A Dissertation. entitled. Star Cluster Populations in the Spiral Galaxy M101. Lesley A. Simanton

Transcription

1 A Dissertation entitled Star Cluster Populations in the Spiral Galaxy M101 by Lesley A. Simanton 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. John-David Smith, Committee Member Dr. Steven Federman, Committee Member Dr. Bo Gao, Committee Member Dr. Bradley Whitmore, Committee Member Dr. Patricia R. Komuniecki, Dean College of Graduate Studies The University of Toledo August 2015

2 Copyright 2015, Lesley A. Simanton This document is copyrighted material. Under copyright law, no parts of this document may be reproduced without the expressed permission of the author.

3 An Abstract of Star Cluster Populations in the Spiral Galaxy M101 by Lesley A. Simanton Submitted to the Graduate Faculty as partial fulfillment of the requirements for the Doctor of Philosophy Degree in Physics The University of Toledo August 2015 Most stars form in groups and clusters. Stars clusters range in age from very young (< 3 Myr, embedded in gas clouds) to some of the most ancient objects in the universe (> 13 Gyr), providing clues to the formation and evolution of their host galaxies. Our knowledge of the diversity of star cluster populations has expanded over the last few decades, especially by being able to examine star clusters outside of the Milky Way (MW). In this dissertation, we continue this expansion of extragalactic star cluster studies by examining the star cluster system of the nearby spiral galaxy M101. We utilize photometry from Hubble Space T elescope images to assess luminosity, color, size, and spatial distributions of old star clusters, and spectroscopy from the Gemini- North telescope to determine ages, metallicities, and velocities of a subset of both young and old clusters in M101. We find that the range of cluster luminosities, ages, and metallicities in M101 is nearly continuous. We discover a population of fairly massive, old disk clusters, and conclude that the disk of M101 may have had a higher rate of cluster formation in the past than in the MW, and that it may be better suited to cluster survival. We find evidence that some clusters in M101 have intermediate ages of several Gyr, whereas the MW has few such clusters. Our analysis of the velocities of young clusters suggests that they rotate with the HI gas disk, while the old globular clusters appear to be in the halo. iii

4 This work is dedicated to my all of family and friends, especially to my parents, Robert and Joyce Simanton, and my great aunt, Donna Simanton, who made my education possible; to my sister and lifelong friend, Wendy Holland; and to my very supportive and motivating fiancé, Jim Coogan.

5 Acknowledgments I would like to thank my advisor, Rupali Chandar, for all of her support, encouragement, and contributions to this work. She inspired my interest and excitement in star clusters and led me through many challenges and difficulties throughout my research. I would also like to thank collaborators Brad Whitmore and Bryan Miller for their contributions to this work and for answering my questions promptly and with genuine interest in my education and discoveries. I would like to thank my other committee members, Steve Federman, J. D. Smith, and Bo Gao for taking the time to give crucial input and criticisms to help shape the results and interpretation presented here. Finally, I would like to thank the graduate students at the University of Toledo Department of Physics and Astronomy who have furthered my understanding of physics and astronomy by helping me find the answers to questions, engaging in discussions, sharing papers, and empathizing in the sometimes frustrating day-to-day necessities of conducting scientific research. v

6 Contents Abstract iii Acknowledgments v Contents vi List of Tables ix List of Figures x List of Abbreviations xii List of Symbols xiii 1 Introduction What are Star Clusters, and What Can They Tell Us about Galaxies? The Current Picture of Galaxy Formation and Evolution: How do spiral galaxies fit in? Models for Studying Extragalactic Star Clusters Two Populations of Old Star Clusters in the Spiral Galaxy M101 Based on HST/ACS Observations Background Observations, Cluster Selection, and Completeness Observations vi

7 2.2.2 Cluster Selection Completeness Results and Analysis Cluster Luminosity Distribution Cluster Colors and Luminosities Sizes Spatial Distribution Discussion Are There Two Populations of Old Clusters in M101? What Are the Faint, Red Clusters in M101? Conclusions Gemini/GMOS Spectra of Old and Young Star Clusters in M101: Ages and Metallicities Background Observations Cluster Candidate Selection with HST Gemini/GMOS Observations Results and Analysis Initial Age Categorization Measuring Ages and Metallicities Using BaSTI Models Ages and Metallicities Spatial-Metallicity Distribution Discussion Are There Young to Intermediate Age GCs in M101? Metallicities of the YMCs and GCs Conclusions vii

8 4 Gemini/GMOS Spectra of Old and Young Star Clusters in M101: Kinematics Background Results and Analysis Velocity Measurements Velocity Distance Comparison Age Velocity Comparison Rotational Velocity Calculation and Comparison Discussion The Structure of M Conclusions Conclusions and Future Work Searching for Faint, Old Star Clusters in Other Spiral Galaxies viii

9 List of Tables 2.1 M101 Old Star Cluster Catalog Coefficients for Surface Density Fits Spatial and Photometric Properties of M101 Clusters with Spectra Spectroscopic Properties of M101 Cluster Spectra Kinematics of M101 Clusters with Spectra Coefficients for v cluster vs. R semi-minor Fits Rotational Velocity and Velocity Dispersion Comparisons ix

10 List of Figures 1-1 Examples of MW Star Clusters Examples of LMC and M101 Star Clusters HST ACS/WFC Fields Covering M Radial Profiles of Stars vs. Clusters HST BV I Color Images of Typical Red M101 Star Clusters Fraction of Selected Artificial Clusters vs. m v Fraction of Selected Artificial Clusters vs. r gc Magnitude and Luminosity Distributions for Red M101 Clusters and MW GCs B V vs. V I for Red Clusters and YMCs in M B I Color Histograms of Red M101 Clusters B V Color Magnitude Diagram for Red M101 Clusters, MW GCs, LMC Intermediate Age Clusters, and MW Old Open Clusters r eff Distribution for Red M101 Clusters and MW GCs M V vs. r eff for Red M101 Clusters and MW GCs M V vs. r eff for Red M101 Clusters and Artificial Clusters Positions of Red M101 Clusters r gc Distribution for Red M101 Clusters Surface Density Distributions of Red M101 Clusters and MW GCs Gemini-North Images of GMOS Mask x

11 3-2 Gemini-North Images of GMOS Mask HST BV I Color Images of M101 YMCs with Spectra HST BV I Color Images of M101 GCs with Spectra GMOS Spectra of Mask 1 YMCs GMOS Spectra of Mask 1 YMCs Continued GMOS Spectra of Mask 2 YMCs GMOS Spectra of Mask 1 GCs GMOS Spectra of Mask 1 GCs Continued GMOS Spectra of Mask 2 GCs Lick/IDS Index Defined Regions Lick/IDS Index Defined Regions Continued Index-index Model Grids From BaSTI Synthetic Spectra for Old Ages Index-index Model Grids From BaSTI Synthetic Spectra for Young Ages Example of Rejected vs. Accepted Lick/IDS Indices for Fitting Ages and Metallicities [Fe/H] vs. log Age for M101 YMCs and GCs with Spectra [Fe/H] vs. r gc for M101 YMCs and GCs with Spectra Example Correlation Functions for an M101 YMC and GC Positions of the M101 YMCs and GCs with Spectra HI Gas Map with Line-of-Sight Velocity Contours v cluster vs. R semi-minor for the M101 YMCs and GCs with Spectra v cluster v disk vs. log Age for M101 YMCs and GCs with Spectra v rot vs. r gc for the M101 YMCs and HI Gas HST /WFC3 Field in NGC xi

12 List of Abbreviations MW Milky Way GC globular cluster YMC young massive cluster LMC Large Magellanic Cloud SSP simple stellar population LF luminosity function HST Hubble Space T elescope WFPC Wide Field Planetary Camera 2 ACS Advanced Camera for Surveys CMD color magnitude diagram WFC Wide Field Channel PSF point spread function S/N signal-to-noise ratio FWHM full width at half maximum BCG brightest cluster galaxy GMOS Gemini Multi-Object Spectrograph BaSTI Bag of Stellar Tracks and Isochrones IDS image dissector scanner xii

13 List of Symbols M solar mass α right ascension δ declination m V apparent V band magnitude M V absolute V band magnitude α exponent for the power law luminosity distribution a slope of a straight line on a log space luminosity distribution plot, related to α by α = (2a + 1) r eff effective radius of a cluster r gc galactocentric radius which is the distance from an object to the center of the host galaxy projected onto the plane of the sky R e effective radius of a galaxy component from fitting a de Vaucouleurs profile R scale radius of a galaxy component from fitting an exponential profile µ ev stellar mass loss rate from a cluster τ age of a cluster population ρ h half light density of a cluster W λ equivalent width of a spectral line F λ flux of a spectral line F C flux of a spectral continuum v cluster line-of-sight velocity of a cluster v disk line-of-sight velocity from HI gas in the disk σ velocity dispersion of a cluster population v rot rotational velocity of a cluster or cluster population xiii

14 Chapter 1 Introduction 1.1 What are Star Clusters, and What Can They Tell Us about Galaxies? Stars typically form in groups and clusters rather than in isolation. Clusters of stars are found to span ages from young (< 3 Myr, embedded in gas clouds) to ancient (> 13 Gyr), making them excellent tracers of the formation and evolution of stars as well as the galaxies in which they reside. We can therefore utilize properties of systems of star clusters to learn about the history of their host galaxies, adding to the body of knowledge on galactic formation and evolution. Star clusters observed within our own Milky Way (MW) galaxy have been studied in great detail due to their proximity and form the rubric for our current understanding of star clusters. There are two main types of star clusters seen in the MW: globular clusters (GCs) and open clusters. MW GCs are massive (10 5 or more stars, M ) and ancient (> 10 Gyr), among the oldest detected objects in the universe, while open clusters are less massive ( stars, < 10 3 M ) and younger (typically < 100 Myr). GCs are located in the spheroidal components of the MW, i.e. the bulge and halo, while open clusters lie in either the thin or thick disks (Portegies Zwart et al., 2010). 1

15 Studying the properties of the MW GC system has led to insights on the history of the MW and galaxy formation. For example, Searle & Zinn (1978) used the abundances and spatial distribution of GCs to first propose the idea that the outer regions of the MW could have been formed from clumpy gas accreted later than the initial formation of the galaxy, thus paving the way for the idea of mergers/accretion forming and evolving galaxies over time (discussed further in 1.2). The MW, however, may not be a typical example of a galaxy star cluster system. Observations from the last two decades have shown that young massive clusters (YMCs, ages < 1 Gyr, masses up to M ) exist, perhaps in large numbers behind the gas and dust within the disk of the MW, and commonly in nearby star forming galaxies, including irregulars (e.g., the Magellanic Clouds), spirals (e.g., M83 and M51), and mergers (e.g., the Antennae and NGC 3256) (Larsen et al., 2011; Chandar et al., 2010; Trancho et al., 2007; Larsen, 2000; Whitmore & Schweizer, 1995; Fischer et al., 1992). Intermediate age GCs have been found in many galaxies such as the Large Magellanic Cloud (LMC, e.g., Palma et al. (2013); Goudfrooij et al. (2011); Piatti et al. (2009); Mucciarelli et al. (2007); Kerber et al. (2007)), M33 (Chandar et al., 1999), and M31 (Puzia et al., 2005b), among other galaxies. Essentially, star clusters appear to have formed fairly continuously in nearby star forming galaxies. The ages, masses, and other properties of the clusters have helped further shape our understanding of typical galaxy histories, structure, and formation. Clusters lose mass throughout their lifetimes. One of the best studied processes is via two-body interactions where stars are flung out at speeds exceeding the escape speed of the cluster (e.g., Spitzer & Harm (1958)). This is, at least in part, how galaxies become populated with field stars. Clusters that lose enough mass to become unbound are considered to be disrupted. Such cluster disruption affects the lowest mass clusters first as their escape speeds are lower, and stars become unbound more readily. For the MW, this explains the apparent lack of low mass clusters with very 2

16 old ages (Fall & Zhang, 2001); however, we are still determining just how universal the process of cluster disruption is. A detailed understanding of the processes that disrupt clusters is an active area of research as there are other possible disruption mechanisms that may be influential, such as disruption from tidal forces or interactions with passing gas clouds (e.g., Fall & Chandar (2012) and references therein). 1.2 The Current Picture of Galaxy Formation and Evolution: How do spiral galaxies fit in? Currently, the most widely accepted scenario of galaxy formation is the hierarchical model, which generally states that galaxies are built up through mergers with some differences in the details of formation scenarios based on the mass ratios of typical mergers and the amount of gas present. The two-phase scenario is well supported (e.g., Dekel et al. (2009); Naab et al. (2009); Zolotov et al. (2009); Oser et al. (2010); Font et al. (2011); Lackner et al. (2012); Navarro-González et al. (2013)). The first phase of the two-phase scenario is early (z > 2) dissipative collapse of the gas that forms the galaxy. The star formation during this phase is considered in-situ and is gas-rich. The second phase is composed of largely gas-poor minor mergers that transfer ex-situ stars to the new host galaxy, and mostly occurs later (z < 2). We expect in-situ components to be located in the inner regions of galaxies and be more metal-rich with steeper metallicity gradients. Conversely, the ex-situ components should lie in the outer regions of galaxies and be metal-poor with a shallower metallicity gradient. This scenario is an oversimplification, as apparent from the diversity of galaxy structure observed in our universe. Galaxy morphological types are commonly categorized on the Hubble sequence (expanded by de Vaucouleurs (1963)), in which early-type galaxies are elliptical galaxies (Es, spheroidal and gas-poor) and lentic- 3

