velocity (Portegies-Zwart, McMillian, & Gieles 2010). As these stars escape, the ve-

locity distribution is altered and a new escape velocity is established, with another

fraction of stars moving faster than escape velocity. This process will continue until the cluster is completely dissolved.

Two-body relaxation inevitably leads to core-collapse, which can be understood

by considering the energy ﬂow in the cluster. Because the system is approximately

virial, it has a negative heat capacity, and therefore any removal of potential energy

(i.e., evaporation) will dynamically heat the cluster. Because this process is slow (on

the order of 108 or 109 yr; Portegies-Zwart, McMillian, & Gieles 2010), the cluster attempts to maintain thermal equilibrium by transferring energy via two-body interactions from the inner region of the cluster toward the outer region. This extraction of energy from the core causes it to contract and eventually collapse, increasing the potential energy and cooling the cluster.

There are a number of mechanisms that heat the cluster, altering the timescale of core collapse. In particular, gravitational interactions between single stars and binary systems in the dense inner regions of the cluster can kick the single star further out in

10 orbit with high velocities, dynamically heating the cluster. Core collapse and binary

heating oppose one another to dictate the fate of the cluster.

1.2.3 Star Cluster Evolution: An Observer’s Perspective

We have discussed some of the physics of star cluster evolution, but how does this translate to what is directly observable? In this work, we compare photometric colors to those predicted by stellar population models to obtain ages and masses of star clusters. This means that age and mass estimates are subject to observational incompleteness, in addition to the eﬀects of fading and mass loss discussed above.

Here, we introduce plots that diagnose these observational eﬀects and are fundamental to understanding the evolution of star clusters.

The top panel of Figure 1-6, taken from Whitmore, Chandar, & Fall (2007), shows the age-mass diagram for a synthetic population of clusters. The middle panel

shows the same diagram after introducing a magnitude threshold to simulate an

observational limit. As Whitmore et al. point out, the absence of points in the lower

right corner of the middle panel could be misinterpreted as a lack of low mass old

clusters, when in reality these clusters likely exist but are simply fainter than the

detection threshold. The observed age (τ) distribution of a population of YMCs, expressed as dN/dτ,

has been found in many galaxies to take the form of τ γ , with γ ≈ −1, (Fall, Chandar,

& Whitmore 2005; Chandar, Fall, & Whitmore 2006; Fall & Chandar 2012). The left

panel of Figure 1-7 shows the age distribution of clusters in the Antennae, taken from

Whitmore, Chandar, & Fall (2005). The limits over which the age distribution is

complete depend on the range of cluster masses and luminosities that are considered. The interpretation of the age distribution is not trivial, as it is a result of clus-

ter formation as well as destruction. Therefore, an age distribution of τ −1 might be interpreted as either a recent burst of cluster formation, or as the combined eﬀects

11 Figure 1-6: Figure 2 of Whitmore, Chandar, & Fall (2007), showing the mass- age and luminosity-age plots for a synthetic population of clusters. of cluster formation plus destruction. For the Antennae, Whitmore, Chandar, & Fall

(2007) make the argument that if the τ −1 relation is due to a burst of cluster formation, this burst would have to be synchronized across the galaxy, because a similar age distribution is found for clusters in diﬀerent portions of the galaxy. Whitmore et al. then established that the time needed for a signal to travel across the galaxy to synchronize the burst would be ∼ 500 Myr, longer than the timescales involved in measuring the age distribution. Cluster destruction, on the other hand, would not rely on an event synchronized across the galaxy, but rather on mechanisms internal to each cluster (as described in Section 1.2.2). If a population of clusters follows a

τ −1 age distribution and we assume a constant rate of cluster formation, then every log unit of time results in 90% of clusters disrupting.

The mass function (dN/dM) for YMCs has been found to take the form of either a simple power law M β (Zhang & Fall 1999; Bik et al. 2003; Lada & Lada 2003;

McCrady, Gilbert, & Graham 2007; Fall, Chandar, & Whitmore 2009; Chandar, Fall,

& Whitmore 2010; Chandar et al. 2010; Goddard et al. 2010; Chandar et al. 2011), or a Schechter-like function (Gieles et al. 2006; Bastian 2008; Larsen 2009; Bastian et

12 Figure 1-7: Left: Figure 2 of Fall, Chandar, & Whitmore (2005), showing the age distribution of Antennae clusters for diﬀerent mass and luminosity cuts. Right: Figure 3 from Fall, Chandar, & Whit- more (2009), which shows the mass function of YMCs in the Antennae for two separate age bins.

β al. 2012) M exp(M/MC ), where MC is the turnover mass. The value of β is ≈ −2 in most galaxies, while the need for the exponential term is still debated (see Chapter

3 for further discussion). The right panel of Figure 1-7 shows the mass function of YMCs in the Antennae, as seen in Fall, Chandar, & Whitmore (2009). Plotted in two independent ranges of cluster age, the mass function takes on a pure power law with exponent β ≈ −2 for both cases. Like the age distribution, the mass function is complete to diﬀerent values for varying age bins, and therefore the older age range does not extend to cluster masses as low as for the younger age bins.

The power law mass function for YMCs is in stark contrast to the shape of the mass function of ancient Milky Way globular clusters, which has a lognormal distribution

5 that peaks at ≈ 2×10 M⊙ (Fall & Zhang 2001). It is possible, perhaps likely, that the mass function of globular clusters started out as a power law, and through the multi- body interactions discussed in Section 1.2.2, lower mass clusters became disrupted, shaping the mass function to its current lognormal form.

13 Both panels in Figure 1-7 feature multiple trend lines with a similar slope. For

the clusters in the Antennae, the age distribution for high and low mass clusters

exhibits nearly the same slope, while young and intermediate age clusters feature a

nearly identical mass function. This suggests that the mass and age distributions are

independent, and can be thought of as a single bivariate distribution g(M, τ) ∝ M βτ γ (Fall, Chandar, & Whitmore 2009). If the mass function and age distribution are truly

independent, it would imply that cluster disruption is largely a mass-independent

process, at least up to τ < 100 Myr. Fall, Chandar, & Whitmore (2009) established a

toy model to test for mass-dependence in disruption for Antennae clusters, and found

that the only model that ﬁts the data involves no mass dependence. Similarities in the

mass functions and age distributions have been found for clusters in the Magellanic Clouds, the spiral galaxies M51 and M83, and the Milky Way (Fall & Chandar 2012),

supporting a lack of mass dependence on cluster disruption.

While the recent studies mentioned in this section ﬁnd evidence of a near universal picture of cluster formation and destruction, a relatively small sample of galaxies has been examined. In this work, we expand the exploration to more extreme star forming environments, by analyzing star clusters in three merging galaxy systems. We start in Chapter 2 with the analysis of cluster populations in the tails of these three mergers. Chapter 3 describes our analysis of the cluster population of NGC

3256 and compares our results with the picture of cluster formation and destruction painted in this section. Chapter 4 details our similar study of two other merging systems, NGC 520 and NGC 2623.

14 Chapter 2

Star Clusters in the Tidal Tails of

Merging Galaxies

2.1 Introduction

Tidal tails form during the interaction between galaxies, and are composed of stars, gas, and dust ejected from the original galactic disks. These interactions are often accompanied by one or more bursts of star and cluster formation in the main bodies of the parent galaxies, according to simulations (e.g., Mihos, Bothun, & Richstone

1993) and observations of star clusters (e.g., Whitmore & Schweizer 1995).

The star cluster formation processes that occur within tidal tails, however, are not well understood. For example, building on pioneering work by Knierman et al. (2003),

Mullan et al. (2011; hereafter M11) used V and I band photometry from WFPC2 on HST to measure the brightness and colors of point-like sources both in and out of the tail region for 23 tidal tails. They found a statistically signiﬁcant number of clusters in 10 out of their 23 tails, and they concluded that the presence of tail clusters depends on several factors, including tail age, HI density, and surface brightness of the tail. In this work we revisit two of these tails using higher quality data taken

15 with the Advanced Camera for Surveys (ACS)1 on HST /Wide Field Camera (WFC)

and detect clusters in both of them for the ﬁrst time.

Even less is known about the ages of star clusters in tidal tails. de Grijs et al. (2003) found ∼ 40 clusters each in the Tadpole galaxy with ages of ≈ 175 ± 25 Myr and in the Mice galaxies (NGC 4676) with ages ≈ 100 ± 20 Myr. Most of these clusters appear to have formed in the tails since their estimated ages are the same age or younger than those of the tails; de Grijs et al. estimated a dynamical age of

400−800 Myr for the Tadpole, and Privon et al. (2013) suggested a formation time of

∼ 175 Myr ago for the Mice. Tran et al. also observed the Tadpole galaxy, and found 42 extremely young tail clusters (∼ 4 − 5 Myr; 2003). Bastian et al. (2005) found young massive star clusters in the tails of NGC 6872 and was able to estimate ages and masses for individual clusters using UBVI and Hα photometry. No star clusters older than the dynamical age of the tidal tail were found (at least down to their completeness limit of MV = −10.4), once again suggesting that the clusters formed from the tidal debris. They also found a large population of clusters that are very young (< 10 Myr). While looking for spatial trends, Bastian et al. found that older tail clusters tend to be located closer to the main bodies, while younger clusters are spread throughout the tail. They argue that these ﬁndings suggest an initial burst of cluster formation as the tails form. After this burst, as the tails expand, the gas clouds cool and condense, eventually becoming sites of individual cluster formation.

While tidal tails provide a unique and interesting environment to study star clusters, they also provide key information for improving simulations of interacting/merging galaxies. Eﬀorts to reproduce mergers using simulations focus on tidal tail morphology to constrain details of the galaxy interaction. Ages of star clusters in tails provide constraints on the merger timescale. The spatial distributions of clusters in the main bodies and tails helps to discriminate between diﬀerent prescriptions

1http://www.stsci.edu/hst/acs

16 Figure 2-1: 10′ x 10′ images of our galaxy merger sample from the Digitized Sky Survey. Overlaid in each image is the WFC ﬁeld of view. for star formation within the simulations. For example, dynamical models of the in-

teracting system NGC 7252 show that density-dependent star formation results in a

more centrally concentrated distribution of main-body clusters, while shock-induced

star formation tends to spread cluster formation throughout the galaxy, including in

the tails (Chien & Barnes 2010).

The goal of this work is to better understand how clusters form and are distributed during the merging of two galaxies. We focus our study on three merging systems,

NGC 520, NGC 2623, and NGC 3256. This chapter is set up in the following manner.

In Section 2.2 we discuss the observations and photometry, as well as cluster selection.

Section 2.3 presents the cluster densities and cluster luminosity functions, as well as colors and ages of clusters and of the diﬀuse light. In Section 2.4 we discuss the galaxy merger history and interpret our results. Section 2.5 summarizes the work and presents conclusions.

2.2 Observations, Data Reduction, and Cluster Se-

lection

2.2.1 Observations and Data Reduction

17 Figure 2-2: ACS/WFC F 814W images showing the sample of all three galaxies and their tails.

18 Table 2.1: Galaxy Properties

19 Galaxy R.A. Dec. Distance Modulus AV E(B − V ) E(V − I) NGC 520 01 24 35.1 +03 47 33 32.46 0.077 0.025 0.035 NGC 2623 08 38 24.1 +25 45 17 34.50 0.113 0.036 0.051 NGC 3256 10 27 51.3 -43 54 13 32.79 0.334 0.107 0.151 Observations of NGC 520 and NGC 2623 were taken with the WFC on the ACS on HST as part of program GO-9735 (PI: Whitmore). NGC 520 was observed in

2004 October using the ﬁlters F 435W (≈ B in the Johnson-Cousins system; exposed for 2400 s), F 555W (≈ V ; 1000 s), and F 814W (≈ I; 1200 s). The entire galaxy, including the majority of its tidal tails, were covered in a single WFC frame. Observa- tions of NGC 2623 were taken in 2004 February using the same ﬁlters with exposure times of 3730, 1206 and 2460 s, respectively. It was also covered in a single WFC frame. Images of NGC 3256 come from two separate observing proposals. F 555W

filter observations of NGC 3256 were obtained in 2003 November as part of the program GO-9735 (2552 s). WFC observations using F 435W and F 814W filters were taken in 2005 November as part of program GO-10592 (PI: Evans) for 1320 and 760 s, respectively. Both fields cover most of NGC 3256’s eastern tail, as well as its main body. Figure 2-1 shows the WFC fields of view used in this study. We note that for these and all of the following images in this thesis, north is up and east is to the left.

The raw data were processed through the standard ACS pipeline. The reduced, multidrizzled images were taken from the Hubble Legacy Archive (HLA) and have a scale of 0.05′′ per pixel. Figure 2-2 shows F 814W images of our targets, with the tidal tail portions studied in this work labeled. We will refer to speciﬁc tails within a galaxy by abbreviating the location of each tail (i.e. the northern tail of NGC 520 will be called NGC 520N). We deﬁne the tail edges as the locations where the counts drop to 3σ above the general background level.

Figure 2-2 shows that NGC 520 has two tidal tails, commonly referred to as the north and south tails. Both of these tails extend beyond the ﬁeld of view in Figure 2-2. Not seen in Figure 2-2 is the dwarf galaxy UGC 957, at about 6′ north-west of

NGC 520. NGC 2623 is in the middle panel of Figure 2-2. It exhibits two nearly symmetric tidal tails, although the northern tail is wider than the southern tail. An extremely bright, triangle-shaped region south of the nucleus of the galaxy, which we

20 call the pie wedge, is also highlighted. NGC 3256 is on the right in Figure 2-2. While

it has two tails (east and west), only part of the eastern tail was observed. The three

mergers are next to each other on the Toomre Sequence, from youngest to oldest, in

the order NGC 520, NGC 2623, and NGC 3256 (Toomre 1977). General properties

of these systems are listed in Table 2.1.

2.2.2 Cluster Selection

Star cluster candidates were found by running the IRAF2 task DAOFIND on the

I band images. A few thousand objects were detected in each galaxy. We perform

BVI aperture photometry on all sources in NGC 520 and NGC 3256 using an aperture size of 3 pixels in radius and a background area of radii 8 to 11 pixels using the

IRAF task PHOT. Because of the greater distance to NGC 2623, we use a smaller aperture of 2 pixels (with the same background area) in order to minimize contamination from nearby sources. In the pie wedge region, we estimate the background level using annuli between 5–8 pixels, in order to minimize the impact of crowding.

Magnitudes were converted to the VEGAMAG system using zeropoints calculated from the ACS zeropoint calculator3. Aperture corrections were performed in NGC

3256 by selecting the brightest ∼ 20 isolated clusters in each filter and finding their mean 3–10 pixel magnitude difference. We then applied the 10 pixel to infinity aperture corrections taken from the encircled energy catalog for ACS/WFC in Table 3 of Sirianni et al. (2005). The final corrections were 0.45, 0.42, and 0.54 magnitudes for B, V, and I, respectively, and they were applied to each cluster. Unfortunately,

the background level was too high in NGC 520 and NGC 2623 for reliable aperture

correction determinations. We ﬁnd that clusters in NGC 520 have a similar range

2IRAF is distributed by the National Optical Astronomy Observatory, which is operated by the Association of Universities for Research in Astronomy (AURA) under a cooperative agreement with the National Science Foundation. 3http://www.stsci.edu/hst/acs/analysis/zeropoints

21 of sizes as those in NGC 3256 and therefore applied the aperture corrections derived

from NGC 3256 to the NGC 520 clusters. Clusters in the more distant NGC 2623 are

nearly indistinguishable from point sources; we assume the 2 pixel to inﬁnity aperture

corrections from Table 3 of Sirianni et al. (2005).

We measure full width at half maximum (FWHM) values for all detected sources using the ISHAPE software (Larsen 1999). ISHAPE measures FWHMs by convolving

the point spread function with a King proﬁle, then performing a χ2 calculation to test the goodness of ﬁt to each individual cluster (King 1966). ISHAPE iterates through

diﬀerent values for the eﬀective radius until a minimum χ2 is found.

