A New Perspective on Galaxy Evolution from the Low Density Outskirts of Galaxies

Home , A New Perspective, Leo Ring, M101 Group, M96 Group, NGC 300, NGC 3077

A NEW PERSPECTIVE ON GALAXY EVOLUTION FROM THE LOW DENSITY OUTSKIRTS OF GALAXIES

by AARON E. WATKINS

Submitted in partial fulﬁllment of the requirements for the degree of Doctor of Philosophy

Department of Astronomy Dissertation Adviser: Dr. J. Christopher Mihos

CASE WESTERN RESERVE UNIVERSITY

August, 2017 CASE WESTERN RESERVE UNIVERSITY SCHOOL OF GRADUATE STUDIES

We hereby approve the dissertation of Aaron E. Watkins candidate for the degree of Doctor of Philosophy.

Committee Chair J. Christopher Mihos

Committee Member Paul Harding

Committee Member Stacy McGaugh

Committee Member Heather Morrison

Committee Member Steven A. Hauck

Dissertation Defense June 19, 2017 Dedicated to the three most important people in my life: Mom, Dad, and Nathaniel Contents

Title 1

Signature Sheet ii

Table of Contents iii

List of Tables vii

List of Figures viii

Preface x

Acknowledgements xi

Abstract xiii

1 Introduction 1 1.1 Tidal Interactions and Environment ...... 4 1.2 Secular Evolution and the Outer Disk ...... 9 1.3 Star Formation At Low Density ...... 14 1.4 Summary, and the Aims of This Thesis ...... 18

2 Searching For Diﬀuse Light in the M96 Galaxy Group 20 2.1 Abstract ...... 20 2.2 Introduction ...... 22 2.3 Observations and Data Reduction ...... 28 2.3.1 Observations ...... 28 2.3.2 Data Reduction ...... 29 2.4 Large Scale Diﬀuse Light ...... 32 2.5 Individual Galaxies ...... 41 2.5.1 M105 ...... 42 2.5.2 NGC 3384 ...... 45 2.5.3 M96 ...... 47

iv 2.5.4 M95 ...... 50 2.6 Discussion ...... 52 2.6.1 The Origin of the Leo Ring ...... 52 2.6.2 The Lack of Intragroup Light ...... 57 2.7 Summary ...... 60 2.8 Acknowledgments ...... 62

3 Deep Imaging of M51: a New View of the Whirlpool’s Extended Tidal Debris 63 3.1 Abstract ...... 63 3.2 Introduction ...... 65 3.3 Observations and Analysis ...... 67 3.4 Results ...... 69 3.4.1 Morphology and Color of Tidal Features ...... 69 3.4.2 Comparison to Simulations ...... 75 3.5 Acknowledgements ...... 77

4 The Red and Featureless Outer Disks of Nearby Spiral Galaxies 78 4.1 Abstract ...... 78 4.2 Introduction ...... 80 4.3 Observational Data ...... 85 4.3.1 Deep Optical Imaging ...... 85 4.3.2 Ancillary Multiwavelength Data Sets ...... 87 4.4 Analysis Techniques ...... 88 4.4.1 Surface Brightness and Color Proﬁles ...... 88 4.4.2 Fourier Analysis ...... 90 4.5 Individual Galaxies ...... 91 4.5.1 M94 (NGC 4736) ...... 96 4.5.2 M64 (NGC 4826) ...... 99 4.5.3 M106 (NGC 4258) ...... 105 4.6 Discussion ...... 112 4.6.1 Outer Disk Stellar Populations ...... 113 4.6.2 Environmental Inﬂuences ...... 116 4.7 Summary ...... 120 4.8 Acknowledgements ...... 122 4.9 Appendix: ON THE BURRELL SCHMIDT PSF ...... 122

5 HII Regions and Diﬀuse Ionized Gas Throughout the M101 Group: Only the Intensity Changes 126

Abstract 127 5.1 Introduction ...... 128 5.2 Observations and Data Reduction ...... 133 5.2.1 Observations ...... 133

v 5.2.2 Data Reduction ...... 135 5.2.3 GALEX data ...... 140 5.2.4 Background/Foreground Contamination ...... 140 5.3 HII Region Photometry ...... 141 5.3.1 Extinction Correction ...... 142 5.3.2 Photometry ...... 144 5.3.3 Statistical Analysis ...... 150 5.3.4 The Unchanging Nature of HII Regions ...... 154 5.4 Diﬀuse Ionized Gas ...... 156 5.4.1 Isolating the DIG ...... 157 5.4.2 Results ...... 158 5.5 Discussion ...... 163 5.5.1 The Connection Between Star Formation and the DIG ...... 164 5.5.2 On the Observed Trends of Integrated FHα/FFUV ...... 167 5.5.3 The M101 Group As a Case Study ...... 170 5.6 Summary ...... 173 5.7 Acknowledgements ...... 175

6 Summary and Future Work 176 6.1 Major Results of This Dissertation ...... 176 6.2 Directions for Future Work ...... 179

Bibliography 183

vi List of Tables

2.1 Properties of Streamlike Features in the M96 Group ...... 30

3.1 Photometry of Tidal Debris in M51 ...... 67

4.1 M94, M64, and M106 Galaxy Properties ...... 92

5.1 Results of Levene’s Test Trials ...... 153 5.2 Integrated Properties of M101 Group Galaxies ...... 168

vii List of Figures

2.1 False-color image of the M96 Group ...... 25 2.2 Binned and labeled broadband mosaics of the M96 Group ...... 33 2.3 HI contours overlaid on the M96 Group V-band mosaic ...... 34 2.4 Schematic of M96 Group features ...... 36 2.5 GALEX and V-band images of Leo Ring star-forming complexes ...... 40 2.6 M105 surface brightness and color profiles ...... 43 2.7 Residuals of model-subtracted NGC 3384 & M105 ...... 44 2.8 NGC 3384 surface brightness and color profiles ...... 48 2.9 M96 surface brightness and color profiles ...... 50 2.10 M95 surface brightness and color profiles ...... 52

3.1 Rescaled B-band image of M51 ...... 66 3.2 M51 pixel-to-pixel colormap ...... 68 3.3 M51 21cm imaging overlay ...... 71 3.4 Surface brightness and color proﬁle of a wedge across M51 ...... 73

4.1 M94 multiwavelength comparison ...... 94 4.2 M94 surface brightness, color, and Fourier analysis profiles ...... 95 4.3 M64 multiwavelength comparison ...... 100 4.4 M64 surface brightness, color, and Fourier analysis profiles ...... 101 4.5 M106 multiwavelength comparison ...... 106 4.6 M106 surface brightness, color, and Fourier analysis profiles ...... 107 4.7 NGC 4248 closeup ...... 108 4.8 Burrell Schmidt B- and V-band PSF radial profiles ...... 124

5.1 Hα difference image mosaic of the M101 Group ...... 134 5.2 Tentative diffuse plume northeast of M101 ...... 138 5.3 Extinction correction comparison ...... 143 5.4 HII region fluxes and FHα/FFUV ratios, M101 ...... 146 5.5 HII region fluxes and FHα/FFUV ratios, NGC 5474 ...... 147 5.6 Inner disk/outer disk boundary in M101 ...... 148 5.7 Evolution of FHα/FFUV in Starburst99 ...... 151 5.8 Statistical comparisons of FHα/FFUV across the M101 Group ...... 151

viii 5.9 Observed and model distributions of FHα/FFUV in M101 ...... 155 5.10 M101 and NGC 5474 DIG reprojections ...... 159 5.11 Radial profiles of FHα/FFUV in the DIG ...... 160 5.12 HII region flux surface density correlation with DIG FHα/FFUV ...... 162 5.13 HII region flux surface density against HI and H2 column densities . . . . . 165

ix Preface

This dissertation presents the culmination of 6 years’ worth of my work as a graduate student. The material presented in Chapters 2, 3, and 4 has previously been published as (respectively): Watkins, Mihos, Harding, & Feldmeier (2014), “Searching for Diﬀuse Light in the M96 Galaxy Group”, ApJ, 791, 38; Watkins, Mihos, & Harding (2015), “Deep Imaging of M51: a New View of the Whirlpool’s Extended Tidal Debris”, ApJL, 800, L3; and Watkins, Mihos, & Harding (2016), “The Red and Featureless Outer Disks of Nearby Spiral Galaxies”, ApJ, 826, 59. Though each was authored by several people, I am the primary author on each and hence am responsible for the majority of the research, analyses, and written content therein.

x Acknowledgements

Maybe it’s an obvious place to start, but I want to ﬁrst thank my adviser, Chris Mihos. I started working with him my ﬁrst day here at CWRU, and, even in the face of occasional bouts of extreme grumpiness in my bumbling early days, he stuck with me and I stuck with him for the next six years. I know that without him, I’d be a vastly sloppier speaker, writer, and scientist. I also want to thank Paul Harding, who is really my other adviser. Paul taught me to pay attention to the nitty-gritty and to never take anything for granted. Until you’ve sat with him for a full 6 hours as he makes minute adjustments to the readout parameters in the CCD controller software, you have no idea just how sensitive an instrument a giant piece of polished glass and metal can be. And that kind of insight is invaluable. I would be remiss in neglecting to thank everyone else in this small department as well, from Idit Zehavi for her endless encouragements and many great conversations, to Heather Morrison for always forcing me to write and to say precisely what I mean to write and to say, Stacy McGaugh for keeping me skeptical of the establishment, and Earle Luck for reminding me that most of science is just rigorous common sense. I also can’t forget John Feldmeier and Pat Durrell of YSU, for valuable input on my research and for letting me in on the process of collaboration. And thanks to Steven Hauck; though I never worked closely with him, he still agreed to read and provide feedback on this massive document I’ve produced. And here’s to Charles Knox, without whom this department would implode, and Agnes Torontali, without whom this department would explode. I want to thank all of the other grad students and postdocs I’ve met throughout the years—Tom Reding, Zhibo “Real” Ma, Jay Franck, John Kerr, Lei Yang, Pengfei Li, Mar- cel Pawlowski, Federico Lelli, and Hong Guo—for keeping this department a community.

xi Finally, inevitably, I want to thank my family and friends for always being there for me, and for legitimately thinking that what I’ve been doing these past six years is super cool and totally worth doing.

xii A New Perspective on Galaxy Evolution From the Low Density Outskirts of Galaxies

Abstract

AARON EMERY WATKINS

In this dissertation, I present a series of studies on the low surface brightness outskirts of galaxies, which contain a record of tidal interactions and secular evolution processes. Each study utilized new deep imaging from the Burrell Schmidt Telescope in either broadband filters or narrow-band filters targeting Hα emission. Regarding tidal interactions, I present a study of the M96 Group (or Leo I Group), as well as deep imaging of the interacting pair M51. I find that the M96 Group’s intragroup light (IGL) consists of only three faint linear streams. I find no stellar counterpart to the group’s H I ring, unusual if it were collisional in origin, and few signs of interaction among its four most massive members, implying a very calm tidal history. In M51, I discover several extremely diffuse plumes of starlight, yet find no stellar counterpart to its H I tail. Additionally, I measure red (B − V ∼0.8) colors in all of its most extended tidal features, implying dominantly old populations and thus a lack of interaction-induced extended star formation. Regarding secular evolution, I conduct a detailed photometric study of three nearby galaxies’ outer disks. Each outer disk lacks both ongoing star formation and the spiral structure necessary to migrate stars from the inner disk, hence it is unclear how these red outer disks formed. Finally, I conduct a study of the H II regions and diffuse ionized gas (DIG) throughout the M101 Group, to determine whether star formation in low density environments occupies a distinct physical regime from its high density counterpart. I find that the distribution of Hα/FUV flux ratios (a tracer of the initial mass function, IMF) is constant among all H II region populations throughout the group. Also, the Hα/FUV ratio xiii in the DIG appears tied only to the local intensity of star formation, leaving little room for changing star formation physics. In total, this dissertation shows that tidal interactions in low-density groups may be infrequent, raises questions about the origin of extended red outer disks, and shows that star formation physics change little with local mass density.

xiv Chapter 1

Introduction

How do galaxies form? This is both a simple question, and a fundamental one. Perhaps it is the first question one might think to pose as soon as one considers studying galaxies, yet it is a question full of facets and nuances. Part of the difficulty in answering this question arises because galaxy evolution cannot be studied longitudinally; we cannot turn our telescopes on a galaxy at the dawn of the universe and take notes while we watch it grow and change. Even if we could, there is no guarantee we would pick a galaxy representative of galaxies as a whole. We want to form a picture that explains why some galaxies appear as randomly orbiting collections of old stars, while others appear as thin, gas-rich disks in which everything orbits in circles, why some galaxies are composed of only a few hundred stars, while others are composed of tens of billions of stars. So because we cannot watch galaxies evolve, we observe as many galaxies as possible and as many aspects of those galaxies as possible in a monumental effort to cobble together a single model that explains everything. And we do claim to have such a model. Based on the ΛCDM cosmological paradigm, it proposes that (aside from rapid formation of elliptical galaxies in dense cluster environments) galaxies form hierarchically; cold dark matter (CDM), a theoretical substance that does not interact via the electromagnetic force, forms the initial gravitational wells into which baryons collapse after the era of recombination (e.g. Hu & Sugiyama 1995; Hu et al. 1997). These wells then mutually attract and merge into successively deeper wells over time, stopping only when the accelerating expansion of the universe (as controlled by Λ, a theoretical energy, dubbed the cosmological constant) prevents further collapse (Percival et

1 al. 2000; Percival 2005). The smallest units thus form first and serve as the building blocks for the largest structures: a hierarchical formation (Press & Schechter 1974; Gott & Rees 1975; Searle & Zinn 1978; White & Rees 1978). In general (nevermind its two large assumptions: the two “dark” components it requires), this model provides a solid path along which galaxies evolve from the clumpy, gas-rich, heavily star-forming entities seen at high redshift to the highly structured disks, lenticulars, and ellipticals seen today (e.g. Bouwens & Illingworth 2006; Akiyama et al. 2008; Conselice & Arnold 2009; Carrasco et al. 2010; Conselice 2014). It also provides an explanation for why most galaxies now live in group or cluster environments (∼60%; Bah- call 2000), and why clusters themselves occupy yet larger structures known as webs and filaments (Davis et al. 1982; de Lapparent et al. 1986; da Costa et al. 1988; Doroshkevich et al. 2004). The ΛCDM model is a solid skeleton, hence only requires verification by way of the details. Yet the details are of primary importance; if it is discovered that some details cannot fit into the model, the model itself must be revised. While focus in the astronomical community has shifted toward large-scale cosmological surveys, effectively treating whole galaxies as single data points, the detailed structure of galaxies is still not fully understood. The faint outskirts of galaxies in particular are only beginning to come under examination; they are in fact mostly ignored in large surveys, which by necessity do not reach the required depth in surface brightness to study them in detail. This is problematic, because galaxy outskirts are sensitive probes not only of environmental factors, such as present or past tidal interactions and merger events (Toomre & Toomre 1972; Malin & Hadley 1997; Ibata et al. 2001; Majewski et al. 2003; Forbes et al. 2003; Bullock & Johnston 2005; Mart´ınez-Delgado et al. 2010), but also of the internal physics of the host galaxy itself, from the outward migration of old inner disk stars through interactions with spiral arms or bars (Sellwood & Binney 2002; Debattista et al. 2006), to the repressed growth of perturbations in low density environments leading to a lack of spiral patterns and new star formation (Toomre 1964; Kennicutt 1989). Outer disks provide a potential bridge between the high-density, spiral arm-dominated regimes already well-studied in the inner disks of massive galaxies, to the less understood regime of low surface brightness (LSB) galaxies and dwarf galaxies. Because outer disks are so faint—orders of magnitude below the surface brightness of

2 the night sky—observing them requires care and patience. In practice, this means long exposure times. This can technically be achieved using survey data; D’Souza et al. (2014), for example, attempted to study faint stellar halos by simply stacking images of thousands of different galaxies. Bakos & Trujillo (2013) and Mart´ın-Navarro et al. (2014) also studied halos using deep imaging available from Stripe 82 in the Sloan Digital Sky Survey (SDSS; York et al. 2000), a strip of sky targeted multiple times throughout the survey’s duration. Both of these methods have the capacity to reach the surface brightness required to study galaxy outskirts, however they both also carry risks. In stacking thousands of different galaxies, individual variations from halo to halo are washed out; because few stellar halos have been observed in detail (i.e. M31, the Milky Way, and Cen A; Belokurov et al. 2006; Lewis et al. 2013; Crnojevic´ et al. 2016), there is currently no way to determine how galaxy- to-galaxy variations might bias a stacking experiment (Monachesi et al. 2016; Harmsen et al. 2017). Also, SDSS was not undertaken with LSB studies in mind, and so it lacks characterization of important sources of uncertainty at low surface brightness, such as the point-spread function (PSF; de Vaucouleurs 1948, 1958; King 1971; Michard 2002), which can mimick stellar halos or thick disks if not properly accounted for (de Jong 2008; Slater et al. 2009; Sandin 2014). The ideal means of investigating the LSB regime is thus to target it explicitly. This is best done using deep exposures of individual galaxies with a single instrument and under highly controlled conditions; hence, the necessary resource committment is large. It is partly for this reason that galaxy outskirts remain a relatively new territory of study, but what has been observed so far is intriguing. Long trails of stars have been observed loop- ing over the planes of disk galaxies, a likely signature of tidally disrupted dwarf satellites (e.g. Mart´ınez-Delgado et al. 2008). Diffuse clouds of starlight inhabit the spaces between galaxies in groups and clusters, a possible record of collisions and interactions that occurred while these agglomerations were still assembling (e.g. Adami et al. 2005; da Rocha et al. 2008). The outer disks of many galaxies also seem to be dominated by old stellar populations, in apparent defiance of the “inside-out” formation of disks implied by ΛCDM (e.g. Bakos et al. 2008). A full understanding of the processes that shape galaxy outskirts is thus necessary for constructing a correct and working theory of galaxy formation and evolution. This thesis

3 tackles this subject through deep observations of local (D . 10 Mpc) and hence highly spatially resolved galaxies done with the Burrell Schmidt Telescope at Kitt Peak National Observatory (KPNO). These observations make use of both broadband and narrow-band filters to study both the stellar populations found in outer disks and tidal streams and the physical properties of star formation when it occurs in low density environments. This thesis presents case studies of three aspects of galaxy evolution: environmental influence, the influence of secular evolution, and star formation in low density environments. To serve as an introduction to the thesis, a brief overview of each of these topics follows, with emphasis on the connection to outer disks.

1.1 Tidal Interactions and Environment

If galaxies indeed form via hierarchical accretion, tidal interactions and merger events ergo must impact galaxy evolution. Also, because galaxies cluster, their mutual separations are often of the same order as their sizes. This implies that interactions should be frequent, and indeed this is observed; many of the most famous galaxies are interacting pairs, such as M51 (the Whirlpool Galaxy, the ﬁrst “nebula” observed to have spiral structure, a con- troversial observation at the time; Parsons 1850; Steinicke 2012), NGC 4038/4039 (the Antennae Galaxies), and the Tadpole Galaxy. Even our own Milky Way Galaxy may have undergone a past collision with our neighbor M31 (Sawa & Fujimoto 2005; Pawlowski et al. 2012b), and, given M31’s near zero proper motion, the Milky Way seems likely to merge with M31 in the distant future (Sohn et al. 2012). Interactions could have been more frequent in the past, at least once large scale structure began to form, simply because the universe was denser. Indeed, observations of high redshift galaxies show largely unsettled systems composed of large, star-forming clumps (e.g. Conselice & Arnold 2009; Shibuya et al. 2016; Bowler et al. 2017). Given the dissipi- tive nature of gas, however, most of a galaxy’s early merger history is erased by the present era; hence, much eﬀort has been expended both theoretically (e.g. Fall & Efstathiou 1980; Lin & Pringle 1987; Ferguson & Clarke 2001; Debattista et al. 2006; Elmegreen & Struck 2016; Struck & Elmegreen 2017; Herpich et al. 2017) and observationally (e.g. Freeman 1970; van der Kruit & Searle 1981a,b; Courteau et al. 1996; Pohlen & Trujillo 2006; Bakos

4 et al. 2008; Laine et al. 2014) in determining the mechanisms that allow for these initial frequent, violent, and chaotic merger histories to result in the stable, thin, rotation-dominated, yet diverse disk galaxies that dominate the local universe. Some clues to a galaxy’s merger history remain in the form of extended streams, arcs, plumes, and other similar features, whether they compose the stellar halo (e.g. Ibata et al. 2001; Bullock & Johnston 2005; Belokurov et al. 2006; Bell et al. 2008; Cooper et al. 2010; Xue et al. 2011; Ma 2015; Janesh et al. 2016) or are stars stripped from the galaxy’s own outer regions (Mart´ınez-Delgado et al. 2010; Atkinson et al. 2013; Mihos et al. 2013b). In galaxy groups and clusters, the small fraction of tidally stripped stars that leave their host galaxies (Mihos 2004) could also become part of the dominant overall group or cluster environment. Such stars, being unbound to any particular galaxy, are referred to as either intragroup light (IGL), intracluster light (ICL), or, more generically, intrahalo light (IHL). IHL thus serves to connect the initial galaxy-by-galaxy steps of hierarchical assembly to the eventual formation of the high-mass agglomerations that must serve as the endpoint of this assembly. Relating these features to the merger history—to estimate the mass of a tidal stream’s progenitor, for example—is not trivial and relies on comparisons between observations and theoretical models. In an early example, Toomre & Toomre (1972) attempted to reproduce the morphology of M51 via a simple ﬂy-by encounter with its companion. In their model, M51’s symmetric tidal tails emerged as the perturbing companion tugged stars from one side, thus weakening the potential felt by the opposite side: a simple explanation. Yet the exact response of any given interaction is sensitive to the parameters of the interaction, from the relative orientations of both galaxies, to the relative directions of rotation, the relative masses, the relative velocities, presence or absence of gas, and so on (e.g. Wright 1972; Barnes 1988, 1992; Hernquist et al. 1993; Howard et al. 1993; Heyl et al. 1994; Mihos 2004; Bournaud et al. 2005; Springel & Hernquist 2005; Hopkins et al. 2009c; Athanas- soula et al. 2016); hence, attempts to reproduce the morphological features of known interacting pairs can ﬁnd degenerate solutions. Consider the many competing models of M51 that have arisen since the Toomres’ initial attempt (Hernquist 1990; Howard & Byrd 1990; Salo & Laurikainen 2000; Durrell et al. 2003; Dobbs et al. 2010, etc.). Gas also acts in a unique way, as, unlike stars, streams of gas can collide, cool, and so

5 radiate energy away as photons. This has important consequences in the face of rapidly changing gravitational potentials. For example, while simulations often show close morphological correspondance between gas and star particles (e.g. Howard et al. 1993; Duc et al. 2000; Salo & Laurikainen 2000; Horellou & Koribalski 2007; Dobbs et al. 2010), torques and dissipation can create large offsets between the two (Mihos 2001). Such torques may cause the gas to lose angular momentum and fall to the center of the potential, where it could undergo further cooling and result in a central starburst (Barnes & Hernquist 1991; Mihos & Hernquist 1994; Barnes & Hernquist 1996). If a significant amount of gas is present during a merger—even a merger of two equal-mass galaxies, which would almost certainly randomize the orbits of existing stars—the disk may also “survive” the encounter by reforming quickly in the dissipating gas (Hopkins et al. 2009c; Athanassoula et al. 2013; Hopkins et al. 2013; Kannan et al. 2015). And, of course, perturbations may also compress otherwise low density gas, inducing star formation in low density environments like outer disks (e.g. Powell et al. 2013; Moreno et al. 2015). Evolution of tidal features becomes more complex when more than one galaxy is involved. In groups and clusters, stars and gas stripped from galaxies during interactions can be subsequently dispersed over time by additional interactions (Gnedin 2003; Willman et al. 2004; Sommer-Larsen 2006; Purcell et al. 2007; Rudick et al. 2009, 2011). Weak potential wells also result in low velocity dispersions; hence, interactions in groups should be slower and therefore more effective at liberating material than in clusters (Rudick et al. 2009). As groups infall into clusters over time, any stars “pre-processed” in this way can go on to join the cluster’s ICL as the infalling group is disrupted (Fujita 2004; Cortese et al. 2006; McGee et al. 2009; Okamoto et al. 2015). These liberated streams should maintain their coherency until they are disrupted by subsequent interactions within their new cluster environment; clusters still in formation should thus have the most structured ICL (still consisting of coherent streams, arcs, etc.), while ICL in the most evolved clusters should be mostly diffuse and randomized (Rudick et al. 2009). Likewise, in galaxy groups, more and more stars should be liberated over time as interactions take place, with the highly evolved “fossil groups” (groups in which the luminosity is dominated by one large central elliptical; Ponman et al. 1994; Jones et al. 2003) predicted to have the largest IGL fractions (Sommer-Larsen 2006).

6 Given this sensitivity to both the interaction parameters and the environment, it becomes important when attempting to reproduce observed tidal features to include as many observational constraints as possible. Stellar or gaseous kinematics, for example, can provide clearer pictures of the 3D geometry of tidal features (which is rarely evident from the 2D projections one sees in the sky, e.g. Rots et al. 1990; Hibbard & Mihos 1995; Salo & Laurikainen 2000; Durrell et al. 2003; Struck et al. 2005). Kinematics also provide fairly clean separation of ICL and the halos of cluster galaxies (e.g. Longobardi et al. 2015), which is not always possible without such information (Kapferer et al. 2010; Puchwein et al. 2010; Rudick et al. 2011). Constraints on stellar populations from broadband colors or color-magnitude diagrams can also help determine the origins of tidal streams and IHL through comparisons with the stellar populations of the galaxies involved (e.g. Zibetti et al. 2005; Williams et al. 2007; da Rocha et al. 2008; Purcell et al. 2008; Rudick et al. 2010a; Mihos et al. 2017). If one is willing to navigate the increased parameter space, the more observational constraints one includes the tighter the eventual solution can become (Theis & Kohle 2001; Struck et al. 2005; Privon et al. 2013; Mortazavi et al. 2016). Simulations and observations thus go hand-in-hand; without simulations, observations are difficult to interpret; without observations, simulations lack constraints and testable hypotheses. But in the low surface brightness regime, such observational constraints are still lacking. Tidal features are difficult to detect. While they can be bright during the initial stages of the interaction, the stars most tightly bound to the host’s potential reach apocenter earliest and begin to fall back toward the host, stretching out the tail and decreasing the surface brightness over time (Toomre & Toomre 1972; Barnes 1988; Mihos 2004). Disrupting satellites suffer a similar fate as the gravity of their host overcomes their internal bind- ing energy (e.g. Bullock & Johnston 2005; Law et al. 2005). Also, in ΛCDM cosmology, satellites are expected to be distributed isotropically around their hosts (e.g. Springel et al. 2005), hence tidal features should have no preferred location or direction. Detecting any but the most obvious tidal streams thus necessitates not only long integration times, but also wide spatial coverage, either via wide-field unresolved imaging (e.g. Malin & Hadley 1997; Forbes et al. 2003; Mart´ınez-Delgado et al. 2010), or through deep resolved imaging with multiple pointings (e.g. Radburn-Smith et al. 2011b; Ibata et al. 2014; Monachesi et al.

7 2016). Many unresolved imaging studies to date have focused only on detection (Malin & Hadley 1997; Forbes et al. 2003; Mart´ınez-Delgado et al. 2008, 2009, 2010), and therefore contain no quantitative information on the stellar populations that compose the identified features. Resolved imaging studies do constrain stellar populations, but are limited in coverage. Perhaps the most valiant efforts to date in studying galaxy outskirts via resolved populations are the GHOSTS Survey (Radburn-Smith et al. 2011b), and the PAndAS Sur- vey (McConnachie et al. 2009; Ibata et al. 2014). The GHOSTS Survey targeted only small portions of the halos of a handful of galaxies, while the PAndAS Survey achieved complete spatial coverage of only one galaxy (M31, though a survey similar to PAndAS is currently underway to study Centaurus A; Crnojevic´ et al. 2016). Regarding studies of IHL, more progress has been made in clusters than in groups. So far, simulations and observations of ICL present a fairly clean picture of cluster evolution. In simulations, ICL loses its coherency over time, hence can be used to estimate a cluster’s age (e.g. Rudick et al. 2010a). Abell 1914, for example, is likely young; it is currently undergoing a cluster merger, hence is still assembling, and has extremely well-structured ICL (Feldmeier et al. 2004; Jones et al. 2005). The Virgo Cluster appears a bit more evolved, with numerous subclusters, and so contains regions of structured ICL and regions of diffuse ICL (Mihos et al. 2017, and references therein). The Coma Cluster seems nearly virialized and contains abundant, centrally-concentrated, and mostly diffuse ICL (which is bright enough to have been detectable some 66 years ago; Zwicky 1951; Welch & Sastry 1971; Thuan & Kormendy 1977; Gregg & West 1998; Adami et al. 2005), hence implying that Coma is one of the most evolved clusters in the local universe. The evolution of diffuse light in galaxy groups is expected to follow a similar path- way, but to date there has been little observational investigation into IGL formation. Most observational work on IGL has been done in compact groups, which can show wide variations in the diffuse light fraction (da Rocha & Mendes de Oliveira 2005; da Rocha et al. 2008). It has been asserted to explain this that compact groups with a higher IGL fraction are more evolved (Durbala et al. 2008), and this has been supported through studies of H I morphology in compact groups (Verdes-Montenegro et al. 2001), the fraction of early type galaxies within the groups (Goto et al. 2003), and through dynamical arguments (Ribeiro et al. 1998). If so, groups with lower density should contain lower IGL fractions, and old

8 groups in which two or more members have already merged through dynamical friction should contain high IGL fractions (e.g. Sommer-Larsen 2006). Yet observations of IGL in loose groups are rare, and observations of IGL in the oldest fossil groups have yet to be done. To build a more solid picture of IGL and group evolution, more observations are needed of groups of differing densities. In summary, given that the morphologies of tidally stripped material depend on the parameters of the interactions as well as the environment, observational studies of stellar halos, IHL, and tidal streams must be done around as large a sample of galaxies as possible. Because of the difficulty of observing LSB features, both due to the necessary exposure times and the necessary wide spatial coverage, this research is only beginning. This thesis aims to expand on previous research in three ways. First, to conduct a study of IGL in a nearby loose group environment—the Leo I Group—to expand our understanding of IGL and galaxy evolution within the hitherto little explored low density group regime. Sec- ond, to conduct the deepest yet multiband observations of the interacting system M51, to seek out undiscovered tidal features that could solidify the interaction history of this iconic system, and to enhance our understanding of how galaxies evolve under such strong tidal forces by studying the stellar populations of the uncovered tidal features. Third, to conduct deep observations of three additional galaxies spanning a range of local environments, to compare and contrast the accretion histories of galaxies evolving in near isolation to those with multiple large satellites. I will present detailed investigations of interaction histories ranging from a bound group containing several massive members, to a galaxy with no ap- preciably large companions whatsoever, and so build a more solid picture of how the local environment affects the growth and evolution of galaxies.

1.2 Secular Evolution and the Outer Disk

While interactions can have signiﬁcant—even spectacular—impacts on galaxy evolution, even a galaxy evolving in complete isolation will show distinct changes over time due to secular processes. Spiral arms might form spontaneously in completely isolated disk galaxies by the growth of giant molecular clouds (Goldreich & Lynden-Bell 1965; Julian & Toomre 1966; Sellwood & Carlberg 1984; Binney & Lacey 1988) or by perturbations

9 from substructure in the dark matter halo (Gauthier et al. 2006; Bush et al. 2010). These arms might then go on to grow into massive overdensities if the disk is unstable (Safronov 1960; Toomre 1964), serving as the sites of enhanced star formation. Observations of H I in disk galaxies argued that most star-forming spiral galaxies are uniformly on the verge of stability (Quirk 1972; Kennicutt 1989), even at fairly large radius (Zasov & Simakov 1988), hence are highly susceptible to the growth of spiral structure. This might explain why most disk galaxies with a significant gas fraction show spiral structure at some level (including many LSB galaxies, e.g. McGaugh et al. 1995). Understanding disk galaxies in the local universe thus requires understanding not just how they form, but also how they change over time when left to their own devices. While these questions remain topics of some contention (Fall & Efstathiou 1980; Lin & Pringle 1987; van der Kruit 1987; Ferguson & Clarke 2001; Debattista et al. 2006; Elmegreen & Struck 2016; Struck & Elmegreen 2017; Herpich et al. 2017), debates tend to hinge on one commonality: all disk galaxies have exponentially declining radial surface brightness profiles (Patterson 1940; de Vaucouleurs 1959). This is true of a disk’s starlight and its molecular gas, the material out of which stars form (e.g. Wang 1990; Regan et al. 2001; Yim & van der Hulst 2016). Yet the slope of the exponential—how large or small the disk appears—changes with wavelength (e.g. Prieto et al. 2001; Fathi et al. 2010; Laine et al. 2016). Partly this may be due to dust extinction, which preferentially scatters and absorbs blue light, but it may also arise if different stellar populations are distributed across the disk in different ways, which might provide clues to how the disk was constructed. The origin and nature of the exponential profile thus serves as a useful anchor for studies of disk galaxy evolution, as it may be linked both to the disk’s initial formation (Elmegreen & Struck 2016; Struck & Elmegreen 2017; Herpich et al. 2017) and to its subsequent secular evolution (Lin & Pringle 1987; Ferguson & Clarke 2001; Debattista et al. 2006). A simple exponential profile is relatively easy to produce; for example, it can form through the collapse of a uniform sphere into a disk, as long as angular momentum is conserved (Mestel 1963). It thus became a point of great interest when it was discovered that most disk galaxies show a sudden change in the slope of their exponential profile at some radius (Freeman 1970; Pohlen & Trujillo 2006; Erwin et al. 2008). This implied that the low surface brightness outer disk was somehow distinct from the inner disk.

10 The first in-depth studies of this phenomenon were done using edge-on galaxies; in such systems, the change in slope is often extremely sharp, implying a truncation in the starlight that is likely set by the maximum angular momentum of the disk (van der Kruit 1979, 1987; Kregel et al. 2002; Kregel & van der Kruit 2004; Mart´ın-Navarro et al. 2012). Subsequent observations in H I showed that these truncations were often coupled to warps in the gas disk, which likely constitutes more recently accreted material with higher angular momentum (e.g. van der Kruit 1979; van der Kruit & Searle 1981a,b; van der Kruit 2007). Yet studies of face-on systems showed more complex behavior. Freeman (1970) dis- tinguished two types of disk galaxies: those best defined using a single-slope exponential (Type I) and those with a downbending ‘break’ in the exponential profile (Type II), which is always shallower than the truncations seen in edge-on disks (Pohlen et al. 2002). In several studies of SDSS galaxies, Erwin et al. (2005), Pohlen & Trujillo (2006), and Erwin et al. (2008) also found many galaxies with upbending profiles, which they thus dubbed Type III breaks. Because breaks do not indicate as sharp a decrease in stellar surface mass density and because they typically occur at smaller radii than truncations (Pohlen & Trujillo 2006; Mart´ın-Navarro et al. 2012), breaks and truncations are thought to be separate phenomena. In fact, breaks have been found in edge-on systems as well (de Grijs et al. 2001)1, and truncations can be found in addition to breaks in face-on systems (Peters et al. 2017), showing that their different behavior is not simply an artifact of inclination. Thus, if truncations mark the true end of the stellar disk, then the question remains: what are breaks? One important clue is that downbending breaks typically occur in close proximity to overdensities such as rings or the ends of bars (Pohlen & Trujillo 2006), even at infrared wavelengths (which more directly trace stellar mass; Laine et al. 2014). This implies a dynamical origin; for example, breaks might occur via angular momentum redistribution caused by a central bar, which builds up a mass (and therefore a luminosity) overdensity at its outer Lindblad resonance (Pohlen & Trujillo 2006; Debattista et al. 2006; Foyle et al. 2008; Minchev et al. 2012; Laurikainen et al. 2013; Laine et al. 2016). Downbending breaks are also less severe in redder wavelengths (Bakos et al. 2008; Laine et al. 2016), implying that star formation may also decrease sharply at the break radius. Gutierrez´ et al.

1It should be noted that de Grijs et al. (2001) used the term “truncations” throughout their paper; they were later dubbed “breaks” by Mart´ın-Navarro et al. (2012) after the publication of the well-cited studies by Pohlen & Trujillo (2006) and Erwin et al. (2008), which illustrates some of the confusion inherent in the use of these two terms. 11 (2011), for example, found that the fraction of galaxies with downbending breaks correlates with morphological type—star-forming, spiral-arm dominated disks have downbending breaks roughly 80% of the time, compared to 50% for the whole sample. This implies that breaks may also have some connection to a star formation threshold (Kennicutt 1989). Or, in fact, breaks may result from both mass redistribution and a star formation threshold. Roskar et al. (2008b) wrote an influential paper on the subject, implicating a process known as radial migration. By this mechanism, stars can adjust the radii of their orbits via resonant interactions with transient spiral arms without a significant change in orbital ellipticity (Sellwood & Binney 2002; Debattista et al. 2006). Migration can thus populate the outer disk with old, inner disk stars; the break marks the boundary between the most heavily star-forming region of the disk and that containing the migrated old populations. In this scenario, the break radius may push outward over time as gas settles onto the disk (Roskar et al. 2008b,a). Observations provide support for this; for example, Azzollini et al. (2008) found that in galaxies of similar stellar mass, the break radius increases between z = 1 and z = 0. In the local universe, galaxies with downbending breaks also frequently have a U-shaped color profile, with a blueward gradient in the inner disk transitioning to a redward gradient in the outer disk, implying mostly evolved stellar populations at large radius (Bakos et al. 2008; Zheng et al. 2015). Radial migration also provides a solution for a problem discovered in the Milky Way; contrary to expectations from chemical evolution models (e.g. Pagel 1997), stars in the solar neighborhood with similar ages show a huge diversity of metallicities (Edvardsson et al. 1993; Haywood 2006). This can be solved if spiral arms mix populations from different parts of the disk over time (Sellwood & Binney 2002; Schonrich¨ & Binney 2009; Minchev & Famaey 2010; Minchev et al. 2013). Sim- ilarly, radial migration has also been invoked to explain the observed radial trends in the metallicities of stars at all ages in the Milky Way (Loebman et al. 2016). The elegance with which radial migration explains these otherwise vexing observations has given it wide acceptance lately. However, because it is a fairly recent theory (it was first proposed by Sellwood & Binney in 2002), it has yet to be thoroughly tested and verified. For example, if downbending breaks also mark the radial extent of spiral arms (Laine et al. 2014), radial migration may not operate within the outer disk itself, limiting its growth (unless spiral arms can persist at extremely large radius only in the HI, e.g. Koribalski &

12 Lopez-S´ anchez´ 2009; Khoperskov & Bertin 2015). Upbending breaks are also entirely unexplained by radial migration, leading to a wide variety of alternative theories of their formation, from a merging companion depositing extra stars directly into the outer disk (Stewart et al. 2009), to disk stars gaining additional angular momentum from a merging companion (Younger et al. 2007), to locally enhanced star formation (Laine et al. 2016). It has also been proposed that they are not part of the disk at all, but instead mark a transition into a thick disk component or a stellar halo (Pohlen & Trujillo 2006). Additionally, some galaxies do not show breaks of either kind (the Type I disks as named by Freeman 1970). The nearby spirals NGC 300 and NGC 7793, for example, show nearly uniform exponential profiles out to at least 10 disk scale lengths (Bland-Hawthorn et al. 2005; Vlajic´ et al. 2009, 2011). Also, if outer disks are typically more stable against spiral arm formation than inner disks (Kennicutt 1989; Elmegreen & Struck 2016), the influence of perturbing satellites on spiral arm formation in low density environments is worth exploring. However, whether or not the environment influences disk break types at all remains unclear (Pohlen & Trujillo 2006; Erwin et al. 2012; Maltby et al. 2012; Roediger et al. 2012). In summary, the distinctness of outer disks compared to the rest of their host galaxy provides important clues to the means by which secular evolution affects how galaxies evolve over time. While radial migration has so far proven an elegant and useful theory to explain many of the facets of outer disks, it has yet to be thoroughly tested. I therefore contribute to this field through a direct test of the radial migration hypothesis, using deep multiband imaging. I study both the radial and azimuthal surface brightness and color profiles of several nearby disk galaxies down to extremely low surface brightness; in so doing, I measure the true radial extent of the disk beyond the break radius, placing limits on the distances to which stars are able to migrate. I also directly search for the spiral patterns that should be migrating the stars that populate these extended outer disks. Additionally, I incorporate both broadband colors and existing multiwavelength data to seek out evidence of any past or ongoing star formation in these outer disks that might account for the stellar mass without the need for migration. Finally, I study galaxies both with and without companions in order to test how the perturbations originating from the local environment might influence the evolution of the outer disk, for example through the tidal generation of extended spiral arms.

13 1.3 Star Formation At Low Density

While the previous section focused mainly on how disk kinematics may build outer disks through radial migration, stars can and do form directly in outer disks (e.g. Courtes` & Cruvellier 1961; Donas et al. 1981; Ferguson et al. 1998b; Cuillandre et al. 2001; de Blok & Walter 2003; Gil de Paz et al. 2005; Thilker et al. 2005, 2007a; Goddard et al. 2010, and others). These extended star-forming regions have received additional attention in recent years due to observations made using the GALEX satellite (Bianchi & GALEX Team 1999, 2000), which found that many galaxies contain signiﬁcant far ultraviolet (FUV) emission at large radius and low surface brightness (Thilker et al. 2005; Gil de Paz et al. 2005; Boissier et al. 2007; Gil de Paz et al. 2007; Thilker et al. 2007a; Zaritsky & Christlein 2007; Hunter et al. 2010; Lemonias et al. 2011), suggesting that outer disk star formation is more frequent than previously suspected. That said, extended ultraviolet (XUV) disks still only occur in .30% of local universe galaxies (Thilker et al. 2007a; Lemonias et al. 2011). It thus behooves us to understand the conditions under which outer disk star formation can occur. This in turn relies on our understanding the physics of star formation in low density environments, and whether it diﬀers from the high density environments from which our current star formation models, star formation rate conversion factors, and star formation laws are derived. Star formation in any environment requires fuel, which comes in the form of gas.

Schmidt (1959) initially proposed that the star formation rate surface density (ΣSFR) scales

as some power of the gas volume density (hence of the gas surface density, Σgas, which is the observable). While Schmidt (1959) calibrated this law using H I clouds in the Milky Way, Sanduleak (1969) and Hartwick (1971) were the first to expand this to other galaxies (in the Small Magellanic Cloud and M31, respectively). Kennicutt (1989, 1998) later expanded the sample to 97 galaxies and found that their integrated H I column densities and star formation rates (SFRs) followed a power law relation with a slope of ∼1.4, providing strong evidence in favor of the so-called Schmidt Law. Since then, much work has been done to determine the true nature of this law, examining, for example, the form of the law as measured on different physical scales (Kennicutt et al. 2007; Bigiel et al. 2008; Leroy et al. 2012; Schruba et al. 2017), the law as measured using different dense gas tracers or SFR indicators (Gordon et al. 2004; Calzetti et al. 2005; Kennicutt et al. 2007; Bigiel et

14 al. 2008; Usero et al. 2015), or the form of the law in different environments (Bigiel et al. 2008; Wyder et al. 2009; Bigiel et al. 2010b; Daddi et al. 2010; Elmegreen & Hunter 2015; Filho et al. 2016). On that last point, Kennicutt (1989) noted that the Hα surface brightness in most spiral galaxies sharply declines in the outer disk (see also Ferguson et al. 1998b; Martin & Kennicutt 2001). Additionally, outer disk H II regions that are present are typically fainter than those farther in (e.g. Ferguson et al. 1998a; Lelievre` & Roy 2000; Thilker et al. 2005; Goddard et al. 2010; Werk et al. 2010). Kennicutt proposed that this may point to an onset of enhanced disk stability, and derived a theoretical gas density below which star formation should not proceed. This threshold is derived from Toomre’s stability criterion, Q (Toomre 1964), an analog to the spherical Jeans stability criterion (Jeans 1902) but for a thin, rotating disk. In short, a perturbation will grow in a disk if the perturbation’s self-gravity (determined by its density) overcomes the competing pressures of velocity dispersion (determined both by random motions within the gas, and by the noncircularity of orbits, hence relatable to epicyclic frequency) and the coriolis force (determined by the rotation speed). Kennicutt (1989) assumed, with some observational support (e.g. Burton 1971; Stark 1984; van der Kruit & Shostak 1984), that gas velocity dispersion in most disk galaxies is constant with radius, hence the conditions for collapse depend only on the density and the rotation speed. A critical density can thus be derived for any galaxy with a measured rotation curve. This threshold would hold for gas located anywhere in the disk so long as it lay below the local threshold density (Figure 14 of Kennicutt 1989), but it has particular relevance in outer disks, where rotation speeds are constant but the gas density may begin to thin out. This implies that most galaxies may have a star formation threshold radius (Kennicutt 1989; Martin & Kennicutt 2001). Star formation clearly does not cease entirely in outer disks, but given its sparseness and rarity, it may be regulated by different mechanisms than star formation in inner disks. In fact, star formation appears to be less efficient in outer disks (SFR per unit gas mass is lower; Madore et al. 1974; Ferguson et al. 1998b; Lelievre` & Roy 2000). Also, while sites of star formation in outer disks are usually spatially coincident with peaks in the gas density, it is questionable whether or not the Schmidt Law still holds there (Thilker et al. 2007b; Bigiel et al. 2008, 2010b). Where the slope of the Schmidt Law has been measured

15 in outer disks, it is always steeper than in higher density environments (Ferguson et al. 1998b; Lelievre` & Roy 2000; Bigiel et al. 2010a,b). Dwarf irregular (dIrr) galaxies are quite similar in this regard (Hunter et al. 1982; Hunter & Plummer 1996; van Zee et al. 1997; Hunter et al. 1998; Elmegreen & Hunter 2015), as are LSB galaxies (van der Hulst et al. 1993; de Blok & van der Hulst 1998; Boissier et al. 2008; Wyder et al. 2009; Schombert et al. 2011), implying that star formation is regulated by similar mechanisms in all three environments, and that this mechanism may differ from what regulates star formation in inner disks. For example, van Zee et al. (1997) proposed that because dIrrs lack spiral structure (possibly because their thicker disks enhance their stability; Vandervoort 1970; Elmegreen 2011), global dynamics must play little role in regulating star formation. Instead, it might be self-regulated through feedback; after star formation initiates, young stars explode as supernovae, creating local compressions in nearby gas that sets off a new star formation episode, which sets off another, and so on until the fuel is consumed. LSB galaxies, dwarf galaxies, and outer disks might also lack cold molecular gas (de Blok & van der Hulst 1998; Leroy et al. 2005; Bigiel et al. 2008, 2010b), implying that the low in-plane density might lead to low gas pressure as well (Elmegreen 2002); this again might impact star formation efficiency by lengthening cooling times. Yet despite the abundance of evidence for different star formation laws in different environments, it remains an open question whether the cloud-to-cloud physics of star formation differs with environment as well. Absent extenuating circumstances (an irregular star formation history–SFH–dust extinction, a non-universal initial mass function, etc.), all star formation tracers should yield identical SFRs (Kennicutt 1983; Donas et al. 1987; Kenni- cutt 1998), yet Hα and UV in particular frequently are discrepent (Buat 1992; Glazebrook et al. 1999; Sullivan et al. 2000; Bell & Kennicutt 2001; Sullivan et al. 2004; Lee et al. 2009; Hunter et al. 2010). The ratio of Hα to FUV emission is typically lower in outer disks than inner disks (Gil de Paz et al. 2005; Boissier et al. 2007; Goddard et al. 2010; Barnes et al. 2011), potentially implying a different ionizing population (with a relative dearth of high-mass stars; Gil de Paz et al. 2005; Meurer et al. 2009). The same appears to be true of dwarf galaxies, as the ratio of Hα to FUV flux, averaged over whole galaxies, declines with decreasing stellar mass (Bell & Kennicutt 2001; Meurer et al. 2009; Lee et al. 2009; Boselli et al. 2009). LSB galaxies may also typically have a low Hα to FUV

16 ratio (Meurer et al. 2009). Also, compared to higher density environments, these environments appear to systematically lack high-luminosity H II regions (e.g. Youngblood & Hunter 1999; Helmboldt et al. 2005, 2009). That said, Hα emission traces populations a factor of 10 younger than FUV emission (10 Myr vs. 100 Myr timescales). The integrated Hα to FUV ratio is thus sensitive to the galaxy’s SFH; bursty histories can result in an older mean age of young clusters (hence lower Hα/FUV) if they happen to be observed after a burst has concluded (Alberts et al. 2011; Weisz et al. 2012; Barnes et al. 2011, 2013). Perhaps the most convincing argument that trends in Hα/FUV are related to a change in star formation physics comes from Meurer et al. (2009), who showed that the ratio also correlates with R-band surface brightness of galaxies, which, given the contributions to the R-band from stars of a large variety of ages, traces the SFH over gigayears. Yet a later study by Weisz et al. (2012) could not replicate this result. Why Hα and FUV emission imply systematically different star formation rates in different environments thus remains a topic of some contention (for a recent review, see Meurer 2017). I contribute to this field via a statistical analysis of H II regions using new deep Hα nar- rowband imaging and archival deep FUV imaging of the M101 Group. M101 is nearby— hence allows for clear separation between H II regions and surrounding diffuse ionized gas (DIG)—and contains Hα-emitting star-forming regions at extremely large radius and low density, as well as two nearby star-forming dwarf companions. This allows for a direct comparison of populations of H II regions across a range of different environments. I thus seek out systematic differences in the Hα to FUV ratio of H II regions among all of these environments, bypassing the star formation history by using only young regions and so targeting the physics of star formation directly. Additionally, I expand on this by studying how the H II regions affect their local ISM; ionizing radiation leaking from H II regions can go on to ionize the gas exterior to them (Reynolds et al. 1977; Walterbos et al. 1994; Ferguson et al. 1996; Greenawalt et al. 1998; Hoopes et al. 2001; Madsen et al. 2006), hence changes in H II region properties might be reflected by changes in the properties of the surrounding diffuse ionized gas (DIG) as well. Studying both H II regions and DIG thus provides a more complete picture of how the star formation physics might be changing with the local mass surface density.

17 1.4 Summary, and the Aims of This Thesis

Given their connection to galaxy merger histories, environmental effects, secular evolution, and star formation, the low surface brightness outskirts of galaxies are ideal for testing and expanding on our theories of how galaxies form and evolve. This thesis contributes to the field of galaxy evolution through deep observations and analyses of the outskirts of galaxies across a range of wavelengths and environments. The following four chapters describe this research, with each chapter focusing on a separate research project. These are summarized as follows. The first project tests the effectiveness of IGL generation in a loose, low-mass group environment via an analysis of the M96 Galaxy Group (also known as the Leo I Group). Using deep, wide-field broadband imaging, I search for interaction tracers within the group; specifically, I search for a stellar counterpart to the Leo Ring, a 200 kpc diameter H I cloud surrounding the central elliptical M105 that may or may not be tidal in origin itself (Schneider 1989). I also conduct a photometric analysis of the group’s four most massive members: I seek out irregular isophotes in its early-type members that might have resulted from recent mergers or tidal interactions, as well as disturbed isophotes or induced star formation in the outer disks of both spiral galaxies, in an effort to determine whether or not any of the group galaxies took part in the formation of the Leo Ring (e.g. as in the simulation by Michel-Dansac et al. 2010). This study hence provides new details on the evolution of galaxies within an as-yet little studied environment: the low mass, loose group. The second project examines the history of the interacting system M51 via a photometric analysis and multiwavelength study of its extended tidal features. Using broadband colors, accurate even at low surface brightness due to the depth of our imaging, I constrain the stellar populations of the various tidal features I identify (including several that are newly discovered) to determine their likely origins—M51’s inner disk, outer disk, or the companion. These constraints serve to illuminate the interaction history of the system for the benefit of future models. Additionally, I incorporate archival 21 cm imaging to constrain the impact that the interaction has had on the system’s gas content and distribution of star formation, including a search for a stellar counterpart to its extended H I tail (Rots et al. 1990). My study thus provides important new constraints to solidify the interaction history of a benchmark system.