17 ular galaxies (S0s, smooth disk structure with dominant spheroidal components) and late-type galaxies are spiral galaxies (Sa-Sd, disk structure with gas-rich spiral arms and varying degrees of spheroidal components). Spiral galaxies themselves are often described as either early or late type depending on how pronounced the spiral structure is and how prominent the inner spheroidal component, known as the bulge, is. Spiral galaxies also undergo secular evolution which is a slow evolution driven by internal structures such as spiral arms and bars rather than being driven solely by mergers. Secular evolution is thought to build up so-called pseudobulges which are smaller bulge-like features in galaxies that are otherwise bulgeless. The extent of the influence of secular evolution in current epochs is somewhat unknown as we see evidence of both mergers and secular evolution with the plausible picture being a transition from a merger dominated universe at early times to a secularly dominated universe far into the future (Kormendy & Kennicutt, 2004). Spiral galaxies with significant bulge components contain metal-rich GC populations associated with their bulges (such as in the MW (Minniti, 1995) and M31 (Brown, 2009)). The presence of a metal-rich GC population in the inner regions of a galaxy can indicate a bulge component of the galaxy as well as provide evidence for whether the bulge is a classical bulge built up through merger events or a pseudobulge driven by secular processes. Pseudobulges will not contain as many bulge GCs as well as have less variations/gradient in the bulge GC metallicities (Minniti et al., 1995). It is important to note that hierarchical and secular evolution can take place simultaneously as well. For example, there is some debate about whether the MW bulge should be defined as classical or pseudo from dynamical studies as there is evidence of both hierarchical and secular evolution (Di Matteo et al., 2015; Shen et al., 2010). Similarly, the kinematics and ages of halo populations of star clusters can reveal the extent of the influence of mergers/accretion on a galaxy s history, such 4

18 as the possible merger of M31 with a massive satellite 6 Gyr ago based on the presence of intermediate age, metal-rich GCs (Brown et al., 2003). In this dissertation, we examine the star clusters of the nearby spiral galaxy M101 in order to better understand the history of this galaxy, compare it to other spiral galaxies, and put it into the context of our picture of galaxy formation and evolution. The specific morphological type of M101 is SABcd (de Vaucouleurs et al., 1991), which means it is between having and not having a bar and has little to no bulge. Disk galaxies which are either bulgeless or have pseudobulges are particularly interesting tests of our current picture of galaxy formation and evolution because they are not well studied despite comprising most of the population of galaxies in the local universe (within 11 Mpc) by number (Fisher & Drory, 2011). M101 is also nearby at 6.4 Mpc (Shappee & Stanek, 2011) and nearly face-on (inclination angle i 18 ) (Bosma et al., 1981) which makes it possible to distinguish star clusters from point sources with the Hubble Space T elescope (HST ) and to examine clusters in both the disk and halo of the galaxy, both crucial elements of this work. 1.3 Models for Studying Extragalactic Star Clusters The rest of this dissertation is organized as follows: In Chapter 2, we examine the photometric and spatial properties as well as sizes of old star clusters in M101. In Chapter 3, we examine the ages and metallicities of a subset of both the old and young star cluster populations of M101 from optical spectroscopy obtained with the 8-m Gemini-North telescope. In Chapter 4, we examine the kinematic properties of the same clusters as examined in Ch. 3. Finally, in Chapter 5, we discuss our conclusions and future work. Our approach in determining the basic parameters for star clusters (e.g., ages, 5

19 metallicities, and extinctions) examined throughout this work will be to compare observations with predictions from simple stellar population (SSP) models. An SSP model simulates the photometric and/or spectroscopic properties of an aggregate stellar population from individual stellar properties by assuming a distribution of masses, metallicities, etc. There is a vast literature on the production of these models, which is well beyond the scope of this work. We use our past experience and comparisons made between observations and models available in the literature to select appropriate models. Integrated light models such as SSPs are necessary analogs in studying extragalactic star clusters because many, including those in M101, cannot be fully resolved. Figure 1-1 shows a GC (left) and an open cluster (right) in the MW, with individual stars distinguishable. Figure 1-2 shows examples of two intermediate age clusters in the LMC ( 50 kpc (Stanek et al., 1998), left) and two M101 clusters examined in this work (6.4 Mpc (Shappee & Stanek, 2011), right), which demonstrate that individual stars become blended together at distances beyond the Local Group. Figure 1-1: Left: Example of a MW GC, M13 (image credit and copyright: Martin Pugh, Right: Example of MW open cluster, NGC 4755 (image credit and copyright: Dieter Willasch, 6

20 Figure 1-2: Left: Examples of two intermediate age LMC clusters from Palma et al. (2013) and observed with the V ictor Blanco telescope at the Cerro Tololo Inter-American Observatory. Right: Example of two M101 star clusters with red (ID 13) and blue (ID 47) colors. These clusters are discussed in Ch. 2, 3, and 4. 7

21 Chapter 2 Two Populations of Old Star Clusters in the Spiral Galaxy M101 Based on HST/ACS Observations The material in this chapter is adapted from Simanton et al. (2015) published in the Astrophysical Journal, volume 805 on page Background Ancient ( 10 Gyr) star clusters formed during the early assembly of most galaxies, and therefore give insight into the broad formation history of their hosts. Elliptical and lenticular galaxies host red, metal-rich GCs believed to be associated with their bulges, and blue, metal-poor GCs believed to be associated with their halos. It is not yet clear whether these different populations have different ages in all early-type galaxies, but an age difference has been observed in several (Park et al., 2012). Spiral galaxies also form metal-poor halo and metal-rich bulge/thick disk clusters, although the fraction of red-to-blue GCs is typically lower than found in similar mass earlytype galaxies. Minniti (1995) and Côté (1999) showed that metal-rich MW GCs in 8

22 the inner regions of the Galaxy are associated with the bulge. In addition to ancient GCs, spiral galaxies form younger clusters with a large range of ages in their disks. In the MW, young (< 100 Myr) open clusters are confined to the thin disk and old (> 3 Gyr) open clusters are found in the thick disk (Portegies Zwart et al., 2010). In fact, Kharchenko et al. (2013) find a smoothly increasing dispersion in the distance of MW star clusters from the Galactic plane with increasing cluster age (see their Figure 5). The GC luminosity functions (LFs) of both elliptical and spiral galaxies, including the MW, are observed to have a peaked shape with a turnover caused by the earlier disruption of lower mass clusters due to evaporation of stars by two-body relaxation in a tidal field (Fall & Zhang, 2001). Small variations in the peak luminosity and width of GCLFs are believed to come from possible differences in cluster ages, cluster densities, or the strength of the tidal field (Villegas et al., 2010). Despite the known small variations, the turnover of GCLFs is nearly universal, which has led to its use as a standard candle for distance determinations (Harris, 2001, p. 223). It is therefore important to study cases where the GCLF does not follow the universal pattern. An interesting example is the so-called faint fuzzies that have been found in the lenticular galaxies NGC 1023 and NGC 3384 (Brodie & Larsen, 2002; Larsen & Brodie, 2000), lenticular and elliptical galaxies in the Virgo cluster (Peng et al., 2006b), and M51 and its companion (Scheepmaker et al., 2007; Hwang & Lee, 2006). These clusters are quite extended with effective radii (r eff ) of 7-15 pc and ages upwards of 7-8 Gyr. Despite their intermediate/old ages, their LF is observed to continue to rise beyond the expected turnover luminosity. Intermediate age clusters found in a few early-type galaxies may be responsible for discrepancies between the distances measured from GCLFs and those measured from surface brightness fluctuations (Richtler, 2003, p. 281). Faint fuzzies also have r eff more than twice that of typical GCs in most ellipicals 9

23 and spirals. Brodie & Larsen (2002) proposed that perhaps faint fuzzies only form in the environment found in lenticular galaxies, or that perhaps they were accreted along with a host dwarf galaxy. However, Chies-Santos et al. (2013) used kinematics and spatial comparisons with planetary nebulae and HI to conclude that faint fuzzies may simply be akin to old open disk clusters. They suggest that the reason faint fuzzies are observed in lenticulars but not in spirals is that these faint clusters are quite difficult to pick out against the strong structure and variable luminosity caused by star formation in spiral disks. Thus far, most detailed studies of GCs have focused on early type galaxies (e.g., Peng et al. (2006a); Villegas et al. (2010)), the MW, M31, and a handful of other, mostly bulge dominated systems (e.g., Goudfrooij et al. (2003)). Late-type spirals, such as M101, have not been as thoroughly studied. Previously, Chandar et al. (2004) studied the GC systems in five spiral galaxies using HST Wide-Field Planetary Camera 2 (WFPC2) observations and compared them with the known distribution of GCs in the MW. They found for M101 and NGC 6946 that the LF continues to rise beyond the expected GCLF turnover; however, this conclusion was based on a small number of clusters resulting from the partial coverage of each galaxy. Barmby et al. (2006) studied a larger sample of M101 clusters (1715 clusters with m V < 23), but they focused on analyzing blue clusters ((B V ) 0 < 0.45) rather than the older, redder clusters we seek to study here. Here, we use the same BV I HST Advanced Camera for Surveys (ACS) observations as Barmby et al. (2006) to examine the LF, colors, sizes, and spatial distributions of the red clusters. The rest of the chapter is organized as follows: We describe the observations, selection criteria and methods, and completeness in 2.2. In 2.3, we show the luminosity distribution, color-color plot, color histograms, color magnitude diagram (CMD), sizes, and spatial distribution of our cluster candidates, and we discuss the results in 2.4. Finally in 2.5, we list our conclusions. 10

24 2.2 Observations, Cluster Selection, and Completeness Observations Ten pointings within M101 were taken with the HST /ACS Wide Field Channel (WFC) in November of 2002 (Program ID: 9490, PI: K. Kuntz) using the F 435W (B), F 555W (V ), and F 814W (I) filters 1. Each field covers (see Fig. 2-1). Assuming a distance to M101 of 6.4 ± 0.2 (random) ±0.5 (systematic) Mpc (Shappee & Stanek, 2011), each field covers a kpc 2 region (each 0.05 pixel covers 1.6 pc at this distance). Images were processed through the HLA MultiDrizzle Pipeline Version 1.0 (Koekemoer et al., 2002), which includes bias subtraction, cosmic ray rejection by combining two sub-exposures from CR-Split observing, dark subtraction, flat fielding, and drizzling. We detect sources, which include star clusters, bright individual stars, and background galaxies, in the V band image using the IRAF task DAOFIND. We detect 383,000 total sources and measure their brightnesses within circular apertures varying from 0.5 to 5 pixels with the background estimated within annuli of 7 to 13 pixels using the PHOT task within IRAF. We determine empirical aperture corrections out to 10 pixels from the curves of growth measured for 25 isolated clusters, and also apply an additional mag, mag, and mag correction from 10 pixels to infinity for B, V, and I respectively (Sirianni et al., 2005). These aperture corrections are added to the measured photometry to obtain instrumental magnitudes. The apparent magnitude (m V ) for each source in the VEGAMAG system is 1 Based on observations made with the NASA/ESA Hubble Space Telescope, and obtained from the Hubble Legacy Archive, which is a collaboration between the Space Telescope Science Institute (STScI/NASA), the Space Telescope European Coordinating Facility (ST-ECF/ESA) and the Canadian Astronomy Data Centre (CADC/NRC/CSA). 11

25 found by applying the following zero points: F 435W = , F 555W = , and F 814W = (Bohlin, 2007; Mack et al., 2007)

26 Figure 2-1: Ground-based optical image of M101 showing the location of all 10 HST ACS/WFC fields used in this work. The 5 scale bar is equivalent to 9.3 kpc. 13

27 2.2.2 Cluster Selection In order to separate ancient star cluster candidates from our full catalog, we made a series of automated selective cuts followed by visual inspection of each object. We use the following automated selection criteria: Clusters brighter than m V < where m V was measured within 3 pixel apertures (i.e. no aperture correction yet applied) to ensure a high signal-tonoise ratio (S/N). Concentration index (CI) > 1.15 where CI is the difference between m V measured within 1 pixel and 3 pixel apertures to eliminate point sources; point sources have CI values that peak around 1.00 with a standard deviation of Colors, i.e < B V < 2.0 and 0.75 < V I < 2.0, similar to those of Galactic GCs. Because CI is a crude measure of object size, we also use the BAOlab/ISHAPE software (Larsen, 1999) to measure the full width at half maximum (FWHM) of each object. ISHAPE fits profiles to each candidate source to determine its FWHM (along with other parameters such as ellipticity). The light profile is a convolution of the point spread function (psf) with a user determined function representing the spread in light from a cluster s non-point-like size. We determined the psf of each of our fields by visually selecting isolated stars in each field. We choose a King profile (King, 1962) with a concentration parameter (ratio of the tidal radius to the core radius) of 30 to represent the cluster-like light profile convolved with the psf for the ISHAPE fitting. All objects with a measured FWHM < 0.2 pixels (r eff 0.46 pc) from ISHAPE were removed from the catalog to further eliminate stars. We then plotted FWHM vs CI and determined a three piece-wise linear fit to the relationship between CI and 14