We select cluster candidates using a combination of automated selection criteria

(listed in Table 2.2) and visual inspection of each candidate (see Bastian et al. 2012 for further discussion on the “hybrid” method). The selection criteria in mV , FWHM, and magnitude error (σV ) remove most foreground stars and background galaxies. We made a cut at the bright end of mV = 20, in order to eliminate saturated foreground stars (brighter than any cluster expected in these mergers). The faint end of our cluster catalog likely suﬀers from incompleteness. We restrict our sample to luminosities brighter than the point where the luminosity function (discussed in Section 2.3.2) turns down rather than continuing to increase in a power law fashion. We use cuts in

FWHM (size) to separate stars and background galaxies from clusters, to the highest degree possible. In NGC 3256, we exclude sources with FWHM values less than 0.2 pixels, because the FWHM of the vast majority of likely ﬁeld stars was measured to be smaller than that value. We also place an upper limit of 2.0 pixels in FWHM to exclude very extended background galaxies. The lower limit on FWHM for clusters in NGC 520 was chosen to be 0.1 pixels because there is a group of young, compact clusters in the northern portion of NGC 520S. At the distance of NGC 2623, clusters are essentially point sources, so FWHM measurements cannot distinguish clusters from bright, individual stars. Therefore, in addition to an upper limit on FWHM of

22 Table 2.2: Selection Cuts

Galaxy mV FWHM σV NGC 520 25.0 0.1 - 2.0 NGC 2623 26.5 ≤ 2.0 0.1 pie wedge 26.5 ≤ 2.0 0.2 NGC 3256 26.0 0.2 - 2.0

2.0, we apply a cutoﬀ in magnitude uncertainty, σV . In the tails, σV < 0.1 did a good job removing obvious non-clusters, but in the pie wedge, we relax our σV cut to 0.2 because the area is more crowded and has a very bright background. Finally, we visually inspect each cluster candidate. For each galaxy we identify

high signal-to-noise ratio (S/N) clusters, as well as obvious foreground stars, as bench-

marks for fainter objects. We choose clusters based on their fuzzy appearance, as well

as wider radial proﬁles than those of stars. Figure 2-3 shows three of the brightest

clusters from each galaxy’s tidal tails, and Figures 2-4 – 2-7 show locations of clusters

in each tail. All possible clusters were identiﬁed independently of the selection cuts

listed above, then later compared to the sample obtained from the selection cuts. We ﬁnd excellent agreement between the two methods in NGC 520 and NGC 3256, and

we only include sources that both ﬁt our selection cuts and visually appear as clusters.

Due to the poor S/N in NGC 2623, several sources appeared visually as likely clusters

but had FWHM and/or σV values outside of the selection criteria. Because of this low S/N, we rely more heavily on visual inspection than selection cuts and included some

sources with FWHM or σV values outside of the cuts. The total number of clusters

found in each tidal tail (Nclus) is compiled in Table 2.3, and the cluster catalog is in the online version of Mulia, Chandar, and Whitmore (2015).

23 Figure 2-3: Some of the brightest star clusters in the tidal tails of each galaxy. Each image is 25 × 25 pixels. Galaxies are separated by column from left to right, in the order NGC 520, NGC 2623 (tail region), NGC 2623 (pie wedge region), and NGC 3256.

24 Figure 2-4: NGC 520 cluster candidates are marked with green circles. On the top is the northern tail. The lower-left panel shows the northern portion of the southern tail. On the lower-right is the southern half of the southern tail.

25 Figure 2-5: NGC 2623 cluster candidates. On top is the northern tail. The southern tail is shown on bottom.

Figure 2-6: NGC 2623 pie wedge cluster candidates.

26 Figure 2-7: NGC 3256E cluster candidates.

27 Table 2.3: Properties of star clusters. τtail [Myr] is the predicted age of the tidal tail as estimated from simulations. Nclus is the total number of clusters found in the 2 region. Σclus is the star cluster density per kpc where MV < −8.5. τclus [Myr] is the approximate age of the bulk cluster formation. Median colors (hB - Vi, hV-Ii) are for clusters younger than the age of the tidal tail. α values (from α the relation dN/dL ∝ L ) listed come from the αconst−num method. 28

Galaxy Tail τtail Nclus Σclus τclus hB - Vi hV-Ii α NGC520 North 300 5 0.00 ∼ 100 0.08 ± 0.05 0.38 ± 0.05 NGC 520 South 300 93 0.49 260 0.19 ± 0.09 0.60 ± 0.17 -2.32 ± 0.36 NGC2623 North 220 75 0.54 230 0.10 ± 0.15 0.48 ± 0.28 -2.39 ± 0.34 NGC 2623 South 220 18 0.32 0.12 ± 0.13 0.43 ± 0.17 NGC 2623 pie wedge ≤ 220 76 8.5 ∼ 100 0.05 ± 0.15 0.43 ± 0.19 -2.02 ± 0.21 NGC3256 East 450 141 0.10 ≥ 260 0.21 ± 0.12 0.58 ± 0.21 -2.61 ± 0.27 Distant, red spheroidal galaxies can have a similar color and appearance to com-

pact star clusters. To assess the possible contamination from background galaxies,

we examine the colors and locations of sources that made it through our automated

cuts, but are located outside of the tidal tail regions, with the reasonable expectation

that most of these are likely background galaxies. However, we ﬁnd no such objects in any of our ﬁelds, only extended, obvious background galaxies. In addition, many

background galaxies should appear elliptical rather than circular. Using ISHAPE, we

measure the ellipticity of each cluster candidate and ﬁnd few elongated red sources.

We conclude that the fraction of background galaxies that have made it through our

cluster selection criteria is negligible.

2.3 Results

We ﬁnd a population of star clusters in each tidal tail in our sample of galaxy mergers. In this section we present the spatial densities, luminosities, and colors of

clusters in each tidal tail and estimate their ages.

2.3.1 Cluster Density

One quantitative measure of a cluster population is its surface number density, which is often used when cluster selection must be done statistically. We measure

−2 the number density of clusters, Σclus (clusters kpc ), for each tidal tail based on

the catalogs in this work. We choose a magnitude cutoff of MV < −8.5 (see Section 2.3.2 for details on absolute magnitude calculation) to directly compare our results to those of M11 (although altering the cutoff to MV = −8.0 and −9.0 yields qualitatively similar results). NGC 520S yielded Σclus =0.49, while NGC 520N yielded Σclus =0.0 because its five clusters were all fainter than MV = −8.5. The pie wedge in NGC 2623 possesses the highest cluster density in our sample by over an order of magnitude at

29 Σclus =8.5, while the north and south tails yielded Σclus =0.54 and 0.32, respectively.

Lastly, NGC 3256E possessed a relatively low density of Σclus = 0.10. We ﬁnd that

while all tidal tails in our sample contain clusters, there is a large variation in Σclus for each tail.

For both NGC 520 and NGC 3256, we are able to directly compare our values of

Σclus to those values found in M11, although we note that we cover diﬀerent areas of the tails for both systems. We also note that we adopt slightly diﬀerent distance

moduli than M11, resulting in small diﬀerences in Σclus. M11 found Σclus = −0.011 ±

0.013 in NGC 520N and Σclus = 0.004 ± 0.082 for NGC 3256E, both of which are consistent with zero. While our results for NGC 520N agree with M11, we find a clear population of clusters in NGC 3256E brighter than MV < −8.5. M11 also measured Σclus for MV < −7.5 and MV < −6.5, however we do not compare Σclus values at these cutoffs due to incompleteness. We find that, in general, younger tails tend to exhibit higher Σclus, in agreement with the results of M11.

2.3.2 Luminosity Functions (LFs)

The LFs of the tail clusters are shown in Figure 2-8. These can be described by a

α power law, dN/dL ∝ L , where α is determined from a linear ﬁt to log (dN/d(MV )). We ﬁt our cluster LFs over the V band magnitude range which we believe to be reasonably complete. Magnitudes include distance moduli from Table 2.1 as well as aperture corrections, and they have been dereddened because of foreground extinction

(see Section 2.3.3.1). We are unable to measure the LF in NGC 520N and NGC 2623S because they contained too few clusters.

We determine the LFs in two different ways, using both a fixed magnitude bin width (denoted αconst−mag) and a fixed number of clusters per bin (αconst−num). The results from both methods are consistent, and we report α values for the latter because it is better at handling low number statistics. In all cases, however, we find that the

30 NGC 520S NGC 2623N 1000 1000

100 100 dN/dL dN/dL 10 10

1 1 -8 -9 -10 -11 -8 -9 -10 -11 -12

MV MV NGC 3256E Pie Wedge 1000 1000

100 100 dN/dL dN/dL 10 10

1 1 -7 -8 -9 -10 -9 -10 -11 -12 -13 -14

MV MV

Figure 2-8: Luminosity functions of clusters in tails where measurements were possible. Filled circles are derived from the constant- number method and are best ﬁt by the dashed line (slope given as αconst−num; see Section 2.3.2 for details). Open circles ﬁt with a dash-dotted line (slope given as αconst−mag) are derived from the constant magnitude bin method.

31 two values for α are the same, within the uncertainties. The typical standard deviation

among α calculated using B, I, and V band magnitudes is ≈ 0.13, with no systematic

trend. The LFs in the B and I bands have similar slopes to those in the V band.

In NGC 520S, we find a best fit of α = −2.32 ± 0.36 down to MV < −8.0. We restrict our fit to this magnitude limit because completeness becomes an issue fainter than this. For NGC 2623, we focus on the pie wedge and northern tail. We fit the luminosity function in both regions down to MV ≈ −9.0, and find α = −2.39 ± 0.34 for the northern tail and α = −2.02±0.21 for the pie wedge. The luminosity function

in NGC 3256E is best ﬁt with α = −2.61 ± 0.27 down to MV = −8.0. We ﬁnd no evidence for a deviation from a simple power law at the bright end of the LF in any of our tail regions. Several studies have shown that the luminosity function can range from −2.8 <

α < −1.9 for clusters in a variety of environments (e.g., Whitmore et al. 2014 and

references therein). We ﬁnd that our results for α are within this range.

2.3.3 Colors and Ages of Clusters in Tidal Tails

We present the color-color diagrams for clusters in all tails in Figure 2-9. The

color coding corresponds to diﬀerent tails within each merging system. We also show with blue symbols the median V − I and B − V colors of clusters that apparently

formed after the tidal tails (we assume tail ages from simulations that are discussed in

Section 2.4.1 and compiled in Table 2.3). The solid line in each panel in Figure 2-9 is a

stellar population model from G. Bruzual & S. Charlot (2006, private communication,

hereafter BC06; and see also Bruzual & Charlot 2003) which predicts the evolution of

star clusters from about 106 – 1010 yr. All models assume a metallicity of Z = 0.02, and a Salpeter (1955) initial mass function (IMF). Numbers mark the logarithmic

age corresponding to the population.

32 6 6 7 7 8 8

9 9

North North tail age tail age South South

6 6 7 8 7 8

9 9 East Pie Wedge tail age tail age

Figure 2-9: Color-color diagrams. Clusters have been corrected for foreground reddening of the Milky Way by using the Aλ values compiled in Table 2.1. Solid black lines are BC06 cluster evolution tracks, and numbers represent log ages of clusters. The reddening vectors are shown by the red arrows for AV = 1. For comparison, the red diamond marks where the tail age (τtail) would fall on this diagram based on the ages shown. Clusters that formed in the tail are younger than the tail and presumably fall above the red diamond on the track. The median colors of these young clusters are marked in blue, represented by either a square (northern/eastern tail), triangle (southern tail), or a cross (pie wedge). The points in the upper right-hand corners indicate typical errors from photometric uncertainties.

33 Figure 2-10: Southern tail of NGC 520. Red and blue circles indicate clusters that are redder and bluer than our color cut (B − V = 0.3), respectively.

Figure 2-11: Eastern tail of NGC 3256. Red and blue circles indicate clusters that are redder and bluer than our color cut (B − V = 0.3), respectively.

34 2.3.3.1 The Impact of Internal Reddening on Cluster Colors

In order to accurately age-date star clusters, we must understand how reddening aﬀects their measured colors. We correct each cluster for foreground reddening of the Milky Way by using Aλ values given in the NASA/IPAC Extragalactic Database (NED), which are compiled in Table 2.1. The reddenings were estimated with the

standard relations between them and the extinction. A reddening vector for AV =1 is shown in Figure 2-9 by the red arrow. The reddening vector is nearly parallel to the predicted evolution of cluster colors, making it challenging to disentangle the eﬀects

of age and reddening in the measured colors. For example, a young cluster with some

extinction will have the same colors as an older cluster with no reddening.

Figure 2-9 shows that the clusters in each tail have a measured color spread.

In principle, there are a few reasons this spread could exist other than photometric

uncertainties, such as internal (and therefore diﬀerential) extinction or a range in cluster ages. Below, we examine each galaxy individually and determine that internal

extinction does not greatly aﬀect the majority of our cluster colors.

We ﬁrst focus on NGC 2623, whose color distributions are shown in the upper-

and lower-right panels of Figure 2-9. The spread in measured colors is the smallest

for any tail studied here, and instead of spreading along the reddening vector, the

cluster colors are fairly isotropically distributed. This pattern suggests that photometric uncertainties are primarily responsible for the color spread. We conﬁrm this

by creating color-color diagrams of NGC 2623 clusters in diﬀerent MV bins, and while the cluster color spread increases with fainter magnitudes, the color distribution is

centered near the median colors quoted in Table 2.3, regardless of brightness. If all

of the clusters in the tails of NGC 2623 formed at the same time, clusters obscured

by dust would exhibit redder colors than unobscured clusters, causing a color spread along the reddening vector. The clusters in both tidal tails of NGC 2623 appear to

have formed over a relatively short period of time. This is also true in the pie wedge,

35 where the narrow color ranges are also indicative of coeval cluster formation.

NGC 520 and NGC 3256, on the other hand, both exhibit a spread in color along the reddening vector. Here, we use the relative locations of blue (B − V < 0.3) versus red (B − V ≥ 0.3) clusters, shown in Figures 2-10 and 2-11, to assess whether age or reddening are more likely to be responsible for the observed spread in color. In the case of NGC 520 (Figure 2-10), the bluer clusters tend to trace out the central, densest portion of the tail, while the redder clusters are more evenly spread throughout the tail. Reddening due to dust would be expected to give the opposite trend, with reddened clusters mostly near the center of the tail. Figure 2-11 shows the locations of blue versus red clusters for NGC 3256. No obvious pattern is observed, but there is no evidence that red sources are primarily found in the brightest, densest portion of the tail. These results support the interpretation that the larger color spread of the clusters in the tails of NGC 520 and NGC 3256 is primarily due to a spread of cluster ages rather than to varying amounts of cluster reddening.

We now compare the colors of red tail clusters to clusters in the halos of NGC

520 and NGC 3256, which are shown in Figure 2-12. Selection and photometry of the halo clusters are performed using the same methods as in Section 2.2. We conclude that they are fairly similar in color, and therefore the red tail clusters are likely to be clusters that formed before the tails but were swept out during the tail formation. All of this evidence supports the interpretation that the red sources are older, unobscured star clusters.

2.3.3.2 Ages

We obtain ages for each cluster candidate by mapping their B − V and V − I colors to the BC06 track and ﬁnding the closest match (Figure 2-9). Ages are fairly accurate in the range 106 <τ < 109 yr. Based on the discussion in Section 2.3.3.1, we assume no internal reddening.

36 Figure 2-12: Locations of star clusters in the halos of NGC 520 (left) and NGC 3256 (right).

Figure 2-9 implies that star clusters in the tails of NGC 520 and NGC 3256 have a broad range of ages up to ∼ 109.5 yr. The tail ages are marked for each system with a red diamond, and clusters with colors below the red diamond are older than the tail age and might be old disk clusters that predate the merger; we will hereafter refer to these as “old” clusters, and we will refer to clusters younger than the tail age as “young” clusters. The tail clusters in NGC 2623 contain ages up to a few ×108 yr. The lack of observed clusters with ages older than a few ×108 yr may simply be because they are too faint to be observed in our data at the distance of NGC 2623.

2.3.4 Diﬀuse Light in Tidal Tails

In addition to distinct stellar clusters, we also study the diffuse stellar light in the tails. Because the tidal tails form from material stripped from the native disks, this diffuse light should consist of a mixture of old disk stars, faint newly formed stars, and dissolved clusters. While individual stars are unresolved, the integrated colors of the diffuse light can help to constrain the age of the stellar populations (e.g., Gallagher et al. 2010). The light will be dominated by the brightest and youngest unresolved objects, so we expect that our age estimates will be lower limits.

37 Figure 2-13: B − I images of NGC 520 (top left), NGC 2623 (top right), and NGC 3256 (bottom left). The scale on bottom gives the B − I color values.