18 The third project explicitly tests theories of outer disk formation—in-situ vs. radial migration—via a detailed multiwavelength study of the outer disks of three nearby galaxies living in different local environments. I first constrain the stellar populations of these outer disks using both broadband colors—measured with low uncertainty at lower surface brightnesses than have yet been achieved in any other unresolved imaging study—and archival FUV and 21 cm imaging to seek out any evidence of recent or ongoing star formation. I measure both radial surface brightness profiles, to assess the physical extent of these outer disks, and azimuthal profiles, to seek out asymmetry indicative of recent perturbations that might have influenced the evolution of outer disk stars and gas. As a direct test of radial migration, I also decompose the azimuthal profiles into Fourier modes as a means of measuring the strength of spiral structure beyond the disk break. In this way, I investigate the origins of outer disks from both the star formation perspective and the radial migration perspective. The fourth and final study tests whether or not star formation proceeds under different underlying physics in low density environments, using a statistical analysis of Hα and FUV emission in HII regions and diffuse ionized gas across the entire disk of M101, as well as across both of its dwarf companions. I compare the median value of and scatter in the Hα/FUV flux ratio of populations of H II regions found in all three environments— inner disk, outer disk, and dwarf galaxies—to search for systematic differences in the star- forming properties of ongoing, hence equal age, star formation. Additionally, to illuminate the interplay between H II regions, O and B star populations outside of H II regions, and the local ISM, I study how the Hα/FUV flux ratio correlates with H II region fluxes and densities across these environments as well. In so making a detailed account of both Hα and FUV emission in both high and low density environments, I further constrain the true origin of observed trends between galaxy mass, surface density, and Hα/FUV. In total, these studies provide a more thorough picture of the stellar populations, star forming properties, and gas content in the low surface brightness regime, from group environments to an ongoing interactions to isolated disk galaxies. This thesis expands on the body of work investigating IGL production, the physics of tidal interactions, and the formation and evolution of both star-forming and non-star-forming outer disks, in the process raising many important questions to be addressed by future such studies.

19 Chapter 2

Searching For Diﬀuse Light in the M96 Galaxy Group

2.1 Abstract

We present deep, wide-ﬁeld imaging of the M96 galaxy group (also known as the Leo

I Group). Down to surface brightness limits of µB = 30.1 and µV = 29.5, we find no diffuse, large-scale optical counterpart to the “Leo Ring,” an extended H I ring surrounding the central elliptical M105 (NGC 3379). However, we do find a number of extremely low surface brightness (µB & 29) small-scale streamlike features, possibly tidal in origin, two of which may be associated with the Ring. In addition we present detailed surface photometry of each of the group’s most massive members—M105, NGC 3384, M96 (NGC 3368), and M95 (NGC 3351)—out to large radius and low surface brightness, where we search for signatures of interaction and accretion events. We find that the outer isophotes of both M105 and M95 appear almost completely undisturbed, in contrast to NGC 3384 which shows a system of diffuse shells indicative of a recent minor merger. We also find photometric evidence that M96 is accreting gas from the H I ring, in agreement with H I data. In general, however, interaction signatures in the M96 Group are extremely subtle for a group environment, and provide some tension with interaction scenarios for the formation of the Leo H I Ring. The lack of a significant component of diffuse intragroup starlight in the M96 Group is consistent with its status as a loose galaxy group in which encounters are

20 relatively mild and infrequent.

21 2.2 Introduction

In the current ΛCDM cosmological paradigm, massive galaxies form hierarchically over time via the continual accretion of smaller objects (Searle & Zinn 1978; Davis et al. 1985; Frenk et al 1988; White & Frenk 1991; Jenkins et al. 2001; Springel et al. 2005). Massive galaxies may interact with one another as well, leaving behind readily observable signatures in the form of tidal streams or drastic changes in morphology, stellar populations, and kinematics (e.g., Arp 1966; Toomre & Toomre 1972; Toomre 1977; Hernquist & Weil 1992; Barnes 1992, 2002). Even after the initial construction of galaxies has ended, however, cosmological simulations indicate that there should remain a wealth of low-mass satellites that continue to interact with the parent galaxy over time (e.g., Frenk et al 1988; Gelb & Bertschinger 1994; Springel et al. 2005, 2008). Observations of the Local Group indeed show a number of low-luminosity satellites around the Milky Way and M31 (Hodge 1971; Feitzinger & Galinski 1985; Mateo 1998; Belokurov et al. 2006; Weisz et al. 2011; McConnachie 2012), lending credence to the idea. However, important discrepancies between theory and observation remain, such as the “missing satellites problem” (Klypin et al. 1999; Moore et al. 1999), demonstrating the importance of continuing to test the theory observationally in a variety of ways. Probing the low surface brightness (LSB) outskirts of galaxies via deep surface photometry provides one such observational test. Under proposed “inside-out” galaxy formation models (Matteucci & François 1989; Bullock & Johnston 2005; Naab & Ostriker 2006; Hopkins et al. 2009a; van Dokkum et al. 2010; Kauffmann et al. 2012), even apparently undisturbed galaxies should show signs of recent accretions or interactions in subtle ways in their outskirts, where restoring forces are low and dynamical times are long. Morpho- logical signatures of accretion such as tidal features, warps, and disk asymmetries (Toomre & Toomre 1972; Bullock & Johnston 2005; Mart´ınez-Delgado et al. 2010) may trace the record of past encounters, while changes in the structural properties and stellar populations of the outer disk can be probed via quantitative photometry and color profiles (e.g. van der Kruit 1979; Pohlen et al. 2002; Trujillo et al. 2009; Mihos et al. 2013b). Finding and cate- gorizing such signatures thus becomes an important step in constraining theories of galaxy formation. Unfortunately, the extended outskirts of galaxies are extremely dim and diffuse. How-

22 ever, with the advent of deep wide-field imaging techniques, over the past decade a wide variety of faint tidal features have been identified around nearby galaxies. These features span a wide range of morphologies, from loops, plumes, and shells (Forbes et al. 2003; Mart´ınez-Delgado et al. 2008, 2009, 2010; Atkinson et al. 2013; Mihos et al. 2013b) to extended streams, diffuse stellar halos, and large-scale intracluster light (ICL; Uson et al. 1991; Scheick & Kuhn 1994; Gonzalez et al. 2000; Feldmeier et al. 2004; Adami et al. 2005; Mihos et al. 2005; Rudick 2010). Gaseous accretion events also appear to influence the properties of disks (Sancisi et al. 2008), fueling new star formation in their low-density outskirts and driving the formation of extended ultraviolet disks (Thilker et al. 2005; Gil de Paz et al. 2005; Thilker et al. 2007a; Lemonias et al. 2011). This extended star formation will in turn affect the age and metallicity distributions, and hence the colors, of the underlying stellar populations in these regions. Meanwhile, internal processes within disks may scatter stars outward, again resulting in changes in the structure and stellar content of the outer disk (Sellwood & Binney 2002; Debattista et al. 2006; Roskar et al. 2008b,a). Local environment also must play a role in the accretion history of galaxies and the evolution of their outer regions. In massive clusters, for example, frequent and rapid interactions between galaxies tend to produce plumes and streams of starlight, which are shredded over time by interactions within the cluster potential to form the more diffuse ICL seen in clusters such as Virgo (Rudick et al. 2009). In group environments, where the velocity dispersions are generally lower, slow encounters lead to strong dynamical friction and rapid merging (Hickson et al. 1977; Barnes 1989; Taranu et al. 2013). Contrary to cluster environments, then, galaxies in groups may exhibit long-lived accretion shells or tidal tails in their outskirts (e.g., Quinn 1984; Hernquist & Quinn 1987; Law et al. 2005). Meanwhile, the low-density environment around isolated field galaxies should lead to fewer recent interactions and accentuate secular evolution processes that are internal to galaxies (e.g. Kraljic et al. 2012). A complete survey of the processes that shape galaxies thus requires an investigation across all environments. Of particular interest is the group environment, the most common environment for galaxies to reside. Intragroup starlight (IGL) in loose groups1 is of significant interest

1Because of their irregular nature, a precise deﬁnition of a “loose group” is elusive. However, characteristic properties are often adopted to classify them, such as a small number of luminous galaxies within a 0.5 Mpc radius and a velocity spread of ∼350 km s−1 (see, e.g., de Vaucouleurs 1975; Geller & Huchra 1983;

23 across many studies of galaxy and galaxy cluster evolution. The fraction of baryons within galaxy groups and clusters constrains models of star formation and cluster assembly, and there is strong evidence that the amount of intragroup and intracluster starlight may be a significant contributor to the baryon fraction of these systems (e.g. Gonzalez et al. 2013; Budzynski et al. 2014). The loose group environment may also dynamically “preprocess” galaxies that later fall into galaxy clusters and produce ICL (Mihos 2004), and substructure within the diffuse IGL should trace this process in dynamically evolving galaxy groups. However, the mean fraction of stars in an IGL component, and how that fraction varies with group/cluster mass, is substantially uncertain. Theoretical studies give conflicting results: some studies have found relatively constant mean IGL/ICL fractions with group/cluster mass, but with large intrinsic scatter (e.g. Sommer-Larsen et al. 2005; Sommer-Larsen 2006; Monaco et al. 2006; Henriques & Thomas 2010; Puchwein et al. 2010; Guo et al. 2011; Contini et al. 2014). Other theoretical studies have found that the mean IGL/ICL fraction (sometimes generalized as the intrahalo stellar fraction; IHL) systematically increases with group/cluster mass (e.g. Murante et al. 2007; Purcell et al. 2008; Watson et al. 2012). Observationally, wide-field imaging studies of 0.01 < z < 0.4 galaxy clusters over a large range of cluster masses have also found different behaviors for the IGL/ICL fraction as a function of cluster mass (e.g. Lin & Mohr 2004; Zibetti et al. 2005; Gonzalez et al. 2007, 2013; Budzynski et al. 2014). Studies of nearby individual loose galaxy groups searching for individual intragroup stars also have mixed results. Observational searches for intragroup red giant stars and intragroup planetary nebulae have generally found very small fractions of IGL (< 2%; e.g. Castro-Rodr´ıguez et al. 2003; Feldmeier et al. 2004; Durrell et al. 2004). Usually, these surveys only observe a portion of the galaxy group, and if the IGL has an uneven spatial distribution (e.g. Sommer-Larsen 2006), the IGL fractions found may not be reflective of the entire group. However, in stark contrast, McGee & +16 Balogh (2010) have estimated that 47%−15 of the stellar mass in galaxy groups is in the IGL component through detections of spectroscopically confirmed hostless intragroup SNe Ia that were within the Sloan Digital Sky Survey (SDSS; York et al. 2000) supernova survey. To improve this situation, more searches for IGL need to be undertaken in galaxy groups

Maia et al. 1998). Leo I has been classiﬁed as a group using these criteria by de Vaucouleurs (1975), Huchra & Geller (1982), and Tully (1987).

24 Figure 2.1: False-color image of the M96 Group, made from a composite of our B and V mosaics. The four largest group members are labeled; the small spiral galaxy south of NGC 3384 is NGC 3389 and is most likely part of a separate group called Leo II, roughly 10 Mpc behind the M96 Group (Stierwalt et al. 2009, and references therein). H I contours from Schneider (1989) are overlaid in white. The image is 1◦.8 × 1◦.8 (350 × 350 kpc2). North is up and east is to the left.

25 over a large range of masses. The M96 Group (sometimes referred to as the Leo I Group), shown in Figure 2.1, helps serve this purpose well. It is the nearest loose group2 that contains both early- and late-type massive galaxies (de Vaucouleurs 1975), and hosts a rather unusual feature—a roughly 200 kpc diameter broken ring of neutral hydrogen, first discovered by Schneider et al. (1983). This ring is fairly diffuse; from Arecibo data, column densities range from 6.4 × 1019 cm−2 to as low as 2 × 1018 cm−2 (Schneider et al. 1989), although smaller clumps with densities of roughly 4×1020 cm−2 were also detected using the Very Large Array (Schneider et al. 1986). These higher density clumps appear to be con- fined to the southernmost portion of the ring, and include a bridge-like feature apparently connecting the ring to M96 (NGC 3368), the most massive spiral galaxy in the group. Two other galaxies—the E1 galaxy M105 (NGC 3379) and the S0 galaxy NGC 3384—appear to lie near the center of the ring, implying some relation between these three galaxies and the intergalactic H I. A fourth galaxy, the barred spiral M95 (NGC 3351), is associated with the group as well, but due to its relatively large projected separation, it may not be associated with the ring complex itself. The group also lies near in the sky (∼8◦ away) and at a somewhat similar distance to the M66 Group, leading to speculation that the two groups are actually each part of a larger complex of galaxies (Stierwalt et al. 2009). The origin of the Leo H I Ring is still under debate. Kinematics gleaned from the H I observations show that it appears to be well-modeled by an elliptical orbit with a focus at the luminosity-weighted centroid of NGC 3384 and M105 (Schneider 1985). The rotational period of this model orbit is approximately 4 Gyr, and if the Ring is primordial, this period presents a rough lower limit on its age (see Schneider 1985). The apparent lack of associated starlight and star formation supports this primordial origin hypothesis (Pierce & Tully 1985; Schneider et al. 1989; Donahue et al. 1995), as does the similarity between the radial velocities of the Ring and the galaxies within the group, both of which trend higher in the northwest and lower in the southeast (Schneider 1985). Additionally, Sil’chenko et al. (2003) argue that circumnuclear stellar and gaseous disks in NGC 3384, M96, and M105 show ages and kinematics consistent with accretion from the Ring taking place on gigayear timescales, which may rule out a more recent formation scenario. Of course, crossing times in the group are much shorter than 4 Gyr (Pierce & Tully 1985),

2We adopt a distance of 11 Mpc for the M96 Group (Graham et al. 1997); at this distance, 10 corresponds to 3.2 kpc.

26 which poses problems for the stability of the ring—maintaining its shape would require a shepherding mechanism, presumably by M96 (Schneider 1985)—and the only galaxy in the group whose disk appears aligned with the Ring is NGC 3384, which is unusual if all of the galaxies formed from a single rotating cloud. An alternative hypothesis was proposed by Rood & Williams (1984), involving the collision of two spiral galaxies. Under this model, the two galaxies collided nearly head- on, stripping most of the gas from one of the galaxies and producing an expanding density wave much like that proposed to explain the morphology of ring galaxies (e.g., Lynds & Toomre 1976; Gerber et al. 1992; Mazzei et al. 1995; Berentzen et al. 2003). A more recent simulation by Michel-Dansac et al. (2010) managed to reproduce qualitatively all of the important features of the Ring (the annular shape, the apparent rotation, the bridge- like feature described above, and the lack of an apparent visible light counterpart). The simulation collided two gaseous disk galaxies, one of which transformed into an S0: a good facsimile of M96 and NGC 3384. In yet another scenario, the total disruption of a gas-rich, LSB galaxy would also leave behind an H I ring with the LSB galaxy’s stellar disk completely unaffected (Bekki et al. 2005). In general, the main problem to solve in a collisional hypothesis is to reproduce the apparent lack of optical light associated with the ring, as slow encounters in the loose group environment should be effective at liberating stellar material from the galaxies as well. To date, neither the primordial origin nor the collisional hypothesis for forming the Leo Ring has been confirmed. No extended diffuse starlight has been found associated with the H I ring; the only known stellar counterpart to the Ring is that discovered by Thilker et al. (2009) in the form of three small far- and near-ultraviolet sources located within the ring’s highest density H I complexes (as measured by Schneider et al. 1986). An optical counterpart to at least one of these clumps was later found by Michel-Dansac et al. (2010), with a measured color of g0 − r0 = 0.2, indicative again of a young stellar population. The presence of these clumps thus raises the intriguing possibility that these stars represent the youngest (and therefore brightest) component of an underlying (and as-yet undetected) diffuse stellar counterpart to the ring. All of this makes the M96 Group a particularly intriguing target in a search for the LSB signatures of historical encounters. This paper presents deep surface photometry of the M96 Group, conducted using Case

27 Western Reserve University’s Burrell Schmidt telescope on Kitt Peak and taken as part of an ongoing project to study the outskirts of nearby galaxies across a range of environments. Our deep imaging, covering approximately nine square degrees in two ﬁlters, probes the diﬀuse outskirts of the group members and the Leo Ring down to a surface brightness of

µV ∼ 29.5, with reliable B − V colors down to µV ∼ 28.0. We use the data to search for diﬀuse light in the Leo Ring that may trace any past interaction events, as well as to study the structure and stellar populations in the outer disks and halos of the group galaxies.

2.3 Observations and Data Reduction

The full description of our observational and data reduction techniques can be found in Rudick et al. (2010a) and Mihos et al. (2013b), however, for the sake of clarity, we describe the most important details here, along with any notable changes to the procedure.

2.3.1 Observations

We observed the M96 Group using the 0.6/0.9m CWRU Burrell Schmidt telescope located at Kitt Peak National Observatory. The telescope has a 1◦.65×1◦.65 field of view and images onto a 4096×4096 STA0500A CCD3 for a pixel scale of 100. 45 pixel−1. We observed in two filters: a modified Johnson B (spring 2012) and Washington M (spring 2013). We used Washington M as a substitute for Johnson V, as it covers a similar bandpass but avoids night sky airglow emission lines (Feldmeier et al. 2002); thus, as Washington M is slightly bluer than Johnson V, we used a modified B filter that is 200 Å bluer than the standard Johnson B in order to maintain a comparable spectral baseline. For the actual analysis, we converted our magnitudes to standard Johnson B and V using observations of Landolt (1992) standard fields. We took data only on moonless photometric nights, although it should be noted that Mars was less than 3◦ from our target fields during part of the spring 2012 run and hence may have contributed some low-level scattered light. Along with target pointings, we also observed offset sky pointings to construct a night sky flat. To minimize uncertainties in the final flat due to changes in sky conditions and telescope flexure, we kept the telescope oriented at roughly the same pointing for flat-field and

3The CCD was backside processed at the University of Arizona Imaging Technology Laboratory.

28 object exposures by alternating between the two. We dithered individual target exposures randomly by up to half a degree to increase the imaging area and reduce artifacts due to scattered light and large-scale flat-fielding errors. Exposure times were 1200 s in B (resulting in sky levels of 700–900 ADU pixel−1) and 900 s in M (1200–1400 ADU pixel−1). Our final B and V mosaics contain 41 exposures each, with roughly 30 accompanying blank sky frames per band taken between 0.25–1 hr in right ascension and 1◦–4◦in declination from the M96 Group. However, we note that the final flat field was produced using blank skies near every object we imaged during the course of this project, for a total of ∼100 frames in B and ∼120 in V (discussed in more detail below). Our total on-target exposure time in the M96 Group is 13.7 hr in B and 10.25 hr in V, although the time spent in each part of the mosaic varies due to dithering across the group’s large angular size (∼ 3◦ × 3◦).

2.3.2 Data Reduction

For each frame, we applied an overscan correction and subtracted a nightly mean bias frame using IRAF4. We applied no crosstalk corrections; close examination of the images revealed that any visible crosstalk was at low enough levels that such a correction was unnecessary. We also corrected each year’s data for nonlinear chip response in each quad- rant of the CCD using a fourth-order polynomial correction. Finally, we applied a world coordinate system to each image to construct the final mosaic. After a preliminary flat-field correction, we derived a photometric zeropoint for each target frame (directly correcting for airmass effects) using SDSS stars as standards, converted from ugriz to Johnson B and V using the conversion formula given by Lupton (2005). We excluded stars outside of the color range B−V = 0 to 1.5 from the fit (Ivezic´ et al. 2007). To derive each filter’s color term, we took exposures of Landolt standard star fields (Landolt 1992) at varying exposure times at the beginning and end of each night, as using Landolt stars avoids any error inherent in a color conversion formula. From our final mosaics, we

were able to recover converted SDSS magnitudes to σV = 0.03 mag and B − V colors to

σB−V = 0.05, using SDSS stars in the magnitude range V = 15–17 and the color range B − V = 0.0–1.5. 4IRAF is distributed by the National Optical Astronomy Observatory, which is operated by the Associa- tion of Universities for Research in Astronomy (AURA), Inc., under cooperative agreement with the National Science Foundation.

29 Table 2.1. Properties of Streamlike Features in the M96 Group

Dimens Dimens 6 Region µB µV MB MV B − V (Range) µB,lim µV,lim 10 L (0) (kpc) ,B

A 1.3 × 9.4 4 × 30 29.3±0.3 28.8±0.3 -12.2 -12.6 0.45(-0.14 – 1.04) 30.9 30.5 12 B 1.1 × 5.2 3.5 × 17 29.2±0.2 28.6±0.3 -11.6 -12.1 0.56(-0.08 – 1.15) 30.9 29.9 7 C North 1.1 × 9.6 3.5 × 31 29.8±1.0 29.6±0.8 -11.7 -11.9 0.21(-1.60 – 2.55) 30.1 29.7 7 C South 1.3 × 7.0 6 × 22 29.9±1.0 29.1±0.8 -11.4 -12.1 0.77(-1.04 – 3.11) 30.1 29.7 6 C Total 1.2 × 60.1 3.5 × 200 29.9±1.0 29.4±0.8 -13.9 -14.4 0.49(-1.32 – 2.83) 30.1 29.7 56a

Note. — aAssumes a continuous structure from C North to C South: see text.

After applying zeropoints and assuming that the night sky has an average B − V color

near unity (Krisciunas 1997; Patat 2003), we measured sky brightnesses from µB = 22.32

to 21.89 and from µV = 21.72 to 21.43. These values changed throughout the course of a night, as well as from night to night, due to changes in solar output, lunar phase, and atmospheric conditions (e.g., dust, humidity, etc.). However, the faintest sky brightness measurement can act as a useful secondary check by comparing our measured sky brightnesses to previously measured values under similar conditions. To measure the actual sky brightness, we ﬁrst removed our airmass corrections of 0.27 mag and 0.16 mag in B and V, respectively, since the source of the sky brightness is mostly within the Earth’s atmosphere.

We then compared our faintest values (µV = 21.88, µB = 22.59) to the empirical relation of Krisciunas (1997, see their Equation (5)), which correlates the average observed night sky brightness to the 10.7cm solar ﬂux. Taking the monthly 10.7cm values from the Do- minion Radio Astrophysical Observatory,5 we ﬁnd an adjusted expected sky brightness of

µV = 21.75 ± 0.06 and µB = 22.65 ± 0.06. Given the uncertainties in this calculation, this shows that our overall results are stable and robust. Accurate flat-fielding is paramount for reliable surface photometry of LSB features. As mentioned previously, we used blank sky fields near our target objects to construct a master night sky flat. For each sky frame, we masked stars and background objects by running IRAF’s objmask task twice—once to generate a basic mask, then again with the first mask applied to mask even fainter sources; both times we masked objects using a threshold of 4σ above background and 5σ below (more details on the masking procedure

5The data are located at http://www.spaceweather.ca/solarﬂux/sx-eng.php

30 can be found in Rudick et al. 2010a). We then calculated the median values in 32×32 pixel blocks, and binned the images using these median values in order to enhance any faint, extended contaminants that were missed by objmask, such as flares from stars off the edge of the chip, extended wings of stars, and internal reflections. These features we masked by hand, and images with obvious large-scale contamination (faint cirrus clouds, unusual electronic noise) were discarded. We median combined all of these masked sky images to create a preliminary sky flat, re-flattened the sky frames using this flat, and modeled and subtracted sky planes from each flattened image. We used these flattened, sky-subtracted images to create a new flat, which we then used as a basis to repeat the process. The flat field converged after five such iterations, resulting in the final generation master sky flat. To test for systematics in the flat-fielding, we used this process to construct three types of master flats: master flats by object (all objects imaged during the course of this project, including targets other than the M96 Group), master flats by observing run, and a master flat made from all skies taken over the course of an observing season. We divided all pairs of these master flats to check for systematic differences, and we determined that there only existed a low-level (0.2%) residual plane that varied from run to run. As such, we created flat fields for each observing run using the master flat generated from all sky images, corrected for each run’s residual plane. In this way, our final B flat combined 98 total sky images, and our final V flat combined 123 total sky images. We note that the consistency of the flats (including even those taken nearer to Mars than the M96 Group) shows that scattered light from Mars was not a problem in constructing the flat field. Bright stars can also introduce extended halos of light due to reflections between the CCD, dewar window, and filter; we modeled and removed these using the process described in Slater et al. (2009), for which we give a brief summary here. To generate visible reflections, we took long (1200 s and 900 s in B and M, respectively) exposures of zeroth magnitude stars (Procyon in B and Regulus in M). We modeled these reflections as annuli of constant brightnesses scaled by the total brightness of the parent star. We also measured and modeled the offset of each reflection from the parent star as a function of the star’s position on the CCD. Lastly, we modeled the star’s point-spread function out to where the

flux from the star drops to 0.3 ADU (µB ∼ 31). We then subtracted these models from each bright star in each image. Bright stars also produce visible off-axis reflections colloquially

31 known as “Schmidt Ghosts” (Yang et al. 2002); we masked these as well for stars brighter than V = 10.5. To obtain accurate photometry for the bright stars producing these various reﬂections, we used short exposures of our target ﬁelds; for the brightest stars we used magnitudes given in the Tycho II catalog (Høg et al. 2000), and in a few cases (for stars

with mB . 8) we had to manually adjust the magnitudes to achieve the best subtraction. We then masked and median binned the images into 32 × 32 pixel blocks and hand- masked any scattered light patterns in each object frame in the same manner as the blank skies, as well as any persistent bright objects in each frame. We fitted a sky plane to each of these binned, masked images, and subtracted this sky from the original images. Finally, we mosaicked all of the resulting images using IRAF’s Wregister and imcombine tasks, scaling each image to a common zeropoint, and created 9 × 9 pixel and 18 × 18 pixel masked and median binned images to help identify faint features and improve the per pixel statistics in the final mosaics.6 The masking process used to generate these images typically only masks pixels brighter than µB ≈ 27 and µV ≈ 26, so that the faint, diffuse signals we are searching for remain visible.

2.4 Large Scale Diﬀuse Light

Figure 2.2 shows our final 18 × 18 pixel (2600× 2600, or 1.4 × 1.4 kpc2) binned B and V mosaics. This binning choice is simply to enhance the clarity of the features of interest; all analyses of the images used either the unbinned image or a 9 × 9 binned image. A schematic is provided in Figure 2.4 to guide the eye. The first thing to note in these images is that we see no large-scale diffuse stellar counterpart to the H I ring. While there is a band of diffuse light running across the image in the B mosaic, the same feature does not appear in the V mosaic, which shows a different pattern of diffuse light elsewhere in the image. Permanent features such as starlight or foreground Galactic dust lanes (“cirrus”) should remain fixed in position and morphology from image to image, whereas scattered light patterns will change with respect to the position of the source. As noted in Section 2.3.1, Mars was less than 3◦ from our fields in spring

6For the sake of eﬃciency, for the remainder of the paper we will refer to the process of masking and calculating the median in blocks as simply “binning”. As such, references to, e.g., the “9 × 9 binned image” refer to the image that has been masked and then median binned into 9 × 9 pixel blocks.

32 Figure 2.2: Our ﬁnal B mosaic (left) and V mosaic (right), with pixels binned 18 × 18 to improve the clarity of the features of interest. The three regions discussed in the text are labeled A, B, and C; due to the faintness of region C, its general position is indicated only by an arrow. C North and C South indicate where the photometry for the feature was performed, with the brackets showing the extent of the regions sampled. M105 and M96 are labeled for reference. In both images, the dimensions are 1◦.9 × 1◦.9 (360 × 360 kpc2), and north is up and east is to the left. Boxes are for illustration purposes only. The contrast is such that pixels saturate below µ ∼ 28.5. Images are masked at high (µ . 24) surface brightness.

33 Figure 2.3: Our ﬁnal V mosaic, binned 18 × 18 as in Figure 2.2, with H I contours from Schneider (1989) and Michel-Dansac et al. (2010) overlaid in yellow and red, respectively. The three regions from Figure 2.2 are labeled for reference, as is the dwarf galaxy KK96 (see the text). The inset image is the unbinned image of KK96, zoomed in for clarity, to show the tidal tails more clearly. Again, the image is 1◦.9 × 1◦.9 (360 × 360 kpc2) with north up and east to the left, and boxes are for illustration purposes only.

34 of 2012 when we observed in the B filter, and hence may have contributed to the scattered light in the final B mosaic. It is less clear where scattered light may have originated in the V mosaic, with the brightest nearby source (Regulus) being ∼ 9◦ to the west. Regardless, as the only diffuse signal we see appears to be scattered light, we find no evidence of actual

diffuse starlight in the Leo H I ring. We place upper limits of µB < 30.1 and µV < 29.5 on the surface brightness of any such component, as these are the average surface brightnesses of the scattered light signals found in the southern portion of the Ring (where the bulk of the H I mass is found). We note that this is consistent with an estimate by Castro-Rodr´ıguez et al. (2003) of µB < 32.8 based on the apparent lack of planetary nebulae associated with the cloud. Despite this lack of large-scale diffuse light in the Ring, we do find three smaller features, which are labeled in Figures 2.2 and 2.3 and shown schematically in Figure 2.4: a streamlike object north of NGC 3384 (A), a shorter streamlike object northeast of M95 (B), and a nearly vertical streamlike object west of NGC 3384 and the background galaxy NGC 3389 (C, most clearly visible in the V-band image). While these features are consistent in morphology and position in both mosaics, they are extremely faint, so to further test their consistency we created three sets of new B and V mosaics using randomly selected half- samples of the object frames used to build the final mosaics in each band. All three objects persisted even in these half-split mosaics. We note also that all three features are visible in a preliminary data set taken in 2010 in the M band, so the three features appear to be robust. Other structures are visible in Figures 2.2 and 2.3, such as a V-shaped structure on the images’ west sides, but these were not consistent in morphology among all half-splits and so were not considered for analysis. Finally, as scattered light from galactic dust (the so-called “Galactic cirrus”) can often show similar diffuse morphology, we cross-checked our mosaics against the IRIS 100 µm map (Miville-Deschenesˆ & Lagache 2005) for the region. As cool dust emits thermally in the far-infrared, any contamination from galactic cirrus should show corresponding emission in the 100 µm map. However, this map shows that the M96 Group lies in a region with relatively little such foreground pollution, and shows no spatially corresponding linear structures. The Planck all-sky thermal dust emission map (?), though it has lower resolution, confirms this as well. This lack of detection in the fariinfrared makes it unlikely that

35 Figure 2.4: Schematic of the three features in relation to the group galaxies and the H I ring, for clarity. these features are merely foreground dust, and thus are likely to be tidal structures located within the M96 Group. In addition to the linear streams, we also detect diﬀuse starlight around the galaxy KK96 (located southwest of NGC 3384). This galaxy, classiﬁed as an LSB dwarf spheroidal (dSph) by Karachentseva & Karachentsev (1998) and Karachentsev et al. (2013), displays clear tidal tails toward the east and west in our images (clearly visible even in the unbinned images; see Figure 2.3), indicating a recent interaction. Due to a lack of distance constraints for this galaxy, its relation to the M96 Group is unclear. However, aside from the M96 Group, it appears to have no obvious nearby neighbors that it may have interacted with. We can attempt to constrain KK96’s location to some degree through its optical prop-

36 erties. The galaxy has an apparent magnitude of V = 16.9, including all light out to the Holmberg radius (0.06; see Karachentsev et al. 2013) and excluding the faint tails. We also 0 measure an effective surface brightness (re = 0.3) of µe,V = 25.8 and central surface brightness of µV ≈ 23. If placed at the distance of the M96 Group, the galaxy would have an absolute magnitude of MV = −13.3, similar to the Fornax dwarf (van den Bergh 2008), and an effective radius of 960 pc, consistent with the empirical MV –µe,V and MV –re rela- tions for Sph-type galaxies from (Kormendy & Bender 2012). If the system is significantly closer, for example within the Local Group at 1 Mpc, the galaxy would be underluminous by ≈ two magnitudes for its surface brightness, given the Kormendy & Bender relation. However, given the scatter in the dSph scaling relationships, the system is also consistent with being a member of the background Leo II galaxy group (at a distance of ∼20 Mpc; see Stierwalt et al. 2009). So while it is plausible that the system is being tidally stripped by its interactions within the M96 Group, without proper distance constraints, it is difficult to state this unequivocally. The properties of the three linear features labeled A, B, and C, including local limiting surface brightnesses and color constraints, are summarized in Table 2.1. Absolute magnitudes and luminosities are calculated assuming a distance of 11 Mpc, and are similar to moderately sized dwarf galaxies, although we point out that we see no actual dwarf galaxies associated with the features themselves. The wide range in the estimated colors is due to the low light levels of the three features; a large uncertainty in surface brightness corresponds to a much larger uncertainty in associated color. As such, we are unable to make any definitive statements regarding the stellar populations in these features. We perform photometry on the three features using apertures that include all of their visible light, subtracting off a locally measured background level, which produces more accurate photometric results (see Rudick et al. 2010a). We measure the background flux by averaging the counts in several (four to six, depending on the proximity of obvious contaminating background or foreground sources) comparably sized blank sky regions surrounding the signal region, and use the dispersion in the median values of the blank regions as our measure of background uncertainty (for examples, see Rudick et al. 2010a; Mihos et al. 2013b). This dispersion in ADU is then converted to a local limiting surface brightness. To determine the likelihood that these features are in any way associated with the H I

37 ring, we overlay H I contours from Schneider (1989) and Michel-Dansac et al. (2010) on our binned V mosaic for comparison. This is shown in Figure 2.3. Interestingly, the northernmost stream, labeled A, roughly follows the northwest arc of the H I ring. The feature may extend farther to the southwest, but is obscured by a foreground star. Assuming an area of 1.03 × 9.04 (4 × 30 kpc2),7 the total flux from the feature yields an absolute B −2 magnitude of MB = −12.2, for a surface brightness of µB = 29.3 ± 0.3 mag arcsec . This 7 corresponds to a luminosity of roughly 1.2 × 10 L ,B. Stream B lies near M95, well away from the H I ring. Given this apparent separation, it may not be associated with the Ring complex at all. Its linear morphology and lack of any central nucleus does imply that it has been tidally disrupted, and so it may be an object similar to KK96 but lower in mass, possibly a satellite of M95 given its apparent proximity 0 0 2 to the galaxy. Assuming an area of 1.1 × 5.2 (3.5 × 17 kpc ), we find MB = −11.6 and 6 µB = 29.2 ± 0.2, for a total luminosity of 7 × 10 L ,B. Finally, Stream C appears to be one long linear feature, extending from just below feature A southward almost 100 before disappearing below NGC 3389. This feature is unlikely to be an instrumental artifact, as it does not align along the rows or columns of the CCD. It lies near a similar vertical linear feature in the H I, although it is displaced to the east and tilted in position angle by ∼ 1◦. The northernmost cloud of this H I feature appears to be associated with NGC 3384 in both proximity and velocity, however the association of the two small clouds just to the south is more ambiguous (Oosterloo et al. 2010). For the northern end of the optical feature (labeled C North), sampling from an area of 1.01 × 9.06 2 (3.5 × 31 kpc ), we find MB = −11.7 and µB = 29.8 ± 1.0. For the southern end (labeled 0 0 2 C South), an area of 1.3 × 7.0 (4 × 22 kpc ) yields MB = −11.4 and µB = 29.9 ± 1.0. Uncertainties are calculated under the assumption that the north and south ends are pieces of one unique structure; as such, dispersion in the “local” background is calculated using apertures in blank regions around both the north and south ends. This is a conservative approach; if the north and south ends are taken individually, the local dispersions are much lower. For the whole structure, if we assume an average surface brightness of µB = 29.9 and a constant shape from the southern tip to the northern tip (a total projected length of ∼

200 kpc), we calculate a total absolute magnitude of MB = −13.9, which corresponds to a

7It should be noted that the boxes depicted in Figures 2.2 and 2.3 do not correspond to the dimensions given here, and are merely for illustrative purposes.

38 7 luminosity of 5.6 × 10 L ,B. The proximity of features A and C to similar-looking features in the H I is intriguing. Similar streams of diﬀuse starlight have been found in the Virgo cluster near M87, which are thought to be collisional in origin (Mihos et al. 2005), so this may well be the case for these features as well—these may be the remnants of small galaxies tidally disrupted by the group potential. In this collisional picture, however, feature B stands out as unusual given its large distance from the central galaxy, M105. Recent observations from the ALFALFA survey (Stierwalt et al. 2009) show no H I detection in its vicinity, making its association with the M96 Group much more ambiguous. Given the rough spatial correspondence between A and C and the H I, it is useful to compare these results with some of the proposed models for the formation of the H I ring. The collision model given by Michel-Dansac et al. (2010), for example, predicts a diﬀuse halo of starlight to have been generated at a surface brightness &29 mag arcsec−2. In terms

of a diffuse component, again, we find nothing up to µB = 30.1 and µV = 29.5; if there is such a diffuse halo component, it must exist at levels even dimmer than this. The model by Bekki et al. (2005), on the other hand, seems to predict no stellar component whatsoever, although it does require the presence of an LSB disk galaxy. In their simulation, this galaxy’s gaseous disk is stripped to smaller radii, with the stellar component left basically intact. No such small isolated gaseous disk has been found in the H I surveys that have been performed on this group (Schneider et al. 1989; Stierwalt et al. 2009), and we discover no heretofore unseen LSB disk in our imaging, despite the simulation by Bekki et al. (2005) predicting that such a remnant should remain within 200 kpc (1◦.5) of the group center (see their Figure 1). Regarding the streamlike features themselves, given their faintness, we can only make general guesses at their physical nature. For example, we might assume, given the luminosities we calculate, that these features are tidally stripped dwarf galaxies. Dwarf galaxies have low mass, and hence typically have low velocity dispersions (of order 10 km s−1; see e.g. Mateo 1998, and references therein). After disruption, the system is no longer gravitionally bound, and so the internal random motions should serve to widen the remnant stream over time. Assuming a characteristic velocity dispersion of 10 km s−1, and given their current widths of roughly 5 kpc, this would imply an interaction timescale on the or-

39 Figure 2.5: Comparison between GALEX FUV (top row), NUV (middle row), and our optical data (V-band, bottom row) for the three dense H I regions discussed in (Thilker et al. 2009) and (Michel- Dansac et al. 2010). The region labels 2, 1, and 2E refer to the nomenclature used by Thilker et al. (2009). Each box is approximately 2.05 on a side (8 kpc). Measured optical colors of the diﬀuse emission (not including point sources) in these regions are as follows. Region 2: B − V = 0.2 ± 0.2; Region 1: B − V = 0.0 ± 0.1; Region 2E: B − V = 0.1 ± 0.2. Uncertainties are due mainly to the small sizes of the regions.

40 der of 500 Myr, shorter than the 1.2 Gyr predicted by Michel-Dansac et al. (2010), and far shorter than the 4–6.5 Gyr proposed by Bekki et al. (2005). However, without kinematic data, it is hard to place robust constraints on formation scenarios for these features. Finally, the only previously known stellar counterparts to the Ring were the three star- forming clumps discovered by Thilker et al. (2009) and later verified by Michel-Dansac et al. (2010). We corroborate this discovery in our imaging as well in the form of three unresolved patches of light coincident with—and with morphologies very similar to—the FUV and NUV emission, and again with blue colors (B − V between ∼-0.1 and ∼0.4, see Figure 2.5). Thilker et al. (2009) and Michel-Dansac et al. (2010) came to opposite conclusions regarding the nature of this star formation, the former finding a best-fit model to their UV colors with a low-metallicity, several hundred Myr burst, and the latter finding that their spectral energy distribution (SED) was fit best by a pre-enriched, 5 Myr instantaneous burst akin to the star-forming region NGC 5291N (within the galaxy NGC 5291). While we cannot constrain either model, it is clear from our data that these star-forming regions are isolated events within the Ring, rather than part of any extended tidal stream of old stars, and thus represent a pure sample of stars that have formed directly out of it. This makes them an ideal target for future observations in the effort to measure the metallicity of the Ring directly. In addition, as these regions are several kpc across, they may represent dwarf galaxies that are in an early stage of formation; obtaining a color magnitude diagram (CMD) of these stars would thus provide further important constraints on the origin of the Leo Ring, and thus further constraints on formation mechanisms of dwarf galaxies (e.g. formation from tidal encounters or out of primordial H I clouds; Lynden-Bell 1976; Barnes & Hernquist 1992; Metz & Kroupa 2007; Pawlowski et al. 2012a).

2.5 Individual Galaxies

The streamlike features discussed in the previous section, while intriguing, remain ambiguous in origin and nature due to poorly constrained colors and lack of distance and velocity data. However, if the collisional origin hypothesis is correct, evidence of the past encounter that produced the features may be found in the group’s galaxies themselves. We now turn to the individual galaxies, searching for accretion signatures and quantifying the

41 structure and stellar populations in their outer regions. Here, we present photometric profiles of the group’s four bright galaxies: M105, NGC 3384, M96, and M95. Each annulus from which the azimuthally averaged photometric values are derived takes into account the changing ellipticity and position angle, measured as a function of radius using IRAF’s ellipse function (for details, see Jedrzejewski 1987). Also for each galaxy, we measure profiles along six constant angular width wedges, using a constant position angle and ellipticity derived from the last isophote fit by ellipse, to look for asymmetries in each galaxy that would otherwise be washed out by azimuthal averaging. All magnitudes and colors are corrected for foreground Galactic extinction

using the values of AV = 0.067 and E(B − V) = 0.02 from Schlaﬂy & Finkbeiner (2011).

2.5.1 M105

M105 (NGC 3379) is a classic elliptical galaxy, most famous for being a keystone of the de Vaucouleurs r1/4 photometric profile (de Vaucouleurs & Capaccioli 1979). Observations in H I and CO have revealed a dearth of atomic and molecular gas (Bregman et al. 1992; Sofue & Wakamatsu 1993; Oosterloo et al. 2010), although there does exist a very small (∼150 M ) dusty disk in the inner 1300 (Pastoriza et al. 2000). The kinematics of the galaxy’s center show some peculiarities, leading to speculation about whether it is actually a face-on triaxial S0 galaxy (Statler & Smecker-Hane 1999), although kinematics gleaned from planetary nebulae show little evidence of rotation out to 33000 (Ciardullo et al. 1993; Douglas et al. 2007). Schweizer & Seitzer (1992) analyzed the galaxy for evidence of recent interactions and found none. In total, aside from the innermost regions, the galaxy appears to be in a very relaxed, pressure-dominated state. Figure 2.6 shows our photometry for this galaxy. It can be clearly seen that the r1/4 model remains a good fit out to the extent of our data (∼ 85000, or 50 kpc) without significant variations within the uncertainty. The B − V color profile shows a slope of ∆(B − V)/∆(log r) = −0.04 mag dex−1 out to r1/4 ≈ 4, or 13.5 kpc (in modest agreement with Goudfrooij et al. 1994). Beyond r1/4 & 4.2 (31000), the level of uncertainty in the colors as indicated by the large dispersion among the different angular profiles makes determination of colors impossible. Out to the 25th magnitude isophote, we measure an integrated color (before reddening correction) of B − V = 0.96, in perfect agreement with the Third

42 r (kpc) 2.1 4.3 8.0 13.7 21.9 33.3 48.8 21 M105 22 23 24

B 25 µ 26 27 28 29

1.00

0.98

0.96 V

− 0.94 B 0.92

0.90

0.88 2.5 3.0 3.5 4.0 4.5 5.0 5.5 r1/4 (arcsec)

Figure 2.6: Left: B-band surface brightness and B − V color profiles for M105. The colored lines show the profiles of equal-angle cuts across the galaxy, which are shown schematically in the upper right of the figure (solid black regions indicate masking). Solid lines indicate values derived from the unbinned mosaic, and dashed lines indicate use of the 9 × 9 binned mosaic. Likewise, black triangles are azimuthally averaged values in the unbinned image, and green squares are azimuthally averaged values in the 9 × 9 binned image. The abscissa shows the length of the semimajor axis in units of arcsec1/4. The physical scale is given along the top axis, assuming a distance of 11 Mpc. Right: diagram of the galaxy with three labeled ellipses for reference. The values shown are in units of r1/4 in arcseconds. The unbinned image is shown inside r1/4 ≈ 4.1, and the 9 × 9 binned image is shown outside this radius. The yellow dotted ellipse indicates where NGC 3384 was masked out.

Reference Catalog (RC3 de Vaucouleurs et al. 1991) value of (B − V)T = 0.96. We see no non-axisymmetric structure in either the surface brightness or color profiles, so to investigate more thoroughly whether M105 shows any signs of interaction, we created an elliptical model of the galaxy using IRAF’s ellipse and bmodel functions and subtracted this from our image. This is the same procedure outlined in Janowiecki et al. (2010), performed under the assumption that local density variations will show up as residuals after subtraction of a smooth profile. The results are shown in Figure 2.7. The only readily apparent structure in the residual image appears near the center of the galaxy, in the form of a pinwheel-like structure. Such residual artifacts are commonly seen (see, e.g., Janowiecki et al. 2010), and are a sign of slight non-ellipticity in the galaxy isophotes rather than any truly distinct structural component. Another much fainter source 43 Figure 2.7: Left: unbinned B image of NGC 3384 and M105 after subtracting out models of both galaxies. Red dotted ellipses correspond to the B = 26.5 magnitude isophotes for both galaxies. North is up and east is to the left. Right: as in the left panel, but median binned into 9×9 pixel bins. We note that independent model subtractions on the V image show similar results. appears, marginally detected, a few arcminutes to the northwest of the galaxy’s center, which may be an extremely faint shell (µB ≈ 29.3, measured from the subtracted image to isolate the feature). We have verified that the isophotal model itself has no artifacts that would imprint such features, and an independent application of the technique on the V-band image also reveals these structures. Other than these weak features, however, the galaxy appears to be well-modeled by smooth elliptical isophotes with no sign of recent accretion. Consistent with previous studies, then, M105 appears remarkable in just how unremarkable it is. Our data imply that if any unusual formation signatures are left to be found in M105, they must lie in the galaxy’s stellar populations. Indeed, Harris et al. (2007) studied the discrete stellar populations in M105’s outer halo at a radius of ∼ 54000 (r1/4 = 4.8) and found evidence for a substantial population of low-metallicity stars. If the outer halo is significantly more metal-poor than the inner regions, this should result in systematically bluer isophotes, modulo differences in mean stellar age between the halo and inner galaxy. If anything, our data suggest that the metallicity gradient becomes more shallow, indicative of a redder stellar population in the halo, but at the extremely LSB characteristic of the

44 outer halo (µB > 28), the uncertainty in our color measurements is extremely large and makes quantitative comparison diﬃcult.

2.5.2 NGC 3384

NGC 3384 is a lenticular galaxy in close proximity (in projection) to M105. Like M105, it too is nearly devoid of gas (Sage & Welch 2006; Oosterloo et al. 2010) and dust (Tomita et al. 2000), and Busarello et al. (1996) found a fairly ﬂat color proﬁle at around B − V ≈ 0.9 out to 4000, typical for S0 galaxies (Li et al. 2011). The galaxy also contains several “faint fuzzies” (Burkert et al. 2005), globular cluster-like objects with

large half-light radii (Reff = 7–15 pc, versus 2–3 pc in normal global clusters), which Burkert et al. (2005) hypothesize may have originated in a low impact parameter galaxy collision. Previous work has also shown the galaxy to be host to a number of other notable peculiarities, including strong isophote twists (Barbon et al. 1976; Busarello et al. 1996), a pseudobulge or double bar (Pinkney et al. 2003; Erwin 2004; Meusinger et al. 2007), and outer boxy isophotes (Busarello et al. 1996). NGC 3384 thus appears much less well-settled than its neighbor M105. Figure 2.8 shows our profiles for this galaxy. It has a roughly exponential surface brightness profile out to ∼32500; beyond this radius, the profile becomes flatter and may simply mark the transition into the local background. Asymmetry is seen near the center (within 4000), resulting from the bulge/bar complex, and from ∼10000–15000, where a hump appears in the azimuthally averaged profile. The color profiles appear mostly flat at B−V = 0.88 out to a radius of ∼20000, after which they redden. This redward color gradient does not appear to be due to influence from the neighbor M105; at this radius, M105’s surface

brightness is µB = 28.5, compared to NGC 3384’s µB = 25.5. The surface brightness profile appears to show a mild anti-truncation at this radius as well, although strong deviations from an exponential profile in the rest of the disk make this difficult to quantify. We measure an integrated color (without extinction correction) of B−V = 0.92, in good agreement with the RC3 value of (B − V)T = 0.93. To further investigate NGC 3384’s outer structure, as we did for M105, we subtracted a smooth model for NGC 3384 to search for residuals. This is also shown in Figure 2.7. Unlike with M105, however, we discover very clear residuals from the smooth profile in

45 the form of two arcs northeast of the galaxy center, as well as a diffuse plume protruding out toward the southwest. Again, all of these features are seen when subtracting a model from the V image as well. The northeast arcs appear much less sharp than arcs often seen in elliptical galaxies (e.g. Malin & Carter 1983), which are thought to be formed via minor mergers (Hernquist & Quinn 1987). The binned image reveals another faint structure toward the south, as well as a possible bridge connecting the innermost arc to the southwest plume, which would imply more of a ring or disk shape. The inner arc appears at a radius of ∼12000, and fades at ∼16500. It has a surface brightness of µB ≈ 27.5 with the underlying disk subtracted. In this region, the northeast (blue) profile from Figure 2.8 is the brightest, followed by the northwest (green), and with the southeast (red and purple) apparently following a more regular exponential disk profile. This behavior appears to reflect the location of the arc, though it is interesting to note that we see no color deviations in this region. The southwestern plume appears at roughly the same radius; however, due to the presence of M105, we can say little about the structure of the disk in its vicinity. The plume has an average surface brightness of µB ≈ 27.8 and a much less regular morphology, making its nature more ambiguous. The outermost arc appears from a radius of ∼20000 to ∼27500, the location of the red color gradient (and pos-

sible anti-truncation). This arc is only slightly fainter than the inner arc: µB ≈ 27.8. These

surface brightnesses correspond to absolute magnitudes of MB ∼ −11.7 and MB ∼ −12.1

for the inner and outer arcs, respectively, and MB ∼ −12.2 for the southwestern plume, which implies progenitors with the luminosities of dwarf galaxies (e.g. Mateo 1998). A “faint extended luminous arc” in this galaxy, so described by Busarello et al. (1996), was photographed by David Malin in 1984 and appears to correspond spatially to this outer arc (Malin 1984, the photograph is shown in Busarello et al. 1996). As such, we would suggest based on our imaging that the arc discovered by Malin is actually part of a system of accretion features in the galaxy. This total system (inner arc, outer arc, and southwestern plume) has a combined luminosity of ∼ 3 × 107L , again indicative of a fairly low stellar mass system. At first glance, the surface brightness profile shown in Figure 2.8 is suggestive of an exponential disk with one or more breaks. However, the presence of the northwest arcs influences the major axis profile, resulting in the hump at 10000. Instead, the profile along the

46 southwest wedge (shown by the red curve), away from the arcs, more accurately follows the true underlying disk structure: a pure exponential out to at least 16 kpc, or approximately 8 scale lengths. Only at larger radii do we see any suggestion of a break in the profile, in the form of a possible anti-truncation of the disk in its outermost regions. NGC 3384 thus provides an interesting contrast to the extremely well-settled M105. We see that the galaxy’s unusual features are not constrained to the innermost regions—boxy isophotes exist as far out as 16 kpc, and the two arcs are strong evidence of disturbance. In some ways, this galaxy is similar to the face-on S0 Arp 227, discussed in Schombert & Wallin (1987). That galaxy showed shells with much less sharp morphology similar to the more standard shells often seen in elliptical galaxies, and with colors that deviated little if at all from the integrated color of the parent galaxy. Schombert & Wallin (1987) explained those features as possibly arising due to an accreted and subsequently evolved hot component from a neighboring massive galaxy (in that case, NGC 470). In NGC 3384’s case, two obvious candidates for such an interaction would be M105 and M96; however, as shown in the previous section, M105 appears to show little sign of any dynamical disturbance in its recent past. At roughly 30000, NGC 3384’s color is red enough (B − V ≈ 1) to suggest an old halo population. The surface brightness profile begins to flatten here as well, as would be expected if an extended halo began to dominate the profile (Mart´ın-Navarro et al. 2014). However, as previously mentioned, it is equally plausible within our uncertainty at these levels that we are simply seeing the contaminating influence of the scattered light in the image.

2.5.3 M96

M96 (NGC 3368) is one of two massive late-type galaxies in the group. The galaxy has long been known to have a faint outer ring or set of spiral arms originating at roughly 9000 (e.g. Schanberg 1973), which we clearly resolve as part of an outer disk with a position angle oﬀset by more than 20◦ from the high surface brightness inner disk. These outer spiral arms have primarily made it (along with NGC 3384) a long-favored candidate in the proposed collisional origin of the H I ring (Rood & Williams 1984; Michel-Dansac et al. 2010), as this level of asymmetry is diﬃcult to produce via secular evolution processes. The galaxy’s connection to the Ring was apparent from the Ring’s initial discovery by

47 r (kpc) 5.3 10.7 16.0 21.3

NGC3384 22

24 B µ 26

1.20 1.15 1.10

V 1.05 −

B 1.00 0.95 0.90 0.85 100 200 300 400 r (arcsec)

Figure 2.8: As in Figure 2.6, but for NGC 3384. The units of the labeled ellipses in the right panel are now in arcseconds, and the yellow dotted circle represents where M105 was masked out. Only full wedges are plotted. We measure a scale length of 3600. 5, or 1.9 kpc.