28 FWHM. Upon visual inspection, objects falling outside of a perpendicular distance of 0.1 from the piece-wise fit only consisted of contaminants (usually faint, distorted patches that cannot be classified, noise, crowded point sources, etc.). Therefore, only objects within a perpendicular distance of 0.1 from the CI-FWHM fits were kept. Finally, obvious background galaxies, chance superpositions, and other contaminants were eliminated via visual inspection. Figure 2-2 shows a few typical radial profiles examined during the visual inspection of six selected candidate clusters and five stars (not included in the catalog). Note the clear difference between stars and the relatively well-resolved clusters. The final catalog consists of 326 candidate clusters (see Fig. 2-3 and Table 2.1) with magnitude, color, and size measurements (six of these do not have FWHM measurements due to ISHAPE fit errors). For simplicity, we refer to the cluster candidates as clusters throughout the rest of the chapter. Table 2.1: M101 Old Star Cluster Catalog ID α 2000 (hms) δ 2000 ( ) V B V V I B I r eff (pc) Continued on next page 15

29 Table 2.1 Continued from previous page ID α 2000 (hms) δ 2000 ( ) V B V V I B I r eff (pc) Continued on next page 16

30 Table 2.1 Continued from previous page ID α 2000 (hms) δ 2000 ( ) V B V V I B I r eff (pc) Continued on next page 17

31 Table 2.1 Continued from previous page ID α 2000 (hms) δ 2000 ( ) V B V V I B I r eff (pc) Continued on next page 18

32 Table 2.1 Continued from previous page ID α 2000 (hms) δ 2000 ( ) V B V V I B I r eff (pc) Continued on next page 19

33 Table 2.1 Continued from previous page ID α 2000 (hms) δ 2000 ( ) V B V V I B I r eff (pc) Continued on next page 20

34 Table 2.1 Continued from previous page ID α 2000 (hms) δ 2000 ( ) V B V V I B I r eff (pc) Continued on next page 21

35 Table 2.1 Continued from previous page ID α 2000 (hms) δ 2000 ( ) V B V V I B I r eff (pc) Continued on next page 22

36 Table 2.1 Continued from previous page ID α 2000 (hms) δ 2000 ( ) V B V V I B I r eff (pc) Continued on next page 23

37 Table 2.1 Continued from previous page ID α 2000 (hms) δ 2000 ( ) V B V V I B I r eff (pc) Continued on next page 24

38 Table 2.1 Continued from previous page ID α 2000 (hms) δ 2000 ( ) V B V V I B I r eff (pc) Continued on next page 25

39 Table 2.1 Continued from previous page ID α 2000 (hms) δ 2000 ( ) V B V V I B I r eff (pc) Continued on next page 26

40 Table 2.1 Continued from previous page ID α 2000 (hms) δ 2000 ( ) V B V V I B I r eff (pc) Continued on next page 27

41 Table 2.1 Continued from previous page ID α 2000 (hms) δ 2000 ( ) V B V V I B I r eff (pc) Continued on next page 28

42 Table 2.1 Continued from previous page ID α 2000 (hms) δ 2000 ( ) V B V V I B I r eff (pc) Completeness To evaluate the completeness of our catalog of clusters, we generate artificial clusters, add them to the M101 images, and re-run the detection and cluster selection methods described in and We generate 4000 artificial clusters using the BAOLAB task MKCMPPSF, which convolves a psf with a user defined function, in this case KING30 profiles with two different input FWHM: 1.0 and 2.0 pixels (the motivation for these sizes is given in 2.3.3). We then use MKSYNTH to randomly place the artificial clusters in one of the images, where the magnitude range matches that of the real clusters. We detect sources with DAOFIND using the same parameters as for the real clusters, measure photometry with PHOT, size measurements with ISHAPE, and run the automated cluster selection criteria and CI-FWHM relation cut (see 2.2.2). Figure 2-4 shows the completeness fraction as a function of apparent magnitude. Both sizes of artificial clusters show a decline in completeness as brightness decreases with the more diffuse clusters declining at slightly brighter magnitudes. Approximately 80% of the artificial clusters brighter than m V = 23.0 make it through our selection pipeline, and approximately 50% of those brighter than m V Although the clusters are randomly placed in the M101 images, we ensured that 2000 are located within galactocentric distance r gc 2 kpc and 500 within r gc 500 pc to ensure that the completeness of the inner regions of M101 are thoroughly 29

43 Figure 2-2: Radial profile showing the enclosed percentage of flux within a given radius (in pixels) for three faint clusters (dashed lines), three bright clusters (solid lines), and five stars (dotted lines). The point sources have significantly steeper profiles. 30

44 Figure 2-3: BV I color images of typical clusters from faint (top panel) and bright (bottom panel) cluster groups. Each panel is 3.7, or 110 pc on a side. 31

45 tested. Figure 2-5 shows the completeness fraction as a function of r gc. While the completeness is slightly better for more compact clusters, there is no significant change in the completeness fraction with r gc except within the innermost 300 pc (dotted line in Fig. 2-5). Figure 2-4: Fraction of selected artificial clusters versus m v for FWHM= 1.0 (solid line) and 2.0 (dashed line) pixels. 32

46 Figure 2-5: Fraction of selected artificial clusters versus r gc for FWHM= 1.0 (solid line) and 2.0 (dashed line) pixels. The dotted line represents the innermost region excluded from the fits in Figure 2-15 (see 2.3.4), which is the only significant drop in completeness as a function of distance. 33

47 2.3 Results and Analysis Cluster Luminosity Distribution Figure 2-6 shows the magnitude and luminosity distributions for the M101 cluster sample (without corrections for completeness) alongside those for MW GCs (taken from the Harris (1996) catalog). The top panel shows a histogram with equal size magnitude bins, while the lower panel shows the luminosity distribution with an equal number of clusters (seven) in each bin. In the magnitude distribution, the M101 clusters have a similar shape to those in the MW (see 2.4 for further discussion) at the bright end, but a drastically different one at the faint end. In the rest of this section, we compare the results of fitting different portions of the cluster luminosity distributions. These fits are performed with a simple least squares regression in logn-m V space. The slope, a, from these fits can be converted into the true power law index α for the LF dn(l V )/dl V L α V by α = (2.5a + 1). We first compare the results of fitting the bright end of the M101 and MW GC distributions (see the bottom panel of Fig. 2-6). The fit range is M V 7.5, approximately 0.2 mag brighter than the peak in the MW GCLF (Ashman et al., 1995). The best fits shown in Figure 2-6 are α M101,bright = 1.97 ± 0.14 and α MW,bright = 1.91 ± The standard deviation for 16 different binnings and ranges are 0.16 for both the M101 and MW data, comparable to the uncertainties. For the faint clusters, we use a magnitude range of 5.4 M V 6.5. Here, completeness is a factor, and fainter than M V 5.4, the luminosity function declines steeply, indicating the probable completeness limit, which matches well with the magnitude at which the completeness fraction of the artificial clusters drops below 50% (see Fig. 2-4). We fit the MW GCs in the full faint range M V > 6.54 (m V > 22.5) since the completeness limit is not a concern for the Harris (1996) catalog. The 34

48 slopes for these fits are α M101,faint = 2.03 ± 0.05 and α MW,faint = 0.46 ± The standard deviation for 16 different binnings and ranges (including more conservative completeness limits brighter than M V 5.4) are 0.25 and 0.22 respectively. The shapes of the faint ends of the M101 and MW GC distributions are very different. Note that any faint clusters not identified in our M101 sample due to incompleteness will only steepen α M101,faint and increase the discrepancy between the M101 and MW GC distributions. Statistical tests confirm the fit results for the luminosity distribution of M101, that the bright end is similar to that in the MW, but the faint end is quite different. We compare the shapes of the M101 and MW luminosity distributions using the twosided Kolmogorov-Smirnov (KS) test and the two-sided Cramer-von Mises (CvM) test. Both tests accept the null hypothesis (p-values > 0.1) that the M101 luminosity distribution and the MW luminosity distribution are drawn from the same distribution at the bright end M V < 6.54, and very strongly reject the null hypothesis (p-values < 0.01) at the faint end M V > In fact, the fits do not produce p-values < 0.05 (strong rejection) until the samples include clusters fainter than M V > Therefore, we divide our sample into two groups throughout the rest of the chapter, where bright clusters have M V < 6.54 (m V < 22.5) and faint clusters have M V > 6.54 (m V > 22.5). There are 90 clusters in the bright group and 236 clusters in the faint group. We perform a final fit to the faint clusters after subtracting the MW GC histogram from the M101 cluster histogram (see the dashed line in the top panel of Figure 2-6). No normalization of the MW GC histogram is performed prior to subtraction because the bright ends of the distributions are well matched (see Fig. 2-6). We find α = 2.60 ± 0.26 (uncertainties from Poisson errors, i.e. 1/ N) in the range 3 Note that the value for α for the faint MW GCs is negative, while a, which is shown in the lower panel of Fig. 2-6, is positive. 35

49 m V = to m V = Nine different combinations of bin sizes and data ranges (always brighter than the completeness limit, m V 23.65) give a median α = 2.57 and a standard deviation of

50 Figure 2-6: Top: Magnitude distribution (top) for our M101 cluster sample (solid line), MW GCs (dotted line), and the difference between the M101 clusters and the MW GCs (dashed line). A power law fit (dashed line) to the dashed line histogram for the bins brighter than the completeness limit gives α = 2.60 ± The solid vertical line represents m V = 22.5 (M V = 6.54), which is where we choose to divide our sample into faint and bright clusters (see text). The dotted vertical line represents the peak of the MW GC distribution. Bottom: Luminosity distribution for our bright clusters (crosses), faint clusters (solid circles), and MW GCs (asterisks) with constant number binning. Fits to subranges of each group are shown with the same line styles as the top panel. 37

51 2.3.2 Cluster Colors and Luminosities In this section, we compare the colors and luminosities of old star clusters in M101 with those of cluster populations in other galaxies. The ages and metallicities of clusters older than 1 Gyr become degenerate, making it difficult to establish differences in age based only on integrated broad-band colors; the colors are a stronger indicator of metallicity than of age in this regime. For bright clusters, the median B V, V I, and B I colors are 0.72, 1.10, and 1.81, respectively, and for the faint clusters they are 0.65, 1.10, and Figure 2-7 shows B V versus V I colors with two metallicity SSP tracks (solar and Z = 0.008) from Bruzual & Charlot (2003) updated 2006 data (from private communication). The cluster colors are approximately centered on the SSP tracks with some spread in the colors. This spread is almost certainly due to photometric errors, based on a comparison between input and measured magnitudes for artificial clusters. Furthermore, the spread in cluster colors around the SSP tracks increases for fainter clusters, as expected when photometric errors are the dominant source of uncertainty. Therefore, the small differences between the median values for the faint vs. bright cluster colors in B V and B I are unlikely to be strongly significant. The B I color has the largest wavelength baseline and is therefore best suited to revealing multiple peaks in the colors, if they exist. Figure 2-8 shows B I histograms. The top plot shows the bright and faint M101 cluster groups, and the bottom plot divides the clusters by distance, either inside or outside of 3 kpc from the center of M101 (Evans et al., 2010). The faint clusters drive this peaked distribution; the bright clusters appear to have a more even color distribution with no obvious peak. Overlaid (dotted lines) are the mean blue and red peak B I colors of eight brightest cluster galaxy 4 (BCG) GC populations determined by Harris et al. (2006). 4 BCGs are the brightest galaxies in a cluster of galaxies. They are typically elliptical galaxies as galaxy clusters are dominated by ellipticals which form through many merger events in such a 38