We create color images: B − V , V − I, and B − I for all three galaxies, and we show the B − I images in Figure 2-13. The images were boxcar-smoothed with a four pixel kernel width. The main bodies of the mergers are very red because they contain large amounts of dust. The tidal tails are easily seen in Figure 2-13 as bluer regions where internal extinction is low (except, perhaps, in NGC 520N, as also suggested by our analysis of clusters in Section 2.3.3.1). We therefore assume that colors in the tail regions are not greatly aﬀected by internal extinction, and we believe diﬀuse light colors trace the ages of stellar populations relatively well.

We measure the V − I and B − V colors in different tail regions (white boxes in Figure 2-13 labeled from “A” up to “E” for each tail) after subtracting the point sources. Colors are calculated from the ratio of fluxes on a pixel by pixel basis and corrected for differences in instrumental zeropoints and foreground extinction. The

38 6

7 8 8.5 Galaxy: Region: 8.8 N520 N A 9 N520 S B N2623 N C N2623 S D N2623 P E N3256 E

Figure 2-14: Diﬀuse light colors for all three galaxies. Regions indicated in the legend correspond to Figure 2-13. Asterisks mark the median color of clusters found in that tidal tail that presumably predate the merger. Asterisks inside squares indicate that there were very few clusters that predate the merger (≤ 3). The solid black line is a BC06 cluster evolution track.

39 diﬀuse light regions were chosen to cover most of the optically bright portions of

the tails while avoiding breaks in ﬁeld coverage and saturated foreground stars. The

values in Table 2.4 are the mean color for each region. Figure 2-14 plots these values

on a color-color diagram overlaid with the BC06 track. Figure 2-13 reveals that the

diﬀuse stellar light in all tidal tails in our sample exhibits a common pattern (with the exception of NGC 3256E). While the tidal tails are fairly blue (B − I ≤ 1.4), we

find that there is a gradient with B −I color; it is redder near the edges of the tail and bluer near its axis. As previously discussed, this is not likely a reddening effect, so we believe that we are observing a gradient in the mean ages of the stars across these tails. In Section 2.3.3.1 we discussed the tendency for younger clusters to be near the center of the tail axis, leaving mostly older clusters along the tail edges, especially in NGC 520S. The diffuse light images further support a lack of young stars near the edges of the tails. We believe this to be the first time such a gradient across the tails

has been observed in these galaxies.

Figure 2-14 reveals that the northern tail of NGC 520 has the reddest diﬀuse light

of any tail in our sample. This is likely due to dust, as the red B − I color in Figure

2-13 extends from the main body straight through the tail. Region A in the southern tail is also almost certainly aﬀected by dust. We believe that the colors of the diﬀuse

light in the other regions of the southern tail are dominated by ages of the underlying

populations and not extinction because of the absence of dust lanes in this tail. These

regions have an age τ of around 800 Myr.

In NGC 2623, most of the regions in the north tail have colors consistent with

stellar populations of ∼ 700 Myr old. Region NC, however, is bluer and yields an age of ∼ 500 Myr. We note that the diﬀuse light in the pie wedge (region P) has B − V

and V − I colors close to region NC, as seen in Table 2.4, which might provide clues

to its formation, as we discuss further in Section 2.4.1.2.

Figure 2-14 shows that the diﬀuse light in NGC 3256 is perhaps the most peculiar

40 Table 2.4: Diﬀuse Light Measurements

Galaxy Region hV - Iidiﬀuse hB - Vidiﬀuse NGC 520 NA 1.13 0.70 NB 1.05 0.60 SA 1.14 0.71 SB 0.94 0.49 SC 0.90 0.47 SD 0.88 0.46 SE 0.97 0.49 NGC 2623 NA 0.96 0.48 NB 1.07 0.42 NC 0.76 0.33 ND 0.90 0.47 SA 0.88 0.45 SB 0.87 0.40 SC 0.92 0.38 SD 0.91 0.41 SE 0.96 0.42 P 0.68 0.31 NGC 3256 EA 0.77 0.82 EB 0.71 0.79 EC 0.78 0.39 ED 0.83 0.76 EE 0.66 0.63

41 of our sample. Only region EC falls near the BC06 track, with an associated age

τ ≈ 650 Myr. This would indicate that the diffuse light population in region EC is similar to the majority of the diffuse populations in the tails of NGC 2623. It would be plausible that region EC falls on a local region of enhanced star formation, causing its diffuse light colors to be bluer than other regions of NGC 3256E. However, region EC does not include any young star clusters, instead containing only one cluster whose colors suggest that it predates the merger. The rest of the regions we measured are gathered around V − I ≈ 0.7 and B − V ≈ 0.8 and are off the BC06 track.

Regardless, these oﬀset colors suggest faint population ages of about 800 – 1000 Myr.

It is likely that most of these diﬀuse light regions do not have colors consistent with the BC06 track because of the low surface brightness of NGC 3256E, resulting in large photometric errors. It is unlikely that Hα line emission could be responsible for the oﬀset from the model, although this emission would move measurements in the observed direction relative to the models, because such recent star formation is usually quite bright and lumpy.

2.4 Discussion

2.4.1 Comparing Cluster Ages with Tail Ages

The age of the tidal tails, which we deﬁne as τtail, refers to the time since perigalac- ticon of the two progenitor galaxies, because they form from material ejected after the galaxies undergo their closest approach. The ages of tails can be estimated from simulations and are probably accurate to within ∼ 20% (∼ 30 − 100 Myr; Barnes & Hibbard 2009). In this section, we brieﬂy summarize previous simulations of our sample of mergers, and we compare tail ages from those simulations with our measured cluster ages.

We ﬁnd that, for most systems, there is a period of increased cluster formation near

42 6 6

7 7 8 8

9 9

6 6

7 7 8 8

9 9

Figure 2-15: Color-color diagrams used to quantify when the peak cluster formation occurred in each system where possible. Right panels show densities of clusters in color-color space, with contours overplotted. The peak is marked with a blue circle. Left panels are similar to Figure 2-9, with density peaks from the right panels included.

43 the time that the tidal tails formed. We quantify the time of this cluster formation

(hereafter denoted τclus) by comparing the density peak of the color distribution with cluster evolution model predictions. These density plots are shown in the right panels

of Figure 2-15. Contours are plotted, and the density peak is marked with a circle.

The colors of the peak are then compared with a BC06 track to obtain an age (shown in left panels of Figure 2-15).

2.4.1.1 NGC 520

The interaction history of NGC 520 is the least well understood of our sample.

Stanford & Balcells (1991) ran the most recent simulation of the system, comparing to infrared (IR) and optical images, as well as the HI velocity ﬁeld. Their best match to

these observations are at a simulated age of ∼ 300 Myr ago. The system is particularly complicated, however, because Stanford & Balcells find that their simulations were not fully consistent with two colliding disk galaxies, due to the presence of a third dwarf galaxy, UGC 957. They conclude that the southern tail formed from the interaction of two massive disk galaxies and that the northern tail is the result of a passage by UGC 957. If the two tails formed via different interactions, we might expect differences in their cluster ages. Despite these complexities, we adopt a best guess of τtail ∼ 300 Myr. The northern tail of NGC 520 contains only three young clusters, making a color density plot impractical here. However, the three clusters are tightly grouped around

100 Myr on the BC06 track; we therefore approximate their ages to be τclus ∼ 100 Myr. The majority of the south tail clusters appear to be older, and, based on their

color distribution in Figure 2-15, they are consistent with a period of cluster formation

that began around τclus ∼ 260 Myr ago. The diﬀerences in the numbers and ages of the clusters detected in the north

and south tails appear to be consistent with the scenario put forth by Stanford &

44 Balcells. The passage of a dwarf galaxy would not necessarily result in a starburst

as intense as one resulting from two equal-mass galaxies, because the dwarf galaxy

would strip less gas to form clusters. We detect many more clusters in the southern

tail, which Stanford & Balcells predict is the result of a disk-disk interaction, than

the northern tail, which likely formed from the passage of the dwarf galaxy UGC 957. Stanford & Balcells speculate that the dwarf galaxy might have formed the northern

tail as the disks were interacting, around the same time the southern tail was forming.

However, the tight age grouping of the three clusters in NGC 520N seems to indicate

that cluster formation took place earlier than in NGC 520S.

2.4.1.2 NGC 2623

Privon et al. (2013) simulated NGC 2623 and estimated that an age τtail = 220±30 Myr best reproduces spatial and kinematic distributions of HI. Privon et al. also successfully recreated the pie wedge region, showing that it is probably material from the northern tail that has fallen back through the main body and is currently on a southward trajectory. We measure clusters in NGC 2623N to have a peak density in color corresponding to an age of τclus ∼ 230 Myr; this supports the scenario that cluster formation began at the same time that the tail formed. There are some clusters that appear to be older than the tail, however, as we discussed in Section 2.3.3.1, we believe this is due to photometric scatter resulting from low S/N. NGC 2623S possesses too few clusters to reliably measure τclus, although it is clear from Figure 2-9 that most of the clusters are slightly bluer than the color associated with the tail age. It is possible that formation of clusters in the south tail was somewhat delayed relative to the tail itself, but uncertainties in tail and cluster ages are too large to deﬁnitively establish such a delay.

Inspection of Figures 2-9 and 2-15 shows that clusters in the pie wedge have a

45 very narrow range of colors, suggesting that they all formed at approximately the same time; we measure τclus to be ∼ 100 Myr. These clusters are clearly bluer and hence younger than clusters in the tails of NGC 2623. This result is consistent with the scenario proposed by Evans et al. (2008), where the pie wedge formed from debris from the inner region of the northern tail. They suggest that the debris fell back through the main body, inducing a burst of star formation (see also Privon et al.

2013). While Privon et al. do not provide an age for the pie wedge, their Figure 8 shows snapshots of the merger at diﬀerent times (Figure 1-3 of this work). At 85 Myr ago, the debris can be seen falling southward, past the main body, roughly consistent with our cluster formation timescale.

2.4.1.3 NGC 3256

Recent unpublished results from simulations place NGC 3256 at τtail = 450 ± 50 Myr (G. Privon, private communication). The simulation is based on spatial and kinematic HI images of the system and was run using the same methodology as for

NGC 2623. Clusters in the eastern tail have the largest spread in color of any system in our sample, with a weak density peak, τclus ∼ 260 Myr. This suggests a longer duration of cluster formation in this tail. These ages are consistent with previous results.

Trancho et al. (2007) obtained spectroscopic ages for three clusters in the western tail, and found two clusters with ages of ∼ 80 Myr, and a third with an estimated age of ∼ 230 Myr. The latter ﬁts well within the range that we ﬁnd for clusters in the eastern tail.

2.4.1.4 The Timescale of Cluster Formation

In the three systems studied here, we have found that cluster formation typically begins with the tidal tail formation, although in a few cases a delay cannot be ruled

46 out. This type of delay was also found by Bastian et al. (2005) in NGC 6872. We

therefore conclude that, in general, τclus . τtail. Very young (τ < 10 Myr) clusters have not formed in any of the tails studied here.

We conﬁrm this by examining HST Hα images of the two most recent mergers, NGC

2623 and NGC 520, and ﬁnd no Hα emission in any of their tails. We speculate that either the tails have used up their gas, or, because the gas is not collisionless, most of the remaining gas was largely dispersed during the interaction. It is worth noting, however, that clusters with age τ < 10 Myr have been reported in the tidal tails of the Tadpole galaxy and NGC 6872 (Tran et al. 2003; Bastian et al. 2005). In the case of NGC 6872, the time since tail formation is ∼ 145 Myr (Horellou & Koribalski

2003), making NGC 6872 signiﬁcantly younger than any interacting system in our sample (except for the pie wedge). It is possible that NGC 6872 has simply not yet dispersed its tidal tail gas, resulting in newly formed clusters.

Based on our cluster age distributions, we suggest the following scenario: initially, the progenitor galaxies form clusters at a lower, approximately constant rate. Once the two galaxies begin their interaction, many of these disk clusters, along with other disk material, are stripped and form the tidal tails. The stripped gas typically begins forming clusters immediately, resulting in an increased rate of cluster formation. The process continues in the tails at a lower rate until the gas reservoir is exhausted or the gas is dispersed.

2.4.2 Comparing Cluster Ages Across Galaxies

We suggest the following sequence of merger ages: NGC 2623 is the most recent of the three mergers, followed by NGC 520, with NGC 3256 being the oldest. This sequence is supported by both our estimated cluster ages and simulated ages of tidal tails (and thus ages of the mergers themselves). We note that the age sequence suggested by these simulations is diﬀerent than what the Toomre Sequence suggests,

47 as NGC 520 and NGC 2623 are ﬂipped.

We find that clusters in the pie wedge have ages around τclus ∼ 100 Myr. The north tail of NGC 2623 contains clusters with ages at τclus ∼ 230 Myr, while the few south tail clusters are slightly younger, although we do not attempt to quantify their ages. NGC 520, the next youngest merger, exhibits a slight age difference among clusters in different tails. NGC 520N has clusters around τclus ∼ 100 Myr, while NGC

520S clusters has τclus ∼ 260 Myr. The oldest merger, NGC 3256, contains clusters in the eastern tail with ages τclus ∼ 260 Myr, although we caution that the broad age distribution in NGC 3256E suggests that cluster formation likely began earlier than

260 Myr ago. For a direct comparison, all the values of τtail and τclus are listed in Table 2.3.

2.4.3 How are the Tidal Tail Age and Luminosity Function

Related?

We have measured the LFs for all tails where possible, and find tentative evidence that younger mergers exhibit shallower LFs, although our sample size is small. M11 found that a LF with −2.5 <α< −2 provided qualitatively acceptable fits to their statistically determined cluster sample in tails where the measurement was possible, although they were not able to directly fit the LF with a power law due to the poor quality of the data. Whitmore et al. (2014) measured the luminosity function in a variety of spiral galaxies in search of various correlations between α and galactic environment. A weak negative correlation was found between star formation rate

(SFR) and α (i.e. galaxies with high SFR tend to have ﬂatter LFs). For instance, the Antennae, a major merger with a relatively high SFR, had the second lowest α value in their sample, at α = −2.07 ± 0.03. No measurements of the cluster LF were made in tidal tails, however. While α

48 tends to decrease with the tail age, all α values in our sample are within the range

observed in a wide variety of other environments (Whitmore et al. 2014 and references

therein).

2.5 Conclusions

We have used ACS/WFC observations from HST to directly observe star clusters in the tidal tails of three nearby mergers. We draw the following conclusions.

1. Every tidal tail in our sample, as well as the pie wedge, contains a population

of star clusters. We note that NGC 520N and NGC 3256E were reported to

have no statistically signiﬁcant cluster presence in M11.

2. We estimated the ages of clusters in each tidal tail and compared with estimated ages for the tails themselves from simulations. The tidal tails in NGC

2623 contain a population of clusters which appear to have formed during the

formation of the tails. Tails of NGC 520 and NGC 3256 possess several clusters

that appear to predate the merger, in addition to a population of clusters that

formed at, or soon after the formation of the tails.

3. In every tidal tail in our sample, cluster formation lasted for several tens of

millions of years, but no very recent formation (in the last few million years)

has occurred in any of the tails.

4. Simulations place the formation of tails in the following sequence, from youngest to oldest: the pie wedge, NGC 2623, NGC 520, NGC 3256. Ages obtained from

clusters generally agree with this sequence.

5. The diﬀuse light in the tails of NGC 520 and NGC 2623 exhibits a gradient in

color across the tail (as opposed to along it), which is likely indicative of system-

atic patterns in the ages of the diﬀuse stellar population. To our knowledge, 49 such color gradients in the diﬀuse stellar light in tidal tails of these galaxies

has not been previously reported. The gradient is loosely traced by the spatial

distribution of young and old star clusters within the tail. We interpret this

gradient as a superposition of broadly distributed older stellar material and

younger stars and clusters that formed along the center of the tail, where gas is densest.

6. The LFs of clusters in our tidal tail sample are similar to those found in a variety

of galactic environments. We tentatively ﬁnd that as the merger age increases,

the luminosity function tends to become steeper.