Schneider et al. (1983) onward, with Schneider (1989) noting that, since most of its H I content exists in its outer disk, M96 may actually be accreting matter from the Ring. Figure 2.9 shows our surface brightness and color profiles for M96. The galaxy shows a clear dip in its surface brightness profile between 8500 and 20000 (this ‘dip’ is what has been previously described as a Type II OLR, or Outer Lindblad Resonance, ‘break’ by Erwin et al. 2008). This appears to reflect a gap between the inner and outer disks. It is interesting to note that along the southeast and northwest profiles (the purple and green wedges, respectively), we do not see the dip, and in fact these profiles appear to have the same smooth exponential behavior as the outer disk, implying one continuous structure rather than, for example, an inner disk embedded in an offset outer ring. The color profiles show an asymmetry in this “dip” region as well, bluer in the northeast than the southwest just outside of the inner disk. Another such asymmetric dip appears around 20000; these are most likely caused by H II regions in the northern arm, which are not visible in the much fainter southern arm. Beyond this radius, the color profile appears flat with some evidence of a continued northeast/southwest asymmetry even at extended radii, and the exponential

48 surface brightness profile continues unimpeded to the extent of our measurements. Beyond 10 kpc, there is no evidence for a disk break in the outer disk, which shows a smooth exponential profile out to 25 kpc (7.5 scale lengths, as measured using the profile beyond 20000). Again, the integrated (uncorrected) color of B − V = 0.85 that we measure is in good agreement with the RC3 color of (B − V)T = 0.86. Most (> 60%) disk galaxies show breaks or truncations in their surface brightness profiles (Pohlen & Trujillo 2006), and while the inner disk of the galaxy does show a break, the lack of a truncation in the outer disk (beyond 10 kpc) is interesting. If we assume, based on the arguments of Pohlen & Trujillo (2006) and Erwin et al. (2008), that truncated disks are the normal end state for spiral galaxies, the lack of truncation might imply that M96’s disk is still being built. For example, mergers or accretion processes have been used to explain anti-truncations (Type III disks, e.g., Penarrubia et al. 2006; Younger et al. 2007) via depo- sition of excess light at large radius; it may thus be that mild star formation in the outer disk of M96 caused by accretion from the H I ring is maintaining an exponential profile out to very large radii. The northeast/southwest asymmetry in our color profiles may reflect this process: the northeast side, which shows bluer colors at ∼8500 and ∼20000, is the direction from which the H I appears to be accreting (Schneider 1989). That these colors are only mildly bluer (B − V = 0.7, compared to 0.8 in the remaining angular wedges) and spatially localized (little azimuthal mixing even at the innermost radius, 8500) would imply that such star formation, if occurring, would be rather weak and somewhat recent, as it has not had time to azimuthally mix. The H I bridge is also visible in more detailed H I observations from Stierwalt et al. (2009), which also indicates that M96’s H I disk is highly extended and lopsided, making the accretion hypothesis quite plausible (Bournaud et al. 2005). That said, the overall impression from the data is that M96’s outer stellar disk appears morphologically very smooth. The isophotes appear very regular, with the centroid of the outermost isophotes drifting south only by ∼1500 (a change of only ∼2.5%). If M96’s outer disk is currently being built by accretion of gas from the Ring, it is apparently doing so in a very smooth and ordered way, despite the one-sided nature of the accretion and the lopsidedness of the accreted gas. A more plausible scenario may be that the accreted gas is collecting in the outer spiral arms and forming stars in those regions: Schneider (1989) showed that most of the galaxy’s H I content is located within a dense ring located

49 r (kpc) 0.0 5.3 10.7 16.0 21.3

20 M96

B 24 µ

1.1

1.0 V

− 0.9 B

0.8

0.7

0 100 200 300 400 r (arcsec)

Figure 2.9: As in Figure 2.6, but for M96. We measure a scale length of 6100. 04, or 3.3 kpc (assuming a pure exponential proﬁle).

at roughly ∼20000(see their Figure 3), the location of the outer spiral arms (see Figure 2.9). This scenario does not easily explain the single-exponential nature of the outer disk, however.

2.5.4 M95

M95 (NGC 3351) is a barred spiral galaxy hosting a well-studied, star-forming circumnuclear ring (e.g., Alloin & Nieto 1982; Colina et al. 1997; Elmegreen et al. 1997; Comeron´ et al. 2010), as well as a larger ring of H II regions encircling the bar. It is the most isolated member of the group, and appears mostly undisturbed but for the star-forming activity in the inner disk. Figure 2.10 shows our surface brightness and color profiles for this galaxy. The bar and ring clearly influence the profiles inside 9000, with asymmetries resulting from spiral arms between 9000 and 13000 (again, the dip seen near 13000 has previously been classified as a Type II OLR break; Erwin et al. 2008). Beyond 13000, we see a smooth exponential disk (scale length 3.4 kpc) coupled with a mild blue color gradient, until a break and reversal in the gradient at ∼20000 (scale length 2.3 kpc). This broken exponential describes the disk 50 morphology well to the extent of our data (40000 or 25 kpc). The outer isophotes of this galaxy appear quite regular, well-fit by smooth ellipses. No plumes or other asymmetries are visible even at extremely low surface brightness (&29 mag arcsecond−2). Once again, our measured integrated uncorrected color of B − V = 0.78 agrees well with the RC3 value of (B − V)T = 0.80. Given the presence of a distinct bar, a ring, and spiral arms in this galaxy, as well as the regularity of the outer isophotes (in both visible light and H I, e.g. Stierwalt et al. 2009), one possible explanation for the upturn in the color profile is an age effect due to outward radial migration of stars, driven by the non-axisymmetries in the inner disk (e.g., Roskar et al. 2008b,a). Such an effect has been observed in other galaxies; for example, consider the case of NGC 2403, in which Williams et al. (2013) discovered a flattening of the metallicity gradient beyond ∼12 kpc. Radial migration preferentially moves stars outward (Roskar et al. 2008b); the farther out a population of stars travels, the longer the travel time and hence the older the population must be (increasing age with increasing radius). A sample of uniform metallicity stars like in NGC 2403 showing such a radial age gradient would tend to show a red color gradient as well. Thus, a similar effect may be behind the behavior we see in M95’s outer disk. M95 thus appears to have no strong signatures of recent accretion in its outer disk. While its inner regions are tumultuous—Sersic´ & Pastoriza (1967) classified it as a ‘hotspot’ galaxy—most of the star formation and other activity is well-described by resonances with, for example, the stellar bar (Devereux et al. 1992; Helfer et al. 2003). Given the galaxy’s membership in the M96 Group, any interactions with other group members must have been extremely weak or have happened in the distant past, long enough ago that the outer disk has had time to recover. In Section 2.4, we suggested that Stream B (200or 65 kpc away from M95, see Figure 2.3), rather than being associated with the Leo H I ring, may have resulted from the tidal disruption of a satellite galaxy interacting with M95. If so, the encounter appears to have made little impact on M95’s outer disk.

51 r (kpc) 5.3 10.7 16.0 21.3 20 M95 22

24 B µ 26

1.2

1.0 V

− 0.8 B

0.6

0.4 100 200 300 400 r (arcsec)

Figure 2.10: As in Figure 2.6, now for M95. We measure an inner scale length of 6400. 26 (3.4 kpc), and an outer scale length of 4200. 32 (2.3 kpc) beyond the break at 20000.

2.6 Discussion

2.6.1 The Origin of the Leo Ring

In the context of the formation of the Leo Ring, arguably the most intriguing aspect of the M96 Group, the lack of either a diﬀuse stellar counterpart or strong interaction signatures in any of the group’s four most massive galaxies is extremely puzzling. The imprint of strong interactions on the outskirts of galaxies (in the form of tidal features or distorted isophotes) should remain over the long dynamical timescales of the outer disks, yet we see no sign of such features in our imaging. This makes it hard to envision dynamical scenarios that invoke recent interactions to produce the Leo Ring. Even with our deep imaging, however, conclusive evidence favoring any particular formation scenario remains lacking. We describe below in more detail the connection between the observed structure of the Leo Group galaxies and the various formation scenarios for the Leo Ring.

52 The Case of M96 and NGC 3384: the Collisional Origin

M96 and NGC 3384 (the two galaxies commonly proposed to have generated the Ring) demonstrate the difficulty in reconciling our results with collisional origin models for the Ring; while both galaxies show signs of disturbance (the arcs found in NGC 3384, and the position angle offset between the inner and outer disk of M96), it is unclear whether or not these could be the remnants of an interaction of the type proposed by Rood & Williams (1984) and Michel-Dansac et al. (2010). For instance, the type of morphology seen in M96 has occasionally been referred to as an “oval distortion” and may actually be a result of secular evolution (Kormendy & Kennicutt 2004). The galaxy M94 (NGC 4736), for comparison, has a very similar morphology to M96, yet appears to have no nearby companions of significant mass (Trujillo et al. 2009). However, the mildness of the interaction signatures we do find in the outer disks must constrain the timescale of any interaction model, as clearly any significant perturbation caused by an interaction has by now been erased. The simulation by Michel-Dansac et al. (2010) required 1.2 Gyr for the Ring to achieve its current size, which is roughly the orbital period of M96’s extreme outer disk (based on the gas velocities of Schneider 1989). Yet M96’s outer disk, the region most loosely bound and hence most easily perturbed, appears mostly smooth, with the exception of the slight southern skew shown in Figure 2.9. It is difficult to imagine how the galaxy’s outer disk could settle so rapidly (in one orbital period) after such a strong encounter as envisioned by Michel-Dansac et al. (2010). To displace such a massive amount of gas into the Leo Ring while only leaving behind such mild signatures in M96’s outer stellar disk would require a high interaction velocity (Spitzer & Baade 1951), while the group’s low measured velocity dispersion (Pierce & Tully 1985; Stierwalt et al. 2009) would seem to imply a more tightly bound system. The timescale of 6.5 Gyr proposed by Bekki et al. (2005) may be more realistic in this regard, but again, the expected signature their model would leave behind (a gas-deprived, LSB galaxy) does not appear to be present. The original scenario proposed by Rood & Williams (1984) appears even less likely due to its much shorter timescale (500 Myr). Alternatively, rather than an encounter disturbing any pre-existing outer disk, it is con- ceivable that M96’s outer disk was created by the encounter. The apparent accretion of H I onto M96 from the Ring is interesting in this context: recall the bifurcation in the color

53 profile, with slightly bluer colors on the side nearest the Ring (B − V = 0.7) than on the farthest side (B − V = 0.8). While these colors are similar to the more quiescent, early-type disk galaxies (Sa-Sab and S0a, respectively; see: Roberts & Haynes 1994), these are colors averaged over wedges; the northern spiral arm itself shows a number of large patches with colors of roughly B − V = 0.4, while the southern arm is nearly devoid of such regions. This is reflected in the GALEX FUV data as well (Martin et al. 2005), in which both arms are clearly visible at 20000, but the southern is visibly fainter than the northern. Measuring the FUV flux in the northern arm from the GALEX image and adopting the conversion to star formation rate (SFR) given by Kennicutt (1998), we find a total SFR of 0.002 M yr−1over an area of roughly 45 kpc2, whereas in the southern arm we find only 5×10−4 M yr−1 over an area of 20 kpc2. This thus implies star formation surface densities of 4 × 10−5 M yr−1 kpc−2 in the north and 2 × 10−5 M yr−1 kpc−2 in the south, a factor of two lower. That said, these derived SFRs do not take into account dust attenuation, which can be a significant factor in deriving SFRs from UV flux due to scattering and absorption effects; Buat et al. (2005), for example, found a mean FUV attenuation of 1.6 mag in their galaxy sample (which was dominated by late-type galaxies). A more detailed analysis of the SFR in M96 should of course take these effects into account. Nonetheless, this is still apparently fairly low-level star formation that could conceiv- ably be induced by slow accretion from the H I ring, which itself totals only 109 M . We do also see star formation continuing within the southern arm, so if we assume that this is left over from the initial accretion where the H I ring meets the galaxy’s disk, we can place a lower limit on the timescale of accretion of one orbital time at the 10 kpc radius where we see the spiral arms, which is roughly 400 Myr. This does not seem to pose a problem for the 1.2 Gyr interaction timescale proposed by (Michel-Dansac et al. 2010), or even the 500 Myr timescale proposed by Rood & Williams (1984) if we assume gas began accreting immediately after the time of the interaction. However, it is also true that the current SFR is not high enough to build the outer disk in a reasonable time frame; at this rate, to produce the total disk luminosity beyond 20000 would require longer than a Hubble time. As such, the accretion timescale given here cannot differentiate between a collisional scenario, in which M96 began accreting H I shortly after it had been ejected from NGC 3384, and a non-collisional scenario, in which M96 began accreting H I from a pre-existing cloud 400

54 Myr ago or earlier by simply passing near it. Given all of these ambiguities, it is important simply to reiterate that we find no obvious indications in our data for this galaxy that any particular collisional model is the correct one. Regarding NGC 3384, the embedded arcs discussed in Section 2.5.2 are unusually broad, but again, they do not appear to be signatures of a major interaction. As stated previously, the symmetric morphology of these features strongly resembles shell systems or caustics formed via minor mergers. Such artifacts can be formed from mergers involving a low-luminosity disk or spheroidal companion (Quinn 1984; Hernquist & Quinn 1987), or simply from accretion of material from a passing galaxy (Hernquist & Quinn 1987). While it can be difficult to constrain the type of the progenitor from the morphology alone, they better reflect the models of dwarf elliptical disruption shown by Hernquist & Quinn (1988). To conserve phase-space volume, shells formed this way should evolve to become sharper over time (Hernquist & Quinn 1988), so the thickness of the shells may thus imply a rather recent merger (or mergers). Minor mergers are also capable of heating disks (Walker et al. 1996) and leaving behind long-lasting (multiple orbit) warps (Quinn et al. 1993) in outer regions. Also, thick-disk formation models using minor mergers tend to show increasing boxiness with decreasing surface brightness (Villalobos & Helmi 2008), so the presence of such signatures in NGC 3384 again does not immediately imply a major, gas-stripping interaction. The nearest extended tidal feature to the galaxy, feature C in Figures 2.2 and 2.3, is not at all aligned with the galaxy’s disk, making it unlikely that it constitutes stripped material from NGC 3384. The only H I clearly associated with the galaxy is also ambiguous in nature and origin (Oosterloo et al. 2010). Thus, once again, we find no particularly clear evidence in favor of any given interaction model based on either of these galaxies’ properties.

The Case of M95 and M105: Passively Evolving Systems

The other large spiral in the group, M95, shows no signs of recent interactions; instead, its appearance is completely consistent with mild secular evolution. The only suggestion of a past interaction is the possible connection with feature B seen in Figures 2.2 and 2.3. Given that this galaxy lies on the outskirts of the group, its undisturbed morphology may simply be another manifestation of the mechanisms responsible for the morphology–

55 density relation (Oemler 1974; Dressler 1980). It is somewhat intriguing, however, that M95 is apparently not host to any satellites capable of visibly disturbing the isophotes out to 25 kpc (given their smoothness at this radius). Regardless, the galaxy appears to have no connection to the H I ring. This leaves the elliptical galaxy M105, sitting at the center of the group and surrounded almost perfectly by the H I ring. Yet, as our deep imaging shows, it is extremely smooth and relaxed. We find only a marginal detection of any interaction signature, in the form of the faint shell-like structure marked in Figure 2.7. Considering the plethora of low-mass members of this group discovered in the H I (Stierwalt et al. 2009), it is truly remarkable just how unperturbed this galaxy is. If the Leo Ring was generated via strong interactions within the group, it is perplexing that the central galaxy shows no evidence of such an encounter. Yet, given the uncertain nature of the interaction model for forming the Leo Ring, it may be that a more reasonable explanation for the Ring’s origin lies with M105 itself. Indeed, extended complexes of neutral hydrogen surrounding normal early-type galaxies are not rare (e.g., van Gorkom et al. 1986; Franx et al. 1994; Appleton et al. 1990; Schiminovich et al. 2000; Oosterloo et al. 2007, 2010; Serra et al. 2012). While the original “primordial origin” concept for the Leo Ring argued that the H I is a pristine remnant from the early universe out of which the group members formed, modern galaxy formation models have ellipticals forming hierarchically, through major mergers of smaller objects. This process can be quite messy and, if gas-rich galaxies are involved, can eject a significant amount of H I out of the remnant in the form of extended tidal tails. If this gas could settle into an extended ring, this could place the formation of the Leo Ring at an intermediate age— likely many Gyr ago, given the relaxed state of M105 and the lack of any observed tidal features around it. However, this idea still suffers from the problem of longevity (how the Ring can be stable for gigayears given the short group crossing times: Schneider 1985). Additionally, it suffers from an angular momentum argument: gas in merger simulations tends mostly to shock and lose angular momentum, sinking to sub-kiloparsec scales and initiating starbursts, with only a fraction remaining in intermediate-scale (similar to Cen A: van Gorkom et al. 1990) disks (e.g. Mihos & Hernquist 1996). It would thus seem extremely difficult to create a 200 kpc diameter ring of low column density gas in this way.

56 2.6.2 The Lack of Intragroup Light

Our deep imaging shows no extended diffuse intragroup light within the M96 Group, −2 down to a limit of µB = 30 mag arcsec , save for a few small LSB streams reminiscent of tidally disrupted dwarfs. These objects aside, the lack of a significant extended IGL component in this group is puzzling. For example, Sommer-Larsen (2006) predicted via group simulations that anywhere from 12% to 45% of the light found in groups should exist in an IGL component by the present day, with a spatially patchy distribution. We do not find anything similar to this in the M96 group. Excluding any undetected diffuse component, the amount of IGL we find constitutes an essentially negligible fraction of the total group light (< 0.01%). Sommer-Larsen (2006) did find that the amount of IGL increases as the group evolves, with the so-called “fossil groups” (as defined by D’Onghia et al. 2005) having suffered the most processing. The M96 Group does not qualify as a fossil group by the D’Onghia et al. (2005) criteria,8 so it may simply be that the group is not dynamically evolved enough to have generated a substantial amount of intragroup light. However, this conclusion is somewhat at odds with M105’s very relaxed state, which suggests that that galaxy is at least a well evolved system. An alternative model comes from simulations by Kapferer et al. (2010), which argue that the fraction of intragroup stellar mass (and hence IGL) should decrease over time due to low frequency of interactions; new stars form in the galaxies over time, but few new stars are ejected to contribute to the IGL. Even so, these models predict that the intragroup stars still contribute from 3% to 30% of the total group stellar mass, which remains higher than what we find. In general, it seems that the M96 Group simply does not easily fall into either of the paradigms described by Kapferer et al. (2010) and Sommer-Larsen (2006). Comparing observations to simulations is not always straightforward, however. For example, the stream-like features found in our imaging data, if indeed part of the M96 Group, have luminosities that fall well below the mass resolution in either the Kapferer et al. (2010) or Sommer-Larsen (2006) models. Furthermore, the brightest galaxies in the Sommer-Larsen (2006) simulations, around which the IGL is centered, are apparently much brighter than M105. Those model galaxies show surface brightnesses of µB . 26.5 at

8Namely, the second brightest galaxy is at least 2 mag fainter than the brightest galaxy; in the M96 Group, all four galaxies are within 1 mag of each other.

57 a radius of 39 kpc (see their Figure 3), while M105 is already below µB = 28 at this radius. This raises the possibility that the M96 Group simply hosts a correspondingly dimmer IGL halo; however, if this is the case, it would be at such an LSB that it would not contribute more than a few percent of the total group light, again suggesting a much lower IGL fraction than that predicted by the simulations. Of course, from the observational perspective, quantitative comparisons to simulations or even to other data depends strongly on how one defines IGL in the first place (good demonstrations of this are shown in Kapferer et al. 2010; Rudick 2010), as well as how one defines a “group” of galaxies. If, for example, we assume a projected area of the M96 Group of ∼ 0.13 Mpc2 (using a rough radius of 200 kpc; see the captions of Figures 2.2

and 2.3) and use our constraints on the observed IGL (an upper limit of µB = 30.1; the light of the three detected streams is negligible), the upper limit on the diffuse starlight of this group comes to ∼4% of the total. If we assume instead that the M96 Group and the M66 Group are a single, larger group (as has been suggested due to their low mutual velocity dispersion and similar distances; see Stierwalt et al. 2009), and that diffuse starlight extends uniformly throughout a circular area encompassing both subgroups (a radius of about 4◦, or 800 kpc), the IGL fraction increases to 20%, although it is important to reiterate that these are absolute upper limits; we in fact make no such detection. Including the extended disks or halos of the group galaxies would artificially add light to the IGL, hence the often rather large ranges of such quantities found in the literature. Given this, it is perhaps more reasonable to compare to other observations of group environments on a more qualitative level. Previous observations of the loose M101 Group showed a similar lack of an extended, diffuse IGL component down to a limiting surface brightness of µB = 29.5 (Mihos et al. 2013b). However, observations of compact groups by da Rocha & Mendes de Oliveira (2005) and da Rocha et al. (2008) show smooth envelopes of diffuse light surrounding most of the galaxy groups in their sample. Observations of more massive, even denser environments such as galaxy clusters can show even larger fractions of diffuse intergalactic light (e.g. Adami et al. 2005), so it may simply be that local density plays the largest role. Intrahalo light (IHL) in general is thought to be mostly composed of stars that have been gravitationally stripped from their host galaxies (e.g. Napolitano et al. 2003; Murante

58 et al. 2004; Rudick et al. 2006; Purcell et al. 2007). It is thus reasonable to assume that the two most influential factors in the generation of IHL in any given environment are the strength and frequency of the interactions. These two factors in turn depend on two physical quantities: mass and density of the cluster or group (mass being essentially a proxy for the number of galaxies; Dressler 1980). For example, the Coma Cluster is one of the densest and most massive clusters in the nearby universe, and the diffuse component of this cluster is dense and bright enough to have been identified in early photographic imaging by Zwicky (1951), although substructure was not detected until much later (Gregg & West 1998; Adami et al. 2005). The Virgo Cluster, by contrast, is less massive and dense than Coma, and hosts a fainter ICL component, again with notable substructure (Mihos et al. 2005; Janowiecki et al. 2010; Rudick et al. 2010a). On the low-mass, high-density end, large fractions of IGL (20%–50%) have been found via imaging compact groups of galaxies (Nishiura et al. 2000; White et al. 2003), but with considerable scatter from group to group (da Rocha & Mendes de Oliveira 2005; da Rocha et al. 2008) and with some compact groups having little or no IGL (Aguerri et al. 2006; da Rocha et al. 2008). With fewer galaxies in play, interactions should be less frequent; this implies that the actual fraction of stellar mass that is liberated from host galaxies will depend more strongly on the details of each individual interaction (e.g. relative masses, inclination angles, and velocities of the interacting galaxies; Toomre & Toomre 1972; Negroponte & White 1983), thus giving rise to higher dispersion in IGL properties. The M96 Group may thus fit into this picture by occupying the low-mass, low-density regime of galaxy environments. Measuring the total mass of the M96 Group is not trivial, since the entire Leo I Group has considerable substructure spatially and kinematically (e.g. Stierwalt et al. 2009). A simple application of the virial theorem using the velocity dispersions and harmonic mean radii from Stierwalt et al. (2009) yields masses for the M96 12 Group in the range of 2–6 ×10 M . Using the calibration between velocity dispersion and group mass given by Yang et al. (2007, their Equation (6)) gives a rough halo mass 12 of 8 ×10 M . While it is highly unlikely that the M96 Group is virialized, it seems clear that the M96 Group is less massive than the mass range of previous loose group studies 13 15 (typically 10 M – 10 M ; Castro-Rodr´ıguez et al. 2003; Feldmeier et al. 2004; Durrell et al. 2004). These more massive loose groups have low IGL fractions as well (2%), yet

59 still seem to contain a higher fraction than that detected in the M96 Group, or the still less massive M101 Group (Mihos et al. 2013b). We may thus be seeing a continuation of this pattern: lower density and lower masses yield lower IHL fractions. In this context, the Leo Ring may thus serve as a vital clue regarding the types of dynamics seen in low-mass loose groups such as M96. This is a rare system in the local universe, in terms of the size and mass of the H I complex. A somewhat similar structure resides in the NGC 5291 complex located 58 Mpc away at the edge of the A3574 cluster, but this H I ring is much more massive and contains a number of large, obviously star- forming clouds (Malphrus et al. 1997; Boquien et al. 2007). As discussed in Section 2.4, star formation in the Leo Ring is much more subtle: it seems to only amount to the three kiloparsec-scale clumps discovered by Thilker et al. (2009). Bot et al. (2009) did make a tentative discovery of dust within the Ring, near (but not precisely coincident with) the region labeled 1 in Thilker et al. (2009). These discoveries are beneficial, as obtaining a metallicity for the ring would be instrumental in constraining the Ring’s origin. Again, however, the analyses of these star-forming clumps by Thilker et al. (2009) and Michel-Dansac et al. (2010) produced conflicting results (1/50 solar metallicity in the former and 1/2 solar in the latter), and Bot et al. (2009) considered their dust detection too uncertain to make any definitive conclusion. The Ring also appears devoid of obvious H II regions (Donahue et al. 1995) and planetary nebulae (Castro-Rodr´ıguez et al. 2003). As such, these very localized star-forming regions may be the extent of star formation within the Leo Ring. Other avenues for studying star formation and the stellar content of the Leo Ring include the possibility of deep space-based imaging to search for discrete stellar populations in the Ring, and even deeper ground- based searches for diffuse Hα indicative of additional ongoing star formation throughout the Ring. Such studies are clearly needed to understand the formation and evolution of this structure, and the galaxy group surrounding it.

2.7 Summary

−2 We have performed deep (µB,lim ≈ 30 mag arcsec ) imaging of the M96 Group in order to search for LSB intragroup light, as well as to study the morphologies and stellar

60 properties of the outer disks and halos of galaxies in the group. We ﬁnd no diﬀuse stellar

counterpart to the group’s 200 kpc diameter H I ring down to µB = 30.1, but identify three extremely faint (µB ∼ 29.2 to 29.9) streamlike features apparently associated with the group. Two of these features may be directly associated with the H I ring, as they show some amount of coincidence with similar-looking features in the distribution of neutral hydrogen. We constructed surface brightness and color profiles (both azimuthally averaged and in discrete angular wedges) for the group’s four most massive galaxies—M105, NGC 3384, M96, and M95—out to ∼25 kpc in each disk galaxy, and out to ∼50 kpc in the elliptical M105. We find no evidence of recent disturbances in either M105 or M95, though the latter shows a clear disk truncation and redward color gradient beginning around 12 kpc, possibly due to the effects of radial stellar migration. We also find two arcs embedded in NGC 3384’s disk, as well as a strong redward gradient in the outer isophotes reaching old halo-like colors (B − V ≈ 1), which appears to be associated with the outermost arc. M96 shows mild asymmetry in its extreme outer disk, as well as a color asymmetry apparently reflecting mild star formation likely induced by accretion from the H I ring; no disk break is found in the outer disk out to the extent of our data, possibly indicating that this component of M96 is still being built. The lack of strong tidal disturbances in the outskirts of the group galaxies, coupled with the absence of significant starlight associated with the Leo Ring provides some tension for models that rely on recent strong encounters between group galaxies to form the H I ring. The connection between the Ring and the individual galaxies remains unclear; however, further study of the small diffuse structures detected in our imaging may allow for further testing of the various theories proposed to explain the origin of the Ring. Finally, unless the M96 Group has been caught in the very early stages of evolution—unlikely, given the relaxed state of its central elliptical, M105—the extremely low fraction of intragroup light we measure also places constraints on the ability of groups of this mass and density to act as “pre-processors” for the more commonly seen ICL found in massive galaxy clusters.

61 2.8 Acknowledgments

This work has been supported by the National Science Foundation via grants 1108964 (J. C. M.) and 0807873 (J. J. F.) and Research Corporation grant 7732 (J. J. F.). We also thank Tom Oosterloo for the use of his WRST H I data in the construction of Figure 2.3, and Pat Durrell for many useful discussions. This research made use of NumPy, SciPy (Oliphant 2007), and Matplotlib (Hunter 2007). Figures 2.6 through 2.10 made use of Min-Su Shin’s (University of Michigan) publicly available code img scale.py 9. Facility: CWRU:Schmidt

9http://dept.astro.lsa.umich.edu/∼msshin/science/code/Python ﬁts image/ 62 Chapter 3

Deep Imaging of M51: a New View of the Whirlpool’s Extended Tidal Debris

3.1 Abstract

We present deep, wide-ﬁeld imaging of the M51 system using CWRU’s Burrell Schmidt telescope at KPNO to study the faint tidal features that constrain its interaction history. Our

images trace M51’s tidal morphology down to a limiting surface brightness of µB,lim ∼30 −2 mag arcsec , and provide accurate colors (σB−V < 0.1) down to µB ∼ 28. We identify two new tidal streams in the system (the south and northeast plumes) with surface bright- 6 nesses of µB = 29 and luminosities of ∼ 10 L ,B. While the northeast plume may be a faint outer extension of the tidal “crown” north of NGC 5195 (M51b), the south plume has no analog in any existing M51 simulation and may represent a distinct tidal stream or disrupted dwarf galaxy. We also trace the extremely diffuse northwest plume out to a total extent of 200 (43 kpc) from NGC 5194 (M51a), and show it to be physically distinct from the overlapping bright tidal streams from M51b. The northwest plume’s morphology and red color (B−V = 0.8) instead argue that it originated from tidal stripping of M51a’s extreme outer disk. Finally, we confirm the strong segregation of gas and stars in the southeast tail, and do not detect any diffuse stellar component in the H I portion of the tail. Extant simulations of M51 have difficulty matching both the wealth of tidal structure in the system and the lack of stars in the H I tail, motivating new modeling campaigns to study the dynamical

63 evolution of this classic interacting system.

64 3.2 Introduction

Galaxy interactions form the basis of the current hierarchical accretion paradigm for galaxy evolution, and have been linked to the formation of bars, spiral structure, and induced star formation. However, all these features may also result from purely secular processess (see the review by Athanassoula 2010), entangling the various evolutionary effects in individual systems. Understanding interaction histories of nearby galaxies, from which much of our detailed understanding of physical processes comes, is thus particularly important. The galaxy pair M51 is one of the best known nearby interacting systems, comprised of the grand-design spiral NGC 5194 (M51a) and its early-type SB0 companion NGC 5195 (M51b). Deep imaging in the optical and infrared (e.g. Burkhead 1978; Smith et al. 1990; Mutchler et al. 2005) coupled with kinematic studies in optical (e.g. Tully 1974; Durrell et al. 2003) and radio (e.g. Appleton et al. 1986; Rots et al. 1990), have revealed extended tidal streams and complex kinematics, including an apparently counter-rotating tail of gas on its southern end (Rots et al. 1990). M51’s proximity (7.5 Mpc; Ciardullo et al. 2002; Bose & Kumar 2014) and well- deﬁned spiral structure has made it a frequent target for studies of the interplay between spiral dynamics, star formation, and the atomic and molecular ISM (e.g. Calzetti et al. 2005; Hughes et al. 2013). If much of the star formation present in the system is interaction- induced, models become an important means of constraining the physical mechanisms and timescales involved, and hence, for example, the calibration of star formation tracers in galaxies (e.g. Calzetti et al. 2005) and determination of generalized star formation laws (e.g. Kennicutt et al. 2007). This has made the pair a favorite subject of interaction models (e.g. Toomre & Toomre 1972; Howard & Byrd 1990; Salo & Laurikainen 2000; Durrell et al. 2003; Dobbs et al. 2010, and many others), constrained by the morphology and kinematics of the observed tidal features. These modeling campaigns have led to two classes of models for the system: one involving a single ﬂy-by passage (Toomre & Toomre 1972; Hernquist 1990; Durrell et al. 2003), and a second involving multiple passes on a more tightly bound orbit (Salo & Laurikainen 2000; Dobbs et al. 2010). Discriminating between these scenarios can constrain both the timescale and strength of the perturbation that set up the grand-design spiral

65 Figure 3.1: Left: a subset of our B band mosaic, rescaled in intensity for diﬀerent surface brightness ranges: µB< 24.6, 24.6 <µB< 26.5, and µB> 26.5. At the lowest surface brightnesses, the image has been median binned to enhance faint structures. Right: as left, marking regions detailed in Table 3.1 and discussed in the text. Each image is 390 (84 kpc) on a side. and induced star formation in the M51 system. Deeper observations can prove useful in locating additional interaction signatures to further constrain the models. Particularly in M51, in which both galaxies have vastly different star formation rates (Lee et al. 2011, 2012), broadband colors may be used to help determine the origin of many interaction signatures, and add additional constraints from stellar populations. Combining these observations with the kinematics and spatial distribution of the gas in the system can also greatly narrow down the available parameter space for the dynamical models. In this Letter, we present deep multiband optical imaging of M51, taken as part of a survey of low surface brightness features of nearby galaxies, using the 0.6/0.9m CWRU Burrell Schmidt Telescope at KPNO (Watkins et al. 2014). We report on the morphology and colors of both known and newly discovered tidal features which will help constrain the dynamical evolution of the M51 system.

66 Table 3.1. Photometry of Tidal Debris in M51

6 Region µB,peak MB B−V 10 L ,B Origin Gas Content

west stream—north 25.43 -14.3 0.75 45 M51b Gas-poor west stream—south 25.46 -14.2 0.73 39 M51b Gas-poor east stream—north 25.78 -14.1 0.73 39 M51b Gas-poor east stream—south 25.63 -13.9 0.72 30 M51b Gas-poor southeast tail 25.80 -14.8 0.64 68 M51a Gas-rich, oﬀset northwest plume 27.59 -14.5 0.81 101 M51a? Patchy? northeast plume 29.0±0.3 -11.2 — 4 M51a? Gas-poor south plume 29.2±0.7 -11.8 — 8 Unknown Gas-poor

Note. — Absolute photometric errors are σB ∼ 0.04 and σB−V ∼ 0.05.

3.3 Observations and Analysis

We give a brief account of our observation and data reduction procedures here; more details can be found in Watkins et al. (2014) and references therein. We observed M51 in spring 2010 and spring 2012, with a setup identical to that described in Watkins et al. (2014). We observed only on moonless photometric nights, in two filters: modified Johnson B (31×1200s exposures) and Washington M (39×900s exposures). We generated dark sky flats using a comparable number of offset night sky images in each filter (Watkins et al. 2014). Photometric conversion to standard Johnson B and V used observations of Landolt (1992) standard fields to determine color terms, and on-frame SDSS (DR8; Aihara et al. 2011) stars to derive direct photometric zeropoints. Data reduction consisted of standard reduction procedures and modeling and subtraction of reflections and extended wings of bright stars (see Slater et al. 2009). Following sky subtraction, each image was spatially registered and scaled to a common photometric zeropoint before median combining all images into the final B and V mosaics. We then masked the final mosaics of bright discrete sources and spatially rebinned and median filtered them in 9×9 pixel blocks to enhance faint, diffuse features. Our final B image and our colormap of M51 are shown in Figures 3.1 and 3.2, respectively. Our limiting magnitude is µB∼ 30. In the quantitative photometric analysis that follows, we use polygonal apertures to define regions of interest in M51 (see Figure 3.1 and Table 3.1). In these regions, contamina-

67 Figure 3.2: B − V colormap of the M51 system. The image spans 360 × 270 (78 kpc × 58 kpc).

68 tion from background sources can be significant, so we subtract off a local background flux (see Rudick et al. 2010a; Mihos et al. 2013b, for details). At extremely low surface brightnesses, this background contamination dominates the uncertainty, and is non-Gaussian.

At surface brightnesses of µB< 27.5, the relative uncertainties are small, but at extremely low surface brightness (µB> 28.5) they become signiﬁcant and preclude measurement of meaningful colors. All values have also been corrected for the local foreground extinction

(AB = 0.126, E(B−V) = 0.031; Schlafly & Finkbeiner 2011). In searching for tidal features in deep surface photometry, an additional source of contamination is scattered light from foreground Galactic dust, which can mimic diffuse structures (see Rudick et al. 2010a). However, such dust radiates in the mid- and far-infrared, and a comparison of our images with the WISE 12µm and IRIS 100µm images (Meisner & Finkbeiner 2014; Miville-Deschenesˆ & Lagache 2005, respectively) shows no evidence for strong dust contamination cospatial with the diffuse features we identify in Figure 3.1 (to limits of 3.58 MJy sr−1/1.11 MJy sr−1 in WISE/IRIS).

3.4 Results

3.4.1 Morphology and Color of Tidal Features

An examination of Figure 3.1 shows a wealth of tidal debris around M51. At higher surface brightness (µB< 26.5) we see many well-known features, including M51’s southern tidal tail, the east and west tidal streams emanating from M51b’s disk, and the three- pronged structure (the “crown”) just north of M51b. At extremely low surface brightness, we trace the extended northwest plume, and identify two new, extremely faint features—the northeast and south plumes. We also show in Figure 3.3 a comparison between our optical image and the H I data from Rots et al. (1990). The most extended feature is the northwest plume. Originally detected by Burkhead (1978), its tip lies 200 (43 kpc) from the center of M51a. While it partially blends with the brighter west streams extending from M51b, a variety of arguments indicate it is a separate feature. First, the plume is morphologically distinct—it extends almost directly west, while the northern west stream extends ﬁrst northwest and then curves toward the north-

69 east (Figure 3.1). The northwest plume is also ∼0.06 mag redder1 than the adjacent west streams (Table 3.1; we note that the plume shows a uniform color from end to end), implying distinct (albeit similar) stellar populations. Durrell et al. (2003) also note bimodality in the PNe velocities northwest of M51 (although not cospatial with the plume), arguing for kinematically distinct populations. Indeed, we note a possible third component, a slight protrusion just to the northwest of M51a (marked with an arrow in Figure 3.1) which may be a stellar stream overlapping M51a’s outer disk. The THINGS H I map of M51 (Walter et al. 2008) also shows gas extending roughly in this direction. There is very little high column density H I associated with the northwest plume (Figure 3.3), but evidence exists for extremely diﬀuse, high-velocity gas in the vicinity (Appleton et al. 1986; Rots et al. 1990, Pisano 2015, private communication). Diﬀuse H I associated with the northwest plume would also imply an origin with the gas-rich spiral M51a, as the companion galaxy M51b appears to be extremely gas-poor. Dynamically, the northwest plume appears several 100 Myr old; its linear extent (∼40 kpc in projection) would require an unrealistic expansion velocity if, for example, it had formed during the most recent passage proposed by Salo & Laurikainen (2000) (50–100 Myr ago) or Howard & Byrd (1990) (70 Myr ago). At the rotation speed of M51a (210 km/s; Appleton et al. 1986), a timescale of ∼200 Myr is more likely. We have also used stellar population modeling (Kotulla et al. 2009) to show the plume is dominated by old stars. Exponentially declining star formation histories typical of Sb/Sc galaxies (τ =5–9 Gyr) produce colors too blue (B − V . 0.6) to match the plume; its red color demands much older star formation histories (τ =3 Gyr). Similarly, any subsequent interaction- induced star formation in the plume must be weak (<5% of the plume’s stellar mass), else the colors would too blue by >0.15 mags even after 500 Myr. As dust extinction appears negligible due to the low gas column density (Appleton et al. 1986; Rots et al. 1990), our inference of old stellar populations appears sound. Partially overlapping with the northwest plume, but with contrasting morphology and color, are the east and west streams. Both streams are bifurcated, and all branches have very similar colors (B−V= 0.73), which are in turn similar to M51b’s average color (B−V= 0.7;

1While this diﬀerence is comparable to our absolute photometric uncertainty, relative color diﬀerences between adjacent features are constrained to ∼ 0.01 mag uncertainty, due to the stability of the background corrections.

70 Figure 3.3: Comparison between our optical image and (in green contours) the H I data of Rots et al. (1990).

71 Lee et al. 2012). Given this, and the streams’ symmetry around and proximity to M51b, it seems likely that these streams originated from that galaxy’s disk. A variety of processes may give rise to bifurcated tidal streams, including kinematic caustics (Struck & Smith 2012; Smith et al. 2014) and differences in the kinematic evolution of the gas and stellar components (Mihos 2001). In the latter mechanism, stellar bifurcations would only arise via subsequent star formation in the displaced gas tail, which would then show younger stellar populations than the collisionless tail. The lack of H I and the uniformly red colors of these streams would seem to argue against such a scenario in M51b, and instead favor collisionless kinematic caustics. Based on the colors shown in Figure 3.2, much of the crown north of M51b also appears associated with the companion, excepting the bluer middle prong. Lee et al. (2012) showed that the latter is an extension of M51a’s northern tidal arm; indeed, its color is similar to that of the southeast tail (B − V= 0.64), the blueness of which is expected from its origin in M51a’s disk. We also note the incredible sharpness of the ridge formed by the crown’s eastern prong, indicative of a strong caustic. In the southeast tail, the clear separation between the gas and stars (Figure 3.3, noted previously by Howard & Byrd 1990 and Rots et al. 1990) is of particular interest. The stellar tail curves sharply northward and disappears into the tidal detritus near the east streams, while the more loosely wrapped H I tail is much more extended and contains no diffuse stellar counterpart down to our limit of µB= 30.5. Simulations have shown that such gas/star offsets in tidal tails can arise from differences in the radial extent of the two components in the pre-encounter disk (Mihos 2001), where the stellar component of a tail consists preferentially of material with low specific angular momentum and thus “trails” the high angular momentum gaseous tail. While this may explain the tail’s lack of an old stellar population, the lack of either ongoing star formation (Gil de Paz et al. 2007) or a young stellar population is more challenging, as the compressive nature of tidal interactions is known to form massive star clusters and tidal dwarf galaxies (Barnes 1992; Jarrett et al. 2006) in many tidal tails. The lack of a young population in M51’s tail may indicate a highly inclined encounter, where compressive forces in the tail are greatly reduced. As shown by Rots et al. (1990), an inclined encounter also explains the tail’s observed counter-rotation, and is reproduced well by interaction models of Salo & Laurikainen (2000).

72 18

22 B µ 24

28 0.8

0.7 V − 0.6 B

0.5

0.4 0 100 200 300 400 r (arcsec)

Figure 3.4: Surface brightness and color proﬁles along the wedge highlighted in the inset, plotted as a function of the average radial distance from M51a’s center. Error bars in both plots are calculated using the background variance, and are often smaller than the point size.

73 Aside from the structure of the tidal features, our data also shed light on the structure of M51a’s outer disk. The colors get redder by 0.2–0.4 mag outside of the spiral arm region, visible in the colormap of Figure 3.2 and explicitly in Figure 3.4, which shows photometric proﬁles along an elliptical wedge which follows the curvature of the outermost spiral arm westward into the outer disk. We see a dramatic drop in surface brightness between 22000 and 25000, corresponding precisely with a change in the B − V color from 0.46 to 0.65 (roughly, the integrated colors of an Sc and Sa galaxy, respectively; Roberts & Haynes 1994). This region marks the edge of the outermost spiral arm; beyond, the redward gradient continues, ultimately reaching colors similar to those of M51b. The dramatic drop in surface brightness is similar to the arm-interarm contrast measured in the same region in M51a by Schweizer (1976), and seen in the outer regions of other spiral galaxies (Schweizer 1976; Tacconi & Young 1990). The red colors, smooth surface brightness proﬁle, and round outer isophotes in this region might be interpreted as marking the transition to a kinematically hot spheroidal component, but we reject this notion for a variety of reasons. M51a lacks a strong bulge

((B:T)I = 0.27; Pompei & Natali 1997), arguing that the outer component is not likely to be extended bulge light. Furthermore, such a bright halo component is unlikely given the

radius (∼10 kpc) and surface brightness (µB ≈ 25) of the component; by comparison, M31’s smooth halo has the much fainter surface brightness of µB ∼ 29–30 at 10kpc (extrapolating from the power law ﬁts of Ibata et al. 2014). The most reasonable interpretation for this component is that we are simply seeing the outer, redder stellar disk in M51a. An outer disk origin for this light is also suggested by its similarity in color to the more extended northwest plume. The linear structure of the plume would require a dynamical origin from material with coherent motion and high angular momentum, as would be found in an outer disk. Such disks are also expected to show redder colors due to eﬀects from radial migration (e.g. Roskar et al. 2008a), and such colors are frequently observed in outer disks (e.g. Bakos et al. 2008, who found typical outer disk colors of g − r ∼ 0.5–0.6 = 0.7– 0.8 in B − V). Hence, it seems likely that the red outer regions of M51a are simply part of its outer disk, and that the northwest plume may have been ejected from the outer disk during the interaction.

74 3.4.2 Comparison to Simulations

Simulations of the M51 system fall broadly into two classes: parabolic single-pass models with typical timescales of > 300Myr (Toomre & Toomre 1972; Hernquist 1990; Durrell et al. 2003), and more tightly bound, multiple passage models with one encounter several 100 Myr ago, and a second passage .100 Myr ago (Howard & Byrd 1990; Salo & Laurikainen 2000; Dobbs et al. 2010). In general, these models were developed to match the structure and kinematics of the H I and the bright tidal debris (the east and west streams and southern tail). The degree to which they successfully reproduce the fainter, more extended debris (the northwest, northeast, and south plumes) is less clear. Here we develop new insights gleaned from a comparison between our deep imaging data and the existing simulations. Perhaps most useful is a comparison involving the northwest plume. Similar features arise in simulations of M51 from the material nearest the companion at pericenter, which receives the most dispersive tidal kick and is ejected as a broad tidal plume (in contrast, material on the opposite side of the disk forms a coherent tidal tail). Tidal features resembling the northwest plume appear in a variety of M51 simulations, but single-passage models tend to reproduce the plume’s morphology better than the best-fit multipassage models of Salo & Laurikainen (2000) or Dobbs et al. (2010): see, for example, Figures 1 and 8 of (Salo & Laurikainen 2000; Howard & Byrd 1990), respectively. However, we note that both Salo & Laurikainen (2000) and Dobbs et al. (2010) elaborate on only one such model; different parameters in a multi-passage encounter may reproduce the feature more effectively. While the northwest plume cannot unambiguously differentiate between the model classes, the dynamical age implied by its linear extent argues that the initial passage of this encounter (or the sole passage) must have taken place at least a few 100 Myr ago (in more agreement with Howard & Byrd 1990). Unfortunately, the other faint features—the northeast and south plumes—have no clear analogs in any published simulation. The northeast plume perhaps most resembles the diffuse northern component seen in the multiple passage models by Salo & Laurikainen (2000). Single passage models generally completely fail to reproduce this northern component (Salo & Laurikainen 2000; Durrell et al. 2003), so the presence of both the crown and this northeast plume seem to favor the multiple-passage models. Indeed, this feature

75 may simply be a faint outer extension of material in the crown itself; its low luminosity 6 (∼ 10 L ,B) would place it below the mass resolution of most published simulations, and could explain why they miss this feature. The south plume, on the other hand, shows no morphological continuity with any adjacent feature, either in optical or in H I, and has no counterpart in any simulation published to date. While Appleton et al. (1986) detected some diffuse H I on the galaxy’s southwestern side, it does not overlap with the south plume itself. Given all this, the plume would seem to be either a distinct tidal feature not replicated in previous dynamical modeling (although like the northeast plume, it may be too low mass to discern in published simulations), or the remains of a disrupted low luminosity dwarf galaxy in the system. Regarding the southeast H I tail, the lack of an old, co-spatial stellar component may argue that the gas came from an extended gas-rich disk belonging to M51a. Simulations of the M51 system favor a highly inclined encounter to explain the kinematics of the tail (Rots et al. 1990; Salo & Laurikainen 2000); such a configuration would suppress caustic formation in the tail and explain its paucity of star formation and young stars. However, the gas/star offset observed in the tail is still problematic, as such features should be most prominent in highly prograde encounters where the orbital and rotational angular momentum is well- aligned (Mihos 2001). Thus the data still leave some ambiguity in the dynamical evolution of the southeast tail. In summary, no single model of M51 provides a perfect match to the additional constraints derived from our imaging. The single-passage models appear to better reproduce the diffuse northwest plume, while the multi-passage models reproduce better the diffuse northern material (the crown and northeast plume). No simulations to date have produced anything resembling the south plume, although its low luminosity suggests that higher resolution simulations may be necessary to model its structure. The lack of a stellar counterpart in M51’s southeast H I tail argues that the tail was drawn almost exclusively from the outermost, gas-rich regions of M51a’s disk, or that extensive kinematic segregation of stars and gas has occurred, without triggering any wide-spread star formation in the tidally stripped gas. Future modeling campaigns thus are warranted, and have a variety of new constraints and challenges ahead. New simulations must recover the newly discovered tidal plumes,

76 ensure that the northwest plume is correctly oriented and gas-free, reproduce the strong oﬀset between the stars and gas in the southeast tail, and bifurcate M51b’s east and west streams. Successful simulations will then be able to pin down the timescale and geometry of the encounter—critical inputs to studies of spiral structure, ISM evolution, and star formation in this classic interacting system.

3.5 Acknowledgements

The work was supported by the NSF through grant 1108964 to J.C.M.

77 Chapter 4

The Red and Featureless Outer Disks of Nearby Spiral Galaxies

4.1 Abstract

We present results from deep, wide-field surface photometry of three nearby (D =4– 7 Mpc) spiral galaxies: M94 (NGC 4736), M64 (NGC 4826), and M106 (NGC 4258). −2 Our imaging reaches a limiting surface brightness of µB∼28 – 30 mag arcsec and probes −2 colors down to µB∼27.5 mag arcsec . We compare our broadband optical data to available ultraviolet and high column density H I data to better constrain the star-forming history and stellar populations of the outermost parts of each galaxy’s disk. Each galaxy has a well- defined radius beyond which little star formation occurs and the disk light appears both azimuthally smooth and red in color, suggestive of old, well-mixed stellar populations. Given the lack of ongoing star formation or blue stellar populations in these galaxies’ outer disks, the most likely mechanisms for their formation are dynamical processes such as disk heating or radial migration, rather than inside-out growth of the disks. This is also implied by the similarity in outer disk properties despite each galaxy showing distinct levels of environmental influence, from a purely isolated galaxy (M94) to one experiencing weak tidal perturbations from its satellite galaxies (M106) to a galaxy recovering from a recent merger (M64), suggesting that a variety of evolutionary histories can yield similar outer disk structure. While this suggests a common secular mechanism for outer disk formation,

78 the large extent of these smooth, red stellar populations—which reach several disk scale lengths beyond the galaxies’ spiral structure—may challenge models of radial migration given the lack of any nonaxisymmetric forcing at such large radii.