52 Note that we have not converted from the VEGAMAG system (using HST filters) to the Johnson-Cousins system. We estimate from Figure 21 in Sirianni et al. (2005) that with a conversion to the Johnson-Cousins system, the M101 B I colors would shift blueward by at most It is then clear that the total cluster histogram strongly peaks close to the blue BCG peak, with a weak tail of clusters extending to the red BCG peak. This red tail may be caused by reddening within the host galaxy itself for which we have not accounted. In the bottom panel of Figure 2-8, the clusters within 3 kpc of the center of M101 are slightly redder than those outside of 3 kpc, although they do not have a peak centered on the red peak of the BCGs. The redder color could result from higher extinction or possibly higher metallicity of these more centrally located clusters, but we cannot distinguish between these two possibilities with the currently available data. Figure 2-9 shows the CMD for our cluster sample with MW GCs, Large Magellanic Cloud (LMC) intermediate age clusters (1-3 Gyr, magnitudes and colors from Bica et al. (1996) and ages from Palma et al. (2013), Goudfrooij et al. (2011), Piatti et al. (2009), Mucciarelli et al. (2007), and Kerber et al. (2007)), and old open MW clusters (1-8 Gyr, magnitudes, colors, and ages from Lata et al. (2002)) included for comparison. Again, note that a conversion to the Johnson-Cousins filters has not been applied to the M101 clusters; we estimate B V colors would shift redward by at most (Sirianni et al., 2005). The M101 cluster sample, as a whole, has colors similar to the MW GCs, but with significantly more clusters fainter than M V = It is of note that the MW GC sample plotted here is missing B V for 40 clusters, 34 of which are in the faint cluster region. Even taking this into account, however, there are still many more M101 faint clusters (236) than MW GCs (53) in this region. The LMC intermediate age clusters also lie largely in the region of the CMD galaxy dense environment. 39

53 dimmer than M V = 6.54, but within the B V color range of the M101 clusters. Old open clusters in the MW occupy a similar color space as the LMC clusters, but with even fainter magnitudes. The median colors for the faint M101 clusters, intermediate age LMC clusters, the old open clusters, and MW GCs fainter than M V = 6.54 are B V = 0.65, 0.67, 0.71, and 0.74 respectively, with the M101 clusters shifting up to with conversion to the Johnson-Cousins system. All of these groups may have consistent B V colors as the MW GC median value is affected by the 34 missing values. Also shown in Figure 2-7 are the colors for 25 YMCs ( 100s Myr) that are discussed further in 2.4.1; they are part of the spectroscopic sample of clusters in M101 analyzed further in chapters 3 and 4. Their categorization as YMCs is derived from the strength of the Balmer lines seen in Gemini/GMOS spectra. The YMC colors shown in Figure 2-7, however, are measured from the HST BV I images studied here with the same treatment as the rest of the cluster catalog. It is clear that the YMCs overall are much bluer than the faint and bright populations with a median B V = 0.17 and V I =

54 Figure 2-7: Color index plot showing B V vs. V I colors for our full M101 cluster sample faint (solid circles) and bright (crosses) groups and the M101 YMCs with spectra (pluses). The lines show the evolution of Bruzual & Charlot (2003) (updated 2006 data) SSP models with solar (solid line) and Z = (dashed line) metallicities. The 1 Gyr age on each SSP track is indicated by an open diamond. The M101 clusters follow both SSP tracks well with some spread most likely due to photometric uncertainties. The arrow represents typical Galactic reddening for A V = 0.5. Galactic reddening in the direction of M101 is very low (E(B V ) < 0.01, Chandar et al. (2004)), and we do not correct for it. 41

55 Figure 2-8: Top: B I color histogram with M101 clusters divided into bright (dashed-dotted line) and faint (dashed line) groups. Vertical dotted lines are the mean blue and red peak B I colors for eight BCG GC populations (Harris et al., 2006). Bottom: Same as above, except the clusters are now divided by distance with the dashed line representing clusters outside of 3 kpc from the center of the galaxy and the dashed-dotted line representing inside of 3 kpc. 42

56 Figure 2-9: B V color magnitude diagram showing our M101 sample (solid circles), MW GCs (asterisks), LMC intermediate age clusters (1-3 Gyr, open squares), and old open clusters (1-8 Gyr, open triangles). The dotted line corresponds to the luminosity division at M V = 6.54 imposed on our M101 sample (see 2.3.1). 43

57 2.3.3 Sizes Figure 2-10 shows the histogram of cluster sizes (r eff ) for the bright and faint cluster subsamples. The FWHM of each cluster was measured using the ISHAPE software, as described in 2.2, then converted to r eff assuming a distance to M101 of 6.4 Mpc (Shappee & Stanek, 2011). The figure shows that the median r eff values for the bright (2.41 pc) versus faint (4.27 pc) clusters are different, with fainter clusters tending to have larger sizes. MW GCs have a median r eff of 2.98 pc, similar to that found for the bright clusters in M101. Figure 2-11 plots cluster luminosity versus r eff for the bright and faint M101 samples, and for MW GCs. The MW GCs cover a similar parameter space as those in M101, with the bright clusters in M101 being more compact with less scatter to larger radii than the faint clusters. In order to investigate whether or not there are systematic biases in the sizes measured by ISHAPE for bright versus faint clusters, we compare ISHAPE measurements of artificial clusters of varying input sizes added to the M101 images (see for details). We use 2000 of the artificial clusters from the completeness testing with FWHM = 1.0 and 2.0 pixels in addition to 2000 artificial clusters with FWHM = 0.5 and 4.0 pixels. Figure 2-12 plots the cluster luminosity versus measured r eff for the M101 clusters and the FWHM= 1.0 and 2.0 pixel (r eff = 2.3 pc and 4.6 pc) artificial clusters. While it is apparent from the artificial clusters that there is a spread in the measured r eff which increases at fainter magnitudes, it is also apparent that this spread cannot fully describe the number of real, faint clusters with large sizes. We compare the results from artificial clusters with the observations using the Hodges-Lehmann (HL) estimator, which determines the shifts between the location parameters of two data sets, in order to quantitatively establish whether the observed 44

58 difference in cluster sizes between the bright and faint subsamples is real or the result of systematic and random errors. The results from the statistical tests and artificial cluster experiments described below suggest that there is a physical difference in the sizes of bright and faint clusters in M101. The HL shift estimates the difference between the median values of two data sets if the data sets are symmetric about the median, or the difference between the pseudo-medians if the data sets are not symmetric about the median. It does this by computing the median of the differences between each pair of the values in the two data sets. It also gives confidence intervals and p-values for the significance of the separation as part of the Wilcoxon rank sum test. The HL estimator indicates a shift between the median sizes of bright vs. faint clusters in M101 of (errors are 95% confidence intervals). For the same magnitude ranges, there is a smaller shift of ( ) for bright vs. faint artificial clusters with input FWHM values of 1.0 pixels or r eff = 2.3 pc (FWHM= 2.0, r eff = 4.6 pc). We also tested the HL estimators for bright versus faint clusters for two faint binnings that excluded the faintest clusters and therefore highest uncertainty FWHM measurements (faint bins from 5.6 > M V > 6.54 and 6.1 > M V > 6.54; bright bins still M V 6.54). This only increased the difference between the real and artificial cluster shift estimators. P-values for the real and artificial clusters for all binnings show that the shifts are significant with values ranging from to 0.03 (strongly accepting the alternative hypothesis that the true location shift is not equal to 0). We conclude that there is a physical difference between the sizes of bright and faint clusters in M101, with the latter tending to be larger (by pc). 45

59 Figure 2-10: r eff or half-light radius histogram for our M101 sample divided into bright (solid line) and faint (dashed line) clusters with the MW GCs (dotted line) also plotted. The median values of the distributions are shown by the vertical lines at the top of the plot with their respective line styles for each group. The median r eff of the bright M101 clusters (2.41 pc) is similar to that of the MW GCs (2.98 pc) while the faint M101 clusters have a much larger median size (4.27 pc). 46

60 Figure 2-11: Absolute magnitude verses r eff for our M101 catalog (solid circles) and MW GCs (asterisks). The horizontal dotted line represents the luminosity division at M V = 6.54 imposed on our sample (see 2.3.1). The density of the faint M101 clusters is much greater than that of the faint MW clusters. (Although the MW clusters are a more complete sample down to fainter magnitudes.) The brighter clusters have less spread in sizes and skew towards being more compact for both the MW and M

61 Figure 2-12: The same as Fig for the M101 clusters (large solid circles) and artificially generated clusters (small solid circles). The vertical solid lines are the input sizes of the artificial clusters (top FWHM= 1.0 and bottom FWHM= 2.0). In the top plot, the real data are reasonably well matched by artificial clusters at bright magnitudes, but at faint magnitudes, it becomes apparent that too many real clusters have large sizes that cannot be explained by the spread due to measurement errors alone. In the bottom plot, we see that the faint clusters are better described by larger input size artificial clusters. 48

62 2.3.4 Spatial Distribution The locations of our clusters within M101 in RA and Dec are shown in Figure Figure 2-14 shows the r gc distribution for the faint versus bright clusters. The bright clusters are largely found within 9 kpc of the center, and are quite centrally concentrated, while the faint clusters are more evenly distributed. Figure 2-15 shows the surface density plots in logn-r 1/4 gc and lnn-r gc space. N for the surface density plots is derived by counting the number of clusters within the annulus inner and outer radii covered by the r gc bin. N is then divided by the area of the annulus to determine the cluster surface density. Here, r gc and N do not include the innermost 10 ( 310 pc at the distance of M101) of the galaxy since this region is too bright to detect clusters down to the same magnitude limit as the rest of the galaxy. Beyond this region, the background level drops sharply and incompleteness does not vary strongly with r gc. This can be seen in Figure 2-5 which shows the completeness fraction as a function of r gc. We fit a de Vaucouleurs law of the form logn = b Vauc + c Vauc r 1/4 gc and an exponential of the form lnn = b exp + c exp r gc to each of the three data sets which are reported in Table 2.2 (errors are Poisson errors based on the count in N). The de Vaucouleurs surface brightness profile is typically used to model spheroidal components and elliptical galaxies (de Vaucouleurs, 1948), while exponential profiles provide a good description of disk components of galaxies (Patterson, 1940). We also convert the slopes of the de Vaucouleurs fits into effective radii (R e ), the radius within which half of the total cluster population lies in projection, with the formula R e = ( 3.33/c Vauc ) 4 (de Vaucouleurs & Buta, 1978). The slopes of the exponential fits are converted into scale radii (R 0 ) by R 0 = 1/c exp. Both the R e and R 0 are almost identical for the bright M101 clusters and MW GCs whereas the values for the faint M101 clusters differ significantly from those 49

63 of the MW GCs. Furthermore, the faint M101 clusters do not appear to follow a straight line in logn-r 1/4 gc space (see the left panels of Fig. 2-15), and so may not be well represented by the de Vaucouleurs law. However, they follow a straighter line path on the lnn-r gc space surface density plot, indicating an exponential law better describes the distribution of the faint clusters. The bright clusters appear to follow a slightly straighter line in the logn-r 1/4 gc space than lnn-r gc space, as expected from the high degree of steepness in linear space, and hence, the de Vaucouleurs law gives a better fit. Figure 2-13: Positions in RA and Dec for our cluster sample. The bright clusters (crosses) are more centrally concentrated than the faint clusters (solid circles). 50

64 Figure 2-14: Histogram of r gc for each cluster in our sample. The bright clusters (solid line) are concentrated towards the center while the faint clusters (dashed line) have a broader distribution. 51

65 Figure 2-15: Surface density plots showing the area normalized number of clusters at different binned radii from the center of the host galaxy in log N-rgc 1/4 (left) and lnn-r gc space (right). The symbols show the mid point of the radii bin they represent with either connecting lines to guide the eye (top) or best fit lines over laid (bottom). In all plots, the bright M101 clusters (crosses with solid line) follow a centrally concentrated distribution, similar to the MW GCs (asterisks with dotted line), while the faint M101 clusters (solid circles with dashed line) are much less centrally concentrated. Slopes of the best fits to the data are shown in Table

66 Table 2.2: Coefficients for Surface Density Fits 53 Data Set b Vauc c Vauc R e (kpc) a b exp c exp R 0 (kpc) b Bright M ± ± ± ± ± ±0.13 Faint M ± ± ± ± ± ±0.32 MW GCs 3.07 ± ± ± ± ± ±0.14 a Effective radii calculated with the formula R e = ( 3.33/c V auc ) 4 (de Vaucouleurs & Buta, 1978). b Scale radii calculated with the formula R 0 = 1/c exp.

