University of Groningen

The influence of the environment on the evolution of Sikkema, Geert

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below. Document Version Publisher's PDF, also known as Version of record

Publication date: 2009

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA): Sikkema, G. (2009). The influence of the environment on the evolution of galaxies. s.n.

Copyright Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

The publication may also be distributed here under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license. More information can be found on the University of Groningen website: https://www.rug.nl/library/open-access/self-archiving-pure/taverne- amendment.

Take-down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.

Download date: 07-10-2021 The Influence of the Environment on the Evolution of Galaxies

Proefschrift

ter verkrijging van het doctoraat in de Wiskunde en Natuurwetenschappen aan de Rijksuniversiteit Groningen op gezag van de Rector Magnificus, dr. F. Zwarts, in het openbaar te verdedigen op vrijdag 13 maart 2009 om 13.15 uur

door

Geert Sikkema

geboren op 19 maart 1973 Delfzijl Promotores: Prof. dr. R. F. Peletier Prof. dr. E. A. Valentijn

Beoordelingscommissie: Prof. dr. D. Carter Prof. dr. S. Trager Dr. M. Balcells

ISBN 978-90-367-3749-4 ISBN 978-90-367-3748-7 (electronic version) If you think you understand quan- tum physics you don’t understand quantum physics.

–Richard Feynman Cover page: Drawing by Gert Sikkema

Contact information: Gert Sikkema [email protected] Contents

1 Introduction 9 1.1 Background ...... 10 1.1.1 Morphology and evolution of galaxies ...... 10 1.1.2 Bars in disk dominated galaxies ...... 11 1.1.3 Shell galaxies ...... 13 1.1.4 systems ...... 14 1.2 This Thesis ...... 15 1.2.1 Part I: properties as a function of environment ...... 15 1.2.2 Part II: HST ACS observations of six shell galaxies ...... 16

2 Environmental Influences on the Evolution of Galaxies 17 2.1 Introduction ...... 18 2.2 Observations ...... 19 2.2.1 Targets ...... 19 2.3 Data Reduction ...... 21 2.3.1 The Calibration Pipeline ...... 24 2.3.2 Photometric Calibrations ...... 26 2.3.3 Image Pipeline ...... 28 2.4 Data Analysis ...... 29 2.4.1 Source catalogues ...... 29 2.4.2 Source Selections ...... 30 2.4.3 Galaxy Density Calculations...... 32 2.4.4 Structural/Morphological parameters ...... 36 2.4.5 Sample Definitions ...... 40 2.4.6 Classification ...... 43 2.5 Results ...... 46 2.5.1 Classification vs. parameters ...... 46 2.5.2 Morphology Density Relation ...... 46 2.5.3 Parameters as a function of environment ...... 47 2.5.4 Colour Selected subsamples: red spirals and blue ellipticals . . . 49 2.6 Discussion ...... 51 2.6.1 Suppression of formation ...... 51 2.6.2 Origin of S0s ...... 54 vi CONTENTS

2.6.3 The red spirals ...... 55 2.7 Summary ...... 56

3 Bars in large scale structures at z=0.11 ± 0.02 59 3.1 Introduction ...... 59 3.2 Data Analysis ...... 61 3.2.1 Detection of bars ...... 61 3.3 Results ...... 64 3.3.1 Bars ...... 64 3.4 Discussion ...... 65 3.4.1 Bar frequencies in previous studies ...... 66 3.4.2 Environmental related bar frequency in spirals vs. bar formation theories ...... 67 3.4.3 Bars in S0s ...... 68 3.5 Summary ...... 68

4 HST/ACS observations of shell galaxies. 71 4.1 Introduction ...... 72 4.2 Observations and Data Reduction ...... 73 4.3 Data Analysis ...... 75 4.3.1 Global parameters ...... 75 4.3.2 Shell radii ...... 76 4.3.3 Shell fluxes ...... 76 4.3.4 Shell colours ...... 79 4.3.5 Galaxy colours ...... 79 4.3.6 Shell radial profiles ...... 80 4.4 Notes on the individual systems ...... 81 4.4.1 NGC 474 ...... 81 4.4.2 NGC 1344 ...... 84 4.4.3 NGC 2865 ...... 84 4.4.4 NGC 3923 ...... 86 4.4.5 NGC 5982 ...... 86 4.4.6 NGC 7626 ...... 87 4.5 General results ...... 89 4.5.1 Shell radial distributions ...... 89 4.5.2 Shell brightness profiles ...... 90 4.5.3 Shell colours ...... 91 4.5.4 Dust in the centres of shell galaxies ...... 92 4.5.5 Dust origin ...... 93 4.6 Summary ...... 95 Appendix 4.A Results for NGC 474 ...... 96 Appendix 4.B Results for NGC 1344 ...... 100 Appendix 4.C Results for NGC 2865...... 104 Appendix 4.D Results for NGC 3923 ...... 107 Appendix 4.E Results for NGC 5982 ...... 111 Appendix 4.F Results for NGC 7626 ...... 114 CONTENTS vii

5 Globular Clusters of Shell Galaxies 119 5.1 Introduction ...... 119 5.2 Observations and Data Reduction ...... 122 5.3 Data Analysis ...... 124 5.3.1 GALPHOT ...... 124 5.3.2 Globular cluster candidates ...... 125 5.3.3 Photometry ...... 126 5.3.4 Completeness ...... 126 5.3.5 Photometric errors and aperture selection ...... 126 5.3.6 Selection of GCCs ...... 127 5.4 V-I distributions and spatial distributions ...... 130 5.4.1 V-I distributions ...... 130 5.4.2 Components of the colour distributions...... 131 5.4.3 Spatial distributions of the globular clusters ...... 132 5.4.4 Globular cluster surface densities ...... 136 5.5 Globular cluster luminosity function in I ...... 137 5.5.1 Determination ...... 140 5.5.2 GCLF as a distance estimator for NGC 1344 and NGC 3923 . . 140 5.5.3 Total numbers of GCs ...... 140 5.6 Total number of globular clusters and specific frequencies ...... 142 5.7 Discussion ...... 143 5.7.1 The shell galaxies ...... 143 5.7.2 Comparison of the GC systems with normal ellipticals ...... 144 5.7.3 Possible evidence for recent GC formation in NGC 7626 and NGC 2865 ...... 147 5.7.4 Ages and minor mergers...... 148 5.8 Conclusions ...... 149

Bibliography 151

Nederlandse samenvatting 163

Acknowledgements 175

Chapter 1 Introduction

New discoveries in astronomy depend very much on improvements in technology, begin- ning of course with the invention and astronomical usage of the telescope, now about 400 years ago. Since then, the telescopes have grown dramatically in size, allowing to detect fainter sources. This culminated in the construction of the 5m Hale telescope, which was completed in 1948. For 45 years, this was the best optical telescope on . Building larger mirrors, with the correct shape was impossible, since their huge weight caused unacceptable deformations. In the 1980s, engineers had developed new building techniques and materials, which allowed the construction of a much larger class of telescopes. They introduced light-weight mirrors that keep their shape using active optics: this technology continuously adjusts the shape of the mirror by using actuators behind it. This resulted in a class of ≈ 10m telescopes with segmented mirrors like Keck, GTC and SALT. Slightly smaller, but consisting of single mirrors, are the 8m class telescopes like VLT, Subaru and Gemini. This development has not stopped yet and the nearby future promises the realisation of 25m class telescopes or even larger: the Extremely Large Telescope (ELT, 42m), Thirty Meter Telescope (TMT, 30m) and the Giant Magellan Telescope (GMT, 25m). These will all have segmented mirrors. The space era has brought many telescopes beyond the Earth’s atmosphere, notably the (HST), providing an undisturbed view of the Universe ∗. Meanwhile, also the detection methods have been improved, making it possible to detect fainter sources with the same telescope as well as reaching a much larger wavelength range. First, with the invention of photographic plates, culminating in the Palomar Observatory Sky Survey (POSS), which maps the whole sky. Photographic plates were followed up by the invention and usage of charged coupled devices (CCDs). Although, these are much more sensitive to and better in quantifying light from astronomical sources than photographic plates, they only cover a small portion of the sky. Since their invention, a few decades ago, CCDs have improved dramatically: they have be- come more sensitive, larger and cheaper to build. The latter development means that, we can now align many more CCDs in the focus of a dedicated telescope; this technol- ogy is called wide field imaging. It provides a larger field of view of the sky and enables astronomers for the first time to map the whole sky with CCDs. Examples of such

∗ In this introduction, we focus on optical telescopes only. 10 chapter 1: Introduction

CCD surveys are the (SDSS, http://www.sdss.org/), Mega- Cam (http://www.cfht.hawaii.edu/Instruments/Imaging/MegaPrime/ and VISTA (http://www.vista.ac.uk/). Two, even larger projects are WASP and PanStarrs. WASP (urlhttp://www.superwasp.org/index.html observes the sky a few times a night in search for extra-solar planets. PanStarrs (http://pan-starrs.ifa.hawaii.edu/ public/ will observe more than half of the sky several times each month using cameras which have a total of 1.4 billion pixels. This year, another new wide field camera consisting of 32 CCDs, called OmegaCAM, is expected to become operational on the ESO VLT Survey Telescope (VST). To deal with the huge amount of data, automated data reduction and analysis systems have to be developed. Such a system has been developed at the Kapteyn Institute in Groningen, called astro-wise (Valentijn et al. 2007). astro-wise is not only able to reduce OmegaCAM data but also has the capabilities to reduce data from many other wide field instruments. In this thesis, we will also make use of observation from modern astronomical in- struments. The thesis be divided in two parts according to data-set. The first part, consisting of Chapters 2 and 3, uses optical wide field imaging data from the ground. These data were used as a testbed for astro-wise, which will process data from the upcoming much larger wide field camera OmegaCAM. In this part, we analyse various observational properties of the three major galaxy types: ellipticals, spirals and lentic- ulars (S0s) as a function of environment,characterised by density. We try to contribute to important ’how’ questions like: how do S0 galaxies form, how does depend on environment and how do bars in disk dominated galaxies form. The second part, consisting of Chapter 4 and 5, uses high resolution optical data of the Advanced Camera for Surveys (ACS) on the space based HST. Here we analyse and try to un- derstand the optical properties of six nearby elliptical shell galaxies (Malin & Carter 1980).

1.1 Background

This thesis, called: The influence of the environment on the evolution of galaxies, treats a broad range of topics: from large scale structure of the Universe to globular cluster populations in galaxies. We present background information about these each of these topics separately in this Section.

1.1.1 Morphology and evolution of galaxies One of the most important discoveries in astronomy is the expansion of the Universe (Hubble 1929). This discovery is directly related to the question of the formation of galaxies. We predominantly find three major classes of galaxies in the nearby Universe, these are elliptical, spiral and lenticular (S0) galaxies. Apart from these classes, there is also the class of dwarf galaxies, objects that are much fainter than the objects in the other classes. We will, however, not consider them here. This classification scheme (Hubble 1926) is still often used today (Figure 1.1). The featureless ellipticals can be seen on the left side. Ellipticals have almost no gas and look like featureless, oblate or prolate-like structures. The scarcity of gas implies that no significant star formation is occurring in these galaxies any more. Many are on radial orbits, but the majority 1.1: Background 11 of ellipticals also posses some degree of rotation (Emsellem et al. 2004). The right side of Fig 1.1 shows the gas-rich spiral class, which is subdivided in barred spirals (lower branch) and non-barred spirals (upper branch). Spirals are characterised by having a thin stellar disk rotating around a central bulge. The central bulge region often contains a bar, which is probably very important in the evolution of a (Combes 2007). Star formation happens mainly in the the spiral arms in the disk and is due to the fact that significant amounts of gas are still available in these galaxies. In Figure 1.1, the S0s are located at the midpoint connecting the three branches. S0s are galaxies with properties belonging to both spirals and ellipticals. They do have a rotating disk, but do not show significant star formation due to a lack of gas. No spiral arms are visible. Not shown in the figure , but known from observations, e.g. Erwin (2005), is that S0s can also contain bars. How does the expansion of the Universe relate to the different galaxy types? The majority of ellipticals are thought of being very old (z >2) galaxy systems, (see Ellis et al. 1997; van Dokkum et al. 2004; Renzini 2006 and references therein). These were formed when small clumps in overdensities collided with each other (Springel et al. 2005). Spiral galaxies likely formed somewhat later in more empty regions of space, where external gravitational disturbances will not destroy the disk. Contrary to ellipticals, S0 galaxies show significant evolution in the past few Gyrs: S0s are now much more prevalent now than at z=0.3 (Dressler et al. 1997; Couch et al. 1998). This evolution is well observed in clusters (Dressler et al. 1997; Treu et al. 2003; Moran et al. 2007). Since S0s have many properties in common with spirals (bars, rotating disks, bulges), it is thought that these evolve from spirals. So, how and where do spirals transform into S0s? First, I discuss the question what physical processes are responsible for this evolution. S0s do not show significant star formation and lack spiral structure. Therefore, if S0s are formed from spirals, S0s must have lost their gas somehow. Starvation (Larson et al. 1980; Bekki et al. 2002) and ram pressure stripping (Gunn & Gott 1972) are able to remove the gas from a spiral galaxy. These processes can not be solely responsible for transformation of spirals to S0s: S0s show similar or larger K-band luminosities than spirals (Burstein et al. 2005). Moreover, S0s are equally prevalent in group environments, where gas removal processes are less important (Wilman et al. 2008). Therefore other processes, probably related to interactions may be more important. One candidate is slow interactions in galaxy groups (pre-processing Zabludoff & Mulchaey 1998, 2000). This latter notion may then answer our second question: ’where’ does the transformation mainly happen. It is still possible however, that the S0 transformation is much faster in cluster regions. Here, fast flybys from many small galaxies may transform a large gas-stripped spiral galaxy to be smoothed into a S0 (harassment Moore et al. 1999). If so, intermediate forms may be found between spiral and S0 in clusters. Several candidates have been proposed: red passive spirals in clusters (Couch et al. 1998; Poggianti et al. 1999; Wolf et al. 2008), blue bright compact galaxies (Braglia et al. 2007) and E+A galaxies (Poggianti et al. 2004). Whether one of these candidates is an intermediate form, remains to be seen.

1.1.2 Bars in disk dominated galaxies Bars, pictured in lower branch of Figure 1.1, are found in the majority of spiral galaxies (Eskridge et al. 2000). Bars play a major role in the evolution of disk galaxies. Sim- ulations show that bars can form spontaneously in dynamically cold pure stellar disks 12 chapter 1: Introduction

Figure 1.1: Hubble Tuning Fork, with in the left side the elliptical galaxies; right are the spirals split into two branches according to bar-presence (lower branch). At the point connecting the spiral branches and ellipticals, we see an S0 galaxy.

(Hohl 1971; Kalnajs 1972; Sellwood 1981), but also after mergers and encounters (Byrd et al. 1986; Noguchi 1987; Gerin et al. 1990). Gas dynamics drives the evolution of bars (Combes 2007). Their simulations imply several cycles of bar formation and destruction in the lifetime of a galaxy, where each subsequent bar will increase in size. Here, bars dramatically redistribute matter, changing the appearance of disk-dominated galaxies in time. This bar-cycle scenario is able to explain the existence pseudo-bulges, which look like normal bulges but have more in common with disks (Kormendy & Kennicutt 2004). Other observational evidence for this scenario comes from: 1) the increasing average size of bars going from late to early type galaxies (Erwin 2005), with the early type disk galaxies probably more massive and older than the late type galaxies. 2) gas inflows along bars (Richter & Sancisi 1994; Zaritsky & Rix 1997) and 3) evolution in the last few Gyrs of bar sizes and frequencies (Sheth et al. 2008). These latter findings are still a matter of debate, however, Jogee et al. (2004) and Barazza et al. (2008) do not find an evolutionary trend in the bar fraction. Comparing properties of bars in clusters and field may give more important clues about stability and formation of bars. Disk galaxies near centres of clusters are devoid of gas due to ram pressure stripping (Gunn & Gott 1972) and/or starvation (Larson et al. 1.1: Background 13

1980; Bekki et al. 2002), possibly lowering the probability of bar destruction. On the other hand: the density of galaxies is much higher in cluster regions, which will heat up disks (harassment Moore et al. (1999)), with the probability that the properties of the bars are changed. Only a few studies are published to date, which analyse bar properties as a function of cluster environment. These studies show contradictory results; there is no consensus about bar occurrence in different environments. Thompson (1981) and Andersen (1996), find an increasing bar fraction towards the Coma and Virgo cluster central regions. These studies are at odds with a more recent study van den Bergh (2002), who did not find any correlation. Finally, the presence of bars in S0s supports the statement that S0 evolved from spirals. Bar frequencies in S0s may thus give clues about their past and bar evolution.

1.1.3 Shell galaxies

In lower galaxy density areas in the Universe we find large numbers of shell galaxies (Colbert et al. 2001), which are elliptical galaxies containing faint, sharp-edged features (shells, see Figure 1.2). Shell galaxies were discovered on processed photographic plates to accentuate faint optical features (Malin & Carter 1980). Numerical simulations provide the most widely accepted framework for interpreting the shell morphologies in terms of mergers (Quinn 1984; Hernquist & Quinn 1987, 1989; Dupraz & Combes 1986). Here, the shells are the debris of a small intruder galaxy. Models for shell formation not based on mergers have been proposed as well, most important of these are the interaction models (Thomson & Wright 1990; Thomson 1991), which state that shells are formed in the host galaxy after a fly-by of another galaxy. In order to form shells, it is necessary for the host galaxy to have some rotation. Using such a model, numerical simulations can also reproduce some of the observed shell features. Shells are mostly observed in galaxies located in isolated environments; this may indicate either a lower formation rate or a shorter lifetime in denser environments; it may also indicate younger ages for shell systems in empty environments, since the shells will then be brighter (Colbert et al. 2001). Forbes & Thomson (1992) noted that many galaxies which contain a kinematically decoupled core (KDC), also show shells, however, a relation between these galaxy properties has not yet been proven. More recent work (McDermid et al. 2006) shows two types of KDCs, large sized KDC which old (>8 Gyrs) in ellipticals with little rotation. Small KDCs, with ages between 0.5 and 15 Gyr can be found in ellipticals with more rotation. It is unclear yet how shells relate to these two types of KDCs. Several observational diagnostics may be used to test the various theories, these include: determination of the shell colours, dust presence and shell morphologies. The merger model predicts colour differences between shells and host galaxy in a significant number of shell galaxies, whereas the interaction models do not have this. Simulations of shell formation predict differences in the shell morphologies for the different models (Quinn 1984; Thomson & Wright 1990; Thomson 1991). We will compare simulated shell morphologies from the literature with our detailed data. We also compare the prevalence of dust in shell galaxies with normal ellipticals. Large amounts of dust may indicate recent star formation, which would happen if a small, gas rich galaxy had been ’eaten’ by the larger elliptical. This would support the merger model 14 chapter 1: Introduction

Figure 1.2: Shells in a galaxy subtracted image.

1.1.4 Globular cluster systems Globular clusters (GCs) can be found around each galaxy, although there are many more globular cluster per luminosity unit for ellipticals than spirals. The majority of GCs is old (Stetson et al. 1996). Their origin is still unclear, however. An important diagnostic, in this respect, is the existence of bimodality in the colour distribution of GCs, present in many early type galaxies Zepf & Ashman (1993); Whitmore et al. (1995). One of these populations, with colour V-I≈0.90, is present in all galaxies and seems to be a universal population. Generally, the colour of the other population is mostly red and explained as being due to the stars within the GC having a much higher . Sometimes, mainly in young merger remnants, the colour of the second population is also blue (Schweizer 1987). The latter population is best explained, using merger scenarios (Toomre 1977; Schweizer 1987; Ashman & Zepf 1992). These propose that the second population are metal-rich GCs, created in gas rich mergers. Since most star formation occurred at early epochs, this means in general that the metal-rich GCs are also old. However, this scenario also suggests that GCs can still be forming today in mergers. This is supported by observations of young cluster-like objects in current mergers in action like NGC 4038/39 (Whitmore & Schweizer 1995; Whitmore et al. 1999). Others models explaining bimodality are the accretion model (Cote et al. 1998) and the multiphase formation model (Forbes et al. 1997). In both models, all GCs are old. The first model produces bimodality by accreting and mixing metal poor GCs from dwarf galaxies with the more metal rich GCs of the massive host galaxy. Cannibalism by our own galaxy of the Sagittarius and Canis Major dwarf galaxies and their clusters (Ibata et al. 1995; Forbes et al. 2004) and observations of large numbers of dwarf galaxies around giant galaxies are cited as supporting this scenario. The 1.2: This Thesis 15 second model explains the bimodality as the result of two phases of GC formation in the initial collapse and formation of a galaxy. The metal poor clusters, and a small proportion of the stars form in the initial gravitational collapse, then the metal rich clusters and the bulk of the stellar component form from enriched gas in a second collapse phase about one or two Gyr later. A combination of these different scenarios is used in the hierarchical merging model of Beasley et al. (2002), who undertook semi- analytical simulations of GC formation. In this model the metal-poor GCs are old and formed in cold gas clumps, the metal-rich ones are created later in merger events. In the hierarchical build up of galaxies, accretion of GCs will also take place. These simulations are able to reproduce the many variations in the colour distributions of GC systems observed in elliptical galaxies. A technique to investigate possible age differences between different GC populations involve measuring radial density profiles. Radial density profiles generally show a flattening of the globular cluster density profile near the centre, when compared with the profile of the background galaxy light (Lauer & Kormendy 1986; Capuzzo-Dolcetta & Donnarumma 2001). Mechanisms which could cause a depletion of the cluster population near the centre are dynamical friction, which causes clusters to spiral in towards the centre (Tremaine et al. 1975; Pesce et al. 1992), and destruction by tidal shocks as the clusters pass close to the nucleus (Ostriker et al. 1989; Capuzzo-Dolcetta & Tesseri 1997). This second process operates preferentially in triaxial potentials, and in clusters on radial orbits. If we can identify a population of younger clusters, created during a recent merger, then the density profile might extend further into the centre as the clusters have had less time to be disrupted. However, this does depend upon the orbital structures to be the same, if one or other population were on predominantly radial orbits, then this would cause a stronger flattening of the core.

1.2 This Thesis

In this thesis, we cover and analyse many observational properties of galaxies and their relation to environment. Since we use two datasets, with very different characteristics, the thesis can be split in two parts, each part related to one particular dataset. In Part 1 (Chapters 2 and 3), we use ground based wide field data and analyse various properties of the three major galaxy types as a function of environment. In Part 2 (Chapters 3 and 4), we use high resolution HST, ACS data, we analyse the optical properties of six shell galaxies (Malin & Carter 1980) and try to contribute to the questions: how were the shells formed, do shell galaxies deviate from elliptical galaxies in other ways (Chapter 4) and are the properties of globular cluster systems of shell galaxies similar to normal elliptical galaxies (Chapter 5).

1.2.1 Part I: Galaxy properties as a function of environment In Chapter 2, we analyse many galaxy properties in a region on the sky of four square degrees in two passbands (V and I). The region contains four Abell clusters at z=0.11 ±0.02 and the lower density regions in between. With this data set, we try to contribute to the important ’how’ questions like: how do S0 galaxies form and how does star formation depend on environment. We characterise environment by attributing a galaxy density to each galaxy, using the 2dfGRS dataset (Colless et al. 2001), which provides a large number of accurate spectroscopic for bright galaxies in this region. Our 16 chapter 1: Introduction observations are a pilot project, to test a recently developed reduction pipeline for wide field imaging astro-wise (Valentijn et al. 2007). astro-wise was build to handle the huge amount of data which will come from the upcoming OmegaCAM mounted on the VLT Survey Telescope. We classify galaxies into three types: elliptical, S0s and spirals. We discriminate between disk dominated galaxies (S0s and spirals) and ellipticals using the Sèrsic parameter. Subsequent separation of S0s and spirals is done by eye. We find a population of red, possibly passive spirals in high density regions, which have intermediate properties between spirals and S0s. In Chapter 3, we continue using our wide field dataset. Here, we use the well defined disk dominated galaxy sample from Chapter 2 and analyse the occurrence, size and strength of large, strong bars (semi-major axis abar > 3.6 kpc and ellipticity larger than 0.40) and their correlations with galaxy density and other galactic properties. The bar detections are done using well known ellipse fitting techniques. The results are used to address questions like: how do bars in disk dominated galaxies form and do bar properties relate to galaxy density. Using theoretical bar formation and destruction scenarios, we try to answer these questions. We find that large, strong bars in spiral galaxies are more common in higher density areas. Strong large bars in S0s are far less common compared to bars in spiral galaxies We find slightly larger bars in higher density regions. Future studies, using, for instance, OmegaCAM observations of somewhat more nearby large scale structures, will give more insight in bar formation scenarios.

1.2.2 Part II: HST ACS observations of six shell galaxies Part II consists of Chapters 4 and 5. Using HST ACS data, six nearby shell galaxies were observed in V and I. In Chapter 4, we determine shell V-I colours and local underlying galaxy V-I colours, using new a technique. Comparison of these colours yields important information about the origin of the shells. We apply the technique of Voronoi binning to our data to get smoothed colour maps of the galaxies. The shell morphologies are compared with simulations available from the literature. Finally, we determine dust presence and estimate dust masses in our shell galaxies. We find that dust is present in all our six shell galaxies. The probability for this to happen for normal ellipticals is below 5%. Putting all results together, we conclude that the merger model better describes our observations than the interaction model. In Chapter 5, we again use the HST ACS dataset to analyse the Globular Cluster (GC) systems of the six shell elliptical galaxies and try to find possible effects of small mergers upon the GC formation history. For the globular clusters, we determine the V-I colour distributions, luminosity functions and radial surface density profiles. In four shell galaxies, the properties of the clusters are similar to those observed in other, non-shell, elliptical galaxies. Two shell galaxies provide possible evidence for recent GC formation; deep follow-up spectroscopy is necessary to confirm these results. We also find that the radial surface density profiles of the GCs are more flattened than the galaxy light in the cores. This effect is stronger for the universal population. Chapter 2 Environmental Influences on the Evolution of Galaxies

ABSTRACT ∗ In this Chapter, we present the data and results involving a pilot project to test a recently developed data reduction system called astro-wise, which main task is to reduce optical wide field imaging data coming from the OmegaCam instrument on the newly built VST telescope at Paranal. We study a sample of galaxies in a low (z=0.11 ±0.02) filament of the cosmic web using ground based wide field imaging data, to analyse galaxy properties as a function of environment. Our observations, taken with the 2.2m telescope at La Silla in filters V, R and I, cover a 4 square degree region near the South Galactic Pole containing 6 Abell clusters. The observations were taken in Omegacam observing mode and were reduced with the astro-wise data reduction system. We create several magnitude limited galaxy samples. Sample 1 is used to create a 2D density map, showing our clusters and filamentary structures. For Sample 2, a magnitude limited sample of 895 galaxies with MI < -20.3, we calculate 12 morphological quantities. Sample 3, with MI < -20.8, consists of 570 galaxies with redshift data, allows us to calculate accurate 3D densities and is used for the environmental analysis. Galaxies in Sample 3 are given a morphological classification using a combination of visibility of spiral structure and Sérsic index. The Sérsic parameter is highly effective (90%) to recover the visually classified spiral galaxies. We define 4 density regimes between the extreme empty field and dense inner cluster core regions to characterize the environment. We calculate averaged parameters for each galaxy class in each density regime. Furthermore, we define several subclasses of galaxies: red spirals, blue S0s and blue ellipticals, depending whether they lie outside their ’normal’ place with respect to the red sequence. Results: The fraction of red spirals increases continuously with increasing density. At the same time, the average asymmetry and star formation rate of spirals decrease. In the two highest density areas considered, i.e. within < 1.5Rvir from cluster centres, we see that on the average, spiral galaxies are significantly redder, which occurs for all Hubble types. S0s and elliptical galaxies have lower star formation than their relatives

∗ Authors: G.Sikkema, E.A.Valentijn & R.F.Peletier (in prep.) 18 chapter 2: Environmental Influences on the Evolution of Galaxies

in the field. We attribute these last two findings by gas depletion processes like starvation and ram-pressure stripping which are apparently felt in the inner regions of clusters. This is supported by the location of blue S0s and blue ellipticals, which are mainly found outside clusters and may provide evidence for continued external gas inflows. The physical properties, presented in this paper, of blue spirals, red spirals and S0s and seem to form a continuum, while masses of these three classes are similar. This is consistent with red spirals being an intermediate form between normal spirals and S0s.

2.1 Introduction

Analysis of galactic structural and morphological parameters in different environments continues to be an important subject in astronomy (see below). This is due to ongoing improvements in sensitivity and sky coverage of various optical and surveys like the SDSS (Loveday 2002) and UKIDSS ∗ and the availability of large redshift surveys: SDSS (Loveday 2002) and 2dFGRS (Colless et al. 2001). Furthermore, thanks to ever increasing detailed simulations, such as the Millennium simulations (Springel et al. 2005), better comparisons can be made of the properties of galaxies as a function of the large scale environment during the cause of time. Observational studies (Bower et al. 1992) show that most early type galaxies in clusters were already in place at z >2, probably due to collisions of clumps of matter which do not resemble our galaxies today. Formation of spiral galaxies likely comes much later (however, see Genzel et al. (2006).) While the average star formation rate has currently decreased by several orders of magnitude, galaxy evolution is still discernible in clusters between z=0.3 and z=0.1. This evolution mainly concerns the spiral galaxies, which somehow transform to S0s during the past few Gyrs: Dressler et al. (1997) and Couch et al. (1998) show that S0s are much more prevalent now in the centres of clusters than at z=0.3. This evolution is mostly affects lower mass systems (van der Wel et al. 2007; Holden et al. 2007) Star forming galaxies also seem to be more prevalent in the past than now (Butcher & Oemler 1984; Pei & Fall 1995; Madau et al. 1996; Margoniner et al. 2001; Braglia et al. 2007). The important question to solve is, what physical process are responsible for these recent evolutionary effects. S0s do not show significant star formation and lack spiral structure. Therefore, these galaxies must have lost their gas somehow. Red passive spirals in clusters (van den Bergh 1976; Couch et al. 1998; Poggianti et al. 1999; Wolf et al. 2008) may be an intermediate stage between star forming spirals and S0s. Several processes have been proposed to explain the S0 phenomenon. The effectiveness of these all depend on galaxy density. Starvation (Larson et al. 1980; Bekki et al. 2002) and ram-pressure stripping (Gunn & Gott 1972) are able to remove the gas from a spiral galaxy. However these processes can not be solely responsible for transformation of spirals to S0s: S0s show similar or larger luminosities than spirals (Burstein et al. 2005). Therefore, other processes, probably related to interactions must also be important. Candidates are merging and harassment (Moore et al. 1996, 1999). These processes occur mainly outside the cluster regions, in infalling groups.

∗ http://www.ukidss.org/surveys/surveys.html 2.2: Observations 19

In this chapter, we try to find clues to these questions, by analysing galaxies in, near and between several Abell clusters at z = 0.11, using wide field photometry. The advantage of such observations at these redshifts, compared to nearby clusters like Coma or Virgo, is that the 2D large scale structures are directly visible using only a few pointings of wide field images. Various structural and morphological parameters are analysed as a function of density in a continuous 4 square degree region, which is also covered by the 2dFGRS, providing distances and rough stellar populations (Colless et al. 2001). The data ∗ were also used to test a reduction pipeline, recently developed to process data from the upcoming 1 square degree camera on the VST. The distance of z=0.11 is near the limit about what is technically possible from the ground studying optical properties of galaxies, given the PSF and size of the telescope. Using these data, we can explore what is possible with the coming large surveys like KIDS†, which aims at structures somewhat more nearby, e.g. in the cluster at z=0.05 The data presented in this paper were reduced with a recently developed reduction system called astro-wise (Valentijn & Kuijken 2004; Valentijn et al. 2007), see http: //www.astrowise.org/. This reduction system is tailored to the upcoming data of OmegaCam at VST, but it can also be used for many other wide field instruments. The observations presented here, taken with the Wide Field Imager (WFI) at the 2.2m telescope at La Silla, were actually used as a test bed for the system and during the observation run, we mimicked the OmegaCam observing mode. In Section 2.2, we describe our observations. Section 2.3 is a detailed description of the data reduction. A quite detailed description is given, since such a description does not yet exist in the literature for astro-wise. In Section 2.4, we apply various selections to the data, derive galaxy densities, calculate the structural parameters and apply galaxy classification. All the results relevant for this work are presented in Section 2.5. We discuss our results in Section 2.6 in the context of previous work and derive the conclusions in Section 2.7. −1 −1 Throughout this paper we use Ω0 = 0.3, ΩΛ = 0.7 and H0=70 km s Mpc .

2.2 Observations

2.2.1 Targets

The observations consist of 16 pointings (see Table 2.1) centred on a region containing the Abell clusters 2798, 2801, 2804, 2811, 2814 and 2829. These clusters have all dis- tances between z=0.10 and z=0.12. In the remainder we will refer to these 16 pointings as 2dF fields, because these pointings are also covered by the 2dFGRS survey (Colless et al. 2001). Our 2dF fields overlap each other by about 10%. In this way, continuous coverage is achieved over the whole field of view. In addition the Chandra Deep Field South (CDFS) was observed each night, using exactly the same observational set-up. The CDFS observations are used for cross-calibration purposes.