79 4.2 Introduction

Disk galaxies present some of the cleanest laboratories to test theories of galaxy formation and evolution. At first glance, their stellar populations appear distinctly segregated, both spatially and kinematically, with the older, kinematically hotter stars forming central bulges or bars, and the younger, kinematically colder stars forming disks around these bulges. In the hierarchical accretion model of galaxy formation (e.g. Searle & Zinn 1978; White & Frenk 1991; Springel et al. 2005; Vogelsberger et al. 2014), this structure is explained by an “inside-out” formation mechanism, leading to the prediction that the mean age of the stellar populations will be lowest in the disk outskirts. Stars near the galaxy center, formed from primordial gas early in the universe’s history, would quickly enrich their local ISM (McClure 1969; Wyse & Gilmore 1992), while in the outskirts the lower gas densities and star formation rates (SFRs) result in reduced enrichment efficiencies (Schmidt 1959; Kennicutt 1998; Bigiel et al. 2008; Krumholz et al. 2012). Thus, there should exist separate, well-defined age-metallicity relationships (AMRs) at individual radii throughout the disk, with an overall negative radial metallicity gradient for stellar populations of a given age (Twarog 1980; Chiappini et al. 1997; Naab & Ostriker 2006). Beyond a certain radius (when the gas density falls below a “critical” value; Kennicutt 1989), one would expect to find very few, if any, stars formed in situ. This simple and elegant picture, however, is not fully supported by observation. Mea- surements of the AMR in the solar neighborhood, for example, have shown a much larger dispersion than is expected under such a model (Edvardsson et al. 1993; Haywood 2006). The Milky Way’s metallicity gradient also appears to flatten beyond ∼ 10kpc (e.g. Twarog et al. 1997; Yong et al. 2005; Maciel & Costa 2010, but see Luck & Lambert 2011; Lemasle et al. 2013); other disk galaxies seem to show similar behavior (e.g. Bresolin et al. 2009; Vlajic´ et al. 2009, 2011; Sanchez´ et al. 2014). Also, while many disks show negative age gradients in their stellar populations (e.g. Sanchez´ et al. 2014), this trend may often re- verse beyond a certain radius (e.g. Bakos et al. 2008; Zheng et al. 2015). Finally, disk stars are found to inhabit the very extended outer reaches of the host galaxies, sometimes beyond the apparent star formation threshold radius (e.g. Tiede et al. 2004; Davidge 2006, 2007; Azzollini et al. 2008; Bakos et al. 2008; Vlajic´ et al. 2009; Mart´ın-Navarro et al. 2014; Okamoto et al. 2015; Zheng et al. 2015), including patches of new star formation

80 (e.g. Thilker et al. 2005; Gil de Paz et al. 2005; Thilker et al. 2007a; Lemonias et al. 2011; Yıldız et al. 2015). Under the simple model described above, the presence of these stars is mystifying. Reconciling these inconsistencies with theory has prompted detailed investigation into the inner workings of disk galaxies, and some amount of consensus is beginning to emerge. Radial migration of stars appears to be important, wherein stars move radially throughout the disk via resonances with inner bars or spiral arms, while maintaining their reasonably circular orbits (Sellwood & Binney 2002; Debattista et al. 2006). Numerous simulations have shown that this process can move stars in excess of a kiloparsec from their birthplaces (Sellwood & Binney 2002; Roskar et al. 2008b; Sanchez-Bl´ azquez´ et al. 2009; Schonrich¨ & Binney 2009; Minchev et al. 2011, 2013), effectively flattening the AMR (Sanchez-´ Blazquez´ et al. 2009; Schonrich¨ & Binney 2009; Minchev et al. 2013). Additionally, migration can serve as a means of growing the disk radially even in the presence of a star formation threshold (Roskar et al. 2008b; Sanchez-Bl´ azquez´ et al. 2009; Minchev et al. 2012; Roskarˇ et al. 2012; Minchev et al. 2013). While the detailed mechnisms differ in the simulations, a common prediction is the creation of a “U-shaped” age gradient in the stellar populations, where older disk stars are found inhabiting both the inner regions of the galaxy and its extended outer disk. Coupled with a flat metallicity gradient in the disk periphery, this would give rise to the “U-shaped” radial color profile observed in most optical bands. In the simulations by Roskar et al. (2008b), the break in the age gradient is always coupled to a break in the radial mass and luminosity surface density profiles and is pushed farther out over time as the stellar disk gains mass (but see Sanchez-Bl´ azquez´ et al. 2009). Increasingly, observations appear to support this picture. Bakos et al. (2008), using galaxies in the frequently observed Stripe 82 of the Sloan Digital Sky Survey (SDSS; York et al. 2000), found that this predicted “U-shaped” color profile is in fact quite common in disk galaxies and is also commonly coupled to a so-called “Type II” (or downbending; see Pohlen & Trujillo 2006; Erwin et al. 2008) break in surface brightness, in apparent agreement with the simulations of Roskar et al. (2008b). However, they found no such break in the mass surface density, in better agreement with Sanchez-Bl´ azquez´ et al. (2009). Follow-up studies showed similar behavior (Bakos et al. 2008; Gutierrez´ et al. 2011; Laine et al. 2014; Mart´ın-Navarro et al. 2014). Studies of outer disks using resolved stars also

81 frequently reveal a shorter scale length for younger, main-sequence populations than for red giant branch (RGB) stars (e.g. Davidge 2006; Vlajic´ et al. 2009; Radburn-Smith et al. 2011a), again implying outer disk populations dominated by old stars, although a break in surface brightness is not always present (as in the case of NGC 300 and NGC 7793; Bland-Hawthorn et al. 2005; Vlajic´ et al. 2009, 2011). This schema of radial migration hence appears to be mostly sound, yet exceptions do exist to complicate it. One obvious example is the existence of (again using the nomenclature of Erwin et al. 2008) “Type I” disks (a constant-slope exponential surface brightness profile) and “Type III” disks (an upbending break), both of which often show flat color profiles, at a relatively bluer color than the downbending “Type II” disks (Zheng et al. 2015). If the model proposed by Roskar et al. (2008b) is universal, such galaxies would require other mechanisms to shape their disks, such as the accretion of low-mass satellites (e.g. Younger et al. 2007). However, while Type III disks do show evidence of environmental influences (Laine et al. 2014), the correlation between environment and Type I and Type II disk fractions is unclear (e.g. Pohlen & Trujillo 2006). Extended ultraviolet (XUV) disks (Thilker et al. 2007a) also pose intriguing questions regarding the assumption of a star formation threshold, a seemingly necessary aspect of the models described above. Zaritsky & Christlein (2007) used Galaxy Evolution Explorer (GALEX) imaging to show that that perhaps as many as 50% of disk galaxies contain young

(< 400 Myr) star clusters between 1.25R25 and 2R25 (out to surface brightnesses of roughly

µV ∼29), which, given their number density and assuming a constant star formation history over 10 Gyr, could fully account for the measured surface brightnesses of starlight at these radii. This would imply that radial migration may not be universally necessary to build outer disks. Indeed, a recent spectroscopic study by Morelli et al. (2015) found mostly negative metallicity gradients in the oldest stellar populations in their sample of disk galaxies, difficult to explain in the context of significant radial migration. Addition- ally, disk metallicity gradients do not appear to be affected by the presence or absence of bars (Sanchez´ et al. 2014), in apparent contradiction to the simulations of Minchev et al. (2011, 2012, 2013). Satellite interactions can also drive evolution in the properties of disks. These interactions can transfer angular momentum to disk stars either directly from the satellite compan-

82 ion (e.g. Walker et al. 1996) or from disk gas driven inward by the effects of the interaction (e.g. Hernquist & Mihos 1995), leading to a radial spreading of the disk (Younger et al. 2007). Tidal stripping of the satellite companion can also deposit stars in the outer disk of the host as well (e.g. Stewart et al. 2009). Tidal encounters may also induce localized star formation in the disk outskirts (e.g. Whitmore & Schweizer 1995; Weilbacher et al. 2000; Smith et al. 2008; Powell et al. 2013), or potentially generate extremely extended spiral arms (Koribalski & Lopez-S´ anchez´ 2009; Khoperskov & Bertin 2015), which may then form new stars and drive migration of older stars further outward in the disk. The affects of accretion and tidal interaction, however, depend strongly on the orbital parameters of the encounter (e.g. mass ratios of the interacting galaxies, prograde vs. retrograde orbits, etc.; Toomre & Toomre 1972; Barnes 1988; Quinn et al. 1993; Walker et al. 1996; Bournaud et al. 2005; D’Onghia et al. 2010). This, in combination with progenitors with potentially different structural properties, star formation histories, and metallicity distributions, implies that the influence of accretion and interaction events on disk galaxies ought to be stochastic in nature. A common thread amongst most of the observational studies cited above, save for those using star counts, is the use of azimuthal averaging when constructing one-dimensional radial profiles of the galaxy light. Such studies measure the surface brightness or color of the disk in successively larger radial bins, thereby maintaining a high signal-to-noise (S/N) ratio even in the faint outer isophotes of the galaxy. This method has proven extremely useful for large statistical studies (e.g. Pohlen & Trujillo 2006; Bakos et al. 2008; Erwin et al. 2008; Mart´ın-Navarro et al. 2014; Zheng et al. 2015), but it suffers from a number of pitfalls when applied on a galaxy-by-galaxy basis. These techniques “average out” azimuthal asymmetries in the surface brightness and color of the outer disk, which often hold important clues about its dynamical history (see, e.g., the case of M101; Mihos et al. 2013b). These asymmetries may also skew the results of azimuthal averaging by mixing disk light with regions of blank background sky, a particular problem when radial bins with constant ellipticity and position angle are used at all radii (which is often the case; Pohlen & Tru- jillo 2006; Erwin et al. 2008). In some cases, inferences drawn from azimuthal averaging may even depend on the choice of metric used to construct the profile. For example, in the outskirts of an XUV disk, a luminosity-weighted average surface brightness will present a

83 much different story than an areal-weighted median surface brightness, as most of the light in the outer disk will be contained in just a few blue pockets of star formation (e.g. M83; Thilker et al. 2005). The exact importance of these various pitfalls to azimuthal averaging is not yet clear. Given these complications, more detailed studies of individual galaxies may provide important new tests for the current paradigm of disk galaxy evolution. For example, if weak spiral arms persist beyond the so-called truncation radius (e.g. Khoperskov & Bertin 2015), this may drive outer disk star formation (Bush et al. 2008) and lead to the radial growth of disks with time; such features may be washed out by an azimuthally averaged photometric analysis. Bright, nearby disk galaxies provide the best targets for such work; their proximity allows us to study them at high spatial resolution and also permits follow- up studies of their discrete stellar populations. While many spatially resolved studies have been done at high surface brightness (µB. 26) in the past (for just a few examples, see Schweizer 1976; Okamura 1978; Yuan & Grosbol 1981; Kennicutt & Edgar 1986; Tacconi & Young 1990), recent improvements in deep imaging techniques now allow us to probe the outer disks of these galaxies using similar techniques down to the much lower surface brightnesses characteristic of their extreme outer disks. Here we present deep surface photometry of three large nearby disk galaxies — M106, M94, and M64 — to explore the structure and stellar populations in their outer disks. Taken using the Burrell Schmidt Telescope at Kitt Peak National Observatory, our data reach lim- −2 iting surface brightnesses of µB∼28–30 mag arcsec in B and V, and we combine our data with extant GALEX and 21cm neutral hydrogen maps of each system to explore the efficacy of different formation mechanisms for outer disks. In Section 4.3 we present our observation and data reduction strategies, in Section 4.4 we discuss our methods for ex- tracting and analyzing the surface brightness and color profiles of the galaxies, in Section 4.5 we present and discuss our results on a galaxy-by-galaxy basis, in Section 4.6 we discuss the implications of these results in the context of galactic evolution, and in Section 4.7 we present a summary of our results and conclusions.

84 4.3 Observational Data

4.3.1 Deep Optical Imaging

We obtained deep broadband imaging of the galaxies M64, M94, and M106 using CWRU’s Burrell Schmidt Telescope at KPNO on moonless, photometric nights in Spring 2012 and Spring 2013. Our observing strategy and data reduction techniques are described in detail in Watkins et al. (2014, and references therein), and we repeat only the most important details here. The telescope’s field of view is 1◦.65×1◦.65, imaged onto a 4096×4096 STA0500A CCD, for a pixel scale of 100. 45 pixel−1. We observed in two filters: a modified Johnson B (2012), and Washington M (2013). The latter filter is a proxy for Johnson V; it is similar in width but ∼200Å bluer, and effectively cuts out diffuse airglow from the bright O I λ5577 line (Feldmeier et al. 2002). Each exposure was 1200 s in B and 900s in M, with ∼0◦.5 dithers between exposures to reduce contamination from large-scale artifacts such as scattered light and flat-fielding errors. For each galaxy, the total exposure times are as follows: for M94, 24 × 1200 s (B) and 32 × 900 s (M); for M64, 23 × 1200 s (B) and 30 × 900 s (M); for M106, 27 × 1200 s (B) and 38 × 900 s (M). Sky levels in each exposure were 700–900 ADU pixel−1 in B and 1200–1400 ADU pixel−1 in M. In addition to the object frames, we also observed offset blank sky pointings for use in constructing night-sky flats. We alternated between observing object frames and blank sky frames, in order to maintain similar observing conditions between the two and minimize systematic differences due to changes in telescope flexure and night-sky conditions. However, during data reduction, we found that the only measurable difference in flat fields constructed from the various subsets of sky frames (taken object by object or run by run) was a mild seasonal gradient that was easily corrected for (for details, see Section 2.2 in Watkins et al. 2014). Thus, in the end we constructed our final sky flat using all sky exposures taken throughout each observing season, resulting in ∼100 sky frames in B and ∼120 in M. Finally, during each season, we also observed Landolt standard fields (Landolt 1992) to derive color terms for each filter, along with deep images of Procyon (1200 s in B) and Regulus (900 s in M) to measure the extended point-spread function (PSF) and characterize reflections between the CCD, dewar window, and filter (Slater et al. 2009).

85 We begin the data reduction by first applying standard overscan and bias subtraction, then correcting for nonlinear chip response, and adding a world coordinate system (WCS) to each image. We derive photometric zero points for each image using SDSS DR8 (Aihara et al. 2011) stars located in the field, converting their ugriz magnitudes to Johnson B and V by adopting the prescription of Lupton (2005) and only using stars within the color range B − V= 0–1.5. We use these zero points and the color terms derived from the Landolt standard stars to convert our magnitudes into standard Johnson B and V magnitudes, which we use in all of our analyses throughout this paper. In our final mosaics, we are able to recover converted SDSS magnitudes of SDSS stars in-frame to σV = 0.03 and σB−V = 0.04 for all three galaxies. These are hence the absolute photometric uncertainties on any magnitudes and colors we quote in this paper; it should be noted that these include the intrinsic scatter both in the transformation between SDSS and Johnson photometric systems and in the transformation from our custom filters to the Johnson system. However, relative photometric uncertainties within a single mosaic are typically much lower than this at high surface brightness (σV < 0.01); at low surface brightness, the relative photometric uncertainty is dominated by uncertainty in the sky-subtracted background. In each mosaic, this background uncertainty (in the vicinity of each galaxy) is typically of order ±1ADU (∼ 0.1% of sky; see above), which implies a global limiting surface brightness of µB,lim ∼ 29.5, although local limiting surface brightnesses vary across each mosaic. The limiting surface brightness in the mosaic of M64 is significantly brighter than the other two (µB,lim ∼ 28.0) due to the presence of foreground Galactic cirrus; we discuss this in more depth in Section 4.5.2. We constructed flat fields in each filter using the offset night-sky frames. For each sky image, we applied an initial mask using IRAF’s1 objmask task, hand-masked any diffuse light missed by objmask (typically scattered light from stars located just off frame), and combined the images into a preliminary flat. We then flattened each sky frame using this preliminary flat, modeled and subtracted sky planes from each flattened image, and created a new flat from the sky-subtracted images. We repeated this step five times, at which point the resulting flat field converged. We then corrected these master flats for the seasonal

1IRAF is distributed by the National Optical Astronomy Observatory, which is operated by the Associa- tion of Universities for Research in Astronomy (AURA), Inc., under cooperative agreement with the National Science Foundation.

86 residual planes described above before applying them to the images. The last steps of the reduction process consist of star and sky subtraction, followed by final mosaicing. We first subtract the diffuse halos around bright stars following the technique of Slater et al. (2009). These halos arise from both the extended stellar PSF and reflections between the CCD, dewar window, and filter. We create a model for these halos by measuring them from our deep imaging of Procyon and Regulus and then scale and subtract these models from each star brighter than V=10.5 in each frame. We then mask each image of all bright stars and galaxies and fit planes to the diffuse night-sky background. After subtracting these planes from each image, the images are ready to be combined into the final mosaics. We use IRAF’s wregister and imcombine tasks to create these mosaics, using a median combine after scaling each image to a common photometric zero point. FWHM values of the stellar PSFs are nearly the same on all mosaics: ∼2.2 pixels, or 300. 2. This large value is a combination of registration error and seeing variations throughout the observing runs; FWHM on individual exposures is typically much smaller (∼100. 5–200). Once the final mosaics are complete, we also create masked and rebinned versions to improve S/N at low surface brightness and better reveal faint extended features in the outer regions of each galaxy. The masking process typically only masks pixels brighter than

µB ≈ 27 and µV ≈ 26 and leaves fainter pixels untouched. After masking, we rebin the mosaic into 9 × 9 pixel (1300× 1300) blocks and calculate the median in each block to create these “low surface brightness enhanced” mosaics.

4.3.2 Ancillary Multiwavelength Data Sets

To supplement our broadband imaging, and to study the star-forming properties and neutral hydrogen distribution in each of our galaxies, we have obtained ancillary ultraviolet and 21 cm radio data from a variety of sources. The ultraviolet data come from several diﬀerent surveys done by the GALEX (Martin et al. 2005) mission, downloaded from the GR6/GR7 data release.2 The far ultraviolet (FUV; 1350–1750 Å) emission traces recent (< 50 Myr) star formation, while near ultraviolet (NUV; 1750–2750 Å) emission traces slightly older (< 100 Myr) populations, along

2http://galex.stsci.edu/GR6/

87 with some contribution from evolved horizontal branch populations. We use the deepest available images for each galaxy, which come from different surveys. For M94, we used the Nearby Galaxies Survey (NGS; Bianchi et al. 2003); for M64, we used the Calibra- tion Imaging Survey, in which it appears serendipitously at the edge of the field containing the white dwarf WDST GD 153 0003 (as such, the M64 UV data are the shallowest); for M106, we used the Deep Imaging Survey. Additionally, we obtained H I data from The H I Nearby Galaxies Survey (THINGS; Walter et al. 2008) for M94 and M64 and from the Westerbork H I Survey of Irregular and SPiral Galaxies (WHISP; van der Hulst et al. 2001) for M106. We note that both surveys are interferometric and hence trace only the relatively high (> 1019 cm−2) column density H I. Extended diffuse H I in the outskirts of these galaxies — where we are most interested — may be systematically missed in such surveys.

4.4 Analysis Techniques

Our goal is to measure the spatially resolved properties of each galaxy without the need for complete azimuthal averaging. This allows us to study whether the disk outskirts are azimuthally mixed (as might arise from radial migration scenarios) or show signiﬁcant azimuthal variations (as might occur if outer disks are shaped by recent accretion events or stochastic star formation). Given the large angular size of our galaxies, as well as the low pixel-to-pixel noise in our ﬁnal mosaic images, we achieve high photometric accuracy

(σB−V ∼ 0.1 mag at µB = 27.5) over scales of ∼ 1 kpc. In analyzing each galaxy, we construct azimuthally distinct radial surface brightness and color proﬁles and decompose the azimuthal surface brightness variations of the disks into their low-order Fourier modes. We give details on each method here. All surface brightnesses are calculated using “asinh magnitudes” (Lupton et al. 1999), which are equivalent to regular magnitudes at high ﬂux levels, but better behaved at low S/N.

4.4.1 Surface Brightness and Color Proﬁles

For each galaxy, we measure the radial surface brightness and color proﬁles along six equal-angle radial wedges in the disk plane, increasing the radial width of each bin with

88 radius in order to preserve high S/N in the faint outer regions. By necessity, we use a constant position angle and ellipticity at each radial bin for these profiles; the galaxies often show significant variation in isophotal position angle and ellipticity, and interpretation of the profiles becomes extremely confused if the isophotal bins are allowed to wander. Still, care must be taken in interpreting each profile, particularly in the disk outskirts where background can begin to mix with starlight in portions of each wedge due to misalignment of the aperture with the true, frequently asymmetric isophotes. We also measure the azimuthally averaged surface brightness and color profiles for comparison. In this case, we do allow the isophotes to wander, which typically has minimal effect on the qualitative profile shape. However, this choice is occasionally non-negligible; for M106, we found that the surface brightness profiles in both the B and V bands show clear Type II (downbending) breaks at 55000 when using varying isophotal parameters, but that this break disappears completely when using fixed isophotal parameters. Choice of method thus changes the classification of M106’s exponential profile from Type II to Type I. In M94 and M64, profile breaks appear using either method, but the changes in slope are much less abrupt in each case when using fixed parameters over varying parameters. Using varying parameters for the azimuthally averaged profile, but fixed parameters for the angular profiles, also causes the azimuthally averaged values not to follow the average of all six profiles. In M64, for example, the azimuthally averaged profile favors the major axis due to changing ellipticity in the outer isophotes (a property we discuss in more detail in Section 4.5.2). In calculating the profiles, we use the median surface brightness of the pixels in each bin, as it is more robust against contamination from foreground stars or background galaxies. Comparing these profiles to those derived from the total flux in each radial bin shows significant differences in the inner regions, where spiral arms and H II regions dominate the light. In the outskirts, however, the median and luminosity-weighted profiles of the outer regions of each galaxy are nearly identical, due to our masking of bright sources in the disk outskirts. While this masking risks excluding light from bright star-forming knots in the disk outskirts, without high-resolution imaging it is often impossible to differentiate between background galaxies, foreground stars, and compact sources within the galaxy itself, and at such faint levels even one such stray source can dominate the total luminosity of a

89 given annulus or wedge. This is a known limitation of deep surface photometry (see, e.g., Bland-Hawthorn et al. 2005), and thus to avoid ambiguity with the background, we choose to measure the properties only of the diffuse starlight in the outer regions. To mollify the effects of masking, we compared our images of each galaxy by eye with GALEX FUV and NUV images to seek out, for example, potential extended star-forming regions, but found that such objects are rare. Hence, the populations we sample appear to be representative of the outer disk as a whole. Given that we perform surface photometry in the faint outskirts using our masked and median-binned images, we correct our surface brightness profiles using sky-subtracted background values measured from these binned images rather than the unbinned images. To measure these values, we place ∼50 equal-sized boxes in regions near each galaxy free of obvious contamination from unmasked sources (in M64’s case, this leaves few regions where we can accurately sample the sky due to the foreground cirrus contamination) and take the median of the median values of all boxes as the local sky. The sky uncertainty

is hence the dispersion in the medians, which is quite small (σsky ∼0.3 ADU). Within each box, typical pixel-to-pixel variance is found to be ∼1 ADU, with very little variation (σ ∼0.1 ADU) from box to box; hence, we subtract the same sky value from all proﬁles for a given galaxy.

4.4.2 Fourier Analysis

In addition to surface brightness and color profiles, we also conduct a Fourier mode analysis of the azimuthal surface brightness profiles of each galaxy, as a function of radius. This analysis is similar to that described by Zaritsky & Rix (1997), Mihos et al. (2013b), Zaritsky et al. (2013), and others to measure lopsidedness in galaxy disks. We decompose the azimuthal surface brightness profiles as a function of radius into Fourier modes:

X I(θ) = cos(mθ + φm) m

where I is the intensity, m is the Fourier mode, θ is the azimuth angle, and φm is the position angle of the mth Fourier mode. We measure both the m = 1 and m = 2 mode amplitudes in each galaxy, normalized

90 to the m = 0 mode (the mean surface brightness in the annulus), as a function of radius, again using annuli with constant position angle and ellipticity. Typically, m = 1 modes are indicative of galaxy lopsidedness, while m = 2 and higher modes are related to repeat- ing patterns such as bars or spiral arms. As such, a measurement of m = 1 power in the outer disk can be an indication of a tidal disturbance that has not had time to settle (but see Zaritsky et al. 2013), while a measurement of m = 2 power in the outer disk might indicate extended spiral patterns. However, m = 2 modes may also arise from misalignments between the photometric aperture and the true isophotal shape, due to asymmetries such as warps or tidal distortions in the disk. As such, visual inspection is necessary in interpreting this type of modal analysis to avoid drawing false conclusions.

4.5 Individual Galaxies

Here we present the results of our broadband imaging and surface photometry of these three galaxies. For reference, we present various global properties of these galaxies in Table 4.1. Figure 4.1 and subsequent figures show a comparison between our broadband imaging, GALEX FUV and NUV imaging, and THINGS and WHISP H I imaging (see Section 4.3.2). Our broadband imaging is shown in the upper left of the figures, with the intensity scale rewrapped over three ranges of brightness (µB< 24.6, 24.6 <µB< 26.5, and µB> 26.5) to highlight different regions. We show the unbinned, native resolution images inside of the

µB= 26.5 isophote, and the 9 × 9 binned images outside of this isophote in order to enhance faint, extended features. In the upper right of the figures, we show a B − V pixel-to-pixel color map of our broadband data (at native resolution only). The color bars on the right- hand sides give B−V values. UV data from GALEX is shown in the lower left of the figures (FUV in blue and NUV in yellow), while 21 cm emission is shown in the lower right. We overlay white ellipses of various semi-major axis length on each image, to provide a visual scale for the surface brightness and color profiles shown in Figure 4.2 and subsequent figures. Each ellipse uses the parameters (ellipticity and position angle) of the last best-fit isophote of the unbinned image and is labeled in arcseconds. We also plot two red lines to indicate the major and minor axes of these ellipses, labeled 0◦and 90◦, respectively, with

91 Table 4.1. M94, M64, and M106 Galaxy Properties

M94 M64 M106 NGC 4736 NGC 4826 NGC 4258

(1) R.A. (J2000) 12:50:53.0 12:56:43.6 12:18:57.5 (2) Decl. (J2000) +41:07:14 +21:40:59 +47:18:14 (3) Type (R)SA(r)ab (R)SA(rs)ab SAB(s)bc (4) Distance (Mpc) 4.2 a 4.7 b 7.6 c (5) M 0 −19.4 −19.5 −20.9 BT 0 (6) (B − V)T 0.72 0.71 0.55 8 d d f (7) MHI (10 M ) 4.00 5.48 35.9 (8) MHI /LB (M /L ,B) 0.045 0.056 0.101 (9) R25 (arcmin) 5.6 5.0 9.3 (10) R25 (kpc) 6.8 6.8 20.6 −1 d d g (11) W50 (km s ) 208.5 304.0 381 −1 d d e (12) SFRHα (M yr ) 0.43 0.82 3.82 (13) Scale (pc arcsec−1) 20.4 22.8 36.8

Note. — Rows are: R.A. and decl. (1,2), morphological type (3), adopted distance (4), absolute B magnitude (5), B − Vcolor (6), HI mass (7), HI mass per unit blue luminosity (8), µB= 25 isophotal radius in arcminutes (9) and kpc (10), HI line width (11), Hα star formation rate (12), and physical scale (13). All values come from the RC3 (de Vau- couleurs et al. 1991), except for those listed in the following footnotes: aRadburn-Smith et al. (2011b), bJacobs et al. (2009), cHumphreys et al. (2013), dWalter et al. (2008), eKennicutt (1998), fvan der Hulst et al. (2001), gTully et al. (2009).

92 0◦marking the position angle of the major axis. Figure 4.2 and subsequent figures show surface brightness and color profiles of the galaxies, plotted as a function of semi-major axis length (shown in arcseconds and kpc). The colored lines in the top left (B-band surface brightness) and bottom left (B − V color) panels of Figure 4.2 represent profiles measured along the corresponding colored wedges depicted in the inset schematic (solid lines indicate where the unbinned mosaic was used, and dashed lines indicate where the 9 × 9 binned mosaic was used). We overplot the azimuthally averaged profiles of each galaxy as well using black squares (unbinned data) and triangles (9 × 9 binned data). Characteristic error bars are also shown in each figure, dominated by the presence of faint, unmasked background sources. Because this is corre- lated scatter, the error in color is much less than the quadrature sum of errors in surface brightness (see Rudick et al. 2010a). We also include the radial FUV and NUV surface brightness (in AB magnitudes, shifted upward by 2 mag arcsec−2 to avoid stretching the ordinate axis of each graph) for comparison, measured using the same isophotes as the optical data: FUV is shown in purple and NUV is shown in gold, plotted only to where the FUV surface brightness begins to flatten into a constant background value. All surface brightnesses and colors have been corrected for foreground extinction using the extinction maps of Schlegel et al. (1998) as recalibrated by Schlafly & Finkbeiner (2011); we use the coefficients measured by Yuan et al. (2013) to derive the extinction in the two GALEX passbands. Additionally, we show Fourier m = 1 and m = 2 amplitudes (normalized to the m = 0 amplitudes; see Section 4.4.2) as a function of semi-major axis length in the right-hand panels of Figure 4.2 and subsequent figures. m = 1 amplitudes are shown in the upper right, with their corresponding azimuthal angle plotted just below, while m = 2 amplitudes and angles are shown in the bottom right. Angles are measured in the plane of each galaxy; 0◦thus marks the major axis at the galaxy’s position angle, increasing clockwise (shown by the red lines in Figure 4.1 and subsequent figures). Blue symbols are measured from the B-band images, and gold symbols are measured from the V-band images. Both bands typically show good agreement except in regions of low S/N.

93 M94 1.20 BBand B V 1.05 −

0.90 90◦ 90◦

0◦ 0◦ 0.75 510 510 240 240 0.60

0.45

850 850 0.30

1200 1200 0.15

0.00

FUV + NUV HI

1200 90◦ 1200 90◦

510 0◦ 510 0◦

240 240

850 850

Figure 4.1: Images of M94, with ellipses of various semi-major axis length overplotted for reference, labeled in arcseconds. Red lines indicate the major and minor axes of the outermost isophote of the unbinned image; labeled angles give degrees from the chosen galaxy position angle for the Fourier analysis discussed in Section 4.4.2. From the top left: (1) A subset of our B band mosaic, rescaled in intensity to highlight ranges of surface brightness (µB< 24.6, 24.6 <µB< 26.5, and µB> 26.5); the 9×9 median-binned image is shown outside of µB> 26.5 to enhance diﬀuse features. (2) B − V color map; B − V values are shown via the color bar on the right-hand side. (3) FUV and NUV false-color image constructed from GALEX data, speciﬁcally the Nearby Galaxy Survey (NGS; Bianchi et al. 2003); blue denotes FUV data, while yellow denotes NUV data. (4) H I image constructed from THINGS data (Walter et al. 2008); 1σ rms noise of this image is 2.6×1020 cm−2 (Walter et al. 2008). All four plots are shown at an identical angular scale.

94 r (kpc) r (kpc) 0.0 4.1 8.1 12.2 16.3 0.0 4.1 8.1 12.2 16.3 0.60 20 M94 0.45 0 /A

22 0.30 1 A

µB 0.15 24

µ 0.00

320◦ 26

240◦ 1 φ 28 160◦ µ 2 NUV − µ 2 80◦ FUV − 30 0 0 200 400 600 800 ◦ 0.60

0.9 0.45 0 /A

0.30 2 A 0.8 0.15 V

− 0.7 0.00 B 320◦

240 0.6 ◦ 2 φ 160◦

0.5 80◦

0 0 200 400 600 800 0 200 400 600 800 ◦ r (arcsec) r (arcsec)

Figure 4.2: Top Left: B, FUV, and NUV surface brightness profiles of M94. Triangles and squares are azimuthally averaged values; colored lines are averaged over the corresponding colored wedges shown in the inset. Triangles and solid lines denote where unbinned data are used, and squares and dashed lines denote where the 9 × 9 binned data are used (Section 4.3.1). Purple and gold lines show FUV and NUV profiles, respectively, shifted brighter by 2 mag to avoid stretching the y-axis, and cut off just before the FUV profile reaches the background value. All values are corrected for foreground extinction. Representative error bars are shown for µB=28, 29, and 30. Error bars are smaller than the point sizes for all brighter surface brightnesses. Bottom Left: B − V profiles of M94. Symbols and colors are the same as the top left plot, again corrected for extinction. UV colors are not plotted. Representative error bars are shown for µB=26, 27, and 28. Top Right: m = 1 amplitude normalized to m = 0 amplitude (top) and angle (bottom) of the azimuthal intensity profile as a function of semi-major axis length (see Section 4.4.2). Blue squares show values obtained from the B-band image, and gold triangles show values obtained from the V-band image. Bottom Right: m = 2 mode normalized amplitude and angle of azimuthal intensity profiles, as above. We assume a distance of 4.2 Mpc to M94 (Radburn-Smith et al. 2011b), for a disk scale length of 2.3 kpc (based on the outer disk). 95 4.5.1 M94 (NGC 4736)

M94 (NGC 4736) is part of the Canes Venatici I Cloud (Karachentsev 2005), a loose association of galaxies that may be expanding with the Hubble flow (Karachentsev et al. 2003). It is hence fairly isolated: Karachentsev & Kudrya (2014) list its nearest neighbor as IC 3687, a dwarf galaxy located at roughly the same distance as M94 (∼ 4.5 Mpc; Jacobs et al. 2009; Radburn-Smith et al. 2011b), but some ∼ 3◦ (∼ 240 kpc, in projection) away on the sky (see also Geller & Huchra 1983). The galaxy contains an outer star-forming structure, often referred to as a “ring”, at ∼ 20000 (4 kpc), which is offset in position angle from the bright inner disk. This ring is also visible in H I (e.g. Bosma et al. 1977; Mulder & van Driel 1993), where it appears as a set of irregular spiral arms at high column density (Walter et al. 2008, this is also true of its appearance in the UV; e.g. Trujillo et al. 2009). Despite this unusual morphology, the gas kinematics show a monotonic rotation curve from the center out (Bosma et al. 1977; Mulder & van Driel 1993); however, noncircular motions are prevalent throughout the disk at small spatial scales (Walter et al. 2008). There is evidence that these outer spiral arms, as well as a more strongly star-forming inner ring inside of 5000, are the result of Lindblad resonances (Gu et al. 1996, Trujillo et al. 2009; it is also interesting to note that the inner and outer rings both have approximately the same position angle; Mulder & van Driel 1993). Planetary nebula (PN) kinematics also show evidence of flaring in the old stellar populations (Herrmann et al. 2009; Herrmann & Ciardullo 2009) that may imply some past perturbation. Overall, the galaxy is difficult to characterize and shows many asymmetric features indicative of a possible recent interaction, despite its rather isolated neighborhood. Figure 4.1 shows these asymmetric features, all located inside of 50000(10 kpc). The outer spiral arms can be traced in our B − V color map, the GALEX images, and in the H I, and show strong north-south asymmetry and several kinks qualitatively similar to those found in the grand-design spiral arms of NGC 5194 (Dobbs et al. 2010), a galaxy known to be interacting with its S0 companion NGC 5195 (Toomre & Toomre 1972; Salo & Laurikainen 2000; Durrell et al. 2003). Beyond this radius, however, both the UV emission and 21cm emission drop off abruptly, leaving only the smooth optical isophotes. The outer disk profile continues dropping exponentially with no sign of any break out to at at least 20 kpc (∼ 9 outer disk scale lengths). While the azimuthally averaged surface brightness

96 proﬁle of the galaxy shows a mild ﬂattening in the last two points (at µB∼ 30), suggestive of transition into a smooth halo, this is in fact well modeled by a transition from the disk to the local background. We also identify a faint plume visible on the southwest side, the source of an upturn in surface brightness along the southwestern minor axis wedges (yellow and purple curves) shown in Figure 4.2; while this plume may be part of M94’s disk, it is too

faint (µB> 29) to constrain its color. M94’s B − V color profile (bottom left in Figure 4.2) shows a blueward color gradient in the disk between 20000 and 50000 (4–10 kpc), at which point the gradient reverses (this behavior can also be seen in the g − r profile shown by Trujillo et al. 2009, though we trace the red part of the profile some 20000 beyond the apparent limit of their data). The FUV and NUV profiles both show a spike in surface brightness at the tail end of the blue gradient, coincident with the outer spiral arms; beyond this point, however, the UV profiles truncate. This same behavior appears in the high column density H I gas as well, though lower column density gas may well exist beyond this radius. The six angular surface brightness and color profiles of the disk show little spread beyond this radius as well. Each color profile turns redward at roughly the same radius, albeit with varying degrees of sharpness. The southwest side of the galaxy shows the mildest gradients; from Figure 4.1, this is also where the spiral arms appear weakest. Otherwise, the average interquartile spread amongst all six wedges beyond the break in the color profile is only ∆µB ∼0.2 and ∆(B − V) ∼0.02. As such, it appears that the bulk of the recent star formation in the disk ends with the outer spiral arms; if star formation occurs beyond this −5 −1 −2 radius, it is either at extremely low levels (µFUV,AB < 5 × 10 M yr kpc ) or very localized such that we smooth over it in our angular bins as well. However, visual inspection of the FUV image and the B − V color map shows no obvious patches of new star formation in the outer disk. Additionally, the m = 2 amplitude weakens outside of this radius, lending credence to the idea of an azimuthally smooth outer disk. The strong m = 2 mode seen inside of 20000 (4 kpc) is driven by an offset between the inner and outer disk, specifically the gap between the inner disk and the outer spiral arms. We find some evidence of lopsidedness beyond 00 300 (6 kpc), though it is mild (A1/A0 ∼ 0.2). In the final radial bin, the m = 1 and m = 2 amplitudes peak sharply in the B band (and to some extent in V); however, the azimuthal

97 variations in surface brightness in low-S/N regions are sensitive to background fluctuations, and hence caution is warranted in their interpretation. That said, this final radial bin does encompass the southwestern plume, which may be driving some of the nonaxisymmetric power. Given the smoothness of the outer disk, it is worth discussing the distorted inner disk morphology in more detail. Inside of ∼20000 (4 kpc), the disk is offset significantly in position angle from the outer disk, suggestive of a tilt in inclination between the two. However, due in part to the noticeable offset between the H I kinematical major axis and the optical major axis of the inner disk (Bosma et al. 1977; Kormendy & Norman 1979), M94 has long been thought to host an oval distortion, a type of disk instability similar to a bar but larger in physical scale (a good overview of oval distortions can be found in Section 3.2 of Kor- mendy & Kennicutt 2004). Early density-wave models suggested that oval distortions may maintain spiral structure (this was proposed, but not explored, by Toomre 1969), acting in a manner very similar to bars (Kormendy & Norman 1979; Athanassoula 1980; Kormendy & Kennicutt 2004). In simulations, Trujillo et al. (2009) found that an oval distortion in the inner disk provided a good match to M94’s observed structure, lending additional support to the idea. We examine this idea again using our surface photometry. M94’s surface brightness profile is complex; Trujillo et al. (2009) stated that M94 may be considered an antitruncated disk, given the nearly flat profile in the outer spiral arm region (30000 – 40000, 6–8 kpc), unless the inner disk was truly an oval distortion, in which case it would be better classified as a single exponential. In contrast, Herrmann & Ciardullo (2009) used PN kinematics to explain this flattening as an increased importance of a thick-disk component in the outer regions. It is thus interesting to note that the surface brightness profile of the galaxy shows a larger inner disk scale length (0.8 kpc, measured between 8500 and 16500) along the minor axis (green and purple curves in Figure 4.2) than the major axis (0.6 kpc). This difference becomes most notable around 20000(4 kpc), coincident with the gap between the inner disk and outer spiral arms. Thus, M94’s surface brightness profile appears more cleanly exponential when the inner disk/outer disk gap is avoided. The smoothness of M94’s profile thus provides additional evidence in favor of a continuous stellar disk, favoring the oval distortion model over an outside accretion event to

98 explain the galaxy’s structure. This explanation may not contradict the results of Herrmann & Ciardullo (2009); resonances with bars or similar features can drive vertical heating and lead to larger scale heights (Schonrich¨ & Binney 2009, but see Minchev et al. 2012). While the mild lopsidedness of the outer isophotes may be evidence of recent interactions, lopsidedness is extremely common even in isolated galaxies (Zaritsky et al. 2013) and may in fact be a signature of misalignment between the stellar disk and dark matter halo. As such, the idea of M94 as a solitary, isolated galaxy evolving in an almost purely secular way appears sound. M94 may thus serve as a particularly interesting target for future studies investigating the eﬀect of secular processes such as radial migration on outer disks.

4.5.2 M64 (NGC 4826)

M64 (NGC 4826) is a nearby Sa galaxy, alternately referred to as the “Black Eye” or “Evil Eye” galaxy (also the “Sleeping Beauty” galaxy; Rubin 1994) due to the prominent dust lane near the bulge on the northeast side. Interest in the galaxy peaked when Braun et al. (1992) discovered that the inner (< 5000, 1 kpc) and outer (> 5000)H I disks counter- rotated with respect to each other. Subsequent observations by Rix et al. (1995) revealed that the entire stellar disk co-rotates with the inner gas disk, implying that the outer gas disk is an accretion relic. Detailed study of the H I (Rubin 1994) and CO (Garc´ıa-Burillo et al. 2003) showed evidence of shocks and radial inflow in the inner disk originating from the boundary between the two counter-rotating systems. Indeed, simulations reproduce such inflows, as angular momentum cancellation at the boundary between the counter-rotating components leads to gas infall (Quach et al. 2015). The galaxy also shows evidence of both leading and trailing stellar spiral arms (Walterbos et al. 1994), suggestive of some disturbance in the stellar disk as well, but nonetheless the disk shows an extremely regular exponential surface brightness profile out to ∼30000. Despite the galaxy’s counter-rotation and overall low H I surface densities, the H I rotation curve is quite regular as well (de Blok et al. 2008; Walter et al. 2008). We show our imaging of M64 in Figure 4.3, along with the radial surface brightness, color, and Fourier profiles in Figure 4.4. Unfortunately, M64 lies in a region of the sky rife with contamination by foreground Galactic dust (or ‘cirrus’), which severely limits the depth of our mosaics of this galaxy. Whereas we can probe surface brightnesses of µB∼ 30

99 M64 1.20 BBand 90◦ B V 90◦ 1.05 −

0.90 0◦ 0◦ 590 590 0.75 340 340

0.60 170 170 0.45

0.30 840 840

0.15

0.00

FUV + NUV 90◦ HI 90◦

590 590 0◦ 0◦

840 840

170 170

340 340

Figure 4.3: Imaging of M64, using the same layout as Figure 4.1. GALEX data are from the Calibration Imaging survey (CAI); M64 appears serendipitously near the edge of the ﬁeld, and exposure times diﬀer between NUV and FUV (∼7000 s and ∼1000 s, respectively). The 1σ rms noise of the H I image is 3.4×1020 cm−2 (Walter et al. 2008).

100 r (kpc) r (kpc) 0.0 4.6 9.1 13.7 18.2 0.0 4.6 9.1 13.7 18.2 0.60 M64 20 0.45 0 /A

0.30 1 A 22

µB 0.15

µ 0.00 24 320◦

240◦ µNUV 2 26 −

µFUV 2 1

− φ 160◦

28 80◦

0 0 200 400 600 800 ◦ 0.60

1.0 0.45 0 /A

0.30 2 A 0.9 0.15 V

− 0.00 B 0.8 320◦

240◦ 2 φ 0.7 160◦

80◦

0.6 0 0 200 400 600 800 0 200 400 600 800 ◦ r (arcsec) r (arcsec)

Figure 4.4: Photometric analysis of M64, using the same layout as Figure 4.2. The Fourier analysis stops at a smaller radius than the photometric profiles due to contamination from asymmetrically distributed foreground dust (see text). The limiting NUV surface brightness is much lower than the limiting FUV surface brightness due to the difference in exposure times between the two filters. We show representative surface brightness error bars only for µB=27 and 28 for this galaxy, as the limiting surface brightness for this galaxy is µB∼28.5 (see text). We show the same representative color error bars as Figure 4.2, however. We assume a distance of 4.7 Mpc to M64 (Jacobs et al. 2009), for a disk scale length of 1.4 kpc (based on the inner disk).

101 in M94 and M106, here we begin transitioning into the background at µB∼ 28.5 due to the surrounding cirrus. Because of the asymmetric nature of the cirrus, we cannot measure the Fourier modes of the disk beyond 60000, where the dust begins to significantly skew the analysis. Spiral structure is most obviously seen in our B − V color map, and appears constrained to within roughly 20000(4.5 kpc). The disk color profile is also generally much flatter than M94’s. From Figure 4.4, we see a mild blueward gradient between 10000 and 20000 (∼2– 4.5 kpc), beyond which the profile flattens out at B − V∼0.75. A single mild (∼0.02 mag) blueward dip appears at ∼45000 (9.5 kpc), and the color trends continuously redder beyond this radius. The majority of the UV light is found within the spiral arms; both the FUV and 00 NUV profiles show a steep decline with radius, reaching µNUV ∼30 by ∼200 (4.5 kpc). Within this radius, the H I shows weak spiral structure as well, while at larger radius the gas distribution is quite irregular and patchy, with extremely low column density (< 1020cm−2; de Blok et al. 2008). The rapid decline in UV surface brightness and very low outer H I column density both argue that recent star formation in M64 is constrained to the inner 4.5 kpc. The presence of both leading and trailing spiral arms in this galaxy (Walterbos et al. 1994) implies a disturbed stellar disk as well, and indeed, distorted isophotes are visible just outside of 17000(4 kpc), and, more weakly, just outside of 34000(7 kpc) as a slight protrusion on the galaxy’s west side. The angular profiles also show more scatter beyond ∼17500, with mild but significant power in the m = 1 and m = 2 modes at these radii as well. Close examination of the images shows that most of the m = 2 power comes from misalignment of the elliptical aperture with the galaxy isophotes, rather than due to any spiral structure. The slow angular slewing of the m = 2 mode thus implies a gradual shift in the outer isophotes’ position angle with respect to the photometric aperture. Indeed, from our ellipse analysis, we find that the position angle steadily increases by ∼ 20◦ between 20000 and 70000(4.5–15 kpc). The most notable feature we find in M64, however, is the dramatic antitruncation of the profile beginning around 40000 (9 kpc). We note that the same break is seen in the R-band profile of Gutierrez´ et al. (2011). The immediate concern is that this feature is induced by the foreground cirrus; however, a battery of tests demonstrates that this is not

102 the case. While the cirrus contamination seems severe in the northwest and southeast sides

of our optical image, its surface brightness in these regions lies around µB= 28.0–28.5. As this Type III upbending break begins roughly 2 magn brighter than this level, it cannot be caused simply by a transition to this background. Additionally, the break can be seen along every angular cut at the same radius; this would only be true if the cirrus were evenly distributed around the galaxy, which Figure 4.3 shows is not true. The effect of the cirrus on the galaxy’s surface brightness profile thus appears to be mild; indeed, it seems strongest only at the largest radii, where the northwest profiles (blue and red curves) flatten

off most quickly, at the expected background levels (µB∼ 28.0). The eastern major axis profile (cyan) shows the lowest surface brightness and is also the wedge with the least cirrus contamination. This antitruncation thus appears inherent to the galaxy. Inside of the break radius (be- 00 00 tween 200 and 400 ) the interquartile spread between the six angular profiles is ∆µB ∼0.2 and ∆(B − V) ∼0.01, indicating a very uniform stellar population despite the asymmetry in the disk. Beyond the profile break, all the angular profiles trend redward, save for the northern (red) wedge. The reason for the discrepant northern wedge is not evident, even upon close examination of the image. It may be related to the cirrus; we measure a color of B − V∼0.7 in the relatively bright patch just to the northwest of M64, which is also the color at which the northern profile flattens out. While showing some patchiness, the mean color of the background near M64 (which includes contributions from cirrus, unresolved background sources, and residual sky variance) is B − V=0.85. Thus, we tested the cirrus’s influence on all of the color profiles using a simple model—a screen with a uniform surface brightness of µB= 28 and a uniform color of B − V=0.8 overlaid atop a model galaxy with a similar surface brightness and color profile as M64—and found that it begins to affect 00 M64’s color profile at a surface brightness of µB∼27.0, or a radius of ∼450 . As such, it appears that the redward color gradients beyond ∼45000 should be attributed to the foreground cirrus, and not to changes in stellar populations within the galaxy. We thus do not consider the color profile beyond ∼45000 for the remainder of our analysis. If the upbending profile seen in Figure 4.4 is due to a distinct outer disk, it may have been spawned from the interaction that led to the counter-rotating kinematics seen in M64. If so, its red colors rule out, for M64, models where antitruncated outer disks are built

103 through induced star formation in extended gas (a mechanism suggested by Laine et al. 2014) and instead favor scenarios where angular momentum exchange during a merger mi- grates stars into the disk outskirts and forms a Type III break (e.g., Younger et al. 2007). However, the proﬁle presents problems for the latter scenario as well. The models of

Younger et al. (2007) display mild breaks (hout/hin = 1.2 − 1.8) that occur at relatively small radius (Rb/hin = 2.5 − 4), while the break we see in M64 happens at much larger radius (Rb/hin = 6) and is significantly more dramatic (hout/hin = 10). Furthermore, from our ellipse analysis, outside of the break M64’s outer isophotes become much rounder (b/a = 0.7) than in the inner disk (b/a = 0.5), which would not arise from simple radial spreading of the disk. Taken together, these arguments suggest that we are not seeing an outer disk formed through angular momentum transfer during an accretion event, but instead a profile transitioning from a disk component to an outer halo. This alternative interpretation of upbending profiles was also proposed by Mart´ın- Navarro et al. (2014), who argued that no true Type III disks exist, and that upbending profiles simply signal the presence of a stellar halo. Under the more detailed classification scheme proposed by Pohlen & Trujillo (2006), this would include the Type III-s galaxies (spheroidal): galaxies with a Type III break that show progressively rounder isophotes beyond the break radius. To test this idea further regarding M64, we fit a variety of models to its surface brightness profile using emcee, a Python-based Markov Chain Monte Carlo sampling algorithm (Foreman-Mackey et al. 2013). While a double-exponential model provides a good fit to the data, we found nearly equally good fits for a disk+power-law model with a power-law slope of α ∼-2, or disk+Sersic´ profiles with indexes n ∼0.5 and 4. Hence, in the end we find that the fits do not provide discrimination amongst the various models. Given this, the rounding of the outer isophotes and the poor match to models of outer disk formation lead us to prefer the disk+halo interpretation for M64’s overall profile. With all the evidence that M64 has suffered a recent merger, do we see signatures in its photometric structure? While the galaxy shows some lopsidedness within r =20000(4.5 kpc; Figure 4.4), we see no obvious tidal features in the outer disk that might signal a past accretion event, although the foreground cirrus precludes us from probing the faintest levels. While infalling companions can cause significant vertical heating in disks (e.g. Barnes 1988; Toth & Ostriker 1992; Stewart et al. 2008; Kannan et al. 2015), it is less

104 clear how such accretion could have heated M64’s disk to form the smooth, diffuse outer spheroid while also depositing a thin H I disk into the rotational plane. This argues that the outer spheroidal component may predate the merger, rather than forming during the event. M64 thus appears to be a case of merger-induced quenching, where the counter-rotating accretion has disrupted the galaxies’ gaseous disk and shut off the bulk of star formation throughout the system, while leaving the stellar disk largely intact. In this way, we may be witnessing the end stages of M64’s transition into an S0 galaxy (e.g. Borlaff et al. 2014).

4.5.3 M106 (NGC 4258)

M106 (NGC 4258) is the brightest member of the Canes Venatici II Group (Fouque´ et al. 1992) and can be considered a Milky Way analog given its similar luminosity, Hubble type, and local environment (e.g. Kim et al. 2011). While the central regions of M106 have been studied extensively (specifically, to determine the origin of a pair of offset “anoma- lous” spiral arms; Courtes` & Cruvellier 1961; van der Kruit et al. 1972; Pietsch et al. 1994), previous studies on the outer disk are relatively scarce. M106 is the most massive galaxy of the three examined in this paper, and it is also the only one of the three with clearly visible satellites. The galaxy NGC 4248, located to the northwest of M106 (labeled in Figure 4.5), has long been suspected to be a satellite (van Albada 1977), although previous studies found no clear evidence of interaction (e.g. van der Kruit 1979). A Tully–Fisher distance places the galaxy at 7.4 Mpc (Karachentsev et al. 2013), essentially the same distance as M106. The galaxy to the southeast, UGC 7356 (also labeled in Figure 4.5), is also a companion, of much lower mass (Jacobs et al. 2009; Spencer et al. 2014). The velocity spread between the three galaxies is comparable to M106’s rotation velocity, suggesting that the companions are bound to M106. Spencer et al. (2014) also find 7 additional probable satellites of low mass (−12 > MV > −17) within 200 kpc (projected) of M106, indicating a fairly rich local environment. We present multiwavelength images of M106 in Figure 4.5 and the photometric profiles in Figure 4.6. M106 has more prominent UV emission than either M94 or M64, reflecting a higher SFR; indeed, M106’s Hα-derived SFR is 3.82 M yr−1 (Kennicutt 1998), compared to 0.43 M yr−1 and 0.82 M yr−1 for M94 and M64, respectively (Walter et al. 2008). A rough estimate using the RC3 colors and the B − V to M/L conversion factors of Bell & de

105 M106 1.20

1190 1190 BBand B V 1.05 − 0◦ 0◦ NGC4248 0.90

90◦ 90◦ 0.75 670 670

290 290 0.60

0.45

0.30

960 960 0.15 UGC7356

0.00

FUV + NUV HI

0◦ 0◦

90◦ 670 90◦ 670

960 960

290 290

1190 1190

Figure 4.5: Same as Figure 4.1, but for M106. GALEX data is from the Deep Imaging Survey (DIS), and H I data are from the WHISP survey (van der Hulst et al. 2001, 3000 resolution shown); 1σ rms noise is ∼ 1020cm−2. The companion galaxies NGC 4248 and UGC 7356 are labeled in the B-band image (upper left).