67 2.4 Discussion In this section, we first examine the properties (luminosity, color, size, and spatial distributions) of the bright versus faint clusters to determine if our sample consists of two distinct populations of old star clusters. Then we more closely examine the properties of the faint clusters and compare to those in other galaxies Are There Two Populations of Old Clusters in M101? The shape of the LF gives the first indication that there may be two populations of red clusters in M101. As seen in Figure 2-6, the LF rises nearly continuously from the bright end down to the completeness limit, but shows a dip near M V 7.5 to This dip can be explained if the LF is a combination of a peaked distribution (as seen for old GCs) at the bright end and a rising, power law-like distribution (as seen for younger clusters) at the faint end. For old GC systems in the MW and other galaxies, both luminosity and mass distributions have similar, peaked shapes, believed to result from mass-loss over 12 Gyr, mostly driven by the effects of two-body relaxation. When we compare the MW GCLF to that of our M101 cluster sample (see Fig. 2-6 and 2.3.1), we find the shapes quantitatively match well at the bright end, at least down to M V = 7.5, 0.2 mag brighter than the peak in the MW GCLF (Ashman et al., 1995). The CMD (see Fig. 2-9) shows that the colors of the majority of the bright clusters are similar to those of MW GCs. The median r eff of the bright clusters is similar to the median size of the MW GCs (but smaller than that of the faint clusters; see 2.3.3). The bright M101 clusters are centrally concentrated towards the center of the galaxy, as are the MW GCs (see 2.3.4), which is expected for a spherically distributed population. The similarity between the luminosity, magnitude, color, size, and spatial distributions of the bright M101 clusters and the MW GCs strongly suggests that the 54

68 former are most likely a typical old ( 12 Gyr) population of GCs in M101. Furthermore, our results suggest the M101 bright clusters are largely metal-poor, consistent with the galaxy s nearly bulge-less morphology. Figure 2-8 shows that the bright cluster colors skew towards the blue peak of BCGs (Harris et al., 2006). Corrections to the photometry that put our clusters on the same photometric system as the BCG data would push the colors further blueward as would any unaccounted for reddening corrections. Thus, the few clusters near the red peak of the BCGs most likely do not account for a separate metal-rich population of clusters as seen in more bulge-dominated/spheriodal galaxies. Turning to the faint clusters, their LF can be described by a rising power law with an average slope of α = 2.6 ± 0.3 (see and Fig. 2-6). This is similar to what is found for young cluster populations in other galaxies. Larsen (2002) found that young star cluster LFs in 6 spirals follow a power law of dn(l V )/dl V L α V where α ranges from 2.0 to 2.6. Whitmore et al. (2014) examined the LFs of clusters in 20 nearby (4 30 Mpc) star-forming galaxies, and found an average α = 2.37 with rms scatter of In checking for similarities between the faint clusters and the MW GCs, we find the median r eff of the MW GCs is smaller than that of the faint clusters, and their spatial distributions are very different. While the MW GCs and bright red M101 clusters are concentrated toward the center of their host galaxies, the faint clusters in M101 are more uniformly distributed as expected if they reside in the disk. We conclude that the faint, red clusters form a population distinct from the typical old GCs, one which is associated with the disk of M What Are the Faint, Red Clusters in M101? Discovering the true nature of the faint clusters depends on determining their ages, metallicities, and masses. It is clear from Figure 2-7 that exact ages and metallicities 55

69 cannot be determined accurately for each cluster from optical colors alone since the SSP tracks lie on top of each other in this color space. Nonetheless, we can compare the colors, luminosities, and sizes of these clusters with those in other galaxies where the cluster properties are better known. The faint clusters may be a population similar to the LMC intermediate age clusters or the brightest old open clusters since they share a similar color-magnitude space (see Fig. 2-9), although the faint M101 clusters do extend to brighter magnitudes than the old open clusters. The faint fuzzies discovered by Brodie & Larsen (2002) and Larsen & Brodie (2000) in lenticular galaxies have V I which is similar to that of the faint clusters, 1.1. Despite the similar colors and magnitudes between faint fuzzies and our faint clusters, only 48 of the 230 faint clusters with FWHM measurements have r eff > 7 pc, but the faint clusters do still have a large median size, greater than the bright clusters or MW GCs. We can use the observed power law shape of the LF to constrain the ages of the faint clusters. As mentioned previously, the peaked shape found for old GCs is driven by stellar mass loss due to internal relaxation. Figure 3 in Fall & Zhang (2001) shows how a peak develops at the low end of the mass function and then moves to higher masses as the population ages, while the high mass end continues to have a power law shape. Since we do not observe a peak in the LF of the faint, red clusters, it must occur below our completeness limit. When luminosities are converted to masses, the age at which the peak of the mass function is just below the completeness limit is the maximum age of the faint clusters. We make the following assumptions for our calculations: a stellar-mass loss rate for each cluster of µ ev = 10 5 M yr 1, a typical value for MW GCs (e.g., Fall & Zhang (2001)), and a single age τ for the cluster population. This gives a predicted peak mass, M p = µ ev τ, at different ages. We then use the SSP model predicted, agedependent M/L V to convert our observed cluster luminosities to masses at different 56

70 assumed ages, and determine the mass equivalent of our completeness limit, M lim. Based on this methodology, the maximum age for our clusters is constrained by the value where M p = M lim. We find the maximum age to be 700 Myr ( 1 Gyr) for Z = (Z = 0.02). Since the faint clusters are found in the disk, we might expect the maximum age based on the more metal-rich model to give a better estimate, but MW old open clusters span metallicities from just above solar down to [Fe/H] 1 (Friel, 1995). Likewise, faint fuzzies have typical [Fe/H] 0.6 (Brodie & Larsen, 2002), and LMC intermediate age clusters have [Fe/H] 0.4 to 0.7 (Palma et al., 2013). The faint M101 clusters do have a lower density than typical MW GCs, and the mass-loss rate due to relaxation-driven evaporation depends on the internal density of the clusters, µ ev ρ 1/2 h, where ρ h is the half-light density (Chandar et al., 2007; McLaughlin & Fall, 2008). Therefore, a better estimate of the maximum age of the faint M101 clusters might be determined by scaling the mass loss rate to the median size of the faint clusters (µ ev,m101 = µ ev,mw (ρ h,m101 /ρ h,mw ) 1/2 ). Now, we find the maximum age to be 9 10 Gyr for Z = and Gyr for Z = Thus it is possible that the faint clusters might be quite old. We can also put constraints on the minimum age of the faint clusters. Barmby et al. (2006) suggested that the faint, red cluster population in M101 may be reddened, young disk clusters; however, they do not provide any estimated ages. We examine B I color images which highlight the locations of dust lanes in M101 and find no preference for the faint clusters to be embedded in dust. Likewise, we examine archival HST/W F P C2 Hα (F656N) images and find no preference for the faint clusters to be in or near HII regions. This strongly indicates that they are older than 10 Myr, as clusters younger than this are expected to have not yet fully dispersed, leaving the gas and dust clouds from which they formed. Also, we have Gemini/GMOS spectra for 25 YMC clusters which exhibit strong 57

71 Balmer lines with no Hα emission lines, indicating ages of a few hundred Myr. We find that their median colors from our HST data are B V 0.2 and V I 0.5 (see Fig. 2-7), significantly bluer than the faint red clusters, which have B V 0.7 and V I 1.1. Barmby et al. (2006) selected 1260 blue clusters in M101 with (B V ) 0 < 0.45 and V < 23, similar to our spectroscopically confirmed YMCs, and they found these clusters appear to coincide with the spiral structure of the M101 disk. The faint red clusters likely do not follow the spiral structure (despite being in the disk), because they have had sufficient time to disperse throughout the disk and away from their birth sites. We therefore believe that the faint red clusters are older than a few hundred Myr. The faint red clusters are associated with the disk of M101, just as old open clusters are in the MW disk (Portegies Zwart et al., 2010). Note that the scale radius from the exponential fit to the faint M101 clusters ( 4 kpc) is similar to the scale radius of the disk of M101 according to optical photometry (R 0 = kpc, Okamura et al. (1976)). It is possible that the population of old open clusters observed in the MW disk suffers a selection bias. Limiting the sample size of a cluster population limits the maximum brightness of the clusters in the sample (Whitmore et al., 2007), and MW disk clusters are a more limited sample as they are observed edge-on through the obscuring dust of the disk, rather than face-on, like M101. Thus, it is possible that we are missing old open MW clusters with brighter magnitudes and higher masses similar to our M101 old disk clusters. However, Chandar et al. (2004) do not find an excess of faint, red clusters in the spiral galaxies M81, M83, or M51 which brings into question how common old disk clusters in spiral galaxies are and what factors into their formation and/or survival. Interestingly, faint fuzzies have also been found to be associated with the disks of their lenticular host galaxies (Brodie & Larsen, 2002; Chies-Santos et al., 2013; Forbes et al., 2014) despite their ages 7 8 Gyr. Chies-Santos et al. (2013) find that the 58

72 strong association of the NGC 1023 faint fuzzies with their galaxy s disk rather than its bulge to be evidence that the faint fuzzies are simply very old open clusters. They further predict that such clusters should be found in spiral galaxy disks as well, and that the only reason they have thus far only been identified in lenticular galaxies is because of their smooth disks, which make them easier to observe. The excellent resolution of HST /ACS allows us to identify such clusters in M101 for the first time. Scheepmaker et al. (2007) also found six clusters in the spiral galaxy M51 that match the definition of faint fuzzies. We conclude that the faint clusters studied here are old (τ 700 Myr) and part of the disk of M101, similar to but more massive than old open clusters observed in the Milky Way. 2.5 Conclusions M101 appears to have two populations of old star clusters: a typical population of old GCs and a fainter population of intermediate-age to old disk clusters. For the population of old GCs, we find: Their luminosity distribution, colors, sizes, and spatial distribution are similar to those of the MW GCs. Their spatial distribution shows a central concentration which is consistent with a spherically distributed halo population. Their colors are skewed toward the typical blue peak of BCG GC populations which indicates they are most likely a metal-poor dominated population, fitting with the nearly bulge-less morphology of M101. For the fainter population, we find: Their luminosity distribution is similar to the power law shape of young cluster populations in other spiral galaxies, dn(l V )/dl V L α V where typically α 2 59

73 to 2.6. We find the average power law fit to our faint clusters in M101 has a slope of α = 2.6 ± 0.3. Age constraints determined from the shape of the LF indicate that they could be quite old, up to Gyr, although this constraint depends on a number of assumptions. They are older than a few hundred Myr because their colors are much redder than a sample of M101 YMCs with spectroscopic ages of a few hundred Myr. These clusters also do not follow the spiral arm structure that the blue M101 clusters identified by Barmby et al. (2006) do. They have a fairly extended spatial distribution, quite different from the centrally concentrated, bright GCs. They are most likely associated with the disk. They occupy the same luminosity-color space as LMC intermediate age clusters, the brightest old open MW clusters, and faint fuzzies. Old open clusters and faint fuzzies are also located in the disks of their galaxies. Chies-Santos et al. (2013) concluded that faint fuzzies are analogous to old open clusters and should be found in spiral galaxies. We conclude the M101 faint clusters are most likely these old disk clusters. The evidence for a large population of old disk clusters in M101 shows that the peak of the GCLF may not be an accurate distance indicator for all galaxy types, especially not spiral and lenticular galaxies. We would caution against using it as a stand alone measure of distance. Other spiral galaxies need to be examined for faint, disk cluster populations (see Ch.5). In the next two chapters, we discuss the ages, metallicities and kinematics derived from Gemini GMOS spectra of bright, old GCs in M101. Unfortunately, we do not have spectra for any clusters in the faint population to examine in this work. 60

74 Discussion of the feasibility of obtaining spectra of such faint objects can be found in Ch

75 Chapter 3 Gemini/GMOS Spectra of Old and Young Star Clusters in M101: Ages and Metallicities 3.1 Background The integrated colors of star clusters are degenerate with respect to age and metallicity. Spectra however, contain absorption lines of specific elements whose strength is dominantly dependent upon either age or metallicity, but not both, allowing the age-metallicity degeneracy to be broken. Here, we use spectra obtained with the Gemini/GMOS instrument of bright star clusters in the spiral galaxy M101 to determine their ages and metallicities. The ages and metallicities of this sample of star clusters can shed light on the cluster populations of M101 and its history. For example in the MW, the metallicities of GCs along with their spatial distribution are used to determine their proposed formation mechanisms and whether they are associated with the halo or the bulge of the Galaxy with the high metallicity peak ([Fe/H] 1.5) being associated with the bulge and the low metallicity peak associated with the halo ([Fe/H] 0.5). 62

76 More specifically, the range of metallicities observed for bulge clusters in the MW ( 1.0 [Fe/H] +0.5, Minniti et al. (1995)) indicates that its bulge has been built at least partially by mergers (Minniti, 1995). The metal-poor outer halo GCs in the MW have a spread in ages ( Gyr, Leaman et al. (2013)) without a metallicity gradient which also support accretion scenarios (Searle & Zinn, 1978). Intermediate age clusters ( 6 8 Gyr) in the halos of M31 and M33 (a bulgeless spiral similar to M101) also demonstrate mergers/accretion in their histories (Brown, 2009). This rest of the chapter is arranged as follows: We describe the cluster candidate selection criteria, observations, and spectra reduction in 3.2. In 3.3, we measure the Lick indices and determine ages and metallicities of the clusters, and we discuss the results in 3.4. Finally in 3.5, we list our conclusions. 3.2 Observations Cluster Candidate Selection with HST Candidate star clusters, regardless of age, were first identified from HST /ACS/WFC BV I images, similar to the method described in 2.2 but without a cut in color. The result was a catalog containing several thousand young clusters (with ages estimated to be less than 1 Gyr) and the ancient cluster candidates presented in Ch. 2, 86 of which are brighter than m V = 21.5 mag. From this catalog, both young and ancient cluster candidates were selected for spectroscopy. The selection criteria for GCs, which were our highest priority targets, were as follows: Brightness cut-off of m V < 21.5 to ensure high S/N. Objects broader than the point spread function in order to distinguish them from point sources. 63