∗ "Based on observations collected at the European Organisation for Astronomical Research in the Southern Hemisphere, Chile under program ID: 074.A-9001(A) † http://www.astro-wise.org/projects/KIDS/ 20 chapter 2: Environmental Influences on the Evolution of Galaxies 1)Ibn opeeeslmtatrapyn h rttoslcinciei seScin2.4.3) Section (see 3.0 criteria using selection limit two completeness first Instrumental the applying (6,9,12): after arcsec), limit in completeness (FWHM band Seeing I (13) (5,8,11): observation, of Night 2.1: Table DS0:22. -27:48:30.0 03:32:28.0 CDFS il A(20)DEC(J2000) (J2000) RA Field 1 2 (3) (2) (1) 00:82. -28:38:30.0 -28:38:30.0 00:38:24.0 -28:38:30.0 00:40:42.0 -28:38:30.0 20 00:43:00.0 -28:38:30.0 19 00:45:18.0 -28:38:30.0 18 00:47:36.0 -28:38:30.0 17 00:49:54.0 -28:38:30.0 16 00:52:12.0 -29:08:12.0 15 00:54:30.0 14 00:36:06.0 13 10 03:40-29:08:12.0 -29:08:12.0 00:38:24.0 -29:08:12.0 00:40:42.0 -29:08:12.0 00:43:00.0 9 -29:08:12.0 00:45:18.0 8 -29:08:12.0 00:47:36.0 7 -29:08:12.0 00:49:54.0 6 00:52:12.0 5 4 3 rpriso l bevdfils 1 ne ffil seas iue29,(-) ih seso n elnto,(4,7,10): Declination, and Ascension Right (2-3): 2.9), Figure also (see field of index (1) fields. observed all of Properties 01-0408 22.8 21.5 0.85 21.7 10-10-2004 1.47 22.1 09-10-2004 1.38 21.7 09-10-2004 1.24 21.5 09-10-2004 1.34 21.8 09-10-2004 1.50 22.8 09-10-2004 1.37 22.8 09-10-2004 0.86 21.5 09-10-2004 1.01 22.1 09-10-2004 1.50 22.1 10-10-2004 1.11 22.8 10-10-2004 1.10 23.1 10-10-2004 0.99 22.8 10-10-2004 0.94 22.7 10-10-2004 0.94 22.6 10-10-2004 1.00 10-10-2004 1.06 10-10-2004 ih S Comp. PSF Night 4 5 (6) (5) (4) V 91-0410 22.4 21.7 1.04 21.9 09-10-2004 1.18 21.5 08-10-2004 1.28 22.1 08-10-2004 1.50 22.7 08-10-2004 1.32 22.4 08-10-2004 1.00 21.9 08-10-2004 0.98 21.7 08-10-2004 1.25 22.1 08-10-2004 1.31 21.5 08-10-2004 1.16 23.7 10-10-2004 1.34 23.2 10-10-2004 0.63 23.1 04-10-2004 0.78 23.5 04-10-2004 0.75 23.3 04-10-2004 0.67 22.9 04-10-2004 0.76 04-10-2004 0.76 04-10-2004 ih S Comp. PSF Night 7 8 (9) (8) (7) R 31-0408 1819.4 19.1 21.8 19.2 21.5 0.82 19.2 21.6 03-10-2004 0.91 19.4 21.5 04-10-2004 0.89 18.8 22.0 04-10-2004 0.82 19.4 21.5 04-10-2004 0.66 19.5 21.8 04-10-2004 0.70 19.4 21.8 03-10-2004 0.80 19.5 21.8 03-10-2004 0.74 19.5 21.8 03-10-2004 0.81 19.3 21.9 03-10-2004 0.90 19.5 21.8 03-10-2004 0.83 19.2 21.8 03-10-2004 0.76 19.3 21.5 03-10-2004 0.84 19.5 21.7 03-10-2004 0.83 19.5 21.7 03-10-2004 0.77 21.5 03-10-2004 0.73 02-10-2004 0.86 02-10-2004 ih S Cmp. PSF Night 1)(1 1)(13) (12) (11) (10) σ I o-on sources. non-point C pxl 2.3: Data Reduction 21

23" 2046 pix N

1 pix = 0.238" E 4098 pix 1 2 3 4 7" Guiding CCD

8 7 6 5 127 mm = 34'

Figure 2.1: Lay-out of WFI chips (taken from Erben et al. 2005)

All fields were observed with the Wide Field Imager (WFI) camera at the 2.2m ESO/MPI telescope between 02 October 2004 and 11 October 2004 with the filters #843 (V band), #844 (R band) and #879 (I band). The WFI contains eight CCDs of 2048 x 4096 pixels with a size of 000.238 pixel−1 resulting in a field of view of 3400 x 3200. Between the CCDs, there are gaps: 23” along the short and 7” along the long side of the chips (see Figure 2.1). The observations were done in dithering mode, using the OmegaCam observing procedure, which optimally benefits the astro-wise reduction system. The procedure dithers five times with each offset shifted 25” in the horizontal and 15” in the vertical direction relative to the previous position. This procedure will fill the gaps during the co-addition phase (see next Section). Each separate dither has an exposure time of 240 s in all passbands, resulting in a total exposure time of 1200 s for the final co-added image (see next Section). The observing conditions during were not photometric during the night of 09/10/2004. This affects the V-band most, see Table 2.1. This table also shows that the FWHM is worst during that night and varies quite significantly. Table 2.1 also shows the estimated instrumental completeness levels for 3σ galaxy-like sources in all passbands. These levels were calculated by estimating the turn-off point in a histogram of number counts (Figure 2.2). Due to several selection criteria, only galaxies that lie far below our instrumental completeness levels will be analysed (see Section 2.4.3)

2.3 Data Reduction

The astro-wise system (Valentijn & Kuijken 2004; Valentijn et al. 2007) was used to reduce the WFI data. astro-wise is a data reduction system for wide field, multi chip, observations, which currently supports many wide field imaging instruments: Omega- 22 chapter 2: Environmental Influences on the Evolution of Galaxies

Figure 2.2: Histogram of number counts in the I band of field 10, without stellar objects. For this field an instrumental completeness level of 21.8 was chosen

CAM, WFI, WFC and MDM8K at the Hiltner telescope. The system was designed and coded by the astro-wise development team; only two external packages are used in the system: SWarp http://terapix.iap.fr/rubrique.php?id_rubrique=49/ and SExtractor (Bertin & Arnouts 1996). The system, written in Python (see http: //www.python.org/), offers a lot of flexibility, allowing the user to modify the re- ductions to her own needs. This is demonstrated below in the case of the photometric pipeline, where some additional corrections with respect to the standard astro-wise procedure were applied. A fairly detailed description of the reduction procedures is given below. Figure 2.3 shows a data flow diagram of the astro-wise data reduction system. In this diagram, the arrows indicate the dependencies of the targets to the raw data or intermediate products. The system can be split into three main parts: a calibration pipeline, a photometric pipeline and an image pipeline. In this Section we describe in detail which processing steps were applied to our data. Standard calibrations like readout noise, bias, cold and hot pixel determination, cosmic ray and satellite track removal, flat fielding and astrometric calibrations were applied to all data using the standard procedures as described in detail on the astro- wise portal, see http://www.astrowise.org/portal/. The standard fringing correc- tion procedure was applied to the I band data. The result of all the reductions is a co-added image with continuous coverage, i.e. all the gaps are filled. In the remainder of this Section, some standard astro-wise procedures are described as well as differences and data specific calibrations with respect to the standard procedures. To the latter category belong: an illumination correction and additional photometric calibrations. 2.3: Data Reduction 23 make lists sExtractor combine lists sAssociate Tell me everything USER Specific make lists sExtractor Seq631 Statistics Seq632 Statistics Seq633 Individual weight image WeightFrame Seq634/C/R Astrometric parameters Seq634 + Photometric parameters Seq636/W/C/cat Calibrated Img Statistics RegriddedFrame seq631 Ingest & statistics RawScienceFrame seq632 de−Bias & Flat ReducedScienceFrame seq633 Individual weight seq634 / req555 Apply astrometry ScienceFrame seq635 Apply photometry ScienceFrame seq636 Coadd images CoaddedFrame Pipeline Cal541 Bias BiasFrame Cal546 Flat field MasterFlatFrame Cal545 Fringe flat FringeFrame Cal546W Weight WeightFrame Cal523 ADU conv. GainLinearity Cal in1 Astrometric reference Cal563 Zeropoint + extinction PhotometricParameters Cal564 Zeropoint + extinction PhotometricParameters req546 MASTER FLAT MASTER WEIGHT Symbol structure req563 − doit ZEROPNT KEY req564 ZEROPNT USER Cal535M Mask MaskFrame Cal/req/seq no. Name / function Class name in code Cal542L Dome Lamp Cal522 Hot pixels HotPixelFrame Cal535 Cold pixels ColdPixelFrame Cal542 Dome flat DomeFlatFrame Cal543 Twilight flat TwilightFlatFrame Cal544 Night sky flat NightSkyFlatFrame Cal545 Fringe flat FringeFrame Cal569 Secondary standards Cal565 User−>Key Bias pipeline Flat field pipeline Image pipeline Photometry pipeline User specific req522 HOT PIXELS req535 COLD PIXELS req533 LINEARITY req548 ILLUMIN. CORR. A data flow diagram. Composites of pipelines: brown: gain/bias/linearity pipeline; red within rectangle: flatfield pipeline; req541 − doit BIAS req542 DOME FLAT req543 TWILIGHT FLAT req544 NIGHT SKY FLAT req545 FRINGE FLAT req523 GAIN req569 2NDARY STAND. Calibration Cal562/PAF Extinction report Figure 2.3: other red: photometric(intermediate) pipeline; data products. yellow: image pipeline. Blue: user choice on extracted sources. Small boxes with soft colours represent 24 chapter 2: Environmental Influences on the Evolution of Galaxies

2.3.1 The Calibration Pipeline

Overscan Correction and Bias Subtraction

Every raw scientific CCD exposure contains a non-zero component, called bias. This extra component needs to be subtracted from every exposure. In astro-wise, there are several ways to subtract this bias level from the images. First, the bias levels contained in the overscan region, are subtracted. Because of the capabilities of astro-wise to reduce multi-instrument data, there are eight different methods to do an overscan correction. As a result, the overscan method is available as a free parameter in the system; the chosen method will be applied to all frames. The advantage of this is that it solves for short timescale variations in the bias level, which are happening in WFI observational set-up. For our data set, method 8 turned out to be the best way: first, it determines a per-row average value of the overscan region in X direction and then applies a boxcar smoothing function that averages over 50 rows in the Y direction. After subtracting the overscan levels from the raw bias images, a master bias is constructed, which contains any remaining structure but has an average count level of 0. This master bias frame is then subtracted from all subsequent science images.

Hot and Cold pixel determination

Hot pixels are pixels which have high count rates despite not being illuminated. In astro-wise, these pixels are detected from bias images (which have an exposure time of 0 seconds). More precisely: >5 sigma outliers in bias are defined as hot pixels. Cold pixels are broken pixels which have low or zero counts even when illuminated. These pixels are determined from dome flat-field exposures because those have high counts across the image. In astro-wise all pixels that deviate substantially, i.e. more than 4% of its surroundings, from the other pixels in the (dome) flat-field are considered cold even though brighter pixels are also detected. All deviant pixels are flagged in weightmaps.

Flat Fielding

In the astro-wise system there are different ways to construct a flat field. We chose to use both dome and twilight flatfields. Dome flat-fields are created by pointing the telescope at a screen on the inside of the dome which is illuminated by lamps. Dome flat fields have the advantage that it is easy to repeatedly obtain a high signal to noise level. Disadvantages are that the direction in which light enters the telescope may be different than during night time observations, that the colour of the dome lamp differs from the colour of the night sky and that it is very difficult to illuminate a screen in such a way that it is a source of uniform radiation. A dome flat field is useful for tracing small scale structure variations. For twilight flat fields, taken during twilight, all the (dis)advantages valid for dome flatfields, are reversed. Another disadvantage for twilight flats is that they can already contain objects like stars during exposures, which should be corrected for by dithering the twilight flats. Twilight flat fields thus are better in tracing large scale structure variations. These considerations result in the desire to combine dome flats and twilight flats by spatially filtering the two types of flat fields. 2.3: Data Reduction 25

The procedure in astro-wise to obtain the final master flat, which flat fields all subsequent science images, is as follows: first, a minimum of five raw dome (twilight) flats are combined into a master dome (twilight) flat. This is done by normalising the raw flats to 1 and averaging each pixel value, rejecting any outliers. Next, low (high) Fourier filters are applied to the master dome (twilight) flat, keeping the small (large) scale structures and taking into account the weightmaps of the previous Section. Finally, the master dome and twilight flatfields are combined, by again doing a normalisation, resulting in the final master flat field.

Fringing Corrections In exposures of (back illuminated, thinned-) CCD detectors while using broad-band or near infra-red filters, “fringes” may be visible. These look like the structures seen in oil as it reflects light (thin film interference). Photons incident on a back illuminated CCD detector enter a silicon layer where they liberate electrons, which are trapped by the pixel structures. In these CCDs, for photons of sufficiently low energy (near infra-red), the silicon layer can have an optical depth of more than twice the thickness of the silicon layer. When this is the case, photons that reflect in the CCD can interfere with those incident on it, leading to an interference pattern across the chip, the shape of which depends on variations in the thickness of the silicon layer and the nature of the incident light. It is known that in the case of broad band filters, where normally any fringing would be invisible against the background of broad band radiation, the origin of the visible fringes are the distinct atmospheric emission lines of some molecules and ions in the near-infrared (mostly OH and O2). The fringe pattern and its stability, particularly the amplitude of the fringes across the CCD, varies from (telescope-) site to site and no satisfactory models for their behaviour exist. The amplitude of the fringes can be up to about 15% of the sky background, depending on lines, passband and chip thickness. The nasty aspect of fringes is that they represent an additive term like the bias, but that they are only seen combined with the flat-field response, which is a multiplicative term. This makes the separation difficult. Fringes form structures on the scale of sev- eral arcseconds, (dozens of pixels in the case of WFI), at least in the direction across the fringes, which means they are hard to computationally distinguish from stars, and hence source extraction programs can deliver unreliable results.

Fringe maps are constructed from a set of partially reduced science frames. After overscan trimming, de-biasing and flat-fielding an science image, it is assumed that any remaining systematic effects (i.e. aside from the astronomical objects) are either bad pixels or fringes. In order to create a fringe map that can be used to correct all images taken through the same filter during a night, a set of science images is reduced and normalised. These images are placed in a cube and a median average is calculated. The median average corrects very well for any outlying values, so any stars, cosmic ray events or satellite tracks are removed, at least when sufficient images are provided to base the median average on (∼ 10). During a night the brightness of the emission lines will change, especially near evening and morning twilight. The result of this is that the amplitude of the observed fringes will change. Therefore, fringe maps should be scaled to fit the amplitude of the fringes in each science frame. We calculate this from the standard deviation in a 26 chapter 2: Environmental Influences on the Evolution of Galaxies science image, which is derived from all non-bad pixels that have values within a given threshold from the median background level. It is assumed that this standard deviation depends on the amplitude of the fringes.

2.3.2 Photometric Calibrations The photometric calibrations were applied in three steps. The first step is part of the astro-wise system and provides a basic zeropoint calibration for each chip during a night using standard fields. This works only if the night is photometric, which is not valid for 09-10-2004 (see Table 2.1). Zero point calibration was done using the stars within the selected areas SA107 and SA110, which were observed each night. Magni- tudes of stars were extracted from several photometric standard catalogues available in the system (Verdoes Kleijn et al. 2007), which cover these SA fields (i.e. Sloan, astro- wise secondary, standards, Landolt). After comparison of the instrumental magnitudes with the standard star magnitudes, the zeropoint was determined for each chip. The second step accounts for small errors in the zeropoint between different chips. Here we make use of the dither pattern: neighbouring chips are used to compare the magnitudes of sources present in the overlapping regions. Usually there are a sufficiently number of bright sources present in these regions. After association, the magnitudes can be compared. This results in a self-consistent way to verify the quality of the zero points in each chip. The corrections were usually of the order of several hundredths of magnitudes. By taking into account all offsets between all possible overlapping regions, we minimised the differences by applying the method of Maddox et al. (1990). After the zero-point corrections are applied to each chip, a co-added image can be produced which has a more homogeneous zeropoint over its field of view. All magnitudes appearing in this paper were converted to the Johnson-Cousins sys- tem and were corrected for galactic extinction. To convert to the Johnson-Cousins system we applied the following conversion formulae, directly taken from the EIS web- pages ∗. Note the large colour term in V.

VJ − VWFI = −0.12(VJ − RJC ) (2.1)

RJC − RWFI = 0.01(RJC − IJC ) (2.2)

IJC − IWFI = 0.03(RJC − IJC ) (2.3)

Here VWFI , RWFI and IWFI are the WFI instrumental magnitudes and VJ , RJC and IJC are in the Johnson-Cousins system. The 2dF pointings are located about 2 degrees from the South Galactic Pole (SGP). Across the entire observed field small variations in extinction of the order 0.01 magnitude in each passband are present, as shown by the Galactic extinction maps of (Schlegel et al. 1998). We adopted the following averaged extinction values of 0.070, 0.056 and 0.035 magnitudes in the three bands respectively. The galactic extinction values in the CDFS field are much lower, i.e. 0.026, 0.021 and 0.015 respectively. In the final step, we match all coadded frames to the same V-I colour by matching colour-colour plots of point-like sources present in each coadded frame. This is necessary, because we expect still significant differences in zeropoints between various coadded frames, mainly due to non-photometric conditions in the V band. We used point-like

∗ http://www.eso.org/science/projects/eis/surveys/readme/70000027 2.3: Data Reduction 27

Figure 2.4: Colour-Colour diagrams: V-I vs. R-I plots of point-like objects present in the 16 2dF fields. Offsets are visible, which are due to systematic photometric zeropoint errors most likely due to extinction variations in V. Horizontal lines indicate the levels in V-I of the intrinsically brightest stars; corrections in V-I for each field were applied to match the reference V-I level of field 18. sources are with CLASS_STAR > 0.85, magnitudes fainter than 16 (no saturation) and brighter than 19.5 (stars are still easily recognised by the SExtractor neural network with output CLASS_STAR). Figure 2.4 shows the results. There are indeed offsets in V-I visible; these offsets in V-I were determined and matched to field 18 as the reference field.

Illumination Correction astro-wise has the ability to apply an illumination correction. WFI suffers from il- lumination variations across the field of view (Manfroid & Selman 2001; Koch et al. 28 chapter 2: Environmental Influences on the Evolution of Galaxies

Figure 2.5: Illumination correction in I. The amplitude from edge to centre is 0.09 magnitude

2004), which is caused by reflections off the telescope corrector. This causes an addi- tional photometric gradient across the field of view with an amplitude of about 0.10 magnitude. Moreover, these variations change each time the telescope configuration is modified, which happened quite often in the lifetime of this telescope ∗. After each configuration change, a new illumination correction should be applied. According to the WFI logs, the latest change made before our observation period was done in April 2004. To determine an illumination correction, we downloaded suitable deep data from the WFI archive (i.e. standard fields with sufficient exposure time to get at least 100 stars per frame, taken between April 2004 and October 2004,) and applied the astro-wise illumination algorithm. Figure 2.5 shows the resulting illumination correction for the I band image. The amplitude of the changes is about 0.10 magnitude which is typical for the WFI camera (see again Manfroid & Selman 2001 and Koch et al. 2004)

2.3.3 Image Pipeline Once the calibration pipeline and photometric pipeline have been passed, the final stage: the image pipeline, results in a set co-added images. In this Section, we present a verification of the astrometry of co-added images..

Astrometric Verification To verify the astrometry of a Coadded image, the RMS errors were calculated with respect to the USNO catalogue (Monet et al. 2003). These errors are typically of the order of 1 pixel (0.238 arcsec). Figure 2.6 (left) shows a typical example for the CDFS field vs. the USNO catalogue in the R band. The RMS errors of 0.22” consist of two parts which contribute to the final RMS error. First, there is an external error. This

∗ http://www.eso.org/sci/facilities/lasilla/instruments/wfi/inst/HistoryOfChanges2p2.html 2.4: Data Analysis 29

1 1

0.5 0.5

0 0

-0.5 -0.5

-1 -1 -1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1

Figure 2.6: Left: RMS error of coadded CDFS field vs. the USNO catalogue in the R band. RMS error is 0.22.Right: Internal astro-wise error after comparison of V and R band of CDFS field taken at the same night. RMS is here 0.06 arcsec. error contributes the most, it is present within the USNO catalogue and is about 0.20 arcsec. The second source of error is the internal error of the system caused by LDAC deformations to correct for image distortions. Here, the internal error was derived by comparing the positions of sources present in two co-added frames, which were derived from observations of the CDFS field, taken in the same night in the V and R bands. Figure 2.6 (right) shows the results: an small internal rms error of about 0.06” is derived. For more information regarding the quality of the astro-wise astrometry, the reader is referred to http://www.astro-wise.org/Public/Astro-WISEAstrometryReport. pdf

2.4 Data Analysis

2.4.1 Source catalogues Catalogues of each co-added image were made by running SExtractor (version 2.4.4, Bertin & Arnouts 1996) on each image. For each passband, we combined all cata- logues into one big catalogue. The co-added images have signficant overlap (10%), therefore, we only kept sources with the highest S/N, in the overlapping regions. The lower S/N is due to the lower number of co-added frames or dithers there, but the overlap is always large enough to have objects with constructed from five co-added frames. SExtractor configuration parameters can be important as they can affect the magnitudes and classification of objects. First of all we set the background estima- tion parameters BACK_SIZE and BACK_FILTER size to 128 and 4. In this way, only very large scale gradients are removed without affecting regions around larger galaxies like cD galaxies. After a lot of fine-tuning, we set DETECT_THRESH=3.0, DEBLEND_MINCONT=0.003 and DEBLEND_NTHRESH=32, so that most faint 30 chapter 2: Environmental Influences on the Evolution of Galaxies objects are kept and deblending levels are not too low or too high. These three pa- rameters are important to construct good quality masks, which are needed to calculate the morphological parameters (Section 2.4.4). Finally we put the SEEING_FWHM at a value valid for the particular night. SEEING_FWHM is used within SExtractor to calculate the output CLASS_STAR, which discriminates between stellar-like objects and galaxy-like objects. We let SExtractor also produce three additional images (SEG- MENTATION, BACKGROUND_RMS, -OBJECTS) which are used at various places in the structural parameter analysis (Section 2.4.4). The detected objects are shown in the lower part of Figure 2.9: the Abell clusters are quite clearly visible as dense concentrations of point sources. Four big structures clearly stand out: at redshift 0.102, Abell 2829 (RA=12.85 degrees, Dec=-28.5 degrees); at redshift 0.112, Abell 2811 and 2814 (RA=10.66 degrees, Dec=-28.6 degrees); Abell 2804 (RA=9.8 degrees, Dec=-29.0 degrees); Abell 2801 (z=0.112) and 2798 (z=0.107) at RA=9.5 degrees and Dec=-28.5 degrees.

2.4.2 Source Selections Several stages of selection criteria were applied to the source catalogues derived using SExtractor. These criteria will remove most bad and spurious objects, stars as well as galaxies beyond our target redshift interval of z=0.11 ± 0.02

1. Association. The three big source catalogues were associated using the association method in astro-wise. We associated using a maximum separation of 1” in chain mode. This separation is larger than the astrometric error of 0.26”, because the astromet- ric errors of extended objects are roughly 2 times compared to those of pointlike objects, i.e. 0.26”. This associated catalogue will be used in the remainder of this paper.

2. First order SExtractor criteria. A first order data selection was done by constraining several SExtractor param- eters. This selection was only applied to the I band data, which were observed in photometric conditions and have the smallest and constant PSF along all fields. After some experimentation, we arrived at several constraints. First, we kept only objects with FLAG < 4. According to the FLAG definition, objects with FLAG values 1-3 are blended with other objects. In our case these could be of potential interest in case they are real merger events. Furthermore, objects within clusters lie often close to other galaxies, and we certainly do not want to throw these away. We also keep only objects which have at least 100 pixels detected above the detection threshold (DETECT_THRESH=3.0× S/N). We note that the objects with a size of 100 pixels are still very small: with a pixel scale of about 0.238, 100 pixels correspond to objects having a theoretical circular diameter of about 2.7 arcsec, which is equal to 5.4 kpc at z=0.11. Finally, we applied the selection CLASS_STAR < 0.85. Using these criteria, still a lot of bright saturated stars pass the selection criteria: these were easily identified by eye and removed from the list. As a last check we ensured that no important objects are missed using these selection criteria by looking at labels of the selected objects which were overplotted on the coadded images. 2.4: Data Analysis 31

3. Colour criterion and redshift information. A second pass on the data selection involves redshift information. Rough red- shift information can be obtained by looking at the colour-magnitude diagram of the objects. Since our target galaxies lie at z = 0.11 ± 0.02, we can still remove a lot of objects by applying a colour criterion. A first insight regarding red- shift information is gained by using the CDFS data, which have similar observing characteristics. For this field, the only available nearby redshifts are photometric redshifts from COMBO-17 (Wolf et al. 2004), which field of view almost com- pletely coincides with our CDFS observations. Their photometric redshifts have much higher uncertainties (≈ 0.01) compared to our spectroscopic redshifts, but a rough comparison for galaxies at z=0.11 is still possible. The advantage is that photometric redshift data is available for 96% out of 404 selected galaxies in the CDFS field, giving distance information about almost all fainter galaxies in the colour-magnitude diagram. Figure 2.7 shows the colour-magnitude diagram for the CDFS field, with colour coding corresponding to target redshift information. With red triangles representing galaxies being further away than z=0.14 shows that all points with V-I > 1.44 are background objects. Contamination from background objects with respect to target objects (blue squares) becomes severe at I ≈> 18.

Figure 2.8 shows the colour-magnitude diagram of all our galaxies selected in our four square degree 2dFGRS field so far. The red sequence (see Baum 1959; Bower et al. 1992) of passive early type galaxies is clearly visible. A regression fit for elliptical galaxies with spectroscopic redshifts near the obviously visible red sequence strip galaxies gives (V −I) = −0.015∗I+1.58 with half-width ± 0.07 in V- I. The calculated width is a combination of the intrinsic width (HST data on Coma cluster red sequence galaxies give an intrinsic half-width of about 0.05 Bower et al. (1999)) and photometric errors. Our calculated thickness of the red sequence thickness implies a photometric error equal to 0.05. These photometric errors are due to small photometric off-sets in the different fields, remaining illumination correction features and photometric measuring errors f.i. due to background or blending problems. We finally define the red sequence using (V − I) = −0.015 ∗ I + 1.58 with half-width ± 0.10 in V-I and this was used in the remaining part of this Chapter. We use a slightly larger half-width to encompass more outliers.

The colouring in Figure 2.8, is similar to Figure 2.7 and shows that our redshift information is far from complete. Due to the 2dFGRS selection of sources in B, redshifts from red galaxies become incomplete starting from I > 17.3. On the other hand, all our available redshifts are spectroscopic redshifts with very small errors << 0.01, giving a more effective selection tool compared to the CDFS data. Combining the information of the CDFS field, we select all galaxies below the previously determined red sequence upper border: (V −I) < −0.015∗I +1.68, shown in the plot as the black line. The few galaxies above the line which are within our target redshift range were also included. Their red colour might result from dust reddening or the above mentioned photometric errors. 32 chapter 2: Environmental Influences on the Evolution of Galaxies

3

2

1

0

14 16 18 20

Figure 2.7: Colour-Magnitude diagram: I vs. V-I of the CDFS area with COMBO-17 redshifts. All associated 404 objects are plotted. Blue squares correspond to galaxies with redshifts 0.08 < z < 0.14, green starred symbols represent foreground galaxies with z < 0.08, red triangles are background galaxies with z > 0.14; the few black crosses represent objects for which no redshifts were found in the literature

2.4.3 Galaxy Density Calculations. Reliable measurements of densities are very important because these will be used to characterise the environment. In Section 2.5, we will plot various galactic properties as a function of density. In this Section, three density related quantities are calculated: a 2D projected surface density for each position in the field of view, a 2D background surface density estimate which is relevant to define overdensities and a 2D projected surface density for each galaxy based on 3D information (i.e. redshifts).

2D projected density In this Section a surface density map of the galaxies is derived. This map assigns a density to each point in our four square degrees 2dF field. Before proceeding, the limiting magnitudes in each 2dF pointing should be considered. It is important to use a common magnitude level for all 2dF pointings, otherwise artificial full or empty areas could be introduced in the map. The limiting magnitude in each field was calculated using our first two selection criteria. The first selection, i.e. associations between V and I band objects, implies that the highly variable V band data quality is introduced. Table 2.1 shows that our V band data have much worse quality than our I band data, with varying PSF and instrumental completeness levels. Therefore, it is expected that we will find variations in the limiting magnitudes limits for each pointing. After making histograms of number counts for each pointing: the maximum number count bin was 2.4: Data Analysis 33

3

2

1

14 16 18 20

Figure 2.8: Colour-Magnitude diagram: I vs. V-I of the whole 4 square degree 2dF area. The sources are depicted as follows: blue squares correspond to galaxies having the target redshifts (0.09 < z < 0.13), green starred symbols represent foreground galaxies with z < 0.09, red open triangles are background galaxies with z > 0.13; black crosses represent objects without redshift information. 34 chapter 2: Environmental Influences on the Evolution of Galaxies taken to be limiting magnitude. These limiting magnitudes for the I band are listed in the last Column of Table 2.1. It is clear that field 16 has the worst quality, which is very likely related to the large the V-band PSF of 1.5 arcsec. The magnitude limit of Field 16: 18.8 in I, was used for all other fields. After applying all three selection criteria plus the magnitude limit of 18.8 in I, there remain 2022 sources. These sources, shown in the lower part of Figure 2.9, were used to derive a density map for the whole four square degrees region. The 10th nearest neighbour approach of Dressler (1980) was used to calculate the densities. For each point in the region, the algorithm takes the 10 nearest galaxies. The radius of the 10th 10 galaxy translates into a value for the surface density according to: Σ10 = 2 . If the πdN radius falls outside the field of view we only consider the area visible in the field of view to calculate the density. Doing this for the whole 2dF field of view results in the upper part of Figure 2.9. From this map a density value can be assigned for each point in the whole four square degrees region which will be used later in the analysis.

Background Surface Density We use a background surface density estimation to define the high density regions. For this purpose, an average number count of the most empty regions from the density map (2dF fields 3, 4, and 16) was calculated. This resulted in a background density of 3.94 ± 0.37 galaxies per Mpc−2. This number is only valid using the criteria used to obtain the density map, i.e. I < 18.8 etc. (see above). Field galaxies and cluster galaxies were defined as galaxies residing in the low and high density regime respectively, where the separation was set at five times the background surface density: 19.7 galaxies per Mpc−2. At this density, we have at most 20% contamination from fore and background galaxies. Using this value for the background surface density, we define four different density regimes used throughout this paper. Density regime 1 (’the field’) equals 0- 2.5 times the background density. Density 2 (’intermediate density’) define regions between the cluster and field, e.g. loose group environments, and equal 2.5-5 times the background density. Density 3 objects belong to outer cluster regions with 5-10 times (19.7-39.4 galaxies per Mpc−2) the background density while density 4 objects reside in the centres of clusters with densities > 10 times the background density (> 39.4 galaxies per Mpc −2). These densities are similar to those found by Dressler (1980), who use a fainter limiting magnitude MV = −19.7 compared to our estimated MV =-20.3.

Surface Densities using velocity information Our redshift interval from z=0.09 to z=0.13 corresponds to a linear dimension of ap- proximately 120 Mpc. A better density estimate can be made for all those galaxies for which redshifts are available, if those redshift are used as well. 96% of these galaxies belong to the 2dFGRS sample, which is roughly complete until B = 19.0 (Colless et al. 2001). A slightly modified version of the density calculation algorithm of Section 2.4.3 was applied to these galaxies by calculating the distance to the 5th nearest neighbour for each galaxy within ± 1000 km/s (this is about 2 times the velocity dispersion of our galaxy clusters) and a radius of 2.0 Mpc. If fewer than 5 galaxies were found, a density of zero was assigned. Similar to Section 2.4.3, here we also define four regimes. Since no background value is given here, each regime is defined by demanding that it contains a similar fraction of galaxies with respect to each 2D density regime (see 2.4: Data Analysis 35 : distribution Bottom . 2 − 19.7 galaxies per Mpc > : Density map using the 10th nearest neighbour recipe (Dressler 1980), using the colour selected sources with (2D Sample 1). The white regions correspond to high surface densities: 8 . 18 < JC Figure 2.9: Top I of sources from 2D Sample 1 (see Section 2.4.5). Both RA and Dec are in degrees. The map shows a region of 7.5 by 30 Mpc. 36 chapter 2: Environmental Influences on the Evolution of Galaxies previous Section). The results in density regime 1 (’the field’) having galaxies with densities lower than 0.6 galaxies per Mpc−2. Density 2 (’intermediate density’) corre- sponds to 0.6-2.0 galaxies per Mpc−2. Density 3 objects belong to outer cluster regions with 2.0-5.9 galaxies per Mpc−2 while density 4 objects have more than 5.9 galaxies per Mpc−2. The densities calculated are similar to those found by Goto et al. (2003), who apply a fainter magnitude limit of MR = -20.6 compared with our MI ≈ -20.8 and use somewhat more nearby galaxies between 0.05 < z < 0.10.