106 r (kpc) r (kpc) 0.0 7.4 14.7 22.1 29.5 36.8 44.2 0.0 7.4 14.7 22.1 29.5 36.8 44.2 0.60 20 M106 0.45 0 /A 22 0.30 1 A

µB 0.15 24

µ 0.00

320◦

26 µNUV 2 − µ 2 240◦ FUV − 1 φ 160◦ 28

80◦

30 0 0 200 400 600 800 1000 1200 ◦ 0.60

0.9 0.45 0 /A

0.30 2 A

0.8 0.15 V

− 0.00 B 320◦ 0.7

240◦ 2 φ 160◦ 0.6

80◦

0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 ◦ r (arcsec) r (arcsec)

Figure 4.6: Same as Figure 4.2, but for M106. We assume a distance of 7.6 Mpc to M106 (Humphreys et al. 2013), for a disk scale length of 6.0 kpc (based on the inner disk).

107 1.20 BBand B V − 1.05

0.90

0.75

0.60

0.45

0.30

0.15

0.00

Figure 4.7: Close-up of NGC 4248. Left: B-band, shown at three levels of brightness (µB< 23.5, 23.5 <µB< 25.5, and µB> 25.5) to highlight the warped inner isophotes, the tidal features on the galaxy’s north and south sides, and the very boxy outer isophotes. Right: B − V color map, showing the centrally concentrated star formation.

Jong (2001) also implies that M106 has a specific SFR roughly a factor of three higher than the other two galaxies. Thilker et al. (2007a) classified M106 as a Type 1 XUV disk, defined as a galaxy containing highly structured UV complexes beyond where the UV-derived SFR −4 −1 −2 surface density drops below ΣSFR = 3 × 10 M yr kpc . Again, our B − V colors trace the UV emission well; of note are the extended spiral arms beyond 30000 (10 kpc), and the plume of UV emission and H I on the galaxy’s south side (near the companion UGC 7356). This plume, as well as another on the galaxy’s north side (including some of the diffuse UV light inside of 60000), drives the XUV disk classification. Three diffuse patches of UV emission can also be found tracing the extremely faint tail of H I extending from the galaxy’s east side (seen crossing the 96500 ellipse in Figure 4.5) and are also visible as small patches of diffuse light in our optical imaging (we note that these are not visible in Figure 4.5, as each patch is only a few arcseconds in radius and quite faint). We see a sharp decline in the UV emission beyond the XUV disk, at roughly 60000 (22 kpc), yet the optical light continues to show an exponential profile well beyond this radius. From Figure 4.6, the B − V color profile turns sharply redward at this radius as well. As with M94 (and to some extent M64), all angular cuts show the same behavior;

108 the spread amongst the cuts, however, is much greater than in M94 or M64: ∆µB ∼0.5 and ∆(B − V) ∼0.07. This is due to irregular isophotes; the minima in the color profiles occur at increasingly large radius from the northeast (red) curve to the west (purple) curve. This trend follows the morphology of the spiral arms, which are stronger on the east side of the galaxy. The northmost (blue) profile breaks the pattern; however, it is clear from Figure 4.5 that the spiral structure is much weaker along this cut. These red disk outskirts extend to at least 40 kpc, some 20 kpc beyond the apparent UV cutoff radius in this XUV galaxy, and again appear devoid of strong spiral structure: note the decline in the m = 2 amplitude beyond ∼60000 (22 kpc), as seen in Figure 4.6. The two bright dwarf companions present one very clear difference between M106 and the other two spirals in our study. Our deep imaging shows clear evidence of tidal distortion in the brighter companion NGC 4248 (Figure 4.7): its low surface brightness outer isophotes are extremely boxy and offset in position angle from the dwarf’s inner regions by almost 45◦. In this aspect, it appears quite similar to NGC 205, a satellite companion to M31 of comparable luminosity. Like NGC 4248, NGC 205 also shows recent star formation very near its center (Cappellari et al. 1999), along with isophotal twisting (Choi et al. 2002) and a strongly boxy morphology in its outer regions as well (seen in an image taken by S. van den Bergh, presented in Kormendy 1982). A study of RGB star kinematics in NGC 205’s outskirts revealed strong high-velocity tails and a reversal in the direction of rotation beyond 1 kpc, indicative of tidal interactions with M31 (Geha et al. 2006). The boxiness in NGC 4248’s outer isophotes appears to result from an overlap of an extended elliptical component with symmetric tidal features extending from the north and south sides of the galaxy. Though NGC 4248 is tentatively classified as irregular (de Vaucouleurs et al. 1991), its warped inner isophotes (Figure 4.7) and rotating H I (van Albada 1977) imply a disklike structure. This galaxy may thus serve as an example of a low-mass disk being tidally transformed into a dwarf elliptical more akin to NGC 205, an intriguing idea in the context of M106 as a Milky Way (or M31) analog. In turn, the effect of the companions on M106’s outer disk may be seen in the disk’s visible warp, seen both in the H I kinematics (van Eymeren et al. 2011) and in the generally high m = 1 amplitude at all radii in this galaxy (Figure 4.6). The m = 1 mode peaks around 60000, near the outer spiral arm radius; this peak is due to the southern plume near

109 UGC 7356. Plumes of diffuse starlight and H I are also present on the galaxy’s north and south sides along the major axis; the west half of the southernmost plume shows fingers of UV light and blue colors indicating induced star formation, and its proximity to UGC 7356 is suggestive of tidal interaction with that companion (although proximity to tidal features is not an unambiguous indication that a satellite has generated the disturbance; a clear counterexample is the southern tail of M51, generated by the interaction with its companion on the galaxy’s north side; Rots et al. 1990; Salo & Laurikainen 2000). As stated in Section 4.4.1, M106’s classification as either a Type II (downbending) or Type I (unbroken) disk changes depending on the method used to measure its surface brightness profile. This can be seen in the behavior of the angular cuts in Figure 4.6, in that the eastern side of the galaxy (red and green profiles), where the spiral arms appear weakest, shows a profile contiguous with the outer disk, while the spiral arms induce an excess of light over this profile in all of the other angular cuts that appears as a Type II break. This behavior hearkens to the study by Laine et al. (2014), who found that Type II disk breaks tend to follow morphological features such as spiral arms, lenses, or rings. The severity of the break thus depends on how closely such features are followed in the isophotal analysis; in some cases the choice of isophotes can mask the presence of a disk break entirely or introduce one where none exists. These effects are most obvious in bluer wavelengths due to the presence of high-mass stars in the star-forming regions along spiral arms. In the near-infrared (a fairly robust tracer of stellar mass; Sheth et al. 2010), Laine et al. (2014) note similarity between the inner disk scale lengths of Type I and Type III (antitruncated) disks and the outer disk scale length in Type II disks. Given that M106’s profile appears unbroken absent the presence of any spiral arms, one might postulate that the term “break” is a misnomer and that the inner disks of Type II galaxies are actually elevated in surface brightness over the baseline outer disk due to, e.g., recent star formation. This would imply that the differences between Type I and II disks are purely morphological, rather than the product of different formation histories; Type III disks would thus be the true outliers, which is consistent with their comparitive rarity (only ∼20%–30% of disks show Type III breaks; Erwin et al. 2008; Laine et al. 2014). Examining the very outermost regions of M106’s disk, we again see a leveling off of the azimuthally averaged surface brightness profile in the last few data points; however, as

110 with M94 it appears to be well modeled simply by a transition into the local background. However, we note that the profiles that follow the galaxy’s minor axis (red, green, purple, and yellow) are elevated in surface brightness above the major axis at this extended radius. The patchy diffuse light that can be seen outside of 96000 in Figure 4.5 likely accounts for this behavior. Given the asymmetry in M106’s isophotes, it is unclear whether this patchy light is part of M106’s outer disk or instead represents an inner halo or thick disk; regardless, we can confidently state that M106’s disk extends to at least 110000 (40 kpc, or ∼6.5 disk scale lengths), making it nearly twice as large in physical size as either M94 or M64. The tidal features in this galaxy — the southern and northern optical plumes and weak H I tails — are most likely to have originated through tidal interactions with its nearby, bound companions, rather than through a flyby interaction with a more massive companion. A stronger encounter would likely induce more dramatic tidal response, but we see no

evidence of elongated tidal tails in M106’s vicinity to a limiting surface brightness of µB=

29.5. The nearest bright galaxy to M106 is NGC 4144 (MB ∼ −18, assuming a distance of 7.5 Mpc; de Vaucouleurs et al. 1991; Seth et al. 2005, but see Jacobs et al. 2009), located some 240 kpc from M106 on the sky (Karachentsev & Kudrya 2014). Given this separation and NGC 4144’s relatively low luminosity, M106’s nearby companions certainly have the strongest tidal influence, and NGC 4248’s boxy outer isophotes confirm that it is tidally interacting with M106 at some level. While strong encounters tend to drive centrally concentrated starburst activity (e.g. Barnes & Hernquist 1991; Hernquist & Weil 1992; Hernquist & Mihos 1995; Cox et al. 2008; Hopkins et al. 2009b; Powell et al. 2013; Moreno et al. 2015), weaker tidal interactions with satellite galaxies may incite a less dramatic but longer-lived response in the disk outskirts as they orbit the primary over longer timescales. While these low-mass interactions may be less efficient at inducing star formation throughout the host (Cox et al. 2008), even a weak starburst in the outer disk may transform the structure and stellar populations of these low surface brightness regions. It may thus be interesting to consider the possibility that NGC 4248 (and to a lesser extent UGC 7356) may be shepherding the gas in M106 in such a way as to produce these outer spiral arms and, at least potentially, trigger star formation in the otherwise low-density outer H I. However, compared to the total extent of

111 the disk, the star formation in M106 is not greatly extended; while H I is present at large radius (Wolfinger et al. 2013), only within 20 kpc (∼3 disk scale lengths) is the gas dense enough to form stars. This stands in contrast to the case in the nearby face-on spiral M101, where interactions with its nearby companions have triggered star formation in the galaxy’s diffuse outer disk (Waller et al. 1997; Mihos et al. 2013b). Why, then, was star formation triggered in the outskirts of M101, but seemingly not in M106? The answer may lie in the fact that in addition to driving tidal resonances in galaxy disks, interactions can also drive nonplanar responses including warps and disk heating, which have the potential to shut down star formation. The relative efficacy of these different processes depends not only on the mass ratio of the encounter but also on the orbital properties of the encounter. While M101 has a single close satellite (NGC 5477), the galaxy’s marked asymmetry and its H I kinematics both argue for a single prograde encounter with the more massive and distant companion galaxy NGC 5474 (Mihos et al. 2012, 2013b). In contrast, M106 has two close companions, one of which (NGC 4248) is more massive than M101’s close satellite NGC 5477. If the orbital geometry of these satellites is highly nonplanar, the two working in concert may tip the dynamical balance toward disk heating rather than tidal compression, suppressing star formation in the outer disk. M106 thus may be an interesting test case concering the influence of fairly massive dwarf satellite galaxies on the star-forming properties of the host, which may be of particular interest in Milky Way studies given the presence of the Magellanic Clouds.

4.6 Discussion

Despite different local environmental conditions and interaction histories, we see consistent behavior in the photometric properties of these three galaxies’ outer disks. In M106 and M94, the onsets of redward gradients in their color profiles correspond to truncations in the UV surface brightness and 21cm emission tracing high column density H I gas. In M64, the UV emission is constrained to the central disk, and the colors flatten beyond the UV truncation to a similarly red color. The high column-density H I is more extended in this galaxy than in the other two, but is globally at much lower density and hence non-star- forming. What is consistent across all three galaxies is a lack of strong azimuthal color

112 variation in the outer disks, with the interquartile spread in color beyond the break radius always < 0.1 mag (and significantly less in the cases of M94 and M64). Spiral features also seem to vanish beyond the UV truncation radius. All three galaxies show only mild evidence of azimuthal asymmetry in their outer isophotes, the strongest present in M106, with no evidence of faint extended spiral structure. In these three galaxies, at least, this appears to imply a natural division between the “inner” and “outer” disks; “outer” disks may be defined as the region beyond any evident spiral features and devoid of new star formation, yet still following an exponential surface brightness profile. Here we address the constraints placed by our deep surface photometry on the stellar populations in these outer disks and compare to studies of outer disk populations in other galaxies. We also consider the role local environment plays in shaping each galaxy’s outer disk; in tandem with the inferred stellar populations, these constraints can provide useful clues to the formation and evolutionary histories of outer disks.

4.6.1 Outer Disk Stellar Populations

The similarity in outer disk colors for each galaxy implies a similarity in stellar populations. In all three galaxies, the outer disks display B−Vcolors of approximately 0.75–0.8 at a surface brightness of µB∼ 27.5. These colors appear robust against the color uncertainty, which is dominated by fluctuation in the background of the order σB−V ∼ ±0.1 mag at these surface brightnesses (see Section 4.3.1). These colors also appear independent of the mean background color (as introduced by faint cirrus, unresolved background sources, and residual sky variance); while the background near M64 is fairly red (B − V= 0.85), near both M106 and M94 it is significantly bluer, B − V= 0.4–0.5. Indeed, we find similarly red colors in the outskirts of three other disk galaxies we recently studied3 — M96, M95 (Watkins et al. 2014), and M51 (Watkins et al. 2015) — making this color of B − V= 0.8 a natural anchor point from which to study the outer disk populations of our galaxies. While broadband colors suffer from the well-known age-metallicity degeneracy (Worthey 1994), these colors can still place some constraints on the outer disk stellar populations. As a

3We note, however, that in the outer disk of M101, we find significantly bluer colors (B − V= 0.3–0.5; Mihos et al. 2013b), in congruence with the galaxy’s classification as an XUV disk (Thilker et al. 2007a). This argues that red outer disk colors are not a universal and systematic artifact of our instrumental setup, such as the extended wings of the PSF. See the Appendix for more details.

113 fiducial reference, B − V= 0.8 is a typical integrated color of an S0a-type galaxy (Roberts & Haynes 1994), implying a fairly evolved population. To explore population constraints in more detail, we model the integrated colors of stellar populations built via a variety of star formation histories and metallicities, constructed using the software SMpy, a Python-based SED modeling code based on the Bruzual & Charlot (2003) population synthesis models (described in Duncan & Conselice 2015). We constrain these models using the surface brightness and color of the outer disks, as well as the upper limits on their inferred SFRs. At the radius where the disk colors reach B−V= 0.8, we do not detect significant FUV flux in any of the galaxies; in M106 and M94 this places a limit on the SFR of .(3–5)×10−5 M yr−1 kpc−2. While the limit is higher for M64 due to the FUV image’s short exposure time (∼ 10−4 M yr−1 kpc−2), it is not so high as to significantly alter our conclusions for this galaxy. Applying these constraints, we run models using both exponentially declining histories (SFR(t) ∝ e−t/τ) and delayed exponential histories (SFR(t) ∝ te−t/τ; Lee et al. 2010; Schaerer et al. 2013). We adopt varying decay rates (τ) and metallicities for a 10 Gyr time span, assuming a Chabrier (2003) IMF. A constant star formation history is ruled out by the low current SFR; the time span required to build enough stars to match the total V-band luminosity within the µB= 27.5 annulus in each galaxy is > 20 Gyr. Between exponential and delayed exponential histories, consistent behavior emerged: for solar metallicity and below, current-day colors become too blue if τ & 2 Gyr, signifying a stellar population dominated by old stars. Metallicities below [Fe/H] ∼ -0.7 are ruled out, as these populations produce colors that are too blue regardless of the choice of τ. While these colors suggest old and only moderately metal poor ([Fe/H > -0.7) populations, significant ambiguity remains due to the age-metallicity degeneracy inherent in broadband colors. How then do these results compare to other studies of the outskirts of nearby disks using resolved stars, which more directly probe the ages and metallicities of stellar populations? Resolved imaging studies show a variety of stellar populations present in the outskirts of disk galaxies, indicative of diverse star-forming histories. For example, the outskirts of NGC 300 and NGC 7793 are populated almost entirely by RGB stars (Vlajic´ et al. 2009, 2011), while a sizable AGB population was found in the outskirts of NGC 2403 and M33

114 (Davidge 2003; Barker et al. 2007). M83, a galaxy known to have highly extended star formation (Thilker et al. 2005; Bigiel et al. 2010a), contains RGB, AGB, and red supergiant (RSG) stars in its outskirts (Davidge 2010). The inferred metallicities of resolved outer disk populations also show significant variation from galaxy to galaxy. In NGC 300, Vlajic´ et al. (2009) found evidence for a metallicity gradient in the outer disk, with metallicities spanning the range [Fe/H]= -0.5 to −1.0. Similarly low metallicities ([Fe/H]∼ -1.0) have been discovered in the resolved outer disk populations of NGC 2403 and M33 (Davidge 2003; Barker et al. 2007), and even lower metallicites are inferred in NGC 7793’s outer disk ([Fe/H]∼ −1.5 Vlajic´ et al. 2011). Such low metallicities, if characteristic of outer disk populations in general, would lead to colors much bluer than we find in M106, M94, and M64, even taking into account age effects. However, those studies focused on fairly low mass systems, smaller in mass than the three galaxies in this study (estimated from their maximum rotation velocities listed in the HYPERLEDA catalog; Makarov et al. 2014), though NGC 2403 is very near in mass to M94 and M64. If we assume that outer disk populations follow their host galaxies’ behavior on the well-known galaxy mass-metallicity relationship (e.g. Tremonti et al. 2004), these galaxies would have a higher mean metallicity in their outskirts. Indeed, higher metallicities are inferred in the outer disk populations of the bright spirals M31 ([M/H]∼ -0.3 to -0.5; Worthey et al. 2005, Gregersen et al. 2015), M81 ([M/H]∼ -0.4 to -0.7; Williams et al. 2009), and M83 (metallicities ranging from ∼20% solar to nearly solar; Davidge 2010). At these metallicities, the integrated colors of the disk would be significantly redder, in line with our deep surface photometry presented here. We also note that metal-rich populations in resolved star studies can be systematically missed due to the faintness of the metal-rich RGB (e.g. Rejkuba et al. 2005; Harris et al. 2007), complicating comparisons between those studies and deep surface photometry. It may thus be of interest to do more studies directly comparing integrated light colors with resolved photometry in order to better constrain the biases inherent in both methods. We discuss one such bias, the galaxy PSF’s influence on our measured colors in the outer isophotes, in the Appendix, though we believe it to be small within µB< 27.5. The uniformly red colors, azimuthally smooth distribution, and inferred old, moderately (but not extremely) metal-poor stellar populations at large radius in these galaxies thus

115 place constraints on the formation history of their outer disks. Disk building via continual low-level star formation in the outer disk appears ruled out: such a model would lead to much bluer colors than we observe, and the current rate of star formation is too low to build the amount of light we see in the disk outskirts in a Hubble time. Instead, radial migration (Roskar et al. 2008b) emerges as the most likely candidate for disk building at large radius, given the red stellar populations in all three galaxy outskirts, as well as the U-shaped color profiles in the two galaxies still actively forming stars in their inner regions. That said, it is unclear just how far out radial migration can drive stellar populations. Radial migration requires the presence of nonaxisymmetric structure such as bars or spiral arms (Sellwood & Binney 2002); significant migration into the outer disk would require the same mechanisms (Minchev et al. 2012; Roskarˇ et al. 2012). In these three galaxies, we find stars extending out to 3–4 scale lengths beyond the edge of the spiral arms; if this behavior is common in other galaxies, it may present a challenge for disk migration models as well. While additional spreading of the outer disk may arise from transient, tidally driven outer spirals, the galaxies studied here live in fairly low density environments and display no such tidal features. We therefore look forward to new dynamical modeling of disk galaxies that will examine these issues in more detail.

4.6.2 Environmental Inﬂuences

Under the hierarchical accretion paradigm, galaxy disks are built continually over time, from the inside out, as material from the surrounding environment (both baryonic and not) continually bombards the disk. This accretion can grow disks by depositing stars in disk outskirts (Stewart et al. 2009), triggering extended disk star formation (Whitmore & Schweizer 1995; Weilbacher et al. 2000; Smith et al. 2008; Powell et al. 2013), or moving stars outward through tidal heating or radial migration (Roskar et al. 2008b; Koribalski & Lopez-S´ anchez´ 2009; Khoperskov & Bertin 2015). Yet regardless of how outer disks are built, one would expect to see signatures of this process in these very faint, highly extended regions, where dynamical times are long and material is more loosely bound. However, in the three galaxies studied here, evidence for such accretion signatures is lacking. In M94, the most isolated galaxy of the three, we see only mild lopsidedness in its isophotes, and one extremely faint plume at the very outer edge of the disk, implying a much less chaotic

116 formation history. While M64’s past interaction history seems to have greatly damaged the gaseous disk, driving H I inward and shutting oﬀ disk-wide star formation, we again see no evidence for discrete tidal streams. Finally, in the case of M106, a large and luminous disk galaxy with two known and many more suspected satellites, the tidal signatures we do observe are rather weak and likely driven by the two luminous satellites. While we see no strong tidal features in any of the galaxies studied here, the similarity in the star-forming properties and stellar populations of their outer disks raises the question of what role, if any, the local environments might play in shaping their outer disks. On large scales, the environments of the three galaxies are similiarly devoid of massive companions. Within 1 Mpc, M94 has only a handful of neighbors, all signiﬁcantly lower in luminosity (the most luminous being NGC 4365, with MV ∼ −18; de Vaucouleurs et al. 1991; Jacobs et al. 2009). M64 may be even more isolated than M94; its brightest neighbor

is the dwarf NGC 4789A, with MV ∼ −14 (de Vaucouleurs et al. 1991; Jacobs et al. 2009). M106 resides in a somewhat richer environment, with several modestly bright companions within 1 Mpc, although none approach M106 in luminosity. M94 and M64 thus might be considered extremely isolated, while M106 resides in a moderately denser but still fairly sparse environment, more similar to the Local Group (though with no massive companion analogous to M31). On smaller scales, however, the local environments of the galaxies do appear different. Looking for satellite galaxies with ∼100 kpc, a scale comparable to the Milky Way’s satellite system, neither M94 nor M64 has luminous satellites, while M106 has the two mentioned previously: NGC 4248 and UGC 7356. We thus find three levels of environmental influence amongst these three galaxies: M94, being very isolated and apparently undisturbed (Figure 4.1), may be evolving purely secularly; M64, while also very isolated, likely suffered a recent merger that greatly affected its own morphology and star-forming properties (Section 4.5.2); and M106, living in a denser environment, is presumably being influenced most by its dwarf companions. Despite their different local environment, M94 and M106 both have similar outer disk structure: a set of extended, star-forming outer spiral arms, beyond which the disk is smoothly distributed and contains an old stellar population. If M94’s outer spiral arms are formed secularly, and if M106’s outer spiral arms are formed via weak interactions, this

117 implies two very different paths toward a qualitatively similar result. We see no sign of a strongly perturbed outer disk in M106, despite the presence of its satellites. This contrasts with the case of M81, which is similar in luminosity and SFR to M106 (Kennicutt 1998), but where its two more massive companions M82 and NGC 3077 have disrupted its disk outskirts (van der Hulst 1979; Yun et al. 1994; Okamoto et al. 2015). The effect of the satellites on M106’s disk appears much gentler — the galaxy may lie in something of a sweet spot, with companions massive enough to drive spiral structure (Weinberg 1995; Oh et al. 2008; Choi et al. 2015) and mediate radial migration to build the outer disk, but not massive enough to significantly disrupt it once formed. The situation for M64 is somewhat more muddled. At first glance, the presence of a Type III upbending break in a post-merger galaxy is consistent with the idea that Type III breaks are driven by strong interactions (Laine et al. 2014) or accretion events (Younger et al. 2007). However, as argued in Section 4.5.2, the properties of M64’s outer component are a poor match for either the induced star formation model or the angular momentum transfer model (Younger et al. 2007). Instead, the changing photometric profile is better explained as a disk–halo transition. That said, the halo is relatively bright: with µB∼27 at 10 kpc, it is significantly higher in surface brightness than that of the Milky Way or M31 (Morrison 1993; Gilbert et al. 2012). If this outer profile is indeed a simply a stellar halo, then in M64 we are seeing a smooth and largely unbroken Type I exponential disk extending all the way out to where it becomes lost in the halo light, at 6 disk scale lengths. The disk is red and azimuthally smooth, save for the very inner regions where some residual star formation continues. With star formation otherwise quenched in the galaxy, M64 may be in the process of becoming an S0 galaxy. Its surface brightness profile is in fact remakably similar to that of ESO 383- 45, an S0 galaxy also suspected of having suffered a recent merger (Kemp et al. 2005). S0 galaxies show antitruncations more frequently than other disk types (Borlaff et al. 2014; Maltby et al. 2015); if mergers drive evolutionary transition from spirals to S0, they may also lead to “spheroidal” antitruncations (denoted Type III-s breaks in Pohlen & Trujillo 2006) by growing the galaxy’s halo component. In these cases, however, the halo-like component forming the antitruncations would by necessity be a different kind of halo than that surrounding the Milky Way, which appears to

118 have been built up over time via satellite disruption rather than from heating of the stellar disk (e.g. Morrison et al. 2000; Bullock & Johnston 2005; Cooper et al. 2010; Ma 2015). Stellar populations in the spheroid beyond the profile breaks would also appear very similar to those in the disk (where they originated), which would explain the relatively flat color profiles such antitruncated galaxies (including M64) typically exhibit (Zheng et al. 2015). Also, if the halo-like component arose due to heating of the thin disk, and no new thin disk formed from an existing gaseous disk, the antitruncation would also appear in the mass profile of the galaxy; this in fact seems to be the case for Type III disks generally (Bakos et al. 2008; Zheng et al. 2015). If merger-spawned spheroids are the root cause of these antitruncated profiles, such galaxies also should appear more frequently in dense environments (Laine et al. 2014) either because of the heightened rate of interactions or simply because of the morphology–density relationship raising the likelihood that galaxies will have significant halo components. However, the fact that M64 is apparently quite isolated serves to demonstrate that a dense local environment is not a necessary condition for their formation — one merger event may be sufficient. Finally, the fact that the outer disks of these three galaxies consist of predominantly old and well-mixed populations may simply reflect their host galaxies’ local environment. All three galaxies live in low-density regions — even the group environment of M106 is sparse, with no large companion galaxies nearby. Weak interactions with low-mass satellites may not be sufficient to trigger widespread star formation in outer disks; hence, a denser group environment may be more conducive to triggering outer disk star formation. However, even in denser groups the evidence is mixed: M101, the dominant galaxy of its dynamically active group, shows young blue populations in its outer disk (Mihos et al. 2012, 2013b), but in the Leo group, the spirals M95 and M96 both show red outskirts (Watkins et al. 2015). This ambiguity is present in larger surveys as well; Maltby et al. (2012), for example, found little difference in outer disk structure between field and cluster galaxies, while Erwin et al. (2012) found significant differences between field and cluster S0 galaxies (including a complete lack of disk truncations in cluster S0 galaxies). Roediger et al. (2012) found that cluster disk galaxies are distributed equally amongst the three disk break types, a significant difference compared to field galaxies (Pohlen & Trujillo 2006), with significant U-shaped age gradients present in all three types, in apparent contradiction to the photometric results

119 of Bakos et al. (2008). Some conflict thus appears to be present regarding environmental influence on outer disk evolution, which may be partially resolved if the immediate, local environment is in fact the driving influence, rather than the global environment.

4.7 Summary

We have performed deep surface photometry (µB=28–30) of the nearby galaxies M94, M64, and M106, and incorporated archival UV and 21cm H I data to probe the formation histories of the galaxies’ outer disks. All three galaxies exhibit red outer disks beyond a radius corresponding to a truncation in star forming activity and high column density H I gas in the disk. A Fourier analysis of the azimuthal surface brightness and color profiles of each galaxy’s outer disk shows that these components are smooth and well mixed, devoid of spiral arms or significant nonaxisymmetric structure. New star formation in M94 is truncated at ∼ 10kpc, beyond which the disk appears azimuthally smooth and red but for some mild lopsidedness. The stellar disk, which seems to be continuous despite the offset inner and outer isophotes, extends to at least ∼20kpc, or ∼9 scale lengths, with no emergence of a stellar halo down to a surface brightness of

µB∼30. Given M94’s isolation and smooth undisturbed outer disk, our data favor secularly driven radial migration of disk populations to explain the galaxy’s outer structure. This, combined with its relatively close distance (∼4 Mpc) makes M94 an ideal test bed for follow-up studies investigating how secular evolution processes such as radial migration affect outer disk formation. M64 shows a stark star formation truncation only a few kiloparsecs from the center, with a low H I column density beyond this radius and a sharp antitruncation in the stellar surface brightness beginning around 40000 (9kpc). We trace this antitruncated disk to ∼19kpc, or ∼13 inner disk scale lengths. M64’s strongly antitruncated profile is likely the signature of a transition from the galaxy’s disk to its diffuse stellar halo rather than being a true upbending of the disk surface brightness profile. The recent merger event in M64 appears to have disrupted its gas disk and truncated star formation in all but the inner few kiloparsecs, leading to the galaxy’s very flat and red color profile. M64 thus appears to be undergoing a transition from a spiral to an S0 galaxy, an interesting example of merger-driven galaxy

120 transformation in an otherwise isolated environment. Despite elevated levels of star formation, M106 still shows a clear star formation truncation radius associated with the end of its outer spiral arms at ∼60000 (22kpc). Its stellar disk extends roughly twice this distance beyond this truncation radius, with signs of interaction with its two brightest companion galaxies. We trace M106’s stellar disk to ∼40 kpc, or ∼6.5 scale lengths. Although M106 possesses a more robust satellite system than M64 or M94, its smooth outer disk and fairly weak tidal structure argue that these satellites are not dramatically reshaping the disk — instead, they may have helped drive the outward migration of stars in M106’s disk without completely disrupting the disk outskirts. M106 may serve as an interesting comparison to the Mily Way’s own satellite-driven evolution, given the similarity in morphology, luminosity, and local environment between the two galaxies. The red colors of these galaxies’ outer disks (B − V∼0.8 in their outermost regions) imply predominantly old stellar populations. For exponentially declining star formation histories, colors this red cannot be achieved for decay rates longer than τ=2 Gyr and cannot be achieved for any τ if metallicities are below [Fe/H]= -0.7. These properties, along with the smoothness of the outer disks, suggest that these parts of the galaxies are not formed through ongoing or sporadic star formation, but rather dynamical processes such as heating or radial migration of stars from the inner disk. The lack of a significant young stellar population in these galaxies’ outskirts may reflect the sparseness of their local environment; stronger or repeated encounters may be needed to trigger widespread and sustained star formation in outer disks. Additional studies of the detailed stellar populations in outer disks over a wider range of environment would be informative. However, while all three of the galaxies studied here do live in low-density environments, they also appear to have different interaction histories. In this sense, it is interesting that similarly old and smooth stellar populations exist in the outer disks of each galaxy irrespective of the influence of their local environments and recent interaction history — secular processes that operate in a completely isolated galaxy (M94) produce a very similar looking outer disk population to those in a galaxy interacting with companions (M106) or recovering from a recent merger (M64). Furthermore, the large physical extent of these azimuthally smooth outer disks implies a very high efficiency with which stars can be trans- ported via radial migration; whether such extended disks can be built this way remains

121 unclear. Finally, while red outer disk colors and U-shaped color proﬁles are frequently cited as evidence of radial migration processes (Bakos et al. 2008; Mart´ın-Navarro et al. 2014; Zheng et al. 2015), broadband colors leave a great deal of ambiguity regarding the actual stellar populations producing them. Ambiguity is present even in many resolved population studies; without a halo ﬁeld to compare to, for example, halo star contamination fractions in outer disk studies remain unconstrained. Measuring stellar kinematics in these extended regions would be ideal to break the disk/halo ambiguity; however, this remains infeasible for galaxies beyond ∼1 Mpc. Until such studies are possible, combining data from low- resolution, deep surface photometry (to derive the morphology and integrated properties of extended regions in galaxies) with resolved star studies (to deconstruct the detailed stellar populations and star formation histories of these regions) seems the best option for future studies of outer disks.

4.8 Acknowledgements

This work has been supported by a Jason J. Nassau Graduate Fellowship to A.E.W., and by the National Science Foundation through award 1108964 to J.C.M. This work made use of Numpy, SciPy (Oliphant 2007), and MatPlotlib (Hunter 2007). This work made use of THINGS, ‘The HI Nearby Galaxy Survey’ (Walter et al. 2008). Figures 4.1, 4.3, 4.5, and 4.7 made use of Min-Su Shin’s publicly available code img scale.py4. We would also like to thank Ken Duncan for the use of his code SMpy. Facilities: CWRU:Schmidt—The Burrell Schmidt of the Warner and Swasey Observa- tory, Case Western Reserve University.

4.9 Appendix: ON THE BURRELL SCHMIDT PSF

As studies such as this one begin to breach lower and lower surface brightness limits, concerns about instrumental artifacts become much more important. Specifically, these concerns focus on the influence of scattered light, in the form of internal reflections and

4http://dept.astro.lsa.umich.edu/∼msshin/science/code/Python ﬁts image/

122 the extended wings of the PSF, which can skew radial surface brightness and color profiles at low surface brightness (for a recent discussion of this problem, see Sandin 2014, 2015). While aggressive antireflection coatings on both the filters and dewar windows of the Burrell Schmidt minimize bright reflections in our data, the effect of the extended PSF still remains at a low level, which we explore in more detail here. Before embarking on quantitative tests of the influence of the PSF on the derived photometric profiles, we first note that we have seen no evidence of systematic reddening of galaxy profiles in previous studies using the Burrell Schmidt. While the three galaxies studied here show red outer disks, Schmidt imaging of the spiral galaxy M101 (Mihos et al. 2013b) revealed blue outer isophotes, while deep imaging of the Virgo Cluster showed blueward gradients in the diffuse outer halos of the massive ellipticals M87 (Rudick et al. 2010a) and M49 (Mihos et al. 2013a). Redward gradients therefore do not seem to be a systematic result of our imaging techniques. However, to assess this effect more quantitatively, in this appendix we convolve an updated and more accurate measurement of the Burrell Schmidt PSF with a variety of galaxy profiles to quantify its effect on the extracted surface brightness and color profiles of our galaxies. As discussed briefly in Section 4.3.1, we measured the Burrell Schmidt PSF using long exposures of bright stars in order to subtract bright reflection halos and the extended PSF wings around bright stars. We show the Burrell Schmidt PSF radial profile in the B and V bands in Figure 4.8 out to one degree. An earlier measurement of the V-band PSF radial profile was published in Slater et al. (2009, hereafter S09) and Janowiecki et al. (2010); in the outer wings, this more recent measurement compares well to the older profile, despite being taken several years later and using a different CCD. We note, however, that the inner core profile shown in Figure 4.8 does differ significantly from that of S09. That earlier study focused on proper subtraction of the outer wings of the PSF, which is insensitive to the shape of the inner core. As such, the core profile of S09 was largely illustrative and not well determined. In the present study we have worked to produce a much more accurate measurement of the PSF core (r .1000) by using four stars of different brightnesses to ensure that all pieces of the profile link up correctly. Our updated profile shows that the core profile illustrated in S09 had actually been underestimated significantly. While this difference has little effect on scaled subtraction of the outer PSF, it has a dramatic effect on

123 0 B Band PSF V Band PSF + 3

5 ) 2 − , mag arcsec

0 10 µ − µ

20 Relative Surface Brightness (

0.1 1.0 10.0 60 Radius (arcmin)

Figure 4.8: The B- and V-Band (offset by 3 mag for clarity) radial profiles of the Burrell Schmidt PSF. Solid lines show the profiles including reflections, as measured from our bright star exposures, while dashed lines show the underlying profile wings (with reflections subtracted out). Profiles are normalized such that µ0 = 0.

124 the normalized PSF used for image convolution. This is simply due to the fact that there is a range of ∼22 mag in brightness between the core and the PSF wings at r ∼1◦; hence, most of the flux is contained within the core. As such, an error in the core profile can create significant variation in the intensity of the wings after normalization. With a more accurate PSF measurement in hand, we tested its influence on the derived photometric profiles by convolving several model galaxies with our normalized PSFs in both bands and measuring the resulting surface brightness and color profiles. The model galaxies were constructed to represent idealized versions (smooth exponential disks of constant B − V color, with bulges of a constant redder color) of M106, M64, and M101, in order to test the influence of the galaxies’ angular sizes and central surface brightnesses. The

M106 and M64 models had similar values of µ0, but diﬀerent scale lengths (see Figures

4.4 and 4.6), while the M101 model had a much lower value of µ0 and large angular size (M101 is nearly face-on; see Mihos et al. 2013b). We found that for the M106 and M64 facsimiles, the PSF induces a color change of

∆B − V∼ +0.1 by a surface brightness of µB∼28.0. This change in color occurs beyond where our photometry is noise limited by 0.5 mag arcsec−2; brighter than this surface brightness, the color change induced by the PSF is much smaller than that seen in the data.

For example, between µB∼25.5 and 26.5, M106 shows a color change of 0.08 mag, while the convolved model galaxy shows a change of only 0.015 mag. Thus, while some of the redward gradient in these galaxies’ outer disks may be attributable to the PSF, it is clear that most of the gradient is attributable to changing stellar populations. It should also be noted that, despite its relatively smaller angular size, we see no evidence that the PSF is inducing the antitruncation seen in M64’s outer disk; a signiﬁcant PSF-induced antitruncation is

only seen in the convolved model of M64 beyond µB∼30. Finally, in the M101 facsimile, we see the same 0.1 mag color change setting in, but at a much lower surface brightness

of µB∼30.0. This is simply due to M101’s lower central surface brightness, which scatters less light to large radius in the PSF. Taken as a whole, the results of these various tests thus show that the scientific results presented in this paper (and in previous papers using data taken with the Burrell Schmidt) are robust to PSF influence; the error budget is dominated by photometric uncertainties quantified in Sections 4.3 and 4.4.

125 Chapter 5

HII Regions and Diﬀuse Ionized Gas Throughout the M101 Group: Only the Intensity Changes

126 A New Perspective on Galaxy Evolution From the Low Density Outskirts of Galaxies

Abstract

AARON EMERY WATKINS

We present a multiwavelength study of star formation within the nearby M101 Group, including new deep Hα imaging of M101 and its two companions. We perform a statistical analysis of the Hα to FUV flux ratio in H II regions located in three different environments: M101’s inner disk, M101’s outer disk, and the lower mass galaxy NGC 5474. We find that, once bulk radial trends in extinction are taken into account, both the median and scatter

in FHα/FFUV in H II regions is constant with environment. We also study the behavior

of the diﬀuse ionized gas (DIG), and ﬁnd that unlike the H II regions, FHα/FFUV in the

DIG does show distinct trends with environment. We show that FHα/FFUV in the DIG correlates with H II region ﬂux surface density in high density environments, but not in low density environments where H I is the dominant gas phase. We propose that in low density environments, star formation is weak enough and sparse enough that it cannot ionize the ISM on large scales, leaving the H I intact; this explains why H I correlates with the star formation rate in such environments, but not in higher density environments. In total, because H II region populations appear invariant with environment but for luminosity, and because the properties of the DIG seem to depend only on local star formation rate density, we see no need to invoke changes to the star formation physics (such as e.g. truncations in upper mass end of the IMF) to explain the properties of star formation across the entirety of the M101 Group.

127 5.1 Introduction

The extended, low surface brightness (LSB) outer disks of galaxies are a poor fit to idealized models of galaxy formation theory. Absent extenuating circumstances, ΛCDM predicts that galaxies form “inside-out”, hence are youngest at their largest radii. Yet real galaxies’ often smooth, red outer isophotes imply the opposite (e.g. Bakos et al. 2008; Zheng et al. 2015; Laine et al. 2016). In fact, old red giant branch (RGB) stars typically have longer scale lengths than main sequence stars (e.g. Davidge 2003; Vlajic´ et al. 2009, 2011), and any young stars present in outer disks tend to be sparsely distributed (e.g. Barker et al. 2007; Davidge 2010). Outer disks are not simply an LSB continuation of inner disks. Star formation is also inefficient in outer disks, with gas consumption timescales ex- ceeding a Hubble time (Thilker et al. 2007a; Bigiel et al. 2010b). This is similar to LSB galaxies (e.g. McGaugh & Bothun 1994; Burkholder et al. 2001; Boissier et al. 2008), suggesting that star formation physics changes in low density environments. Jeans stability criteria suggest that low gas column density results in depressed or truncated star formation 20 21 −2 (with an apparent threshold below around ΣHI ∼ 10 –10 cm , e.g. Hunter & Gallagher 1986; Skillman 1987; van der Hulst et al. 1987), but star formation may also be suppressed on large scales via dynamically induced stability (e.g. Zasov & Simakov 1988; Kennicutt 1989). The latter suggests disks should have a star formation truncation radius (Martin & Kennicutt 2001), with star formation taking place beyond this only in local high density pockets (e.g. Courtes` & Cruvellier 1961; Ferguson et al. 1998a; Gil de Paz et al. 2005; Thilker et al. 2005). Despite its scarcity and inefficiency, this in situ outer disk star formation could fully account for all of the outer disk stellar mass in some galaxies (depending on the star formation history, SFH; Zaritsky & Christlein 2007). However, outer disk star formation is present in only ∼4%–14% of star-forming galaxies out to z = 0.05 (Lemonias et al. 2011), hence it may not be sufficient to explain outer disk formation in general. It also may not be necessary: many authors have proposed that much outer disk stellar mass can be accounted for through radial migration (Sellwood & Binney 2002; Debattista et al. 2006), which can migrate early generations of inner disk stars outward via resonances with transient spiral arms, bars, or couplings thereof (e.g. Roskar et al. 2008a; Sanchez-Bl´ azquez´ et al. 2009; Schonrich¨ & Binney 2009; Minchev et al. 2011; Roskarˇ et al. 2012).

128 Because our empirical star formation laws (e.g. the conversion of Hα flux to star formation rate, SFR; Kennicutt et al. 1994) were derived in high density environments, accounting for the fraction of stellar mass that formed in-situ in outer disks assumes that these laws remain unaltered in low density environments. If this is not true, conclusions drawn from typical star formation indicators about gas consumption timescales, star formation efficiency, and so on will be erroneous in outer disks and other similar environments. Consider, for example, two star-forming regions of equal mass and age, and so equal in predicted SFR. Hα emission is sensitive to the initial mass function (IMF, e.g. Sullivan et al. 2004); hence, if one region lacks massive O stars, it will emit fewer ionizing photons, resulting in lower Hα flux. Measuring its SFR using a standard Hα–SFR conversion factor will thus underestimate its true SFR. It remains an open question if star formation physics changes in low density environments. Whether or not such a change occurs depends on whether or not changes in the underlying structure of the disk—surface mass density, gas velocity dispersion, gas phase, turbulence, etc.—affects the formation and subsequent evolution of molecular clouds and star clusters. For example, (Meurer et al. 2009) argue that the formation of dense bound clusters is inhibited in regions of low mass surface density because the midplane pressure in the disk influences internal cloud pressures (see e.g. Dopita & Sutherland 2003). If massive stars form via competitive accretion (Larson 1973), in which interactions between protostars drive mass segregation and subsequent gas accretion in high density cluster cores, protostars in low-density clusters would suffer fewer interactions and accrete less mass, inhibiting the growth of high-mass stars (e.g. Bonnell et al. 2004). Seeking out changes to the IMF in populations of young clusters could thus help determine how sensitive star formation within dense cores and molecular clouds is to the surrounding environment. Some evidence does indicate that the cloud-to-cloud physics of star formation may be influenced by the local surface density of the disk. In inner disks, star formation follows a α power law of the form ΣSFR ∝ Σgas with the measured value of α ranging between ∼1 and 1.5 (as originally proposed by Schmidt 1959, and subsequently confirmed observationally, e.g. Kennicutt 1989, 1998; Kennicutt et al. 2007; Bigiel et al. 2008). Such studies have been much rarer in outer disks and other low density environments, partly because of lack of CO emission (likely due to low metallicity or changes in ISM pressure; Elmegreen & Hunter

129 2015). However, those that have broached this regime find a significantly steeper value of α (∼2–3; Bigiel et al. 2008, 2010b; Bolatto et al. 2011; Schruba et al. 2011), implying different physical conditions for star formation than found in the inner disk. Clues to this difference may come from dwarf irregular (dIrr) or LSB galaxies, which, like outer disks, are often gas-dominated and low in mass surface density (McGaugh & de Blok 1997; van Zee et al. 1997; Hunter et al. 2011). Stellar and gaseous disks in dIrr galaxies are also thicker than normal spirals (Elmegreen & Hunter 2015), which can help stabilize them (Vandervoort 1970); outer disks may again be similar, as they are frequently warped (Sancisi 1976; van der Kruit 1987; Bottema et al. 1987; Garc´ıa-Ruiz et al. 2002; van Eymeren et al. 2011). In a case study of the dIrr Sextans A, Hunter & Plummer (1996) found that stars still form at a slow rate in the peaks of the gas distribution even though dynamical arguments suggest this should not be the case (e.g. Toomre 1964; Kennicutt 1989). van Zee et al. (e.g. 1997) found similar results for six additional LSB dwarf galaxies. These galaxies lack interaction signatures, hence van Zee et al. (1997) proposed that star formation therein is likely regulated by feedback, such as stellar winds or supernovae, locally compressing gas. Such a mechanism may be necessary to sustain star formation in environments that lack the periodic forcing provided by spiral arms or bars, which may also be absent in outer disks (Watkins et al. 2016). One might thus consider whether these differing mechanisms yield observationally distinct populations of young clusters and H II regions. This is currently a topic of considerable discussion, and some previous studies have uncovered hints to this effect. Hoversten & Glazebrook (2008), for example, found that integrated colors of dwarf and LSB galaxies suggest a deficiency in high-mass stars; this may be related to their low integrated SFRs (Gunawardhana et al. 2011). A lack of high-mass stars may also account for the lack of high-luminosity H II regions in dwarfs and LSB galaxies (Helmboldt et al. 2005, 2009). Yet Schombert et al. (2013) found that when all 54 LSB galaxies in their sample were taken as a whole, the H II region luminosity function (LF) was the same as that found in normal spirals, hence the lack of bright H II regions in LSB galaxies could be merely a sampling effect given the intrinsic rareness of high-luminosity H II regions in general. One means of informing this debate is to compare and contrast different star formation tracers. SFR conversion factors assume the following: that stars are sampled from a univer-

130 sal IMF, that the SFH is constant over Gyr timescales, and that there is no attenuation by dust (Kennicutt 1983; Donas et al. 1987). Under those assumptions, different SF indicators should yield identical SFRs. Conversely, if different SF indicators yield different SFRs, one or more of those assumptions must be invalid. For example, when properly accounting for dust, Hα emission traces mainly O stars with masses M∗ & 10M , while far ultraviolet (FUV) emission traces O and B stars down to M∗ ∼ 3M (Kennicutt & Evans 2012); hence, variation in the Hα to FUV flux ratio (hereafter FHα/FFUV) can be used to study the behavior of the high mass end of the IMF in young clusters (e.g. Lee et al. 2009). This ratio also shows trends that may hint at environmentally dependent star formation physics: globally averaged FHα/FFUV correlates with galaxy stellar mass (Boselli et al. 2009; Lee et al. 2009), with R band surface brightness (Meurer et al. 2009, but see Weisz et al. 2012), and with radius in some galaxies (Thilker et al. 2005; Goddard et al. 2010;

Hunter et al. 2010). Unfortunately, FHα/FFUV is sensitive to a large number of variables, which makes interpretation of these trends diﬃcult. In addition to dust extinction (in fact,

FHα/FFUV correlates extremely well with extinction, to the point that it can itself be used as an extinction estimator; Cortese et al. 2006; Koyama et al. 2015), FHα/FFUV decreases rapidly with age (e.g. Leroy et al. 2012) as the high-mass stars traced by Hα emission die off. IMF sampling effects play a similar role, and introduce stochasticity in Hα emission at low mass, where a given H II region may be powered by a single O or B star (Lee et al. 2009, 2011). These degeneracies have led to much discussion regarding the true origin of the observed FHα/FFUV trends, with explanations ranging from a changing IMF at low density (Pflamm-Altenburg & Kroupa 2008; Meurer et al. 2009), to age effects (Alberts et al. 2011), to stochastic sampling (Goddard et al. 2010; Hermanowicz et al. 2013) or non-uniform SFHs (Weisz et al. 2012).

Hα emission is not unique to H II regions, however, so any study of FHα/FFUV in galaxies must also consider the diffuse ionized gas (DIG, or, in the Milky Way, the warm ionized medium or Reynolds Layer; Reynolds 1990), as well as its connection to star formation in different environments. The DIG is pervasive; it can provide ∼40%–60% of a galaxy’s total Hα flux (Thilker et al. 2002; Oey et al. 2007). Studies of line ratios of the DIG or the WIM have also confirmed that it is physically distinguishable from H II regions (e.g. Reynolds 1985; Rand 1997; Haffner et al. 1999; Madsen et al. 2006), having a different ionization

131 state and temperature. Early studies investigated the contribution of cosmic rays and soft X-rays to its ionization budget (e.g. Spitzer & Tomasko 1968; Silk & Werner 1969), but later observations found these sources to be inadequate to explain it fully (Spitzer & Jenkins 1975). The discovery of [O III] λ5007 emission in the Milky Way’s warm ionized medium (Reynolds 1985) implicated some contribution from shocks as well. Scattered light from H II regions may also contribute upwards of 20% of the observed line emission at high latitude (Wood & Reynolds 1999). Yet emission line ratios show that the DIG is primarily ionized by radiation in the 14–60 eV range (Reynolds et al. 1977), implicating O and B stars as the primary source. While field O and B stars may thus contribute to the DIG (e.g. Torres-Peimbert et al. 1974; Hoopes et al. 2001; Crocker et al. 2015), ionizing photons might also leak from H II regions. In fact, DIG rarely strays far from H II regions (e.g. Walterbos & Braun 1994; Ferguson et al. 1996; Greenawalt et al. 1998), implying that leakage is perhaps the dominant ionizing source. In a study of edge-on galaxies, Hoopes et al. (1999) found that extra-planar Hα emission was strongest and most extended in galaxies with the highest SFRs (in agreement with a previous study by Rand 1996). This close connection between the DIG and H II regions thus implies that changes in H II region properties should be reflected somehow in changes in the DIG, hence making DIG a useful complement in studying how star formation changes with environment. The nearby face-on spiral M101 (NGC 5457) provides a unique target for investigating the connection between star formation, diffuse gas, and local environment. Broadband imaging by Mihos et al. (2013b) found extremely blue (B − V∼0.2–0.4) colors in the extended LSB outer disk of the galaxy, implying a significant population of young stars at large radius. This is also apparent from deep GALEX FUV and near ultraviolet (NUV) imaging, which show that the galaxy has an XUV disk (Thilker et al. 2007a). Given its disturbed morphology, this extended star formation likely resulted from an interaction with one or both of its companions, NGC 5477 and NGC 5474 (Mihos et al. 2013b). Both companions are star-forming themselves, and nearby on the sky. The M101 galaxy group thus provides examples of three different kinds of star-forming environments in close proximity; a high-mass star-forming disk, an LSB star-forming outer disk, and two star-forming companion galaxies with lower mass. Additionally, the near face-on inclination of both

132 M101 and its companion NGC 5474 allows for a direct line of sight to the in-plane DIG, making for easy association between DIG and H II regions in both galaxies. As such, we targeted the M101 Group for deep narrow-band Hα imaging with the Bur- rell Schmidt Telescope at Kitt Peak National Observatory (KPNO). The Burrell Schmidt’s wide ﬁeld of view allows for a direct comparison of all three galaxies in the M101 Group in a single mosaic image. We use our Hα narrow-band imaging data in conjunction with the deepest available GALEX FUV and NUV images of M101 and its companions in order to investigate the statistical properties of the FHα/FFUV ratio in both the H II regions and the DIG as a function of these three environments. The paper is thus split into two major sections, focusing on the H II regions ﬁrst, then moving on to discuss the DIG. In Section 5.2, we give a brief overview of our observation and data reduction procedures. In Section 5.3, we describe our analysis of the H II regions, including analysis techniques (extinction correction, H II region selection, and photometry) and results. In Section 5.4, we describe our analysis of the DIG in a similar manner. Section 5.5 presents a discussion of the connection between H II regions and the DIG, as well as a discussion of the broader applicability of our results. We conclude with a summary in Section 5.6.

5.2 Observations and Data Reduction

Here we present a discussion of our observing strategy and data reduction techniques. We brieﬂy review these here; for an exhaustive description, we refer the reader to our previous work (Watkins et al. 2014; Mihos et al. 2017, and references therein). However, this previous work used broadband ﬁlters, hence we focus in this section on adjustments to these procedures necessary in shifting to narrow-band imaging data.