77 Colors similar to those of Galactic GCs, i.e < B V < 2.0 and 0.75 < V I < 2.5 to help exclude background galaxies. Cluster candidates are then selected based on positions to fill masks with multiple slits for obtaining spectra of many objects at once. Candidate positions were selected from pre-imaging taken by Gemini in the 2007A semester and designed to sample the target clusters described above. Targets had to have different y-coordinates along the CCD so that the spectra do not overlap on the detector. Our final slit masks for spectroscopy (see Fig. 3-1 and 3-2) contained 55 candidates, 23 of which have red colors typical of Galactic GCs. Color postage stamp images taken with HST in the BV I filters showing all GCs and YMCs for which spectra were obtained are shown in Figures 3-3 and 3-4 and spatial and photometric properties are listed in Table 3.1. Table 3.1: Spatial and Photometric Properties of M101 Clusters with Spectra ID α 2000 ( ) δ 2000 ( ) M V B V V I B I r eff (pc) Mask 1 3 a Continued on next page 64

78 Table 3.1 Continued from previous page ID α 2000 ( ) δ 2000 ( ) M V B V V I B I r eff (pc) a a a a a a Mask Continued on next page 65

79 Table 3.1 Continued from previous page ID α 2000 ( ) δ 2000 ( ) M V B V V I B I r eff (pc) a Also on Mask Gemini/GMOS Observations The spectroscopic observations were taken with the GMOS-North instrument (GMOS-N) over several nights in the 2008A semester with seeing that varied between 0.61 and The slow read mode was utilized with the B600-G5303 grating and 4 2 binning (dispersion spatial). The slit widths were fixed at 1.0, and the slit lengths were variable to accommodate many objects on each mask. Observing conditions allowed for spectroscopic observations for two of the four masks requested in the program to be taken. Figures 3-1 and 3-2 show the pre-imaging for the two observed masks overlaid with the positions of the objects. Labels show the ID numbers used to refer to each object throughout the rest of this chapter and the next. There were 10 exposures of mask 1 and four exposures of mask 2 taken, each with an exposure time of 3600 s. 66

80 Figure 3-1: Gemini-North preimaging of M101 mask 1 with objects labeled by ID number. 67

81 Figure 3-2: Gemini-North preimaging of M101 mask 2 with objects labeled by ID number. 68

82 Figure 3-3: HST combined BV I images of YMCs. Dimensions are approximately The top image shows clusters located on mask 1 of the GMOS data, and the bottom image shows clusters located on mask 2. 69

83 Figure 3-4: HST combined BV I images of GCs. Dimensions are approximately The top image shows clusters located on mask 1 of the GMOS data, and the bottom image shows clusters located on mask 2. 70

84 Reduction of the spectra is done with PyRaf 1 routines supplied by Gemini and Bryan Miller as well as IDL 2 routines written by Bryan Miller. A brief outline of these steps is listed below, but for a more detailed description of reduction steps, see appendix A2 of Trancho et al. (2007). Bias frames are combined for each night of observations using GBIAS and subtracted from the corresponding science frames using GSREDUCE. A custom pipeline called MOSPROC 3 is used to output one-dimensional spectra with bad pixel masks and a signal-to-noise ratio (SNR) plane. The spectra also receive flat-fielding, wavelength calibration, cosmic ray cleaning, quantum efficiency correction, and background subtraction in MOSPROC. It is of note that MOSPROC is run several times in order to improve the wavelength calibration, background subtraction, and to extract one dimensional spectra for each object. We find the best wavelength calibrations are obtained by fitting a sixth order spline3 function along the dispersion axis and a third order spline3 function along the spatial axis. Some light is lost for each object during observing due to a combination of atmospheric dispersion and seeing conditions with the largest losses being for wavelengths less than 4500 Å at high parallactic angles. The position angle of the slit for each object is designed to minimize the parallactic angle; however, for the multiple exposures taken at different times, small differences persist due to changes in the targets position on the sky. We correct for atmospheric dispersion and image quality differences between exposures by the following procedure: SLITCORR 4 is used to calculate the slit angle for each exposure of an object. Then, the exposure with the smallest slit angle is used as a reference to compare to the other exposures of the same object using PCOMPARE 5. Finally, the SLITCORR parameter xoffset, which corresponds to Custom PyRaf script written by Bryan Miller. 5 Custom IDL script written by Bryan Miller. 71

85 the centering of the object in the slit, is altered by no more than ±0.2 until differences between the tilt of the spectra in different exposures is minimized. Changes to each spectrum are multiplicative to correct for light losses, not shifts along the dispersion axis, and should have no impact on either the wavelength regions used in measuring Lick indices (see 3.3.2) or the measurement of velocities (see 4.2.1). The resulting exposures of the same source on each mask are combined by a weighted average using GSHIFTCOMB. CLEANSPEC 6 is used to mask any remaining bad pixels as well as spurious lines from incomplete background subtraction, typically of sky lines, and to apply a relative flux calibration based on the sensitivity function of the detector.. Figures 3-5 through 3-10 show the final spectra with continuum subtraction also applied, divided by the initial age categorization as described in Table 3.2 lists properties of the spectra such as their full wavelength range, S/N ratio, and equivalent width of the H beta line (W Hβ, used in categorizing spectra into YMCs and GCs in 3.3.1). Table 3.2: Spectroscopic Properties of M101 Cluster Spectra ID λ min -λ max Count a S/N b W Hβ (Å) Type Mask 1 3 c GC GC GC GC GC GC Continued on next page 6 Custom PyRaf script written by Bryan Miller. 72

86 Table 3.2 Continued from previous page ID λ min -λ max Count a S/N b W Hβ (Å) Type GC GC GC GC 15 c GC GC GC 23 c YMC YMC YMC YMC 29 c YMC 30 c YMC 31 c YMC YMC 36 c YMC GC YMC YMC YMC YMC YMC YMC YMC Continued on next page 73

87 Table 3.2 Continued from previous page ID λ min -λ max Count a S/N b W Hβ (Å) Type YMC Mask YMC YMC GC YMC YMC YMC GC YMC YMC YMC a Non-calibrated flux count at 5000 Å. b S/N estimated by N N where N is the flux of the continuum at 5000 Å. c Also on mask Results and Analysis Initial Age Categorization We first visually categorize each cluster as either young or old. Clusters roughly less than 1 Gyr old are considered young and are called young massive clusters (YMCs). Clusters older than 1 Gyr are considered old and are called globular clusters (GCs). The criteria used to determine whether an object is a GC or YMC are primarily 74

88 Figure 3-5: GMOS spectra of the YMCs on mask 1. All spectra shown here through Fig are continuum subtracted. 75

89 Figure 3-6: Continued GMOS spectra of the YMCs on mask 1. 76

90 Figure 3-7: GMOS spectra of the YMCs on mask 2. 77

91 Figure 3-8: GMOS spectra of the GCs on mask 1. 78

92 Figure 3-9: Continued GMOS spectra of the GCs on mask 1. 79

93 Figure 3-10: GMOS spectra of the GCs on mask 2. visual with some constraints from the equivalent width of the Hβ Balmer line (W Hβ ), roughly measured with the SPLOT task of IRAF (see Table 3.2). The visual cues for identifying a spectrum as a YMC include a strong series of Balmer lines. Figures 3-5 through 3-7 clearly show a strong Balmer line series in the spectra of YMCs, and Table 3.2 lists the W Hβ 5.34 Å for YMCs. The cues for GCs include weaker Balmer lines and the presence of Ca H and K ( 3900 Å), G band ( 4300 Å), and Fe5170 lines. Figures 3-8 through 3-10 show the weaker Balmer lines, G band, and other metal lines. 14 objects on the masks are not included in our final sample. They include five background galaxies we identify based on redshifted Ca H and K lines. These galaxies also have very reddened photometry. Two objects appear to be red supergiant stars in the MW. One object is rejected because its S/N is so low no lines can be discerned. Also, six clusters show emission lines that indicate either embedding within or close proximity to H II regions. This leaves 25 YMCs and 16 GCs for our spectroscopic sample. 80

94 3.3.2 Measuring Ages and Metallicities Using BaSTI Models We measure the ages and metallicities of the clusters by comparing spectral indices of absorption lines in the cluster spectra to those of synthetic spectra of SSPs from the Bag of Stellar Tracks and Isochrones (BaSTI) library 7 (Percival et al., 2009). Spectral indices are equivalent widths of (in this case absorption) lines, given by W λ = λ 2 λ 1 (1 F λ F C )dλ where λ 1 and λ 2 define the region where the line is measured (the passband), F λ is the flux of the absorption line and F C is the flux of the continuum. Here, the indices are measured with respect to a pseudo continuum drawn between regions redward and blueward of the passband. We use the index system defined from several decades of study of stellar, cluster, and galaxy spectra taken on the Lick Observatory/image dissector scanner (Lick/IDS indices) (Faber et al., 1985; Burstein et al., 1986; Worthey et al., 1994; Worthey & Ottaviani, 1997). We first shift the cluster spectra according to the centers of several strong absorption lines found with the SPLOT task of IRAF to account for doppler shifts that would otherwise offset the passband regions of the indices. We then measure Lick/IDS indices of both the cluster spectra and BaSTI library of synthetic spectra using the LICK EW IDL function written by Genevieve Graves 8, which smooths the input spectrum to the 8 11 Å/pixel resolution of the Lick/IDS index system and determines the W λ of each index in the smoothed spectrum. Figure 3-11 and 3-12 show the twelve out of 25 indices used here. These indices are chosen because they are strong enough to be observable above the noise in most of the cluster spectra and are dependent on either age (Hβ, HγA, HγF, HδA, and HδF) or metallicity (Fe4531, Fe5015, Mg2, Mgb, Fe5270, Fe5335, and Fe5406). Note that the Hγ, Hδ, and Mg indices have different definitions with different widths of the passbands and/or pseudo 7 Available at 8 Part of the EZ AGES IDL code package described in Graves & Schiavon (2008) and available at 81

95 continua. Mg2 is considered a measurement of a molecular band, rather than an atomic band, with a magnitude index measurement, rather than W λ index in Å, given by Mg2 = 2.5 log[( 1 λ 2 λ 1 ) λ 2 λ 1 F λ F C dλ]. Also, three of the metal indices are combined to form a single index, [MgFe] = Mgb (0.72 Fe Fe5335), known for being particularly α-element insensitive (Puzia et al., 2005a). By measuring the Lick/IDS indices for BaSTI model spectra, we can create grids of index values associated with known ages and metallicities. We choose BaSTI models with solar α-element abundance 9 and Reimers (1975) red giant mass loss parameter η = 0.4. The BaSTI spectra are generated by combining individual stellar spectra according to a Kroupa (2001) initial mass function, ξ(m) m α i where α 0 = 0.3±0.7 (0.01 m/m < 0.08), α 1 = 1.3 ± 0.5 (0.08 m/m < 0.5), α 2 = 2.3 ± 0.3 (0.5 m/m < 1.00), and α 3 = 2.3 ± 0.7 (1.00 m/m ), to form an integrated spectrum of a single age and metallicity population (an SSP). The BaSTI library contains both low and high resolution spectra, but we only use the high resolution spectra here (1.0 Å/pixel). We utilize SSPs with ten different metallicities (Z = 0.04, 0.03, (solar), 0.01, 0.008, 0.004, 0.002, 0.001, , and with [Fe/H] = 0.40, 0.26, 0.06, -0.25, -0.35, -0.66, -0.96, -1.27, -1.49, and -1.79) and ages per metallicity spanning from 50 Myr to 13.5 Gyr. Figure 3-13 shows some example index-index grids of constant ages and metallicities using two age sensitive Balmer indices and two metallicity sensitive indices for old ages and less than solar metallicities. The overlap of the 13 Gyr age line is caused by the effect of the mass loss parameter η = 0.4 which mimics clusters with a blue horizontal branch (HB), rather than setting η = 0.2 for clusters whose evolved red giant branch (RGB) stars form a red clump. Bluer HBs generally correlate with lower metallicities; however, there is a second parameter effect which causes some higher 9 α-elements (C, O, Ne, Mg, Si, S, Ar, and Ca) are created during the alpha process from He burning during nuclear fusion. 82

96 metallicity clusters to also have blue HBs (Rich et al., 1997; Sweigart & Catelan, 1998). Thus, despite the inconvenience of the overlapping 13 Gyr model, we choose to use the η = 0.4 models for all measurements. Note that overlap of the 13 Gyr model the blue HB is less pronounced at the lowest metallicities because blue HBs at these metallicities form extended tails down to faint magnitudes which contribute less total flux to the Balmer lines, despite hot temperatures. For determining the ages and metallicities of younger clusters we use Lick/IDS indices of SSPs with younger ages and include higher metallicity tracks. Figure 3-14 shows an index-index plot and illustrates that at younger ages, the uniform grid configuration of the constant age and metallicity modeled indices breaks down. It is still possible to fit age and metallicities to the YMC spectra, however, by doing a least χ 2 fit to multiple indices at once (Trancho et al., 2007). This is a more robust method than plotting clusters on the index-index grids alone, and we employ it for both the YMCs and GCs (see 3.3.3). 83