50 Chapter 3

The Cluster Population of NGC

3256

3.1 Introduction

NGC 3256 is a merging system ≈ 36 Mpc away. Dynamical simulations suggest that the system began interacting ≈ 450 Myr ago, and it has since undergone a period of major star and cluster formation. It exhibits two tidal tails that are rich with young massive stellar clusters (Mulia, Chandar, & Whitmore 2015). The main body of NGC 3256 contains a dense population of clusters, many of which are embedded in the galaxy’s dusty interstellar medium. The galaxy’s intense star formation make it an extreme environment to study cluster formation and destruction.

The cluster population of NGC 3256 has been studied in a number of previous works. Zepf et al. (1999) used B and I band images taken with the Wide Field Planetary Camera 2 on HST to examine the colors and luminosities of main body

clusters. Using the fraction of blue light that they found in clusters, they estimated

that the eﬃciency of cluster formation in NGC 3256 is ∼ 20%. Goddard et al.

(2010) estimated ages and masses of NGC 3256 clusters from UBVI photometry using

HST ’s Advanced Camera for Surveys and measured the cluster formation eﬃciency

51 (hereafter referred to as Γ; Bastian 2008). Their method involved estimating the

cluster formation rate, taken from the total mass of clusters younger than 10 Myr,

+7.3 and dividing by the galaxy’s SFR, ﬁnding Γ = 22.9%−9.8. Some works, including Goddard et al., have searched for a relation between Γ and star formation rate density. Such a relation would imply that the conditions for cluster formation are not universal, but dependent on the host galaxy. Chandar, Fall,

& Whitmore (2015), on the other hand, measured the cluster mass function (CMF) in multiple galaxies and found that it can nearly be normalized by the star formation rate of the host galaxy. The CMF/SFR statistic implies that cluster formation is similar in many galaxies, regardless of SFR. We apply both Γ and, for the ﬁrst time, the CMF/SFR statistic, to NGC 3256.

This chapter is set up as follows. Section 3.2 describes the observations and cluster selection. Section 3.3 gives the method for obtaining ages and masses of clusters. Section 3.4 presents the luminosity functions, age distributions, and mass functions for NGC 3256. In Section 3.5, we quantify the eﬀects that distance has on the measured luminosity function and age distribution of clusters. We apply the

CMF/SFR statistic and Γ to NGC 3256 in Section 3.6. We summarize our results and state conclusions in Section 3.7.

3.2 Observations, Data Reduction, and Cluster Se-

lection

3.2.1 Observations and Data Reduction

Our observations come from the ACS on HST. NGC 3256 was observed using the

ﬁlters F 555W (≈ V in the Johnson-Cousins system; exposed for 2552 s), F R656N

(Hα; 2552 s), and F 330W (≈ U; 11358 s) as part of the program GO-9735 (PI:

52 Whitmore). The V and U band images were taken in 2003 November using the Wide

Field Camera and High Resolution Camera (HRC), respectively. The Hα observations were taken in 2004 March. WFC observations using F 435W (≈ B) and F 814W (≈ I)

ﬁlters were taken in 2005 November as part of program GO-10592 (PI: Evans) for

1320 and 760 s, respectively.

The raw data were processed through the standard ACS pipeline. The reduced, multidrizzled WFC images were taken from the HLA and have a scale of 0.05′′ per pixel, while U-band images taken with HRC have 0.025′′ per pixel. A BVIHα color image of NGC 3256 is shown in Figure 3-1.

Figure 3-1: BVI color image of NGC 3256, with Hα overlaid. The circle indicates where we divide our sample into inner and outer regions and is ≈ 9.5′′ (2 kpc) in radius.

53 3.2.2 Cluster Selection

We use the IRAF task DAOFIND on the I band image to make an initial catalog of point sources in NGC 3256. Using the PHOT task in IRAF, we run photometry on the B, V , and I band images using a 3 pixel aperture in radius and a background area between 5 and 8 pixels in radius. For the U band image, we use an aperture radius of

6 pixels and a background area of radii 10 to 16 pixels, since the HRC has twice the resolution of the WFC. We use the ACS zeropoint calculator to convert instrumental magnitudes to the VEGAMAG system. We use 3 pixel to inﬁnity aperture corrections in B, V , and I bands taken from isolated star clusters found in the tidal tails of NGC

3256 reported in Mulia, Chandar, and Whitmore (2015). For the U band aperture correction, we find ∼ 15 isolated bright clusters and measure their mean 6 – 20 pixel magnitude. We use the encircled energy catalog found in Table 4 of Sirianni et al. (2005) for the 20 pixel to infinity aperture correction. The final corrections were 0.63,

0.45, 0.42, and 0.54 for U, B, V , and I band ﬁlters respectively. We also correct the photometry in each ﬁlter for foreground extinction, 0.528, 0.441, 0.334, 0.264, and

0.183 for U, B, V , Hα, and I, respectively, taken from NED.

We require UBVIHα photometry to accurately age-date star clusters and therefore throw out sources not covered in all ﬁve bands. This has a minimal impact on our catalog and removes only a small fraction (< 10%) of our sources. We visually inspect all sources to remove cosmic rays and hot pixels from the catalog; using the

IRAF task IMEXAMINE, we manually separate obvious foreground stars from cluster candidates by their sharper radial proﬁles.

The low galactic latitude of NGC 3256 means that there will be a signiﬁcant number of foreground stars from the Milky Way in our images. Luckily, at the distance of NGC 3256, we expect most clusters to be at least partially resolved in the

HST /WFC images. We measure the concentration index, C, to aid our separation

54 Figure 3-2: Concentration index-brightness plot for all sources found in NGC 3256. The blue vertical line at C =2.2 illustrates that all likely field stars are more centrally concentrated than C = 2.2. We note that the diagonal incompleteness to right of the dashed line is below our cutoff of mV = 23.0. of foreground stars and cluster candidates. For a single source, C is measured as the difference in magnitude taken at two different apertures (typically done in V band at 0.5 and 3.0 pixels). C has proven to be a robust method in determining the compactness of an object and has therefore often been used to remove stellar contamination from cluster catalogs in many previous works (see Chandar et al. 2010 for a more in-depth discussion on C). Figure 3-2 shows mV versus C for all cluster candidates and likely foreground stars, shown as circles and asterisks, respectively.

With the exception of the very brightest star shown here (which we later conﬁrmed to be saturated), all stars fall to the left of the vertical dashed line at C = 2.2. We therefore only include cluster candidates with C > 2.2.

We address incompleteness of our catalog by making a cut where the mV distri-

55 bution begins to ﬂatten at the faint end, rather than continue in a power law fashion.

Because the central portion of NGC 3256 is dustier and has a higher background level than the outer regions, we use separate completeness limits for the inner and outer regions. Clusters in the inner region are complete down to mV = 21.5, while the outer region is complete down to mV = 23.0. The separation of these two regions is shown in Figure 3-1. We ﬁnd that contamination from background galaxies is negligible in the main body of NGC 3256. In all, we ﬁnd 505 cluster candidates that meet our selection criteria.

We brieﬂy compare our catalog to that of Goddard et al. (2010; hereafter, G10), who studied clusters in the very center of NGC 3256 using UVI ﬁlters taken with the HRC and B band taken with the WFC from HST /ACS. They covered the very central ∼ 0.5′ × 0.5′ and found 904 cluster candidates. Their sample was reduced to

276 clusters after including detection and magnitude uncertainty criteria, as well as criteria regarding goodness of ﬁt to simple stellar populations (SSPs). The portion of

NGC 3256 studied here extends the G10 coverage to include the northern spiral arm, as well as many clusters just outside the central, brightest portion of the system. In addition, we include Hα observations, which can be critical for breaking the degeneracy between age and extinction, a degeneracy that aﬀects a number of clusters in this merging system (see Section 3.3.2 for more details).

3.3 Cluster Age and Mass Determination

We present the color-color diagram for star clusters in NGC 3256 in Figure 3-3.

The solid line is a stellar population model from BC06 (described in Section 2.3.3) which predicts the evolution of star clusters from about 106 – 1010 yr. All models assume a metallicity of Z = 0.02 (i.e., solar metallicity), and a Salpeter (1955) IMF. Numbers mark the logarithmic age corresponding to the population.

56 Figure 3-3: Color-color diagram of main body clusters in NGC 3256. For clarity, we only show the brightest clusters (mV < 21.5). The red arrow shows the reddening vector for AV = 1. Various ages in log(τ/yr) are shown by numbered red squares.

57 3.3.1 SED Fitting

Using UBVIHα photometry, we ﬁt photometric spectral energy distributions

(SEDs) of clusters to SEDs from BC06 cluster evolution tracks in order to determine the age of each cluster. Reddening due to dust can yield inaccurate ages if not accounted for, as redder colors imply older ages. We therefore adopt the common technique of varying age and E(B − V ) in order to minimize the function

2 obs mod 2 χ (τ, E(B − V )) = X Wλ(mλ − mλ ) . (3.1) λ

obs Here, mλ is the observed magnitude and has been corrected for both aperture and

mod for foreground extinction (see Section 3.2.2 for details), and the BC06 models (mλ )

are normalized to each cluster’s apparent V band magnitude. The function Wλ =

2 2 −1 (σλ +0.05 ) weights each photometric measurement in the ﬁt by accounting for the

photometric uncertainty (σλ) output by PHOT. We add 0.05 in quadrature to σλ to ensure that no single measurement dominates the SED ﬁt.

The BC06 model predicts a mass-to-light ratio (for the V band luminosity) for each age, which we use to estimate a mass for each cluster. LV is calculated from each cluster’s MV magnitude (corrected for internal extinction) and the distance modulus of NGC 3256. Masses computed using a Salpeter vs. Chabrier (2003) IMF are known to differ by ≈ 40%, independent of age. In order to be consistent with other works that use a Chabrier IMF, we correct for this effect by dividing our masses by 1.38. In addition, the age range and mass-to-light ratio will differ among various SSP models.

For instance, the GALEV models used by G10 do not extend to ages younger than log(τ/yr) = 6.5. Clusters that are in reality younger than log(τ/yr) = 6.5 have masses that are underpredicted by the GALEV models. We will revisit this issue in Section

3.6.1, where we discuss the eﬀect this has on the cluster mass function.

58 3.3.2 Calibrating the Hα Filter

Broad band colors alone make it diﬃcult to diﬀerentiate clusters that are red

because they are older from those which are young but appear redder due to the

presence of dust. This is known as the age-extinction degeneracy. Including Hα

photometry as a ﬁfth point in the SED can help break this degeneracy, and has been

used in a number of studies to age-date clusters (i.e. Fall, Chandar, & Whitmore 2005; Chandar et al. 2010). Unfortunately, there is no F 658N image to serve as the continuum to be subtracted from the F R656N image. In this section we describe our method for calibrating the Hα ﬁlter in order to use it in the SED ﬁtting method.

We first determine the zero point of the F R656N filter described above, since it is not publicly available. We identify ≈ 20 somewhat older star clusters, those that lack Hα emission, by comparing objects in the F 555W and F R656N filters. These clusters should only have continuum emission from stars (i.e., no line emission from ionized gas). We estimate the age and extinction of each of these line free clusters by performing the χ2 fit described in 3.3.1 using the UVBI filters and use these to predict the clusters’ F R656N magnitude. Comparison between the predicted and observed magnitudes gives a zeropoint of 22.4 ± 0.1. We find that shifting the zero point by 0.1 alters the ages of only ∼ 5% of the clusters by ≥ 0.3 in log(τ/yr).

The brightness due to Hα line emission (LHα) can be predicted from the ﬂux of Hydrogen-ionizing photons (NLy; given in BC06 as a function of age) from the equation

−12 LHα = NLy × EHα × jHα =1.36 × 10 NLy, (3.2)

where EHα is the energy of the Hα emission line (taken from Leitherer & Heckman

1995) and jHα is the fraction of Hydrogen-ionizing photons that go in to Hα, which is

24%. By adding the ﬂux taken from LHα to the ﬂux from F 658N predicted by BC06,

59 we obtain the total Hα brightness that each cluster should emit.

Typically, applying the calibration above should overestimate the observed Hα

ﬂux, because some fraction of the photons will escape the area covered by our aperture before encountering a Hydrogen atom to ionize. This is corrected for by introducing

−12 an escape fraction fe, to Equation 3.2 (i.e., LHα = 1.36 × 10 NLy × [1 − fe]). To measure the escape fraction, we ﬁnd a new sample of clusters that clearly exhibit

Hα line emission (i.e., clusters that are much brighter in F R656N than F 555W ; we will temporarily refer to these clusters as “young.”). As with the “old” clusters, we obtain an initial estimate of their age from the UBVI SED fitting. By comparing the F R656N magnitudes for the young clusters to the F 658N magnitudes predicted from their age, we can adjust fe to obtain a best fit. However, for NGC 3256, plotting the calibrated F R656N magnitudes against the predicted F 658N magnitudes reveals reasonable scatter, rendering the exact value of fe somewhat uncertain. In order to understand the impact that the escape fraction has on the cluster age estimation, we run the UBVIHα SED fitting with two values of fe; we use the best

fit fe of 0.25, as well as fe=0.0. We find that < 5% of clusters have a difference in age of log(τ/yr) > 0.1 between age estimations from the two escape fractions, and adopt fe=0.0 for simplicity due to its negligible impact on age estimates. We test that cluster ages are reasonable by performing some simple checks. In most galaxies, it has been found that the brightest clusters tend to be young (i.e.,

Whitmore et al. 1999). We test our sample by picking the brightest 15 clusters after correction for internal extinction, and ﬁnd that they are all younger than 10 Myr.

In addition, Figure 3-4 shows a continuum-subtracted Hα map of NGC 3256, with clusters overlaid that are color-coded by age. Clusters younger than 10 Myr (blue circles) tend to clump together and follow the brightest Hα regions, while 10-100 Myr

(green) and 100-400 Myr (red) year old clusters are more sparse and fall on areas with no Hα emission.

60 Figure 3-4: A map of Hα line emission, with continuum emission subtracted out. The image has been inverted so that black regions show Hα line emission. Blue, green, and red circles indicate locations of clusters with ages log(τ/yr) < 7, 7 < log(τ/yr) < 8, and 8 < log(τ/yr) < 8.6, respectively.

61 Figure 3-5: Luminosity functions of NGC 3256, uncorrected for internal extinction within NGC 3256. On the left is the LF for all coverage of the main body. The middle panel shows the LF for the inner clusters, while the right panel shows the LF for the outer clusters.

3.4 Results

3.4.1 Luminosity Function

We present the luminosity functions for clusters in NGC 3256 in Figure 3-5. The

LF is best described by a power law, dN/dL ∝ Lα, where α is determined from a linear

ﬁt to log (dN/d(mV )). The mV values have been aperture corrected and corrected for foreground extinction in the Galaxy. The three panels of Figure 3-5 show the LF for the entire sample, as well as for clusters in the inner and outer regions of NGC

3256, separately. We find α = −2.23 ± 0.07 for the LF for the combined sample for mV ≤ 21.5. The LF for the inner and outer regions individually yielded α ≈ −2.1 in both cases. We also compare the results for the B and I filters and find that the power law index α remains within our uncertainty calculated for the V band. Zepf et al. (1999) previously presented a catalog of clusters in NGC 3256 using poorer quality (in both depth and resolution) B and I band WFPC2 images from

62 HST. Using their cluster catalog to measure the B band LF, we ﬁnd α =2.15 ± 0.07,

in agreement with the LF presented in this work. In addition, the LF is similar to α

found in many other galaxies (Whitmore et al. 2014).

We also compare the LF found in this work to that of clusters in the eastern tidal

tail of NGC 3256, as measured by Mulia, Chandar, & Whitmore (2015). They obtain α = −2.61 ± 0.27, notably steeper than the LF in the main body, for clusters brighter

than MV ≈ −8.0.

3.4.2 Age Distribution

Figure 3-6 shows the mass-age relation, separated for clusters in the inner and

outer regions. The solid lines show our brightness limits of mV = 23.0 (MV = −9.79

with a distance modulus of 32.79) for the outer region and mV = 21.5(MV = −11.29) for the inner region. The diﬀerent completeness limits of the inner and outer regions

result in diﬀerent mass regimes in which the age distribution can be determined. The

two dashed lines indicate the masses at which both samples are complete over the

age ranges used in the age distribution (Figure 3-7). The inner cluster sample is

complete to log(τ/yr) ≈ 8 for log(M/M⊙) > 5.6, while the outer sample is complete

to log(τ/yr) ≈ 8.4 for log(M/M⊙) > 5.2. The apparent lack of clusters in the range 7 < log(τ/yr) < 7.5 is an artifact of the SED ﬁtting, partially due to the BC06 track forming a loop in color space.