5.2.1 Observations

We observed M101 with the Burrell Schmidt telescope at KPNO in spring of 2014, using two custom narrow-band interference filters. The two filters have central wavelengths at 6589 Å and 6726 Å (hereafter the on-band and off-band filters, respectively), with ∼100 Å widths, necessitated by the Schmidt’s fast f/3.5 beam. The on-band covers Hα at M101’s

133 Figure 5.1: A view of our diﬀerence image mosaic, showing Hα emission in M101 and its companions. Insets are shown of NGC 5477, NGC 5474, and the eastern side of M101 containing the giant H II region complexes NGC 5471 (center frame) and NGC 5462 (at the lower right), to showcase the wealth of low surface brightness structure we detect. Pixels saturate (white) in this image at ∼2.85×10−16 ergs s−1 cm−2 arcsec−2. North is up and east is to the left.

134 velocity (∼240 km s−1; de Vaucouleurs et al. 1991), while the off-band filter covers the adjacent stellar continuum; given M101’s low inclination, all Hα emission from the galaxy lies within a region of the on-band filter with ∼96% transmission. The on-band filter bandpass is wide enough to include Milky Way emission; however, M101 is located at a high Galactic latitude in a field relatively free of Galactic cirrus (Schlegel et al. 1998; Schlafly & Finkbeiner 2011), limiting contamination. We observed only on moonless, photometric nights, using exposure times of 1200 s for both filters, with dithers of ∼0◦.5 between exposures to remove large-scale artifacts such as flat-fielding errors and scattered light. This resulted in sky levels of 200–300 ADU in the on-band filter, and 150–250 ADU in the off-band. In total, we observed M101 in each filter for 71 × 1200 s (nearly 24 hours per filter). Due to low sky counts in the narrow-band filters, we could not construct flats from night-sky frames alone. To construct the flats, we started with twilight exposures; however, given our large field of view, these twilight flats contained noticeable gradients induced by the setting sun. We therefore also produced flats without gradients using offset night-sky frames with exposure times equal to our object frames (1200 s for both filters), as we did in constructing flat fields for our broadband imaging (see Watkins et al. 2014; Mihos et al. 2017). The final twilight flats consisted of ∼110 individual exposures per filter, averaging ∼20000 ADU px−1, while final night sky flats totaled 82 × 1200 s exposures in the on-band, and 74 × 1200 s exposures in the off-band. We defer a discussion of how we used both of these flats for the final reduction to the next section. Finally, we observed spectrophotometric standard stars from Massey et al. (1988) for photometric calibration, along with several 1200 s exposures of Arcturus in order to model internal reflections and the extended wings of the Schmidt point-spread function (PSF; see Slater et al. 2009).

5.2.2 Data Reduction

We began data reduction by applying a standard overscan and bias subtraction, correcting for nonlinear chip response, and applying a WCS to each frame. Flat-fielding took place in stages. We first constructed master twilight flats by median- combining all ∼110 twilight exposures per filter. To remove gradients in the twilight flats,

135 we then constructed night-sky flats as described in previous works (Watkins et al. 2014; Mihos et al. 2017). In short, for each frame, we created an initial mask using the IRAF1 task objmask, hand-masked any remaining artifacts (typically light scattered by stars just off-frame), and combined the resulting masked frames into a preliminary flat. We then flattened and sky-subtracted all night-sky frames using this preliminary flat, combined the flattened and sky-subtracted images into a new flat, and repeated for 5 iterations, until the flat field converged. We isolated the twilight flat gradients through division by the gradientless night-sky flats. We then modeled and divided the planes out of the twilight flats, resulting in final generation flat fields. This is mathematically equivalent to using the night-sky flats (modulo uncertainty in the gradient fits), but with the improved Poisson statistics of the twilight flats on small scales. Mild fringing is visible in all of our on-band images at an amplitude of ∼0.1%, but absent in the off-band images. As M101 is far from the ecliptic plane (hence from zodial light contributions), the main contributor of this fringing is telluric emission lines (OH; Massey & Foltz 2000), which are not present in the off-band filter. We thus measure and correct for fringing in on-band frames only. Because scattered sunlight dominates the telluric emission in the twilight frames, the twilight flats lack the fringe pattern. Hence, to isolate the pattern, we divide the night-sky flat (which does contain the pattern) by the twilight. We then scale a normalized version of this fringe map to the sky level of each on-band frame (corrected for large-scale gradients) and subtract it from each frame. Because this fringing is present on all on-band night-sky frames, we reconstruct the on-band night-sky flat after fringe removal and rederive the on-band twilight flat gradient before flat-fielding the on-band object frames. For our final flux calibration, we observed spectrophotometric standard stars from the Massey et al. (1988) catalog. We derived photometric zero points by convolving our filter transmission curves over the spectra of these stars to derive filter magnitudes (defined as

−2.5 log(Ffilt) for simplicity, where Ffilt is the total flux in ADU of the star through the filter), which we compared with instrumental magnitudes derived through photometry of

136 each exposure of each star. In each observing run, we observed 12 unique standard stars, several of which we observed multiple times to improve the ﬁnal zero points. Due to uncooperative weather, we did not achieve adequate airmass coverage from these standard star observations; instead, we derived airmass terms for each ﬁlter using photometry of SDSS DR8 (Aihara et al. 2011) stars found in the individual exposures of M101 (this is described in more detail below). The photometric zero points are thus simply:

ZP = −2.5 log(Fﬁlt) − (minst − κ sec z)

where minst is the instrumental magnitude and κ is the airmass term. For each ﬁlter, we take as the zero point the median value of the zero points derived √ from each star. The standard error on the median is 1.253σ/ N, hence errors on the two

filter zero points are σon = 0.006 mag and σoff = 0.003 mag; this translates to an error of ∼2% on M101’s total flux. In our final mosaic of M101, 1 ADU per pixel per 1200 −18 −1 −2 −2 s is equal to an Hα surface brightness of ΣHα = 3.557 × 10 ergs s cm arcsec , or an emission measure of EM∼1.78 cm−6 pc. Using this flux calibration, we find good agreement (to within ∼3%) with the value of M101’s total flux published by Kennicutt et al. (2008), measured within their value of R25. To reduce scattered light artifacts, we also remove reflections and diffuse halos around bright stars in all frames in the manner described by Slater et al. (2009). Briefly, we use deep (1200 s) exposures of Arcturus at different positions on the chip to measure and model these reflections and halos, then scale and subtract them from all stars brighter than V= 10.5 found in each frame. We do this scaling via a rough photometric calibration using SDSS stars found in each field, assuming our on-band filter is equivalent to SDSS r with no color term. This produces fairly robust scalings for the reflection- and halo-subtraction process; only for the brightest stars (V > 8) did we need to tweak the derived magnitudes by hand in order to produce an acceptable subtraction. Given this stability, the large number of SDSS stars in each frame, and the improved airmass coverage, we choose to use the airmass terms derived in this way over those derived from the standard star exposures for our flux calibration. This choice has little effect on the calibration, as the airmass terms are quite small (. 0.1) for both filters. Finally, we sky-subtract each frame by masking all bright stars and galaxies, fitting 137 Figure 5.2: A masked, 9×9 pixel median-binned image of our difference mosaic, showcasing a plume of extremely diffuse Hα emission. North is up and east is to the left. sky planes to each masked image, and subtracting these planes from the frames. To preserve precise flux scaling, we then scale these images to zero airmass and median-combine them into two final mosaics (an on-band and off-band) using the IRAF tasks wregister and imcombine. Because these two mosaics combine many exposures taken under variable observing conditions, a direct subtraction of the two does not produce a clean difference image, making it difficult to identify LSB regions. Hence we create a third mosaic using individual pairs of images taken back-to-back. We align both images to within 0.1 pixels, photometrically scale and subtract the off-band images from the on-band, and combine the individual difference images into one mosaic, as before. While we use this difference mosaic to display our data, all Hα fluxes quoted henceforth are measured from the on-band and off-band mosaics, which preserve the flux calibration most accurately.

138 In our previous work, the background sky scatter was dominated by unresolved sources (background galaxies and foreground stars; see Rudick et al. 2010a), however the grand majority of these sources have no emission lines that fall within our two narrow-band filters and thus cleanly subtract out. This, combined with our large total exposure time, results in extremely low background noise. We calculate the background sky uncertainty as the dispersion in the median count levels measured in 50–100 blank apertures with radius 15 pixels (2200) chosen adjacent to the target galaxies. Near M101, the background scatter in −19 the difference image is σ ∼0.15 ADU, giving a limiting depth of ΣHα,lim ∼5.34×10 ergs s−1 cm−2 arcsec−2 (EM∼ 0.27). The scatter is slightly lower near NGC 5474 (σ ∼0.13 ADU) despite it being nearer the edge of the mosaic; this is due to the presence of several slightly imperfectly subtracted reflections from bright stars near M101. Figure 5.1 shows a subset of our full difference mosaic, with several areas of interest zoomed in to showcase the wealth of LSB Hα emission we detect. We also tentatively identify an extremely extended and LSB plume of Hα-emitting gas northeast of M101. While barely visible in Figure 5.1, we show an enhanced image of it in Figure 5.2, which shows our difference image masked of bright pixels (masks shown in white) and median- binned into 9×9 pixel bins. The plume spans a length of ∼30 kpc, and has a characteristic surface brightness of −18 ΣHα = 1.4 × 10 (EM∼0.7), extending from the diffuse star-forming northeast plume region discussed by Mihos et al. (2013b). When compared to adjacent background regions of similar size and shape (see Rudick et al. 2010a; Watkins et al. 2014), this surface brightness amounts to roughly a 2σ detection. So well removed from M101’s star-forming disk, the ionization source for this plume is unclear. One possibility is that it is gas ionized by the metagalactic ionizing background, however the feature’s Hα surface brightness is roughly an order of magnitude higher than expected for this phenomenon (Vogel et al. 1995). Ad- ditionally, we see no evidence of DIG in the long, low column density H I feature on the opposite, southwest side of M101 (Mihos et al. 2012) as might be expected if the ionization was from the metagalactic background. A more mundane explanation might be that the plume is diffuse Hα located within our own Milky Way galaxy. The velocity width of our filter also covers Galactic ISM velocities, and an examination of the H I data cube of Mihos et al. (2012) shows copious diffuse Galactic H I projected across the M101 Group. If this

139 Milky Way gas is ionized, it would show as a patchy screen of diﬀuse Hα across our image. However, the spatial coincidence of the Hα tail with the NE plume in M101’s tidally distorted outer disk, and the lack of any comparable features elsewhere in our mosaic (which covers 2◦× 2◦), remains intriguing.

5.2.3 GALEX data

In order to measure the FHα/FFUV ratio, as well as to correct for extinction, we use the deepest available GALEX FUV and NUV images of M101 and its companion NGC 5474. The images of M101 were taken as part of the guest observing program in 2008 (GI3 05), and were first published in (Bigiel et al. 2010b). These images have exposure times of ∼13300 s in both FUV and NUV. The images of NGC 5474 were taken as part of the Nearby Galaxy Survey (NGS; Bianchi et al. 2003) and have exposure times of 1610 s in both FUV and NUV, hence are shallower than those of M101. We calculate all FUV and NUV fluxes directly from the intensity maps, while we calculate photometric errors on these fluxes as Poisson errors using the associated high resolution relative response maps (as discussed in Morrissey et al. 2007).

5.2.4 Background/Foreground Contamination

Given the width of our filters, we detect Hα emission from sources at a large range of redshifts (we cover Hα-emitting sources at 10% transmission out to ∼4300 km s−1 in our on-band filter), resulting in both background and foreground contamination. While background spiral and elliptical galaxies are typically resolved, hence identifiable by eye, we also find many point sources in the difference mosaic that are not obviously associated with the M101 Group galaxies. We investigated the origins of these point sources using the method described by Kellar et al. (2012). Briefly, they define a quantity ∆m = mHα −mR, where mHα is the magnitude of a source in their filters targeting Hα emission and mR is the magnitude of the same source in their continuum R band filter, scaled such that ∆m = 0 for sources with no emission present in the Hα filter. They label unresolved sources with ∆m < 0 “Hα dots”, which are simply point sources that are bright in their difference images. As we use a narrow-band

140 continuum filter instead of R, in our case ∆m = mon − moff. We utilize the same cutoff limit as Kellar et al. (2012) for “dot” selection, that being sources with emission line equivalent widths & 30Å. This corresponds to ∆m . −0.3 for our filter widths of 100Å. While Kellar et al. (2012) obtained follow-up spectroscopy of the Hα bright point sources in their fields, such follow-up is beyond the scope of our project. Hence, we investigated the Hα dots in our field by cross-referencing them with SDSS and plotting their g−r vs. r − i colors. We find that the majority of the Hα dots in our final mosaic lie in the region of color-color space occupied by M stars (Figure 1 of Finlator et al. 2000), while only a select few have colors bluer than this. This M star contamination results from the width and placement of our filters; typical M star spectra contain broad TiO absorption features, and our on-band filter’s central wavelength (∼6600Å) happens to often lie on a peak in between

two such features, while our off-band filter (λcen ∼6700Å) lies in an adjacent trough. This gives M stars the appearance of an emission line source in the difference mosaic. Thankfully, these stars are readily identifiable as being bright in the difference mosaic but strongly lacking in FUV emission, as well as through available SDSS photometry.

We hence reject all sources with FHα/FFUV> 2.0 (this cutoﬀ is also justiﬁed by Starburst99

models, which never reach FHα/FFUV higher than this; Leitherer et al. 1999), g−r > 1.2, and r−i > 0.8. The handful of dots with bluer colors are likely unresolved background galaxies, unresolved star-forming dwarfs near M101, or intergalactic H II regions (Kellar et al. 2012). For example, SDSS spectra of two of the sources shows that they are quasars at z = 1.34007 (α =211◦.8225, δ =53◦.75559) and at z = 1.34536 (α =211◦.13981, δ=53◦.40635); we detect redshifted Mg emission from both of these sources. These bluer sources are rare, however (we find 8 across our entire field of view, for a surface density of ∼2 per square degree), hence have a negligible effect on our analyses.

5.3 HII Region Photometry

First we present our analysis of H II regions in the M101 Group. We begin by discussing our extinction correction method, then we discuss how we identify H II regions against the DIG background, and conclude with the results of this analysis.

141 5.3.1 Extinction Correction

Given that we focus much of this study on the ratio FHα/FFUV, the components of which are separated by some ∼5000Å in wavelength, some manner of extinction correction is called for. Ideally, this would be done using direct tracers of nebular extinction such as the Balmer decrement (the Hα/Hβ ﬂux ratio). While Balmer decrements have been published for ∼200 of the brighter H II regions in M101 (Scowen et al. 1992), we need an extinction correction we can apply across the entire dataset, and so we choose to employ the GALEX- calibrated radial IRX-β extinction correction method described by Goddard et al. (2010, in their Section 3.6). We recap this method brieﬂy here. IRX-β is an empirical relationship between the ratio of the infrared and UV luminosities (the infrared excess, IRX) and the slope of the UV continuum (β). It works under the assumption that all of the non-ionizing UV radiation that is absorbed by intervening dust is reprocessed into the IR (Heckman et al. 1995; Meurer et al. 1995, 1999). IRX-β can be calibrated for the GALEX passbands into the following form:

AFUV = C(FUV − NUV) + ZP

(Calzetti 2001; Seibert et al. 2005; Cortese et al. 2006; Goddard et al. 2010). For normal star-forming galaxies, Cortese et al. (2006) give C = 5.12, while Seibert et al. (2005) give a value of C = 4.37. This value depends on the assumed star formation history (e.g. Calzetti et al. 2005), which affects the transformation from β to FUV−NUV color. The value of ZP depends on the age of the regions of interest, and is relatively constant for populations aged between ∼0–30 Myr (Figure 9 in Goddard et al. 2010). Following Goddard et al. (2010), we make bulk radial extinction corrections using the median FUV−NUV color of H II regions (hence excluding DIG and field O and B stars) in both M101 and NGC 5474. For ease of comparison, we adopt the same values of C = 4.82 and ZP = 0.0 as Goddard et al. (2010), which are, respectively, the average of the values of C published in Calzetti (2001); Seibert et al. (2005); Cortese et al. (2006), and the typical color of ∼10 Myr old populations (Figure 9 in Goddard et al. 2010). We find that our results are not sensitive to these choices for reasonable values of both. The primary purpose of this correction is not to accurately account for dust effects from H II region to H II region, but rather to make a reasonable bulk

142 2.5 Scowen + 1992 FUV-NUV color 2.0

1.5

(mags) 1.0 Hα A 0.5

0.0

0.5 − 0 100 200 300 400 500 600 700 Radius (arcsec)

Figure 5.3: Hα extinction values derived from the GALEX FUV-NUV color IRX-β relation (red triangles), compared with the extinctions of H II regions in M101 as derived from the Balmer decrement given by Scowen et al. (1992) (black points). correction that places the inner and outer disks at the same mean extinction level for a more consistent comparison among environments. This is particularly pertinent in our study, in which we measure the scatter in FHα/FFUV from environment to environment; because we are comparing populations across large radial expanses (e.g. M101’s inner vs. outer disk). A strong gradient could increase the scatter in a given radial range. For comparison, we employed an alternative correction in M101 using the extinction values published by Scowen et al. (1992), derived from the Balmer decrement. We show this comparison in Figure 5.3 by overplotting our UV color–derived values of AHα to the values for H II regions from Scowen et al. (1992), plotted as a function of radius in M101.

While the two are broadly consistent, the UV color–derived AHα values are consistently lower by ∼0.1 mag. This is sensible, because the UV emission is directly tracing the stellar populations, which may not always lie behind a screen of dust depending on the relative dust geometry (for a beautiful demonstration of this, see Figure 1 of Whitmore et al. 2011). We find through application of both methods that this small offset does not affect the conclusions of this paper. We therefore use the UV color–derived values throughout to maintain consistency.

143 5.3.2 Photometry

Region identiﬁcation

We use SEXtractor (Bertin & Arnouts 1996) to identify H II regions directly from the Hα difference mosaic. Because we are selecting regions based on their Hα emission, we are focusing our study only on regions with ongoing star formation. Our interest in this particular study is in comparing physical differences in star formation (for example, changes in the IMF) across environments, hence by focusing on such short timescales, we avoid complications introduced by aging populations, such as the dissolution of Hα- emitting regions by stellar winds (Whitmore et al. 2011). We perform photometry on all regions using a 400. 5 (150 pc) aperture, which is the typical FWHM of the GALEX FUV PSF (the Burrell Schmidt PSFs in the on- and off-band images have FWHM∼300, hence the use of the FUV FWHM is warranted). This is large enough to contain multiple H II regions at M101’s distance (see, for example, Quireza et al. 2006, for sizes of Milky Way H II regions); we discuss how this affects our conclusions in Section 5.5. However, our statistical analyses are also robust to moderate adjustments to the aperture size. To efficiently pick out both outer-disk and inner-disk H II regions, we run SEXtractor at a 2σ threshold on an unsharp-masked version of our difference mosaic, without deblending. This turns SEXtractor into something of a local peak-finding algorithm, hence is useful for identifying the often densely packed inner-disk H II regions against the smooth background DIG. That said, it results in many spurious detections, thus we employ several rejection criteria. First, we run SEXtractor in dual-image mode, measuring fluxes of difference image detections from the FUV images; we reject all regions with FFUV ≤ σsky,FUV , where

σsky,FUV is the pixel-to-pixel background dispersion in the FUV images (measured from the intensity maps in the manner described in Section 5.2.3). We also reject any sources with FHα/FFUV > 2.0, which is set by the maximum FHα/FFUV value we find in Starburst99, from a zero-age cluster with 1/50 solar metallicity (lower than the lowest metallicity found in M101; Croxall et al. 2016). We also reject sources with g − r > 1.2 and r − i > 0.8 to remove M stars (Section 5.2.4). Finally, we reject all sources >144000 (48 kpc) in radius from M101, and >36000 (12 kpc) in radius from NGC 5474. These cuts remove the bulk of the contaminating sources. However, running SEXtractor 144 with no deblending detects not only H II regions, but also local peaks in the DIG. These regions are identifiable by eye as being more uniform in flux across the photometry aperture (as opposed to the point source–like H II regions). However, to reduce subjectivity, we make a first-round rejection of such regions via an automated procedure. We define a concentration parameter:

c50 = 1 − fpx,50 where fpx,50 is the fraction of pixels in the photometry aperture containing 50% of the total

flux (c50 is defined such that high values correspond to higher concentration). We iterate the threshold value of c50 until we see a reasonable rejection of diffuse regions, then reject the few remaining DIG regions by hand. We choose not to reject diffuse-looking regions in the outer disk; H II regions expand until they reach pressure equilibrium with the ISM (Dyson & Williams 1980; Garcia-Segura & Franco 1996), hence in low density environments they can potentially grow quite large. The statistical analyses we discuss below are robust to this rejection procedure, as diffuse-looking regions most often have anomalously low FHα/FFUV (which further implies they are mostly DIG; Hoopes et al. 2001), and are rejected as outliers in the statistical metrics we use.

Results

We show the results of our H II region photometry in Figures 5.4 and 5.5 for M101 and NGC 5474, respectively. The final sample contains 1525 H II regions in M101 and 156 regions in NGC 5474. We show radial profiles of log(FHα) on the left and log(FHα/FFUV) on the right. For comparison, we show radial profiles before and after we apply the extinction correction described in Section 5.3.1 (top and bottom plots, respectively). The grey dashed lines in Figure 5.4 mark M101’s outer disk, which we define as >3 times the azimuthally averaged disk scale length (43000, 14.5 kpc; Mihos et al. 2013b). In Figure 5.6, we show this outer disk demarcation and a potential alternative on both our difference mosaic and on the V-band data from Mihos et al. (2013b), for reference. We discuss how the choice of outer disk boundary affects our results in Section 5.3.3. It should be noted here that NGC 5474 has a strongly offset bulge (van der Hulst & Huchtmeier 1979; Kornreich et al. 1998), hence the definition of its “center” is not entirely clear. We define its center as the centroid of the circular outer isophotes (at 18000, or 6

145 M101 Radius (kpc) Radius (kpc) 0.0 6.7 13.4 20.1 26.8 33.5 40.1 46.8 0.0 6.7 13.4 20.1 26.8 33.5 40.1 46.8 2.0 Uncorrected for extinction 12

2 − ) − 1 1.5 − cm ˚ A 1 )( − 13 −

FUV1.0 /F ) ergs s 14 Hα

Hα− F F 0.5 log( log( 15 − 0.0 2.0 Corrected for extinction 12

2 − ) − 1 1.5 − cm ˚ A 1 )( − 13 −

FUV1.0 /F ) ergs s 14 Hα

Hα− F F 0.5 log( log( 15 − 0.0 0 200 400 600 800 1000 1200 1400 0 200 400 600 800 1000 1200 1400 Radius (arcsec) Radius (arcsec)

Figure 5.4: Left: Hα fluxes of M101 H II regions, plotted against radius. The top panels show fluxes uncorrected for extinction, while the bottom panels show fluxes after the correction described in Section 5.3.1 is applied. The colors represent the local density of points in the plot. Black stars represent regions located within the dwarf companion NGC 5477. The gray dotted line shows our chosen inner disk–outer disk demarcation. Right: FHα/FFUV of all M101 H II regions, plotted against radius. Symbol colors, symbol types, and the gray dotted line are the same as in the left plots.

146 NGC 5474 Radius (kpc) Radius (kpc) 0.0 1.7 3.3 5.0 6.7 8.4 0.0 1.7 3.3 5.0 6.7 8.4 2.0 Uncorrected for extinction 12

2 − ) − 1 1.5 − cm ˚ A 1 )( − 13 −

1FUV .0 /F ) ergs s 14 Hα

−Hα F F 0.5 log( log( 15 − 0.0 2.0 Corrected for extinction 12

2 − ) − 1 1.5 − cm ˚ A 1 )( − 13 −

1FUV .0 /F ) ergs s 14 Hα

−Hα F F 0.5 log( log( 15 − 0.0 0 50 100 150 200 250 0 50 100 150 200 250 Radius (arcsec) Radius (arcsec)

Figure 5.5: As in Figure 5.4, but for H II regions in NGC 5474. We have used the same scale on the y-axes for ease of comparison.

147 Figure 5.6: Choices of inner disk/outer disk boundary in M101, overlaid on the Hα difference image on the left, and the V-band image (Mihos et al. 2013b) on the right. The solid line marks our primary choice, which is 3 times the azimuthally averaged disk scale length (43000, or 14.5 kpc). The dashed line marks an alternative (30000, or 10kpc), located where the Hα surface brightness profile begins to decline (Martin & Kennicutt 2001). kpc) on our on-band mosaic, which is very close to the kinematic center of its (strangely regular) H I velocity field (van der Hulst & Huchtmeier 1979). This choice does not affect the qualitative behavior of the radial profiles, however the flux profile does show more scatter with radius when centered on the bulge. This implies that the isophotal center is the more appropriate choice regarding star formation in this galaxy. The extinction correction has the expected behavior: Hα fluxes increase absent extinction, and FHα/FFUV decreases given stronger attenuation for FUV. The correction applied at all radii in the lower metallicity companion NGC 5474 (which has a central O abundance of 12 + log(O/H) = 8.19, vs. 8.71 in M101; Pilyugin et al. 2014) is less severe than that applied in the dustier central regions of M101. Additionally, we plot the values of log(FHα) and log(FHα/FFUV) for the dwarf irregular (dIrr) companion NGC 5477 (due east of M101; see Figure 5.1) in the same plots as M101 using black stars. Despite its much smaller mass, NGC 5477’s H II regions span the same range of luminosity as those in M101’s inner disk, implying similar LFs between the two environments. The same appears true of NGC 5474; we found it possible to reproduce NGC 5474’s global LF by resampling from M101. Each

148 galaxy contains pockets of high column density gas (of order 1021 cm−2; van der Hulst & Huchtmeier 1979; van der Hulst et al. 2001; Walter et al. 2008), which may account for the similarity. Regardless, that all three galaxies have qualitatively similar LFs is reminiscent of the study by Schombert et al. (2013), which found that the lack of bright H II regions in LSB galaxies can be explained as an artifact of small number statistics, rather than as a change in the LF itself. Yet though each galaxy’s integrated LF appears similar, there are strong radial gradients in mean Hα luminosity in both M101 and NGC 5474. This is most likely a demonstration of the Schmidt Law: molecular gas density in M101 declines exponentially with radius (e.g. Kenney et al. 1991), hence the SFR declines accordingly (Kennicutt et al. 2007; Bigiel et al. 2008). Also, the azimuthally-averaged SFR and gas density within galaxies have a power law relationship (up to the threshold density; Kennicutt 1998), hence it is not surprising that we see general radial declines in mean Hα flux with a large region-to-region scatter. A comparison with the THINGS H I map of M101 (Walter et al. 2008) also shows that regions with the highest Hα flux for their radius always cluster around high H I column density peaks. That the global H II region LFs of M101 and NGC 5474 (and possibly NGC 5477) appear similar thus seems a consequence of each having a similar density structure within its ISM. If gas density alone imposes the radial dependence of Hα flux, it should affect the FUV flux in a similar way, assuming no dramatic changes in e.g. the IMF. Indeed, Figure 5.4 shows that the radial gradient in FHα/FFUV in M101 is strongly reduced after the extinction correction is applied. NGC 5474 contains no strong gradient before correction; this remains mostly true after a correction is applied, although a mild positive gradient is induced, implying that perhaps we are slightly overcorrecting for extinction in this galaxy. Therefore, it may be that any radial trend in mean FHα/FFUV in either galaxy can be attributed to extinction.

The scatter in FHα/FFUV also appears roughly constant with environment, from M101’s inner disk, to its outer disk, to NGC 5474, and possibly even NGC 5477 (though with only 14 total H II regions, any measure of scatter in this galaxy will be highly uncertain). In tandem, this implies that star formation is ignorant of the global environment; other than the available fuel, it does not seem to know whether it is taking place in a low mass galaxy,

149 a high density inner disk, or a low density outer disk. We test these observations explicitly in the next section.

5.3.3 Statistical Analysis

The intrinsic FHα/FFUV ratio is mainly driven by the number of massive O and B stars. If present, they are the primary source of the ionizing radiation that powers the Hα emission. A truncated IMF would result in fewer massive stars being born, reducing the maximum possible FHα/FFUV. We show this in Figure 5.7 via evolutionary tracks of FHα/FFUV from Starburst99 (Leitherer et al. 1999) for two diﬀerent metallicities. We show both a standard

Kroupa IMF (solid lines; Kroupa 2001), and a Kroupa IMF truncated at 20 M (which has been suggested; Bruzzese et al. 2015). The tracks diverge clearly at early times (.6 Myr, set by the lifespans of the most massive stars), with the truncated IMF tracks peaking at much lower FHα/FFUV as expected. If a change in the IMF occurs at a given radius in a galaxy, the distribution of allowed values of FHα/FFUV in the H II region population will adjust accordingly. Lower variation in the region-to-region dust content in outer disks would result in a similar change, once bulk radial trends are taken into account. Such behavior ought to be observable, therefore, in the scatter of bulk extinction-corrected FHα/FFUV within diﬀerent populations of H II regions, assuming that variations in the median FHα/FFUV can be fully attributed to extinction eﬀects.

We display the medians and two measures of the scatter in FHα/FFUV in Figure 5.8 for three regions: M101’s inner disk (inside of 3h), M101’s outer disk (outside of 3h), and the more massive companion NGC 5474. The inner/outer disk boundaries in M101 are marked on both Figures 5.4 and 5.6, for reference. In the top panel of Figure 5.8, we show box and whisker diagrams for these three regions. As a reminder, the boxes span the 1st through 3rd quartiles of the data (Q1 and Q3), and the whiskers span to ±1.5× the interquartile range. Medians are shown in red, and outliers as +’s. In the bottom panel, we show the values of the trimmed standard deviation

(σt) for the same three regions. This is the standard deviation of the sample trimmed of its top and bottom 5% of values, multiplied by a corrective factor (1/0.789 for 5% trimming;

Breiman 1973; Huber 1981; Morrison et al. 1990) to ensure that σt and σ (the standard deviation of the whole sample) are measuring the same parameter in the case of purely

150 2.0 Z = Z

Z = Z , trunc.

Z = 1/3 Z 1.5 Z = 1/3 Z , trunc.

) FUV

/F 1.0 Hα log(F

0.5

0.0 0 2 4 6 8 10 Time (Myr)

Figure 5.7: Starburst99 models of the time evolution of FHα/FFUV for two diﬀerent metallicities (using Padova isochrones; Bressan et al. 2012). Solid lines show the evolution for a standard Kroupa IMF; dashed lines show a Kroupa IMF with a truncation at 20 M . ) 1

− 2.0 ˚ A

)( 1.5

FUV 1.0 /F

Hα 0.5 F (

10 0.0

log 0.34 0.32 σ 0.30 0.28

Trimmed 0.26 0.24 M101: 000–43000 M101: 43000–97500 NGC 5474

Figure 5.8: Top: box and whisker plots showing the distribution of FHα/FFUV values in three regions in the M101 Group: M101’s inner disk, M101’s outer disk, and NGC 5474 as a whole. Bottom: values of the√ trimmed standard deviation in FHα/FFUV for the three regions described above. Error bars are 1/ 2N for sample size N. The radius 43000 is 3× the scale length of M101.

151 Gaussian data. We use σt over σ for its robustness to outliers, such as extremely luminous H II regions or the handful of DIG regions that might have made it into the ﬁnal sample; other such robust estimators of scatter (such as the median absolute deviation) give similar

results. The error bars on σt are simply the standard error on the standard deviation, which √ is equal to 1/ 2N for sample size N.

After we apply our extinction correction, the median values of FHα/FFUV for all three regions are: 1.047±0.013, 0.903±0.015, and 0.832±0.029, respectively. While this implies statistically significant differences in the medians from region to region, we give only the standard errors (which are equivalent to bootstrapped errors, despite the slight non- Gaussianity of the data). Systematic errors on the GALEX-calibrated IRX-β extinction correction are larger (of order 0.1 mag, excluding uncertainties in the transformation from FUV−NUV color to β; Cortese et al. 2006), which does not include methodological uncertainty inherent in applying this correction on average in radial bins. The differences in the medians between all three regions are also smaller than the standard deviations in

FHα/FFUV (σ ∼ 0.3), again implying that most of the gradient in FHα/FFUV in M101 can likely be explained by extinction alone.

Similar box widths in Figure 5.8, as well as similar values of σt, also suggest that the

scatter in FHα/FFUV among the three regions are equal. We therefore compare the sample variances using Levene’s Test (Levene 1960). This test assesses whether or not the quantity zi j = |xi j − x¯i|, wherex ¯i is the mean of the i-th group, is equal between groups. It is hence similar to the F-test in that it assesses equality of variances between populations, but it is more robust to non-Gaussianity and higher in statistical power (e.g. Lim & Loh 1996). The mean can be replaced with a more robust statistic, such as the median (e.g. Brown & Forsythe 1974); we use the trimmed mean, defined analogously to the trimmed standard deviation. We show the results of this test in Table 5.1 for the following comparisons: M101’s inner disk to its outer disk, M101’s inner disk to NGC 5474, M101’s outer disk to NGC 5474, and all three simultaneously. W is the value of the test statistic, while the p-value is defined in the standard way for confidence 1 − α. In all four tests, we cannot reject the null hypothesis that the variances in FHα/FFUV in all three environments are equal. While, for philosophical reasons, this does not by itself prove that the variances are equal, these results

152 Table 5.1. Results of Levene’s Test Trials

TEST: In-Out In-5474 Out-5474 All

W 0.293 0.009 0.052 0.147 p-value 0.588 0.924 0.821 0.863

in conjunction with the similarity in values of σt and widths of the box plots for each region strongly imply that this is the case. We verified that this result is not sensitive to the definition of the outer disk; the conclusion remains true for choices anywhere between 30000 (the point at which the Hα surface brightness profile begins to decline; Martin & Kennicutt 00 2001) and 600 (roughly the Holmberg radius, R26.5; Mihos et al. 2013b). While the comparisons between the regions of M101 and NGC 5474 seem immune to the choice of inner disk–outer disk boundary, large uncertainties in regions with smaller sample sizes could make it harder to draw such a strong conclusion. We thus further tested this through a bootstrapping experiment. For each definition of the inner disk–outer disk boundary, we randomly sampled N values of FHα/FFUV from either the inner or outer disk, with N equal to NGC 5474’s sample size. We then ran Levene’s Test again between the downsampled M101 population and NGC 5474. We repeated each sampling test 10000 times; in all tests, the resulting p-values were >0.05 between 93% and 97% of the time, providing evidence that the results of the previous tests using the full samples were not an artifact of sample size. While these results are robust to the choice of inner disk–outer disk boundary, we find that the lowest p-value was obtained using 30000 rather than 43000 (p=0.14 vs. 0.59, respectively). By splitting the disk into three parts, we found that the region within 30000–43000 does have significantly higher scatter in FHα/FFUV. Figure 5.4 shows that this region has a low density of H II regions relative to the rest of the disk. It also appears dynamically distinct; it lies roughly at co-rotation with the inner disk spiral arms (Waller et al. 1997), and is the site of a severe kink in the H I rotation curve (Meidt et al. 2009). This is also the location of a pocket of high velocity gas in the galaxy’s northeast (Walter et al. 2008; Mihos et al. 2012) and a region with a high velocity dispersion (Walter et al. 2008). Dynamical effects may thus have influenced the H II region population in this particular area (a high

153 gas velocity dispersion, for example, may inhibit star formation; Kennicutt 1989). Aside from this unusual region, however, we ﬁnd that once extinction is taken into

account, both the median FHα/FFUV and the scatter in FHα/FFUV shows no signiﬁcant variation with environment in the M101 Group. This supports our initial conjecture that, aside from gas density (which aﬀects the intensity of the star formation), star formation on short timescales is blind to environment.

5.3.4 The Unchanging Nature of HII Regions

If both the median and scatter in the FHα/FFUV ratio in H II region populations is constant with environment, once extinction is taken into account, one might question how much room is left for variations in the IMF. We explore this question through comparisons with Starburst99 (Leitherer et al. 1999) models, which we show in Figure 5.9.

Blue histograms in Figure 5.9 show the distributions of FHα/FFUV in H II regions in four radial bins within M101, uncorrected for extinction. In order to compare our data with Starburst99 models, we chose these bins such that their mean metallicities (measured from the H II region metallicity values supplied by Scowen et al. 1992, which range from ∼5× solar to ∼ 1/5 solar) corresponded to the metallicity options available in Starburst99. For each bin, we randomly sampled matching values of Hα and FUV from the appropriate metallicity Starburst99 model based on age, sampling between 0–10 Myr (with timesteps of 0.1 Myr). We then randomly applied extinctions to these values, sampling from the measured H II region extinctions found in the appropriate radial bins from the dataset by

Scowen et al. (1992), and calculated the resulting attenuated Starburst99 FHα/FFUV values. We plot the resulting distributions as empty histograms in Figure 5.9. It can be easily seen in Figure 5.9 that these distributions provide a poor match to the data in all radial bins. As noted in Section 5.3.2, our choice of photometry aperture (400. 5) corresponds to ∼150 pc at M101’s distance, and is hence large enough to potentially include multiple H II regions (as well as surrounding DIG). We veriﬁed this through visual comparison with archival HST Hα imaging of M101 (GO13773, PI Chandar), and found that our apertures contain anywhere from 1-8 individual H II regions. Adjacent H II regions should be similar in metallicity, but may not be uniform in age; the Orion Nebula complex, for example, contains four stellar associations within a ∼100 pc radius that span ages from

154 R=14kpc – 21kpc R=7.5kpc – 14kpc 140 90 Observed Observed 80 120 SB99 (Z=0.004) SB99 (Z=0.008) SB99 (avg. 4) 70 SB99 (avg. 4) 100 60 80 50 N 60 40 30 40 20 20 10 0 0 0.5 0.0 0.5 1.0 1.5 2.0 0.5 0.0 0.5 1.0 1.5 2.0 − − R=3.5kpc – 7.5kpc R < 3.5kpc 80 30 Observed Observed 70 SB99 (Z=0.02) SB99 (Z=0.05) 25 SB99 (avg. 4) SB99 (avg. 4) 60 20 50

N 40 15

30 10 20 5 10

0 0 0.5 0.0 0.5 1.0 1.5 2.0 0.5 0.0 0.5 1.0 1.5 2.0 − − log(FHα/FFUV ) log(FHα/FFUV )

Figure 5.9: Comparisons of the observed distributions of H II region FHα/FFUV (blue histograms) in M101 in diﬀerent radial bins with model distributions from Starburst99. Empty histograms show Starburst99 models with uniform sampling of single model regions, sampled from models with metallicities representative of their respective radial bins, while green histograms show averages of 4 samplings.

155 0–10 Myr (Brown et al. 1994), arguing that any individual H II region complex identiﬁed in our sample may actually consist of multiple clusters with varying ages. Indeed, we found that we could reproduce the observed distributions much more successfully by averaging together four Starburst99 models of varying ages, shown via the green histograms in Figure 5.9. We adopted here a standard Kroupa IMF, hence we are able to produce good qualitative matches to the data without the need for a changing IMF. However, the agreement between the model and the data remains somewhat poor in the outermost (lowest metallicity) bin; the model distribution, averaging over four model regions, consistently has a higher median and higher kurtosis. While we can reduce the kurtosis by averaging more than four regions, we cannot adjust the median in this way. Hence we attempted to determine whether or not a truncated IMF provided a better match to this bin.

We found that, while we can produce a better match using an IMF truncated at 30M , the median is still consistently high by ∼0.2. In all other radial bins, truncations as low as

40M can also produce qualitatively good ﬁts to the observed distributions, while trunca-

tions 20M and below provide a poor match at any radial bin. Hence, while truncated IMFs

are not necessary to explain our results, we also cannot rule out truncations as low as 30M . In summary, it is not necessary to invoke truncated IMFs to reproduce the observed distributions of FHα/FFUV in H II regions throughout M101. We now consider whether or not any indication of unusual cloud-to-cloud star formation physics can be seen in the impact H II regions have on their surrounding ISM.

5.4 Diﬀuse Ionized Gas

We now consider star formation physics from a different perspective: the influence of H II regions on diffuse ionized gas in the ISM. We begin by discussing how we identify the

DIG, then move on to our analysis of the DIG’s FHα/FFUV across different environments in the M101 Group, both to better understand the connection between the DIG and H II regions, and to seek out evidence of changing star formation physics from the influence of the H II regions on the ISM in different environments.

156 5.4.1 Isolating the DIG

Diffuse ionized gas permeates the volume outside of H II regions. Hence, to identify the DIG, we mask all of the sources that we identified as H II regions in the previous section, as well as all foreground stars. For the purposes of this study, we are concerned mainly with the relative values of the DIG fraction (hereafter fDIG, the total Hα flux in the DIG divided by the total Hα flux in the galaxy) between environments, hence it is important only that we measure fDIG in a consistent manner for each environment. However, given that we base our mask on the low-resolution GALEX imaging, it is useful to compare our value of fDIG with that found in other studies, to estimate how much of the DIG directly adjacent to H II regions we could be masking. Given our canonical mask (400. 5 apertures), we find a DIG fraction of 33%. Our masking thus appears more aggressive than previous studies of the DIG (e.g. Thilker et al. 2002, who found a DIG fraction of 43% in M101), implying that with our canonical mask we are isolating the most diffuse part of the DIG. Changing the 00 mask aperture by ±1. 5 yields changes in fDIG of ±20%. We find a DIG fraction of ∼17% in NGC 5474, using our canonical mask. This may be quite low for a low-mass galaxy; in a sample of 24 dwarfs, Oey et al. (2007) found an average fDIG of 65%±18%, with none below 45%. As Oey et al. (2007) uses the same method to measure fDIG as Thilker et al. (2002), our value is once again more conservative. However, this fraction remains comparatively low even assuming a ∼20% uncertainty in our value of fDIG based on mask size. NGC 5474 is luminous enough that it may not be considered a dwarf (MV ∼ −18; de Vaucouleurs et al. 1991), which may partly explain its low DIG fraction relative to dwarfs, yet we measure a similar fraction in the dIrr NGC 5477 (20%). However, Oey et al. (2007) found that very few galaxy properties (SFR, surface brightness, morphological type, etc.) correlate well with fDIG; only starburst galaxies (defined as galax- −1 −2 ies with an effective Hα surface brightness of log(ΣHα) > 39.4 ergs s kpc ) consistently showed low fDIG compared to the rest of the sample. As neither NGC 5474 nor NGC 5477 are starburst galaxies by this definition, they may both simply fall on the low end of the fDIG distributions of normal star forming galaxies. A more careful comparison with their DIG fractions may be more revealing, but is beyond the scope of this study.

To study the properties of the DIG, we again turn to the FHα/FFUV ratio. Because DIG is usually amorphous, hence less amenable to region photometry, we remap our Hα images

157 to the FUV pixel scale, convolve our on-band and oﬀ-band mosaics with the larger FUV

PSF, and measure the FHα/FFUV ratio on a per-pixel basis. Noise in the GALEX FUV background obeys Poisson statistics, hence we reject all pixels with counts that fall below 3× the square root of the FUV sky background (as measured from the respective galaxies’ sky background images), as well as pixels that fall below 3σsky in the Hα on- and oﬀ-band images.

5.4.2 Results

Radial trends of FHα/FFUV

Because the isophotes in both M101 and NGC 5474 are distorted and asymmetric, we ﬁnd that radial trends in the amorphous DIG are more clearly demonstrated using reprojections of both galaxies, which we show in Figure 5.10. These plots show radial reprojections of the per-pixel FHα/FFUV values of the DIG in both galaxies, median binned in bins of azimuthal angle (1◦.2 and 2◦.4 in M101 and NGC 5474, respectively) and radius (300. 6 and 200, or 120 pc and 67 pc) using the astroML (VanderPlas et al. 2012) package ‘binned statistic 2d’. The ordinate axis is radius, and the abscissa is azimuthal angle; the galaxy centers are at R = 0. North is 0◦, increasing counterclockwise; for reference, the large H II region complex to the east of M101 (the zoomed inset in Figure 5.1) appears as the red and yellow overdensity at roughly (90◦, 60000) in the top panel of Figure 5.10. H II region and foreground star masks are shown in gray, while bins with more than 50% of their pixels below our 3σ signal-to-noise threshold in Hα and FUV emission are not shown

(i.e. white). Values of FHα/FFUV have not been corrected for extinction.

The link between H II regions and the FHα/FFUV value of the DIG is apparent from the plots: FHα/FFUV peaks near H II regions and tapers oﬀ with distance (seen as red around the gray masks). Tests with model point sources indicate that this is not simply a result of differing PSFs between the FUV and Hα bands. Additionally, the DIG is always concentrated near H II regions and H II region complexes; this is also demonstrated by the sensitivity of the DIG fraction to the mask size (and was also, for example, seen in the Sculptor Group galaxies by Hoopes et al. 1996). It is also noteworthy, however, that FHα/FFUV in the DIG in both galaxies appears to decline overall with radius. We show these trends in Figure

158 1000 2.50 M101

800 2.00

600

1.50

R (arcsec) 400

1.00 200 ) FUV

300 0.50 /F Hα F NGC 5474 ( log 250 0.00 200

150 0.50 − R (arcsec)

100 1.00 − 50

0 1.50 0 50 100 150 200 250 300 350 − φ (degrees)

Figure 5.10: Reprojections of M101 (left) and NGC 5474 (right), showing median-binned per-pixel FHα/FFUV ratios of the DIG plotted against radius in both galaxies. The ordinate is radius from the galaxy center (note diﬀerence in radial scale), while the abscissa is azimuthal angle in degrees: due north is 0◦, increasing counterclockwise (north through east). Gray indicates a mask, and white indicates that more than 50% of the pixels in the bin were below our 3σ signal-to-noise threshold in both Hα and FUV emission.

159 M101 NGC 5474 ) 1.0 1.0 (No extinction correction) (No extinction correction) FUV 0.8 0.8 /F Hα

F 0.6 0.6

0.4 0.4

0.2 0.2 Median log( 0.0 0.0

(Extinction corrected) (Extinction corrected) ) 1.0 1.0

FUV 0.8 0.8 /F Hα

F 0.6 0.6

0.4 0.4

0.2 0.2 Median log( 0.0 0.0 200 400 600 800 1000 50 100 150 200 250 R (arcsec) R (arcsec)

Figure 5.11: Radial profiles of the median FHα/FFUV in the DIG. M101 is on the left, and NGC 5474 is on the right; profiles shown in the top panels have no extinction correction applied, while profiles shown in the bottom panels use the extinction correction derived from the galaxies’ H II regions (Section 5.3.1). Errorbars (typically smaller than the point size) are the standard error on the median.

160 5.11. The top panels show the median FHα/FFUV in the DIG against radius in both galaxies, with no extinction correction applied. Both galaxies show clear downward trends, though it is much more severe in M101 (at least out to ∼45000, or 15 kpc, beyond which it flattens). We find that this trend is robust to our choice of mask in identifying the DIG as well. We applied a larger H II region mask using SEXtractor, masking all pixels SEXtractor determined to be 2σ above the local background in the Hα difference image (resulting in an unreasonably low DIG fraction of 6%), and found that the radial trends persisted even with this aggressive mask. Hence, even the DIG farthest from H II regions shows a radial trend in FHα/FFUV.

We noted in Section 5.3.2, however, that the radial trend in FHα/FFUV in the H II regions was likely an extinction effect. To make a fair comparison requires an extinction correction in the DIG as well. A proper correction is not straight-forward, however, as on large scales dust is distributed in a complex and filamentary way (e.g. Schlegel et al. 1998). The spectral types of interarm UV-emitting populations may also differ in the field compared to H II regions, depending on whether they formed there (Oey et al. 2013) or migrated there (Crocker et al. 2015), hence deriving extinction from UV color is also more uncertain. However, for consistency, we apply an extinction correction to the DIG using the radial values we derived from the H II regions. We show this in the bottom panels of Figure 5.11. This is likely an over-correction; IRX-β is based on a galaxy-wide metric, hence is biased toward H II regions with higher extinction. Also, from Figure 5.3, we showed that extinctions derived from the UV color are systematically lower than those derived from the Balmer decrement, which is explainable if the UV-emitting populations are not always obscured by dust along the line of sight. By symmetry, UV-emitting populations should on average be less obscured by dust in interarm regions.

Even with this overly severe extinction correction, however, the radial trend in FHα/FFUV persists, implying that it is not solely an extinction eﬀect. If most of the ionizing ﬂux in the DIG comes from photons leaking from H II regions (Ferguson et al. 1996; Oey &

Kennicutt 1997; Hoopes et al. 2001), one might naively expect that FHα/FFUV in the DIG

would closely track FHα/FFUV in the H II regions, perhaps oﬀset to lower values due to the presence of ﬁeld O and B stars contributing additional FUV emission (Hoopes & Walterbos 2000; Hoopes et al. 2001; Oey et al. 2004). However, as we showed in Section 5.3.2, the

161 1.4 16

1.2 14

1.0 12 ) 0.8

FUV 10 /F 0.6 Hα

8 N 0.4 6 0.2 Median log(F 4 0.0

0.2 M101 NGC 5474 2 − 0.4 0 − 19 18 17 16 15 19 18 17 16 15 − − −1 2 − 2 − − − −1 2 − 2 − log(ΣHα,c) (ergs s− cm− arcsec− ) log(ΣHα,c) (ergs s− cm− arcsec− )

Figure 5.12: Correlation of FHα/FFUV in the DIG with ﬂux surface density of Hα emission from compact sources (H II regions, denoted ΣHα,c) in M101 (left) and NGC 5474 (right). 2D histogram 00 bins show log(ΣHα,c) and the median log(FHα/FFUV) in 75 (2.5 kpc) wide boxes, centered at randomly selected pixels within 87000 (29 kpc) of M101 and 25000 (8.5 kpc) of NGC 5474. To maintain consistent relative areal coverage between the two galaxies, we sample 3500 boxes near M101 and 300 boxes near NGC 5474.

H II regions show no discernable radial trend in FHα/FFUV. In the following section, we explore how these discrepant radial trends may be related to the declining density of H II regions with radius in the disk.

The DIG and HII region density

The close spatial correspondance between the DIG and H II regions argues that most of the ionizing flux in the DIG comes from Lyman continuum photons leaking from H II regions (Ferguson et al. 1996; Hoopes et al. 1996; Zurita et al. 2002). The ionizing flux— hence the Hα emission—in the DIG should thus decrease with distance from H II regions. Analogously, a higher spatial density of H II regions should result in a higher Hα flux surface density in the surrounding DIG, as any parcel of such gas would be flooded with

ionizing radiation from multiple H II regions. This may thus explain why FHα/FFUV in the DIG has a radial trend in both M101 and NGC 5474, despite similar ionizing populations in H II regions in all environments.

We test this in Figure 5.12, which shows the median FHα/FFUV ratio in the DIG against

162 ﬂux surface density of Hα emission from H II regions (hereafter, ΣHα,c, where “c” stands for “compact” sources, in units of ergs s−1 cm−2 arcsec−2), as measured in boxes 7500 (2.5 kpc) on a side, a scale corresponding roughly to the thickness of a spiral arm. To create an unbiased sample, we randomly centered these boxes within 87000 (29 kpc) of M101 and within 25000 (8.5 kpc) of NGC 5474, radii chosen to include the most extended H II regions in each galaxy. Because of its much smaller area, we sample only 300 boxes around NGC 5474; using a number larger than this provides no additional information, as boxes begin to signiﬁcantly overlap. Thus, to maintain the same relative areal coverage, we sample from 3500 boxes around M101.

In the relationship between FHα/FFUV and ΣHα,c, both galaxies show similar trends, though it is less deﬁned in NGC 5474 due to its smaller sample size. In both galaxies,

FHα/FFUV does correlate with ΣHα,c, but only at the high density end. We see no evidence −17 −17 of a correlation below ΣHα,c∼ 10 . The slopes of the correlation above ΣHα,c∼ 10 appear similar, although differences in the dispersion between the two galaxies, as well as the more limited range of ΣHα,c and FHα/FFUV in NGC 5474, makes a comparison difficult. The higher dispersion in M101 is partly physical: it originates both from the larger sample of boxes we measure within M101, and from a wider diversity of DIG environments than can be found in NGC 5474. For example, we are able to find local regions within M101 with similar distributions of both DIG FHα/FFUV and ΣHα,c as we find in NGC 5474 (e.g., in M101’s far north), but also local regions in M101 with vastly different distributions of both (e.g. in M101’s inner disk). Regardless, this strong correlation above 10−17 between Hα flux surface density from

H II regions and FHα/FFUV in the DIG concurs with the idea that the bulk of the ionizing radiation contributing to the DIG can in fact be traced to leakage from H II regions. We explore the details of this correlation in Section 5.5.2, including the lack of a correlation −17 below ΣHα,c∼ 10 .