97 Figure 3-11: Six of the 12 Lick/IDS index regions highlighted on a BaSTI synthetic spectrum (black line, Z = 0.01 and age = 1 Gyr SSP) utilized in fitting the age and metallicities of the cluster spectra. The smoothed spectrum is overplotted in red. The passband regions are within the solid blue lines (yellow lines filling the area used to measure W λ ), and the bounds of the pseudo continua are shown in green. 84

98 Figure 3-12: The other six of the 12 Lick/IDS index regions highlighted on the same BaSTI synthetic spectrum with the same color coding as Fig

99 Figure 3-13: Example index-index grids showing lines of constant ages (solid lines, from top to bottom: 1, 2, 3, 4, 5, 7, 10, and 13 Gyr) and constant Z (dotted lines, from left to right: , , 0.001, 0.002, 0.004, 0.008, 0.01, (solar, not available for the bottom plots) or [Fe/H]: -1.79, -1.49, -1.27, -0.96, -0.66, -0.35, -0.25, 0.06). 86

100 Figure 3-14: Example index-index grid showing lines of constant ages (solid lines, labeled: 50, 100, 400, 800 Myr) and constant Z (crossing dotted lines: 0.002, 0.008, (solar), 0.04 or [Fe/H]: -0.96, -0.35, 0.06, 0.40). 87

101 3.3.3 Ages and Metallicities In order to fit the ages and metallicities of the clusters, we run a least χ 2 minimization fit of each set of twelve cluster indices with a sufficiently high quality (based on individual visual inspection to remove lines with too low S/N for the line to be clearly distinguished from the noise of the spectrum) to the model BaSTI indices. The χ 2 is divided by the sum of the squares of the errors for each index output by the LICK EW code which are based on the S/N plane of each spectrum. To add robustness to the fits, we run the χ 2 minimization 5000 times for each cluster with artificial errors to each index that follow a normal distribution with a width, σ, equal to the LICK EW error of the index. We then take the final age and metallicities of the clusters as the mean of the 5000 minimized χ 2 best fitting models with errors as the standard deviation of the best fitting models. For the YMCs, we only fit to models with ages up to 2 Gyr, and for the GCs, we only fit models with ages down to 600 Myr and metallicities up to solar. Also note, that Fe5015 was of poor or questionable quality in all of the GCs, so only up to eleven indices were fit for each GC. We compare the age and metallicity results of the χ 2 fits to the closest age and metallicity tracks on index-index plots, and find similar results which give us added confidence in the accuracy of the χ 2 fits. Figure 3-16 shows the [Fe/H] versus the log of the ages for both GCs and YMCs with error bars over plotted. 88

102 Figure 3-15: HδA (top) and HδF (bottom) indices for GC ID 10 illustrating an index rejected from the χ 2 minimization fitting (top) and an index kept in the fitting (bottom). 89

103 Figure 3-16: [Fe/H] vs. log Age for all clusters. Open squares with blue error bars represent YMCs, and solid circles with red error bars represent GCs. The two dotted lines at [Fe/H] = 0.55 and 1.56, which correspond to the metal rich and poor peaks respectively of the MW GC system. 90

104 3.3.4 Spatial-Metallicity Distribution Figure 3-17 shows the [Fe/H] versus the distance of each cluster as projected onto the plane of the sky (r gc ). Note that the sample of GCs follows the centrally concentrated distribution of bright, red clusters seen in while the YMCs are spread to further distances, limited by the extent of the two GMOS masks for the observations. Figure 3-17: [Fe/H] vs. r gc for all clusters. Open squares with blue error bars represent YMCs, and solid circles with red error bars represent GCs. 91

105 3.4 Discussion Are There Young to Intermediate Age GCs in M101? From Figure 3-16, we see that the ages for the sample of M101 star clusters is almost continuous, most surprisingly even for ages greater than 1 Gyr. Specifically, there are six clusters with ages 4 5 Gyr (IDs 4, 11, 13, 15, 17, and 37) and four clusters with ages 1 2 Gyr (IDs 3, 14, 18, and 28 on mask 2). Before definitively determining if M101 has young to intermediate age GCs, we must consider possible sources of error on the ages. From the model fit errors alone, we see that the ages of ID 28 and ID 15 are not well constrained (note that these clusters both have very poor Balmer lines with little to no age lines possible to fit), and we cannot definitively conclude that they are young to intermediate age. Percival & Salaris (2009) analyze the level of error in measured ages and abundances of clusters caused by differences in the model indices from T eff ( 100 K offsets), log g ( 0.25 dex offsets), and [Fe/H] ( 0.15 dex offsets). They find these differences inherent in chosen models affect the ages (for Hβ up to 20%). However, the affects of the model differences on the other Balmer lines as well as between the different metal lines are often in opposition (one index strength may increase while another decreases). Thus, fitting many indices at once, as we have done here, can reduce systematic shifts due to the choice of models. Percival & Salaris (2009) also caution against over interpreting mismatching results from fitting only a single age and metal index at a time as the result of multiple stellar populations or α-element enhancement 10. We do not assume α-element enhancement here. Only ID 37 of the remaining young and intermediate age candidate GCs has an error much greater than the up to 20% deviation between model choices, so we 10 α-element enhancement is the enhancement of α-element (C, O, Ne, Mg, Si, S, Ar, and Ca) abundance ratios over solar values which is seen in the spectra of some stars and clusters. 92

106 exclude this cluster from being definitely categorized as intermediate age. We also hesitate to report a definitive age for ID 3 as it contains more starkly mismatched equivalent widths for the Balmer series of lines with Hβ indicating a much younger age (< 700 Myr) than Hγ or Hδ as well as very strong, sharp absorption lines at Fe5015 and λ 4964 Å(see Fig. 3-8). We excluded Hβ from the age determination of ID 3, rather than the other three available Balmer Lick indices, in since the cluster shows some observable G4300 absorption, which typically indicates an age of at least 3 Gyr. Therefore, there are at least four GCs (IDs 4, 11, 13, and 17) with ages most likely within the intermediate age range of 3 8 Gyr and two GCs (IDs 14 and 18) with even younger ages of 1 2 Gyr Metallicities of the YMCs and GCs Figure 3-16 shows that the metallicities of the YMCs and GCs roughly separate into metal-rich and metal-poor populations as expected if the YMCs formed more recently out of metal enriched gas in the disk of M101 and the GCs in the spheroidal components formed earlier out of gas that had fewer metals. Interestingly, the transition between the young to old clusters follows an almost continuous trend of decreasing metallicity. This hints at more or less continuous cluster formation, possibly aided by mergers/interactions, in M101 s past. Half of the GC sample here have metallicities above the lower limit of the radial abundance gradient observed for MW bulge ([Fe/H] = 1.0, Minniti et al. (1995)). While the mean of all of the GC metallicities is [Fe/H] = 1.0 ± 0.4, dividing our sample in half along [Fe/H] = 1.0 gives a metal-poor GC mean [Fe/H] = 1.3±0.1 and metal-rich mean [Fe/H] = 0.7 ± 0.3. While this division is artificially imposed, it suggests that these GCs may be sampling both a bulge/thick disk population of GCs as well as an older more metal poor halo population. 93

107 From Figure 3-17 we see that the GCs studied here are entirely within the inner portions of M101 with r gc < 8 kpc. Three of the eight GCs with [Fe/H] > 1.0 (IDs 10, 14 on mask 1, and 28 on mask 2) and two of the eight GCs with [Fe/H] < 1.0 (IDs 8 and 11) are at r gc < 1.2 kpc, which is the upper limit of the effective radius of the bulge of M101, r e = pc (Fisher & Drory, 2010). While the presence of clusters with a spread in metallicities (including metal-rich clusters) could support a bulge formed through mergers rather than by secular processes, our sample size is not sufficiently large enough to rule out a secularly formed pseudo-bulge, which is the type of bulge expected for M101 given its morphology. The metal-poor half of the GCs have a mean [Fe/H] greater than observed for the MW GCs ([Fe/H] 1.5). These GCs are also overall younger (mean age 7±3 Gyr) than the very ancient MW GCs ( 13 Gyr). This combination of elevated metallicities and an age spread encompassing younger ages most likely indicates a rich history of mergers and accretion in the halo of M101, possibly blurring between a thick disk population of GCs and inner halo GCs. 3.5 Conclusions We find that the massive star clusters in the spiral galaxy M101 have a fairly continuous spread of ages and metallicities. The YMC and GC populations do separate into younger and more metal-rich versus older and more metal-poor populations as expected for a typical spiral galaxy structure of disk and spheroidal components. However, the transition between the two categories of GCs is not stark, and there are at least four GCs with intermediate ages 3 8 Gyr and two GCs with young ages 1 2 Gyr, unlike clusters in the MW. Our overall conclusion is that M101 has a rich, possibly continuous history of cluster and star formation. The metallicities and spatial distribution of the GCs indicate that the cluster formation is most likely 94

108 driven by mergers/accretion. 95

109 Chapter 4 Gemini/GMOS Spectra of Old and Young Star Clusters in M101: Kinematics 4.1 Background The kinematics of a population of star clusters allow us to disentangle the components and structure of their host galaxies. In the MW, old GCs are found in the bulge and halo components while younger open clusters are found in the thin or thick disks (Portegies Zwart et al., 2010). In M33, another bulgeless spiral galaxy (of much lower mass than M101), Chandar et al. (2002) find a population of old GCs associated with the disk/pseudobulge and old GCs in the halo as well as a population of intermediate age clusters with evidence of disk motions. In this chapter we use velocities measured from Gemini/GMOS spectra of a sample of M101 YMCs and GCs to disentangle the structure of the galaxy, and add insight into the formation history. Both the line-of-sight velocities as well as the rotational velocity, v rot, can reveal the structure of a cluster population. For nearly face-on galaxies such as M101, high dispersion in the line-of-sight velocities (σ) are indicative 96

110 of a spheroidal component such as a bulge or halo, while low σ correspond to disk populations. In the MW, old open clusters have σ old, open = 28 km/s, disk/bulge GCs have σ bulge,gcs = 67 km/s and halo GCs have σ halo,gcs = 114 km/s (Chandar et al., 2002). The ratio v rot /σ is another useful measurement as values > 1 are rotationally supported systems while v rot /σ < 1 indicate a pressure supported system (such as a halo). Kormendy et al. (2010) use HST photometry and high resolution (R λ/fwhm 15, 000) spectra from the Hobby-Eberly Telescope to decompose and determine velocity dispersions of the nuclear regions of several late-type spiral galaxies, including M101. They find the inner velocity dispersion (most likely corresponding to the pseudobulge) to be low, σ pseudobulge = 27±4 km/s, which may mean it is more of an inner disk than a bulge in structure. van Dokkum et al. (2014) examined the surface brightness profile of M101 out to r gc = 70 kpc in order to fit the bulge, disk, and halo components. They find a surprisingly low halo mass fraction f halo = M halo /M tot = compared to that of the MW f halo In this chapter, we see how the cluster kinematics in M101 fit into this apparently highly disk dominated galaxy. The rest of this chapter is arranged as follows: In 4.2, we describe measuring the cluster velocities and compare the velocities to the cluster spatial distribution and ages. We discuss the results in 4.3. Finally in 4.4, we list our conclusions. Note that we do not discuss the observations here as we are using the same spectra examined in the previous chapter with reduction as discussed in 3.2 and general age categorization as described in

111 4.2 Results and Analysis Velocity Measurements Velocity measurements were obtained using the PyRaf task FXCOR to perform a Fourier cross-correlation of each cluster spectrum with a template spectrum. The template spectra are good S/N clusters from within our own sample so that the velocities measured for each object are robust on a relative scale. We choose one YMC (ID 36 on mask 1, see Fig. 3-5) and one GC (ID 11, see Fig. 3-9) as templates for measuring the relative velocities for the YMCs and GCs respectively. By using templates within our own sample, we ensure that the template spectra are a decent match to the object spectra, and remove any complications from mismatches in resolution and other effects that can occur if the template spectra come from models or are taken on a different instrument or in a separate program. Both the object and template spectra for each FXCOR run were first continuum subtracted and various wavelength ranges were tested. The wavelength ranges chosen for each object were those with the largest range that contained good S/N, clear lines, and lacked masked regions. The three important parameters of the cross-correlation of the object and template spectra are the position, height, and width of the strongest peak of the correlation function (see Fig. 4-1). The position gives the shift between the spectra, and thus the relative velocity. The height indicates the goodness of the fit (the best possible value being for a spectrum cross-correlated with itself, which has a peak height of 1). Velocity errors are computed in FXCOR from the ratio of the height of the strongest peak to the height of the average peak in the correlation (i.e. the antisymmetric noise of the correlation function) according to the methods of Tonry & Davis (1979). The width of the peak combined with the width of the template reveal the velocity dispersion of the object, which is mainly relevant for galaxy spectra as 98