We ﬁnd that the age distribution follows a power law with dN/dτ ∝ τ γ , with

γ ≈ −0.67 ± 0.08 for two independent mass ranges in both the inner and outer regions. The combined catalog from both regions yields γ ≈ −0.5 for both mass ranges. Fall & Chandar (2012) ﬁnd the cluster age distribution for multiple galaxies to be a power law with γ between ≈ −0.6 to −1.0.

The slope of the age distribution fluctuates somewhat depending on how the fit is performed. However, this fluctuation agrees well with the fitting errors shown in

63 8

7 )] . O

5 Mass [log(M/M

6.0 6.5 7.0 7.5 8.0 8.5 9.0 Age [log(τ/yr)]

Figure 3-6: Mass-age relation for clusters in the inner and outer regions of NGC 3256. The solid lines illustrate our completeness limits of mV = 21.5 (blue) and mV = 23.0 (red) for the inner and outer catalogs, respectively.

64 Combined Inner Outer -4 6.00 < Log(M) < 6.40 6.00 < Log(M) < 6.40 5.50 < Log(M) < 5.80 5.60 < Log(M) < 6.00 5.60 < Log(M) < 6.00 5.20 < Log(M) < 5.50 -5 ) τ

γ = -0.45 ± 0.07 -6 γ = -0.71 ± 0.13 γ = -0.66 ± 0.06

log(dN/d γ ± -7 = -0.52 0.12 γ = -0.64 ± 0.06 γ = -0.66 ± 0.06

-8 6 7 8 9 7 8 9 7 8 9 log(τ/yr) log(τ/yr) log(τ/yr)

Figure 3-7: Age distributions of NGC 3256 for two mass regimes, with best fits shown as dashed lines. Different panels are for different catalogs.

Figure 3-7, as well as the uncertainties listed in Fall & Chandar (2012). We conclude that γ for NGC 3256 is similar to the values obtained for the cluster populations in

other galaxies.

3.4.3 Mass Function

We show the mass functions in Figure 3-8, which can be described by a power law in the relation dN/dM ∝ M β. We measure β for three age ranges for clusters in the inner and outer areas of the galaxy, as well as for the combined sample. The given uncertainties only reﬂect the formal uncertainty in the ﬁt; we discuss other sources of error in Section 3.4.3.1.

In the age range 6 < log(τ/yr) < 7, we find that the combined sample of inner and outer clusters yields β = −1.86±0.14. Separating the sample yields β = −1.84±0.14 for outer clusters, but a shallower β = −1.61 ± 0.13 for inner clusters. The somewhat shallower slope is likely a result of crowding, which can artificially flatten the mass function. Studies of other galaxies have also found β ≈ −2 (e.g., Zhang & Fall 1999;

65 -1 10 6 < log(τ/yr) < 7 7 < log(τ/yr) < 8 8 < log(τ/yr) < 8.6 β = -1.61 ± 0.13 β = -1.31 ± 0.15 β = -2.08 ± 0.45 -2 β = -1.86 ± 0.14 10 β = -1.84 ± 0.14 10-3

10-4

-5 dN/dM + const. 10 10-6 4 5 6 7 5 6 7 5 6 7

Log(M/MO .) Log(M/MO .) Log(M/MO .)

Figure 3-8: Cluster mass functions for NGC 3256 for three age ranges. Dif- ferent symbols indicate catalogs from the inner area of the main body (squares), the outer area (circles), and both catalogs combined (triangles). The inner and combined catalogs contained too few clusters for reasonable ﬁts to mass functions for ages log(τ/yr) > 7 and are excluded from the middle and right panels. Fits to the mass functions are done for a constant number of clusters per bin. See Section 3.4.3 for details.

66 Chandar, Fall, & Whitmore 2010; Chandar et al. 2010; Chandar, Fall, & Whitmore

2015). We note that G10 ﬁnd β = −1.85 ± 0.12 for the same age range in NGC 3256,

quite similar to our result.

We determine the mass functions for older clusters in the outer region, since there are very few such clusters in the inner region. We find β = −1.31 ± 0.36 in the age range 7 < log(τ/yr) < 8. This age range is also the most difficult to fit, partially due to the loop in the BC06 track discussed in Section 3.4.2. Clusters with ages 8 < log(τ/yr) < 8.6 produce a mass function with β = −2.08 ± 0.45, where the large

uncertainty in ﬁtting is due to small number statistics.

3.4.3.1 Uncertainties on β

Several diﬀerent factors can contribute to the uncertainty in the shape of the

cluster mass function. We mentioned some of these factors already in Section 3.3.2,

such as the impact that small changes in the zero point and escape fraction for

the Hα ﬁlter have on β. In this section, we consider additional contributors to the

uncertainties on β. Incompleteness can impact β, causing it to become artiﬁcially shallow. This is es-

pecially true when considering β for a very small age range, where brightness and mass

are correlated, and a loss of faint clusters is equivalent to a loss of low mass clusters.

We discussed completeness of our catalog in Section 3.2.2, and we choose selection

limits that are suﬃciently bright that the impact of incompleteness is minimal.

Details of the fitting procedure can also affect β. We find that the specific mass range used in the fits affects β by ±0.05, except in the cases where the number of

clusters is small (. 40). We also measure the impact of bin size on β and again ﬁnd

β to fluctuate by ±0.05. Overall we find that all of these sources of uncertainty affect

β within the formal uncertainty given in Figure 3-8 for each age range.

67 3.4.3.2 Is There an Upper Mass Cutoﬀ?

Some works have found that mass functions for clusters in spiral galaxies deviate from a single power law and have a downturn at the high mass end (e.g. Bastian et al.

2012). These works suggest that the mass function is better described by a Schechter

β function, dN/dM ∝ M exp(−M/MC ), where MC is the upper mass cutoff. While the lower-mass end of the Schechter function converges to a power law with index β, the high-mass end drops off from a power law quickly above MC . A single power law provides a good fit to the observed cluster mass function, since no obvious downturn is observed at the high mass end. Therefore, our distribution is only consistent with a Schechter function with a high MC . Below, we determine a lower limit to MC . For simplicity, we compare our 6 < log(τ/yr) < 7 and 8 < log(τ/yr) < 8.6 cluster

mass functions to Schechter functions with a ﬁxed β = −2.0 and varying MC values, although there is no reason β has to be −2.0. For the latter age range, we use only the sample taken from the outer region of the central galaxy.

The mass function of clusters with ages 6 < log(τ/yr) < 7 can only be reasonably

ﬁt for a Schechter function with β = −2 and log(MC /M⊙) values above 7.0. Because

the mass function is ﬁt out to log(M/M⊙) ∼ 6.5, it is not surprising that Schechter functions with log(MC /M⊙) < 7.0 begin to fall below the mass function at the high mass end. The left panel of Figure 3-9 shows the mass function with multiple MC values.

Fitting the 8 < log(τ/yr) < 8.6 mass function with a Schechter function is indistinguishable from fitting a pure power law for log(MC ) ≥ 6.5. This is shown in the right panel of Figure 3-9, where MC = 6.5 and MC = 7.0 provide good fits to the mass function. The mass function is fit out to log(M/M⊙) ∼ 6.25, and again we find that the Schechter functions with log(MC /M⊙) < 6.5 begin to fall below the mass function at the high mass end.

We conclude that while both mass functions tested here can be ﬁt with a Schechter

68 0 β = -2.00 β = -2.00

Log(Mc) = 6.00 Log(Mc) = 5.50 -2 Log(Mc) = 6.50 Log(Mc) = 6.00 Log(Mc) = 7.00 Log(Mc) = 6.50 -4

Log(dN/dM) -6 6 < log(τ/yr) < 7 8 < log(τ/yr) < 8.6 -8 5 6 7 5 6 7

Log(M/MO .) Log(M/MO .)

Figure 3-9: Cluster mass functions fit with Schechter functions, represented β by dN/dM ∝ M exp(−M/MC ). On the left, clusters in the range 6 < log(τ/yr) < 7 are well fit by β = −2 and log(MC ) & 6.5. On the right, the 8 < log(τ/yr) < 8.6 mass function is fit by β = −2 and log(MC ) & 6.0. We find that the Schechter function is not necessary to fit the mass function, however (see Figure 3-8).

function, it is not required and a single power law provides a good fit to each. We therefore favor a pure power law in fitting the mass functions, and find little evidence for a truncation of the mass function at the high end.

3.5 How Does Distance Aﬀect the Observed Clus-

ter Distributions?

At a distance of 36 Mpc, NGC 3256 is nearly twice as far as the Antennae, the nearest example of a pair of actively merging galaxies. In order to assess how distance might impact the LFs, mass functions, and age distributions of cluster populations in

NGC 3256, we use the well studied HST images of the Antennae, and compare the resulting cluster distributions when we simulate the entire system to lie a factor of two further away.

We download the Antennae data from HST /WFC F 435W , F 550M, and F 814W

69 Figure 3-10: F 550M images of a portion of the Antennae taken with HST /WFC, however the image on the right has been artiﬁcially smoothed (resolution is half the image to the left) to simulate the distance of NGC 3256. Circles indicate source detections from DAOFIND. By smoothing the image, many faint sources and sources in crowded areas are not detected.

ﬁlters, as well as the WFC3 F 336W ﬁlter from the HLA. The combined image for each

ﬁlter is then boxcar smoothed with a kernel width of two pixels. We run DAOFIND on both the original and degraded F 550M image, tuning only the FWHM parameter

in both images, and we detect 8405 and 3173 objects, respectively. We refer to the

cluster catalog produced from the unaltered and degraded images as the “original”

and “image-smoothed” catalogs, respectively. We perform aperture photometry on

all detected sources and correct for foreground extinction. In addition, we measure

and apply separate aperture corrections to each catalog. Figure 3-10 shows a comparison of the original and image-smoothed F 550M images. The circles indicate detections from DAOFIND. It is evident that the image- smoothed source catalog is missing many of the very faint clusters in addition to some of the bright sources. We illustrate this in Figure 3-11, where the bottom panel shows the relative fraction of sources found in the image-smoothed catalog compared to the original catalog. We ﬁnd that ≈ 85% of the 3173 image-smoothed detections share coordinates (within one pixel) with their original image counterparts.

The top and middle panels of Figure 3-11 show the LFs for the original and smoothed-image cluster catalogs. Both LFs are ﬁt for mV < 22.7 (MV < −9), where

70 105 α = -2.13 ± 0.04 104

103

dN/dL 10

101 Original 100 105 α = -2.10 ± 0.04 104

103

2 dN/dL 10

101 Smoothed 100 100 80

Relative Completeness 0 24 22 20 18

mv Figure 3-11: Top: The luminosity function for the Antennae galaxies, constructed from a run using DAOFIND on the original F 550M image. Middle: Same as the top panel, but for the smoothed image. Note that nearly the same α value is obtained in both cases. Bottom: Relative fraction of sources found in the image- smoothed catalog compared to the original catalog.

71 the mV distribution begins to flatten out at the faint end. We find values for α of −2.13±0.04 for the clusters detected in the original image, very similar to those found in previous works. For the smoothed image sample we also find α = −2.10 ± 0.04. A similar and more extensive study was performed by Randriamanakoto et al. (2013), where it was also determined that α is not greatly impacted by resolution. They performed various convolutions on their image in order to simulate different galaxy distances, finding that α differed by no more than 0.2.

The color distributions of clusters brighter than MV < −9 from the original and smoothed images are quite similar. While there are more points from the original catalog present in Figure 3-12, the scatter in the colors among the two catalogs is roughly the same. Since ages are estimated from colors, similarity of the color distributions indicates that the age distributions for these two catalogs are also quite similar. We run SED ﬁtting on both catalogs and construct their age distributions

(not shown), ﬁnding γ ≈ −0.8 in both cases.

We ﬁnd that the LF, color distributions, and age distributions for the original and image smoothed catalogs are nearly the same, and we therefore conclude that distance does not signiﬁcantly hamper our ability to study clusters out to ≈ 40 Mpc.

3.6 How Eﬃciently Does NGC 3256 Form Clus-

ters?

Several works have suggested that galaxies with high rates of speciﬁc star formation (SFR of a galaxy divided by its mass) form clusters more eﬃciently than more quiescent galaxies (e.g., G10; Kruijssen et al. 2012). Two independent methods have been developed to test this, the CMF/SFR statistic (Chandar, Fall, & Whitmore

2015; hereafter CFW15) and Γ (Bastian et al. 2012). We apply each method to our clusters in NGC 3256 compared with those from previous works and diﬀerent

72 Figure 3-12: Color-color diagrams of the Antennae for clusters with MV < −9. Blue and red points indicate colors from the original and image-smoothed catalogs, respectively.

73 galaxies, and then discuss the results.

3.6.1 CMF/SFR Statistic

CFW15 compared the CMF between several diﬀerent galaxies by normalizing each by the SFR of its host galaxy, and comparing the amplitudes of the CMF/SFR statistic. The sample included spiral galaxies (M83 and M51), irregular galaxies (Large and Small Magellanic Clouds, NGC 4214, and NGC 4449), and the merging galaxy the

Antennae. Previous studies of these galaxies found that they ﬁt a “quasi-universal” picture of cluster formation and destruction, independent of their environment (Fall

& Chandar 2012). The CMFs for these galaxies, which have a large range of amplitudes, collapse when divided by their respective SFRs. The diﬀerence in CMF/SFR statistics is only σ ≈ 0.2, similar to the expected uncertainties. CFW15 concluded that cluster formation and disruption rates are similar across diﬀerent galaxies, and that these processes can be scaled by each galaxy’s SFR.

Here, we determine the CMF/SFR statistic for NGC 3256 for the ﬁrst time, and compare with the rest of the galaxies in the CFW15 sample. The mass functions discussed in Section 3.4.3 were shown to follow a power law with index β. We measure the CMF/SFR statistic only for clusters with ages in the range 6 < log(τ/yr) < 7,

−1 due to the smaller number of older clusters. We adopt an SFR of 50M⊙ yr from

Sakamoto et al. (2014). In Figure 3-13, we compare the CMF/SFR statistic for NGC

3256 with those for the seven galaxies shown in CFW15 and ﬁnd that it is similar to the other galaxies.

CFW15 quantify the scatter in the CMF/SFR statistics among the galaxies in their sample by ﬁtting the CMFs to

4 −1.9 dN/dM = A × SFR × (M/10 M⊙) (3.3)

74 and ﬁnd the scatter in the normalization coeﬃcient A to be σ(logA)=0.21. Using

NGC 3256 as an eighth galaxy, we remeasure the coeﬃcient and ﬁnd σ(logA)=0.24. The uncertainties in the CMF/SFR statistic come from uncertainties in the mass function and the SFR. Here we focus on the latter, and we refer the reader to Sec- tion 3.4.3.1 for further discussion on uncertainties regarding the mass function. The

−1 SFR of 50M⊙ yr from Sakamoto et al. (2014) was calculated using the bolometric luminosity calibration of Murphy et al. (2011) over an area comparable to that cov-

−1 ered by our data. This is quite similar to the SFR (46.17M⊙ yr ) adopted in G10, calculated from the total IR luminosity. We ﬁnd other modern calculations of the

−1 −1 SFR of NGC 3256 in the 45 − 50M⊙ yr range, with a high value of 85 M⊙ yr using IR luminosity (Rodr´ıguez-Zaur´ınet al. 2011). This is consistent with an SFR accuracy to at least a factor of two, as also found in CFW15. Also of importance is that while the coverage of these SFR determinations vary from the inner region to the entire galaxy, they are all in reasonable agreement because the vast majority of the IR luminosity resides in the center where this work focuses.

Figure 3-13 also includes the CMF/SFR statistic as calculated from the mass function and SFR used in G10. However, the amplitude of their CMF/SFR statistic does not agree as well with the other galaxies. Instead, it falls below the rest of the galaxies, and leftward of our own CMF/SFR statistic for NGC 3256. While this is in part due to their finding fewer clusters, the difference in amplitude is primarily caused by an offset in mass estimates, which is a result of their use of different SSP models

(see discussion in Section 3.3.1). We confirm this by performing two tests. We first use cluster photometry presented in G10 in our SED fitting procedure, finding that this produces a CMF/SFR statistic that falls much closer to the CMF/SFR statistics presented in CFW15 as well as this work. Secondly, in order to test the effect of the different cluster selection and different area covered, we find that if we restrict our catalog to only clusters that overlap with the area covered in G10, then the amplitude

75 3 NGC 3256 10 CFW15 G10 102

101

dN/dlogM 10

10-1 10-2 102 103 104 105 106 107

Log(M/MO .)