5.5 Discussion

We have shown that the distribution of the FHα/FFUV ratio in H II region populations, aside from extinction eﬀects, does not change with environment in the M101 Group. We

163 have also shown that we can model the observed distributions of FHα/FFUV in H II regions throughout the M101 Group without invoking a truncated IMF. The makeup of stellar populations ionizing H II regions throughout the M101 Group therefore appears ignorant of the local surface mass density; only the intensity of star formation changes.

On the other hand, the FHα/FFUV ratio of the DIG does show a distinct environmental trend in both M101 and NGC 5474, regardless of extinction eﬀects. We have also shown

that FHα/FFUV in the DIG correlates with H II region ﬂux surface density, but only when the H II region ﬂux surface density is high. We thus more carefully consider the connection between star formation and the DIG by comparing the DIG’s FHα/FFUV ratio with the local makeup of the ISM (neutral vs. molecular hydrogen gas). Also, given that the IMF seems to be universal, and bearing in mind the connection

between star formation and the DIG, we consider the possible origin of trends in FHα/FFUV with e.g. galaxy stellar mass and central surface brightness found by other authors (e.g. Lee et al. 2009; Meurer et al. 2009). Because these studies focus on the integrated Hα and FUV fluxes of galaxies—which includes compact H II regions, DIG, and diffuse FUV emission—we consider how the DIG, diffuse FUV emission, and bias in measurement

techniques might each contribute to the observed trends in integrated FHα/FFUV of whole galaxies. Finally, we discuss our results in the context of the M101 Group itself, particularly its tidal interaction history, and consider whether or not our results can be generalized to other systems.

5.5.1 The Connection Between Star Formation and the DIG

While H II region populations appear similar across all environments, this does not appear to be true regarding the DIG. Because the DIG is a part of the ISM, we explore this further by comparing its properties to two other components of the ISM: neutral and molecular hydrogen. For this purpose, we use H I data of M101 from the THINGS survey (Walter et al. 2008) and CO data of M101 from the HERACLES survey (Leroy et al. 2009, 2012).

We showed in Section 5.4.2 that the Hα ﬂux surface density from H II regions, ΣHα,c, −17 correlates with FHα/FFUV in the DIG only above ΣHα,c∼ 10 , while below this value there appears to be no correlation at all. To a point, this agrees with previous studies, which

164 15.5 100 −

16.0 − ) 2 − 16.5 − arcsec

2 17.0 − − cm 1

− 17.5 − 10 N

18.0 ) (ergs s −

HII,c 18.5 − HI H2 log(Σ 19.0 All radii Inner disk −

19.5 − 1 0 2 4 6 8 10 12 14 0 5 10 15 20 25 30 2 2 ΣHI (M pc− ) ΣH2 (M pc− )

Figure 5.13: Left: correlation of mean H I surface density with logΣHα,c(as defined in Figure 5.12), measured within 15000×15000 (5kpc×5kpc) boxes across the entire disk of M101. Right: correlation 00 00 of logΣHα,c with molecular gas surface density, measured within 75 ×75 (2.5kpc×2.5kpc) boxes, only in M101’s inner disk (R<30000, or 10kpc). found that leaking Lyman continuum photons from H II regions contribute a significant fraction of the flux required to ionize the DIG (Ferguson et al. 1996; Hoopes et al. 1996,

2001; Oey et al. 2004). But given the lack of correlation between FHα/FFUV and ΣHα,c at low flux surface density, this may only be true in regions with a high star formation density. This behavior is in fact reminiscent of results from Bigiel et al. (2010b), who found two regimes of star formation: an extremely low efficiency regime, mostly found in outer disks and other low density environments, in which the star formation rate is defined by the H I column density; and a much higher efficiency regime in which the SFR is defined instead by the molecular gas. If this model is correct, low efficiency star formation in outer disks might lead to systematically fainter H II regions which are not as capable of ionizing the surrounding ISM: see, for example, the radial decline of H II region flux within M101 and NGC 5474, as shown in Figures 5.4 and 5.5. This would then lead to a weak correlation between FHα/FFUV in the DIG and H II region density on scales larger than a kiloparsec, as we showed in Figure 5.12. We thus compare ΣHα,c against ΣHI and ΣH2 in M101, to determine whether or not different star forming regimes might also explain the trends in

FHα/FFUV in the DIG.

165 We show this in Figure 5.13. The left panel shows logΣHα,c plotted against ΣHI; we measured both in 20000 boxes 15000 (5 kpc) in width randomly placed throughout the entirety of M101’s disk. In the right panel, we show logΣHα,c plotted against ΣH2 , measured in 20000 boxes 7500 (2.5 kpc) in width within M101’s inner disk only (<30000; CO emission was not detected in M101’s outer disk in the HERACLES survey; Leroy et al. 2012). This sample number is roughly set by the ratio of total sampling area to box width, as it was in the construction of Figure 5.12.

Three things are of note in Figure 5.13. First, ΣHα,c clearly correlates with ΣHI below −17 ΣHα,c∼ 10 , but this correlation becomes more ambiguous above this value. This is the same ΣHα,c demarcation we found in the correlation between ΣHα,c and FHα/FFUV in the

DIG. While we find that a handful of the boxes with low ΣHI and high ΣHα,c straddle spiral arms in the outer disk—hence resulting in artificially low mean ΣHI within the boxes as most of the flux comes from H II regions on the edge of the bin—most of them lie fully within M101’s inner disk, indicating that this lack of correlation at high ΣHα,c is real. −17 Second, ΣHα,c appears to also correlate with ΣH2 above ΣHα,c∼ 10 . In this case, all of

the boxes with low ΣH2 and high ΣHα,c straddle the outer edge of M101’s molecular disk,

hence are artiﬁcially depressed in ΣH2 . Ignoring these boxes, we ﬁnd that the correlation

between ΣH2 and logΣHα,c is well fit by a power law with a slope of 1.3, very near the slope of the Kennicutt-Schmidt Law (∼1 - 1.5; Kennicutt 1998; Kennicutt et al. 2007; Bigiel et al. 2008), implying that this correlation is real. −17 Third, we find very few regions in M101 with ΣHα,c. 10 that have accompanying CO detections. While this may be due to a limited depth in the HERACLES CO map of M101, because Bigiel et al. (2010b) based their conclusions on the HERACLES CO data −17 we can be confident that the flux surface density ΣHα,c∼ 10 also serves to demarcate

their proposed H I- and H2-dominated star formation regimes in M101. Hence, FHα/FFUV in the DIG does not correlate with H II region flux surface density in the low efficiency, H I-dominated regime proposed by Bigiel et al. (2010b). This picture is thus self-consistent. In high-density environments, where star formation is controlled by molecular gas, star formation is efficient and many bright H II regions form in close proximity. Because these H II regions are packed so closely together, they act in concert to ionize their surrounding ISM on large scales via leaking Lyman continuum

166 photons, with the intensity of the resulting diffuse Hα emission—the DIG—scaling with the H II region flux surface density. This also results in any extant H I becoming strongly ionized throughout the disk, hence H I no longer correlates with SFR on any physical scale in such environments. At extremely high star formation density, as in starburst galaxies, supernova feedback may create holes in the ISM, increasing the transparency of the galaxy to Lyman continuum photons and decreasing the DIG fraction (Oey et al. 2007). In low density environments, star formation is less efficient, resulting in fainter (note again that H II regions decline in flux with radius; Figures 5.4 and 5.5) and more sparsely distributed H II regions, which are unable to ionize anything but the ISM in their immediate vicinities.

H I is thus preserved on large scales (>2.5 kpc), and so ΣHI correlates with ΣSFR on these

scales as well (and up to galaxy-wide scales, as in Kennicutt 1989). While FHα/FFUV in the DIG may correlate with H II region ﬂuxes in their immediate surroundings, it no longer correlates on roughly kiloparsec scales, as demonstrated in Figure 5.12.

The dependence of the DIG’s FHα/FFUV with environment thus requires only an environmental dependence on H II region ﬂux surface density, not any physical change with the H II region populations themselves beyond their average luminosity. Any explanation

for why the integrated FHα/FFUV—the ratio of the total Hα flux to total FUV flux, diffuse and otherwise, in whole galaxies or in large radial bins within galaxies—is lower in outer disks and low mass galaxies (Gil de Paz et al. 2005; Thilker et al. 2005; Lee et al. 2009; Goddard et al. 2010), then, depends not on the H II regions but on the diffuse Hα emission

and diffuse FUV emission, as due to low ΣHα,c, the diffuse emission becomes the dominant source of the flux in such environments.

5.5.2 On the Observed Trends of Integrated FHα/FFUV

If the IMF does not change with environment, as we have argued, why then do many

studies ﬁnd that FHα/FFUV integrated over galaxies, or azimuthally averaged in wide radial bins, is lower in low density environments (e.g. Gil de Paz et al. 2005; Thilker et al. 2005;

Lee et al. 2009; Meurer et al. 2009; Goddard et al. 2010)? Because the integrated FHα/FFUV includes all sources of Hα and FUV emission, from H II regions to DIG to diﬀuse FUV

with no Hα counterpart, changes in integrated FHα/FFUV can result from many diﬀerent factors, from variations in the IMF, to stochastic sampling of the IMF in low mass clusters,

167 Table 5.2. Integrated Properties of M101 Group Galaxies

Region: M101 Inner M101 Outer M101 NGC 5474 NGC 5477

fDIG 0.33 0.34 0.28 0.17 0.20 fDUV 0.56 0.51 0.60 0.42 0.45 FHα/FFUV 1.20 1.23 1.11 0.89 1.02 FHα/FFUV, corr. 0.98 0.97 0.97 0.70 0.96

Note. — Rows are: fraction of Hα flux from the DIG (1), fraction of diffuse FUV emission (2), integrated FHα/FFUV (3), and integrated FHα/FFUV corrected for extinction (4). Systematic uncertainties, which dominate, are discussed in the text. to a non-uniform SFH (e.g. Lee et al. 2009; Alberts et al. 2011; Barnes et al. 2011, 2013; Weisz et al. 2012; da Silva et al. 2014). We have shown that variation in the IMF is unlikely within the M101 Group. Given our low resolution, most (>90%) of the H II region complexes we identified have fluxes above where stochastic sampling ought to be important (e.g. Hermanowicz et al. 2013). However, a non-uniform SFH could be identifiable in the form of populations of FUV-emitting stars with no Hα counterpart, specifically if such populations are remnants of a fading burst of star formation. Therefore, we test for an overabundance of FUV relative to Hα by comparing the fractions of diffuse Hα emission (or the DIG fraction, fDIG) and diffuse FUV emission, which we define in a similar manner as DIG, as FUV emission located outside of H II regions (hereafter, fDUV). While the expectation is that high fDUV relative to

fDIG should yield lower integrated FHα/FFUV ratios, we find that this is not always the case, implying that there may be methodological bias at play as well in measuring integrated properties of galaxies and regions of galaxies. Specifically, a bias may be incurred when using flux-weighted values over, e.g., areal-weighted values of Hα and FUV flux. In Table 5.2, we give diffuse fractions in five environments in the M101 Group: M101 as a whole, M101’s inner disk, M101’s outer disk, its more distant companion NGC 5474, and its nearby dIrr companion NGC 5477. We measure both fDIG and fDUV in an identical manner, hence they are comparable regardless of uncertainty in e.g. the choice of H II region mask. Additionally, for each region we give integrated values of FHα/FFUV before and after applying an extinction correction. In this case, we apply an integrated correction, measured using the integrated FUV−NUV colors of each region, as is typically done in 168 galaxy survey studies (e.g., Lee et al. 2009; Meurer et al. 2009). Comparison of the diffuse fractions in Hα and FUV indicate that diffuse FUV emission is more prevalent compared to DIG in M101’s outer disk and in both companion galaxies. This concurs with a visual examination of the images; in M101’s outer disk, for example, we find many large (several kiloparsec wide) patches of diffuse FUV emission that have no Hα counterpart in our difference image. The areal covering fraction of diffuse FUV emission appears larger than the DIG covering fraction across the whole outer disk, while in the inner disk, the covering fractions of both are roughly equal. Quantitatively, this is observable as a larger outer disk scalelength in the FUV compared to the Hα, such as is typically seen in other XUV disks (Gil de Paz et al. 2005; Thilker et al. 2005; Goddard et al. 2010).

However, the integrated values of FHα/FFUV do not reflect this. Despite the larger fractions of older FUV-emitting populations in M101’s outer disk and in the two companion galaxies, after correcting for extinction, only NGC 5474 shows a significantly different

value of integrated FHα/FFUV. This appears to be an artifact of the flux-weighted measurement; in NGC 5474, we find that the brightest 10% of H II regions (only 16 regions) contribute nearly 60% of the galaxy’s total Hα flux. Thus, if something is systematically different about these few regions—age, dust content—compared to the remaining H II re-

gions in the galaxy, this diﬀerence will drive the galaxy’s ﬂux-weighted mean FHα/FFUV ratio to an unrepresentative value. In NGC 5474, the brightest H II regions have redder FUV−NUV colors (∼0.05 compared to ∼ −0.1 in the dimmer regions). Because we derive the extinction based on the UV color, these regions are measured as dustier environments; if so, the extinction correction may be overcompensating for dust throughout NGC 5474

and driving the integrated FHα/FFUV down. This is demonstrated in an alternative way in M101’s outer disk. As in NGC 5474, the brightest regions in M101’s outer disk are redder in UV color (∼ −0.05 compared to ∼ −0.2), hence potentially dustier, and again contribute a large fraction total Hα ﬂux

(40%). Before applying an extinction correction, the median FHα/FFUV value of all of the H II regions in M101’s outer disk is 0.90, but the ﬂux-weighted mean value of H II regions

is 1.31. Flux-weighting thus drives the integrated FHα/FFUV ratio of H II regions in M101’s outer disk to a higher value, as it is biased by the brighter, redder regions, in direct analogy

169 with the integrated FHα/FFUV value of NGC 5474.

As such, it is unclear whether or not the trends in integrated FHα/FFUV with stellar mass, SFR, and surface brightness noted in other studies result from physical changes or purely from systematics induced by the flux-weighted measurements. Regardless, the M101 Group is a well-studied system, with constraints on stellar populations throughout its disk and constraints on its tidal interaction history with its companions (e.g. Beale & Davies 1969; Rownd et al. 1994; Waller et al. 1997; Mihos et al. 2013b). We can therefore make more specific conclusions about how M101’s local environment may have influenced the star formation taking place in its outer disk and companions, and consider whether or not these conclusions can be generalized to other similar systems. We discuss this further in the following section.

5.5.3 The M101 Group As a Case Study

We have shown that in the M101 Group, H II regions have roughly constant FHα/FFUV distributions regardless of their environment. We have also shown that diffuse FUV emission, with no Hα counterpart, is abundant in M101’s outer disk and its two companions, implying widespread populations of slightly older O and B stars in the field, similar to other XUV disks (e.g. Gil de Paz et al. 2005; Thilker et al. 2005). We argue here that this can be explained in the context of M101’s interaction history, and consider whether or not star formation in the low density environments of the M101 Group could be representative of low density environments as a whole. In general, the origin of field O and B stars is not yet clear. They may form in-situ (de Wit et al. 2005; Lamb et al. 2010; Oey et al. 2013), form within H II regions but be ejected at high velocity (Gies 1987; Moffat et al. 1998; de Wit et al. 2005), or they may be young clusters that have fully succeeded in clearing out gas and dust from their birth H II regions. In a study of diffuse FUV emission in the interarm regions of M101’s inner disk, Crocker et al. (2015) found that the majority is likely emitted by 10–50 Myr old stellar populations that have drifted from their birthplaces in spiral arms. Because these stars are carried by the disk’s underlying rotation, the difference between the rotation speed and the spiral arm pattern speed determines how far they might travel from a given spiral arm; one might expect stars to remain very close to spiral arms near corotation, for instance.

170 UV light scattered into our line of sight by dust contributes a sizeable fraction of the diffuse UV as well (upward of ∼60%; Crocker et al. 2015), but only in the vicinity of spiral arms; in a field adjacent to a spiral arm, Crocker et al. (2015) estimate that the UV flux contributed by scattered light drops by a factor of roughly 1.5 over a distance of ∼1.5 kpc. This, along with lower dust content, implies that diffuse FUV in M101’s outer disk contains very little scattered light. For example, in the galaxy’s northeast, we find large patches (several kpc on a side) of diffuse FUV located some 5–10 kpc from the nearest spiral arm, and some a similar distance from the nearest H II region. This FUV emission thus appears to be a remnant of a previous episode of star formation, which either formed in-situ or migrated from elsewhere in the disk. The largest such patch (∼2 kpc in radius, detected at > 10σ significance in the FUV) has an FUV−NUV color of ∼0.6; in a model of color evolution in integrated populations by Boissier et al. (2008), young populations maintain an FUV−NUV color of ∼0.0 while SF is ongoing, and reach ∼0.6 roughly 200 Myr after star formation begins to decline (neglecting extinction, although extinction may be safely neglected in outer disks). In the Milky Way, populations of O and B stars have radial velocity dispersions of order ∼10 km s−1 (Binney & Merrifield 1998), thus can easily disperse over ∼2 kpc in radius in 200 Myr. This diffuse FUV-emitting starlight thus likely formed in a localized burst a few hundred Myr ago and is now beginning to fade. We find many other such patches of diffuse FUV throughout M101’s outer disk with similarly red colors (∼0.4–0.6), implying similar origins. M101’s disturbed morphology implies that it suffered a recent tidal interaction. From integrated B−V colors in its outer disk, Mihos et al. (2013b) proposed that this morphology resulted from a fly-by encounter with its more distant companion NGC 5474 some ∼300 Myr ago, resulting in a brief and currently fading burst of star formation. After 300 Myr, even the NUV light begins to fade; Hα emission would thus be scarce, as it is in the diffuse FUV patches discussed above. Follow-up HST imaging of stellar populations in M101’s northeast plume region are consistent with this star formation timeline (Mihos et al. in prep.), providing strong support that star formation in M101’s outer disk was induced by an interaction. This in turn shows that M101’s outer disk does not have a uniform SFH. If NGC 5474 was the culprit in the interaction, it too should have seen a starburst on the same timescale, hence it too should have a non-uniform SFH. The M101 Group hence

171 provides a fairly clear example of an FUV-dominated outer disk resulting from a fading, tidally induced starburst; it is not necessary to invoke changes in the IMF to explain the star-forming properties of the M101 Group. Is this scenario generalizable to other systems? XUV disks are often suggested to be tidal in origin (Gil de Paz et al. 2005; Thilker et al. 2005, 2007a), or else are created through gas accretion into the outer disk (Lemonias et al. 2011). Also, the UV emission in XUV disks is typically concentrated in filments reminiscent of spiral structure (Thilker et al. 2007a). While outer disks may typically be stable against spiral arm formation, we can consider the longevity of a set of spiral arms induced in an outer disk by a tidal interaction, hence the longevity of XUV disks in general. As a rough estimate, let us assume that spiral arms in outer disks are not self-sustaining due to high disk stability (e.g. Kennicutt 1989) and so lose their coherency over one dynamical time; in M101 at 16 kpc (roughly where we demarcate its outer disk), this is ∼500 Myr (assuming a rotation speed of ∼190 km s−1; Meidt et al. 2009). Star formation persists in M101 out to ∼40 kpc, where a dynamical time is ∼1.3 Gyr. As such, if galaxies like M101 suffer only one interaction in their lifetimes capable of producing an XUV disk, assuming a total lifetime of ∼10 Gyr, there would be a ∼5–13% probability that we would witness it in this state at z = 0. A study by Lemonias et al. (2011) found that XUV disks exist in 4–14% of galaxies out to z = 0.05; if M101 can be considered representative (it is slightly brighter than L∗ in the V-band; de Vaucouleurs et al. 1991), then the average galaxy need suffer only 1–2 interactions capable of producing XUV disks in their lifetimes to explain the frequency of XUV disks in the local universe. Whether or not this is reasonable depends on how specific the parameters of the interactions must be in order to produce an XUV disk (mass ratio, inclination, velocities, etc.), but this simple argument suggests that all XUV disks may be explainable through tidal interactions with satellites. Even in purely isolated systems, the global stability of outer disks requires that some manner of perturbation is still necessary to initiate star formation there (e.g. substructure in the dark matter halo; Bush et al. 2010). If star formation in dwarf galaxies results mainly from e.g. supernova feedback (van Zee et al. 1997), some manner of perturbation would be required to initiate it in the first place. This general dependence on external forces, rather than on potentially long-lived, regularly rotating spiral features or bars, implies that

172 star formation in low density environments may always be subject to stochasticity, hence an assumption of constant star formation over Gyr timescales in such environments could be highly suspect. As more and more systems are studied, and as ﬁner and ﬁner resolution SFHs are obtained of these systems, the nature of star formation and the evolution of galaxies should begin to become clear.

5.6 Summary

We have performed a study of star formation across all environments in the nearby M101 Group—M101’s inner disk, its outer disk, and its two lower mass companions— using both deep Hα narrow-band imaging, as well as archival UV (GALEX NGS and PI data;Bianchi et al. 2003; Bigiel et al. 2010b), 21cm (THINGS; Walter et al. 2008), and CO data (HERACLES; Leroy et al. 2009), in order to test whether or not star formation physics—specifically the IMF—change with environment. We have studied both the sites of ongoing star formation, H II regions, as well as the impact of star formation on the ISM via the creation of diffuse ionized gas (DIG). In H II regions, we have shown that both the median and the scatter in the Hα to FUV flux ratio does not vary with environment in the M101 Group, once bulk radial extinction trends are taken into account. While typical H II region fluxes do decline with radius in M101 and its larger companion NGC 5474, the near constant FHα/FFUV ratio implies that the populations of ionizing stars even in the fainter outer disk H II regions are being sampled from the same IMF as in the inner disk. Also, using Starburst99 (Leitherer et al. 1999) models drawn from a standard Kroupa IMF (hence without resorting to truncated

IMFs), we are able to qualitatively reproduce the distribution of FHα/FFUV in H II regions in the M101 Group even at low surface brightness. Therefore, in H II regions—which are young and hence still contain their most massive members—we see no evidence that the IMF changes in low density environments. Only the intensity of star formation appears to change with environment, and not the cloud-to-cloud physics.

The FHα/FFUV ratio in the DIG, on the other hand, does decline with radius, even taking extinction into account. The FHα/FFUV in the DIG, however, appears to be driven only by the local intensity of star formation. We have found that FHα/FFUV in the DIG correlates

173 with the flux surface density of H II regions (ΣHα,c), and that this correlation weakens at the same gas density that marks the transition to the low efficiency, H I-dominated regime of star formation proposed by Bigiel et al. (2010b). The lack of correlation between ΣHα,c and FHα/FFUV we observe in the DIG within this regime hence results from lower star formation efficiency, or specifically the more sparsely distributed and fainter H II regions in outer disks, which are less able to ionize their surrounding ISM. This implies that it is mainly the flux surface density of Lyman continuum photons leaking from H II regions that controls the FHα/FFUV ratio in the DIG. Therefore, despite showing a correlation with environment, the FHα/FFUV in the DIG also does not require unusual star formation physics.

Assuming the IMF is universal, the origin of trends in integrated FHα/FFUV with surface brightness, SFR, and stellar mass (e.g. Thilker et al. 2005; Boselli et al. 2009; Lee et al. 2009; Meurer et al. 2009; Goddard et al. 2010) must lie with either the star formation history, or with stochastic sampling from the IMF becoming important at low SFR. However, we have also shown that using flux-weighted FHα/FFUV or flux-weighted extinction corrections can bias the value of the integrated FHα/FFUV in galaxies, particularly if the bulk of the Hα or FUV intensity emerges from a small number of bright H II regions. We thus advise caution in future such studies with regard to how integrated FHα/FFUV is measured. Star formation in M101’s outer disk was likely triggered by a tidal interaction several hundred Myr ago (Mihos et al. 2013b). Therefore, the presence of abundant diffuse FUV emission with no Hα counterpart in M101’s outer disk and companions is most likely the fading remnant of a recent burst of star formation, and not a sudden truncation in the IMF at high mass. Whether this can be generalized to other galaxies with outer disk star formation (e.g. XUV disks) remains to be seen, but we have shown that if interaction-induced star formation in outer disks persists over only one dynamical time, it may still be long-lived enough to account for the low frequency of XUV disks observed in the local universe. If so, this would imply that star formation in low density environments only differs from star formation in high density environments in that it requires outside perturbation to be initiated.

174 5.7 Acknowledgements

This work has been supported by the National Science Foundation, through award 1108964 to J.C.M., as well as through award 1211144 to P.H. This work made use of Numpy, Scipy (Oliphant 2007), MatPlotLib (Hunter 2007), astroML (VanderPlas et al. 2012), and the community-developed core Python package for astronomy, Astropy (As- tropy Collaboration et al. 2013). This work also made use of THINGS, ‘The H I Nearby Galaxy Survey’ (Walter et al. 2008), and HERACLES, ‘The HERA CO-Line Extragalactic Survey’ (Leroy et al. 2009). We would also like to thank Heather Morrison for many helpful discussions regarding the statistical analyses presented in this work, as well as Sally Oey and Daniella Calzetti for helpful discussions about extinction, star formation physics, and star formation tracers.

175 Chapter 6

Summary and Future Work

Galaxies’ outer regions are faint, low in mass, and tenuously held to their hosts, making them sensitive to the slightest perturbation, whether this comes from outside or from within. They trace their hosts’ interaction histories and secular evolution over hundreds of millions to billions of years. This, and their distinct diﬀerences from the comparitively well studied inner regions, makes them a frontier in galaxy evolution studies. A galaxy formation theory that cannot simultaneously explain the star-forming properties, stellar populations, gas content, and dynamics of both the inner and outer regions of galaxies is inaccurate.

6.1 Major Results of This Dissertation

This dissertation seeks to understand how galaxies form and evolve through a series of in-depth case studies of the low surface brightness outskirts of a variety of nearby galaxies. A study of the M96 Group explores how tidal interactions influence galaxy growth in a low density, low mass group environment—a regime as-yet mostly unexplored, but which provides an important link in the context of IGL and ICL formation. New observations of highly extended, faint tidal streams in the famous interacting pair M51 provide important constraints on the system’s dynamical history, including how the interaction has affected the system’s gas disk and star-forming properties. A deep imaging survey showcasing three systems in more isolated environments explores how both secular processes and low-mass satellites influence galaxies’ radial growth. Finally, an analysis of M101’s extended star- forming disk and dwarf companions seeks to understand whether or not star formation in 176 tenuous environments proceeds under different physics than in high-density environments. The main results of these studies are summarized below, and discussed in the context of galaxy evolution theory.

Pre-processing is inefficient in low mass, low density groups. Despite having four massive galaxies in close proximity to each other, IGL in the M96 Galaxy Group makes up much less than 1% of the group’s total luminosity. All of the group’s IGL is in the form of extremely diffuse linear streams, whose total luminosities imply dwarf progenitors. Signs of interaction are also lacking on a galaxy-to-galaxy basis; the central elliptical M105, in particular, appears almost perfectly settled. Hence, while slow- moving tidal interactions in such low mass group environments ought to be effective at liberating and dispersing starlight, the evident lack of even one such interaction in the M96 Group implies that IGL production is highly inefficient in such environments.

The local—not the global—environment inﬂuences galaxy evolution. Despite belonging to a single bound system, the galaxies in the M96 Group are evolving as though they were on their own. Even at very low surface brightness, tidal tails and strongly lopsided isophotes are only visible in the galaxies we observed that have obvious and bright companions, such as M51 and M106, or in galaxies with clear evidence of a past interaction, such as the counter-rotating H I disk in M64. Whether or not a galaxy belongs to a group or cluster, then, appears to inﬂuence its tidal interaction history only insofar as group or cluster galaxies are more likely to have a nearby neighbor with which to interact. The local environment hence appears most important when it comes to galaxy evolution.

Radial migration may not be suﬃcient to populate old outer disks. Non-star-forming, old outer disks can extend several scale lengths beyond the edge of star-forming spiral arms. In the case of M106, this amounts to nearly 20 kpc in radius. Three possibilities exist to explain the extent of these red outer disks: either there exist low-amplitude spiral arms undetectable via the Fourier analysis presented in this dissertation, or else faint spiral arms exist at extended radius and low column density in the gas disk under the THINGS and WHISP detection limit (and assuming that extremely low density

177 spiral arms can migrate stars eﬃciently enough to populate the outer disk), or else all of the galaxies we observed did have extended spiral structure in the past that is now faded. Without past or present extended spiral structure, it is diﬃcult to understand how these stars could have migrated to their current radii.

Outer disk star formation induced by tidal interactions could be rare. A lack of a stellar counterpart to M51’s extended H I tail, coupled with its kinematics, implies that the system’s tidal interaction imparted a great deal of momentum to the gas without compressing it; therefore, despite its coherency and length, the tail contains no pockets of induced star formation. Additionally, despite having two interacting companions, M106’s extended outer disk is dominated by old stellar populations; non-planar motions by the two satellites may thus be heating the outermost gas disk, inhibiting star formation there. On the other hand, star formation has apparently been induced in M101’s outer disk through interaction with its two companions. It is thus likely that star formation may be induced in low density environments only via interactions with very speciﬁc conﬁgurations.

Where it does occur, star formation at low density appears no diﬀerent than star for-

mation at high density. Contrary to integrated studies of galaxies, the ratio FHα/FFUV in populations of H II regions in M101’s outer disk, inner disk, and dwarf companions are statistically indistinguishable once bulk extinction eﬀects are taken into account.

Also, while the diﬀuse ionized gas does show radial trends in FHα/FFUV , it is explainable as a result of systematically fainter and more sparsely distributed H II regions in outer disks, not any change in the kinds of stellar populations that inhabit these fainter regions. Aside from the amount of available fuel, then, star formation does not appear to be aware of its local environment.

Systematically low integrated FHα/FFUV in low density environments likely results from a non-uniform star formation history. Because most diﬀuse Hα emission is di-

rectly traceable to leakage from H II regions, systematically low integrated FHα/FFUV in M101’s outer disk and dwarf companions results from their relatively large fractions of diﬀuse FUV emission with no Hα counterpart. In M101’s case, this FUV

178 emission can be readily traced to a starburst that took place several hundred Myr ago and is now beginning to fade, hence is a result of a non-uniform star formation history. Assuming the aforementioned uniformity in H II region populations persists in other systems, non-uniform SFHs may well account for this behavior generally, though the mechanism that induces the star formation may change with circumstances.

6.2 Directions for Future Work

The results presented in this dissertation contribute to a wide swath of galaxy evolution theory, so of course much work remains to be done. Below are listed possibilities on how to expand on this research in the future.

IGL in fossil groups

Though this dissertation ﬁlls in at least one gap in our understanding of IGL/ICL production and pre-processing—that being the low density, low mass group end—to date, few or no studies have been published on IGL in fossil groups. First deﬁned by Ponman et al. (1994), fossil groups contain both X-ray emitting gas and a massive, central elliptical galaxy that is at least 2 magnitudes brighter than the second brightest group galaxy.

D’Onghia et al. (2005) hence proposed that they assemble early, leaving time for L∗ galaxies within the group to fall to the group center through dynamical friction and merge. Sommer-Larsen (2006) showed that such groups ought to contain a large fraction of IGL, but this has not been observationally conﬁrmed. IGL studies in general are much rarer than ICL studies. Those that have been done tend to focus on “normal” or compact groups (e.g. White et al. 2003; da Rocha & Mendes de Oliveira 2005). Most of the work to date on fossil groups has either been theoretical (Sommer-Larsen 2006; Dariush et al. 2007; von Benda-Beckmann et al. 2008; Cui et al. 2011; Farhang et al. 2017), or has focused on, e.g., the luminosity functions of galaxies within them (e.g. Mendes de Oliveira et al. 2006; Proctor et al. 2011; Gozaliasl et al. 2014; Zarattini et al. 2015) rather than on the IGL. Thus, observing and characterizing the IGL in one or more fossil groups would go far in forming a complete picture of pre-processing in groups of all ages and densities. In turn, this would expand on our knowledge of galaxy

179 evolution within groups; an observed low fraction of IGL in a fossil group, for example, would call for a re-evaluation of their presumed “fossil” status. Such a study would require a telescope with a higher resolution than the Burrell Schmidt. The nearest fossil group is the NGC 6482 Group (Khosroshahi et al. 2004), which is at a redshift of z = 0.0131, or a luminosity distance of ∼57 Mpc (under standard ΛCDM cosmology; this concurs with several of the group’s distance estimates collected in the NASA Extragalactic Database). This group also lies behind a thick patch of Milky Way cirrus, hence is not an ideal target for IGL studies (AV =0.277; Schlegel et al. 1998; Schlaﬂy & Finkbeiner 2011). However, a good target could be found by searching through one or more fossil group catalogs (e.g. Zarattini et al. 2014), which could then be targeted for deep observation using a large aperture telescope.

Frequency of Spiral Arms in Outer Disks

This dissertation showed that a handful of nearby spiral galaxies with non-star-forming outer disks also seem to lack the extended spiral structure that is purported to build them through radial migration. In order to build a general picture of outer disk formation, this small sample needs to be expanded. This may not require as deep observations as were conducted for this dissertation. Spiral arms in the galaxies discussed herein are truncated at the break radius, an observation that is in accordance with previous investigations of breaks in galaxy surveys like SDSS (Pohlen & Trujillo 2006) and S4G (Laine et al. 2014). As such, this may simply require systematically measuring the m = 2 mode amplitude beyond the break radius in all galaxies previously analyzed for disk breaks, hence would require no new observations. The more limited surface brightness depths of these surveys would, however, require a more careful treatment of uncertainty on the m = 2 mode amplitude. This might be done through modeling. Speciﬁcally, the primary sources of noise in the m = 2 amplitude in this dissertation’s study came from isophotes with changing position angles, as well as lopsidedness; one might thus consider constructing and analyzing model galaxies with increasing levels of lopsidedness in the outer disk, with increasingly strong spiral arms overlaid, in order to place limits on the detectability of outer disk spiral arms in these surveys. Morphology may also provide some uncertainty; tightly wound spiral arms may

180 not present a strong m = 2 amplitude, for example, hence this would also need to be taken into account. These limits can then be compared with radial migration models to determine whether or not migration is eﬃcient enough to fully account for the mass of these outer disks. Finally, there remains the possibility that while stellar spiral arms are weak in outer disks, they may be stronger in H I (Koribalski & Lopez-S´ anchez´ 2009; Khoperskov & Bertin 2015). This dissertation examined H I data from interferometric surveys—THINGS and WHISP—which are not ideal for studying low column density gas. It would thus be of great interest to target the galaxies studied in this dissertation (particularly M106) with deeper follow-up H I observations in order to seek out evidence of extended spiral arms in the gas disk as well.

The True Origin of Outer Disk Star Formation

While M101’s disturbed morphology and nearby dwarf companions insistently points toward a tidal origin for its outer disk star formation, it is not clear this is always the case for XUV disks, or disks with extended H II regions. Yet while an interaction may not be called for, evidence for enhanced disk stability in outer disks (Kennicutt 1989; Martin & Kennicutt 2001) implies that some kind of perturbation is required. A more detailed investigation into the origins of extended star formation is thus called for. A similar statistical analysis of H II regions as presented in this dissertation might useful across a much larger sample of galaxies. The data for this type of project already exists; in fact, the similarity in the range of values of FHα/FFUV published in Figure 10 of God- dard et al. (2010) implies that the median and scatter in FHα/FFUV in star forming regions could be constant across the 21 galaxies in their sample as well. This would merely need to be tested in a similar manner as I have done for M101, through systematic comparisons of H II region populations within and between these galaxies. As the study by Goddard et al. (2010) focused on spiral galaxies, data from dwarf galaxies or low surface brightness galaxies should be included as well, whether from the literature or newly observed. Additionally, while it has frequently been suggested that XUV disks are the result of interactions (Gil de Paz et al. 2005; Thilker et al. 2005, 2007a), this has not yet been ver- iﬁed. XUV disks can theoretically also appear in isolated galaxies (Bush et al. 2010), for

181 example via perturbations from substructure in the dark matter halo. Hence, one might consider a systematic investigation into the local environments of XUV disk galaxies, in order to determine the fraction of such galaxies with no companion massive enough to perturb their outer disks. The discovery of even one such isolated XUV disk would be extremely illuminating, as it would verify that star formation in otherwise stable environments does not require a perturbing companion.

182 Bibliography

Adami, C., Slezak, E., Durret, F., et al. 2005, A&A, 429, 39

Aguerri, J. A. L., Castro-Rodr´ıguez, N., Napolitano, N., Arnaboldi, M., & Gerhard, O. 2006, A&A, 457, 771

Aihara, H., Allende Prieto, C., An, D., et al. 2011, ApJS, 193, 29

Akiyama, M., Minowa, Y., Kobayashi, N., et al. 2008, ApJS, 175, 1-28

Alberts, S., Calzetti, D., Dong, H., et al. 2011, ApJ, 731, 28

Alloin, D., & Nieto, J. -L. 1982, A&AS, 50, 491

Appleton, P. N., Foster, P. A., & Davies, R. D. 1986, MNRAS, 221, 393

Appleton, P. N., Pedlar, A., & Wilkinson, A. 1990, ApJ, 357, 426

Arp, H. 1966, Atlas of Peculiar Galaxies (Pasadena, CA: California Institute of Technol- ogy)

Astropy Collaboration, Robitaille, T. P., Tollerud, E. J., et al. 2013, A&A, 558, A33

Athanassoula, E. 1980, A&A, 88, 184

Athanassoula, E. 2010, in ASP Conf. Ser. 421, Galaxies in Isolation: Exploring Nature Versus Nuture, ed. L. Verdes-Montenegro, et al. (San Francisco, CA: ASP), 157

Athanassoula, E., Machado, R. E. G., & Rodionov, S. A. 2013, MNRAS, 429, 1949

Athanassoula, E., Rodionov, S. A., Peschken, N., & Lambert, J. C. 2016, ApJ, 821, 90

183 Atkinson, A. M., Abraham, R. G., & Ferguson, A. M. N. 2013, ApJ, 765, 28

Azzollini, R., Trujillo, I., & Beckman, J. E. 2008, ApJL, 679, L69

Bahcall, N. A. 2000, in Allen’s Astrophysical Quantities, ed. A. Cox, AIP Press (New York)

Bakos, J., & Trujillo, I. 2013, Memorie della Societa Astronomica Italiana Supplementi, 25, 21

Bakos, J., Trujillo, I., & Pohlen, M. 2008, ApJL, 683, L103

Barbon, R., Benacchio, L., & Capaccioli, M. 1976, A&A, 51, 25

Barker, M. K., Sarajedini, A., Geisler, D., Harding, P., & Schommer, R. 2007, AJ, 133, 1125

Barnes, J. E. 1988, ApJ, 331, 699

Barnes, J. E. 1989, Nature, 338, 123

Barnes, J. E. 1992, ApJ, 393, 484

Barnes, J. E. 2002, MNRAS, 333, 481

Barnes, J. E., & Hernquist, L. E. 1991, ApJL, 370, L65

Barnes, J. E., & Hernquist, L. 1992, Nature, 360, 715

Barnes, J. E., & Hernquist, L. 1996, ApJ, 471, 115

Barnes, K. L., van Zee, L., & Skillman, E. D. 2011, ApJ, 743, 137

Barnes, K. L., van Zee, L., & Dowell, J. D. 2013, ApJ, 775, 40

Beale, J. S., & Davies, R. D. 1969, Nature, 221, 531

Bekki, K., Koribalski, B.S., Ryder, S.D., & Couch, W.J. 2005, MNRAS, 357, L21

Bell, E. F., & de Jong, R. S. 2001, ApJ, 550, 212

Bell, E. F., & Kennicutt, R. C., Jr. 2001, ApJ, 548, 681

184 Bell, E. F., Zucker, D. B., Belokurov, V., et al. 2008, ApJ, 680, 295-311

Belokurov, V., Zucker, D. B., Evans, N. W., et al. 2006, ApJL, 642, L137

Berentzen, I., Athanassoula, E., Heller, C. H., & Fricke, K. J. 2003, MNRAS, 341, 343

Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393

Bianchi, L., & GALEX Team 1999, Mem. Soc. Astron. Italiana, 70

Bianchi, L., & GALEX Team 2000, Mem. Soc. Astron. Italiana, 71, 1117

Bianchi, L., Madore, B., Thilker, D., Gil de Paz, A., & GALEX Science Team 2003, Bul- letin of the American Astronomical Society, 35, 91.12

Bigiel, F., Leroy, A., Seibert, M., et al. 2010, ApJL, 720, L31

Bigiel, F., Leroy, A., Walter, F., et al. 2008, AJ, 136, 2846

Bigiel, F., Leroy, A., Walter, F., et al. 2010, AJ, 140, 1194

Binney, J., & Lacey, C. 1988, MNRAS, 230, 597

Binney, J., & Merriﬁeld, M. 1998, Galactic astronomy / James Binney and Michael Merri- ﬁeld. Princeton, NJ : Princeton University Press, 1998. (Princeton series in astrophysics)

Bland-Hawthorn, J., Vlajic,´ M., Freeman, K. C., & Draine, B. T. 2005, ApJ, 629, 239

Boissier, S., Gil de Paz, A., Boselli, A., et al. 2007, ApJS, 173, 524

Boissier, S., Gil de Paz, A., Boselli, A., et al. 2008, ApJ, 681, 244-257

Bolatto, A. D., Leroy, A. K., Jameson, K., et al. 2011, ApJ, 741, 12

Bonnell, I. A., Vine, S. G., & Bate, M. R. 2004, MNRAS, 349, 735

Boquien, M., Duc, P. -A., Braine, J., et al. 2007, A&A, 467, 93

Borlaﬀ, A., Eliche-Moral, M. C., Rodr´ıguez-Perez,´ C., et al. 2014, A&A, 570, 103

Bose, S., & Kumar, B. 2014, ApJ, 782, 98

185 Boselli, A., Boissier, S., Cortese, L., et al. 2009, ApJ, 706, 1527

Bosma, A., van der Hulst, J. M., & Sullivan, W. T. 1977, A&A, 57, 373

Bot, C., Helou, G., Latter, W. B., et al. 2009, AJ, 138, 452

Bottema, R., Shostak, G. S., & van der Kruit, P. C. 1987, Nature, 328, 401

Bournaud, F., Combes, F., Jog, C. J., & Puerari, I. 2005, A&A, 438, 507

Bournaud, F., Jog, C. J., & Combes, F. 2005, A&A, 437, 69

Bouwens, R., & Illingworth, G. 2006, NAR, 50, 152

Bowler, R. A. A., Dunlop, J. S., McLure, R. J., & McLeod, D. J. 2017, MNRAS, 466, 3612

Braun, R., Walterbos, R. A. M., & Kennicutt, R. C. 1992, Nature, 360, 442

Bregman, J. N., Hogg, D. E., & Roberts, M. S. 1992, ApJ, 387, 484

Breiman, L. 1973, Statistics with a View toward Applications (Houghton Miﬄin, Boston)

Bresolin, F., Ryan-Weber, E., Kennicutt, R. C., Goddard, Q. 2009, ApJ, 695, 580

Bressan, A., Marigo, P., Girardi, L., et al. 2012, MNRAS, 427, 127

Brown, A. G. A., de Geus, E. J., & de Zeeuw, P. T. 1994, A&A, 289, 101

Brown, B. W., & Forsythe, A. B. 1974, Robust tests for equality of variances, J. Amer. Statist. Assoc., 69, 364

Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000

Bruzzese, S. M., Meurer, G. R., Lagos, C. D. P., et al. 2015, MNRAS, 447, 618

Buat, V. 1992, A&A, 264, 444

Buat, V., Iglesias-Paramo,´ J., Seibert, M., et al. 2005, ApJL, 619, L51

Budzynski, J. M., Koposov, S. E., McCarthy, I. G., & Belokurov, V. 2014, MNRAS, 437, 1362

186 Bullock, J. S., & Johnston, K. V. 2005, ApJ, 635, 931

Burkert, A., Brodie, J., & Larsen, S. 2005, ApJ, 628, 231

Burkhead, M. S. 1978, ApJS, 38, 147

Burkholder, V., Impey, C., & Sprayberry, D. 2001, AJ, 122, 2318

Burton, W. B. 1971, A&A, 10, 76

Busarello, G., Capaccioli, M., D’Onofrio, M., et al. 1996, A&A, 314, 32

Bush, S. J., Cox, T. J., Hayward, C. C., et al. 2010, ApJ, 713, 780

Bush, S. J., Cox, T. J., Hernquist, L., Thilker, D., & Younger, J. D. 2008, ApJL, 683, L13

Calzetti, D. 2001, PASP, 113, 1449

Calzetti, D., Kennicutt, R. C., Jr., Bianchi, L., et al. 2005, ApJ, 633, 871

Cappellari, M., Bertola, F., Burstein, D., Buson, L. M., Greggio, L., & Renzini, A. 1999, ApJL, 515, L17

Carrasco, E. R., Conselice, C. J., & Trujillo, I. 2010, MNRAS, 405, 2253

Castro-Rodr´ıguez, N., Aguerri, J. A. L., Arnaboldi, M., et al. 2003, A&A, 405, 803

Chabrier, G. 2003, PASP, 115, 763

Chiappini, C., Matteucci, F., & Gratton, R. 1997, ApJ, 477, 765

Choi, P. I., Guhathakurta, P., & Johnston, K. V. 2002, AJ, 124, 310

Choi, Y., Dalcanton, J. J., Williams, B. F., et al. 2015, ApJ, 810, 9

Ciardullo, R., Feldmeier, J. J., Jacoby, G. H., et al. 2002, ApJ, 577, 31

Ciardullo, R., Jacoby, G. H., & Dejonghe, H. B. 1993, ApJ, 414, 454

Colina, L., Garcia Vargas, M. L., Mas-Hesse, J. M., Alberdi, A., & Krabbe, A. 1997, ApJL, 484, L41

187 Comeron,´ S., Knapen, J. H., Beckman, J. E., et al. 2010, MNRAS, 402, 2462

Conselice, C. J. 2014, ARA&A, 52, 291

Conselice, C. J., & Arnold, J. 2009, MNRAS, 397, 208

Contini, E., De Lucia, G., Villalobos, A.,´ & Borgani, S. 2014, MNRAS, 437, 3787

Cooper, A. P., Cole, S., Frenk, C. S., et al. 2010, MNRAS, 406, 744

Cortese, L., Boselli, A., Buat, V., et al. 2006, ApJ, 637, 242

Cortese, L., Gavazzi, G., Boselli, A., et al. 2006, A&A, 453, 847

Courteau, S., de Jong, R. S., & Broeils, A. H. 1996, ApJL, 457, L73

Courtes,` G., & Cruvellier, P. 1961, Compt. Rend. Acad. Sci. Paris, 253, 218

Cox, T. J., Jonsson, P., Somerville, R. S., Primack, J. R., & Dekel, A. 2008, MNRAS, 384, 386

Crnojevic,´ D., Sand, D. J., Spekkens, K., et al. 2016, ApJ, 823, 19

Crocker, A. F., Chandar, R., Calzetti, D., et al. 2015, ApJ, 808, 76

Croxall, K. V., Pogge, R. W., Berg, D. A., Skillman, E. D., & Moustakas, J. 2016, ApJ, 830, 4

Cui, W., Springel, V., Yang, X., De Lucia, G., & Borgani, S. 2011, MNRAS, 416, 2997

Cuillandre, J.-C., Lequeux, J., Allen, R. J., Mellier, Y., & Bertin, E. 2001, ApJ, 554, 190 da Costa, L., Pellegrini, P. S., Sargent, W. L. W., et al. 1988, ApJ, 327, 544

Daddi, E., Elbaz, D., Walter, F., et al. 2010, ApJL, 714, L118

Dariush, A., Khosroshahi, H. G., Ponman, T. J., et al. 2007, MNRAS, 382, 433 da Rocha, C., & Mendes de Oliveira, C. 2005, MNRAS, 364, 1069 da Rocha, C., Ziegler, B. L., & Mendes de Oliveira, C. 2008, MNRAS, 388, 1433

188 da Silva, R. L., Fumagalli, M., & Krumholz, M. R. 2014, MNRAS, 444, 3275

Davidge, T. J. 2003, AJ, 125, 3046

Davidge, T. J. 2006, ApJ, 641, 822

Davidge, T. J. 2007, ApJ, 664, 820

Davidge, T. J. 2010, ApJ, 718, 1428

Davis, M., Efstathiou, G., Frenk, C. S., White, S. D. M. 1985, ApJ, 292, 371

Davis, M., Huchra, J., Latham, D. W., & Tonry, J. 1982, ApJ, 253, 423

Debattista, V. P., Mayer, L., Carollo, C. M., et al. 2006, ApJ, 645, 209

de Blok, W. J. G., & van der Hulst, J. M. 1998, A&A, 336, 49

de Blok, W. J. G., & Walter, F. 2003, MNRAS, 341, L39

de Blok, W. J. G., Walter, F., Brinks, E., et al. 2008, AJ, 136, 2648

de Grijs, R., Kregel, M., & Wesson, K. H. 2001, MNRAS, 324, 1074

de Jong, R. S. 2008, MNRAS, 388, 1521

de Lapparent, V., Geller, M. J., & Huchra, J. P. 1986, ApJL, 302, L1

de Vaucouleurs, G. 1948, Annales d’Astrophysique, 11, 247

de Vaucouleurs, G. 1958, ApJ, 128, 465

de Vaucouleurs, G. 1959, Handbuch der Physik, 53, 311

de Vaucouleurs, G. 1975, in Stars and Stellar Systems, Vol 9, Galaxies and the Universe, ed. A. Sandage, M. Sandage, & J. Kristian (Chicago, IL: Univ. of Chicago Press), 557 de Vaucouleurs, G., & Capaccioli, M. 1979, ApJS, 40, 699 de Vaucouleurs, G., de Vaucouleurs, A., Corwin, H. G. Jr., et al. 1991, Third Reference Catalogue of Bright Galaxies (New York: Springer)

189 Devereux, N. A., Kenney, J. D. P., & Young, J. S. 1992, AJ, 103, 784 de Wit, W. J., Testi, L., Palla, F., & Zinnecker, H. 2005, A&A, 437, 247

Dobbs, C. L., Theis, C., Pringle, J. E., & Bate, M. R. 2010, MNRAS, 403, 625

Donahue, M., Aldering, G., & Stocke, J. T. 1995, ApJL, 450, 45

Donas, J., Deharveng, J. M., Laget, M., Milliard, B., & Huguenin, D. 1987, A&A, 180, 12

Donas, J., Milliard, B., Laget, M., & Deharveng, J. M. 1981, A&A, 97, L7

D’Onghia, E., Sommer-Larsen, J., Romeo, A. D., et al. 2005, ApJL, 630, L109

D’Onghia, E., Vogelsberger, M., Faucher-Giguere, C.-A., & Hernquist, L. 2010, ApJ, 725, 353

Dopita, M. A., & Sutherland, R. S. 2003, Astrophysics of the diﬀuse universe, Berlin, New York: Springer, 2003

Doroshkevich, A., Tucker, D. L., Allam, S., & Way, M. J. 2004, A&A, 418, 7

Douglas, N. G., Napolitano, N. R., Romanowsky, A. J., et al. 2007, ApJ, 664, 257

Dressler, A. 1980, ApJ, 236, 351

D’Souza, R., Kauﬀman, G., Wang, J., & Vegetti, S. 2014, MNRAS, 443, 1433

Duc, P.-A., Brinks, E., Springel, V., et al. 2000, AJ, 120, 1238

Duncan, K., & Conselice, C. J. 2015, MNRAS, 451, 2030

Durbala, A., del Olmo, A., Yun, M. S., et al. 2008, AJ, 135, 130

Durrell, P. R., Decesar, M. E., Ciardullo, R., Hurley-Keller, D., & Feldmeier, J. J. 2004, in IAU Symp. 217, Recycling Intergalactic and Interstellar Matter, ed. P.-A. Duc, J. Braine, & E. Brinks (Cambridge: Cambridge Univ. Press), 90

Durrell, P. R., Mihos, J. C., Feldmeier, J. J., Jacoby, G. H., & Ciardullo, R. 2003, ApJ, 582, 170

190 Dyson, J. E., & Williams, D. A. 1980, New York, Halsted Press, 1980. 204 p.

Edvardsson, B., Andersen, J., Gustafsson, B., et al. 1993, A&A, 275, 101

Elmegreen, B. G. 2002, ApJ, 577, 206

Elmegreen, B. G. 2011, ApJ, 737, 10

Elmegreen, D. M., Chromey, F. R., Santos, M., & Marshall, D. 1997, ApJ, 114, 1850