3D density regimes vs. cluster virial radii

While density is the major parameter to analyse various galaxy properties as a function of environment, many authors compare these properties with the cluster virial radius Rvir. We think it is not appropriate to do this: many clusters have irregular shapes (or are not virialized) and therefore, a radius becomes an inconsistent parameter to be used in the analysis of galaxy properties. However, in order to compare with other authors, we try to give a rough estimate how our density regimes relate to Rvir. We calculated Rvir for four clusters using the recipes of Hilton et al. (2005) and Girardi et al. (1995). With the recipe of Hilton et al. (2005), we find 30% smaller Rvir than with the recipe Girardi et al. (1995). We just took the average of both recipes and find a Rvir for Abell 2804 and 2829 of ≈ 1.5 Mpc ±0.25. For Abell 2798 and 2811, we find Rvir ≈ 1.75 ± 0.25 Mpc. Looking at our distribution of galaxies in each density regime, we find that our density regime 4 corresponds to regions at roughly < 0.6×Rvir; density regime 3: 0.6 × Rvir < R < 1.5 × Rvir; density regime 2 corresponds to outer cluster regions or filaments and groups while density regime 1 corresponds to isolated galaxies.

2.4.4 Structural/Morphological parameters In this Section we present the structural parameters which are included in the catalogue; these are: asymmetry, concentration, Gini coefficient, ellipticity, M20, Sérsic index, R50, surface brightness µ50 at R50, D23 and ∆(V −I) colour gradients. The meaning of each parameter is described briefly below. Except for ∆(V − I), each parameter was cal- culated in the I band, where the seeing was relatively stable (see Table 2.1). Except for ellipticity, all parameters were calculated from small thumbnail images with ac- companying masks. The masks were derived from the SEGMENTATION SExtractor checkimage, by identifying, within the SEGMENTATION thumbnail, all pixels belong- ing to other sources .

Asymmetry

Asymmetry A (Conselice et al. 2000; Lotz et al. 2004) measures the degree of rotational symmetry of the light distribution by calculating the normalised difference between the galaxy image and the image rotated by 180◦. Low values near zero indicate a very high degree of symmetry, while large values (up to 0.3) indicate large asymmetry. A measures gravitational perturbations and star formation, and is a good indicator to separate late type disk galaxies from early type galaxies. A was calculated in an elliptical aperture (ellipticity and position angle from SEx- tractor) with a semi-major axis of 1 petrosian radius using the formula: 2.4: Data Analysis 37

P x,y |I(x,y) − I180(x,y)| A = P − B180; (2.4) |I(x,y)| o where I is the galaxy flux in pixel (x, y), I180 is the image rotated by 180 about the galaxy’s central pixel, and B180 is the average asymmetry of the background. The correction for background noise, B180, was calculated from a nearby noise thumbnail image. The noise thumbnail image was taken from the SExtractor -OBJECTS image (i.e. the original image with all detected objects subtracted), by choosing a similar sized area from the eight neighbouring regions with the most equal noise characteristics as our thumbnail image. The best candidate was chosen by: 1) taking the image with the most comparable noise RMS. In this way, we account for sources that lie in regions with varying noise characteristics, in particular:the WFI gaps which have larger noise RMS. The RMS of our target galaxy thumbnail image was taken from the SExtractor checkimage BACKGROUND_RMS thumbnail. The RMS values in this global checkim- age are smoothed (with SExtractor parameters BACK_SIZE and BACK_FILTER size to 128 and 4, see Section 2.4) and are not expected to be influenced by the (small) tar- get galaxy. 2) the noise thumbnail must contain not more than 10% pixels which are occupied by other sources. If there are some pixels occupied by a source in our final chosen noise thumbnail image, these were replaced by randomly choosing neighbouring noise pixels.

Concentration This parameter gives an indication of galactic light distribution. To calculate this we used the formula given in Bershady et al. (2000) applied to an elliptical aperture: r C = 5 log 80 (2.5) r20 with r_x the semi-major axis containing x percent of the galaxy light contained within 1.0 Petrosian radius. C typically ranges between 1 and 5 for low and high concentration respectively. Concentration seems to be correlated with mass (van der Wel 2008); bright early type elliptical galaxies usually have high concentration values.

Ellipticity The ellipticity,  = 1−b/a, with b and a the minor and major axis respectively, is taken from the SExtractor output. The ellipticity also defines the inclination angle for disk galaxies using: s 2 − 2 sin(i) = 2 (2.6) 1 − q0

with i the inclination and q0 the intrinsic disk axis ratio (i.e. the thickness of the disk). Here, q0 = 0.2 was used (Erwin 2005, originally from Lambas et al. 1992).

Gini coefficient The Gini coefficient, G, describes how uniformly the flux is distributed among galaxy pixels (Abraham et al. 2003). The Gini takes a value of 0, if the galaxy light is equally 38 chapter 2: Environmental Influences on the Evolution of Galaxies distributed among all pixels within a certain radius. The Gini coefficient is 1 if all the light is concentrated in one pixel. Using an elliptical aperture (ellipticity and position angle from SExtractor), the Gini coefficient was calculated within 1.0 petrosian radius using:

n 1 X G = (2i − n − 1)X , (2.7) ¯ i Xn(n − 1) i where n is the number of pixels within 1.0 petrosian radii, and X¯ is the mean pixel value

M20 M20 (Lotz et al. 2004) is the second order moment of the brightest 20% of the galaxy flux defined by:

X 2 M20 = (log( fi × di ))/Mtot (2.8)

with Mtot the total second order moment, fi the flux value of pixel i at distance P di from the galaxy centre valid for all i with fi <= 20%, where flux values fi have been sorted towards decreasing values. M20 is low for centrally concentrated objects and high for late type disk galaxies containing bright off–centered star forming regions. M20 was calculated within 1.0 petrosian radius using a circular aperture.

Sérsic index The Sérsic index n, characterising the steepness of the surface brightness distribution, is often used to classify galaxies, with disk dominated galaxies have light profiles near n=1 and early type,elliptical galaxies having n near 4. The ellipse fitting tool GALPHOT (Jørgensen et al. 1992) was applied to the thumbnail images of the galaxies using the previously determined thumbnail masks. GALPHOT returns information such as ellipticity, position angle, surface brightness and the C3,C4,S3,S4 coefficients (Carter 1978), all as a function of radius. A model galaxy subtracted residual image is returned as well. The best ellipse fits were obtained by allowing the centre, position angle and ellipticity to vary. Surface brightness profiles were obtained by plotting the surface brightness for each fitted ellipse as function of radius. The 1D, I band surface brightness profiles were fitted by a Gaussian (with a FWHM of 1 arcsec) convolved Sérsic profile. The Sérsic profile is given by: " #  r 1/nSer µ(r) = µe + cn − 1 (2.9) rc

with cn = 2.5(0.868n − 0.142) (valid for 0.5 < n < 16.5; Caon et al. 1993).

Fits were made on data points within a certain range from the centre of the galaxy. The starting point was set at that point where the semi-minor axis of the ellipse as obtained by SExtractor reaches 1.5 arcsec. This is slightly more than three times the seeing PSF radius. This radius has been chosen first, to minimise PSF smearing issues and secondly, to minimise the influence of bulges (Graham & Driver 2005). At z≈ 0.11, 2.4: Data Analysis 39

1.5 arcseconds is about equal to 3 kpc. which is a larger than an effective radius of typical bulges of spiral galaxies (see Figure 6 of de Jong et al. 2004). The outer limit was set at that point where the flux drops below 5× the error in the fitted elliptical ring as determined by GALPHOT or if the surface brightness reaches 25.0 mag arcsec−2

R50, half light semi-major axis

R50 represents the half light semi-major axis: within this radius, half of the flux of the galaxy resides. This parameter was calculated using fixed elliptical apertures, applying fixed values for the centre, ellipticity and position angle, taken from SExtractor. All galaxy thumbnails were rebinned to 10 times smaller pixels. This increases the precision of the estimate for R50, especially for low values of R50. Masks were also taken into account by replacing masked pixels with the average of the remaining pixels within the elliptical ring.

µ50, average surface brightness within R50

µ50 is the average surface brightness within the in the previous Section calculated el- liptical aperture R50.

Colour gradient ∆(V-I).

To calculate valid V-I colour gradients, all galaxy thumbnail pairs must have similar PSF sizes. Table 2.1 shows, that for all 2dF pointings, the PSF of the I band is smaller than for the V band. Therefore, we convolved each I band coadded image with a p 2 2 Gaussian of size PSFV − PSFI , in order to reach similar ’seeing’ in I band. Using again fixed values for the centre, ellipticity and position angle taken from SExtractor, profiles were calculated for all thumbnails taken from V band image and newly generated thumbnails for the PSF corrected I band coadded image. Finally a linear fit was made to the the colour profile using a log scale in radius, between 2” from the centre and until a points where the I band or V band surface reaches 24 magnitude or 25.4 magnitude respectively. Figure 2.10 shows an example for a galaxy with magnitude I=16.8.

2dF Star formation indicator η

This parameter is only available for galaxies with redshifts taken from the 2dFGRS and is defined by Madgwick et al. (2002). The parameter indicates the level of star formation based on the averaged emission and absorption line strength in the galaxy rest frame spectrum. It is correlated with the equivalent width of the Hα emission line, which is in turn related to star formation. Values for η range between -5 to +10; with values of η lower than -2 having no measurable Hα emission at all while n≈7 corresponds to an equivalent width of 50 angstrom.

D23

D23 is the semi-major axis within the elliptical aperture, where the surface brightness in I reaches 23 magnitude per square arcsec−2. 40 chapter 2: Environmental Influences on the Evolution of Galaxies

1.5

1

0.5

0 0.5 Figure 2.10: Example of V-I profile for a I=16.8 magnitude galaxy, with vertically the V-I colour and horizontally a log(radius[arcsec]) scale. Vertical dotted lines show the fit range. The fit is also plotted as an dotted line with slightly negative slope.

2.4.5 Sample Definitions

In this Section, we summarise and define various data samples used throughout this work. The main goal is to obtain several working data sets with high completeness levels regarding redshift information. For many galaxies, no redshift information is available and other considerations, e.g. density, have to be taken into account whether to include these galaxies or not in a particular data set. Table 2.2 shows, for several IJC magnitude bins, the number counts and several statistics for objects inside our target redshift range (z = 0.11±0.02), outside our target redshift range and for objects without redshift information. These numbers are directly related to Figure 2.8. The left and right side of Table 2.2 show these numbers for low and high density regime respectively. The two density regimes are separated at densities 5 times the background density = 5 × 3.94 = 19.7 galaxies per Mpc−2 (see Section 2.4.3). The high density regime corresponds to the white regions in the density map of Figure 2.9, which correspond roughly to few times the virial radius for the clusters. Columns 5 (and 10) shows the following percentage: #column3/(#column3 + column4). This is the percentage of galaxies with unknown redshifts versus the galaxies sample of galaxies within z = 0.11 ± 0.02 plus unknown redshifts in each magnitude bin. For the low density regime, the redshift completeness is high (> 70%) for all mag- nitude bins lower than IJC < 17.3. Columns 6 (and 11) list the percentage of galaxies within z = 0.11 ± 0.02 compared all galaxies with redshifts known within the particular magnitude interval. From Figure 2.8, it is obvious that the redshift completeness is higher for blue galaxies: this is due to the magnitude limit of the 2dF survey, which selects galaxies up to MB = 19.0, which results in intrinsic red galaxies fainter than about I=17.4 not being selected in 2dF. For blue ((V − I) < 1.1) galaxies this implies that our completeness is slightly higher at IJC = 17.8: this is shown in the lower part 2.4: Data Analysis 41

Low Density. High Density Mag. bin out unkn. in % unkn. % in out unkn. in % unkn. % in (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) < 14.3 4 0 0 - 0.0 1 1 1 0.5 0.5 14.3 <= IJC < 14.8 4 1 0 1.0 0.0 0 0 4 0.0 1.0 14.8 <= IJC < 15.3 11 7 6 0.54 0.35 1 1 11 0.08 0.92 15.3 <= IJC < 15.8 6 8 35 0.19 0.85 1 1 26 0.04 0.96 15.8 <= IJC < 16.3 25 12 56 0.18 0.69 7 12 33 0.27 0.82 16.3 <= IJC < 16.8 54 15 69 0.18 0.56 5 7 46 0.13 0.9 16.8 <= IJC < 17.3 68 38 105 0.27 0.61 11 27 48 0.36 0.81 17.3 <= IJC < 17.8 87 150 74 0.67 0.46 12 74 27 0.73 0.69 17.8 <= IJC < 18.3 31 315 38 0.89 0.55 3 124 8 0.94 0.73 18.3 <= IJC < 18.8 11 487 7 0.99 0.39 4 146 2 0.99 0.33 V-I<1.1 galaxies < 14.3 3 0 0 - 0.0 0 0 0 - - 14.3 <= IJC < 14.8 0 0 0 - - 0 0 0 - - 14.8 <= IJC < 15.3 4 1 0 - 0.0 0 0 1 0.0 1.0 15.3 <= IJC < 15.8 4 1 3 0.25 0.43 0 0 0 - - 15.8 <= IJC < 16.3 7 5 3 0.63 0.3 0 2 2 0.5 1.0 16.3 <= IJC < 16.8 25 5 14 0.26 0.36 2 1 2 0.33 0.5 16.8 <= IJC < 17.3 17 5 23 0.18 0.57 2 2 8 0.2 0.8 17.3 <= IJC < 17.8 46 16 44 0.27 0.49 6 5 9 0.36 0.6 17.8 <= IJC < 18.3 28 67 37 0.64 0.57 2 20 6 0.77 0.75 18.3 <= IJC < 18.8 10 222 6 0.97 0.38 4 47 2 0.96 0.33

Table 2.2: Galaxy counts: Upper part: all galaxies, Lower part: blue galaxies with V-I < 1.1 (1) magnitude bin), (2-6): Statistics for field galaxies with density lower 19.7 galaxies/Mpc−2, (7-11).Statistics for the high density regime (> 19.7 galaxies/Mpc−2). Columns 2 and 7: number of galaxies outside redshift range 0.09 < z < 0.13; Columns 3 and 8: number of galaxies without redshift information; columns 4 and 9: number of galaxies within redshift range 0.09 < z < 0.13; Columns 5 and 10: percentage of galaxies with unknown redshifts with respect to all galaxies within redshift range plus galaxies with unknown redshifts; columns 6 and 11: percentage of galaxies within redshift range with respect to all galaxies with redshifts known. Second part below dividing line: similar for blue galaxies with V-I < 1.1 where we can go 0.5 mag deeper (see Figure 2.8) of the Table 2.2 for galaxies which have V-I < 1.1. The 130 galaxies without redshift information at IJC < 17.3 can be divided into three classes. About one third of the galaxies are located near bright stars: these were masked in the 2dFGRS. Another third is located in centres of clusters: here the galaxy density is very high and the 2dF fibres were too few to take these into account. The last third include some close pairs of galaxies and other galaxies which were missed for unknown reasons. Whether these galaxies are included in a particular sample is discussed below:

1. 2D Sample 1: 2D density sample This sample has been used in Section 2.4.3 to derive the density map shown in Figure 2.9. It consists of all those 2022 galaxies with IJC < 18.8 (equal to MI =- 42 chapter 2: Environmental Influences on the Evolution of Galaxies

Figure 2.11: Comparison distributions of four parameters with literature data. The black (dotted) histograms corresponds to our data; red (dashed) histograms to Zamojski et al. (2007), blue (dashed-dotted) histograms to Conselice et al. (2000) and green (solid) histograms to Lotz et al. (2004). Except for asymmetry, our data correlates best with Zamojski’s. The differences with the other authors come from definition and sample differences.

19.3 in I at z=0.11) and applying the colour criterion (Section 3) or are within our target redshift range z = 0.11 ± 0.02

2. 2D Sample 2: Morphological Parameters Sample This is a subsample of Sample 1 and contain 895 galaxies with IJC < 17.8. The magnitude limit used here corresponds to the limit where visible classifications are still possible and where reliable values for the morphological quantities from Section 2.4.4 can be calculated. The reliability of the calculated parameters is assessed in Figure 2.11, where the distribution of our calculated values of this sample for four morphological parameters (Concentration, M20, Gini and Asym- metry) is compared to distributions from other ground-based data sets (Conselice et al. 2000; Lotz et al. 2004) and ACS data of galaxies at ≈ z = 0.7 (Zamojski et al. 2007). Using only non-inclined (< 60 degrees) galaxies below IJC < 17.8 with redshifts between z = 0.11 ± 0.02, we see that our distributions, except for Asymmetry, are best matched with Zamojski’s data. For the other two data-sets, Conselice and Lotz, there are offsets. These offsets are likely explained by using different definitions to calculate the parameters. For example, Lotz uses a circular aperture to calculate the Gini coefficient. For Gini, this can make a large differ- ence: more elongated objects will include areas with much lower intensity valued pixels resulting in a lower value for Gini. Changing our aperture to circular shape, results in a similar distribution as Lotz. The huge difference in Asymmetry with Zamojski is likely explained as a resolution difference between these data sets.

3. 2D Sample 3: Highly redshift-complete sample. This sample consists of all 570 galaxies below IJC < 17.3 (corresponding to MI =- 20.8 at z=0.11) and is redshift complete, at a few percent level, for densities above 2.5 times the background density. This density occurs at several times 2.4: Data Analysis 43

the virial radius of our clusters. 130 of the galaxies below this magnitude do not have redshifts available. Out of these 130, 49 galaxies are located in high density regions, i.e., > 5 times background (summing galaxies with IJC < 17.3 in column 8 of Table 2.2). Comparing the surface density (see 2.4.3) assigned to each galaxy with the constant value of the galaxy background surface density, we expect that about 5 galaxies of these 49 will lie outside our target redshift range. This is a very small number compared to all galaxies within the target redshift range in the high density regime (218 galaxies). For the 81 galaxies lying in the low density regime, we estimated probabilities that a galaxy belongs to our target redshift range or not: 13 galaxies lie in fore- ground z < 0.09 groups or background z > 0.13 groups and are removed from the list. 47 other galaxies lie at regions bordering the target clusters which still have significant overdensities: 2.5 to 5 times the background density. Again, taking into account the background density, we expect about 10 out of these 47 to lie in the background. For the remaining 21 galaxies, lying in < 2.5 times background density region (i.e. ≈ 10 galaxies per Mpc−2), 14 galaxies are expected to be back- or foreground galaxies. With such a large ’membership’ uncertainty, these 21 galaxies are not used in the analysis using this sample, which results in less completeness for regions with <2.5 times the background . 4. 3D Sample. Based on only those galaxies with redshift information (602), another sample can be defined: the 3D sample. 3D densities were already calculated in Section 2.4.3. These surface densities based on velocities will separate foreground galaxies and background galaxies from cluster galaxies within the 0.09 and 0.13 redshift inter- val. The large majority of these galaxies belong to the 2dFGRS sample, which is highly complete until B = 19.0 (Colless et al. 2001). Therefore, the sample is biased towards bluer objects which have a higher probability belonging to the field (see also Figure 2.8). The 2dFGRS galaxies also contains a star forming parameter (Madgwick et al. 2002) based on the averaged emission and absorption line strength in the galaxy rest frame spectrum. For this sample, we also calculate galaxy masses according to: Mass has been determined from the standard model of Bell & de Jong (2001), using the equation:

 M  log = −1.204 + 1.347(V − I) (2.10) LI

,with M the galaxy mass in M ; LI the absolute I band luminosity in L . K- corrections were ignored because these are very small in the V and I bands for various types of galaxies at our target redshifts (Fukugita et al. 1995) and will not influence comparison of galaxy masses, which we will do in Section 2.5.4.

2.4.6 Classification

Classification was applied on all ’2D Sample 2’ galaxies (i.e. IJC < 17.8). Based on inclination, eyeball classification and Sérsic index, all galaxies were distributed among four different classes. First, all highly inclined galaxies (i > 60 degrees) were put into the ’edge-on’ class. The objects belong to this class are likely disk dominated galaxies, 44 chapter 2: Environmental Influences on the Evolution of Galaxies

All Target Redshifts Nearby Far Away Unknown Sér. E Sér. D Sér. E Sér. D Sér. E Sér. D Sér. E Sér. D Sér. E Sér. D Eyeball E 276 269 150 100 17 19 35 23 74 127 Eyeball D 23 273 10 116 4 26 4 81 5 50 Edge-on 320 161 52 31 76

Table 2.3: Classification statistics for eyeball and Sérsic classification for Sample 2 (IJC < 17.8) galaxies. Spirals have Eyeball D, ’S0s’ are listed in this Table as objects with Eyeball E and Sérsic D., while ellipticals have Eyeball E and Sérsic E. The ’edge- on’ class is a blend of mainly S0s and disks probably mixed with a few ellipticals. Comparison of the numbers of Eyeball D + Sérsic E and Eyeball D + Sérsic D shows that Sérsic recovers 90% of the spirals as labeled by eye.

e.g. S0s or spiral galaxies. Their calculated morphological parameters are often highly disturbed because of dust and other projection effects, and classifying a galaxy into an S0 or a spiral galaxy using visual means or a Sérsic index is a real challenge here.

For the remaining galaxies, eyeball classification was applied on thumbnail images in the I-band. All galaxies with spiral, ring-like or irregular structure were put in a separate class called the ’spiral’ class. The remaining featureless galaxies were further separated into two classes, based on their Sérsic index. The Sérsic (see Section 2.4.4) index is a reliable indicator to discriminate between disk dominated galaxies (exponential profile, n=1) and elliptical galaxies (n=4). Within the class of featureless galaxies, we define disk-dominated galaxies as those galaxies with Sérsic index <= 2.5, resulting in the ’S0’ class. These likely correspond to the featureless disk dominated S0s or may be low surface brightness disk galaxies. Because of our fairly bright magnitude limit IJC < 17.8 ≈ MI = -20.3, no faint dwarf E galaxies, which also have low Sérsic index, are expected to be included in this class. All remaining objects with Sérsic index > 2.5 were put in the ’elliptical’ class. Early type galaxies were defined as the combination of the ’S0’ and ’elliptical’ class.

The reliability of the Sérsic index to detect spiral (and thus disk dominated) systems can be derived from Table 2.3. This Table lists the number counts of galaxies classified by eye and Sérsic for Sample 2 galaxies, far away galaxies, nearby galaxies and galaxies with unknown redshifts. The combination of 2 classification results in 3 galaxy classes. Spirals are all galaxies as classified by eye; ’S0s’ are listed in this Table as objects with Eyeball E and Sérsic D., while ellipticals have Eyeball E and Sérsic E. The ’edge-on’ class is a blend of mainly S0s and disks probably mixed with a few ellipticals. The numbers of S0s, Ellipticals and spirals in this Table are comparable with Figure 4 of Dressler (1980) for densities higher than 30 galaxies per Mpc−2 [at (log(ρ)=1.5]. For our target redshifts, we find that 116 spiral galaxies are recovered by Sérsic, while 10 spirals are not recovered. This means that about 90% of eyeball-classified spiral galaxies are recovered with the Sérsic method (see Table 2.3). Sérsic spiral recovery is still doing fine for nearby galaxies (z < 0.09), despite the expected larger angular size of spiral bulges, which will disturb our automatic Sérsic index calculation (Section 2.4.4). 2.4: Data Analysis 45

Figure 2.12: Plots showing separation of our galaxy classes using various parameters (parameter definitions: see Section 2.4.4). Red open circles represent the ’elliptical class’, blue asterisks are ’spiral’ class objects, while green triangles belong to the ’S0’ class (see Section 2.4.6). Left: colour-magnitude diagram. Right Asymmetry vs. 2dF parameter η.

Figure 2.13: Plots showing separation of our galaxy classes using various parameters (parameter definitions: see Section 2.4.4). Symbols similar to Figure 2.12. Left: V-I Colour gradient vs. average surface brightness in I within R50. Right: Effective radius R50 (kpc) vs. concentration. 46 chapter 2: Environmental Influences on the Evolution of Galaxies 2.5 Results

In this Section, the (average) properties of our three types of galaxies as a function of density are presented. We also present results for smaller subsets of galaxies. We begin with an assessment of our classification scheme:

2.5.1 Classification vs. parameters More insight in our classification scheme can be gained by looking at the plots in Figure 2.12 and 2.13, which show the best separations between the ’elliptical’, ’S0’ and ’spiral’ classes in several parameter spaces. Figure 2.12 shows on the left side a colour-magnitude diagram. Although our classification is purely based on morphological information, a good separation between spirals and ellipticals is visible in this plot. The majority of ellipticals belong to the red sequence (see Bower et al. 1992). The ’S0’ class seems to be much redder on average than the spirals and bluer than ellipticals, as expected for S0s (Sandage & Visvanathan 1978). The luminosities of the ’S0’ class objects seem similar to disk like objects. The right part of Figure 2.12 shows the 2dF star formation parameter η vs. asym- metry. The combination of both parameters gives a good separation between spirals and ellipticals plus S0s. This is no surprise since both parameters trace star formation (Conselice et al. 2000; Madgwick et al. 2002). Asymmetry is related to our eyeball classification, which separated galaxies containing spiral features from smooth look- ing objects. η alone gives a good separation, with spirals and ellipticals separating at ≈ η < −2. The S0s mostly resemble the elliptical class for both parameters, although a relative larger proportion of S0s with respect to ellipticals are asymmetric and lie in the spiral space of this plot. On the left side of Figure 2.13, the colour gradient vs. the average surface brightness is plotted. Colour gradients for ellipticals are centred around ∆(V-I)=-0.15. These values are similar to results found by Ferreras et al. (2005). Colour gradients for spirals in our Figure have on average much lower values. This due to a dust gradient in the inner parts of spiral galaxies in combination with an age and metallicity gradient in the disk (Peletier et al. 1995; de Jong 1996b; Holwerda et al. 2005). The S0 colour gradients seem to lie somewhere in between. µ50 discriminates between ellipticals and spirals, with the spirals µ50 being much fainter, in agreement with e.g. de Jong (1996a). The right panel shows concentration C vs. the effective radius R50. Concentration alone is not a efficient separator for the various galaxy classes in our dataset. This is also visible in the work of Ferreras et al. (2005); Conselice et al. (2008). R50 does a better job: R50 separates spirals from S0s reasonably well at about 4 kpc. Six cD galaxies stand out at large R50 and higher concentrations.

2.5.2 Morphology Density Relation Table 2.4 presents the well-known morphology-density relation (Dressler 1980) for the ’3D Sample’ and ’2D Sample 3’ in the four different density regimes defined in Sec- tion 2.4.3 and 2.4.3. We list only the early type fraction (’elliptical’ plus ’S0’ class) and do not include the ’edge-on’ class. The latter class also contains about 50% S0s, therefore, the real early type fractions are estimated to deviate by a few percent, by not including these S0s. The morphology-density relation is clearly discernible and not 2.5: Results 47

Sample Regime 1 Regime 2 Regime 3 Regime 4 (1) (2) (3) (4) (5)

2D3 0.60 ±0.05 0.71 ±0.04 0.65 ±0.05 0.82 ±0.04 Dressler 0.53 ±0.02 0.63 ±0.02 0.71 ±0.02 0.80 ±0.04 3D 0.58 ±0.06 0.69 ±0.05 0.75 ±0.05 0.78 ±0.05 Goto 0.60 ±0.02 0.66 ±0.03 0.70 ±0.04 0.78 ±0.06 3D blue ell 0.22 ±0.07 0.11 ±0.05 0.08 ±0.04 0.03 ±0.03 3D red sp 0.12 ±0.06 0.30 ±0.08 0.53 ±0.11 0.71 ±0.12 3D blue s0 0.36 ±0.13 0.20 ±0.07 0.05 ±0.05 0.21 ±0.11

Table 2.4: Fractions of early type galaxies (’Elliptical’ plus ’S0’ class) in four different density regimes (see Sections 2.4.3 and 2.4.3) for the 2D3 and 3D sample, compared with Dressler (1980) and Goto et al. (2003). Lower part: fractions of three colour selected subsamples as a function of density. much difference is seen between the two sample definitions. The gradient of the relation is similar in both relations. For the ’3D Sample’, there is good agreement with Goto et al. (2003) (our Table 2.4 uses information from their Fig. 11). They use similar projected densities with velocity information, apply a slightly fainter magnitude limit (MR = -20.6) compared with our MI ≈ -20.8 and use somewhat more nearby galaxies between 0.05 < z < 0.10. For the ’2D Sample 3’, we compare with Dressler (1980), although their limiting magnitude MV = −19.7 is comparable to ours, we find small offsets.

2.5.3 Parameters as a function of environment Figure 2.14 shows the averages of 12 parameters for the ’elliptical’, ’S0’ and ’spiral’ classes from Sample 3 in the four different density regimes defined for this sample in Section 2.4.3. The errorbars represent the dispersion divided by the square root of the number of galaxies (statistical error) used for that particular data point. Strong variations with environment are visible for the parameters which are probably most related to star formation: η, V-I, and asymmetry. The 2dF parameter η is directly related to star formation. Values of η lower than -2 indicate no star formation at all, while larger values indicate increasing star formation. The dividing line between early and late type galaxies lies somewhere at η=-1.4 (Porter et al. 2008). Therefore, the plot shows that some S0s and E-type galaxies may still have some star formation in lower density areas. η drops quickly for these types of galaxies in the highest density areas. Star formation ceases continuously for spirals with increasing density, in agreement with Gómez et al. (2003). The colour V-I changes for all types of galaxies, but is most conspicuous for spirals, where it is more than 0.1 redder in the two highest density regimes (i.e. > 1.5Rvir). The colours of S0s and ellipticals change slowly with increasing density. A similar behaviour is shown by R50. It drops for spirals only in the two highest density regimes. Asymmetry changes significantly only for spirals in all density regimes: spirals become continuously more symmetric with increasing density. Asymmetry measures merging as well as star formation (Conselice et al. 2000), which implies that our results 48 chapter 2: Environmental Influences on the Evolution of Galaxies

Figure 2.14: Averaged parameters for different types of galaxy classes for the 3D sample: solid line: ellipticals, dashed line: spirals, dotted line: S0s. The 3D projected density regimes: 1 = 0-2 times the background density (’field’); 2: 2-5 times back- ground (intermediate region); 3: 5-10 times background (cluster region) ; 4: > 10 times background (cluster centres). Parameters plotted: from left to right, Top Row: ef- fective surface brightness µ50, Sérsic, D23, star formation parameter η; Middle Row: Asymmetry, colour V-I, effective radius R50, colour gradient ∆(V − I); Bottom Row: Concentration, M20, Gini, absolute magnitude. 2.5: Results 49

Ellipticals Spirals S0s Param. ell err blue ell. err. spirals err. red spiral err. s0s err. blue s0s err. (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) C 3.26 0.04 3.03 0.10 2.85 0.05 3.10 0.07 3.00 0.04 2.79 0.09 M20 -1.80 0.02 -1.67 0.05 -1.79 0.02 -1.78 0.03 -1.71 0.03 -1.56 0.07 Gini 0.55 0.00 0.53 0.01 0.46 0.01 0.50 0.01 0.52 0.01 0.49 0.01 A 0.03 0.00 0.04 0.01 0.07 0.01 0.04 0.00 0.02 0.00 0.03 0.00 R50 (kpc) 4.17 0.18 3.58 0.25 5.70 0.19 4.72 0.30 3.63 0.14 3.45 0.23 ∆(V − I) -0.18 0.02 -0.24 0.10 -0.51 0.03 -0.34 0.04 -0.23 0.03 -0.42 0.09 µ50 18.86 0.04 18.93 0.18 19.84 0.07 19.31 0.10 19.03 0.06 19.25 0.14 I -21.98 0.06 -21.65 0.15 -21.59 0.07 -21.58 0.1 -21.49 0.07 -21.29 0.09 D23 (kpc) 13.04 0.48 10.80 0.86 11.92 0.40 11.90 0.62 10.10 0.42 8.78 0.50 η -2.82 0.07 -1.09 0.65 0.27 0.28 -1.99 0.18 -2.71 0.12 -0.33 0.66

Table 2.5: Averaged properties + errors of normal ellipticals and blue ellipticals (left side), normal spirals and red spirals (middle) and normal and blue S0s. indicate that these processes are less effective at higher densities. The lower right panel shows averaged luminosities. We see that the ellipticals are the brightest galaxy class at all density regimes, and become brighter in the two highest density regime. On average, S0s seem to be a little fainter than spirals, contrary to results in the K-band found by Burstein et al. (2005). Note also that no extinction correction was used for spirals, which would imply even brighter I-band luminosities for spirals. Although, it is expected that the extinction in I, is low (Giovanelli et al. 1995). Concentration-like parameters: Gini, M20 and concentration are observed not to vary significantly with environment for the three galaxy classes. Sérsic does vary for ellipticals, but this is likely an arbitrary effect due the lack of constraints on n during the Sérsic fitting process.

2.5.4 Colour Selected subsamples: red spirals and blue ellipti- cals

The previous Section showed that star forming parameters are sensitive to density. In this Section, we look at three colour selected subsamples: red spirals, blue ellipticals and blue S0s.