Figure 3-13: The normalized mass function for NGC 3256 clusters younger than log(τ/yr) < 7, shown in dark grey, while the lighter grey points show the normalized mass functions of several other galaxies calculated in Chandar, Fall, & Whitmore (2015). Red points show the CMF presented in G10, normalized to their quoted SFR.

of our mass function is the same as the mass function in G10 (with the horizontal oﬀset in masses still present).

3.6.2 The Γ Statistic

A second, similar statistic that has been widely used in the literature is Γ, deﬁned as the mass formed in bound clusters divided by the total stellar mass. The total mass in clusters in a given age range is typically found by summing the mass of all observed clusters above a given threshold value (which is determined from the completeness limit) and extrapolating to lower masses. This extrapolation is accomplished by either assuming a power law or Schechter function, or by using a model. Γ has been

76 estimated for several galaxies, and is typically plotted against the SFR density, ΣSFR,

of the host galaxy. G10 found a linear relation between log(Γ) and log(ΣSFR), while a theoretical study by Kruijssen (2012) showed that though the ﬁt may not be linear, a strong relation between the two quantities still exists, such that the higher the SFR density, the higher the value of Γ. This implies that galaxies with higher SFRs will contain a larger proportion of their total mass in clusters than those with lower SFRs.

This is diﬀerent from the result of CFW15, who found little variation from galaxy to galaxy for CMF/SFR.

+7.3 G10 estimated Γ for NGC 3256, and found Γ = 22.9%−9.8. This was calculated from the cluster formation rate (CFR) divided by the SFR, and it included only

log(τ/yr) < 7 clusters. We summarize their method below, and refer the reader to G10

for details. G10 found the total mass in clusters above a cutoﬀ of log(Mcut/M⊙)=4.7 and used a synthetic cluster population to ﬁnd the fraction of mass contained in

clusters that is above this cutoﬀ. They obtained the CFR by dividing their total

mass in clusters by 7 Myr, rather than 10 Myr, due to the expectation that the

clusters are embedded for the ﬁrst 3 Myr and hence are missing from the sample.

Lastly, they divided the CFR by the SFR and obtained Γ = 0.12. They made several corrections to Γ that were due to various underestimates of cluster mass. These came

from SED ﬁtting, uncertainty in metallicity, and cluster selection eﬀects. Each of

these correction factors increased Γ by 25%, resulting in a ﬁnal value of Γ = 0.23.

We follow the G10 method to estimate Γ from our larger cluster catalog in NGC

3256. We ﬁrst consider the mass in clusters above a cutoﬀ mass log(Mcut/M⊙)=5.0

7 and younger than 10 Myr, arriving at M(M > Mcut)=7.23 × 10 M⊙. We use a similar synthetic cluster population as G10 to calculate the fraction of mass contained

−1 in clusters for M(M > Mcut). We arrive at a cluster formation rate of 16.7M⊙ yr

by dividing Mtot by 7 Myr, and we calculate Γ = CFR/SFR = 0.33, assuming an

−1 SFR of 50 M⊙ yr . Lastly, we apply the same correction factors as G10, and arrive

77 atΓ=0.65. Our larger value is mainly due to the higher normalization of our CMF

(since we detect more clusters), as well as the diﬀerence in masses predicted by SSP

models (discussed further in Section 3.6.1).

Different assumptions have been made by previous authors when estimating Γ in previous works. Here, we assess the impact that different assumptions can have on the determined value of Γ. One of the biggest sources of uncertainty is the assumption in the slope of the cluster mass function, β. G10 found that a change as small as 0.2 in β can change the estimated value of Γ by as much as 50%. Further, Γ is highly sensitive to the mass range used in the extrapolation. If β = −2.0, then each decade in mass included in the mass function adds an equal percentage of total mass. For example, extrapolating down to a lower mass cutoff of, say, 1000M⊙ rather than 100M⊙ would increase the fraction of M(M > Mcut) by ∼ 10%. In Table 3.1, we compile different estimates of Γ made using different sets of assumptions. The estimates of Γ vary from

0.19 − 0.70, depending on the speciﬁc choice of parameters.

Given the large uncertainties in estimates of Γ for reasonable sets of assumptions,

we believe that the CMF/SFR statistic is a better method for comparing the relative

cluster and star forming eﬃciencies.

3.7 Summary and Conclusions

We have used ACS/WFC from HST to measure the properties of star clusters in the main body of NGC 3256. We draw the following conclusions.

1. The luminosity function follows a power law with index α = −2.23±0.07 where

mV < 21.5 for the sample combining inner and outer regions of NGC 3256. Measuring α for the inner and outer regions separately yields α ≈ −2.1. These

values agree with previous work.

2. We ﬁnd that the age distribution can be described by a power law with index 78 Table 3.1: Diﬀerent estimates of the fraction of stars forming in clusters (Γ) in NGC 3256.

β SFR (M⊙) ∆t (Myr) Γ -2.00 46.17 7 0.70 -2.00 46.17 10 0.49 -2.00 50.00 7 0.65 -2.00 50.00 10 0.46 -2.00 85.00 7 0.38 -2.00 85.00 10 0.27 -1.86 46.17 7 0.49 -1.86 46.17 10 0.34 -1.86 50.00 7 0.45 -1.86 50.00 10 0.32 -1.86 85.00 7 0.27 -1.86 85.00 10 0.19

γ ≈ −0.65 for independent mass ranges when considering catalogs from inner

and outer areas of the galaxy separately. Combining these two samples results

in γ ≈ −0.50. Our γ values are consistent with, although somewhat shallower

than, typical values of γ = −0.8 found in other systems.

3. The mass functions in various cluster age ranges are well described by a power

law with index β. Young (log(τ/yr) < 7) clusters follow a robust β = −1.86 ±

0.34. The mass functions for older clusters suﬀer from low number statistics,

and we measure β for clusters in the outer region of the center. We ﬁnd 7 < log(τ/yr) < 8 clusters are better described by β = −1.31 ± −0.36, while 8 <

log(τ/yr) < 8.6 clusters follow β = −2.08 ± 0.45. We investigate a number of

sources of uncertainty in β and ﬁnd that uncertainties agree with the formal

uncertainties given in Figure 3-8.

4. In order to test for the eﬀect that image resolution can have on cluster proper-

ties, we artiﬁcially degraded an image of the Antennae and created independent

catalogs from the degraded and original images. While the degraded image 79 found less than half of the sources detected in the original image, the LFs, color

distributions, and age distributions produced from both catalogs were very sim-

ilar. We conclude that reliable measurement of the ages and luminosities of star

clusters is not signiﬁcantly hampered by the distance to NGC 3256.

5. We considered two diﬀerent methods that measure the eﬃciency with which clusters form in a galaxy. The CMF/SFR statistic has been measured in other

systems and implies that cluster formation and destruction is similar in many

galaxies, regardless of SFR. This statistic was applied to NGC 3256 and found

to agree well with previous results. We calculated Γ for NGC 3256 and found

a range of possible values (0.19 < Γ < 0.70), while our best value of Γ = 0.65

disagrees with the previous Γ estimate by over a factor of two.

80 Chapter 4

The Cluster Populations of NGC

520 and NGC 2623

4.1 Introduction

In Chapter 3, we discussed NGC 3256, a major merger with ongoing star formation. We also brieﬂy discussed the ongoing star formation in the Antennae in

Chapter 1. In this Chapter, we consider two merging systems that seem to no longer be forming stars very actively, namely NGC 520 and NGC 2623.

NGC 520 is a dusty luminous infrared galaxy (LIRG) that is ≈ 31 Mpc away.

It exhibits two nuclei, one of which is enshrouded in dust (commonly referred to as the primary nucleus) and another to the northwest (referred to as the secondary nucleus). In addition to the cluster population in the tidal tails discussed in Chapter

2, it features clusters throughout its main body as well. These clusters have been studied previously, and rough ages have been determined but with conﬂicting results.

Stanford (1991) used optical spectroscopy to age-date the clusters and found that the clusters in the secondary nucleus (SN) are ∼ 100 Myr old, while the primary nucleus (PN) hosts ∼ 50 Myr old clusters. The dust surrounding the PN complicates the age-dating process, however, as Kotilainen et al. (2001) used near-infrared (NIR)

81 imaging to age-date the PN clusters as < 10 Myr old.

NGC 2623 is a more distant LIRG, at ≈ 79 Mpc. The system exhibits two tidal

tails and an oﬀ nuclear region of star formation (the pie wedge), all of which host star

clusters. The cluster population of the main body has not been well studied, likely

due to its distance and obscuration by dust. The conditions of star formation that determine the duration and intensity of a

merger-induced starburst can be predicted from simulations, but they rely on ob-

servations of the system as it appears currently. However, recent studies of cluster

systems in galaxies have established a relation between the cluster mass function and

star formation rate of the host galaxy (the CMF/SFR statistic discussed in Section

3.6.1). By turning the statistic around, we have established a method to estimate the SFR during the initial starburst based on the CMF. In this work, we study the

cluster populations in NGC 520 and NGC 2623 using HST data, and we use our new

method to estimate the SFR in past episodes of star formation.

This chapter is set up in the following manner. Section 4.2 covers the observations and cluster selection. Section 4.3 presents the luminosity functions, color distributions, and age distributions for the cluster sample. We estimate and present SFRs in select regions of these systems in Section 4.4. We discuss and interpret our results in

Section 4.5, and we present our conclusions in Section 4.6.

4.2 Observations and Cluster Selection

Observations of NGC 520 and NGC 2623 were taken with the ACS/WFC on HST as part of program GO-9735 (PI: Whitmore). Details of the F 435W , F 555W , and

F 814W observations for both systems were given in Section 2.2.1. Observations of both systems were taken with two additional ﬁlters as part of Program GO-9735; F 330W HRC observations were taken of NGC 520 in September 2004 (exposed for

82 Figure 4-1: I band image of NGC 520 (left) and NGC 2623 (right) showing U band HRC coverage.

5463 s) and of NGC 2623 in 2004 December (5465 s), and F 656N observations were taken of NGC 520 for 2480 s in August 2004 and NGC 2623 for 2400 s in 2004

February. The main body of NGC 520 exhibits clusters in two primary regions, which we refer to as the PN and SN. These regions are highlighted in Figure 4-1, where we indicate HRC U band coverage of both regions. Although there are some clusters outside of these regions, the dusty environment of NGC 520 hampers our ability to age-date clusters without U band, due to the age-extinction degeneracy (see Section

2.3.3.1 for further details on the age-extinction degeneracy). We therefore focus our cluster study only where U band coverage is available.

The entirety of the main body of NGC 2623 (including the pie wedge) was covered

in one HRC frame. Our catalog includes all cluster candidates that are within the

HRC frame that are not included in the tidal tail or pie wedge sample from Chapter

We follow our usual procedure for source detection and photometry, and we refer the reader to Section 2.2.2 for details. We use the BVI aperture corrections quoted in

Section 2.2.2 and calculate U band aperture corrections of 0.630 and 0.665 for NGC

520 and NGC 2623, respectively. Foreground U band extinction values are AU =

83 Figure 4-2: The left panel shows the concentration index versus mV for NGC 520 cluster candidates in the PN (red) and SN (blue). The right panel shows their mV distributions.

0.122 and 0.178 for NGC 520 and NGC 2623, respectively, while BVI extinction

values can be found in Section 2.2.2 and are taken from NED.

We select cluster candidates in NGC 520 with a cut in mV and concentration index, C, and we illustrate these cuts in Figure 4-2. The left panel shows mV versus C for all cluster candidates and likely foreground stars, shown as circles and asterisks,

respectively. The vertical line shows that all stars are more compact than C = 2.3.

The right panel shows the mV distribution for NGC 520 sources (after applying a C > 2.3 cut), and it is apparent that objects fainter than the horizontal line at

mV = 23.5 no longer follow a power law distribution. We therefore implement a cut

at mV = 23.5 to address incompleteness of our sample. Cluster selection in the main body of NGC 2623 suﬀers from multiple issues,

making a complete catalog diﬃcult to compile. As in the tidal tails and pie wedge

discussed in Chapter 2, clusters in NGC 2623 are eﬀectively point sources, rendering

84 Figure 4-3: BVI images of two nuclei of NGC 520, with detected clusters shown as green circles. The left panel shows the dusty PN of NGC 520. On the right is the SN with several unobscured clusters. the concentration index ineﬀective in distinguishing clusters from foreground stars.

While we attempt to implement a completeness cut in mV , the relatively low number of detected sources in the main body due to the high background make this task difficult. Regardless, we settle on a cut at mV = 24.5. The locations of NGC 520 cluster candidates that remain after implementing the preceding cuts are shown in Figure 4-3, and NGC 2623 cluster candidates are shown in 4-4. We find a total of 49 and 128 clusters in the PN and SN of NGC 520, respectively, while we find 65 clusters in the main body of NGC 2623.

4.3 Results

4.3.1 Luminosity Functions

Here we present the LFs of clusters in the main bodies of NGC 520 and NGC 2623,

α which are best described by dN/dL ∝ L . The LF is measured using the αconst−num method described in Section 2.3.2. The LF of the SN region of NGC 520 is shown in the left panel of Figure 4-5, with α = −2.28 ± 0.17. We are unable to measure the

LF for the PN region of NGC 520, because we detect too few clusters.

The slope of the LF of clusters in the main body of NGC 2623 was measured to be

85 Figure 4-4: BVI image of the main body of NGC 2623, with clusters marked as green circles.

NGC 520 - SN NGC 2623 - Main Body 105 103

α ± 104 = -2.28 0.17 α= -1.92 ± 0.33 102 103

dN/dL 102 dN/dL 101 101

100 100 24 23 22 21 20 19 25 24 23 22

mV mV

Figure 4-5: The luminosity function of cluster candidates in the SN of NGC 520 (left) and the main body of NGC 2623 (right).

86 α = −1.92 ± 0.33 and is shown in the right panel of Figure 4-5. Despite the relatively large ﬁtting uncertainty and cluster selection diﬃculties discussed in Section 4.2, the value here is toward the shallow end of what is found in many galaxies (Larsen 2002;

Whitmore et al. 2014).

4.3.2 Color Distributions

We present the color-color diagrams of NGC 520 and NGC 2623 clusters in Figure

4-6. On the left, red and blue colors correspond to PN and SN clusters of NGC 520, respectively. NGC 2623 clusters are shown on the right, with main body clusters shown in blue. We have overlaid the pie wedge color-color diagram in red for comparison. The BC06 track as presented in Section 2.3.3 is overlaid in each diagram to show the evolution of cluster color with age. The mean errors in color, which are calculated from the photometric uncertainty output by PHOT, are shown for each cluster population in the upper-left of each panel.

The reddening vectors in the B −V versus V −I color distributions (top panels of

Figure 4-6) are parallel to the BC06 track for log(τ/yr) > 8, resulting in a degeneracy of cluster age and reddening. With the inclusion of U band, however, this degeneracy can be broken; the reddening vectors in the bottom panels of Figure 4-6 are not parallel to the BC06 track. Therefore, we can use Figure 4-6 to obtain a rough estimate of the ages of clusters in NGC 520 and NGC 2623.

Figure 4-6 shows that clusters in the SN of NGC 520 have colors that are generally redder than PN clusters. This suggests that SN clusters are mostly older than PN clusters. In addition, PN clusters lie further from the BC06 track than SN clusters, suggesting a higher degree of reddening in the PN. This is consistent with the PN’s dusty appearance in Figure 4-3.

The fairly tight color grouping in the top-right panel of Figure 4-6 suggests that both the main body and pie wedge formed clusters over a relatively short period of

87 6 6 8 7 8 7 8.5 8.5 9 9 SN Main Body PN Pie Wedge

6 6 7 7

8 8

8.59 8.59

Figure 4-6: On the left are color-color diagrams of cluster candidates in and around both nuclei of NGC 520. On the right are color-color diagrams of cluster candidates in the nucleus (blue) and pie wedge (red) of NGC 2623. Reddening vectors are shown for AV = 1. In the upper left corner of each panel are the mean errors in color.