Elmegreen, B. G., & Hunter, D. A. 2015, ApJ, 805, 145

Elmegreen, B. G., & Struck, C. 2016, ApJ, 830, 115

Erwin, P. 2004, A&A, 415, 941

Erwin, P., Beckman, J. E., & Pohlen, M. 2005, ApJL, 626, L81

Erwin, P., Gutierrez,´ L., & Beckman, J. E. 2012, ApJL, 744, L11

Erwin, P., Pohlen, M., & Beckman, J. E. 2008, AJ, 135, 20

Fall, S. M., & Efstathiou, G. 1980, MNRAS, 193, 189

Farhang, A., Khosroshahi, H. G., Mamon, G. A., Dariush, A. A., & Raouf, M. 2017, ApJ, 840, 58

Fathi, K., Allen, M., Boch, T., Hatziminaoglou, E., & Peletier, R. F. 2010, MNRAS, 406, 1595

Feitzinger, J. V., & Galinski, T. 1985, A&AS, 61, 503

Feldmeier, J. J., Ciardullo, R., Jacoby, G. H., Durrell, P. R., & Mihos, J. C. 2004, in IAU Symp. 217, Recycling Intergalactic and Interstellar Matter, ed. P.-A. Duc, J. Braine, & E. Brinks (Cambridge: Cambridge Univ. Press), 64

Feldmeier, J. J., Mihos, J. C., Morrison, H. L., Rodney, S. A., & Harding, P. 2002, ApJ, 575, 587

Feldmeier, J. J., Mihos, J. C., Morrison, H. L., et al. 2004, ApJ, 609, 617

191 Ferguson, A. M. N., & Clarke, C. J. 2001, MNRAS, 325, 781

Ferguson, A. M. N., Gallagher, J. S., & Wyse, R. F. G. 1998, AJ, 116, 673

Ferguson, A. M. N., Wyse, R. F. G., Gallagher, J. S., III, & Hunter, D. A. 1996, AJ, 111, 2265

Ferguson, A. M. N., Wyse, R. F. G., Gallagher, J. S., & Hunter, D. A. 1998, ApJL, 506, L19

Filho, M. E., Sanchez´ Almeida, J., Amor´ın, R., et al. 2016, ApJ, 820, 109

Finlator, K., Ivezic,´ Z.,ˇ Fan, X., et al. 2000, AJ, 120, 2615

Forbes, D. A., Beasley, M. A., Bekki, K., Brodie, J. P., & Strader, J. 2003, Sci, 301, 1217

Foreman-Mackey, D., Hogg, D. W., Lang, D., & Goodman, J. 2013, PASP, 125, 925

Fouque,´ P., Gourgoulhon, E., Chamaraux, P., & Paturel, G. 1992, A&AS, 93, 211

Foyle, K., Courteau, S., & Thacker, R. J. 2008, MNRAS, 386, 1821

Franx, M., van Gorkum, J. H., & de Zeeuw, P. T. 1994, ApJ, 436, 642

Freeman, K. C. 1970, ApJ, 160, 811

Frenk, C. S., White, S. D. M., Davis, M., & Efstathiou, G. 1988, ApJ, 327, 507

Fujita, Y. 2004, PASJ, 56, 29

Garc´ıa-Burillo, S., Combes, F., Hunt, L. K., et al. 2003, A&A, 407, 485

Garc´ıa-Ruiz, I., Sancisi, R., & Kuijken, K. 2002, A&A, 394, 769

Garcia-Segura, G., & Franco, J. 1996, ApJ, 469, 171

Gauthier, J.-R., Dubinski, J., & Widrow, L. M. 2006, ApJ, 653, 1180

Geha, M., Guhathakurta, P., Rich, R. M., & Cooper, M. C. 2006, AJ, 131, 332

Gelb, J. M., & Bertschinger, E. 1994, ApJ, 436, 467

192 Geller, M. J., & Huchra, J. P. 1983, ApJS, 52, 61

Gerber, R. A., Lamb, S. A., & Balsara, D. S. 1992, ApJL, 399, L51

Gies, D. R. 1987, ApJS, 64, 545

Gilbert, K. M., Guhathakurta, P., Beaton, R. L., et al. 2012, ApJ, 760, 76

Gil de Paz, A., Boissier, S., Madore, B. F., et al. 2007, ApJS, 173, 185

Gil de Paz, A., Madore, B. F., Boissier, S., et al. 2005, ApJL, 627, L29

Glazebrook, K., Blake, C., Economou, F., Lilly, S., & Colless, M. 1999, MNRAS, 306, 843

Gnedin, O. Y. 2003, ApJ, 589, 752

Goddard, Q. E., Kennicutt, R. C., & Ryan-Weber, E. V. 2010, MNRAS, 405, 2791

Goldreich, P., & Lynden-Bell, D. 1965, MNRAS, 130, 125

Gonzalez, A. H., Sivanandam, S., Zabludoﬀ, A. I., & Zaritsky, D. 2013, ApJ, 778, 14

Gonzalez, A. H., Zabludoﬀ, A. I., Zaritsky, D., & Dalcanton, J. J. 2000, ApJ, 536, 561

Gonzalez, A. H., Zaritsky, D., & Zubludoﬀ, A. I. 2007, ApJ, 666, 147

Gordon, K. D., Perez-Gonz´ alez,´ P. G., Misselt, K. A., et al. 2004, ApJS, 154, 215

Goto, T., Yamauchi, C., Fujita, Y., et al. 2003, MNRAS, 346, 601

Gott, J. R., & Rees, M. J. 1975, A&A, 45, 365

Goudfrooij, P., Hansen, L., Jørgensen, H .E., et al. 1994, ApJS, 104, 179

Gozaliasl, G., Khosroshahi, H. G., Dariush, A. A., et al. 2014, A&A, 571, A49

Graham et al. 1997, ApJ, 477, 535

Greenawalt, B., Walterbos, R. A. M., Thilker, D., & Hoopes, C. G. 1998, ApJ, 506, 135

Gregg, M. D., & West, M. J. 1998, Nature, 396, 549

193 Gu, Q.-S., Liao, X.-H., Huang, J.-H., Qu, Q.-Y., & Su, H.-J. 1996, A&A, 314, 18

Gunawardhana, M. L. P., Hopkins, A. M., Sharp, R. G., et al. 2011, MNRAS, 415, 1647

Guo, Q., White, S., Boylan-Kolchin, M., et al. 2011, MNRAS, 413, 101

Gutierrez,´ L., Erwin, P., Aladro, R., & Beckman, J. E. 2011, AJ, 142, 145

Haﬀner, L. M., Reynolds, R. J., & Tufte, S. L. 1999, ApJ, 523, 223

Harmsen, B., Monachesi, A., Bell, E. F., et al. 2017, MNRAS, 466, 1491

Harris, W. E., Harris, G. L. H., Layden, A. C., & Wehner, E. M. H. 2007, ApJ, 666, 903

Hartwick, F. D. A. 1971, ApJ, 163, 431

Haywood, M. 2006, MNRAS, 371, 1760

Heckman, T., Krolik, J., Meurer, G., et al. 1995, ApJ, 452, 549

Helfer, T. T., Thornley, M. D., Regan, M. W., et al. 2003, ApJS, 145, 259

Helmboldt, J. F., Walterbos, R. A. M., Bothun, G. D., & O’Neil, K. 2005, ApJ, 630, 824

Helmboldt, J. F., Walterbos, R. A. M., Bothun, G. D., O’Neil, K., & Oey, M. S. 2009, MNRAS, 393, 478

Henriques, B. M. B., & Thomas, P. A. 2010, MNRAS, 403, 768

Hermanowicz, M. T., Kennicutt, R. C., & Eldridge, J. J. 2013, MNRAS, 432, 3097

Hernquist, L. 1990, in Dynamics and Interactions of Galaxies, ed. R. Wielen (Berlin: Springer), 108

Hernquist, L. 1992, ApJ, 400, 460

Hernquist, L., & Mihos, J. C. 1995, 448, 41

Hernquist, L. & Quinn, P. J. 1987, ApJ, 312, 1

Hernquist, L. & Quinn, P. J. 1988, ApJ, 331, 682

194 Hernquist, L., Spergel, D. N., & Heyl, J. S. 1993, ApJ, 416, 415

Hernquist, L., & Weil, M. L. 1992, Nature, 358, 734

Herpich, J., Tremaine, S., & Rix, H. W. 2017, MNRAS, 467, 5022

Herrmann, K. A. & Ciardullo, R. 2009, ApJ, 705, 1686

Herrmann, K. A., Ciardullo, R., & Sigurdsson, S. 2009, ApJL, 693, L19

Heyl, J. S., Hernquist, L., & Spergel, D. N. 1994, ApJ, 427, 165

Hibbard, J. E., & Mihos, J. C. 1995, AJ, 110, 140

Hickson, P., Richstone, D. O., & Turner, E. L. 1977, ApJ, 213, 323

Hodge, P. W. 1971, ARA&A, 9, 35

Høg, E., Fabricius, C., Makarov, V. V., et al. 2000, A&A, 355, L27

Hoopes, C. G., & Walterbos, R. A. M. 2000, ApJ, 541, 597

Hoopes, C. G., Walterbos, R. A. M., & Bothun, G. D. 2001, ApJ, 559, 878

Hoopes, C. G., Walterbos, R. A. M., & Greenwalt, B. E. 1996, AJ, 112, 1429

Hoopes, C. G., Walterbos, R. A. M., & Rand, R. J. 1999, ApJ, 522, 669

Hopkins, P. F., Bundy, K., Murray, N., et al. 2009, MNRAS, 398, 898

Hopkins, P. F., Cox, T. J., Hernquist, L., et al. 2013, MNRAS, 430, 1901

Hopkins, P. F., Cox, T. J., Younger, J. D., & Hernquist, L. 2009, ApJ, 691, 1168

Hopkins, P. F., Somerville, R. S., Cox, T. J., et al. 2009, MNRAS, 397, 802

Horellou, C., & Koribalski, B. 2007, A&A, 464, 155

Hoversten, E. A., & Glazebrook, K. 2008, ApJ, 675, 163-187

Howard, S., & Byrd, G. G. 1990, AJ, 99, 1798

195 Howard, S., Keel, W. C., Byrd, G., & Burkey, J. 1993, ApJ, 417, 502

Hu, W., & Sugiyama, N. 1995, ApJ, 444, 489

Hu, W., Sugiyama, N., & Silk, J. 1997, Nature, 386, 6620, 37

Huber, P. J. 1981, Robust Statistics (Wiley, New York)

Huchra, J. P., & Geller, M. J. 1982, ApJ, 257, 423

Hughes, A. et al. 2013, ApJ, 779, 46

Humphreys, E. M. L., Reid, M. J., Moran, J. M., Greenhill, L. J., & Argon, A. L. 2013, ApJ, 775, 13

Hunter, D. A., Elmegreen, B. G., & Baker, A. L. 1998, ApJ, 493, 595

Hunter, D. A., Elmegreen, B. G., & Ludka, B. C. 2010, AJ, 139, 447

Hunter, D. A., Elmegreen, B. G., Oh, S.-H., et al. 2011, AJ, 142, 121

Hunter, D. A., & Gallagher, J. S., III 1986, PASP, 98, 5

Hunter, D. A., Gallagher, J. S., & Rautenkranz, D. 1982, ApJS, 49, 53

Hunter, D. A., & Plummer, J. D. 1996, ApJ, 462, 732

Hunter, J. D. 2007, CSE, 9, 90

Ibata, R., Irwin, M., Lewis, G., Ferguson, A. M. N., & Tanvir, N. 2001, Nature, 412, 49

Ibata, R. A., Lewis, G. F., McConnachie, A. W., et al. 2014, ApJ, 780, 128

Ivezic,´ Z.,ˇ Smith, J. A., Miknaitis,G., et al. 2007, AJ, 134, 973

Jacobs, B. A., Rizzi, L., Tully, R. B., et al. 2009, AJ, 138, 332

Janesh, W., Morrison, H. L., Ma, Z., et al. 2016, ApJ, 816, 80

Janowiecki, S., Mihos, J. C., Harding, P., et al. 2010, ApJ, 715, 972

Jarrett, T. H. et al. 2006, AJ, 131, 261 196 Jeans, J. H. 1902, Philosophical Transactions of the Royal Society of London Series A, 199, 1

Jedrzejewski, R. I. 1987, MNRAS, 226, 747

Jenkins, A., Frenk, C. S., White, S. D. M., et al. 2001, MNRAS, 321, 372

Jones, M. E., Edge, A. C., Grainge, K., et al. 2005, MNRAS, 357, 518

Jones, L. R., Ponman, T. J., Horton, A., et al. 2003, MNRAS, 343, 627

Julian, W. H., & Toomre, A. 1966, ApJ, 146, 810

Kannan, R., Maccio,` A. V., Fontanot, F., et al. 2015, MNRAS, 452, 4347

Kapferer, W., Schindler, S., Knollmann, S. R., & van Kampen, E. 2010, A&A, 516, A41

Karachentsev, I. D. 2005, AJ, 129, 178

Karachentseva, V. E., & Karachentsev, I. D. 1998, A&AS, 127, 409

Karachentsev, I. D., & Kudrya, Y. N. 2014, AJ, 148, 50

Karachentsev, I. D., Makarov, D. I., & Kaisina, E. I. 2013, AJ, 145, 101

Karachentsev, I. D., Sharina, M. E., Dolphin, A. E., et al. 2003, A&A, 398, 467

Kauﬀmann, G., Li, C., Fu, J. et al. 2012, MNRAS, 422, 997

Kellar, J. A., Salzer, J. J., Wegner, G., Gronwall, C., & Williams, A. 2012, AJ, 143, 145

Kemp, S. N., de la Fuente, E., Franco-Baldera, A., & Meaburn, J. 2005, ApJ, 624, 680

Kenney, J. D. P., Scoville, N. Z., & Wilson, C. D. 1991, ApJ, 366, 432

Kennicutt, R. C., Jr. 1983, ApJ, 272, 54

Kennicutt, R. C. 1989, ApJ, 344, 685

Kennicutt, R. C., Jr. 1998, ApJ, 498, 541

Kennicutt, R. C., Jr., Calzetti, D., Walter, F., et al. 2007, ApJ, 671, 333 197 Kennicutt, R. C., & Edgar, B. K. 1986, ApJ, 300, 132

Kennicutt, R. C., & Evans, N. J. 2012, ARA&A, 50, 531

Kennicutt, R. C., Jr., Lee, J. C., Funes, J. G., et al. 2008, ApJS, 178, 247-279

Kennicutt, R. C., Jr., Tamblyn, P., & Congdon, C. E. 1994, ApJ, 435, 22

Khoperskov, S. A., & Bertin, G. 2015, MNRAS, 451, 2889

Khosroshahi, H. G., Jones, L. R., & Ponman, T. J. 2004, MNRAS, 349, 1240

Kim, E., Kim, M., Hwang, N., Lee, M. G., Chun, M.-Y., & Ann, H. B. 2011, MNRAS, 412, 1881

King, I. R. 1971, PASP, 83, 199

Klypin, A., Kravtsov, A. V., Valenzuela, O., & Prada, F. 1999, ApJ, 522, 82

Knapen, J. H., Peters, S. P. C., van der Kruit, P. C., et al. 2016, The General Assembly of Galaxy Halos: Structure, Origin and Evolution, 317, 39

Koribalski, B. S., & Lopez-S´ anchez,´ A.´ R. 2009, MNRAS, 400, 1749

Kormendy, J. 1982, in Saas-Fee Advanced Course 12: Morphology and Dynamics of Galaxies, ed. L. Martinet & M. Mayor (Sauverny, Switzerland: Observatoire de Geneve),` 113

Kormendy, J., & Bender, R. 2012, ApJS, 198, 2

Kormendy, J., & Kennicutt, R. C. Jr. 2004, ARA&A, 42, 603

Kormendy, J., & Norman, C. A. 1979, ApJ, 233, 539

Kornreich, D. A., Haynes, M. P., & Lovelace, R. V. E. 1998, AJ, 116, 2154

Kotulla, R., Fritze, U., Weilbacher, P., & Anders, P. 2009, MNRAS, 396, 462

Koyama, Y., Kodama, T., Hayashi, M., et al. 2015, MNRAS, 453, 879

Kraljic, K., Bournaud, F., & Martig, M. 2012, ApJ, 757, 60 198 Kregel, M., & van der Kruit, P. C. 2004, MNRAS, 355, 143

Kregel, M., van der Kruit, P. C., & de Grijs, R. 2002, MNRAS, 334, 646

Krisciunas, K. 1997, PASP, 109, 1181

Kroupa, P. 2001, MNRAS, 322, 231

Krumholz, M. R., Dekel, A., McKee, C. E. 2012, ApJ, 745, 69

Laine, J., Laurikainen, E., & Salo, H. 2016, A&A, 596, A25

Laine, J., Laurikainen, E., Salo, H., et al. 2014, MNRAS, 441, 1992

Lamb, J. B., Oey, M. S., Werk, J. K., & Ingleby, L. D. 2010, ApJ, 725, 1886

Landolt, A. U. 1992, AJ, 104, 340

Larson, R. B. 1973, MNRAS, 161, 133

Laurikainen, E., Salo, H., Athanassoula, E., et al. 2013, MNRAS, 430, 3489

Law, D. R., Johnston, K. V., & Majewski, S. R. 2005, ApJ, 619, 807

Lee, J. C., Gil de Paz, A., Kennicutt, R. C., Jr., et al. 2011, ApJS, 192, 6

Lee, J. C., Gil de Paz, A., Tremonti, C., et al. 2009, ApJ, 706, 599-613

Lee, J. H., Kim, S. C., Park, H. S., et al. 2011, ApJ, 740, 42

Lee, J. H., Kim, S. C., Ree, C. H., et al. 2012, ApJ, 754, 80

Lee, S.-K., Ferguson, H. C., Somerville, R. S., Wiklind, T., & Giavalisco, M. 2010, ApJ, 725, 1644

Leitherer, C., Schaerer, D., Goldader, J. D., et al. 1999, ApJS, 123, 3

Lelievre,` M., & Roy, J.-R. 2000, AJ, 120, 1306

Lemasle, B., Franc¸ois, P., Genovali, K., et al. 2013, A&A, 558, A31

Lemonias, J. J., Schiminovich, D., Thilker, D., et al. 2011, ApJ, 733, 74 199 Leroy, A. K., Bigiel, F., de Blok, W. J. G., et al. 2012, AJ, 144, 3

Leroy, A. K., Bolatto, A. D., Simon, J. D., & Blitz, L. 2005, ApJ, 625, 763

Leroy, A. K., Walter, F., Bigiel, F., et al. 2009, AJ, 137, 4670

Levene, H. 1960, Robust tests for equality of variances. In Contributions to Probability and Statistics (I. Olkin, ed.) 278-292, Stanford Univ. Press, Palo Alto, CA

Lewis, G. F., Braun, R., McConnachie, A. W., et al. 2013, ApJ, 763, 4

Li, Z., Ho, L. C., Barth, A. J., & Peng, C. Y. 2011, ApJS, 197, 22

Lim, T. S., & Loh, W. Y. 1996, Computational Statistics & Data Analysis, 22, 287

Lin, Y.-T., & Mohr, J. J. 2004, ApJ, 617, 879

Lin, D. N. C., & Pringle, J. E. 1987, ApJL, 320, L87

Loebman, S. R., Debattista, V. P., Nidever, D. L., et al. 2016, ApJL, 818, L6

Longobardi, A., Arnaboldi, M., Gerhard, O., & Hanuschik, R. 2015, A&A, 579, A135

Luck, R. E., & Lambert, D. L. 2011, AJ, 142, 136

Lupton, R. H. 2005, SDSS Data Release 8, Transformations Between SDSS Magnitudes and Other Systems, www.sdss3.org/dr8/algorithms/sdssUBVRITransform.php#Lupton2005

Lupton, R. H., Gunn, J. E., & Szalay, A. S. 1999, AJ, 118, 1406

Lynden-Bell, D. 1976, MNRAS, 174, 695

Lynds, R. & Toomre, A. 1976, ApJ, 209, 382

Ma, Z. 2015, PhD thesis, Case Western Reserve Univ.

Maciel, W. J., & Costa, R. D. D. 2010, in IAU Symp. S265, Chemical Abundances in the Universe: Connecting First Stars to Planets, ed. K. Cunha, M. Spite, & B. Barbuy (Cambridge: Cambridge Univ. Press), 317

200 Madore, B. F., van den Bergh, S., & Rogstad, D. H. 1974, ApJ, 191, 317

Madsen, G. J., Reynolds, R. J., & Haﬀner, L. M. 2006, ApJ, 652, 401

Maia, M. A. G., Willmer, C. N. A., & da Costa, L. N. 1998, AJ, 115, 49

Majewski, S. R., Skrutskie, M. F., Weinberg, M. D., & Ostheimer, J. C. 2003, ApJ, 599, 1082

Makarov, D., Prugniel, P., Terekhova, N. Courtois, H., & Vauglin, I. 2014, A&A, 570, A13

Malin, D. F. 1984, in IAU Coll. 58, Astronomy with Schmidt Type Telescopes, ed. M. Capaccioli (Reidel: Dordrecht), 73

Malin, D. F., & Carter, D. 1983, ApJ, 274, 534

Malin, D., & Hadley, B. 1997, Publications of the Astron. Soc. of Australia, 14, 52

Malphrus, B. K., Simpson, C. E., Gottesman, S. T., & Hawarden, T. G. 1997, AJ, 114, 142

Maltby, D. T., Aragon-Salamanca,´ A., Gray, M. E., et al. 2015, MNRAS, 447, 1506

Maltby, D. T., Gray, M. E., Aragon-Salamanca,´ A., et al. 2012, MNRAS, 419, 669

Martin, C. L., & Kennicutt, R. C., Jr. 2001, ApJ, 555, 301

Martin, D. C., Fanson, J., Schiminovich, D., et al. 2005, ApJL, 619, L1

Mart´ınez-Delgado, D., Gabany, R. J., Crawford, K., et al. 2010, AJ, 140, 962

Mart´ınez-Delgado, D., Penarrubia,˜ J., Gabany, R. J., et al. 2008, ApJ, 689, 184

Mart´ınez-Delgado, D., Pohlen, M., Gabany, R. J., et al. 2009, ApJ, 692, 955

Mart´ın-Navarro, I., Bakos, J., Trujillo, I., et al. 2012, MNRAS, 427, 1102

Mart´ın-Navarro, I., Trujillo, I., Knapen, J. H., Bakos, J., & Fliri, J. 2014, MNRAS, 441, 2809

Massey, P., & Foltz, C. B. 2000, PASP, 112, 566

201 Massey, P., Strobel, K., Barnes, J. V., & Anderson, E. 1988, ApJ, 328, 315

Mateo, M. 1998, ARA&A, 36, 435

Matteucci, F., & Franc¸ois, P. 1989, MNRAS, 239, 885

Mazzei, P., Curir, A., & Bonoli, C. 1995, AJ, 110, 559

McClure, R. D. 1969, AJ, 74, 50

McConnachie, A. W. 2012, AJ, 144, 4

McConnachie, A. W., Irwin, M. J., Ibata, R. A., et al. 2009, Nature, 461, 66

McGaugh, S. S., & Bothun, G. D. 1994, AJ, 107, 530

McGaugh, S. S., & de Blok, W. J. G. 1997, ApJ, 481, 689

McGaugh, S. S., Schombert, J. M., & Bothun, G. D. 1995, AJ, 109, 2019

McGee, S. L., & Balogh, M. L. 2010, MNRAS, 403, L79

McGee, S. L., Balogh, M. L., Bower, R. G., Font, A. S., & McCarthy, I. G. 2009, MNRAS, 400, 937

Meidt, S. E., Rand, R. J., & Merriﬁeld, M. R. 2009, ApJ, 702, 277

Meisner, A. M., & Finkbeiner, D. P. 2014, ApJ, 781, 5

Mendes de Oliveira, C. L., Cypriano, E. S., & Sodre,´ L., Jr. 2006, AJ, 131, 158

Mestel, L. 1963, MNRAS, 126, 553

Metz, M., & Kroupa, P. 2007, MNRAS, 376, 387

Meurer, G. R. 2017, Formation and Evolution of Galaxy Outskirts, 321, 172

Meurer, G. R., Heckman, T. M., & Calzetti, D. 1999, ApJ, 521, 64

Meurer, G. R., Heckman, T. M., Leitherer, C., et al. 1995, AJ, 110, 2665

Meurer, G. R., Wong, O. I., Kim, J. H., et al. 2009, ApJ, 695, 765 202 Meusinger, H., Ismail, H. A., & Notni, P. 2007, AN, 328, 562

Michard, R. 2002, A&A, 384, 763

Michel-Dansac, L., Duc, P.- A., Bournaud, F., et al. 2010, ApJL, 717, L143

Mihos, J. C. 2001, ApJ, 550, 94

Mihos, J. C. 2004, in IAU Symp. 217, Recycling Intergalactic and Interstellar Matter, ed. P.-A. Duc, J. Braine, & E. Brinks (Cambridge: Cambridge Univ. Press), 390

Mihos, J. C., Harding, P., Feldmeier, J. J., & Morrison, H. 2005, ApJ, 631, L41

Mihos, J. C., Harding, P., Feldmeier, J. J., et al. 2017, ApJ, 834, 16

Mihos, J. C., Harding, P., Rudick, C. S., & Feldmeier, J. J. 2013, ApJL, 764, L20

Mihos, J. C., Harding, P., Spengler, C. E., Rudick, C. S., & Feldmeier, J. J. 2013, ApJ, 762, 82

Mihos, J. C., & Hernquist, L. 1994, ApJL, 437, L47

Mihos, J. C., & Hernquist, L. 1996, ApJ, 464, 641

Mihos, J. C., Keating, K. M., Holley-Bockelmann, K., Pisano, D. J., & Kassim, N. E. 2012, ApJ, 761, 186

Minchev, I., Chiappini, C., & Martig, M. 2013, A&A, 558, A9

Minchev, I., & Famaey, B. 2010, ApJ, 722, 112

Minchev, I., Famaey, B., Combes, F., et al. 2011, A&A, 527, A147

Minchev, I., Famaey, B., Quillen, A. C., et al. 2012, A&A, 548, 126

Miville-Deschenes,ˆ M. -A., & Lagache, G. 2005, ApJS, 157, 302

Moﬀat, A. F. J., Marchenko, S. V., Seggewiss, W., et al. 1998, A&A, 331, 949

Monachesi, A., Bell, E. F., Radburn-Smith, D. J., et al. 2016 MNRAS, 457, 1419

203 Monaco, P., Murante, G., Borgani, S., & Fontanot, F. 2006, ApJL, 652, L89

Moore, B., Ghigna, S., Governato, F., Lake, G., & Quinn, T. 1999, ApJL, 524, L19

Morelli, L., Corsini, E. M., Pizzella, A., et al. 2015, MNRAS, 452, 1128

Moreno, J., Torrey, P., Ellison, S. L., et al. 2015, MNRAS, 448, 1107

Morrison, H. L. 1993, AJ, 106, 578

Morrison, H. L., Flynn, C., & Freeman, K. C. 1990, AJ, 100, 1191

Morrison, H. L., Mateo, M., Olszewski, E. W., et al. 2000, AJ, 119, 2254

Morrissey, P., Conrow, T., Barlow, T. A., et al. 2007, ApJS, 173, 682

Mortazavi, S. A., Lotz, J. M., Barnes, J. E., & Snyder, G. F. 2016, MNRAS, 455, 3058

Mulder, P. S. & van Driel, W. 1993, A&A, 272, 63

Murante, G., Arnaboldi, M., Gerhard, O., et al. 2004, ApJ, 607, L83

Murante, G., Giovalli, M., Gerhard, O., et al. 2007, MNRAS, 377, 2

Mutchler, M., Beckwith, S. V. W., Bond, H., et al. 2005, BAAS, 206, 1307

Naab, T., Oser, L., Emsellem, E., et al. 2014, MNRAS, 444, 3357

Naab, T., & Ostriker, J. P. 2006, MNRAS, 366, 899

Napolitano, N. R., Pannella, M., Arnaboldi, M., et al. 2003, ApJ, 594, 172

Negroponte, J., & White, S. D. M. 1983, MNRAS, 205, 1009

Nishiura, S., Murayama, T., Shimada, M., et al. 2000, AJ, 120, 2355

Oemler, A. Jr. 1974, ApJ, 194, 1

Oey, M. S., & Kennicutt, R. C., Jr. 1997, MNRAS, 291, 827

Oey, M. S., King, N. L., & Parker, J. W. 2004, AJ, 127, 1632

204 Oey, M. S., Lamb, J. B., Kushner, C. T., Pellegrini, E. W., & Graus, A. S. 2013, ApJ, 768, 66

Oey, M. S., Meurer, G. R., Yelda, S., et al. 2007, ApJ, 661, 801

Oh, S. H., Kim, W.-T., Lee, H. M., & Kim, J. 2008, ApJ, 683, 94

Okamoto, S., Arimoto, N., Ferguson, A. M. N., et al. 2015, ApJL, 809, L1

Okamura, S. 1978, PASJ, 30, 91

Oliphant, T. E. 2007, CSE, 9, 10

Oosterloo, T. A., Morganti, R., Crocker, A., et al. 2010, MNRAS, 409, 500

Oosterloo, T. A., Morganti, R., Sadler, E. M., van der Hulst, T., & Serra, P. 2007, A&A, 465, 787

Pagel, B. E. J. 1997, Nucleosynthesis and Chemical Evolution of Galaxies, by Bernard E. J. Pagel, pp. 392. ISBN 0521550610. Cambridge, UK: Cambridge University Press, October 1997., 392

Parson, W. 1850, Observations of Nebulae, Philosophical Transactions of the Royal Soci- ety, 140, 499

Pastoriza, M. G., Winge, C., Ferrari, F., Macchetto, F. D., & Caon, N. 2000, ApJ, 529, 866

Patat, F. 2003, A&A, 400, 1183

Patterson, F. S. 1940, Harvard College Observatory Bulletin, 914, 9

Pawlowski, M. S., Kroupa, P., Angus, G., et al. 2012, MNRAS, 424, 80

Pawlowski, M. S., Pﬂamm-Altenburg, J., & Kroupa, P. 2012, MNRAS, 423, 1109

Penarrubia,˜ J., McConnachie, A., & Babul, A. 2006, ApJL, 650, L33

Percival, W. J. 2005, A&A, 443, 819

Percival, W. J., Miller, L., & Peacock, J. A. 2000, MNRAS, 318, 273

205 Peters, S. P. C., van der Kruit, P. C., Knapen, J. H., et al. 2017, arXiv:1705.03555

Pﬂamm-Altenburg, J., & Kroupa, P. 2008, Nature, 455, 641

Pierce, M. J., & Tully, R. B. 1985, AJ, 90, 450

Pietsch, W., Vogler, A., Kahabka, P., Jain, A., & Klein, U. 1994, A&A, 284, 386

Pilyugin, L. S., Grebel, E. K., & Kniazev, A. Y. 2014, AJ, 147, 131

Pinkney, J., Gebhardt, K., Bender, R., et al. 2003, ApJ, 596, 903

Planck Collaboration, Abergel, A., Ade, P. A. R., et al. 2014, A&A, 571, A11

Pohlen, M., Dettmar, R.-J., Lutticke,¨ R., & Aronica, G. 2002, A&A, 392, 807

Pohlen, M., & Trujillo, I. 2006, A&A, 454, 759

Pompei, E. & Natali, G. 1997, A&AS, 124, 129

Ponman, T. J., Allan, D. J., Jones, L. R., et al. 1994, Nature, 369, 462

Powell, L. C., Bournaud, F., Chapon, D., & Teyssier, R. 2013, MNRAS, 434, 1028

Press, W. H., & Schechter, P. 1974, ApJ, 187, 425

Prieto, M., Aguerri, J. A. L., Varela, A. M., & Munoz-Tu˜ n˜on,´ C. 2001, A&A, 367, 405

Privon, G. C., Barnes, J. E., Evans, A. S., et al. 2013, ApJ, 771, 120

Proctor, R. N., de Oliveira, C. M., Dupke, R., et al. 2011, MNRAS, 418, 2054

Puchwein, E., Springel, V., Sijacki, D., & Dolag, K. 2010, MNRAS, 406, 936

Purcell, C. W., Bullock, J. S., & Zentner, A. R. 2007, ApJ, 666, 20

Purcell, C. W., Bullock, J. S., & Zentner, A. R. 2008, MNRAS, 391, 550

Quach, D., Dyda, S., & Lovelace, R. V. E. 2015, MNRAS, 446, 622

Quinn, P. J. 1984, ApJ, 279, 596

206 Quinn, P. J., Hernquist, L., & Fullagar, D. P. 1993, ApJ, 403, 74

Quireza, C., Rood, R. T., Bania, T. M., Balser, D. S., & Maciel, W. J. 2006, ApJ, 653, 1226

Quirk, W. J. 1972, ApJL, 176, L9

Radburn-Smith, D. J., Roskar,ˇ R., Debattista, V. P., et al. 2011, ApJ, 753, 138

Radburn-Smith, D. J., de Jong, R. S., Seth, A. C., et al. 2011, ApJS, 195, 18

Rand, R. J. 1996, ApJ, 462, 712

Rand, R. J. 1997, ApJ, 474, 129

Regan, M. W., Thornley, M. D., Helfer, T. T., et al. 2001, ApJ, 561, 218

Rejkuba, M., Greggio, L., Harris, W. E., Harris, G. L. H., & Peng, E. W. 2005, ApJ, 631, 262

Relano,˜ M., Kennicutt, R. C., Jr., Eldridge, J. J., Lee, J. C., & Verley, S. 2012, MNRAS, 423, 2933

Reynolds, R. J. 1985, ApJL, 298, L27

Reynolds, R. J. 1990, The Galactic and Extragalactic Background Radiation, 139, 157

Reynolds, R. J., Roesler, F. L., & Scherb, F. 1977, ApJ, 211, 115

Ribeiro, A. L. B., de Carvalho, R. R., Capelato, H. V., & Zepf, S. E. 1998, ApJ, 497, 72

Rix, H.-W. R., Kennicutt, R. C., Braun, R., & Walterbos, R. A. M. 1995, ApJ, 438, 155

Roberts, M. S., & Haynes, M. P. 1994, ARA&A, 32, 115

Roediger, J. C., Courteau, S., Sanchez-Bl´ azquez,´ P., & McDonald, M. 2012, ApJ, 758, 41

Rood, H. J. & Williams, B. A. 1984, ApJL, 285, 5

Roskar,ˇ R., Debattista, V. P., Quinn, T. R., Stinson, G. S., & Wadsley, J. 2008, ApJL, 684, L79

207 Roskar,ˇ R., Debattista, V. P., Quinn, T. R., & Wadsley, J. 2012, MNRAS, 426, 2089

Roskar,ˇ R., Debattista, V. P., Stinson, G. S., et al. 2008, ApJ, 675, L65

Rots, A H., Bosma, A., van der Hulst, J. M., Athanassoula, E., & Crane, P. C. 1990, AJ, 100, 387

Rownd, B. K., Dickey, J. M., & Helou, G. 1994, AJ, 108, 1638

Rubin, V. C. 1994, AJ, 107, 173

Rudick, C. S. 2010, PhD thesis, Case Western Reserve University

Rudick, C. S., Mihos, J. C., Frey, L. H., & McBride, C. K. 2009, ApJ, 699, 1518

Rudick, C. S., Mihos, J. C., Harding, P., et al. 2010, ApJ, 720, 569

Rudick, C. S., Mihos, J. C., & McBride, C. 2006, ApJ, 648, 936

Rudick, C. S., Mihos, J. C., & McBride, C. K. 2011, ApJ, 732, 48

Safronov, V. S. 1960, Annales d’Astrophysique, 23, 979

Sage, L. J., & Welch, G. A. 2006, ApJ, 644, 850

Salo, H., & Laurikainen, E. 2000, MNRAS, 319, 377

Sanchez,´ S. F., Rosales-Ortega, F. F., Iglesias-Paramo,´ J., et al. 2014, A&A, 563, A49

Sanchez-Bl´ azquez,´ P., Courty, S., Gibson, B. K., & Brook, C. B. 2009, MNRAS, 398, 591

Sancisi, R. 1976, A&A, 53, 159

Sancisi, R., Fraternali, F., Oosterloo, T., & van der Hulst, T. 2008, A&ARv, 15, 189

Sandin, C. 2014, A&A, 567, 97

Sandin, C. 2015, A&A, 577, 106

Sanduleak, N. 1969, AJ, 74, 47

Sawa, T., & Fujimoto, M. 2005, PASJ, 57, 429 208 Schaerer, D., de Barros, S., & Sklias, P. 2013, A&A, 549, A4

Schanberg, B. C. 1973, ApJS, 26, 115

Scheick, S., & Kuhn, J. R. 1994, ApJ, 423, 566

Schiminovich, D., van Gorkum, J. H., Dijkstra, M., et al. 2000, in ASP Conf. Ser. 240, Gas and Galaxy Evolution, ed. J. E. Hibbard, M. Rupen, & J. H. van Gorkum (San Francisco, CA: ASP), 864

Schlaﬂy, E. F., & Finkbeiner, D. P. 2011, ApJ, 737, 103

Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500, 525

Schmidt, M. 1959, ApJ, 129, 243

Schneider, S. E. 1985, ApJL, 288, L33

Schneider, S. E. 1989, ApJ, 343, 94

Schneider, S. E., Helou, G., Salpeter, E. E., & Terzian, Y. 1983, ApJL, 273, L1

Schneider, S. E., Salpeter, E. E., & Terzian, Y. 1986, AJ, 91, 13

Schneider, S. E. Skrutskie, M. F., Hacking, P. B., et al. 1989, AJ, 97, 666

Schombert, J., Maciel, T., & McGaugh, S. 2011, Advances in Astronomy, 2011, 143698

Schombert, J., McGaugh, S., & Maciel, T. 2013, AJ, 146, 41

Schombert, J. M., & Wallin, J. F. 1987, AJ, 94, 300

Schonrich,¨ R., & Binney, J. 2009, MNRAS, 399, 1145

Schruba, A., Leroy, A. K., Kruijssen, J. M. D., et al. 2017, ApJ, 835, 278

Schruba, A., Leroy, A. K., Walter, F., et al. 2011, AJ, 142, 37

Schweizer, F. 1976, ApJS, 31, 313

Schweizer, F., & Seitzer, P. 1992, AJ, 104, 1039

209 Scowen, P. A., Dufour, R. J., & Hester, J. J. 1992, AJ, 104, 92

Searle, L. & Zinn, R. 1978, ApJ, 225, 357

Seibert, M., Martin, D. C., Heckman, T. M., et al. 2005, ApJL, 619, L55

Sellwood, J. A., & Binney, J. J. 2002, MNRAS, 336, 785

Sellwood, J. A., & Carlberg, R. G. 1984, ApJ, 282, 61

Serra, P., Oosterloo, T., Morganti, R., et al. 2012, MNRAS, 422, 1835

Sersic,´ J. L., & Pastoriza, M. 1967, PASP, 79, 152

Seth, A. C., Dalcanton, J. J., & de Jong, R. S. 2005, AJ, 129, 1331

Sheth, K., Regan, M., Hinz, J. L, et al. 2010, PASP, 122, 1397

Shibuya, T., Ouchi, M., Kubo, M., & Harikane, Y. 2016, ApJ, 821, 72

Sil’chenko, O. K., Moiseev, A. V., Afanasiev, V. L., Chavushyan, V. H., & Valdes, J. R. 2003, ApJ, 591, 185

Silk, J., & Werner, M. W. 1969, ApJ, 158, 185

Skillman, E. D. 1987, NASA Conference Publication, 2466

Slater, C. T., Harding, P., & Mihos, J. C. 2009, PASP, 121, 1267

Smith, B. J., Soria, R., Struck, C., et al. 2014, AJ, 147, 60

Smith, B. J., Struck, C., Hancock, M., et al. 2008, AJ, 135, 2406

Smith, J., Gehrz, R. D., Grasdalen, G. L., et al. 1990, ApJ, 362, 455

Sofue, Y., & Wakamatsu, K.-I. 1993, PASJ, 45, 529

Sohn, S. T., Anderson, J., & van der Marel, R. P. 2012, ApJ, 753, 7

Sommer-Larsen, J. 2006, MNRAS, 369, 958

Sommer-Larsen, J., Romeo, A. D., & Portinari, L. 2005, MNRAS, 357, 478 210 Spencer, M., Loebman, S., & Yoachim, P. 2014, AJ, 788, 146

Spitzer, L. Jr., & Baade, W. 1951, ApJ, 113, 413

Spitzer, L., Jr., & Jenkins, E. B. 1975, ARA&A, 13, 133

Spitzer, L., Jr., & Tomasko, M. G. 1968, ApJ, 152, 971

Springel, V., & Hernquist, L. 2005, ApJL, 622, L9

Springel, V., Wang, J., Vogelsberger, M., et al. 2008, MNRAS, 391, 1685

Springel, White, S. D. M., Jenkins, A., et al. 2005, Nature, 435, 629

Stark, A. A. 1984, ApJ, 281, 624

Statler, T. S., & Smecker-Hane, T. 1999, AJ, 117, 839

Steinicke, W. 2012, Journal of Astronomical Heritage, 15, 19

Stewart, K. R., Bullock, J. S., Wechsler, R. H., & Maller, A. H. 2009, ApJ, 702, 307

Stewart, K. R., Bullock, J. S., Wechsler, R. H., Maller, A. H., & Zentner, A. R. 2008, ApJ, 683, 597

Stierwalt, S., Haynes, M. P., Giovanelli, R., et al. 2009, AJ, 138, 338

Struck, C., & Elmegreen, B. G. 2017, MNRAS, 464, 1482

Struck, C., Kaufman, M., Brinks, E., et al. 2005, MNRAS, 364, 69

Struck, C., & Smith, B. J. 2012, MNRAS, 422, 2444

Sullivan, M., Treyer, M. A., Ellis, R. S., & Mobasher, B. 2004, MNRAS, 350, 21

Sullivan, M., Treyer, M. A., Ellis, R. S., et al. 2000, MNRAS, 312, 442

Tacconi, L. J., & Young, J. S. 1990, ApJ, 352, 595

Taranu, D. S., Dubinski, J., & Yee, H. K. C. 2013, ApJ, 778, 61

Theis, C., & Kohle, S. 2001, A&A, 370, 365 211 Thilker, D. A., Bianchi, L., Meurer, G., et al. 2007, ApJS, 173, 538

Thilker, D. A., Bianchi, L., Boissier, S., et al. 2005, ApJL, 619, L79

Thilker, D. A., Boissier, S., Bianchi, L., et al. 2007, ApJS, 173, 572

Thilker, D. A., Donovan, J., Schiminovich, D., et al. 2009, Nature, 7232, 990

Thilker, D. A., Walterbos, R. A. M., Braun, R., & Hoopes, C. G. 2002, AJ, 124, 3118

Thuan, T. X., & Kormendy, J. 1977, PASP, 89, 466

Tiede, G. P., Sarajedini, A., & Barker, M. K. 2004, AJ, 128, 224

Tomita, A., Aoki, K., Watanabe, M., Takata, T., & Ichikawa, S.-I. 2000, AJ, 120, 123

Toomre, A. 1964, ApJ, 139, 1217

Toomre, A. 1969, ApJ, 158, 899

Toomre, A. 1977, in The Evolution of Galaxies and Stellar Populations, ed. B, Tinsley & R. Larson (New Haven, CT: Yale Univ. Press), 401

Toomre, A. & Toomre, J. 1972, ApJ, 178, 623

Torres-Peimbert, S., Lazcano-Araujo, A., & Peimbert, M. 1974, ApJ, 191, 401

Toth, G., & Ostriker, J. P. 1992, ApJ, 389, 5

Tremonti, C. A., Heckman, T. A., Kauﬀmann, G., et al. 2004, ApJ, 613, 898

Trujillo, I., Martinez-Valpuesta, I., Mart´ınez-Delgado, D., et al. 2009, ApJ, 704, 618

Tully, R. B. 1974, ApJS, 27, 415

Tully, R. B. 1987, ApJ, 321, 280

Tully, B. R., Rizzi, L., Shaya, E. J., et al. 2009, AJ, 138, 323

Twarog, B. A. 1980, ApJ, 242, 242

Twarog, B. A., Ashman, K. M., Anthony-Twarog, B. J. 1997, AJ, 114, 2556 212 Usero, A., Leroy, A. K., Walter, F., et al. 2015, AJ, 150, 115

Uson, J. M., Boughn, S. P., & Kuhn, J. R. 1991, ApJ, 369, 46 van Albada, G. D. 1977, A&A, 61, 297 van den Bergh, S. 2008, MNRAS, 390, L51 van der Hulst, J. M. 1979, A&A, 75, 97 van der Hulst, J. M., & Huchtmeier, W. K. 1979, A&A, 78, 82 van der Hulst, J. M., Skillman, E. D., Kennicutt, R. C., & Bothun, G. D. 1987, A&A, 177, 63 van der Hulst, J. M., Skillman, E. D., Smith, T. R., et al. 1993, AJ, 106, 548 van der Hulst, J. M., van Albada, T. S., & Sancisi, R. 2001, ASP Conf. Ser. 240, Gas and Galaxy Evolution, ed. J.E. Hibbard, M. Rupen, & J.H. van Gorkum (San Francisco, CA: ASP), 451 van der Kruit, P. C. 1979, A&AS, 38, 15 van der Kruit, P. C. 1987, A&A, 173, 59 van der Kruit, P. C. 2007, A&A, 466, 883 van der Kruit, P. C., Oort, J. H., & Mathewson, D. S. 1972, A&A, 21, 169 van der Kruit, P. C., & Searle, L. 1981, A&A, 95, 105 van der Kruit, P. C., & Searle, L. 1981, A&A, 95, 116 van der Kruit, P. C., & Shostak, G. S. 1984, A&A, 134, 258

VanderPlas, J., Connolly, A. J., Ivezic, Z., & Gray, A. 2012, Proceedings of Conference on Intelligent Data Understanding (CIDU), pp. 47-54, 2012., 47

Vandervoort, P. O. 1970, ApJ, 161, 87 van Dokkum, P. G., Whitaker, K. E., Brammer, G., et al. 2010, ApJ, 709, 1018 213 van Eymeren, J., Jutte,¨ E., Jog, C. J., Stein, Y., & Dettmar, R.-J. 2011, A&A, 530, A29 van Eymeren, J., Jutte,¨ E., Jog, C. J., Stein, Y., & Dettmar, R.-J. 2011, A&A, 530, A30 van Gorkom, J. H., Knapp, G. R., Raimond, E., Faber, S. M., & Gallagher, J. S. 1986, AJ, 91, 791 van Gorkom, J. H., van der Hulst, J. M., Haschick, A. D., & Tubbs, A. D. 1990, AJ, 99, 1781 van Zee, L., Haynes, M. P., Salzer, J. J., & Broeils, A. H. 1997, AJ, 113, 1618

Verdes-Montenegro, L., Yun, M. S., Williams, B. A., et al. 2001, A&A, 377, 812

Villalobos, A.,´ & Helmi, A. 2008, MNRAS, 391, 1806

Vlajic,´ M., Bland-Hawthorn, J., & Freeman, K. C. 2009, ApJ, 697, 361

Vlaijc,´ M., Bland-Hawthorn, J., & Freeman, K. C. 2011, ApJ, 732, 7

Vogel, S. N., Weymann, R., Rauch, M., & Hamilton, T. 1995, ApJ, 441, 162

Vogelsberger, M., Genel, S., Springel, V., et al. 2014, MNRAS, 444, 1518 von Benda-Beckmann, A. M., D’Onghia, E., Gottlober,¨ S., et al. 2008, MNRAS, 386, 2345

Walker, I. R., Mihos, J. C., & Hernquist, L. 1996, ApJ, 460, 121

Waller, W. H., Bohlin, R. C., Cornett, R. H., et al. 1997, ApJ, 481, 169

Walter, F., Brinks, E., de Blok, W. J. G., et al. 2008, AJ, 136, 2563

Walterbos, R. A. M., & Braun, R. 1994, ApJ, 431, 156

Walterbos, R. A. M., Braun, R., & Kennicutt, R. C. 1994, AJ, 107, 184

Wang, Z. 1990, ApJ, 360, 543

Watkins, A. E., Mihos, J. C., & Harding, P. 2015, ApJL, 800, L3

Watkins, A. E., Mihos, J. C., & Harding, P. 2016, ApJ, 826, 59

214 Watkins, A. E., Mihos, J. C., Harding, P., & Feldmeier, J. J. 2014, ApJ, 791, 38

Watson, D. F., Berlind, A. A., & Zentner, A. R. 2012, ApJ, 754, 90

Weilbacher, P. M., Duc, P.-A., Fritze von Alvensleben, U., Martin, P., & Fricke, K. J. 2000, A&A, 358, 819

Weinberg, M. D. 1995, ApJL, 455, L31

Weisz, D. R., Dalcanton, J. J., Williams, B. F., et al. 2011, ApJ, 739, 5

Weisz, D. R., Johnson, B. D., Johnson, L. C., et al. 2012, ApJ, 744, 44

Welch, G. A., & Sastry, G. N. 1971, ApJL, 169, L3

Werk, J. K., Putman, M. E., Meurer, G. R., et al. 2010, AJ, 279

White, P. M., Bothun, G., Guerrero, M. A., West, M. J., & Barkhouse, W. A. 2003, ApJ, 585, 739

White, S. D. M. & Frenk, C. S. 1991, ApJ, 379, 52

White, S. D. M., & Rees, M. J. 1978, MNRAS, 183, 341

Whitmore, B. C., Chandar, R., Kim, H., et al. 2011, ApJ, 729, 78

Whitmore, B. C., & Schweizer, F. 1995, AJ, 109, 960

Williams, B. F., Ciardullo, R., Durrell, P. R., et al. 2007, ApJ, 656, 756

Williams, B. F., Dalcanton, J., Seth, A. C., et al. 2009, AJ, 137, 419

Williams, B. F., Delcanton, J. J., Stilp, A., et al. 2013, ApJ, 765, 120

Willman, B., Governato, F., Wadsley, J., & Quinn, T. 2004, MNRAS, 355, 159

Wolﬁnger, K., Kilborn, V. A., Koribalski, B. S., et al. 2013, MNRAS, 428, 1790

Wood, K., & Reynolds, R. J. 1999, ApJ, 525, 799

Worthey, G. 1994, ApJS, 95, 107

215 Worthey, G., Espana,˜ A., MacArthur, L. A., & Courteau, S. 2005, 631, 820

Wright, A. E. 1972, MNRAS, 157, 309

Wyder, T. K., Martin, D. C., Barlow, T. A., et al. 2009, ApJ, 696, 1834

Wyse, R. F. G., & Gilmore, G. 1992, AJ, 104, 144

Xue, X.-X., Rix, H.-W., Yanny, B., et al. 2011, ApJ, 738, 79

Yang, B., Zhu, J., & Song, Y.-Y. 2002, ChJAA, 2, 474

Yang, X., Mo, H. J., van den Bosch, F. C., et al. 2007, ApJ, 671, 153

Yıldız, M. K., Serra, P., Oosterloo, T. A., et al. 2015, MNRAS, 451, 103

Yim, K., & van der Hulst, J. M. 2016, MNRAS, 463, 2092

Yong, D., Carney, B. W., Teixera de Almeida, M. L. 2005, AJ, 130, 597

York, D. G., Adelman, J., Anderson, J. E., Jr., et al. 2000, AJ, 120, 1579

Youngblood, A. J., & Hunter, D. A. 1999, ApJ, 519, 55

Younger, J. D., Cox, T. J., Seth, A. C., & Hernquist, L. 2007, ApJ, 670, 269

Yuan, C., & Grosbol, P. 1981, ApJ, 243, 432

Yuan, H. B., Liu, X. W., & Xiang, M. S. 2013, MNRAS, 430, 2188

Yun, M. S., Ho, P. T. P., & Lo, K. Y. 1994, Nature, 372, 530

Zarattini, S., Aguerri, J. A. L., Sanchez-Janssen,´ R., et al. 2015, A&A, 581, A16

Zarattini, S., Barrena, R., Girardi, M., et al. 2014, A&A, 565, A116

Zaritsky, D., & Christlein, D. 2007, AJ, 134, 135

Zaritsky, D., & Rix, H.-W. 1997, ApJ, 477, 118

Zaritsky, D., Salo, H., Laurikainen, E., et al. 2013, ApJ, 772, 135

216 Zasov, A. V., & Simakov, S. G. 1988, Astrophysics, 29, 518

Zheng, Z., Thilker, D. A., Heckman, T. M., et al. 2015, ApJ, 800, 120

Zibetti, S., White, S. D. M., Schneider, D. P., & Brinkmann, J. 2005, MNRAS, 358, 949

Zurita, A., Beckman, J. E., Rozas, M., & Ryder, S. 2002, A&A, 386, 801

Zwicky, F. 1951, PASP, 63, 61

217