• Environmental differences Figure 2.15 shows in the top panel CM diagrams for the four density regimes defined for Sample 3D. Red sequence galaxies (see Section 3) are defined to be the points plotted between the solid lines. Red spirals are defined as spirals belonging to the red sequence, while blue elliptical galaxies and S0s are bluer than their red sequence counterparts. In the bottom panel, we show colour-mass diagrams for the same environments. Qualitatively, it can be seen in Fig. 2.15 that red spirals and blue ellipticals are more ubiquitous in high and low density areas respectively. Quantitatively, this is shown in the lower part of Table 2.4, which lists the fraction of blue spirals, ellipticals and S0s. Blue ellipticals seem to prefer low density areas while red spirals dominate the spiral class in the densest cluster areas. Blue S0s also prefer low density areas, but in the cores of clusters, i.e. in the highest density regime (Rvir < 0.7), we find somewhat more blue S0s. 50 chapter 2: Environmental Influences on the Evolution of Galaxies

-24 -23 -22 -21

12.5 12 11.5 11 10.5

Figure 2.15: Colour-magnitude plots (Top) and Colour-mass plots (Bottom) for four different environments. Red crosses are elliptical galaxies, blue circles are spirals and green triangles are S0s. Surface densities in galaxies per Mpc−2 per 1000 km/s. 2.6: Discussion 51

• Physical differences Table 2.5 shows averaged values for 10 parameters for red spirals, blue ellipticals and blue S0s vs. ’normal’ spirals, ellipticals and S0s. Qualitatively, except for R50 for spirals, differences between parameters for red and blue subclasses are similar. Blue ellipticals have, as expected, some star formation, as shown by η. The differences in R50 and D23 between these elliptical subsamples are likely explained by small I-band luminosity differences. Differences between red and normal spirals are more pronounced. The largest differences occur in asymmetry, R50, ∆(V − I), µ50 and η. While having similar D23 (size) and luminosities I, red spirals seem to be more concentrated with their smaller R50, which means that their bulge to disk ratio is probably higher. Their lower star forming rate is likely reflected in lower asymmetries. Increased ∆(V − I) for red spirals with respect to normal spirals is likely due to relatively more dust present in the disk with less active star formation (Wolf et al. 2008) resulting in lower age differences along the disk (de Jong 1996b). Finally, for the S0 subsamples, there are significant differences for η and ∆(V − I).

2.6 Discussion

Here, we will discuss our results with respect to what is already known from the lit- erature. Topics which are discussed: environmental dependencies of star formation, global changes of morphological properties of galaxies and characteristics of red spi- rals and their relation to S0 galaxies. The environment is characterised by the density parameter.

2.6.1 Suppression of star formation In this Section, we discuss our results in terms of star formation in different environ- ments. It has already long been known that higher density regions show fewer star forming galaxies (Dressler 1980) and suppression of star formation (Gómez et al. 2003). Figure 2.16 and Figure 2.17, zoom-ins of Fig 2.14, shows averages of η and Asymmetry (left and right panel in Figure 2.16) and µ50 and colour V-I (left and right panel Fig- ure 2.17)for different types of galaxy classes as a function of environment. For spirals, we see a very strong dependency of the star formation parameter η (Left) on the local environment. Remarkably, this dependency is not only significant when comparing re- gions inside clusters with those outside, but it also pertains both within clusters and outside clusters with a strong dependency of the star formation rate in all these density regimes on the local galaxy density. A first notion of this kind was presented by Gómez et al. (2003); our results confirm this with a strong ’signal’ over a large range of galaxy densities. Star formation suppression in lower-density areas may be caused by physical processes like galaxy harassment for higher mass galaxies (Moore et al. 1999) or slow interactions in groups (pre-processing Zabludoff & Mulchaey 1998, 2000; Wilman et al. 2008). Additional star formation suppression processes most likely occur at density regimes 3 and 4. For these regimes, the average star formation η and its statistical error becomes much lower for S0s and ellipticals, while the average V-I colour for spiral galaxies becomes suddenly significantly redder. This break occurs at slightly different parameters compared to that found by Gómez et al. (2003): our break occurs at 1.5 Rvir 52 chapter 2: Environmental Influences on the Evolution of Galaxies

Figure 2.16: Zoom-in of Figure 2.14 for parameters η and asymmetry.

Figure 2.17: Zoom-in of Figure 2.14 for parameters µ50 and V-I.

−2 and density > 2 Mpc while they find a break more outwards at 3.5 Rvir and density −2 1 Mpc . The absolute magnitude limits are similar (MR=-20.4 vs. our MI =-20.8) but our sample lies at roughly 1.5 times larger distances. The break is likely due to gas depletion: our S0s and ellipticals show a minimal value for η here, which probably means that no gas is present or infalling anymore onto these galaxies. The most likely processes responsible for gas depletion in these high density areas at our redshift are starvation (Larson et al. 1980; Bekki et al. 2002) and ram-pressure stripping (Gunn & Gott 1972). Gas depletion in high density regions is further supported by looking at the change in asymmetry for spirals, which continuously decreases with increasing density. Large asymmetries of spirals are proven to be related 2.6: Discussion 53

Figure 2.18: Deviations of colour as a function of environment. Left: µ50 vs. V-I colour for 3D sample spirals. The straight line is a least squares fit to the data. Symbols are represent spirals located in different densities: going towards higher density areas these are triangles, squares, circles and asterisks. Right: residuals with respect to the fit vs. 3D density. Points with errorbars are averages with statistical errors. A slight trend towards redder spirals for a given Hubble type in denser regions is visible. to by gas accretion from filaments (Richter & Sancisi 1994; Zaritsky & Rix 1997). Evidence of significant gas infall onto field galaxies is numerous (see Sancisi et al. (2008) and references therein). Although asymmetry could also be caused by mergers, this is probably not reflected in our data, as we did not include the few merger systems in our galaxy sample. We conclude that gas infall onto galaxies is largely stopped in cluster regions at radii smaller than 1.5 Rvir. For the spiral galaxies we also detect a significant trend in the surface brightness at the effective radius µ50 with the local galaxy density: high density regions contain on average higher surface brightness spirals. This is an interesting result, however, it reflects to some extend the morphology-density relation, with in dense regions a higher fraction of early type spiral galaxies with higher surface brightness, redder colours and less asymmetry. In order to assess how much of the observed trends are caused by the morphology density relation and how much is caused by other effects we make use of the detailed surface brightness photometry obtained in this work. Figure 2.18, Left, shows the V-I of spiral galaxies of the 3D sample versus the µ50. The progression, using least squares fitting, in the plot corresponds to early type spirals (Hubble type Sa, type=1, de Vaucouleurs et al. 1991) with relatively red colours and bright µ50 to late type spirals (Hubble type Sd, type=8) with relatively blue colours and fainter µ50 (e.g. Valentijn 1990). We use this relation to characterise spiral Hubble subtype, which we actually could not determine visually. In Figure 2.18, Right panel, we plot the residuals of (V-I) with respect to our fitted relation vs. density where for each galaxy. The residuals (V-I) were computed by:

residual(V − I) = observed(V − I) − predicted(V − I) (2.11) 54 chapter 2: Environmental Influences on the Evolution of Galaxies where predicted (V-I) comes from the relation plotted in the Left panel of Fig 2.18.

In the right panel, we plot the averaged offsets with statistical errors in V-I with respect to the fitted colour-surface brightness relation for each of the 4 3D density regimes. The two highest density regimes show an increased red colour with respect to the colour-surface brightness relation, a result already found by Kennicutt (1983), but here we are more precise by specifying the exact locations. The averaged ’normalized’ V-I colours show a nearly identical increase (0.13 mag instead of 0.17 mag) with density as the ’original ’ V-I - density relation (right panel of Figure 2.17). The slightly reduced amplitude reflects the expected effect of the morphological density relation with rela- tively more early types in dense regions. We conclude that the V-I density relation at grosso mode is NOT a result of a different composition of morphological types at various environments but holds in fact for each sub-type. Thus for various subtypes, redder colours are observed in high density regions. This relation is next to the well-know morphology density relation and requires a physical interpretation. The left panel of Fig 2.18 shows that the increased red colours in dense regions is the strongest for the later Hubble types (asterisks).

2.6.2 Origin of S0s Can our findings contribute something to the problem of the origin of S0 galaxies? A large increase is seen in the S0 fraction in clusters since the last few Gyr (Dressler et al. 1997). To explain such a fast, recent S0 evolution, it is believed that spirals transform into S0s (e.g. Aragón-Salamanca et al. 2006). Several physical processes have been proposed to explain S0 transformation from spirals. S0s show no or little star formation. To arrive at such a state starting from a spiral, one necessary ingredient must be gas removal. Starvation (Larson et al. 1980; Bekki et al. 2002) and ram-pressure stripping (Gunn & Gott 1972) are possible candidates. The latter process is only felt in the very inner regions (≈< 1Rvir) of clusters (Treu et al. 2003), while in the outer regions, the much slower process of starvation is applicable. Ram-pressure stripping is also less effective for more massive spirals. This is demonstrated by Crowl & Kenney (2008), who show that gas stripping is only effective in the outer regions of massive spirals in the Virgo cluster at the virial radius. Assuming that the gas is removed and a S0 is formed somehow, the S0 galaxy will then slowly fade. This fading-scenario has been verified by Aragón-Salamanca et al. (2006). They also compared the globular cluster specific frequency (number of globular clusters per unit V-band luminosity) of S0s with those of spirals. They found a specific frequency for S0s which is a factor of 3 higher than for spirals. Assuming that no globular clusters are created during the S0 transformation process, this factor is consistent with the lower zero point of about 1,2 magnitude (≈ a factor 3) for the Tully-Fisher relation for S0s. However, many problems remain with a simple gas removal scenario. There are differences in the observational properties between spirals and S0s, which can not be explained after simply removing gas: S0s show on average similar or larger K band lumi- nosities than spirals (Burstein et al. 2005) and have larger bulges than spirals (Christlein & Zabludoff 2004). Counterarguments against this may be related to a down-sizing ef- fect: currently observed spirals in clusters have lower mass than currently observed S0s, which may have formed from higher mass spirals in the past (Kauffmann et al. 2004; Smith et al. 2005). Using a mass limited sample, van der Wel et al. (2007) even showed 2.6: Discussion 55 that for high mass early type galaxies, no evolution is found in the morphological density relation since z=0.8. Furthermore, S0s are also found in large numbers in the field (Goto et al. 2003; Wilman et al. 2008). Here, gas removal processes like starvation and ram-pressure strip- ping are not present. Therefore, other processes, probably related to interactions may be more important. Candidates are: merging and harassment for high mass galaxies (Moore et al. 1999; Christlein & Zabludoff 2004). These processes occur mainly out- side the cluster regions, necessary to explain the existence of S0s in low density areas. These processes will likely cause enhanced central star formation, probably supported by growth and destruction of large bars, which may in turn accelerate the starvation process and transformation towards S0s. Recent work of Wilman et al. (2008) shows that S0s are equally prevalent in groups and clusters. They also provide evidence that S0s are more prevalent at group peripheries. This implies that dense environments speed up S0 formation. If S0s did really evolve quickly from spirals in clusters in the recent past and if the cluster environment is the place where fast S0 evolution is happening, we should expect the averaged parameters of spirals and S0s to converge with increasing den- sity. Looking at Figure 2.14, such a convergence is indeed visible for all parameters. This is visible for: R50, D23, asymmetry, V-I and ∆(V − I). A lowered R50 for spi- rals, for equal luminosities, implies a larger bulge, precisely what is observed for S0s (Christlein & Zabludoff 2004). Lower asymmetries for spirals might point towards a process of smoothing the spirals disks, may be due to starvation ram-pressure stripping or harassment of higher-mass galaxies (Moore et al. 1999).

2.6.3 The red spirals If S0s are still forming at z=0.11, as suggested by Dressler et al. (1997), we should be able to identify intermediary products or transition objects. Several authors do propose galaxy populations as transition objects between spiral and S0 galaxies: blue flaming compact galaxies at z=0.4 (Braglia et al. 2007), spirals with enhanced central star formation (Bamford et al. 2007) and interacting galaxies within groups (Moss 2006). Another candidate class of galaxies are, within cluster regions, the so-called (passive) red spirals. We will discuss these galaxies in the context of S0 formation in this Section. The red spirals, already identified by van den Bergh (1976), Couch et al. (1998) and Poggianti et al. (1999), are in fact two different classes as discovered by Wolf et al. (2008) and Gallazzi et al. (2009). They used multi-wavelength data including U band and far-infrared Spitzer data centred on Abell 901/2 at redshift 0.16, covering all galaxies located within 1 virial radius of the cluster centres. They found one class of truly passive red spirals with (almost) no star formation and a class of obscured, dusty, star forming spirals within one virial radius of Abell 901/2. The latter class of red spirals form stars at a rate at about a quarter of blue spirals. Unfortunately, to detect these star-forming red spirals, far-infrared data are requiered. Bamford et al. (2008), using a sample of ≈ 130000 galaxies and covering lower density regions, find a progressive reddening starting from very low density areas. Our results agree with this. The increase in V-I colour or reddening of spiral galaxies, mentioned in Section 2.6.1, is reflected in the lower part of Table 2.4, which shows a strong increase of the red spiral fraction among spirals towards higher density areas. This increasing fraction is also directly visible in Figure 2.15. Since both Table 2.4 and Fig. 2.15 do not include the 56 chapter 2: Environmental Influences on the Evolution of Galaxies edge-on class, this implies that the reddening in these spirals is not due to inclination effects, which is confirmed by recent work on Abell 901/902 (Wolf et al. 2008). If red spirals are intermediate between spirals and S0s, their morphological trans- formation is likely a slow process for larger spirals, taking one or several Gyr (Wolf et al. 2008). In order to see whether red spirals may be intermediate forms between S0s and normal spirals, we compare our calculated averaged properties of spirals, red spirals and S0s. It is expected that the averaged properties of red spirals are in between normal spirals and S0s. Columns 6, 8 and 10 of Table 2.5 show averaged physical prop- erties for normal spirals, red spirals and normal (red) S0s respectively (the blue S0s, column 12 are found mainly in the field). Looking at the properties, we see that the red spirals are indeed in between the normal spirals and S0s, supporting the scenario in which red spirals are intermediary products. Holden et al. (2007) and van der Wel et al. (2007) note that in order to compare morphological classes, this is better to be done with similar masses than with similar luminosities. Masses correlate with colour and luminosity (Bell & de Jong 2001). We compare the location of spirals, red spirals and S0s in a colour-mass diagram, which is shown in the lower part of Figure 2.15 and see that normal spirals, red spirals and S0s do have comparable masses in all density regimes, which is at least consistent with spirals and S0s being related to each other. If red spirals are really pre-S0 galaxies, the following scenario must be valid for cluster S0 formation. Here, first the gas is removed from spirals in a few Gyr after the spiral enters a cluster. Afterwards, the red spiral will slowly loose its star formation. Subsequent harassment of our (high mass) spirals by many fast small flyby galaxies will do the ’smoothing’ process to produce the S0 (Moore et al. 1999). Such a scenario, however, is at odds with S0 formation in groups: Moss (2006) and Wilman et al. (2008) find that only harassment and other interaction like processes are responsible for S0 formation. This is no problem however, if S0 progenitors in groups can retain their gas longer than cluster S0 progenitors, different properties between field S0s and cluster S0s are to be expected. Separate scenarios for S0 formation in low and high density areas were suggested by e.g. Quilis et al. (2000) and Poggianti et al. (2001a).

2.7 Summary

In this chapter, we have reduced and analyzed 16 wide field imaging pointings in three passbands, as part of a pilot project to test the astro-wise data reduction system. We analysed many properties of galaxies as a function of environment, characterised by density. Our relatively small number of target galaxies at z=0.11±0.02 allowed for visible classification of spirals. Using the Sérsic parameter, 90% of all visually classified spirals were also classified as spirals automatically. We see a significant reddening of spiral galaxies going towards the highest densities. This is accompanied with ever decreasing star formation rates going from empty field regions towards the cluster cores. The reddening of spiral galaxies occurs for all Hubble types, which increases fractions of red spirals towards the higher density areas. Red spirals may be an intermediate state between spirals and S0s, which is supported by their averaged morphological properties which are in between normal spirals and S0s. Gas removal processes like starvation and ram-pressure stripping are likely to be active only within the innner 1.5 virial radii of our clusters. In the future, larger wide field surveys with OmegaCam targeted at more nearby clusters will give more information about the real nature of red spirals, 2.7: Summary 57

S0s and other classes of galaxies. In the next Chapter we will study barred galaxies in our sample as a function of environment. 58 chapter 2: Environmental Influences on the Evolution of Galaxies Chapter 3 Bars in large scale structures at z=0.11 ± 0.02

ABSTRACT ∗ We study the environmental dependencies, i.e. occurrence and size, of strong ∗ ( > 0.4), large bars (semi-major axis abar > 3.6 kpc) in MI > MI + 0.8 mag. disk-dominated galaxies in a region of size 120 x 30 x 7.5 Mpc, at z=0.11, which contains six Abell clusters. We use a well-defined ground based dataset of 4 square degrees of I-band imaging, containing 173 disk-dominated galaxies located in large scale structures at z=0.11 ± 0.02. The bar detections are done using well known ellipse fitting techniques. Results: large, strong bars in spiral galaxies are slightly more common in higher density areas (0.31 ± 0.06) than in field regions (0.18 ± 0.06). High density regions also tend to contain larger bars, which is due to the larger dispersion of bar sizes in these regions compared to low-density regions. These environmental differences are naturally explained when cluster specific physical processes are taken into account, like ram-pressure stripping and harassment. Strong large bars in galaxies without any visible spiral structure (i.e. S0s) are far less common compared to bars in spiral galaxies (fraction 0.07 ± 0.03 vs. fraction 0.26 ± 0.04). Harassment, applied to our massive galaxies, likely explains this effect. Compared with higher-redshift bar fractions, we find no significant evidence for evolution of the bar fraction. Since galaxy fractions of lower mass S0s and spirals differ at different redshifts, care should be taken when comparing bar fractions using a combined sample containing lower mass disk-dominated galaxies.

3.1 Introduction

A majority of disk-dominated galaxies have central bars (Eskridge et al. 2000). Early simulations have shown that bars form spontaneously in dynamically-cold pure stellar disks (Hohl 1971; Kalnajs 1972; Sellwood 1981), but also after mergers and encounters (Byrd et al. 1986; Noguchi 1987; Gerin et al. 1990). Interactions can strengthen bars

∗ Authors: G.Sikkema, R.F. Peletier & E.A.Valentijn (in prep.) 60 chapter 3: Bars in large scale structures at z=0.11 ± 0.02 which are already present (Berentzen et al. 2004). The evolution or (in)stability of bars depends on the redistribution of angular momentum in the bar region: relative massive halos with respect to the disk seem to favour bar stability (Combes & Sanders 1981; Athanassoula & Misiriotis 2002; Athanassoula et al. 2005; Holley- Bockelmann et al. 2005). The observation of larger sized bars in earlier type disk galaxies supports this notion (Elmegreen & Elmegreen 1985; Erwin 2005). Destruction of bars is encouraged by central mass concentrations (CMCs) (Norman et al. 1996; Friedli & Benz 1993; Athanassoula et al. 2005). The availability of gas in the bar regions may also weaken or destroy bars (Bournaud & Combes 2002), even in the presence of a massive halo (Berentzen et al. 2007). On the other hand, external gas accretion from intergalactic filaments is a prerequisite for bar growth (Block et al. 2002; Combes 2007). Observational evidence for external gas accretion in (isolated) disk galaxies is numerous (Richter & Sancisi 1994; Zaritsky & Rix 1997; Varela et al. 2004; Sancisi et al. (2008) and references therein). Considering all these bar evolution processes, a bar-cycle scenario can be imagined, where several bar destruction and resurrection episodes can change the appearance of disk galaxies during their lifetime (Bournaud & Combes 2002; Combes 2007). Such a scenario implies that each subsequent bar will increase in size and will have a slower angular rotation speed (Combes 2007). This is supported by the presence of larger sized bars in earlier type disk galaxies (Elmegreen & Elmegreen 1985; Erwin 2005). The bar-cycle scenario may also explain the presence and properties of ’pseudo-bulges’, which seem to be related to bars (Kormendy & Kennicutt 2004). Matter redistribution caused by a bar might give an explanation for the presence of so-called truncations, a sudden downwards break in the exponential profile, seen in many disk galaxies (van der Kruit 1979; Pohlen & Trujillo 2006). Evolution of bar sizes and bar presence is still a matter of debate. Sheth et al. (2008) do show significant evolution in the last few Gyr. Jogee et al. (2004) and Barazza et al. (2008) do not find any evolutionary trend in the bar fraction. Comparing the properties of barred galaxies in different environments, like in clusters and in the field with their different galaxy densities, may give more important insight in the various formation and destruction processes of bars. Many disk galaxies near centres of clusters have seen their gas removed by processes like ram-pressure stripping (Gunn & Gott 1972) or starvation (Larson et al. 1980; Bekki et al. 2002), which might change bar occurrences and properties. Interactions in higher density areas may encourage bar formation. On the other hand, the higher density of galaxies in clusters will heat up disks through harassment (Moore et al. 1996), possibly destroying bars in time. Probably, an interplay of processes will occur, but which process is most important is unclear at the moment. Currently, there is even no consensus about bar occurrence in different environments. Thompson (1981) and Andersen (1996), found an increasing bar fraction towards the central regions of the Coma and Virgo cluster. These studies are at odds with a more recent study of van den Bergh (2002), who did not find any correlation (field 25 ± 3% vs cluster 28 ± 3%). Another study found somewhat larger bars in early-type disks in Virgo, however, not at high significance (Erwin 2005). Finally, the presence of bars in S0s may give important clues about the origin of S0s. If these S0s have evolved from spirals, bar frequencies in S0s may give important insight into their past. In this chapter, we analyse the occurrence and properties of large bars (semi-major axis abar > 3.6 kpc) in disk-dominated galaxies in a large scale structure (120 x 30 x 7.5 Mpc) at z=0.11, containing six Abell clusters. The properties of the disk galaxies in 3.2: Data Analysis 61 this area were already presented in the previous chapter. We determine occurrence, size and strength of large bars in this region and analyse correlations with galaxy density and other galactic properties. Section 2 discusses the data and the techniques used to find bars. The results are presented in Section 3. Section 4 discusses the implications of our findings. Section 5 summarises all our results and gives some future prospects.

3.2 Data Analysis

The data used throughout this chapter are described in detail in Chapter 2. Here, a brief summary is given. Using the Wide Field Imager on the 2.2m telescope at La Silla, a field of view comprising four square degrees, was observed in the I band near the South Galactic Pole, covering the Abell clusters 2798, 2801, 2804, 2811, 2814, 2829 and all regions in between at z ≈ 0.11 ± 0.02. The observations have a stable seeing, with a PSF of about 0.8 arcsec and reach a surface brightness limit of µ ≈ 25 mag arcsec2. Chapter 2 presents several galaxy-subsamples, depending on magnitude limit, density and availability of redshift data. In this chapter, only the disk-dominated galaxies within the 3D sample are used. The 3D sample is defined (see Chapter 2) as all galaxies below I=17.3 (corresponding to MI = −20.8 at z=0.11) which have redshift data available. The redshift data ensure a reliable local galaxy density estimate (see Chapter 2). The disk-dominated galaxies consist of all galaxies with inclination lower than 60 degrees and are divided into two subclasses. The ’spiral class’: 97 galaxies, which have (spiral) structure as seen by the eye; 91% of these galaxies have a Sérsic index < 2.5. The second class, the ’S0 class’ with 76 galaxies within the 3D sample, consists of all galaxies without spiral structure but with a disk component with Sérsic index < 2.5. For all galaxies used throughout this chapter, many galactic structural parameters were calculated in Chapter 2. These parameters were derived from thumbnail images, extracted from the wide field images. In this Chapter we will use the following parameters: global colour V-I, colour gradient ∆(V-I), D23 (semi-major axis where the galaxy reaches µ = 23 mag arcsec−2) and star formation parameter η as provided by the 2dFGRS survey (Colless et al. 2001).

3.2.1 Detection of bars To detect the bars, the ellipse fitting package GALPHOT (Jørgensen et al. 1992) was applied to the thumbnail I-band images of the disk-dominated galaxies. Other sources present in the thumbnail images were masked using the SEGMENTATION image of SExtractor as described in Chapter 2. Detection of bars was done by analysing the results of a free ellipse fit. This method is used and described extensively in the literature (Wozniak et al. 1995; Peletier et al. 1999b; Jogee et al. 2004; Menéndez-Delmestre et al. 2007; Barazza et al. 2008; Sheth et al. 2008). Here we follow exactly the method and definition of Jogee et al. to define and detect strong bars. They used the following criteria to detect bars: going outwards from the centre of the galaxy, the ellipticity should drop suddenly by at least 0.1 from its maximum. At the same time, the position angle should change by at least 10 degrees. The radius, at which this rapid change occurs, is defined as the bar radius. An example of a bar detection using these methods is shown in Figure 3.1. In addition, an additional criterion was used to select bars: we want to avoid spatial selection effects due to small variations in the shape of the 62 chapter 3: Bars in large scale structures at z=0.11 ± 0.02

Figure 3.1: Example of a bar detection using a free ellipse fit. The ellipticity (lower panel) has a maximum above 0.4 at radius of 2 arcsec. When the ellipticity drops from its maximum value, the position angle changes by at least 10 degrees.

PSF, observed from the ground: we only extract bars, which have a minor axis larger than 3 times the σPSF of about 0.38 arcsec which amounts to 1.14 arcsec. For strong bars, with ellipticities larger than 0.4, this translates to bars having abar larger than 3.6 kpc at z=0.11. We also measure the bar strength, which is defined as the maximum ellipticity of the bar. Our definition of the bar fraction among a galaxy sample, used throughout this paper, is then defined as:

f = #(bars in galaxy sample)/#(galaxies in sample) (3.1)

, with a bar being defined as having at least a maximum bar strength of 0.4 and a minimum semi-major axis of 3.6 kpc. 3.2: Data Analysis 63

Figure 3.2: Normalised bar radius (i.e. bar radius/D23 of galaxy) vs. density (left) and I band magnitude (right). Triangles and circles are spiral galaxies in low and high density areas respectively. Crosses and squares are S0 galaxies in low and high density areas respectively.

Figure 3.3: Normalised bar radius (i.e. bar radius/D23 of galaxy) vs. galaxy colour V-I (left) and V-I colour gradient (right). Description of data points, see Fig. 3.2 64 chapter 3: Bars in large scale structures at z=0.11 ± 0.02

Bars Type Low Density High Density (1) (2) (3) (4) (5) Spiral Galaxies 39 58 Barred 7 0.18 ± 0.06 18 0.31 ± 0.06 S0s 17 59 Barred 1 0.06 ±0.06 4 0.07 ±0.03 All 56 117 Barred 8 0.14 ± 0.05 22 0.19 ± 0.04

Table 3.1: Bar fractions in spiral, S0 and all galaxies in the field and in high density regions.

Figure 3.4: Bar strength vs.2dF star forming indicator ν (left) and density (right). Description of points see caption Figure 3.2

3.3 Results

Using the 3D galaxy sample, four density regimes were defined in Chapter 2, based on redshift information. These (projected) densities were calculated by summing the 5th nearest neighbour projected densities in cylinders with depth 2000 km/s and radius 2 Mpc. Because of the low numbers of disk galaxies and S0s, we narrow the four density regimes down to two density areas by combining density regimes 1 and 2 (these are ’field’ and cluster transition areas (> several virial radii) respectively) to form the ’low density’ area, while density regimes 3 and 4 are combined to form the ’high density’ area (i.e. regions with densities between zero and a few virial radii).

3.3.1 Bars The upper part of Table 1 shows a summary of the statistics for barred spiral galaxies in low and high density areas. It shows that bars in spiral galaxies appear to be more 3.4: Discussion 65 common in the high density areas (0.31 ± 0.06) than low density areas (0.18 ± 0.06). This is in agreement with a recent result of Barazza et al. (2009), who, however, do not mention error percentages. Our findings also agree with the results of Thompson (1981) and Andersen (1996), who find increasing overall bar fractions towards the Coma and Virgo cluster central regions, but is at odds a more recent study van den Bergh (2002), who did not find any correlation. Our results also imply that, overall, strong bars are far less common in S0 type galaxies (global S0 bar fraction 0.07 ± 0.03 vs. global spiral bar fraction 0.26 ± 0.04). This in agreement the work of Erwin (2005), who examined bars in spirals and S0’s. The overall bar fraction of all disks (S0s + spirals) is 0.18 ± 0.03 All errors were calculated using:

∆N = (f(1 − f)/N)0.5 (3.2)

with f the bar fraction and N the number of galaxies. In Fig 3.2 and Fig 3.3, we plot the normalised bar radius and bar strength versus various galactic parameters (see Chapter 2). The normalised bar radius is calculated by dividing the bar radius by the galaxy radius at the point where the I band surface 2 brightness reaches 23.0 magnitude per arcsec (D23 from Chapter 2). This gives a better measure for the bar radius relative to the size of the galaxy. In all Figures, triangles and circles are spiral galaxies in low and high density areas respectively, while crosses and squares are S0 galaxies in low and high density areas respectively. Figure 3.2, left, shows the normalised bar radius vs. the density. We calculated the averaged normalised ¯ bar sizes in low and high density areas to be rbar/D23 = 0.51 ± 0.15 for high density ¯ galaxies vs. rbar/D23 = 0.41 ± 0.08 for low-density galaxies. The difference between these averaged normalised bar sizes is not significant (similar results were found by Erwin 2005). However, the two times larger dispersion in normalised bar size for the high-density area, also found by Erwin (2005), may be important. This is caused by the fact that the seven largest normalised bars (3 in S0s and 4 in spirals) all lie in high density areas (we note that lower values for normalised bar radii are incomplete due to bar detection sensitivity). A larger dispersion may be due to different galaxy populations traced in the low and high density areas. Therefore, we plot in the right panel of Figure 3.2 the host-galaxy’s magnitude vs. normalised bar size. A weak correlation is visible, but more important: we see that most large normalised bars reside in fainter galaxies, with reddish colours, which is derived from Figure 3.3, left panel. This panel shows the normalised bar radius vs the V-I colour of the galaxy. On the right side of Figure 3.3, we depict the normalised bar radius against the colour gradient of the galaxy. A weak correlation is visible with larger bars having lower absolute colour gradients. In Figure 3.4, we plot the the bar strength vs. η (left) and density (right). The distribution of the points in the left panel looks somewhat peculiar with the weakest and strongest bars having low η and intermediate-strength bars having high η. Weaker bars are somewhat more prevalent in higher density areas (right panel).

3.4 Discussion

. 66 chapter 3: Bars in large scale structures at z=0.11 ± 0.02

3.4.1 Bar frequencies in previous studies

In this Section, we compare our bar statistics with previous, evolutionary, bar studies to see how our results fit in the big picture. A proper comparison of bar frequencies among disk-dominated galaxies can only be made if similar observational criteria are used. These criteria depend on how spiral galaxies were defined, which passband was used to obtain the observations, which method was used to detect bars, which abso- lute magnitude limit was used to constrain the galaxy sample, the lookback time and which constraints on bar properties were applied to define the barred galaxy sample. Considering all these criteria, we use the studies of Jogee et al. (2004) and Sheth et al. (2008), since they use also ellipse fitting to detect bars. Jogee used the Sérsic index < 2.5 to select disk-dominated galaxies. If we combine our S0s and spiral class, our sample should be comparable to the Jogee et al. sample. (about 5% of the galaxies in our disk-dominated galaxy class have Sersic > 2.5 but this should not disturb the comparison much). An important difference is that Jogee et al. are sensitive to bars with abar larger than 1.2 kpc, compared to our sensitivity of 3.6 kpc . Our 3D sam- ple has an absolute magnitude limit of MI ≈ -20.8, which equals roughly MV =-19.8, assuming that our spirals have on average V-I=1.0 (see Chapter 2). This magnitude limit is slightly brighter than their faint magnitude limit of MV =-19.3. A difference is that Jogee et al. observe in the redshift range from z=0.25 to z=0.70, corresponding to rest-frame I to V. We observe in the I band at z=0.11. It is not expected that Jogee et al. will miss many large bars between V and I, since a study by Marinova & Jogee (2007) find similar large (abar > 3.6 kpc) bar fractions in passbands B and H. Cutting the bar size in Figure 2c of Jogee et al. to include only bars with abar > 3.6 kpc, we find that their expected bar fraction value for our bar definition is 0.14 ± 0.03. Our bar fraction is slightly higher, i.e. 0.18 ± 0.03, which is within the joint errors. We note that many of our barred galaxies reside in cluster regions. Figure 3.2 shows that there is a hint that in clusters tend to be larger in clusters than field galaxies, resulting in an increased bar fraction. Using only disk-dominated galaxies in our low density regions (lower right Table 3.1), we find a bar fraction of 0.14±0.05, which is very similar to Jogee et al.. Since the barred galaxies of Jogee et al. lie at 0.25 < z < 0.7, while our galaxies lie at z=0.11, this implies no evolution of the bar fraction in the last few Gyrs. These results contradict Sheth et al. (2008), who find a significant increase in the bar fraction going from z=0.84 to z=0.2. We also make a comparison with Sheth et al. (2008), although this will be less accurate. Since Sheth et al. do not use S0s in their sample, the comparison is best done using our spiral sample. They use an absolute magnitude limit, which is about 1.3 magnitude brighter in V. Like Jogee et al. they also define a strong bar sample with  > 0.4. An important difference with our sample is that they are sensitive to all bars with abar larger than 2 kpc, while we are only sensitive to bars with abar larger than 3.6 kpc. They find in their lowest redshift bin, at z=0.2, a strong bar fraction of 0.27 ± 0.05. To make a proper comparison, we apply the barred galaxy statistics of Jogee et al. (2004) and Fig. 10b of Marinova & Jogee (2007), who find that about 40% of barred galaxies with bar sizes larger than 2 kpc have bars with abar > 3.6kpc. Multiplying this percentage with the bar fraction of 0.27 of Sheth et al., we estimate a bar fraction for Sheth et al. of about 0.11 for galaxies with abar > 3.6 kpc. Our total spiral bar fraction is higher, namely 0.26 ± 0.04, but, again, this may be because we include cluster spirals. Our barred fraction of spirals in low density regions is 0.18 3.4: Discussion 67

± 0.06 (see Table 3.1), which is slightly higher, but not inconsistent, with our derived fraction from Sheth et al.. We conclude that we find no significant evolution in the bar fraction in low density regions between z=0.20 and z=0.10.