88 time, though the cluster population in the main body is older than the pie wedge.

The bottom-right panel, however, indicates more of a spread in the color of both

the main body and pie wedge. This spread is likely dominated by a relatively large

uncertainty in U − B color, however, rather than a spread in cluster age.

4.3.2.1 Ages

In Chapter 3, we obtained ages and masses of NGC 3256 clusters by ﬁtting cluster age and extinction simultaneously through the use of high S/N UBVI photometry and carefully calibrated Hα photometry. This treatment is not feasible in NGC 520 and

NGC 2623 due to the lower quality U band photometry and lack of Hα line emission in both galaxies. Instead, we implement the simpliﬁed age-dating methodology that was used in Section 2.3.3.2 (i.e., mapping the U − B and V − I cluster colors to the

BC06 track and ﬁnding the closest match).

In general, we ﬁnd that the majority of cluster ages span a relatively small range.

In the SN of NGC 520, as well as the main body of NGC 2623, clusters have ages between 8 < log(τ/yr) < 8.5. In Section 2.3.3.1, we found that the pie wedge clusters also span a small age range.

The ages in the PN are less clear, due to the large amount of reddening aﬀecting the cluster colors. It is possible that the majority of clusters have ages between 7 < log(τ/yr) < 8, or they are highly reddened log(τ/yr) < 7 clusters, or perhaps some combination. We discuss these possibilities further in Section 4.5.2.

4.4 Constraining Star Formation Rates Using the

CMF/SFR Statistic

In Chapter 3, we discussed the CMF/SFR statistic, which is a tool that has been used to compare cluster formation with the star formation rate of the host galaxy.

89 The CMF/SFR statistic was ﬁrst used in CFW15, where it was found that the cluster

mass function for their sample of seven galaxies can be normalized by the host galaxy’s SFR. In Section 3.6.1, we found that NGC 3256 also ﬁts the statistic of the other

galaxies remarkably well. In this section, we invert the problem and use the measured

CMF to estimate the SFR in diﬀerent regions.

4.4.1 Approximating the Cluster Mass Function

The mass function is constructed from individual cluster masses and takes the form dN/dM ∝ M β. Masses are calculated from the cluster luminosity and its mass-to- light ratio (see Section 3.4.3 for more details on mass calculation). While we are able to accurately measure cluster luminosity, we are only able to obtain rough estimates for ages, and thus mass-to-light ratios. However, in Section 4.3.2.1, we found that most of the regions we cover in NGC 520 and NGC 2623 are (approximately) coeval.

For simpliﬁcation, we will assume cluster formation in these regions is coeval, which is justiﬁed by the relatively small change in the mass-to-light ratio (a factor of 2 − 3) over the estimated range of cluster ages.

By assuming that all clusters for a given region have the same mass-to-light ratio, mass is only dependent on luminosity. In Section 4.3.1, we found that the slope of the LF is ≈ −2, which is similar to the slope of the mass function of many galaxies

(Fall 2006). We construct the mass function by multiplying luminosities of clusters by a single mass-to-light ratio, resulting in α = β. The age of the clusters in a single region is determined from the age associated with the peak in the color distribution, as described in Section 2.4.1. In addition to the ages of bulk cluster formation shown in Figure 2-15, we estimate the age of cluster formation in the SN of NGC 520. The color distribution and density distribution are shown in Figure 4-7, where we ﬁnd two peaks in color density. The primary peak occurs at colors associated with an age of 150 Myr, while the secondary peak occurs

90 6 7

8.5 9

Figure 4-7: Color-color diagram used to quantify when the peak cluster formation occurred in the SN of NGC 520. The right panel shows the density of clusters in color-color space, with contours overplotted. The peak is marked with a blue circle, while a secondary peak is marked with a green circle. The left panel is similar to Figure 4-6, with density peaks from the right panels included.

at 100 Myr. We construct the cluster mass functions from the ages obtained in these density distributions.

4.4.2 SFRs in the Main Bodies of Mergers

Fits are performed by ﬁnding the SFR that minimizes the equation

4 −1.9 ∆(dN/dM)=(A¯ − A) × SFR × (M/10 M⊙) , (4.1)

4 where A is the value of dN/dM evaluated at M = 10 M⊙, and A¯ is the mean value

4 of A evaluated at M = 10 M⊙ for the CFW15 galaxy sample. Uncertainties are determined from SFRs that result in (A¯ − A) = ±0.20, where 0.20 is the typical

dispersion in normalized dN/dM given by CFW15 for 100 <τ < 400 Myr clusters.

Figure 4-8 presents the estimated SFR of clusters in the SN of NGC 520. The

left panel shows the unnormalized cluster mass functions for galaxies in the CFW15

91 NGC 520 NGC 520 102 CFW15

101

100 dN/dlogM 10-1 +3.7 -1 SFR = 6.3 -2.3 MO . yr 10-2 102 103 104 105 106 1072 103 104 105 106 107

Log(M/MO .) Log(M/MO .)

Figure 4-8: The CMF/SFR statistic is used to estimate the SFR of the SN of NGC 520, assuming clusters formed roughly ≈ 150 Myr ago.

sample, with the mass function from NGC 520 overlaid. We show the 100 <τ < 400 Myr CFW15 clusters, because we estimate that the majority of cluster formation in

NGC 520 occurred ≈ 150 Myr. On the right panel of Figure 4-8, we show normalized

CFW15 mass functions in addition to the NGC 520 mass function normalized by the best ﬁt SFR from Equation 4.1. We predict that the SFR at the time of cluster

+3.7 −1 formation in the SN of NGC 520 is 6.3−2.3M⊙yr . We present an estimate of the SFR in the pie wedge in Figure 4-9. The mass function assumes a single burst 100 Myr old, which we obtained in Section 2.4.1.2.

We are unable to estimate the SFR in the main body of NGC 2623 with this method due to the high and variable background and low S/N found there.

4.4.3 SFRs in Tidal Tails of Mergers

In addition to regions of the main bodies of galaxies, we can also apply the

CMF/SFR to tail clusters. The physical mechanism that drives star formation in tidal tails is unclear, and constraining SFRs in tails may help to determine between shock-induced versus density-dependent star formation, as discussed brieﬂy in Section

92 Pie Wedge Pie Wedge 102 CFW15

101

100 dN/dlogM 10-1 +1.0 -1 SFR = 1.6 -0.6 MO . yr 10-2 102 103 104 105 106 1072 103 104 105 106 107

Log(M/MO .) Log(M/MO .)

Figure 4-9: The CMF/SFR statistic is used to estimate the SFR of the pie wedge of NGC 2623 when it formed, roughly ≈ 100 Myr ago.

2.1. The duration of cluster formation is relatively brief in tidal tails. In Chapter 2, we found that cluster formation in our galaxy sample lasts only a few tens of millions of years before shutting oﬀ. Using age estimates for the bulk of this cluster formation that we estimated in Chapter 2, we estimate the SFR for two tidal tails.

Figure 4-10 shows our estimate of the SFR in the southern tail of NGC 520 for a single burst of cluster formation 150 Myr ago. In Figure 4-11, we estimate the SFR in the eastern tail of NGC 3256 for a single episode of cluster formation at 250 Myr.

−1 Interestingly, both estimates of SFRs are ≈ 1M⊙ yr . We ﬁnd that there are too few clusters in the northern tail of NGC 520 as well as in both NGC 2623 tidal tails to apply this method.

4.4.4 Uncertainties

While the errors given for each estimation of SFR take into account the intrin- sic spread of the CMF/SFR statistic given by CFW15, there are other causes of uncertainties that we consider.

93 NGC 520S NGC 520S 102 CFW15

101

100 dN/dlogM 10-1 +0.6 -1 SFR = 1.0 -0.4 MO . yr 10-2 102 103 104 105 106 1072 103 104 105 106 107

Log(M/MO .) Log(M/MO .)

Figure 4-10: The CMF/SFR statistic is used to estimate the SFR of the SN of NGC 520, assuming clusters formed roughly ≈ 150 Myr ago.

NGC 3256E NGC 3256E 102 CFW15

101

100 dN/dlogM 10-1 +0.5 -1 SFR = 1.2 -0.3 MO . yr 10-2 102 103 104 105 106 1072 103 104 105 106 107

Log(M/MO .) Log(M/MO .)

Figure 4-11: The CMF/SFR statistic is used to estimate the SFR of the eastern tidal tail of NGC 3256, assuming clusters formed roughly ≈ 250 Myr ago.

94 As we mention in Section 4.4.1, the assumption of coevality of cluster formation introduces uncertainties in the mass-to-light ratio (and therefore the masses themselves) of a factor of 2−3. However, there are two reasons that this will only minimally impact the SFR estimate. The ﬁrst reason is that we ﬁt the mass function to log(M),

so misestimating a single cluster by a factor of 3 in M will come out to a factor of ≈ 0.5 in log(M). More importantly, while a few cluster masses may be incorrect by up to a factor of three because of uncertainties in their age, the majority are likely well represented by the single age we have estimated. To test the eﬀect that the assumption of coevality has on the estimated SFR, we recalculate cluster masses by using the mass-to-light ratio predicted for each cluster from the age ﬁtting discussed in 4.3.2.1. In every case, the new SFR estimation is well within the quoted errors.

4.5 Discussion

4.5.1 LFs Across the Galaxy Sample

The age and shape of the LF of the cluster population in the SN of NGC 520 are consistent with the statement in Section 2.4.3 that the LF tends to be shallower for younger mergers. We previously found that ages of NGC 520 clusters are well represented by a single burst 150 Myr ago, and their LF has an index of α = −2.28.

The pie wedge is the only younger population in our sample, with α = −2.02. All other tidal tails are older and possess steeper LFs.

It is also interesting to note that the SN and the southern tidal tail of NGC 520 have very similar LF slopes, but their SFRs are quite diﬀerent. In addition, while NGC 520S and NGC 3256E have similar SFRs, their LF values were very diﬀerent.

This is evidence against the weak negative correlation between α and SFR suggested in Whitmore et al. (2014).

95 4.5.2 Do NGC 520 and NGC 2623 Lack Recent Star Forma-

tion?

The main body of NGC 2623 contains many clusters with ages roughly 100 <

τ < 300 Myr, consistent with an enhanced period of cluster formation after the two galaxies began interacting ≈ 220 Myr ago. However, we ﬁnd a lack of τ < 10 Myr old clusters in NGC 2623, indicating no recent star formation. For NGC 520, the ages of clusters are less certain and are potentially 10 <τ < 100 Myr old, indicating little recent star formation. We contrast this to NGC 3256 and the Antennae, where star formation is clearly ongoing, and we address the question of whether NGC 520 and

NGC 2623 lack recent star formation and, if so, what physical mechanisms may be the cause.

4.5.2.1 Gas Dynamics

One initial guess may be simply that more recent interactions should host more recent star formation and older interactions are beyond the timescale for the starburst.

Of the galaxy sample in this study, simulations and cluster ages both support that

NGC 2623 is the youngest merger, followed by NGC 520, with NGC 3256 being the oldest. Most simulations place the interaction age of the Antennae at ∼ 200−260 Myr, making the Antennae as young or younger than any system in this study (although one paper ﬁnds the age to be 600 Myr; Privon et al. 2013 and references therein). Clearly, since the oldest and youngest interacting systems both show recent star formation, an understanding of the star formation timescale involves a more complex examination of the physics of merging galaxies.

Simulations that include gas dynamics indicate that the timing of star formation is dependent on both the structure of the progenitor galaxies (speciﬁcally, the presence of a bulge) and the recency of a close encounter. Multiple N-body simulations of

96 the Antennae (some also including SPH) predict that the Antennae has very recently

undergone or will soon undergo its second close passage, which would explain its presence of very young and massive clusters (Karl et al. 2010; Privon et al. 2013).

NGC 3256 is less well constrained, as there are no published N-body simulations of

the system. It exhibits two close together nuclei, supporting that it is a late stage

merger (English et al. 2003). Mihos & Hernquist (1996) found that simulations of

disk galaxies with bulges have a relatively weak initial burst of star formation, and

because much of the gas is not depleted immediately, a massive starburst occurs just before coalescence as the gas rapidly collapses. Therefore, the ongoing star formation

in NGC 3256 could be the result of the presence of bulges in its two progenitor

galaxies.

The N-body simulation of NGC 2623 predicts that while it has only undergone one passage 220 Myr ago, its nuclei coalesced ≈ 80 Myr ago (Privon et al. 2013). This implies that either the ﬁnal starburst in NGC 2623 recently ended, or perhaps one or both of its progenitor galaxies was bulgeless. In the latter case, the ﬁnal starburst was weak or did not happen at all.

As we discussed in Section 2.4.1.1, the interaction history of NGC 520 is more complicated because of the nearby dwarf galaxy UGC 957, and we found evidence that its passage may have caused the formation of the northern tidal tail. Since the system is not well simulated, we are unable to deduce when the starburst in the PN may have concluded or if it will start again while the galaxies coalesce.

4.5.2.2 Other Considerations

Another possible reason for the termination of star formation is the feedback from

AGN. It is currently unknown whether or not either the Antennae (Ueda et al. 2012 and references therein) or NGC 3256 (Sakamoto et al. 2014 and references therein) host an AGN, although their presence in either system cannot be ruled out. However,

97 NGC 520 also likely does not harbor an AGN, and we therefore do not attribute it

to the lack of recent star formation (Castangia et al. 2008 and references therein).

NGC 2623 is the only merger in our sample that is thought to host an AGN (Evans

et al. 2008 and references therein). This could potentially explain the lack of recent

clusters, even if the bulk of the gas began infalling toward the center recently. However without any simulations that incorporate AGN feedback or gas dynamics, its eﬀect

on the starburst is unclear.

Finally, we consider that NGC 520 could indeed exhibit ongoing star formation.

The PN of NGC 520 features a population of clusters with ages < 100 Myr. However,

as mentioned previously, the large amount of dust prevents a clear interpretation of

the ages of these clusters from UBVI photometry and Hα emission in the nucleus. We ﬁnd one pocket of possible Hα line emission near the edge of the PN that is

spatially coincident with the ﬁve bluest clusters in our sample. On the other hand,

we ﬁnd another possible region with Hα emission coincident with three extremely

red clusters; if these clusters were in fact < 10 Myr old, than they are enshrouded in

AV =2 − 4 magnitudes of extinction. This is entirely possible, as Stanford & Balcells

(1990) predicted that AV = 7 toward the PN. Lastly, Kotilainen et al. (2001) used NIR imaging to age-date clusters in the PN and found typical ages of log(τ/yr) ≈ 6.5.

However, these results are based on ground based observations with 0.6′′ to 0.7′′ seeing

(physically corresponding to 90 to 105 pc in NGC 520), and Bastian et al. (2014) found

that poor resolution (& 45 pc) can lead to artiﬁcially young ages for clusters when

using NIR colors.

It is therefore possible that the PN does indeed contain ongoing star formation. If this is the case, then the diﬀerence between NGC 520 and, say, NGC 3256, is that

nearly all of its clusters are deeply embedded. This could be possible, especially if

the PN is the remnant bulge from the progenitor galaxy that is currently nearly edge

on, as is thought to be the case (Stanford & Balcells 1991).

98 4.5.3 Timescales of Star Formation in NGC 520

We found from Figure 4-7 that the bulk of the star formation in the SN occurred

≈ 100 − 150 Myr ago, which is in agreement with estimates from Stanford (1991).

Stanford & Balcells (1991) predict that the two progenitor galaxies began interacting

≈ 300 Myr ago, putting cluster formation ≥ 150 Myr afterward. This apparent

delay in star formation might be reasonable, because gas outside of the inner region will take time to fall inward and begin the central starburst (Mihos & Hernquist

1996). Further, the lack of young clusters and Hα line emission indicates that very

little recent star formation has occurred, suggesting that the starburst in the SN has

concluded.

The apparent lack of older PN clusters is likely due to dust obscuration, rather

than their absence. Unfortunately, the lack of detection of τclus ∼ τtail clusters prevents our determination of a timescale of when the starburst began.

4.5.4 Conditions of Star Formation in NGC 520

−1 We found in Section 4.4.2 that the SFR in the SN of NGC 520 was ≈ 6.3M⊙ yr at the time of star formation. Stanford (1991) measured the current SFR using Hα

−1 ﬂuxes and far IR luminosities, ﬁnding that the SN has a current SFR of 0.02 M⊙ yr .