112 velocity dispersions within star clusters are quite low ( 2 10 km/s). For each template spectrum, we calculate velocities using Doppler Shift, v = c( λ cluster λ rest λ rest ), from the Hβ, Hγ, and Hδ absorption lines and take the mean as the velocity of the template. The appropriate template velocity is added to the relative velocity of each cluster to estimate an absolute velocity, v cluster, for each cluster. These velocities are reported in Table 4.1 along with the FXCOR errors. Also included in Table 4.1 are velocity errors estimated from the results of running FXCOR on the individual exposures of the combined cluster spectra with the same parameters, wavelength regions, and template spectra as the combined cluster spectra. We take the standard deviation of the velocities of the individual exposures as the error on the v cluster. We find these errors are more comprehensive and realistic than the FXCOR calculated errors as they reflect potential variations in exposures with different random noise patterns and exposures taken on different nights. For example, we expect the YMCs to have higher errors than the GCs since their Balmer series lines are broader and have less sharply defined positions with which to determine shifts in the spectra, and we find that the median FXCOR errors for the GCs 19 km/s and YMCs 11 km/s, while the medians for the velocity errors are 32 and 40 km/s, respectively. Table 4.1: Kinematics of M101 Clusters with Spectra ID v cluster (km/s) FXCOR Error (km/s) v cluster Error (km/s) v disk (km/s) Mask 1 3 a Continued on next page 99

113 Table 4.1 Continued from previous page ID v cluster (km/s) FXCOR Error (km/s) v cluster Error (km/s) v disk (km/s) a a a a a a Continued on next page 100

114 Table 4.1 Continued from previous page ID v cluster (km/s) FXCOR Error (km/s) v cluster Error (km/s) v disk (km/s) Mask a Also on mask Velocity Distance Comparison Figure 4-2 shows the positions for each cluster with the semi-major and minor axes plotted (major axis position angle 39, Bosma et al. (1981)). From this plot we can obtain the perpendicular distance from each cluster to the semi-minor axis, R semi-minor. We then obtained estimates of the disk velocities, v disk, for the RA and Dec of each cluster from a study of M101 s HI gas by Bosma et al. (1981) (see Fig. 4-3 and Table 4.1). Figure 4-4 shows v cluster (or v disk ) versus R semi-minor. Also shown are the linear best fits to the YMCs, GCs, and the HI gas disk velocities. The slopes 101

115 Figure 4-1: Example correlation functions for a YMC (top two plots, ID 23) and a GC (bottom two plots, ID13). The smaller top plots in each panel show the entire correlation function with multiple peaks, and the larger plots are a closer view of the strongest peak of each function showing the best fits to the peak (dashed lines). 102

116 and errors of the best fits are found in Table 4.2. Note that we exclude all clusters with v cluster errors > 100 km/s (ID 42 on mask 1 and IDs 10, 17, 26, and 28 on mask 2) from Fig 4-4 as well as the fits to the GCs and YMCs as these clusters have large uncertainties from noisy spectra and sometimes poor coverage, limiting the wavelength coverage available for the cross-correlation. Figure 4-2: Positions of the YMCs (open squares) and GCs (solid circles) with respect to the semi-major and minor axes of M101 (labeled). 103

117 Figure 4-3: HI gas map with velocity contours overlaid from Bosma et al. (1981). Note the axes are B1950 coordinates (δ B1950 = 54 from bottom to top 25, 35, 45, and 55 and α B1950 = 14h from right to left 0m, 1m, 2m, and 3m). 104

118 Figure 4-4: Velocities versus distance to the semi-minor axis of the YMCs (open squares) and GCs (solid circles). The best fit lines to the GCs (solid line), YMCs (dotted line), and HI gas disk (dashed line, individual points not shown) are overplotted. The YMCs have a similar although slightly shallower slope than the HI gas disk, while the GCs do not match the disk rotation (see Table 4.2). Table 4.2: Coefficients for v cluster vs. R semi-minor Fits Data Set Intercept Slope GCs 207 ±7-5 ±5 YMCs 249 ±8 16 ±3 HI gas 243 ±3 18 ±1 105

119 4.2.3 Age Velocity Comparison The difference between v disk and v cluster roughly corresponds to the distance of each cluster above or below the gas disk. Thus, we show in Figure 4-5 this velocity difference versus the age of the clusters. The spread in velocity differences is greater for the GCs than the YMCs, and even shows the same broadening with increasing age within the YMC population. The standard deviation of the velocity differences of the two populations are the best estimate of the velocity dispersions, σ, of the YMC and GC populations. We find that σ YMC 25 km/s and σ GC 66 km/s. Figure 4-5: The difference between the cluster and disk velocities versus ages of the YMCs (open squares) and GCs (solid circles). 106

120 4.2.4 Rotational Velocity Calculation and Comparison Because the inclination of M101 is not quite zero, i.e. not perfectly face-on, the rotational velocity can be calculated from the line-of-sight velocities for each star cluster associated with the disk and for the HI gas along the semi-major axis. We determine the rotational velocities by v rot = (v los v sys )/cos θ sin i where v los is the line-of-sight velocity (either v cluster or v disk here), v sys is the systemic velocity of the galaxy ( 241 km/s for M101), θ is the PA of each object with respect to the semi-major axis, and i is the inclination of the galaxy ( 18 for M101 (Bosma et al., 1981)). In and 4.2.3, we find that the GCs do not show evidence of association with the disk of M101, and therefore, the formula for v rot (dependent on sin i) is not valid. Thus, we do not include the GCs in our rotational velocity analysis. Table 4.3 shows the mean rotational velocities for the HI gas in the disk and the YMCs along with standard deviations as the errors. We list the values for M101 populations in their entirety (excluding clusters with v cluster error > 100 km/s) as well as only for r gc > 5 kpc as the outer regions of the rotation curve more accurately reflect the peak of the rotation curves. We also list the values for the MW and M33 for comparison. Figure 4-6 shows v rot as a function of distance, i.e. the rotation curves, for the YMCs and HI gas in the disk. The HI gas disk shows an extremely high rotational velocity for the innermost datapoint because of the extreme proximity of this point to the semi-minor axis, which results in a spurious velocity. None of the YMCs are as close as this point to the semi-minor axis. 107

121 Figure 4-6: The rotational velocities versus r gc for the YMCs (open squares) and HI gas (dashed line). 108

122 Table 4.3: Rotational Velocity and Velocity Dispersion Comparisons Data Set v rot (km/s) v rot Error a (km/s) σ (km/s) v rot /σ MW b Young Clusters Old Open Clusters Disk/bulge GCs Halo GCs ± 0.2 M33 b Young Clusters Disk/bulge GCs Halo GCs All M101 HI gas c YMCs GCs 66 M101, r gc > 5 kpc HI gas YMCs a For M101 data, the error is the standard deviation of the individual cluster or gas v rot. b From Chander et al. (2002). For M33, disk/bulge GCs are within 2.25 kpc while halo GCs are outside 2.25 kpc. c Excludes only the innermost disk data point due to close proximity to the semi-minor axis skewing the calculation of v rot. 109

123 4.3 Discussion The Structure of M101 Figure 4-4 and Table 4.2 show that the YMC line-of-sight velocities as a function of the distance follow the HI gas in the disk. Likewise, we find that the M101 YMC v rot /σ ratio is well above a value of one (see Table 4.3), and therefore, shows definite signs of disk-like rotation. However, the YMCs have some scatter about the HI disk fitted v cluster -R semi-minor relation, a lower v rot /σ ratio than the HI gas, and their v rot /σ is more comparable to that of the old open clusters in the MW. Thus, the YMCs lag the HI disk and have a larger velocity dispersion. Interestingly, σ YMC is similar to σ pseudobulge determined by Kormendy et al. (2010) ( 25 km/s compared to 27 ± 4 km/s), which supports the possibility of the M101 pseudobulge having more of an inner disk-like structure. The GCs do not show line-of-sight velocities as function of distance that follow the HI gas in the disk, despite having lower v cluster errors than the YMCs. While this is more indicative of a halo than a disk population, the σ of the GCs is more similar to the disk/bulge GCs of the MW. We cannot rule out the possibility of the GCs sampled here forming a pseudobulge or thick disk population; however, the effective radius determined from light profile fitting of the pseudobulge (r e = pc, Fisher & Drory (2010)) is small enough to call into question these GCs (some at distances out to more than 6 kpc) being associated with the bulge alone. There are most likely some halo GCs (also supported by the more complete photometric GC catalog discussed in Ch. 2), and their existence conflicts with the results of van Dokkum et al. (2014) showing M101 as containing little to no halo component. Figure 4-5 shows that M101 has a continuous increase in its dispersion of the velocity residuals (with respect to the disk velocities) with age. This is similar to the trend seen for M33 clusters (Chandar et al., 2002), and indicates that systems of older 110

124 clusters have undergone a source of heating whether by perturbations from passing giant molecular clouds (a secular process) or mergers/accretion. The lag between the YMCs and the HI disk and the larger velocity dispersion of the YMCs compared to the HI disk also supports this conclusion. 4.4 Conclusions We find the kinematics of the YMC and GC populations differ with the YMCs following the characteristics of the HI gas in the disk and the GCs exhibiting line-ofsight and rotational velocities and velocity dispersion similar to either a bulge/thick disk or halo. The smooth increase in the difference between the cluster velocities and their local disk velocities with age indicates that M101 may have undergone heating of its disk over time or a fairly continuous merger/accretion history. 111

125 Chapter 5 Conclusions and Future Work We find that M101 contains a rich population of star clusters from which clues about the galaxy s formation and history can be obtained, adding to our picture of the formation and evolution of spiral galaxies with little to no bulge. M101 appears to contain an almost continuous range of star cluster luminosities (and most likely masses) and ages, along with a range of metallicities. We specifically find: A population of bright, old halo clusters typical of spiral galaxies and dominated by metal-poor clusters as expected for a nearly bulgeless galaxy such as M101. A large population of faint, old disk clusters, which are not seen in other spiral galaxies like M81, M83, M51, and the MW. A sample of young (< 1 Gyr), massive clusters similar to those seen in other spiral galaxies like M51 and M83 (though discovered in lower numbers within the MW thus far). A sample of young, intermediate, and old age GCs with a wide spread in metallicities, and kinematics/spatial distribution placing them in the bulge/thick disk or halo. Similar populations are seen in spiral galaxies such as M31 and M33. We conclude that the disk of M101 is a more hospitable place for cluster formation and survival than that of MW, possibly more similar to that of M33, another late- 112

126 type spiral. The possibility of bulge clusters in M101 raises the question of whether they formed via secular or hierarchical processes. The morphology of M101 strongly suggests that secular processes are more likely as pseudobulges are thought to form internally. However, the metallicity spread of the old clusters is more ambiguous, and does not definitively rule out a merger/accretion scenario. 5.1 Searching for Faint, Old Star Clusters in Other Spiral Galaxies One of the most important questions that remains from this work is: how long lived are star clusters, especially low mass clusters? This question is crucial to understanding how star clusters evolve and how galaxies become populated by field stars. This question can be answered by determining just how common the newly discovered faint, old M101 clusters discussed in Ch. 2 are in other galaxies, especially spiral galaxies. Thus, the methods employed throughout this work need to be applied to other spiral galaxies in the future. Good candidate spirals include NGC 6946, M94, M106, NGC 300, NGC 1313, NGC 4395, NGC 5055, and NGC 7793 as they are roughly face-on and at distances from 2 8 Mpc away with available archived HST /ACS/WFC or WFC3 B, V, and I images. Imaging with HST is crucial to identify extragalactic star clusters beyond the Local Group, and at the distances of the proposed galaxies, clusters imaged with HST are broader than the point spread function and can be separated from bright field stars. NGC 6946 is a particularly good candidate galaxy for such a search as it has already shown a possible excess of faint, red clusters as identified from HST /WFPC2 images (Chandar et al., 2004). Unfortunately, the WFPC2 observations are not deep enough to definitively identify a population of old disk clusters, so HST imaging from 113

127 deeper, higher resolution instruments is necessary. Recently, observations of one field covering part of the disk of NGC 6946 in B, V, and I filters on HST /WFC3 have become available (see Fig 5-1). More fields are needed to throughly investigate the old star cluster populations of NGC 6946, but searching for faint, old clusters within the available field combined with the previous findings of Chandar et al. (2004) could provide a substantial start to answering the question of how common potentially long lived, low mass star clusters are. In order to determine the ages or metallicities of any faint, red clusters in other spiral galaxies, spectra of the faint objects are needed. This presents a challenge as high S/N spectra are needed for Lick index measurements. However, it is possible to get the required S/N by stacking spectra of different clusters with similar photometric properties (such as color and luminosity). Also, because lower S/N is required for measuring velocities, it would still be possible to measure the velocities of the low S/N individual spectra, and thus insights on the cluster population kinematics are also achievable. 114

128 Figure 5-1: Ground-based optical image of NGC 6946 showing the location of the HST /WFC3 field available. 115