3.4.2 Environmental related bar frequency in spirals vs. bar formation theories

In Section 3.3 we found that more spirals in high-density cluster areas tend to contain strong bars. This is in agreement a recent results of Barazza et al. (2009) and with the findings of Thompson (1981) and Andersen (1996), who found an increasing bar fraction towards the Coma and Virgo cluster central regions respectively. High-density areas also have a larger spread in normalised bar sizes, which results in the larger normalised bars being found preferably in these areas. Larger normalised bars in the central regions of clusters are also reported by Barazza et al. (2009). How do these results fit in bar formation and destruction theories? Using the scheme of Treu et al. (2003, Fig. 10), we will discuss all important physical processes in chrono- logical order, i.e.: starting with the moment that a gas-rich (barred) galaxy enters a cluster for the first time. According to this scheme, merging and low velocity tidal in- teractions are effective in the field and groups. In clusters, this likely occurs most often in the very outer regions (> several virial radii). Gerin et al. (1990) and Berentzen et al. (2004) show that these physical processes tend to increase bar growth and for- mation. However, Berentzen et al. (2003) show that vertical interactions with the disk will destroys bars. The next process felt by the galaxy, is harassment, i.e. frequent high velocity galaxy- galaxy encounters (Moore et al. 1996). Harassment is very effective for lower luminosity (0.2 L∗) spiral galaxies. These low mass spirals, when entering the cluster, quickly form bars, which are then destroyed after one cluster orbit (≈ 1 Gyr), with the end product being a diffuse spheroidal . More massive (L∗) galaxies, with accompanying massive dark matter haloes, are stable to the chaos of cluster formation and tidal encounters. Their discs will heat up and spiral structure is gradually destroyed, while bulge growth is encouraged (Moore et al. 1999). Our luminosity limit lies at 0.5 ∗ ∗ ×L (combining MI =-21.6 (Dale et al. 1999) with our magnitude limit MI =-20.8 (see ∗ ∗ Chapter 2), this translates to MI +0.8 ∼ 0.5 ×L ). This means that our spiral galaxies survive longer, and that the strong bar fraction will probably have increased. The final important process felt by a spiral galaxy is ram-pressure stripping (Gunn & Gott 1972). Although this process is only felt within the densest cluster regions, within 0.5 virial radius of the cluster, it can strip a galaxy almost completely from its HI gas on a relatively short timescale, ≈ 100-200 Myr (see Verheijen 2004 and Koopmann & Kenney 2004 for some examples). The majority of galactic cluster orbits within 1 virial radius are radial (Ghigna et al. 1998), which shows that galaxies in this region are highly susceptible to ram-pressure stripping. Once the gas is stripped, it will not be replenished from external sources (starvation; Larson et al. 1980; Bekki et al. 2002). The absence of gas further prevents already existing bars to be destroyed and will break the bar-cycle scenario (Bournaud & Combes 2002). We conclude that all physical processes somehow increase the frequencies of (strong) bars in clusters and support our findings of a higher strong bar frequency for spirals in dense regions. 68 chapter 3: Bars in large scale structures at z=0.11 ± 0.02

3.4.3 Bars in S0s Our much lower bar frequency in the spiral-less disky S0 galaxies (see Table 3.1) is also visible in previous work: for Hubble types T = -3 or -2 (de Vaucouleurs et al. 1991), we see this effect in Erwin (2005, Fig. 8), Knapen et al. (2000, Fig.1) and Menéndez- Delmestre et al. (2007, Fig. 5a). Here we try to find possible (physical) explanations for the lower prevalence of bars in S0s. Gadotti (2008) showed that this may have to do with the detection technique: barred galaxies with luminous axisymmetric central components will show artificially weak bars after fitting ellipses. However, this artificial effect would likely be similarly valid for our early-type spirals with Hubble T-types between -2 and 0. Testing this with detailed image component decomposition techniques should be able to answer this problem. If the lower prevalence of strong bars in S0s is not due to failing detection methods, it may give some clues about the origin of S0s. Dressler et al. (1997) found a significant increase in the S0 fraction in clusters since the last few Gyrs. Passive, red, spirals in clusters (Couch et al. 1998) may be an intermediate stage in the transformation process. These galaxies may have passed through the centre of the cluster at least one time (taking about 1Gyr: Ghigna et al. 1998), with ram-pressure stripping and starvation exhausting them from their gas (see previous Section). The subsequent slow (several Gyr) process of harassment may transform such large, passive, red spirals into the featureless disky S0s (Moore et al. 1999). Harassment of larger spiral galaxies will heat up the disk and remove instabilities in S0s which favour bar formation (Moore et al. 1999). Observations of S0s do indeed show dynamically hotter disks (Vega Beltrán et al. 2001) and also have larger and more centrally concentrated bulges compared to spirals (Andredakis et al. 1995). Figure 3.4 shows that bars in spirals in high density areas are slightly weaker than in the field. We conclude that harassment may be a good candidate to lower the bar fraction in S0s, if these are the remains of spirals. An additional evolutionary selection effect may also be related to a lower bar fraction in S0s. Although controversial, Sheth et al. (2008) find a much lower bar frequency among higher redshift spiral galaxies . If S0s are the remains of these past spirals with their lower bar fractions, then we would find fewer bars in the S0s currently present in clusters. If S0s are really old, this would also imply a large stability for bars in cluster S0s. However the notion of an evolutionary trend in the bar fraction is still a matter of debate (Jogee et al. 2004). Finally, we note that if the lower prevalence of bars in S0s is real, it is important to take this in account when selecting galaxy samples in evolutionary bar studies. In the past, there were fewer lower mass S0s (Dressler et al. 1997; Holden et al. 2007; van der Wel et al. 2007). If samples, selected at different z, contain lower mass disk-dominated galaxies (i.e. S0s + spirals), then measurements of different bar fractions may be related to different S0 and spiral fractions at these different redshifts.

3.5 Summary

We find that the bar fraction in cluster regions is somewhat larger than in low density regions. Cluster specific physical processes like ram-pressure stripping and harassment all conspire to increase the cluster bar fraction. We find a lower bar fraction in S0s compared to spirals. If S0s are the transformation products of spirals, then the lower bar fractions in S0s suggest that bars in S0s are slowly destroyed, probably by harassment 3.5: Summary 69 effects. Detailed simulations of behaviour of barred galaxies entering or located in cluster regions might give more insight into this problem. Also interesting would be to compare bar fractions in a large sample of field and cluster S0s. If field S0s are less susceptible to harassment, we might expect a larger bar fraction in field S0s. A lower prevalence of bars in S0s must be taken into account when selecting comparing disk-dominated (S0s + spirals) galaxy samples with lower mass systems at different redshifts, since in the past there were fewer lower mass S0s galaxies. 70 chapter 3: Bars in large scale structures at z=0.11 ± 0.02 Chapter 4 HST/ACS observations of shell galaxies.

ABSTRACT ∗

Shells in Elliptical Galaxies are faint, sharp-edged features, believed to provide evidence for a merger event. Accurate photometry at high spatial resolution is needed to learn on presence of inner shells, population properties of shells, and dust in shell galaxies. We want to learn more about the origin of shells and dust in early type galaxies. V-I colours of shells and underlying galaxies are derived, using HST Advanced Camera for Surveys (ACS) data. A galaxy model is made locally in wedges and subtracted to determine shell profiles and colours. We applied Voronoi binning to our data to get smoothed colour maps of the galaxies. Comparison with N-body simulations from the literature gives more insight to the origin of the shell features. Shell positions and dust characteristics are inferred from model galaxy subtracted images. The ACS images reveal shells well within the effective radius in some galaxies (at 0.24 re = 1.7 kpc in the case of NGC 5982). In some cases, strong nuclear dust patches prevent detection of inner shells. Most shells have colours which are similar to the underlying galaxy. Some inner shells are redder than the galaxy. All six shell galaxies show out of dynamical equilibrium dust features, like lanes or patches, in their central regions. Our detection rate for dust in the shell ellipticals is greater than that found from HST archive data for a sample of normal early-type galaxies, at the 95% confidence level. The merger model describes better the shell distributions and morphologies than the interaction model. Red shell colours are most likely due to the presence of dust and/or older stellar populations. The high prevalence and out of dynamical equilibrium morphologies of the central dust features point towards external influences being responsible for visible dust features in early type shell galaxies. Inner shells are able to manifest themselves in relatively old shell systems.

∗ Published as Sikkema, Carter, Peletier, Balcells, Del Burgo & Valentijn, 2007, A&A, 467, 1011 72 chapter 4: HST/ACS observations of shell galaxies. 4.1 Introduction

Shell galaxies (Malin & Carter 1980) have long been recognised as useful laboratories for learning on both the formation processes and the internal structure of elliptical galax- ies. Soon after their discovery, shells were identified as tracers of ”the splatter produced by a merger” (Schweizer 1980), more specifically minor mergers. Hence shell galaxies provided candidate configurations to investigate details of the accretion process, such as the nature of the accreted galaxy, the dominant types of accretion orbits, the radial distribution of the accreted matter, and the connection of accretion events to AGN ac- tivity. If shells trace specific orbit configurations of accreted stars, then they potentially contain information on the three-dimensional shape of the galaxian potential. Prieur (1990) recognised different morphological categories of shell galaxies. Type I shell galaxies have shells antisymmetrically (interleaved) aligned along the major axis. Type II shells are placed all around the galaxy. Type III shells show both or irregular features. Numerical simulation work in the eighties and early nineties provided the most widely accepted framework for interpreting these shell morphologies in terms of mergers. Quinn’s phase-wrapping formalism provides an elegant explanation for Type I, interleaved shells. Phase wrapping recognises the discrete distribution of turning points for radially-injected stars that oscillate back and forth in the galaxian potential (Quinn 1984); the shells themselves are the loci of the turning points, where pile-up leads to increased surface brightness; these loci move outward in the galaxy as a consequence of the proportionality between orbital period and apocentre distance. More complex shell systems of Types II and III can be produced by minor mergers from non-radial orbits, on non-spherical parent galaxies, or due to internal rotation in the accreted galaxy (Hernquist & Quinn 1987, 1989). Shells may also result from ”space wrapping” (Dupraz & Combes 1986) in the absence of radial orbit turn-around, when line-of-sight integration leads to an increase in surface brightness for debris of a satellite accreted on a high-angular momentum orbit. Finally, shells may result from major mergers between two disk galaxies, as a result of the return of tidal tail material (Hernquist & Spergel 1992), whenever the bulge-to-disk ratio of the parent galaxies is low (González-García & Balcells 2005). Models for shell formation not based on mergers have been proposed as well. These include tidal interactions (Thomson & Wright 1990; Thomson 1991) or asymmetric local star formation (Loewenstein et al. 1987). Several observational diagnostics may be used to test the various theories. Shells are mostly observed in isolated environments; this may indicate either a lower formation rate or a shorter lifetime in denser environments; it may indicate younger ages for shell systems in empty environments (Colbert et al. 2001). Forbes & Thomson (1992) noted that almost all galaxies which contain a kinematically decoupled core (KDC), also show shells, suggesting a relation between these galaxy properties. These two properties give support to the merger/accretion model for shells. How close to galaxy centres do shells exist is a matter of current interest. Because shell detection requires some form of unsharp filtering, whether digital or photographic (Malin & Carter 1980), and that works best when brightness gradients are not pro- nounced, shells have more often been detected in the outer parts of galaxies. However, inner shells contain useful information on the shell-making process. The existence of inner shells requires that orbital energy and angular momentum be removed from the accreted stars before these are released into the potential of the larger galaxy, there- fore inner shells imply that dynamical friction operated, and that the accreted galaxy 4.2: Observations and Data Reduction 73 disrupted late into the accretion process. Because shells trace accreted material with orbit apocentres at the shell radii, inner shells would provide strong evidence for the late build-up of the inner regions of ellipticals through accretion of small galaxies. To improve on our ability to detect inner shells, we used the ACS camera on board HST to image the inner parts of six well-known shell galaxies (see Table 4.1). The spatial sampling of the ACS, six times better than typical ground-based cameras, coupled to the absence of atmospheric blurring, provides for a more accurate modelling of the underlying galaxy brightness distribution, and a more accurate mapping of the shell profiles themselves. In this paper we present photometric data in V and I for the galaxies and the detected shells. The HST images also allow a precise determination of shell colours. The latter may provide useful diagnostics on the various shell formation models. The interaction model predicts similar colours for the shells as the host galaxy, whereas significant differences in shell colours are possible in the merger models. To date, observations give a confusing picture on shell colours. Examples are found of shells that are redder; similar, or bluer, than the underlying galaxy. In some cases, different authors report opposite colour differences (shell minus galaxy) for the same shell; we mention specific instances of this in Sect 4.5.3. Colour even seems to change along some shells; examples are NGC 2865 (Fort et al. 1986), NGC 474 (Prieur 1990), and NGC 3656 (Balcells 1997). Errors in shell colours are very sensitive to the correct modelling of the underlying light distribution. HST images allow for a detailed modelling of the galaxy light distribution, especially near the centres, and should provide increased accuracy in the determination of shell colours. Another important issue is the properties of the visible dust in the centres shell galaxies, which might say something about the dust visible in the centres early type galaxies in general. Our observations may help to learn more about dust origin and formation theories (Lauer et al. 2005). The paper presents a photometric analysis of the shells in the six shell galaxies imaged in our HST/ACS program. The systems observed contain three type I galaxies (NGC 1344, NGC 3923 and NGC 5982) and two type II (NGC 474, NGC 2865) and one type III (NGC 7626). The HST/ACS images are analysed by applying the technique of Voronoi binning (Cappellari & Copin 2003) on the single passband images, which yields high-S/N brightness and colour maps to see if the shells influence local colour gradients.. In an companion paper, the properties of the globular clusters were analysed (Sikkema et al. 2006, hereafter Paper I). The data reduction is briefly summarised in Section 2. In Section 3 we describe how we obtained global parameters, the shell fluxes and colours, the production of colour maps and dust properties. In Section 4, we compare our observations with N-body simulations of shell galaxies using different models, discuss implications of shell colours and analyse the dust properties of shell galaxies. We summarise our main conclusions in the last Section.

4.2 Observations and Data Reduction

The six shell galaxies were observed with the ACS_WFC camera between July 2002 and January 2003 with the filters F606W (V-band) and F814W (I band) in CR_SPLIT=2 mode . The camera contains two CCDs of 2048 x 4096 pixels, each pixel having a 74 chapter 4: HST/ACS observations of shell galaxies.

Galaxy type RA (J2000) DEC(J2000) AV AI type mV (1) (2) (3) (4) (5) (6) (7) (8) NGC 474 II 1h20m06s.7 +03◦2405500 0.11 0.07 E/S0 11.39 NGC 1344 I 3h28m19s.7 −31◦0400500 0.06 0.04 E5 10.41 NGC 2865 II 9h23m30s.2 −23◦0904100 0.27 0.16 E3-4 11.30 NGC 3923 I 11h51m01s.8 −28◦4802200 0.27 0.16 E4-5 9.88 NGC 5982 I 15h38m39s.8 +59◦2102100 0.06 0.04 E3 11.20 NGC 7626 III 23h20m42s.3 +08◦1300200 0.24 0.14 Epec 11.25

Galaxy MV σ re skyV skyI nSer r (1) (9) (10) (11) (12) (13) (14) (15) NGC 474 -21.17 164 50 203 129 ∗7.9 1.0 NGC 1344 -21.07 187 13 94 47 4.6 1.9 NGC 2865 -21.59 230 27 146 109 ∗6.3 1.2 NGC 3923 -21.92 249 39 170 103 5.1 1.9 NGC 5982 -21.91 240 34 124 182 5.1 1.0 NGC 7626 -22.16 270 24 119 80 7.0 1.1

Table 4.1: Properties of six shell galaxies. Columns (1-11) give data from the litera- ture (with columns (1-10) from paper I and column (11): de Vaucouleurs et al. 1991); columns (12-14) present data derived in this paper: (1): Galaxy name, (2): shell galaxy type, (3-4): Right Ascension and Declination in degrees, (5-6): Galactic extinction in V and I (in magnitudes), (7-9): Morphological type, apparent V band magnitude and absolute V band magnitude. (10): velocity dispersion σ in km/s, (11): effective radius in arcsec. (12-13): Calculated background values of the ACS V and I images in counts. (14-15): fitted Sérsic index n, starting from radius r (arcsec). *: no stable fit possible. Fits are drawn in Figures A.2-F.2. size of 000.049 pixel−1 resulting in a field of view of 20200 x 20200. Exposure times were on average 1000s. The inner 24 pixels of NGC 2865 and the inner 8 pixels of NGC 474 were saturated in both V and I. Table 4.1 contains the main characteristics of the galaxies like Right Ascension and Declination at Epoch J2000.0, the extinction coefficients, exposure times and adopted distances throughout this paper. Detailed information about the data reduction can be found in Paper I. In addition we found three B band (filter F435W) observations from July 2003 of NGC 7626, associated with program GO9427 in the HST archive with exposure times of 2620, 2620 and 2480 seconds.. We used the standard reduced images and combined them to remove most cosmic rays. We used a galactic extinction value of 0.313 (Schlegel et al. 1998) and applied the following transformation formulae (Sirianni et al. 2005):

2 BJ = m(F 435W ) + 25.709 + 0.108(B − V )JC − 0.068(B − V )JC (4.1)

2 VJ = m(F 606W ) + 26.410 + 0.170(B − V )JC + 0.060(B − V )JC (4.2)

Here =m(F435W) and m(F606W) are the ACS instrumental magnitudes and B and V are in the Johnson-Cousins system. 4.3: Data Analysis 75 4.3 Data Analysis

4.3.1 Global parameters Information about the morphology of the galaxies and location of possible shells was ob- tained by using the ellipse fitting task GALPHOT (see Jørgensen et al. 1992); it returns information such as ellipticity, position angle, surface brightness and the C3,C4,S3,S4 coefficients (Carter 1978), all as a function of radius. A model galaxy subtracted resid- ual image is returned as well, which is shown in Figures A.1 to F.1 for each galaxy. In the GALPHOT processing, background galaxies, point-like objects, dust features, bright pronounced shells and additional bad data were masked out by hand in an iterative way. Remaining faint shell structures, having a brightness typically not more than 5% of the underlying galaxy emission, do not notably affect the results. The best fits were obtained by allowing the centre, position angle and ellipticity as free parameters to vary. In two cases: NGC 2865 and NGC 5982 the central regions do not have reliable fits (see Figures C.1 and E.1 respectively). These regions correspond to rough circles with diameters 2400 for NGC 5982 and 400 for NGC 2865 respectively. The pixels within 0.300 of the centre of the latter galaxy are saturated. Global surface brightness profiles were obtained by plotting the surface brightness for each fitted ellipse as function of radius. The outer parts of the profiles are severely influenced by the background uncertainty. It is difficult to determine reliable back- ground values from the ACS images themselves, since the galaxies fill the whole field. Fortunately, for all galaxies, apart from NGC 5982, we found optical wide field data in the R band in the ESO archive from the Wide Field Imager (WFI) at the 2.2m ESO/MPI telescope. The WFI camera has a field of view of 340 x 330, much larger than the galaxies. We used the ASTRO-WISE system ∗ (Valentijn & Kuijken 2004) to reduce the WFI images. After subtracting a constant background value from the WFI images, GALPHOT was applied on them. The ACS background values were determined by matching ACS surface brightness profile to the WFI surface brightness profiles. For the galaxy without WFI data, NGC 5982, we assumed that the surface brightness profile follows a straight line from a certain point in a r1/4-log(I) plot. The calculated values for the background are listed in Table 4.1. The final GALPHOT results describing the morphology of the six galaxies are shown in Figures A2-F2 in Appendix A to F. These Figures also include the results of the isophotal analysis on P.A. and ellipticity from WFI ground based images. The errors in V-I (top-right panels) were determined using the scatter in the WFI backgrounds. We fitted Sérsic profiles to the I band surface brightness data using of the equation:

" #  r 1/nSer µ(r) = µe + cn − 1 (4.3) rc

with cn = 2.5(0.868n − 0.142) (valid for 0.5 < n < 16.5; Caon et al. 1993). The fit was made from that particular radius, chosen by eye, where the, sometimes visible, inner plateau, will not disturb the fit. These starting radii are drawn in Figures A-F.2 as vertical dashed lines. For two galaxies, NGC 474 and NGC 2865, no stable fits were

∗ www.astro-wise.org/portal 76 chapter 4: HST/ACS observations of shell galaxies. possible, with large variations for n depending on the starting point fitting radius r. The fitted values for n and starting points are listed in column (13-14) of Table 4.1. For NGC 474, adding an outer exponential to the fitting function significantly im- proves the surface brightness profile fit, resulting in a smaller and less concentrated spheroid than that listed in Table 4.1: re = 6.6arcsec and nSer = 3.0, and B/D = 0.71. Such ’bulge-disk’ decomposition of the profile led Schombert & Wallin (1987) to argue that NGC 474 is a face-on S0, a point which we address in Sect. 4.4.1. Residual images are obtained after subtracting the galaxy models obtained by GALPHOT and are shown in Figures A1-F1. They were solely used to identify and locate the shell features, but not to determine the brightness of the shells: the residual images still show some large scale fluctuations in the background which will disturb measurements of faint shell fluxes significantly. This makes it difficult to obtain reliable shell bright- ness from these images. A better approach, described in the next Section, is to work locally, within wedges. Except for NGC 1344, global isophotal analysis of our galaxies with other data has been done before: NGC 474 (ground-based B and V: Pierfederici & Rampazzo 2004; ground based B and R: Turnbull et al. 1999), NGC 2865 (ground based B, V and I: Reid et al. 1994), NGC 3923 (ground based B and R: Jedrzejewski 1987), NGC 3923, NGC 5982 and 7626 (ground based V, R and I: Bender et al. 1988) and NGC 5982 and NGC 7626 (HST WFPC2, V and I: Carollo et al. 1997; NICMOS 1.6 µm: Quillen et al. 2000). As mentioned before we can also compare with WFI archive data for NGC 474, 1344, 2865, 3923 and 7626. Comparison with these data give similar results.

4.3.2 Shell radii Shell positions were determined by two of us (DC and GS) by visual inspection on an image display of the residual images described in § 4.3.1. Shell positions are listed in Table 4.2. Following Prieur, we list radii corresponding to the outermost edge of each shell. Shell radii are discussed in § 4.5.1.

4.3.3 Shell fluxes We have developed a three-step procedure to determine the shell brightness. Common to all the steps is that we work locally, within wedges, that are carefully placed over parts of the shells. For a particular galaxy the same wedges, shown in Figures A.1-F.1, were used in the two passbands. Next follows a detailed description of the procedure. 1. Determining the local surface brightness profile of the galaxy. First the surface brightness profile within the wedge was derived. Data points of the curve were calculated by averaging the pixel values within the partial elliptical rings covered by the wedge. The elliptical ring segment had a thickness of 2.5 pixels, with a fixed central point and an ellipticity and position angle, whose values were taken by averaging the I-band GALPHOT results of the outer galaxy regions. Pixels within the wedge belonging to GALPHOT masks (see Section 4.3.1) were not used. Remaining pixel outliers in the elliptical ring segment were removed by iterating 10 times over the set of pixel values, each time applying a 4 σ clipping method. The resulting surface brightness profile valid for the wedge is used in the next step. 4.3: Data Analysis 77

Galaxy Direction a (”) × re Comments N 474 1∗ N 27.6 0.55 2 WSW 31.8 0.64 3∗ S 39.7 0.79 shells 3a and 3b 4∗ W 41.3 0.83 5∗ W 60.8 1.22 6∗ S 61.1 1.22 7∗ N 64.0 1.28 shells 7a and 7b, long arc 8 NNW 74.9 1.50 diffuse 9∗ S 76.2 1.52 10 S 77.2 1.54 11∗ W 87.2 1.74 12 SW 99.5 1.99 diffuse 13∗ N 103.0 2.06 shells 13a and 13b, long arc N 1344 1∗ NNW 26.7 2.05 2 SSW 37.0 2.85 bright blob 3∗ NNW 53.3 4.10 4 WSW 57.8 4.45 5 NNW 62.0 4.77 diffuse 6 SSW 65.5 5.04 7 WSW 71.3 5.48 8∗ SSE 93.1 7.16 9 SSE 109.5 8.42 10∗ SSE 122.6 9.42 N 2865 1 SW 77.1 2.86 large 2∗ E 83.0 3.07 bright shell 3 W-E 90 3.33 large scale loop 4 SE 99.0 3.67 diffuse loop N 3923 1∗ S 18.0 0.46 Prieur: 24S, 18.8" 2 N 19.4 0.50 Prieur: 23N, 19.5" 3∗ S 28.7 0.74 Prieur: 22S, 30.0" 4 N 29.3 0.75 Prieur: 21N, 30.0" 5 N 34.3 0.88 diffuse 6 S 37.7 0.97 7 N 41.5 1.06 8∗ S 44.0 1.13 Prieur: 20S. 44.7" 9 N 51.2 1.31

Table 4.2: All shells identified by eye in our GALPHOT residual images. (1) Shell number/label. Asterisks indicate shells which have a colour determination (see also Table 4.3 and the appendix); (2) Shell direction relative to the centre; (3) Semi-major axis of outer shell border in arcseconds as measured from the galaxy centre; (4) Similar to column 3 but now in terms of effective radius; (5) Comments: additional information. The comments for NGC 3923 give Prieur (1988) labels + positions. 78 chapter 4: HST/ACS observations of shell galaxies.

Galaxy Direction a (”) × re Comments N 3923 (continued) 10∗ S 55.5 1.42 Prieur: 18S, 55.7" 11 N 60.4 1.55 12 N 64.1 1.64 13∗ S 67.0 1.72 Prieur: 16S, 67.1" 14 N 72.8 1.87 Prieur: 15N, 73.0" 15∗ S 79.6 2.04 Prieur: 14S, 79.3" 16 N 99.9 2.56 17∗ S 103.6 2.66 Prieur: 12S, 104.7" 18 N 128.1 3.28 Prieur: 11N, 128.1" N 5982 1 E 8.0 0.24 2 E 9.8 0.29 3 W 10.9 0.32 4 E 12.5 0.37 5 E 15.0 0.44 6 E 17.7 0.52 7 E 19.9 0.59 8 W 20.1 0.59 9 E 21.4 0.63 10 W 21.9 0.64 11 E 23.6 0.69 12 W 23.9 0.70 13 E 27.3 0.80 14 W 28.7 0.84 15 W 31.9 0.94 16 E 39.0 1.15 17 W 39.5 1.16 18 W 47.5 1.40 19 E 49.8 1.46 20 E 65.3 1.92 21 W 67.5 1.99 22 W 78.7 2.31 23 W 91.8 2.70 24 NE 100 2.94 N 7626 1 SW 24.1 1.00 2∗ NE 32.4 1.35 3∗ SW 43.8 1.83 4 E 47.6 1.98

Table 4.2: continued 4.3: Data Analysis 79

2. Making a local galaxy model. Several smooth curves were fitted to the surface brightness profile using Legendre polynomials of different degrees, applying the IRAF tool ’CURFIT’. Usually the shells are already visible as small bumps in the profile. Existence of shells is double checked by inspecting the residual images of GALPHOT. Another check is made by inspecting if the bumps are visible at the same position in both passbands. The data points in the profile which are part of the bumps in V and I, were left out in the fitting procedure. The fitted profile gives the model flux values in the partial elliptical rings mentioned in the previous step, which enables us to construct a 2d model image, which is valid locally for the wedge.

3. Obtaining shell surface brightness. After subtracting each model image from the galaxy image, the residual images show the shells in each wedge for each Legendre model. The same recipe, as described in the first step, was applied again to obtain surface brightness profiles for each residual image, but now using the ellipticities of the shells, which, in case of type I shell galaxies (NGC 1344, NGC 3923 and NGC 5982), are much rounder than the underlying galaxy (this was already known, see e.g. Prieur, 1988). The surface brightness profiles now clearly show the shell fluxes and the background is close to zero. The results are shown in the top panels of the Figures A3-F3 in Appendix A to F. Reliable shell fluxes could not be obtained for left-side (North- East) of NGC 3923 and NGC 5982, because they are not well defined features, which is related to their low contrast and S/N w.r.t. the galaxy.

4.3.4 Shell colours By combining the results of two passbands for each Legendre fit, we obtain the average colours in the shell regions. Due to the different fits (i.e. each fit is constructed using a particular Legendre polynomial), the derived colours show variations which increase for fainter shells. Consequently, we use only those shell regions for which all fits showed a stable answer. These regions are indicated by vertical lines in the top panels of Figures A3-F3 in Appendix A to F and typically contain data points with at least 10 counts. The final shell colour within a region is calculated by applying a weighted average, using the values for the colours and their errors derived for each Legendre fit. The resulting values are plotted in the Figures and listed in Table 4.3.

4.3.5 Galaxy colours Global colour maps of the galaxies were obtained by using the adaptive binning al- gorithm (Cappellari & Copin 2003). This algorithm bins two-dimensional data to a constant signal-to-noise ratio per bin by calculating a Voronoi tessellation. For objects with large gradients in S/N, as is the case for galaxies, this will result in smooth 2D colour images in regions with low S/N and therefore will show the colour gradients bet- ter than in the traditional way. A practical example and explanation of this algorithm is given in Ferreras et al. (2005). The procedure to obtain the colour maps consists of three steps: First, the adaptive binning algorithm is applied on the 3x3 binned non-background subtracted I band images by using a S/N of 250 and leaving out masked regions. The 80 chapter 4: HST/ACS observations of shell galaxies. resulting Voronoi tessellation is also used for the V band image in the next steps. The Voronoi tessellation is further processed by applying a Delaunay triangulation to the central positions of the Voronoi cells, assigning the average flux values within the cells to these central positions. Finally, V-I colour maps are obtained by subtracting the appropriate background values and combining the images by applying the appropriate transformation formulae (see Paper I). In the outer regions, where the uncertainties in the background become important, we still see large variations. These outer regions also show discrete offsets between dif- ferent quadrants which increase when going outwards. This is due to random variations of the subtracted bias level as measured in the overscan versus the actual bias level in the science images∗. These differences can be as large as a few counts, which will show up especially in the V-I colour maps with low signal: the outer galaxy regions in our images. Unfortunately, appropriate calibration data to correct for this effect only exists for ACS observations later than November 2004, much later than our observations. NGC 474 and NGC 3923 show elliptical red rings which are approximately 0.05 mag. higher in V-I than its surroundings (Figure 4.1, left and right respectively). We believe that these are caused by artifacts, probably reflections, within the optical system. We double checked this for NGC 3923 by constructing a V-I colour map from ground based VLT-FORS2 images in Bessel V and I. The colour map does not show the ring which is illustrated in the lower panel of Figure D.3 showing, as a green line, the V-I profile of the ground based data: the bump due the ring between r=40” and 60” is not visible. Any correlation between shells and integrated V-I colours is checked by calculating the Voronoi colours in the same wedges as used in the previous section; in general the shell fluxes are so low that little of it is reflected in the integrated V-I colour profiles. This can be seen in the lower parts in Figures A3-F3 in Appendix A to F. We also see that most shell V-I colours are usually similar or sometimes redder than the colour of the galaxy. The signature of the rings of NGC 474 and NGC 3923 is also visible in these Figures as shallow, large scale, bumps with an amplitude about 0.05 at 25 arcsec and 50 arcsec for NGC 474 and NGC 3923 respectively. For NGC 7626, where we also have B data, we only could determine a reliable shell colour for shell 2, which has (B − V )galaxy = 0.99 ± 0.02 and (B − V )shell = 1.11 ± 0.05 (see Figure 4.5).