This implies that the SFR has dropped by a factor of ≈ 300 in the last 100 − 200 Myr. This is similar to the changes in SFR that are experienced by post-starburst galaxies (French et al., in preparation), and agree with predictions from simulations

(e.g., Mihos & Hernquist 1996).

In order to better understand the conditions of the starburst in the SN, we estimate

the gas density (Σgas) present at the time of star formation. We implement the ΣSFR

- Σgas relation found in Daddi et al. (2010) for starbursting galaxies, given as

99 −1 −2 −1 −2 logΣSFR/[M⊙yr kpc ]=1.14 × logΣgas/τdyn/[M⊙yr kpc ] − 0.62, (4.2)

2 where τdyn is the dynamical time. We estimate that the area over the SN is ≈ 4 kpc ,

−1 −2 −1 −2 which gives ΣSFR = 1.5M⊙ yr kpc and Σgas/τdyn = 5.0M⊙ yr kpc . These values are in the range of ΣSFR and Σgas values for (ultraluminous)LIRGS given in Daddi et al.

4.6 Conclusions

Using ACS/HST imaging, we have studied the cluster population in the merging galaxies NGC 520 and NGC 2623. We draw the following conclusions:

1. The main bodies of NGC 520 and NGC 2623 contain YMCs that formed as a

result of the ongoing interaction. The SN in NGC 520 contains clusters that

likely formed from a single burst 100−150 Myr ago, while the formation history

in the PN is less clear due to the dust in which it is enshrouded. The cluster

population of NGC 2623 is 100 − 300 Myr old, roughly consistent with their

formation at the time the galaxies began interacting.

2. Both galaxies do not obviously seem to exhibit current star formation that is

seen in other mergers such as NGC 3256 and the Antennae. We suggest for

NGC 2623 that during the ﬁnal stage of merging, either there was little gas

and the minimal subsequent starburst has since ended, or this star formation

was quickly quenched due to the probable presence of an AGN. In the case of

NGC 520, it is possible that there are young star clusters present in the PN, but the majority of them are highly extinguished from dust. The nearly edge

on orientation of the galaxy that hosts the PN supports this possibility.

100 3. Using the CMF/SFR statistic from CFW15, we estimated the SFR at the time

of cluster formation for the SN and southern tidal tail of NGC 520, the pie wedge of NGC 2623, and the eastern tidal tail of NGC 3256. The highest estimated

+3.7 −1 +1.0 SFR is in the SN, with 6.3−2.3M⊙ yr . The pie wedge has an SFR 1.6−0.6M⊙

−1 +0.5 −1 yr , while the tails NGC 3256E and NGC 520S have SFRs of 1.2−0.3M⊙ yr

+0.6 −1 and 1.0−0.4M⊙ yr , respectively.

4. We estimate the SFR density and gas density in the SN of NGC 520 from its

−1 −2 −1 estimated SFR and ﬁnd ΣSFR =1.5M⊙ yr kpc and Σgas/τdyn =5.0M⊙ yr kpc−2; both of these values lie in the range observed in ULIRGS. In addition,

−1 the current SFR of 0.02M⊙ yr implies that the SFR has dropped by factor of

∼ 300, which is not uncommon for post-starburst galaxies.

5. The LF of clusters in the SN of NGC 520 follows a power law with slope α =

−2.28 ± 0.17, and the LF of NGC 2623 main body clusters is similar with

α = −1.92±0.33. However, the LF of NGC 2623 may be artiﬁcially shallow due to incompleteness. The LF of NGC 520 is consistent with our claim in Chapter

2 that more recent bursts of cluster formation seem to exhibit shallower LFs.

101 Chapter 5

Conclusions and Future Work

In this thesis, we have studied the star cluster populations in both the main body and tidal tails of three nearby merging galaxies. Surprisingly, the processes of cluster formation and disruption appear to be similar to those found in more quiescent systems. Because of this similarity, we are able to use star clusters as a tool to better understand the properties of mergers and the conditions in which they form stars.

5.1 Cluster Evolution in Merging Galaxies

The luminosity function of YMCs in both the main bodies and tidal tails of merging galaxies follows a power law dN/dL ∝ Lα with −2.6 <α< −1.9. These values are similar to those found for clusters in a variety of galaxy environments.

The mass function and age distribution for NGC 3256 clusters suggest that the galaxy follows the “quasi-universal” picture of cluster formation and destruction. The mass function of NGC 3256 clusters follows a power law dN/dM ∝ M β and is roughly independent of cluster age. We found β = −1.86 ± 0.34 for τ < 10 Myr clusters and β = −2.08 ± 0.45 for 100 − 400 Myr old clusters. The age distribution also follows a power law with dN/dτ ∝ τ γ , with γ ≈ −0.67 ± 0.08 and is independent of cluster mass. This is consistent with a bivariate distribution g(M, τ)= M βτ γ with β ≈ −2 and γ ≈ −1 that has been found in a variety of diﬀerent galaxies. 102 The mass function of NGC 3256, when normalized by its SFR, falls amongst those

of seven other galaxies. These galaxies are mostly typical spirals and irregular galax-

ies, suggesting that the cluster formation in NGC 3256 is similar to other galaxies.

We considered the claim that the cluster formation eﬃciency (Γ) increases with a

galaxy’s SFR density, which predicts that NGC 3256 should have a higher value of Γ than normal spirals and irregular galaxies. However, we found that the uncertainties

on Γ were too large to support or reject the prediction.

5.2 Using Clusters as a Tool

Simulations are a useful way to understand the physics of galaxy interactions.

However, simulations must cover a parameter space, and the results usually have degeneracies amongst the parameters as a result of a lack of observational constraints.

An empirical understanding of cluster formation and evolution in mergers leads to better constraints on simulations.

We found that the interaction ages of mergers provided by N-body simulations are well supported by the age of the cluster population in the tidal tails of these galaxies. The ages of these tails are critical to accurate simulations, as their unique morphology provides direct comparisons of simulations and observations. We have also established that cluster formation in the tails takes place along the tail center, where the gas is densest, and it is converted to stars relatively quickly (within several tens of millions of years). In the main bodies of these merging systems, we constrained properties of their progenitor galaxies by studying their SFH.

The link between star and cluster formation in galaxies has allowed us to turn the

CMF/SFR relation around and estimate SFRs for galaxy environments with a well

established cluster population. By constraining SFRs for regions both in the main bodies and in tidal tails of galaxies, we were able to estimate gas densities at the

103 time of cluster formation for these regions. We found that in the main body of NGC

520, the gas densities are well within the typical range of (U)LIRGs, supporting the

accuracy of this method.

These ﬁndings will provide constraints on future detailed N-body and SPH simulations of these galaxies.

5.3 Future Work

While we have made progress toward understanding the stellar populations in merging galaxies, more and better quality data could further solidify our ﬁndings.

The CMF/SFR statistic is a promising method to estimate the SFR of a galaxy in the past. However, the galaxy sample size for which it has been tested is small, and it is imperative that independent measurements of SFRs of starbursting systems be estimated to compare with SFRs predicted from the CMF/SFR statistic. We obtained rough ages of star clusters in tidal tails in Chapter 2 by assuming that there was negligible extinction in the tails. While this is likely true, obtaining U band and Hα photometry with an instrument such as HST /WFC3 would break any

possible degeneracy in age and extinction, as well as much more tightly constrain the

ages of these clusters. With better ages, we would deﬁnitively establish whether or

not an oﬀset exists between the formation of the tidal tail and the onset of cluster formation.

We also found that all of the mergers in our sample host star clusters in their tails, prompting the question of whether or not all tails host star clusters. Some previous studies have claimed that some tidal tails do not form clusters, or at least that there are few there today. NGC 3256E and NGC 520N were suggested to be among these cluster-less tails, however we found clusters in them with better observations. We conclude it is likely that most, and perhaps all, tidal tails form clusters. Good

104 candidates in which to search for tidal tail clusters are NGC 3690, NGC 1614, NGC

7252, and NGC 3921. In addition to ﬁnding them, constructing the mass functions

and estimating SFRs during the formation of these tails could provide insight into

the conditions in which the clusters form. In particular, NGC 7252 could prove to be

interesting, as the system has been fully simulated using SPH+N-body codes, from which its SFH has been predicted (Chien & Barnes 2010).

One of the biggest diﬃculties we faced was detecting and age-dating star clusters in the dusty primary nucleus of NGC 520 and the main body of NGC 2623. An attempt to penetrate this dust in NGC 520 was made by Kotilainen et al. (2001), using ground based NIR imaging, however the limited resolution likely aﬀected their estimated ages. With the advent of the James Webb Space Telescope, it will be possible to obtain accurate NIR colors of embedded clusters with instruments such as the near-infrared camera (NIRcam); NIRcam will have resolutions of ≤ 0.065 ′′/pixel, comparable to the ACS/WFC on HST.

105 References

Barnes, J. E. & Hibbard, J. E. 2009, AJ, 137, 3071

Bastian, N. 2008, MNRAS, 390, 759

Bastian, N., Adamo, A., Gieles, M., et al. 2012, MNRAS, 419, 2606

Bastian, N., Adamo, A., Schirmer, M., et al. 2014, MNRAS, 444, 3829

Bastian, N., Hempel, M., Kissler-Patig, M., Homeier, N. L., & Trancho, G. 2005, A&A, 435, 65

Bik, A., Lamers, H. J. G. L. M., Bastian, N., Panagia, N., & Romaniello, M. 2003, A&A, 397, 473

Bruzual, G. & Charlot, S. 2003, MNRAS, 344, 1000

Cameron, E., & Carollo, C. M., Oesch, P. A., et al. 2011, ApJ, 743, 146

Castangia, P., Tarchi, A., Henkel, C., & Menten, K. M. 2008, A&A, 479, 111

Chabrier, G. 2003, PASP, 115, 763

Chandar, R., Fall, S. M., & Whitmore, B. C. 2010, ApJ, 711, 1263

Chandar, R., Fall, S. M., & Whitmore, B. C. 2015, ApJ, 810, 1

Chandar, R., Whitmore, B. C., Calzetti, D., et al. 2011, ApJ, 727, 88

Chandar, R., Whitmore, B. C., Kim, H., et al. 2010, ApJ, 719, 966

Chien, L.-H. & Barnes, J. E. 2010, MNRAS, 407, 43

Daddi, E., Elbaz, D., Walter, F., et al. 2010, ApJL, 714, 118

106 de Grijs, R., Lee, J. T., Mora Herrera, M. C., Fritze-v. Alvensleben, U., & Anders, P. 2003, NewA, 8 155

Dekel, A., Sari, R., & Ceverino, D. 2009, ApJ, 703, 785

English, J., Norris, R. P., Freeman, K. C., & Booth, R. S. 2003, AJ, 125, 1134

Evans. A. S., Vavilkin, T., Pizagno, J., Modica, F. et al. 2008, ApJ, 675, L69

Fall, S. M. 2006, ApJ, 652, 1129

Fall, S. M. & Chandar, R. 2012, ApJ, 752, 96

Fall, S. M., Chandar, R., & Whitmore, B. C. 2005, ApJ, 631, 133

Fall, S. M., Chandar, R., & Whitmore, B. C. 2009, ApJ, 704, 453

Fall, S. M. & Zhang, Q. 2001, ApJ, 561, 751

Font, A. S., McCarthy, I. G., Crain, R. A., et al. 2011, MNRAS, 416, 2802

Gallagher, S. C., Durrell, P. R., Elmegreen, D. M., et al. 2010, AJ, 139, 545

Gieles, M., Larsen, S. S., Scheepmaker, R. A., Bastian, N., Haas, M. R., & Lamers, H. J. G. L. M. 2006, A&A, 446, 9

Goddard, Q. E., Bastian, N., & Kennicutt, R. C. 2010, MNRAS, 405, 857 (G10)

Horellou, C. & Koribalski, B. 2003, Ap&SS, 284, 499

Karl, S. J., Naab, T., Johansson, P. H., et al. 2010, ApJL, 715, L88

King, I. 1966, AJ, 71, 64

Knierman, K. A., Gallagher, S. C., Charlton, J. C., et al. 2003, AJ, 126, 1227 (K03)

Kotilainen, J. K., Reunanen, J., Laine, S., & Ryder, S. D. 2001, A&A, 366, 439

Kruijssen, J. M. D. 2012, MNRAS, 426, 3008

Lackner, C. N., Cen, R., Ostriker, J. P., & Joung, M. R. 2012, MNRAS, 425, 641

107 Lada, C. J. & Lada, E. A. 2003, ARA&A, 41, 57

Larsen, S. S. 1999, A&AS, 139, 393

Larsen, S. S. 2009, A&AS, 494, 539

Leitherer, C. & Heckman, T. M. 1995, ApJSS, 96, 9

McCrady, N., Gilbert, A. M., Graham, J. R. 2003, ApJ, 596, 240

McKee, C. F. & Ostriker, E. C. 2007, ARA&A, 45, 565

Mihos, C., Bothun, G. D., & Richstone, D. O. 1993, ApJ, 418, 82

Mihos, C. & Hernquist, L. 1994, ApJ, 437, 611

Mihos, C. & Hernquist, L. 1994, ApJ, 464, 641

Mulia, A. J., Chandar, R., & Whitmore, B. C. 2015, ApJ, 805, 99

Mullan, B., Konstantopoulos, I. S., Kepley, A. A., et al. 2011, ApJ, 731, 93 (M11)

Murphy, E. J., Condon, J. J., Schinnerer, E., et al. 2011, ApJ, 737, 67

Naab, T., Johansson, P. H., & Ostriker, J. P. 2009, ApJL, 699, L178

Navarro-Gonz´alez, J., Ricciardelli, E., Quilis, V., & Vazdekis, A. 2013, MNRAS, 436, 3507

Oser, L., Ostriker, J. P., Naab, T., Johansson, P. H., & Burkert, A. 2010, ApJ, 725, 2312

Portegies Zwart, S. F., McMillan, S. L. W., & Gieles, M. 2010, ARA&A, 48, 431

Privon, G. C., Barnes, J. E., Evans, A. S., et al. 2013, ApJ, 771, 120

Randriamanakoto, Z., V¨ais¨anen, P., Ryder, S., Kankare, E., Kotilainen, J., & Mattila, S. 2013 MNRAS 431, 554

Rodr´ıguez-Zaur´ın, J., Arribas, S. Monreal-Ibero, A., Colina, L., Alonso-Herrero, A., & Alfonso-Garz´on, J. 2011, A&A, 527, 60

108 Sakamoto, K., Aalto, S., Combes, F., Evans, A., & Peck, A. 2014, ApJ, 797, 90

Salpeter, E. E. 1955, ApJ, 121, 161

Sirianni, M., Jee, M. J., Ben`ıtez, N., et al. 2005, PASP, 117, 1049

Stanford, S. A. 1990, ApJ, 381, 409

Stanford, S. A. & Balcells, M. 1990, ApJ, 355, 59

Stanford, S. A. & Balcells, M. 1991, ApJ, 370, 118

Toomre, A. 1977, in The Evolution of Galaxies and Stellar Populations, ed. B. M. Tinsley & R. B. Larson (New Haven, CT: Yale Univ. Observatory), 401

Toomre, A. & Toomre, J. 1972, ApJ, 178, 623

Tran, H. D., Sirianni, M., Ford, H. C., et al. 2003, ApJ, 585, 750

Trancho, G., Bastian, N., Schweizer, F., & Miller, B. W. 2007, ApJ, 658, 993

Ueda, J., Iono, D., Petitpas, G., Yun, M. S., Ho, P. T. P., Kawabe, R., et al. 2012, ApJ, 745, 65

Whitmore, B. C., Chandar, R., Bowers, A. S. et al. 2014, AJ, 147, 78

Whitmore, B. C., Chandar, R., & Fall, S. M. 2007, AJ, 133, 1067

Whitmore, B. C. & Schweizer, F. 1995, AJ, 109, 960

Whitmore, B. C., Zhang, Q., Leitherer, C., Fall, S. M., Schweizer, F., & Miller, B. W. 1999, AJ, 118, 1551

Zepf, S. E., Ashman, K. M., English, J., Freeman, K. C., & Sharples, R. M. 1999, AJ, 118, 752

Zhang, Q. & Fall, S. M. 1999, ApJ, 527, 81

Zolotov, A., Willman, B., Brooks, A. M., et al. 2009, ApJ, 702, 1058

109