4.3.6 Shell radial profiles Figure 4.2 shows profiles for 21 bright shells in several galaxies. Some shells show plateaus and are asymmetric: they reach a maximum flux near the outer shell border, often accompanied with a sudden sharp drop (examples: N474_6, N474_13, N2865_2, N3923_8, N3923_10, N7626_2, N7626_3). Other shells have a symmetrical Gaussian- like shape (N474_7, N474_11, N3923_1, N3923_15, N1344_3, N1344_6). Some shells seem to have double peaks (not shown in Figure but see the Appendix). A quantitative characterisation is obtained by applying a Gauss Hermite fitting procedure (van der Marel & Franx 1993) using five free parameters: γ, R0, σ, H3 and H4. The fitted values are given in Table 4.3 and the results are drawn as red curves in Figure 4.2. The shells we see are the result of a projected 3D density distribution in our line of sight. To get more information about the real or intrinsic three dimensional density distribution, we integrated along the line of sight using two different simple 3D density

∗ http://www.stsci.edu/hst/acs/documents/isrs/isr0407.pdf 4.4: Notes on the individual systems 81

ESO NGC474.color.fits ESO NGC3923.color.fits

sikkema/Skycat Dec 08, 2005 at 12:29:16 sikkema/Skycat Dec 08, 2005 at 12:30:12

Figure 4.1: Voronoi binned colour image of NGC 474 (left) and NGC 3923 (right). The rings are probably artifacts; offsets between quadrants are also visible. shell models assuming spherical symmetry and assuming a opening angle φ in our line of sight. Two other parameters used in both models are the points r0 where shells have their maximum stellar density ρ0. −0.5 The first model assumes that the shell has an intrinsic (r0 − r) density distribu- tion which is predicted by theory to describe the inner parts of phase wrapped shells (Dupraz & Combes 1986; Prieur 1988). The second model is supposed to describe spatial wrapped shells (Turnbull & Collett 1999). It assumes that shells are have an intrinsic Gaussian shape with thickness rg. However this model lacks a physical basis. The results are again given in Table 4.3 and shown in Figure 4.2: here the green and blue curves depict the results of the first and second model respectively. In most cases the second model describes the data best. The ’noise’ in the green curve is due to the fact that a discontinuous model (the model has a sudden drop in flux to zero at r0) is fitted to the data: the noise in the model reflects the noise in the data.

4.4 Notes on the individual systems

4.4.1 NGC 474 The low v/σ ratio (this defines the amount of rotation), of 0.18 (Rampazzo et al. 2006) is consistent with NGC 474 being either a near face-on disk, or a near spherical galaxy. NGC 474 is part of a small evolving poor group (Rampazzo et al. 2006). Its shell system is very complicated as shown in Figure 1 of the multi-wavelength study of Rampazzo et al.. In addition to the shells, a loop is visible, which heads east-west, starting from a comma shaped feature (outside the ACS field of view), passing the sideline towards the galaxy centre. In HI, there are signs of tidal interactions with the nearby (at 5.5’) regular spiral galaxy NGC 470. The same study does not make clear, however, if NGC 470 is responsible for or even related to the shell system. 82 chapter 4: HST/ACS observations of shell galaxies.

Figure 4.2: Shell profiles with different fits. Horizontal scale in arcsec. Vertical scale in counts. The red curve is a Gauss Hermite fit to the data giving a quantitative characterisation of the shell shapes. Green and blue curves are the results by integrating −0.5 along our line of sight a spherical symmetrical density model using a (r0 − r) and Gaussian distribution respectively. 4.4: Notes on the individual systems 83

Being classified as a type II shell galaxy, this system has been used to test predictions of the weak interaction model (WIM; Thomson & Wright 1990; Thomson 1991). Here, shells are induced in the outer parts of the host galaxy due to tidal effects resulting from a fly-by of another galaxy. Morphologically, the WIM simulations show the shells as almost complete windings or spirals around the centre, when looking face-on. Looking at our GALPHOT residual image of NGC 474, however, the shells look more like short arcs. The shell positions and shapes (see the residual image Figure A.1) resemble more those of the results of the merger simulations shown in e.g. Figure 5 of Dupraz & Combes (1986), model 4 of Hernquist & Quinn (1989), or model 7 of Hernquist & Quinn (1987): all mergers of low-mass companions on non-radial orbits with a spherical or mildly oblate primary. In these merger simulations the shells can be very old (>5Gyr) and are spatially wrapped around the galaxy. Except for one shell, the colours of the analysed shells (see Table 4.3 and Figure A.3) are similar to the galaxy, consistent with previous studies (in B-V, V-r, R-I (Schombert & Wallin, 1987), in B-R (Turnbull et al, 1999) and in B-V (Pierfederici & Rampazzo 2004)). Some large shells are overlapped by different wedges, enabling us to compare their colours independently: shell 3 is located at ≈ 40” from the centre. We splitted the shell into shells 3a and 3b (see Figure A.1); comparing their colours gives similar V-I colours: 1.21 ±0.13 for 3a and 1.17±0.05 for 3b. Similarly, shell 7 was splitted into shells 7a and 7b (located at about 64” from the centre, see Figure A.1). They have V-I colours of 0.96 ±0.05 and 1.01 ±0.08 respectively. For shell 13a and 13b, at 103” from the centre, we find significantly different colours: 0.93±0.04 and 1.14±0.05. These are probably erroneous values due to low shell fluxes compared to the underlying galaxy. Wilkinson et al. (2000) also find offsets in colour up to 0.30 mag between shell segments a similar type II shell galaxy 0422-476. Another colour determination for this shell is given by Schombert & Wallin (1987), who find redder R-I colours than the galaxy (galaxy R-I=0.88, shell R-I=1.09) and Pierfederici & Rampazzo (2004) who find slightly redder colours in B-V. The innermost sharp edged shell is detected at about 30” from the centre in the South direction. Of all shells analysed in this work, the only really blue shell relative to the integrated galaxy colour is found in this galaxy, which is shell 5. Interestingly, (Turnbull et al. 1999), found that their only really blue shell w.r.t. the integrated galaxy colour is the comma shaped feature at the SW (beyond the field of our ACS images, but see Figure 9 Turnbull et al. 1999). We note that the position of our blue shell 5 lies exactly on the tail or loop, which is connected to the comma shaped feature. We therefore suggest that shell 5 is related to this feature. Blue shell colours in other shell galaxies have been found in young, gas-rich merger remnants such as NGC 3656 (Balcells 1997) and Arp 230 (McGaugh & Bothun, 1990) as well as blue tails in many other interactions (see e.g. Schombert et al. 1990). It is therefore tempting to conclude that this entire feature is the remains of a recent small merger unrelated to the rest of the red shell system. Isophotal analysis of NGC 474 (Figure A.2) shows that the ellipticity is changing fast from 0.08 to 0.24 between 10” and 20”. and back to 0.08 beyond 20”. At the same radii, the position angle is changing from 0 to about 20 and back. The galaxy contains several pronounced dust lanes within the inner 15”, not seen previously (see, e.g., Ravindranath et al. 2001; Sarzi et al. 2006). Ravindranath et al. find a point source in the centre. The top right panel of Figure A.2 shows that this source is 0.05 mag. bluer in colour than its surroundings (V − I = 1.41). NGC 474 contains the largest visible dust mass 4 of our sample (≈ 10 M , Table 4.4) and its centre shows peculiar kinematic behaviour 84 chapter 4: HST/ACS observations of shell galaxies.

(Hau et al. 1996).

4.4.2 NGC 1344

The shells in this galaxy are supposed to be the result of phase wrapping, since NGC 1344 is a type I shell galaxy. The colour of one outer shell of this type I shell galaxy was determined by Carter et al. (1982). This shell appears to be somewhat bluer than the main body of the galaxy. In our GALPHOT residual data the innermost shell is visible at about 27” North from the centre. For the first time, we determined positions and the colours of some inner shells for this galaxy. The shell positions (Table 4.2) seem to show an interleaving pattern, although not as regularly as in NGC 3923. The V-I colours of the Northern shells are redder than the galaxy while they are similar or slightly bluer than the galaxy on the South side. The redder colours on the North side are probably due to dust, since these shells are much nearer towards the centre where dust is more present. The blue colour of one of the Southern shells is at least consistent with the earlier finding of Carter et al. The low number of shells and blue colour of some of these may be evidence for a relatively recent merger event.

4.4.3 NGC 2865

This galaxy is classified as a type II shell galaxy. However, looking at our residual images this system looks more as if it belongs to the type III class, with lots of irregular features and loops. The core is much bluer (V-I ≈ 1.05) than the outer parts (V-I ≈ 1.15). This blue colour is related a young stellar population (0.4-1.7 Gyr) which forms a KDC (Hau et al. 1999). NGC 2865 also contains an incomplete HI disk (Schiminovich et al. 1995). An interesting result is that a bright HI patch is coinciding with a bright shell observed earlier (labelled as shell 2 in this work and called shell 2B by Fort et al. (1986)). Figure 1c of Schiminovich et al. shows an overlay of the HI data and Fort’s schematic shell map. If these are related, this means that the shell is moving towards us, implying a spatially wrapped shell and confirming that this is not a type I shell galaxy. However, the asymmetric shell profile seems more consistent with a phase wrapped shell. Fort et al. derived a slightly redder colour for shell 2 of (V − R)Johnson= 0.84 ±0.09, compared to the galaxy colour (V −R)J =0.75 ±0.03. A similar colour difference shell 2 and the galaxy is found in this work: V-I=1.11 ±0.05 vs local galaxy V-I=0.98 ±0.01. The ’jet’, as reported by Fort et al. looks more like a loop (see top of Figure C.1), which may or may not be connected with the bright shell and HI. Compared with the simulations, the residual image looks similar to situations in for instance Figure 2 or 6 of Hernquist & Quinn (1987), Figure 10 of Hernquist & Quinn (1989), or Figure 11 of Dupraz & Combes (1986). All these simulations use small disk galaxies as the intruders, which is supported by the presence of an HI disk in NGC 2865. The bright shell 2, coinciding with the HI patch, is probably the best candidate for follow-up spectroscopy of our whole sample, because of its high contrast with respect to the galaxy light (≈0.8 mag. higher in V than the galaxy, see Figure 4.3). 4.4: Notes on the individual systems 85 09 . 18 . 0 0 ± ± 05 . 84 0 . 68 . ± = 0 = 0 11 . J J ) ) R R = 1 − (17) − V V ( shell Comments ( ) : : † V † − B ( P.A.=5.0 and ellipticity=0.13 ; (12-16) fitted parameters using P.A.=49.0 and ellipticity=0.35 P.A.=15.0 and ellipticity=0.15 P.A.=163.0 and ellipticity=0.38 P.A.=152.0 and ellipticity=0.27 P.A.=107.0 and ellipticity=0.31 ρ Shell 1 of Shell 2B of and 4 Φ h , 0 3 R σ h and position angle of the outer parts of the galaxy; 0 b a γ R (12) (13) (14) (15) (16) 78.5 57.6 2.498.7 25.167.1 -0.2 2.1 80.768.7 0.1 3.6 98.1 -0.1 -0.2 4.030.3 -0.1 40.7 -0.4 0.0 -0.1 2.054.6 -0.1 -0.0 57.2 3.4 -0.0 -0.1 82.5 47.6 1.7 -0.1 0.0 80.9 42.111.8 1.3 54.226.9 0.7 65.117.7 -0.3 -0.2 1.4 77.6 -0.2 -0.1 1.3 -0.1 -0.2 0.1 -0.1 76.4 39.6 2.8 -0.4 -0.2 152.0 34.3 4.3 -0.0192.6 -0.2 93.4 8.0 -0.3 -0.2 226.6 87.1 4.1 -0.2214.0 -0.2 78.0 4.3324.9 -0.3 14.0102.4 -0.0 1.8 26.8 1.6 1.0 -0.0 -1.2 0.0 106.9 30.9 1.1 -0.3 -0.1 ρ model (see Section 4.3.6): Φ 5 . 0 − 0 ) R (9) (10) (11) r 28.0 30.5 1.5 40.3 40.042.8 0.9 23.061.5 0.4 28.059.8 0.3 17.065.5 0.8 22.3 0.2 83.5 18.0 0.4 49.8 19.593.4 1.3 25.1 1.1 83.2 21.3 1.1 17.9 25.228.4 8.9 24.444.0 2.2 18.755.0 1.8 67.3 9.7 17.879.0 0.5 0.4 14.2 0.3 32.4 19.043.4 2.8 25.0 0.8 103.4 28.0 0.4 − 0 g r ( ρ R Φ ; (17) Comments: Average ellipticity 1- 0 4 R (5) (6) (7) (8) h 27.1 39.1 2.037.4 1.8 41.042.3 7.3 28.060.3 0.6 2.0 32.059.4 0.5 3.8 29.062.5 0.3 1.4 12.5 1.1 4.380.9 0.4 7.0 4.8 0.8 47.991.9 6.9 33.0 2.2 4.0 4.5 0.8 82.3 36.3 2.717.4 1.1 23.527.3 0.9 21.043.7 18.0 2.1 27.055.0 3.5 1.1 17.566.6 2.9 0.6 21.878.1 0.9 1.4 12.5 0.6 1.7 0.4 32.2 28.642.7 0.8 42.2 6.0 2.3 1.0 102.5 43.1 5.7 0.3 13 06 06 08 05 05 11 05 02 05 05 08 04 06 05 08 05 08 04 02 04 03 14 15 ...... and 13 ...... 0 0 0 0 0 0 0 0 0 0 0 0 0 I 0 0 0 0 0 0 0 0 0 0 0 0 = Fort et al. (1986) 3 − ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± h ± † V , ∆ σ 0.00 , -0.04 -0.04 -0.19 -0.07 -0.06 -0.08 -0.05 -0.02 -0.01 -0.07 -0.08 +0.09 +0.25 +0.26 +0.03 +0.25 +0.13 +0.16 +0.10 +0.13 +0.08 +0.01 +0.09 +0.17 0 ; (9-11) fitted parameters of g R 03 02 02 02 02 02 02 02 02 02 02 02 02 01 01 01 01 01 01 01 01 03 02 02 02 R , ...... gal 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 γ I ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± − and ρ V 1.31 1.28 1.21 1.21 1.21 1.20 1.15 1.06 1.03 1.07 1.17 1.01 1.01 1.23 1.20 1.21 1.22 0.98 1.28 1.29 1.28 1.31 1.29 1.31 1.29 , Φ , 13 07 04 05 07 14 15 11 05 02 03 05 05 04 05 06 13 05 06 08 08 05 08 05 04 ...... 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 shell R I ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Shell properties. (1) galaxy NGC number/shell indexation (see Figures A.1-F.1); (2) V-I colour and errors of the shells, (3) − V 3 1.24 1 1.46 456 1.46 0.96 1.09 138 1.39 1.30 1.16 138 1.27 1.37 1.29 2 1.46 10 1.40 3a 1.21 7a 0.96 11 1.42 10 1.20 1315 1.40 17 1.22 1.23 3b 1.17 7b 1.01 (1) (2) (3) (4) 13a 0.93 13b 1.14 2, E 1.11 N 474 N 1344 N 2865 N 3923 N 5982 N 7626 Galaxy Gaussian-Hermite fitting: Table 4.3: local V-I colourSection and 4.3.6): errors of galaxy; (4) Colour difference between shell and local galaxy (5-8) fitted parameters of Gaussian model (see comparison with external data with 86 chapter 4: HST/ACS observations of shell galaxies.

Figure 4.3: Light profile of the galaxy along the wedge containing the brightest shell 2 of NGC 2865. The shell is about 0.8 magnitude brighter in V than the galaxy.

4.4.4 NGC 3923 This is probably the most studied shell galaxy and the prototype of the Type I shell class, where shells are placed in interleaving order along the major axis. Prieur (1988), using ground based CCD data and photographic plates, mapped the whole shell system. A comparison shows that we do not find any other shells than those already given by Prieur, despite our much better resolution and galaxy subtraction near the centre (compare our Figure D.1 with his Figures 1-5). In the last column of Table 4.2, we list and compare his shell positions with our results. There is a good agreement; small offsets between positions are likely due to measurement errors. Fort et al. (1986) give colours for three shells in NGC 3923, with only one of them in our field of view (our shell 8): his shell 1 has a colour of (V − R)Johnson =0.68 ±0.18 with the local galaxy colour being about (V − R)J = 0.82 ± 0.03. Our results for this shell are V-I=1.29 ±0.05 and local galaxy colour of V-I=1.28; the colour difference being consistent with Fort’ s work. On the West side, we find that all shells have a similar or slightly redder colour than the galaxy. We were not able to determine reliable shell colours on the East side. The local models, using different degrees for the Legendre polynomials, do not give stable answers. This is related to the low S/N of the images and badly defined shell features.

4.4.5 NGC 5982 NGC 5982 is a Type I shell galaxy (Figure 4.4). The galaxy is well known for its KDC (Wagner et al. 1988), recently confirmed using 2D central mapping using OASIS (McDermid et al. 2006) and SAURON (Emsellem et al. 2004). Analogy with the KDCs of Emsellem et al. (2004) indicates that the KDC is probably a rotating central feature, i.e. a disk. The ellipticity becomes very round in the inner 2” of the galaxy. We were 4.4: Notes on the individual systems 87 not able to smoothly fit the central regions. Even after using 4 harmonics, a quadrupole with wings containing a flux of about 200 counts (± 1-2 % of the galaxy flux), is still visible. The C4 coefficient in this region is about -0.02 (Figure E.2) indicating a boxy structure (Carter 1978). 2D spectral mapping of this region (Emsellem et al. 2004) show that the stellar velocities exhibit a 90 degree offset in the central region with respect to the outer regions. The bad GALPHOT fit limits us in giving a final answer about how far the shells do extend to the centre. The innermost detectable shell is located about 8” East. Assuming the galaxy is about two times further away than NGC 3923 (see Table 4.1), the distance from the centre of this shell is comparable to the inner shell of NGC 3923. Double checking using other methods (unsharp masking and GALFIT (Peng et al. 2004), which fits symmetric 2D models), show no more inner shells. Unfortunately reliable shell brightness could not be determined, because of their faintness and low contrast. A deeper, 3600s, ground based image, was found in the ING Archive (taken with the 2.5 meter Isaac Newton Telescope INT in July 1989 in the R band). After again using GALPHOT and subtracting galaxy-model, we obtain Figure 4.4 5, which shows the shells with somewhat higher contrast. The outer shell 24 on the minor axis is barely visible in the ACS images and is 90 degrees displaced with respect to the inner shells . The next bright outer shells 20 and 21 are slightly displaced with respect to the inner shells. This peculiar shell morphology resembles the scenario shown in Figure 6c of Dupraz & Combes (1986). In this simulation a small elliptical falls into the potential of a prolate E3.5 galaxy with an impact angle of 90 degrees. The same misalignments occur for the outer shells in this event. This galaxy is classified as an YP galaxy (i.e. containing a young stellar population similar to NGC 2865) with a fine structure index Σ2 of 6.8 (Michard & Prugniel 2004). Michard & Prugniel do not mention shells. Looking at our images we estimate that there are at least eight shells, which would imply a much higher Σ2 of about 11.

4.4.6 NGC 7626

Its core kinematics (Balcells & Carter 1993) and the fact that it has bright globular clusters (Paper I) indicate a (minor) merger event which happened one or several Gyr ago. This is the first time that shell colours and brightness have been obtained for this galaxy. A shell on the East side was detected by Jedrzejewski & Schechter (1988); this shell lies outside our field of view. Another possible shell lying on the S.W. side, was detected by Forbes & Thomson (1992). This shell is also seen in our data. We only detect one other, very sharp, edged shell to the N.E. The structure of these shells looks somewhat like those of the simulations shown in Figure 5 of Dupraz & Combes (1986). Here a a small spiral (1% mass of host) was thrown into the potential of an E3.5 oblate galaxy. Frame 3 (after 4 Gyr) of this Figure looks very similar to the NGC 7626 shell system. The zero rotational momentum encounter should create phase wrapped shells which is probably evidenced by the radial shell shapes (see Section 4.5.2). Further support for a disk intruder galaxy comes from the shell colours. In both (B-V) and (V-I), the brightest, inner shell is redder than the galaxy (see Figure 4.5), which is probably due to dust (see Section 4.5.3). The fainter outer shell has a similar or slightly bluer (V-I) colour than the galaxy. 88 chapter 4: HST/ACS observations of shell galaxies.

Figure 4.4: Ground based residual image of the shell system NGC 5982, with shell labels (see Table 4.2). This morphology resembles the scenario shown in Figure 6c of Dupraz & Combes (1986), where a small galaxy has fallen in the host galaxy at an offset angle of 90 degrees. 4.5: General results 89

Figure 4.5: Colour-colour, B-V vs V-I, diagram of the brightest, inner, shell in NGC 7626. Cross + errorbars represent the local galaxy colour, while the triangle with errorbars represents the derived shell colour

4.5 General results

4.5.1 Shell radial distributions

As discussed in Sect. 4.1, inner shells contain useful clues to the shell formation process, and indeed, one of the central goals of the HST-based imaging program was to determine how close to the galaxy centres we find shells. Our determined shells and shell radii (see Sect. 4.3.2) are listed in Table 4.2, given in arcsec as well as in units of the effective radius re (from col. 14 of Table 4.1). Innermost shell radii span a wide radial range, from rmin/re = 0.24 in NGC 5982 to rmin/re = 2.86 in NGC 2865. We find a slight tendency for type-I shell systems to extend within re and for types II and III to lie in the outer parts: the two innermost shells, in units of re, are in NGC 5982 and 3923, two Type-I galaxies, while the third galaxy with rmin/re < 1, NGC 474 (type-II), has an uncertain value of re: as shown in Sect. 4.3.1, re might be significantly smaller than listed in Table 4.1; adopting such smaller value for re, we would get rmin/re = 4.2. However, not all type-I shell systems show inner shells: for NGC 1344 we find rmin/re > 2. The HST imaging has revealed shells in the inner two kpc for two of the type I shell galaxies: NGC 3923: rmin = 8arcsec = 1.8 kpc (at an assumed distance of 20.0 Mpc) and NGC 5982: rmin = 8arcsec = 1.7 kpc (at an assumed distance of 41.9 Mpc). This is interesting as it clarifies that shells do indeed form near the galaxy nuclei. This region is easily available to spectroscopic kinematic measurements, hence the correspondence between kinematic features and shells may be revealed. For the galaxies with no inner shell detections, the most straightforward interpreta- tion is that shells never formed at those radii. This conclusion may be too simplistic. Three conditions need to concur for the detection of inner shells: (i) shells need to 90 chapter 4: HST/ACS observations of shell galaxies. form; (ii) shells need to survive until the observation epoch; and, (iii) they need to be detectable through the shell-detection methods employed. The second and third conditions listed above make shell detection harder as we look closer to the galaxy centres. We first address shell survival. We expect shell lifetimes to be shorter near the centres: phase mixing scales with dynamical time, hence it is faster near the centre; shells should lose contrast and fade away faster near the centres. Furthermore, galaxy centres are more dynamically active than the outer parts, as any object that merges after the formation of the shell system and reaches the centre, will gravitationally perturb the orbits of the stars that define the shells. In this respect, it is quite suprising that we see inner shells in NGC 3923 and NGC 5982, which have a large number of shells and are relatively old systems (Nulsen 1989) Shell detectability becomes progressively more difficult as we approach galaxy cen- tres. As mentioned above, the pronounced brightness gradients in the inner regions of ellipticals lead to the break-down of unsharp-mask techniques. Because shells near the centres are closer to each other than further out, it may become difficult for the detection algorithm to pick up the underlying, non-shell brightness levels. Finally, el- liptical nuclei are known to be dusty (Lauer et al. 1995; Phillips et al. 1996; Peletier et al. 1999a; Ravindranath et al. 2001). Dust may act in two ways. It may simply hide the shells: examples of such an effect are NGC 474, NGC 1344 and NGC 7626. Dust may also perturb the general light distribution, so that the smooth galaxy model one generates and subtracts to reveal the shells has too strong residuals for the shells to appear. Typically, the underlying light distribution needs to be smooth to within a few percent for the shells to show up. Strong dust patches easily lead to stronger third- or fourth-order Fourier residuals in the isophotes. Examples of this situation are NGC 2865 and NGC 5982. Clearly, HST imaging at NIR wavelengths would allow us to see through the dust and would strongly improve the chances of detecting inner shells in ellipticals.

4.5.2 Shell brightness profiles Looking at Figure 4.2, it is clear that in general the Gaussian model fits the data better −0.5 than the (r0 − r) model. The latter model especially fails for lower r. However, this model is only meant to describe the flux behaviour very close to the shell maximum −0.5 at projected radius rmax. The (r0 − r) model also predicts a fast dropping flux at distances slightly larger than rmax. This is indeed seen at the two bright shells in NGC 7626 (see Figure 4.2). Here the flux drops from its maximum to zero within a small interval relative to the shell size. In general this seems to happen for shells with a plateau, i.e. N474_6, N474_13, N2865_2, N3923_8, N7626_2, N7626_3. −0.5 The (r0 − r) model is supposed to describe phase wrapped shells (Dupraz & Combes 1986; Prieur 1988), but is does not seem to fit well most of the bright shells in type I shell galaxies NGC 1344 and NGC 3923, where shells are expected to be the result of phase wrapping. For NGC 3923 this may be related to the age of the shell system. The large number of shells indicate an old age. The shells may smooth out −0.5 as a result of their older age and will not have razor sharp edges as the (r0 − r) model assumes. On the other hand: NGC 1344, showing only a few shells, indicating a younger age, does not do much better. NGC7626, which is probably very young (see Section 4.4.6), might be a better example where the model works. We conclude that the −0.5 (r0 − r) model works best for younger shells. Older phase wrapping shells probably 4.5: General results 91

−0.5 have a more extended structure and density profile than the theoretical (r0 − r) model, for instance due to internal velocity dispersions in the intruder galaxy. It seems −0.5 that the (r0 − r) is just too simplistic to describe the real shells.

4.5.3 Shell colours In our galaxy sample we find only one shell with blue colours. All other shells have similar or redder colours. Red shell colours are also found by many others like in NGC 7600 (Turnbull et al. 1999), IC 1459 (Forbes et al. 1995b), NGC 7010, NGC 7585 and IC 1575 (Pierfederici & Rampazzo 2004). The redder colours could be explained by at least four scenarios:

• The stars in the shells are on average older than those in the main body of the galaxy. This may occur if the shells belong to older parts of the intruder galaxy, e.g. the bulge, or if the interaction or merger led to the formation of young stars from gas throughout the galaxy. N-body simulations show that the best reproductions of shell morphologies are obtained by using very small intruder galaxies, with only a few percent of the mass of the host. It is therefore not very likely that they will form sufficiently many new stars which could lower the average colour of the whole galaxy.

• The stars in the shells are more metal rich than those in the main body of the galaxy. This is highly unlikely given the expected small mass of the intruder galaxy and usually positive correlation between metallicity and galaxy mass (Sandage & Visvanathan 1978).

• The stars in the shells have different, redder, populations than the underlying galaxy. This scenario only works for very specific conditions. If the progenitor galaxy is a (small) late type (star-forming) galaxy and star formation is truncated after the merger event, then after some several 108 yrs the light of the original stellar population will be dominated by RGB and AGB stars, which will redden the integrated colours of the shells. This reddening effect has been demonstrated by Maraston (2005). Her Figure 27 (middle left panel) shows an enhanced reddening in V-I after several 100 millions of years, mainly due to AGB stars,. However, the amplitude of the reddening she found is not enough to account for our red shell colours, which are sometimes even redder than the colour. Resolved data on the shell or stream in M 83 also shows significant amounts of AGB and RGB stars (de Jong et al. 2007), but his data are not deep enough to calculate a reliable global colour for this stream.

• The shells contain more dust per unit stellar mass than the main body of the galaxy. Here, the problem is to explain why shells have more dust per unit stellar mass. Several possibilities can be thought of. The first possibility is related to the previous item: if RGB and AGB stars make up a significant part of the population, their large mass loss (mostly AGB stars) will result in more dust residing in the shells (Athey et al. 2002). Another possibility is based on theory: it has been shown that it is possible for gas or dust to remain connected with the shell stars after a small merger (Kojima & Noguchi 1997; Charmandaris & Combes 2000). Observational evidence for significant amounts of dust residing in 92 chapter 4: HST/ACS observations of shell galaxies.

a shell was found in NGC 5128 (Stickel et al. 2004). HI gas in shells has been found in M 83, NGC 2865 and NGC 3656 (Schiminovich et al. 1995; Balcells et al. 2001). A third, speculative, idea explaining the presence dust in shells, is that dust is swept up by the shell stars as they pass through the potential of the galaxy. This should be tested using simulations. In the ISO archive, we found ISOCAM (Kessler et al. 1996) data for NGC 1344 and NGC 7626. These observations, taken at wavelengths near 9.5 µm, could in principle detect the warm dust. Although both galaxies show red shells, we find no evidence for enhanced emission at the shell regions in these two galaxies.

4.5.4 Dust in the centres of shell galaxies

All of our galaxies show visible dust features, mainly found in the central parts of the galaxy (see Figures A1-F1 in Appendix A-F). Following Tran et al. (2001) the morphologies of the visible dust features can be divided into two groups, i.e.: 1) nuclear ring or disks-like structures and 2) filamentary structures and/or (small) dust patches. All our galaxies show at least features of group 2 (filaments: NGC 474, NGC 2865, NGC 3923 and NGC 7626; small dust patches: NGC 1344, NGC 2865, NGC 3923 and NGC 5982). Although NGC 5982 was listed before as a dust-free galaxy (Sarzi et al. 2006), we see several patches in the residual frames (two dust patches are visible in Figure E.1 at about 6.5” E and N.E.). NGC 7626 shows both a dust lane and a nuclear ring within the inner arcsec (the ring was already reported by Forbes et al. (1995a)). A combination of these two dust features is not seen very often: Lauer et al. (2005) do not find any example in a sample of 77 early type galaxies. Saturation in the core in NGC 2865 and a bad fit of the inner regions in NGC 5982 hinders a conclusion about the presence of nuclear dust rings in these galaxies. The bad fit in NGC 5982 may be related to the presence of a KDC in the inner regions (Wagner et al. 1988; McDermid et al. 2006). When we apply GALPHOT to much lower resolution ground based data, we see similar (bad) results (see Figure 4.4). Lauer et al. are able to make a better fit, however they do not detect the dust patches and shells, probably due to smoothing in their modelling procedure. All dust morphologies are summarised in column two of Table 4.4. Column three of the same table lists the position angles valid for the filaments and nuclear disk. ’Visible dust masses’ were obtained by using the method of van Dokkum & Franx (1995). They assume that the visible dust acts as a foreground screen w.r.t. the background galaxy light, which will result in a lower limit for the dust mass and will have large errors of the order of 50%. The dust mass is derived with the following expression: −1 Md =< AV > ΣΓ (4.4) with < AV > extinction measured at a pixel, Σ the surface area (using distances listed −6 2 −1 in Table 4.1) and Γ = 6 × 10 mag kpc M the extinction coefficient per unit mass. Only areas were selected where the extinction/dust is visible by eye. AV was calculated for each visible dust feature by dividing the real and model images. Values of AV are all lower than 1. All derived dust masses are listed in column 5 of Table 4.4. The masses are of similar magnitude to those found by other authors (van Dokkum & Franx, 1995; Tran et al. 2001). 4.5: General results 93

Galaxy Morph. P.A.dust P.A.gal Mass MV (1) (2) (3) (4) (5) (6) NGC 474 f 0 0 8.3 -21.17 NGC 1344 p 160 0.3 -21.07 NGC 2865 f,p 40 150 4.0 -21.59 NGC 3923 f,p 45 49 3.9 -21.92 NGC 5982 p 105 0.3 -21.91 NGC 7626 f,d 135 0 2.6 -22.16

Table 4.4: Properties of dust. Column 2: Morphology of the dust where d=disk, f=filament, p=patchy dust; column 3 and 4: P.A. in degrees of main dust feature and 3 galaxy respectively; column 5: dust mass in 10 M as determined using the method of van Dokkum & Franx (1995); column 6: absolute V magnitude of galaxy

4.5.5 Dust origin

Currently, there are at least four scenarios which explain the presence of dust in the centres of early type galaxies. The dust survival time, which depends on the main destruction mechanism (sputtering by the hot X-ray gas), is expected to be relatively low in the centres of early type galaxies (107 − 108 yrs depending on the electron density, (Draine & Salpeter 1979; Tielens et al. 1994). A problem with these timescale calculations is that it does not take into account the effect of self-shielding in dust clouds. This may enhance the survival time considerably. The ubiquitous presence of dust in the centres of early type galaxies is difficult to explain without some rate of replenishment. Mathews & Brighenti (2003) showed that it is possible to form dust clouds in the centres of early type galaxies by accumulating dust from stellar winds. Other scenarios use external influences like accretion from flybys or mergers with other galaxies. We will now discuss the dust properties in the shell galaxy sample, where external influences are evident. Using HST archival data, about half of all elliptical galaxies exhibit visible dust features, which are equally present in power-law and core galaxies (Lauer et al. 2005: 47% of 177 in field galaxies). Assuming that a similar dust detection rate of 50% is representative for our galaxy sample and given the fact that dust is visible in all our galaxies, we can reject the statement that our sample belongs to the parent set of normal early type galaxies with visible dust at the 97.5% level. However we should also take into account the possibility that our sample is biased by considering much higher dust prevalence in certain classes of early type galaxies. This happens for instance in radio-loud galaxies, having dust detection probability of about 90% (van Dokkum & Franx, 1995; Verdoes Kleijn et al. 1999). Our sample hosts one : NGC 7626 (Hibbard & Sansom 2003). Second, we consider a possible bias due to the presence of ionized gas. It is well known that dust is almost always accompanied with ionized gas in early type galaxies (Macchetto et al. 1996; Sarzi et al. 2006). Due to selection effects in the detection of visible dust, the converse is not true, although the probability to detect dust in early type galaxies with ionized gas is still quite high (Tran et al. 2001). Modern instrumentation detect emission in about 75% of early type galaxies (Macchetto et al. 1996; Sarzi et al. 2006). Compared with these detection rates, our sample does not seem to be biased: only three galaxies (NGC 474, NGC 5982 and NGC 7626) have low 94 chapter 4: HST/ACS observations of shell galaxies. levels of Hα+[NII] emission (Verdoes Kleijn et al. 1999, 2002) with luminosities below the median value of 2 × 1039 ergs s−1, determined for a nearly complete magnitude limited sample of nearby galaxies (Ho et al. 1997b). Combining these biases and dust detection probabilities still implies that we can reject the statement at the 95% level, and that shell galaxies have a higher dust prevalence than normal early type galaxies, contrary to an earlier finding by Sadler & Gerhard (1985). Number counts of the morphology of the dust features, occurring in our sample, also differs from normal early type galaxies. Dust features in the centres of early type galaxies come in two types: regular rings and irregular shaped patches or lanes. Number counts give a ratio of 3:5 for the two types of dust features. They are almost never both seen in one galaxy (Lauer et al. 2005), which has been used as evidence for an (episodic) dust settling sequence scenario (Tran et al. 2001; Lauer et al. 2005; Verdoes Kleijn & de Zeeuw 2005). Here, first the irregular dust patches and lanes form in some way, while some time later these dust features move to the centre and form a disk in dynamic equilibrium. All the dust features in our sample seem to be out of dynamical equilibrium as they show up as irregular patches or lanes. The probability for this to happen, assuming a regular to irregular ratio dust feature of 3:5, is 6%. These considerations lead to the conclusion that external influences are responsible for the ubiquitous presence of dust in shell galaxies. We will now briefly discuss the dust properties of the individual shell galaxies and see how they fit into this discussion and shell formation theory. NGC 474 contains several pronounced dust lanes within the inner 15” (see Figure A.1), not detected previously (see e.g. Ravindranath et al. 2001 and Sarzi et al. 2006). 4 It contains the largest visible dust mass of our sample (≈ 10 M ). Sarzi et al. report a misalignment of the stars and ionised gas by 74 ±16 degrees. Comparing the position angle of the dust lanes with the velocity maps of Sarzi et al. show that the dust is aligned with the velocity field of the ionised gas and not with the stars. This probably implies a connection between the dust and the gas and is seen in many other early type galaxies (Goudfrooij et al. 1994; Ferrari et al. 1999; Sarzi et al. 2006). The velocity structure of the gas, the dust morphology of NGC 474 (many dust lanes residing on top of the supposed bulge) and the large offset between stars and gas+dust disturb the picture of this galaxy being an S0, and point towards an external origin for the dust. The well defined type I shell galaxies, NGC 1344, NGC 3923 and NGC 5982, are expected to be created by a minor merger with a small, non-rotating dwarf galaxy. 2 While the dust content of NGC 1344 and NGC 5982 is quite low (several time 10 M , Table 4.4) and corresponds to such a scenario, the dust mass of NGC 3923 is an order of magnitude higher. Most (80%) of the visible dust in NGC 3923 belongs to a large patch, visible in the NE direction at 50 arcsec from the centre (see Figure D.1). This patch was shown to be part of NGC 3923 and also emits small amounts of Hα+[NII] (Pence 1986). At the projected distance of this patch, the electron density (Fukazawa et al. 2006) corresponds to a dust sputtering minimum survival time of 4 × 107 yrs. Another 5% resides in small patches at 4.5 arcsec from the centre (Figure D.1). At this projected distance, the dust sputtering minimum survival time is ≈ 106 yrs. The rest, 15%, is located in a long diffuse dust lane NW from centre. The amount of dust in this galaxy does not conform to the minor merger picture of a small elliptical dwarf galaxy falling into a much larger potential (already noted by Carter et al. 1998). An internal origin for the dust (e.g. ejection from stars) seems to be the most likely scenario. In NGC 2865 and NGC 7626, comparison of simulations and shell structures point 4.6: Summary 95 towards gas rich intruder galaxies (see Section 4.4.3 and 4.4.6 respectively), which has resulted in recent (0.4-1.7 Gyr) star formation in the core of NGC 2865 (Hau et al. 1999) and likely the creation of new globular clusters (Paper I). The dust content of 3 several times 10 M (see Table 4.4) is distributed in diffuse layers and patches near the centre (Figures C.1 and F.1). In NGC 2865 we can again calculate the minimum dust sputtering survival time because the electron density is known (Fukazawa et al. 2006). This results between 3 × 107 and 1 × 108 years for the inner and outer dusty regions respectively. This is again much lower than merger timescale of several 108 to 109 yr.

4.6 Summary

Using observations in V and I with the ACS on board the HST, we analysed the proper- ties of shell systems, in particular their colours, morphologies and dust properties. For most shells listed in this paper, we determined their colour for the first time. For those shells for which their colour had already been determined, we find similar results, giving support to the quality of the result of our methodology. In general we find that colours of shells are similar or redder to the colours of their host galaxies. We attribute the red colour to dust which is physically connected to the shell. In some cases, a different stellar population as a result from a truncation of star formation, may also redden the shells. The only blue shell is found in NGC 474, which is very likely related to a long tail and probably a very recent minor merger event. N-body merger simulations, rather than simulations by the interaction model, describe best the observed morphologies of the shell systems. We detect out of dynamical equilibrium central dust features in all our galaxies. Comparison with a set of ’normal’ elliptical galaxies, which have a dust detection rate of 50%, implies an external origin for central dust found in shell galaxies. However this is in contradiction with theoretical predicted dust survival times. Better models of dust behaviour in centres of early type galaxies, which include self shielding, are needed to solve this problem. The best shell candidate for follow-up spectroscopy has been found in NGC 2865. Innermost shells are found in the type I shell galaxies NGC 3923 and NGC 5982 at a distance about 2 kpc from their centres. These shell have survived for a long time, since both galaxies have relatively old shell systems. Current models to describe the profiles of phase wrapped shells probably work best for young shells. 96 chapter 4: HST/ACS observations of shell galaxies. Appendix 4.A Results for NGC 474 4.A: Results for NGC 474 97 Left: ACS residual image of GALPHOT for NGC 474 in V (202x202 arcsec). Surface brightness of shells were Figure A.1: determined in thedirected wedge in regions the using NS direction ellipticities and as aligned indicated with by the ionized the gas white (to strips. be compared Right: with the Zooming Figures in in Sarzi (40x40 et arcsec). al. (2006)) Dust is 98 chapter 4: HST/ACS observations of shell galaxies.

Figure A.2: Morphological data NGC 474, circles and crosses represent V and I band data respectively. Top left: surface brightness profiles, corrected for background (see text). The black line is a Sérsic fit to the I band surface brightness data right of the vertical dashed line. Middle and bottom left: position angle and ellipticity respectively. Open squares represent WFI data. Top right: Global V-I profile. Top middle and bottom: S4 and C4 respectively. 4.A: Results for NGC 474 99

Figure A.3: Panels with shell V band fluxes in counts (upper parts) and V-I shell colour data points and galaxy colour profiles (lower parts) vs. radius. Fluxes were averaged along wedge using ellipticities as indicated by the white strips as shown in Fig A.1. Shell V-I colours were calculated within the vertical dashed lines. The galaxy V-I colour profile was calculated from a Voronoi binned image. Top-left: Data for wedge covering shells 3b and 9; top-right: Data for wedge covering shells 3a and 6; middle-left: Data for wedge covering shells 7a and 13a; middle-right: Data for wedge covering shells 1, 7b and 13b; bottom-left: Data for wedge covering shells 4, 5 and 11. 100 chapter 4: HST/ACS observations of shell galaxies. Appendix 4.B Results for NGC 1344 4.B: Results for NGC 1344 101 residual image of GALPHOT for NGC 1344 in V with wedges left and right. The field of view is 202 x 202 Inner region residual image of NGC 1344 in V (40x40 arcsec). Dust patches are visible. Right: arcseconds. Figure B.1: Left: 102 chapter 4: HST/ACS observations of shell galaxies.

Figure B.2: Morphological data NGC 1344. Description: see NGC 474, Figure A.3 4.B: Results for NGC 1344 103

Figure B.3: Top-left:NGC 1344 shells 1 and 3 in the wedge on the North side. De- scription: see Figure A.4; Top-right: NGC 1344 shells 8 and 10 in the wedge on the South side. 104 chapter 4: HST/ACS observations of shell galaxies. Appendix 4.C Results for NGC 2865. 4.C: Results for NGC 2865. 105 Inner region Right: Residual image of GALPHOT for NGC 2865 in V with a wedge overlapping a bright shell. A loop is visible in residuals of NGC 2865saturated in pixels. V (40x40 arcsec). Dust is visible in the South direction. Fitting problems in the central region are due to Figure C.1: Left: the NS direction. This morphology was already drawn in Fort et al. (1986). The field of view is 202” x 202”. 106 chapter 4: HST/ACS observations of shell galaxies.

Figure C.2: Morphological data NGC 2865. Description: see NGC 474, Figure A.3

Figure C.3: NGC 2865 shell 3 in the wedge placed at the East side. Description: see NGC 474, Figure A.4 4.D: Results for NGC 3923 107 Appendix 4.D Results for NGC 3923 108 chapter 4: HST/ACS observations of shell galaxies. 4x0ace) eea ml utpthsaevsbe h nems hl iil a lodtce yPiu 18)uigground using (1988) Prieur by detected also 202”. was x detected. visible 202” are shell is shells innermost view inner The of other visible. field more are No The patches data. direction. dust based SW small the Several in arcsec). wedge (40x40 the within lane dust Left: D.1: Figure eiuliaeo APO o G 93i ihwde oetelreds ac tteN n faint and NE the at patch dust large the Note wedge. with V in 3923 NGC for GALPHOT of image Residual Right: ne einrsdaso G 93i V in 3923 NGC of residuals region Inner 4.D: Results for NGC 3923 109

Figure D.2: Morphological data NGC 3923. Description: see NGC 474, Figure A.3 110 chapter 4: HST/ACS observations of shell galaxies.

Figure D.3: NGC 3923 shells 1, 3, 8, 10, 13, 15 and 17 in southern wedge. The lower panel shows shell and local galaxy V-I colours determined within the wedge: the open circles with errorbars are the shell colours. The solid line and open circles represent local V-I profiles from ACS and ground based VLT-FORS2 data respectively. The signature of the ring (the bump between r=40” and 60”) is not visible in ground based data. 4.E: Results for NGC 5982 111 Appendix 4.E Results for NGC 5982 112 chapter 4: HST/ACS observations of shell galaxies. rscns h hl aesaebs iil nteeetoi version. electronic the in visible best are labels shell The arcseconds. Left: E.1: Figure ml utln svsbeo h ao xsi h ieto,amr rnucdsalds ac svsbeN.Tesellabels shell The NE. visible is patch dust small pronounced more a direction, E the in version. axis electronic the major in the visible on best visible are is lane dust small eiuliaeo APO o G 92i .Sel r aeyvsbe h edo iwi 0 202 x 202 is view of field The visible. barely are Shells V. in 5982 NGC for GALPHOT of image Residual Right: ne eiul fNC58 nV(06 rsc.A arcsec). (60x60 V in 5982 NGC of residuals Inner 4.E: Results for NGC 5982 113

Figure E.2: Morphological data NGC 5982.. Description: see NGC 474, Figure A.3 114 chapter 4: HST/ACS observations of shell galaxies. Appendix 4.F Results for NGC 7626 4.F: Results for NGC 7626 115 Inner region residuals of NGC 7626 in V (40x40 arcsec); the two shells are Right: Residual image of GALPHOT for NGC 7626 in V . Two bright shells are visible with the wedges overlapping already outside the field of view, but the dust and a lot of globular clusters are clearly visible. them. The field of view is 202 x 202 arcseconds. Figure F.1: Left: 116 chapter 4: HST/ACS observations of shell galaxies.

Figure F.2: Morphological data NGC 7626. Description: see NGC 474, Figure A.3 4.F: Results for NGC 7626 117

Figure F.3: Top-left: NGC 7626 shell 3 in the wedge at the SW side. Description: see Figure A.4. Top-right: NGC 7626 shell 2 in the wedge at the NE side. Bottom- left: NGC 7626 shell 2 in the wedge at the NE side but now the B band residual flux is drawn in the top panel while B-V colours are drawn in the bottom panel. 118 chapter 4: HST/ACS observations of shell galaxies. Chapter 5 Globular Clusters of Shell Galaxies

ABSTRACT ∗

Shells in Elliptical Galaxies are faint, sharp-edged features, believed to provide evidence of a recent (∼ 0.5 − 2 × 109 years ago) merger event. We analyse the Globular Cluster (GC) systems of six shell elliptical galaxies, to examine the effects of mergers upon the GC formation history. We examine the colour distributions, and investigate differences between red and blue globular cluster populations. We present luminosity functions, spatial distributions and specific frequencies (SN ) at 50 kpc radius for our sample. We present V and I magnitudes for cluster candidates measured with the HST Advanced Camera for Surveys (ACS). Galaxy background light is modelled and removed, and magnitudes are measured in 8 pixel (0.4 arcsec) diameter apertures. Background contamination is removed using counts from Hubble Deep Field South. We find that the colour distributions for NGC 3923 and NGC 5982 have a bimodal form typical of bright ellipticals, with peaks near V − I = 0.92 ± 0.04 and V − I = 1.18 ± 0.06. In NGC 7626, we find in addition a population of abnormally luminous clusters at MI = −12.5. In NGC 2865 we find an unusually blue population, which may also be young. In NGC1344 and NGC474 the red cluster population is marginally detected. The radial surface density profiles are more flattened than the galaxy light in the cores. As already known, in NGC3923, which has a high SN of 5.6, the radial density distribution is more shallower than the diffuse galaxy light. The clusters in NGC 2865 and NGC 7626 provide evidence for formation of a population associated with a recent merger. In the other galaxies, the properties of the clusters are similar to those observed in other, non-shell, elliptical galaxies.

5.1 Introduction

In current galaxy formation models, most ellipticals have already formed at z > 2 (Ellis et al. (1997); Peebles (2002); van Dokkum et al. (2004)). It is unclear whether all globular cluster (GC) systems were also formed at this early epoch or if substantial

∗ Published as Sikkema, Peletier, Carter, Valentijn, Balcells, 2006, A&A, 458, 53 120 chapter 5: Globular Clusters of Shell Galaxies numbers are still forming today. An important diagnostic is the existence of bimodality in the colour distribution of GCs, present in many early type galaxies (Zepf & Ashman (1993), Whitmore et al. (1995)). Generally this is explained as being due to metallicity differences indicating two or more populations of GCs. Several theories have been proposed to explain the origin of bimodality: Merger scenarios (Toomre (1977); Schweizer (1987); Ashman & Zepf (1992)) in which the metal-rich GCs were created in gas rich mergers. Since most star formation occurred at early epochs, this means in general that the metal-rich GCs are also old. However, this scenario also suggests that GCs can still be forming today in mergers. This is supported by observations of young cluster-like objects in current mergers in action or possible merger remnants like NGC 4038/39 (Whitmore & Schweizer (1995); Whitmore et al. (1999)), NGC 3921 (Schweizer et al. (1996)), NGC 7252 (Miller et al. (1997)), NGC 1316 (Goudfrooij et al. (2001b); Goudfrooij et al. (2001a); Goudfrooij et al. (2004)), NGC 1700 and NGC 3610 (Whitmore et al. (1997)). Two others models explaining bimodality are the accretion model (Cote et al. (1998)) and the multiphase formation model (Forbes et al. (1997); Harris et al. 1998). In both models all GCs are old. The first model produces bimodality by accreting and mixing metal poor GCs from dwarf galaxies with the more metal rich GCs of the massive host galaxy. Cannibalism by our own galaxy of the Sagittarius and Canis Major dwarf galaxies and their clusters (Ibata et al. (1995); Forbes et al. (2004)) and observations of large numbers of dwarf galaxies around giant galaxies are cited as supporting this scenario. The second model explains the bimodality as the result of two phases of GC formation in the initial collapse and formation of a galaxy. The metal poor clusters, and a small proportion of the stars form in the initial gravitational collapse, then the metal rich clusters and the bulk of the stellar component form from enriched gas in a second collapse phase about one or two Gyr later. Strader et al. (2004) argue that this “in situ” model of GC formation is in better agreement with their observations of the correlation of the colours of the metal-poor populations with galaxy luminosity. A combination of these different scenarios is used in the hierarchical merging model of Beasley et al. (2002), who undertook semi-analytical simulations of GC formation. In this model the metal-poor GCs are old and formed in cold gas clumps, the metal- rich ones are created later in merger events. In the hierarchical build up of galaxies, accretion of GCs will also take place. These simulations are able to reproduce the many variations in the colour distributions of GC systems observed in elliptical galaxies as well as the observed L − Ntot relation. Recently, Yoon et al. (2006), showed that the apparent bimodality in globular cluster colours not necessarily implies a bimodal metallicity distribution. The nonlinear nature of the metallicity-to-colour transformation could cause a single old population with a unimodal metallicity distribution to look bimodal. This model is attractive because it gives a very simple explanation for the observed distributions and could simplify theories of elliptical galaxy formation. However, the observations of recent GC formation, as mentioned above will likely sometimes disturb the predictions made by this model. There are many examples of multi-colour and spectroscopic data for the GC systems of ’normal’ elliptical galaxies, and these generally give old ages for both blue and red populations (M49: Puzia et al. (1999), Cohen et al. (2003); NGC 1399: Forbes et al. (2001a); M87: Cohen et al. (1998); NGC 1052 and NGC 7332: Forbes et al. (2001b); and in a sample of early-type galaxies: Strader et al. (2005)). This is in contrast to studies of the well-known bimodality of the colour distribution of the GCs of the 5.1: Introduction 121

LMC (Gascoigne & Kron (1952); van den Bergh 1981, 1991) which is primarily an age effect, understood in terms to the evolution with time of clusters in the colour-colour diagram (Frenk & Fall (1982)), possibly combined with a relationship between age and metallicity (Battinelli & Capuzzo-Dolcetta (1989); Girardi et al. (1995)). In ellipticals, the majority of GCs formed at high redshift (z>2.5), whichever formation mechanism is dominant. However it is important to examine evidence that recent merger events can produce enhanced populations of young clusters, as this process may have been much more important in the early universe. This study focused on a sample of elliptical galaxies with faint stellar shells in their envelopes. Shells and ripples in elliptical galaxies (Malin & Carter (1980); Malin & Carter (1983); Schweizer & Seitzer (1992)) are faint, sharp edged stellar features in the envelopes of these galaxies which are the remnants of the stellar components of minor mergers in the comparatively recent history of the galaxy (Quinn (1984)). Typical dynamical ages of the shell systems are ∼ 0.5 − 2 × 109 years (Nulsen (1989); Hernquist & Quinn (1987)). If we can detect a population of clusters then the age of this population will provide an independent estimate of the age since the merger, assuming that the same event produced both the shells and the young clusters. In NGC 2865, one of our sample, Hau et al. (1999) find that the age of a nuclear starburst model for the young stellar population in the core of this galaxy is much older than the dynamical age of the shells, although they do find a better correspondence between ages for a model involving truncation of ongoing star formation. In NGC 1316, Goudfrooij et al. (2001b) find a cluster system with an age of 3.0 ± 0.5 Gyr, consistent with the age of the nuclear stellar population. The shell systems of both NGC 2865 and NGC 1316 are complex, Type II or III systems (Prieur (1990)) whereas the dynamical age estimates are more directly applicable to simple phase wrapped, Type I shell systems. Two techniques to investigate possible age differences between different GC popula- tions involve measuring radial density profiles, and globular cluster luminosity functions (GCLFs). Radial density profiles generally show a flattening of the globular cluster den- sity profile near the centre, when compared with the profile of the background galaxy light (Lauer & Kormendy (1986); Capuzzo-Dolcetta & Donnarumma (2001)). Mecha- nisms which could cause a depletion of the cluster population near the centre are dy- namical friction, which causes clusters to spiral in towards the centre (Tremaine et al. (1975), Pesce et al. (1992)), and destruction by tidal shocks as the clusters pass close to the nucleus (Ostriker et al. (1989); Capuzzo-Dolcetta & Tesseri (1997)). This second process operates preferentially in triaxial potentials, and on clusters on radial orbits. If we can identify a population of younger clusters, created during a recent merger, then the density profile might extend further into the centre as the clusters have had less time to disrupt. However, this does depend upon the orbital structures to be the same, if one or other population were on predominantly radial orbits, then this would cause a stronger flattening of the core. The same process can lead to evolution of the mass function of the clusters and hence the GCLF (Fall & Rees (1977), Gnedin & Ostriker (1997), Fall & Zhang (2001)). Lower mass clusters are preferentially destroyed, leading to the well known turnover in the GCLF. A younger population would contain more low-mass clusters, and the GCLF should be closer to the original mass function, which might be a power law. Observations of young cluster systems such as NGC 1316 (Goudfrooij et al. (2001a)) and NGC 7252 (Miller et al. (1997)) do show power law GCLFs. In this paper we analyse the properties of the GC systems of six shell elliptical 122 chapter 5: Globular Clusters of Shell Galaxies galaxies using optical V and I data from the Advanced Camera for Surveys (ACS) on HST. Although the observations were optimised to study the shell structures rather than the GCs, the magnitudes, colours and spatial distribution will indicate whether GC systems of any of the shell galaxies differ from those of normal early type galaxies. We will also provide data on three GC systems which have not been studied before: NGC 474, NGC 1344 and NGC 2865. The last galaxy is known as a recent merger remnant (Hau et al. (1999); Schiminovich et al. (1995)). The GC systems of three of our galaxies have been studied before (NGC 3923, ground-based (Zepf et al. (1994); Zepf et al. (1995)); NGC 5982 and NGC 7626, WFPC2 on the HST (Forbes et al. (1996)). The larger field of view and higher resolution of ACS will provide more detections and more complete knowledge of these GC systems. In Section 2 and 3 the observations and data reduction are described, which include the detection and selection of the GCs, calculation of completeness levels and photom- etry of the globular clusters. Section 4 presents the V-I colour distributions and spatial distributions of the globular cluster systems. Section 5 describes the globular cluster luminosity function which is used in Section 6, where the specific frequencies are calcu- lated. Finally in Sections 7 and 8 we discuss the results and present the conclusions.

5.2 Observations and Data Reduction

The six shell galaxies (see Table 5.1) were observed with the ACS_WFC camera be- tween July 2002 and January 2003 with the filters F606W (V) and F814W (I), with CR_SPLIT=2. The ACS camera contains two CCDs of 2048 x 4096 pixels, each pixel having a size of 0.04900 pixel−1 resulting in a field of view of 20200 x 20200. Exposure times were on average 1000s, see Table 5.2. The inner 8 pixels of NGC 474 and the inner 24 pixels of NGC 2865 were saturated in both V and I. Standard reduction was carried out in the IRAF+STSDAS∗ environment, using the packages CALACS and PyDrizzle. These are provided by the Space Telescope Science Institute (STScI). CALACS processing includes bias and dark subtraction, removal of the overscan regions, flat fielding and cosmic ray rejection. The default pipeline is not efficient at removing cosmic rays when the image is filled by a large galaxy. To solve this problem we changed the value of CALACS pipeline parameter SCALENSE from 0.3 to 0.0. Setting SCALENSE to 0.0 increases the probability of removing good stellar data†. However, this is only true for empty fields, which is not the case for our data. The resulting images were further processed by PyDrizzle, which removes the geometric distortion of the ACS optical configuration. Finally, after drizzling the images, the IRAF package LA_COSMIC (van Dokkum (2001)) was used to remove any remaining cosmic rays, which still affected several hundreds of pixels in each image. The standard ACS photometric calibration was used (Sirianni et al. (2005)) to obtain Johnson V and Cousins I magnitudes. The following transformation formulae were applied:

2 VJ = m(F 606W ) + 26.331 + 0.340 ∗ (V − I)JC − 0.038 ∗ (V − I)JC (5.1)

∗ IRAF is distributed by the National Optical Astronomy Observatories, which are operated by the Association of Universities for Research in Astronomy, Inc., under cooperative agreement with the National Science Foundation. † http://www.stsci.edu/hst/acs/documents/newsletters/stan0301.html 5.2: Observations and Data Reduction 123

Galaxy RA (J2000) DEC(J2000) l b AV AI mag mag (1) (2) (3) (4) (5) (6) (7) NGC 474 1h20m06s.7 +03◦2405500 136.80 −58.68 0.11 0.07 NGC 1344 3h28m19s.7 −31◦0400500 229.07 −55.68 0.06 0.04 NGC 2865 9h23m30s.2 −23◦0904100 252.95 +18.94 0.27 0.16 NGC 3923 11h51m01s.8 −28◦4802200 286.53 +33.32 0.27 0.16 NGC 5982 15h38m39s.8 +59◦2102100 93.10 +46.92 0.06 0.04 NGC 7626 23h20m42s.3 +08◦1300200 87.86 −48.38 0.24 0.14

Galaxy type mV m-M d Dn-σ σ mag mag Mpc Mpc km/s (1) (8) (9) (10) (11) (12) (13) NGC 474 E? 11.39 32.56 † 32.5 - 164 NGC 1344 E5 10.41 31.48 22.1 ‡ 20 187 NGC 2865 E3-4 11.30 32.89 37.8 25 230 NGC 3923 E4-5 9.88 31.80 20.0 ‡ 21 249 NGC 5982 E3 11.20 33.11 † 41.9 41 240 NGC 7626 Epec 11.25 33.41 † 48.2 46 270

Table 5.1: Properties of six shell galaxies. Data in columns 2-9 from Roberts (1991). 1st column: Galaxy name; 2nd and 3rd column: Right Ascension and Declination; 4th and 5th column: galactic longitude and latitude; 6th and 7th column: extinction coefficients in V and I magnitudes from Schlegel et al. (1998); 8th column: Morpho- logical type (throughout this paper we assumed that NGC 474 is elliptical and not S0 (Hau et al. (1996)); 9th column: total apparent V magnitude; 10th column: SBF distance modulus in I, corrected for extinction. (Tonry et al. (2001)). †: distance moduli taken from Roberts et al. (1991) determined from HI velocity data, corrected for galactic rotation and restframe of the Local Group using H0=75km/s/Mpc. 11th column: distances adopted in this paper using column 10 except ‡: GCLF distances for NGC 1344 and NGC 3923, calculated in Section 5.2.; 12th column: Dn-σ distances from Faber et al. (1989); 13th column Central velocity dispersion σ; typical errors are 10 km/s (HYPERLEDA4) : d(Dn-σ) for NGC 2865 is probably wrong, since NGC 2865 exhibits a central depression in σ due to a rotating disk (Hau et al. (1999)). Ex- trapolating his data gives σ=230km/s (d(Dn-σ)=34.7Mpc), much larger than the value used by Faber et al. (σ=168km/s). 124 chapter 5: Globular Clusters of Shell Galaxies

2 IC = m(F 814W ) + 25.496 − 0.014 ∗ (V − I)JC + 0.015 ∗ (V − I)JC (5.2)

Here m(F606W) and m(F814W) are ACS Vega instrumental magnitudes and V and I are in Johnson and Cousins systems respectively. The Johnson and Cousins magnitudes were corrected for galactic extinction using the values of Schlegel et al. (1998), which are listed in Table 5.1.

5.3 Data Analysis

In this Section we describe how the globular cluster source catalogues were obtained and describe their characteristics in terms of photometric errors and completeness, which are used in the further analysis of the data.

5.3.1 GALPHOT

Information about the morphology of the galaxies was obtained by using the ellipse fitting task GALPHOT (see Jørgensen et al. (1992)); it returns information such as ellipticity, position angle, surface brightness and the C3,C4,S3,S4 coefficients (Carter (1978)), all as a function of radius. A galaxy subtracted residual image is also returned, which we used to extract the GCs. In the GALPHOT processing, background galaxies, point-like objects, dust lanes, bright pronounced shells and additional bad data were masked out in an iterative way. Remaining faint shell structures, having a brightness typically not more than 5% of the galaxy light, do not severely affect the final results. The best fits were obtained by allowing the center, position angle and ellipticity to be free parameters. In two cases: NGC 2865 and NGC 5982 the central regions could not be subtracted in a proper way, these regions correspond to circles with diameters 400 for NGC 2865 and 2400 for NGC 5982 respectively. Note that the inner 0.300 is saturated in the center of the latter galaxy. Light profiles were obtained by plotting the surface brightness for each fitted ellipse as function of radius. The outer parts of the light profiles are severely affected by uncertainty in the determination of the background. It is difficult to determine a reliable background values from the ACS images themselves, since the galaxies fill the whole field. Fortunately, for five galaxies we found optical wide field data in the R band in the ESO archive of [email protected]. The WFI camera has a field of view of 340 x 330, much larger than the galaxies. We used the ASTRO-WISE system ∗ (Valentijn & Kuijken (2004)) to reduce the WFI images. After subtracting a constant background value from the WFI images, GALPHOT was applied to them. The ACS background values were determined by matching ACS light profiles to the WFI light profiles. For the galaxy without WFI data, NGC 5982, we extrapolated from the outer points of the image using a de Vaucouleurs (1948) r1/4 law. The calculated values for the backgrounds are listed in Table 5.2. The final GALPHOT results describing the morphology of the six galaxies are presented in Chapter 4).

∗ www.astro-wise.org/portal 5.3: Data Analysis 125

(1) (2) (3) (4) (5) (6) (7) Galaxy exp. V (s) exp. I (s) 80%V (mag) 80%I (mag) bg. V (cnts) bg. I (cnts) NGC 474 1140 960 25.74 24.78 203 129 NGC 1344 1062 840 25.81 24.76 94 47 NGC 2865 1020 840 25.85 24.78 146 109 NGC 3923 1140 978 25.59 24.47 170 103 NGC 5982 1314 1020 26.08 24.80 124 182 NGC 7626 1140 960 25.96 24.86 119 80

Table 5.2: Observational characteristics. 1st column galaxy name, 2st and 3nd column: V and I exposure times in seconds, 4th and 5th column: 80% completeness levels in V and I, 6th and 7th column: adopted background value in counts

5.3.2 Globular cluster candidates

The galaxy subtracted, residual images were used for source detection and photometry of globular cluster candidates (GCCs). SExtractor (Bertin & Arnouts (1996)) was used for this purpose. The keyword BACK_FILTERSIZE was set to 5; the keyword BACK_SIZE was set to the relatively small value of 24, to account for background variations caused by broad shell regions, diffracted light from bright stars, and dusty regions.

Figure 5.1: Left: completeness-histogram for NGC 3923 I band image, using the gc-type object. Vertically: completeness ratio; horizontal scale: instrumental ACS magnitudes. Dashed lines indicate 80% and 50% completeness levels. Right: Lower part: Photometric errors for artificial pointlike objects in the GALPHOT NGC 3923 I band residual image. Upper part: RMS errors in magnitudes in each bin; bottom: Off-sets from STARLIST artificial objects using an aperture of 8 pixels.

§ http://www-obs.univ-lyon1.fr/hypercat 126 chapter 5: Globular Clusters of Shell Galaxies

5.3.3 Photometry Because we are analysing point like objects, the magnitudes are measured in fixed apertures, and we tested apertures with diameters of 4,6,8,10,12,14,16,18,20 pixels. The quality of the photometry was analysed by adding objects to the residual images using the IRAF task MKOBJECT (the object was extracted from an ACS image and is a typical PSF object with a few thousands counts at its peak). We did this 5 times, each time introducing 350 objects with a uniform magnitude distribution and positions (generated by IRAF/STARLIST. Catalogues were extracted by using SExtractor again with the same parameters as before. The resulting catalogues in V and I were then associated, keeping only associations within positional error ellipses of 1xFWHM=2.4 pixels. Comparing input STARLIST data with output data gives information about photometric errors and completeness levels.

5.3.4 Completeness We define completeness as the ratio of recovered objects divided by the original number of added objects measured in a particular magnitude bin. Figure 5.1 shows a typical result of this procedure for the I band residual image of NGC 3923. On the horizontal axis are the input STARLIST magnitudes. The completeness is plotted on the vertical axis. The horizontal lines indicate 50% and 80% completeness levels. The results for each image are shown in Table 5.2.

5.3.5 Photometric errors and aperture selection The photometric errors were obtained by subtracting input STARLIST magnitudes from output aperture magnitudes, where extreme outliers (mainly due to false associations) in each magnitude bin were removed using 5x sigma clipping. Typical results, using the NGC 3923 I band residual image and an 8 pixel aperture, are shown in Figure 5.1. The upper panel shows RMS errors in magnitudes in each bin; off-sets from STARLIST magnitudes are plotted in the lower panel. The magnitude errors depend on the different apertures in various ways:

• The rms-errors at the 80% completeness levels increase for larger apertures: from 0.08 mag using a 4 pixel aperture, to 0.18 mag using a 20 pixel aperture. • The measured magnitudes show systematic off-sets for each aperture; larger aper- tures give smaller offsets. We assume that the offsets of the real data are the same in all of our twelve images. • The off-sets vary within each magnitude bin; however the variations are larger for large apertures.

Considering these three error sources leads to a optimum aperture diameter of 8 pixels, which was used in all remaining analysis. The photometric error is ≈ 0.10 mag. near the 80% completeness levels and the aperture correction is 0.26 mag. Figure 5.1 shows an example using this aperture for the NGC 3923 I band residual image. Before doing the photometry, we checked whether globular clusters are resolved in our galaxies. Table 5.1 shows that our closest galaxies, NGC 1344 and NGC 3923, have a distance of about 20 Mpc. At this distance, typical galactic globular clusters, which 5.3: Data Analysis 127 have a half total light diameter of 6 pc (van den Bergh (1994)), will have an angular size of somewhat more than 1 ACS pixel. We conclude that most globular clusters will be unresolved.

5.3.6 Selection of GCCs We made use of various SExtractor keywords to select the globular cluster candidates. We kept all objects which have ELONGATION < 1.4 (keep round looking objects), 1.8 < FWHM_IMAGE < 5 (removing most spurious objects and extended objects) and FLAGS > 0 (removing objects which are blended, saturated or placed near other bad pixels). Objects fainter than the 80% completeness levels were also removed. The remaining lists of objects in V and I were associated using positional error ellipses of 1×FWHM=2.4 pixels, which again removed many objects. The final source list was further cleaned by visual inspection of the residual images, thereby removing any false data, for instance bright galaxy cores, sources located near borders or detections in spikes of bright stars. Figure 5.2 shows the distribution of the remaining objects. Before using these data, we note that the source catalogue is still contaminated by two sources: foreground stars and background galaxies. Below, an estimate of these numbers is made.

1. Galactic foreground stars. The number of foreground stars depends upon galactic coordinates. Table 5.1 shows that NGC 2865, with the lowest galactic latitude of 19◦, is likely to be affected most by foreground stars. We estimated the number of stars in each field by using The Besançon model (Robin et al. (2003)). We used the following input for the model:

• An error polynomial = a polynomial was fitted to our rms-error curve (top panel in Fig. 5.1) and given as input. • Magnitude limits = our 80% completeness levels • Field of view = ACS field size • Galactic coordinates.

The model returns catalogues with the expected number of stars and their V and I magnitudes. Generally we find that the number of stars is negligible com- pared with the number of GCs. A more detailed discussion of the effect of this contaminant is given in section 4.1.

2. Unresolved background galaxies. An estimate of the number of these contaminants was made by using the Hubble Deep Field South (Williams et al. (1996)). The images are publicly available and were observed in the same passbands as our data. Since the HDFS images are much deeper than our images, we dimmed the HDFS in order to reach our 80% completeness levels. We did this by dimming the HDFS V and I images with 2.8 and 2.6 magnitudes respectively, by multiplying with 10−<∆mag>/2.5. To restore the noise levels of the original images, we added Gaussian noise. Next, SExtractor was applied with the same parameters as before, selecting objects using the same selection criteria as before and then associating the objects. We find 15 sources within 0.55 < V − I < 1.45, which are evenly spread in V-I. This number, together with the number of foreground stars, will 128 chapter 5: Globular Clusters of Shell Galaxies

ngc474 ngc1344 ngc2865 4000

3000

2000

1000

0 ngc3923 ngc5982 ngc7626 4000

3000

2000

1000

0 0 1000 2000 3000 40000 1000 2000 3000 40000 1000 2000 3000 4000 X [pixel] X [pixel] X [pixel]

Figure 5.2: Location of globular clusters, using ACS pixel coordinates (pixelsize = 0.05 arcsec; the open square represents the center of the galaxy. 5.3: Data Analysis 129

Figure 5.3: Histograms and colour diagram of V-I colours of the globular cluster candidates. All histograms are better fit by two Gaussians than one, but for NGC 1344 and NGC 474 we plot only the blue component as a Gaussian fit, as the numbers in the red peak are not significant. For the other four galaxies we plot the two Gaussians fit by the KMM algorithm. Residual histograms (thick lines) represent the difference between the data and the sum of the Gaussian fits. Vertical, dashed lines represent the separation of the data into red and blue groups at (V − I) = 1.05.

be used in Section 4.4 to estimate a contaminating background density valid for each galaxy. 130 chapter 5: Globular Clusters of Shell Galaxies

Galaxy µblue µred ass. bl. ass. rd. P-value (1) (2) (3) (4) (5) (6) NGC 474 0.90 1.22 224 34 0.035 NGC 1344 0.92 1.21 268 49 0.004 NGC 2865 0.85 1.12 192 111 0.039 NGC 3923 0.94 1.16 253 390 0.036 NGC 5982 0.96 1.24 240 221 0.000 NGC 7626 0.92 1.13 455 448 0.039

Table 5.3: Output of KMM algorithm (explanation: see Section 4.2) with blue and red V-I peaks (columns 2, 3), KMM assignments of GCs to blue and red groups (columns 4, 5) and P-value (column 6), See also Fig. 5.3.

Galaxy log(r) αred err αblue err αall err αgalaxy err (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) NGC 474 1.2-2.2 - - - - -1.50 0.14 -1.87 0.03 NGC 1344 1.5-2.2 - - - - -2.02 0.17 -2.23 0.02 NGC 2865 1.35-2.15 -3.05 0.24 -2.04 0.17 -2.28 0.14 -2.04 0.01 NGC 3923 1.5-2.1 -0.87 0.15 -0.86 0.15 -0.90 0.10 -1.59 0.03 NGC 5982 1.4-2.1 -1.98 0.18 -1.22 0.19 -1.68 0.13 -2.23 0.04 NGC 7626 1.5-2.2 -1.99 0.14 -1.42 0.13 -1.67 0.09 -2.16 0.04

Table 5.4: Slopes α and errors of the surface densities of red, blue (based on a split at V-I=1.05) and all GCs (columns 3-8), determined applying a weighted least squares method on ranges listed in column 2. Columns (9-10): slope + error of galaxy light. Surface densities of all, red and blue groups are shown in Figures 5.7, 5.8 and 5.9

5.4 V-I distributions and spatial distributions

5.4.1 V-I distributions To appear in our histograms an object must be detected in both passbands, so to avoid colour bias in the selection, we selected only objects from our source catalogue which are 0.3 magnitudes brighter in both V and I than the respective 80% completeness limits. Figure 5.1 shows that we reach 100% completeness at these brighter magnitude limits. Another advantage of doing this, is that the photometric errors are smaller at these limiting magnitudes (0.07 magnitude instead of 0.10). The final V-I colour distributions are depicted in the panels of Figure 5.3. The upper part of each panel shows the distribution of objects in the colour - absolute magnitude plane. Here, the vertical axis is absolute I-band magnitude, which was determined using the distances listed in column 11 of Table 5.1. The lower part shows V-I histograms with binsize 0.075 mag. In these panels the upper, thin-line histogram represents the data, the lower, thick-line histogram represents the residuals from the fits described in Section 5.4.2. The expected number of stars in the Besançon catalogues is very low compared to the number of GCCs and will not influence any conclusion drawn from the figures. In the NGC 2865 data, the galaxy at the lowest galactic latitude, there is a small bump 5.4: V-I distributions and spatial distributions 131 near V-I=1.6, which is also expected from the Besançon model. The curves and thick lines are Gaussian fits to the histograms and their residuals respectively (see Section 5.4.2).

5.4.2 Components of the colour distributions. Inspection of the histograms in Figure 5.3 suggests that the colour distributions of GCs for NGC 3923 and NGC 5982 have the bimodal form which is normal for bright ellipticals, with blue and red peaks near (V − I) = 0.92 and (V − I) = 1.18 respectively. This bimodal form is parameterised conventionally as the sum of two Gaussians. In section 5.7 we investigate the ways in which the other four histograms differ from this form. A check for colour bimodality was made by applying the KMM algorithm (Ashman et al. (1994)) and DIP test (Gebhardt & Kissler-Patig (1999)) on data points between 0.55 < V − I < 1.45. KMM was used in standard mode (fitting two Gaussians with equal σ’s). It returns the value P, which indicates if a distribution is better characterised by a sum of the two Gaussians than a single Gaussian. Table 5.3 lists the output which consists of peak values, number counts in each Gaussian and the P-value. The KMM analysis shows that all of the colour distributions are better fit by a double than a single Gaussian, but in NGC 1344 and NGC 474 the numbers of clusters in the red (metal-rich) Gaussian are too small to be statistically significant, so we plot in Figure 5.3 only the blue Gaussian for these two galaxies. For NGC 7626 and NGC 2865 neither a single nor double Gaussian provides a good fit to the histogram, and the structure is more complex. However we do show in Figure 5.3 the blue and red Gaussians generated by the KMM algorithm. The obvious colour bimodality for NGC 3923 and NGC 5982 is confirmed by using the DIP test (Gebhardt & Kissler-Patig (1999)). This test calculates the probability of a dip occurring in a supposedly bimodal distribution. Applying this test to the colour distributions of our six galaxies gives significant bimodality for NGC 3923 and NGC 5982 with the dip probability Pdip = 0.99 and Pdip = 0.92 respectively. NGC 3923 and NGC 5982 have been studied before. NGC 3923 has been observed from the ground in the Washington system (Zepf et al. (1995)). Like us, they found a bimodal distribution, which, converting their data into V-I (Forbes & Forte (2001)) and correcting to our extinction scale, give colours of 1.01 (±0.05) and 1.21 (±0.05), an offset of +0.06 with respect to our data. Data on NGC 5982 (Kundu & Whitmore (2001)) also revealed a bimodal distribution; applying our extinction scale to their data gives V-I colours of 0.96 (±0.03) and 1.15 (±0.03). While the blue peak is the same as ours, their red peak is 0.1 magnitude bluer. We attribute this to our much larger GC sample. In the further analysis we distinguish between red and blue sample by cutting the distribution at (V − I) = 1.05, in order to analyse whether these two populations have different properties. If we fit the V-I histogram of NGC 2865 with 2 Gaussians, as we do for NGC 3923, NGC 5982 and NGC 7626, the blue peak extends much further to the blue than in other galaxies. We suggest in section 5.7 that this can be attributed due to a population of young blue clusters, overlapping in colour with the normal old metal-poor population. For NGC 7626 KMM gives maxima at 0.92 and 1.13; the bimodal nature was already observed by earlier, but less deep, observations (Kundu & Whitmore (2001)). However 132 chapter 5: Globular Clusters of Shell Galaxies

Figure 5.4: Globular clusters in V-I (vertical scale) as a function of radius. inspection and the output of KMM show that the simple bimodal form is not a good fit, instead the histogram shows a broad flat peak. The middle of this peak ( (V −I) ∼ 1.0) is filled by exceptionally bright (MI < −11) and probably young GCs (see Section 5.7.3). The colour magnitude diagram shows that these are distributed in several small clumps in, which cannot be explained by random effects: the errors in V-I at these magnitudes (mV ≈ 23.0) are at most 0.04 (Figure 5.1) and smaller than the distances between the clumps. This galaxy shows that the distribution of globular cluster colours is not necessary well represented by 2 Gaussians.

5.4.3 Spatial distributions of the globular clusters Figure 5.2 shows the spatial distribution of the GCCs. The centers of the galaxies are represented by an open square. The distribution of GCs roughly follows the ellipticities 5.4: V-I distributions and spatial distributions 133

Figure 5.5: Average colours of globular clusters (circles: all GCs) and galaxy colour (curve through triangles) as a function of radius. NGC 2865, NGC 3923, NGC 5982 and NGC 7626 also show data points for blue and red groups. 134 chapter 5: Globular Clusters of Shell Galaxies

Figure 5.6: Ratio of red vs. blue clusters as a function of radius; NGC 5982 and NGC 7626 have strong gradients. 5.4: V-I distributions and spatial distributions 135

Figure 5.7: Surface density of all globular clusters, starlike symbols show galaxy surface brightness (arbitrary scale). The slope of the GC surface density was calculated between the dashed vertical lines. 136 chapter 5: Globular Clusters of Shell Galaxies of the underlying galaxies. The lack of GCs in the central regions is due to dust, bad fits by GALPHOT (see Section 5.2) or increased noise. Figure 5.4 shows V-I colours of individual GCCs as a function of radius. These colours as a function of radius are plotted in bins in Figure 5.5. Here, triangles, represent the galaxy colour returned by GALPHOT, and circles the average colours of all GCs. For those galaxies for which we can separate the population into red and blue groups (Section 5.4.2) we plot as crosses the mean colours of the red and blue groups respectively. Except for NGC 1344, the average colours of all GCs tend to be bluer at large radii, confirming earlier results (Forbes et al. (1996)), and reflecting the colour gradients in the stellar halos. Where there is a significant red subgroup the cluster colours in that group match quite well the galaxy colours, except in NGC 3923 where they are bluer, as is the case with NGC 1052 (Forbes et al. (2001b)). In NGC 5982 and NGC 7626 the strong blue-ward gradient in the mean cluster colour is caused by a gradient in the relative fractions of the red and blue groups, as illustrated by Figure 5.6, where we plot the ratio of the red to the blue population against radius.

5.4.4 Globular cluster surface densities The GC surface densities were calculated in elliptical annuli; their ellipticity was the average value in the outer regions as determined by GALPHOT. Dusty regions, bad pixels and other bad regions were not used when calculating the effective area of each annulus, from which the surface densities are derived. We accounted for contaminating sources by subtracting a background density, calculated as described in Section 5.3.6. These background densities were between 4.1 to 8.6×10−4 arcsec−2 for the six galaxies. In Figure 5.7 we plot the radial surface density of GCs as a function of radius, the galaxy surface brightness distribution is also shown in this plot. In Figures 5.8 and 5.9, we plot the radial density distribution for the clusters in the red group and blue group respectively, for those galaxies with a significant red population. For all galaxies we note a deficit of clusters in the inner regions, with respect to the background light surface brightness distribution. Possible physical causes of this are dis- cussed in Section 5.7.3, but here we consider the possibility of greater incompleteness in the inner regions causing this deficit. Two effects could contribute to greater incom- pleteness in the inner regions: confusion due to crowding; and the increased photon noise level in the higher surface brightness regions. To test the effect of confusion a low density region of 1000 x 1000 pixels, containing 43 clusters, was cut from the residual image of NGC 7626. As in Section 5.3.3, we introduced artificial objects at various number densities, using uniform and power-law surface density distributions. After using exactly the same detection methods and selection criteria as described in Section 5.3.6, we found that we start losing objects due to confusion effects at densities of 0.15 arcsec−2, where 5% are missed. From Figure 5.7 we see that even the highest density of our galaxy sample (i.e. NGC 7626) is still below this value. The effect of the galaxy surface brightness on completeness is due to the increased photon noise in high surface brightness regions. Within the central 10 arcseconds (13.5 arcseconds for NGC 5982) this leads to increased incompleteness and we do not plot points within these regions. Outside this, the background photon noise from the galaxy is at most 2× the background noise, except for NGC 3923 were the photon noise of the inner data point is about 4× the background photon noise. After again introducing 5.5: Globular cluster luminosity function in I 137 artificial objects and using the same selection criteria as before, we find that this effect is significant only for NGC 3923: the inner two points in Figure 5.7 should probably lie somewhat higher. The radial density distributions of most of the GC systems follow the surface bright- ness distributions of the galaxy light in the outer parts, but show a deficit at small radii which we argue is not due to incompleteness due to confusion or photon noise from the galaxy. This is typical of ellipticals in general (e.g. Lauer & Kormendy (1986); Grill- mair et al. (1986); Puzia et al. (2004); Forbes et al. (2001b); Schweizer et al. (1996) and Brown et al. (2000)). The deficit at the centre is often interpreted as evidence for tidal disruption of clusters passing through the core of the galaxy (Fall & Rees (1977)), but to be effective this process requires the clusters to be on predominantly radial orbits (Grillmair et al. (1986); Ostriker et al. (1989)) NGC 3923 shows a different behaviour, the surface density profile is much shallower than the light profile at all radii (already noted by Zepf et al. (1994)). This is typical of the GC systems of some brightest cluster galaxies, e.g. NGC 4874 (Harris et al. (2000)). However it is unusual for a galaxy such as NGC 3923, which is the brightest in a small group. Comparison of Figures 5.8 and 5.9 shows that there are sometimes differences be- tween the surface density profiles of red and blue GCs. To quantify these differences we fit the linear relation log(σ) = b + α × r by applying a linear weighted least squares fit to the outer points of the density curve. These points were chosen to lie between an inner and outer radius. The inner radius was defined as the point where the flattening stops, estimated by eye. The outer radius excludes unreliable data points at larger radii with very low number statistics (usually containing only 1 or 2 objects in the partial elliptical ring located in the very outer corners of the image). We indicate the inner and outer radius in the Figures as vertical dotted lines. Using the same method and radii we also fitted the galaxy light profile. In columns 7 to 15 of table 5.3 we list the radii used and the results of the fits, including errors on the slopes (we used the standard recipe and error formulae, i.e. equations 15.2.6 and 15.2.9 respectively as listed in Press et al. (1992)) In NGC 7626, the slope of the red GCs is significantly steeper than the slope of the blue GCs. Similar but less significant differences are visible in NGC 2865 and NGC 5982. We checked if these differences are due to the choice of our inner radius by shifting the inner radius one data point to the left as well as to the right; we found no significant change in slope differences. The slopes calculated for the set of all GCs will be used in Section 5.6, where the specific frequency is calculated.

5.5 Globular cluster luminosity function in I

In this section, the GCLFs in the I band are determined. If the observations are deep enough to cover the absolute turnover magnitude (TOM) of MI = −8.46±0.03 (Kundu & Whitmore (2001)), the GCLF can be used as a distance estimator. The GCLF can also be used to estimate the number of globular clusters in a system; this number then determines the specific frequency SN (Section 5.6). In those galaxies with a significant red population, we also compare the GCLF of the red and blue samples, which could be different, for instance if the samples have large M/L differences (Whitmore et al. (2002)). 138 chapter 5: Globular Clusters of Shell Galaxies

Figure 5.8: Surface density of red globular clusters; starlike symbols show the galaxy light (arbitrary scale). The slope of the GC surface density was calculated between the dashed vertical lines. 5.5: Globular cluster luminosity function in I 139

Figure 5.9: Surface density of blue globular clusters; starlike symbols show the galaxy light (arbitrary scale). The slope of the GC surface density was calculated between the dashed vertical lines. 140 chapter 5: Globular Clusters of Shell Galaxies

5.5.1 Determination

We calculated the GCLF in the I band, which is less affected by extinction. The GCLFs were constructed from histograms with a binsize of 0.25 magnitudes, approximately twice the photometric error of the faintest objects. The Besançon catalogues were used to correct for contamination by foreground stars. The foreground stars (see Section 5.3.6) were cumulatively subtracted: i.e. if a bin contains only three sources, while there are an expected number of five stars, we subtracted the remaining two stars from the next bin. We only used objects within the 80% completeness levels and also corrected for incompleteness. The blue and red GCLFs were calculated used the red and blue groups defined in Section 5.4.2. The GCLFs are shown in Figure 5.10, where they are divided in blue and red groups, we find no significant differences between the GCLFs of red and blue populations. The peak in the GCLF of NGC 1344 at I=24.5 is due to red objects, caused either by photometric errors near the faint limit, or more likely a local excess in the background.

5.5.2 GCLF as a distance estimator for NGC 1344 and NGC 3923

If the events which formed the shells also formed large numbers of new GCs, then we might find a difference between the TOM of the GCLF, firstly between red and blue populations, and secondly when compared with the canonical value for normal ellipticals of MIpeak = −8.46 (Kundu & Whitmore (2001)). However we only cover this absolute magnitude for the closest two galaxies, NGC 1344 and NGC 3923. Using the IRAF tool NGAUSSFIT we find for NGC 3923 σ = 1.4 ± 0.1, TOMI = 23.04 ± 0.06, and, ignoring the spike in the GCLF at faint magnitude, for NGC 1344, σ = 1.35 ± 0.10, TOMI = 23.26 ± 0.10. Comparing our GCLF distance estimates for NGC 1344 (m-M=31.72) and NGC 3923 (m-M=31.50) with the SBF distance moduli (from column 10 in Table 5.1: 31.48± 0.30 and 31.80 ± 0.28 respectively), shows that the differences of +0.24 and -0.30 do not exceed the typical errors between these two methods (Richtler (2003)). A PNLF distance estimate for NGC 1344 (Teodorescu et al. (2005)) with m − M = 31.40 ± 0.18 is in excellent agreement with our value. In the remainder we will use our calculated distance moduli for NGC 3923 and NGC 1344.

5.5.3 Total numbers of GCs

A Gaussian fit to the GCLF of the other four galaxies is also necessary to estimate the specific frequency (Section 5.6). Because the magnitude limit is below the TOM of the GCLF, we fit only for σ and amplitude, keeping the distance modulus fixed within the error bars (0.26 mag.). Doing this for NGC 5982 and NGC 7626 give σ of 1.5 and 1.25 respectively, and good fits. For NGC 474 and NGC 2865 we find σ ∼ 1.05, which is unusually small, and conclude that the calculated SN will be uncertain, and that deeper data are required. 5.5: Globular cluster luminosity function in I 141

n474 n1344 n2865 40 60

40 30 40

20 20 20 10

0 0 0

n3923 n5982 n7626 80 150 60

60 100 40 40

20 50 20

0 0 0 20 22 24 20 22 24 20 22 24 I I I

Figure 5.10: Globular cluster luminosity functions in the I band. Dashed and dotted histograms represent the GCLF of the red and blue groups respectively. Gaussian fits are also drawn. Only the GCLFs of NGC 3923 and NGC 1344 extend past the TOM. 142 chapter 5: Globular Clusters of Shell Galaxies 5.6 Total number of globular clusters and specific frequencies

In this section we estimate the total number of clusters and specific frequency SN , which is a measure of the number of GCs per unit galaxy luminosity. It is a strong function of galaxy type and environment: late type galaxies and isolated galaxies usually have lower SN than early type and cluster galaxies (Harris (1991)). The calculations are straightforward using the results of previous sections. The specific frequency is defined by (Harris & van den Bergh (1981)):

0.4(MV +15) SN = Ntot10 (5.3)

where Ntot is the total number of GCs, and MV is the absolute V magnitude of the galaxy, derived from the apparent magnitudes and adopted distances (Table 5.1). The total number of GCs is obtained by using:

Ngauss Ntot = (Nmissed + Nfound) (5.4) Nfound

where Nfound is the detected number of GCs, corrected for incompleteness, Ngauss is the expected number of GCs assuming a Gaussian distribution: 1 √ N = Aσ 2π (5.5) gauss binsize with binsize=0.25; amplitude and σ are from the Gaussian fits from the previous section . Nmissed is the number of GCs expected to be missed due to the fact that the galaxy is much larger than the ACS field of view. This number is calculated by extrapolating and integrating the GC surface density as determined in Section√ 5.4.4 from the outer points (listed in column 7 of Table 5.4), to the point where ab reaches 50 kpc and 100kpc. Error sources used in the calculation of SN are: 1) A photometric measurement error of 0.15 mag. 2) A distance error of 15%. This error is taken from Faber et al. (1989) and represents the scatter in the Dn − σ distance estimator. Comparing our adopted distances with Dn − σ distances (see columns 11 and 12 in Table 4.1) shows that this is a reasonable assumption. The distance error was applied to both Nmiss and absolute V. 3) The errors in the slope of the GC surface density, listed in column 13 of Table 5.4. A possible large error is the unknown value of the slope of the GC surface density outside the ACS field of view. We assumed this slope to be equal to the fitted GC density profile in the outer part of the galaxies. Especially for our two most nearby galaxies this assumption is uncertain; these galaxies are only covered by the ACS to 15-17 kpc, while we extrapolate until 50 kpc and even 100 kpc. Data on the slope in the outer regions of our most distant galaxy, NGC 7626, and surface density profiles in the literature, support this assumption (e.g. Rhode & Zepf (2004); Zepf et al. (1994); Harris et al. (2000)). Finally, for NGC 3923, wide field data (Zepf et al. (1994)) show a constant slope to a radius of 5.6 arcmin ( 34 kpc). Applying a least squares fit to their inner data, covering the ACS fieldsize, gives a slope of −0.82 ± 0.34, comparable to our value (−0.90 ± 0.10). Their outer data gives a slightly steeper slope of roughly −1.14 ± 0.12, which we used to calculate the missed GCs in Equation 4. 5.7: Discussion 143

Galaxy Nfnd Nmiss err. Ngauss err. Ntot err. SN50 err. SN100 err. R (kpc) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) NGC 474 284 143 29 405 56 609 94 2.1 0.5 2.7 1.0 2.5-25.0 NGC 1344 352 93 25 367 28 464 44 1.4 0.3 1.5 0.4 3.4-17.0 NGC 2865 350 35 6 647 98 712 108 1.6 0.4 1.7 0.4 4.1-25.9 NGC 3923 660 1254 162 860 67 2494 286 5.6 1.3 8.3 3.6 3.1-15.4 NGC 5982 505 116 18 1218 161 1497 202 2.6 0.6 3.0 0.9 5.1-25.6 NGC 7626 1051 141 16 2498 263 2833 300 3.9 0.9 4.8 1.4 7.4-37.0

Table 5.5: GC number counts and their estimations from extrapolation. Column 2: Detected GCs; columns 3,4: missed GCs + error; columns 5,6: number of GCs + error using the values for the Gaussians determined in section 5.5.2; Columns 7,8: total GCs + error applying equation (4); columns 9,10: Specific frequency SN within 50 kpc with error; Columns 11,12: SN within 100 kpc with error; Columns 13: ACS range R in kpc where slope has been determined

Assuming our assumption is justified, we adopt the final values for SN , which are listed in columns 6, 7 (50kpc) and 8, 9 (100kpc) of Table 5.5. There sometimes are large variations between SN calculated at 50 kpc and 100 kpc, this especially true for NGC 3923 and is due to the shallow slope of the GC density distribution compared with the luminosity distribution. Harris et al. (1998) also noticed such variations and found a overall 40 overall 40 ratio SN /SN = 1.3, with SN the total SN and SN the SN within 40kpc. It is important to keep these different definitions of SN in mind if one compares SN with other authors. Some authors calculate local SN , others extrapolate to various values ranging between 25kpc and 200kpc. Integrated out to 50kpc, we find values of SN typical for isolated elliptical galaxies, except for NGC 3923 and NGC 7626, for which we find SN = 5.6 and SN = 3.9 respectively. These values are more typical of cluster ellipticals, indeed NGC 3923 has the highest SN of any isolated elliptical (Zepf et al. (1994)). Both of these galaxies are in groups: NGC 3923 is a dominant group galaxy embedded in an X-ray envelope (Buote & Canizares (1998); Pellegrini (1999)); and NGC 7626 is the second brightest member of the Pegasus group and is also detected in X-rays (O’Sullivan et al. (2001)).

5.7 Discussion

In this section a short summary of the properties of the shell galaxies is given; a com- parison with the GC systems of ’normal’ early type galaxies is made; the data are compared with predictions of the hierarchical merger scenario simulations (Beasley et al. 2002); and possible signs for recent GC formation and ages are discussed.

5.7.1 The shell galaxies All of our galaxies are located in low density regions; early type galaxies residing in groups and clusters have a much lower probability of exhibiting shells. (Malin & Carter (1983), Colbert et al. (2001)). As well as the presence of the shells, all six galaxies also show visible dust patches and/or lanes, mostly in the central regions. It has been shown that all galaxies, except for NGC 1344, contain a KDC or show otherwise peculiar 144 chapter 5: Globular Clusters of Shell Galaxies kinematic behaviour (Hau et al. (1996); Hau et al. (1999); Carter et al. (1998); Emsellem et al. (2004); Balcells & Carter (1993)). NGC 474, NGC 5982 and NGC 7626 may be LINERS (Ho et al. (1997a)). Except for NGC 474, unresolved X-ray data are available for all galaxies (O’Sullivan et al. (2001)) and even a 2D X-ray map for NGC 3923 (Buote & Canizares (1998); Pellegrini (1999)). In this paper we assume that NGC 474 and 1344 are E-type galaxies, although there is evidence in each case that they night be classified as S0 galaxies. The environmental and other properties of the galaxies are described briefly below (LGG group numbers from Garcia (1993)):

NGC 474 – brightest galaxy of LGG20 (4 members). The galaxy is connected with the small spiral NGC 470 via a HI tidal bridge (van Gorkom & Schiminovich (1997)). Often classified as S0; shows some rotation (Emsellem et al. (2004)). May be a LINER. NGC 1344 – located at the outskirts of the Fornax Cluster, 4.9◦ from the central cluster galaxy NGC 1399. At this distance the density is comparable to the field galaxy density (Kambas et al. (2000)). Sometimes classified as S0; shows rotation within 2 effective radii (Teodorescu et al. (2005)) NGC 2865 – isolated galaxy (Reda et al. (2004)), remains of a rotating HI disk (Schimi- novich et al. (1995)). Hau et al. (1999) found a KDC and evidence for a young (0.4-1.7 Gyr) stellar population; two possible explanations were given: a starburst or a trunca- tion of the star formation. NGC 3923 – brightest galaxy of LGG255 (5 members) NGC 5982 – brightest galaxy of LGG402 (4 members), may be a LINER. NGC 7626 – second brightest member of the Pegasus group, LGG473, of at least 15 members. This probable LINER has a radio-jet, directed NE, and a small HI cloud between 1.5’ and 3.0’ WSW of the center; the galaxy does not contain any HI tidal fea- tures (Hibbard & Sansom (2003)). The core shows orthogonal kinematics to the main body (Balcells & Carter (1993)), with no emission lines, nor signs of nuclear young populations. A dust lane in direction ENE is visible in the inner 15 ACS pixels of the core.

5.7.2 Comparison of the GC systems with normal ellipticals Our sample is selected on morphological grounds to have undergone a recent minor merger. In this section we investigate the effect of the merger event upon the GC systems. All six galaxies have a peak at 0.94 ± 0.04 in their V-I histograms, in common with ’normal’ ellipticals. In four of our six galaxies we see a distribution of galaxies red-wards of this, this second peak again occurs in the majority of normal ellipticals. All galaxies show a flattening in the GC density profile near the central regions. This feature is also a generally seen in other GC systems and is attributed to disruption processes (Ostriker et al. (1989); Pesce et al. (1992)). The properties of the GC systems of NGC 3923 and NGC 5982 are very similar. Both galaxies have blue and red populations of comparable size, which peak at roughly the same V-I values near 0.95 and 1.20. Using the colours of the GCCs and assuming that the red peak has solar metallicity, the evolution models of Fritze-v. Alvensleben (2004) (to get the of the blue clusters) and Fig. 12 from Goudfrooij et al. (2001b) indicate that these systems are old, > 5 Gyr, and evolved systems. 5.7: Discussion 145

NGC 3923 has been extensively studied by Zepf et al. (1995). They already noted the very high SN for a galaxy located in a low density environment, which we confirm. McLaughlin (1999) proposes that high SN galaxies are probably best explained by taking into account the presence of a extended massive X-ray halo. However, NGC 3923 is probably in contradiction with these results. While NGC 3923 has a very shallow GC surface density profile and also contains a X-ray halo (Buote & Canizares (1998), Pellegrini (1999)), its X-ray luminosity is about a factor 10 lower (O’Sullivan et al. (2001), Fukazawa et al. (2006)) than the galaxies studied by McLaughlin: M49, M87, and NGC1399. The colour distributions of NGC 474 and NGC 1344 (Figure 5.3) are similar, and appear either to be unimodal, or to have a very low red (metal-rich) component. Al- though KMM returns a combination of a blue and red peak as a good description of the distribution, there are very few GCs in the red peak. Such colour distributions are very rare in luminous ellipticals (Kundu & Whitmore (2001)). Peng et al. (2006) show that, in early-type Virgo galaxies, the red peak is much more prominent in the more luminous galaxies. However unimodal, blue, histograms only become common for galaxies fainter than MV = -18 (Figures 4 and 6 of Peng et al.), whereas the absolute magnitudes of NGC 474 and NGC 1344 are MV = -21.17 and -21.07 respectively. They are however the lowest luminosity and lowest velocity dispersion galaxies in our sample. Environmental differences in the early history of the galaxy might lead to differences in the GC (V-I) histograms. In the merger model, fewer early mergers would mean less red peak clusters. Fewer early mergers might also cause a galaxy to retain more of its angular momentum, and thus to be an S0 or a rapidly rotating elliptical. NGC 474 and NGC 1344 both have significant rotation (Emsellem et al. (2004) and Teodorescu et al. (2005) respectively), and both are sometimes classified as S0 galaxies. The most luminous galaxies in the Peng et al. (2006) Virgo sample with unimodal blue histograms are NGC 4660, an elliptical with significant rotation (Emsellem et al. (2004)); and NGC 4340, an SB0. However there are counterexamples in the Virgo sample, such as the rotating ellipticals NGC 4564 and NGC 4697 which have red peak dominated colour histograms. Insufficient galaxies have been studied in sufficient depth to analyse any possible correlation between the rotational properties of ellipticals and their GC colour histograms. Kundu & Whitmore (2001b) find a number of unimodal, blue colour histograms among a sample of S0 galaxies studied with WFPC2 (e.g. NGC 2768), but their sample sizes are small. They do however find that a lower proportion of their sample of S0s are significantly bimodal, than their equivalent sample of ellipticals (Kundu & Whitmore (2001)). Finally we compare these objects with some properties predicted by the hierarchical merger model of Beasley et al. (2002). In these semi-analytical simulations, metal- poor GCs are old and formed in cold gas clumps, while the metal-rich are created in time during merger events. Of course, accretion of GCs also takes place during the hierarchical build-up of galaxies. These simulations are able to reproduce the many variations in the colour distributions of GC systems observed in elliptical galaxies. For instance, in Figure 13 of Beasley et al., a blue-peaked V-I distribution is shown which looks similar to our large single peak V-I distributions of NGC 474 and NGC 1344. His model also roughly reproduces the observed LV − Ntot relation. We plot this relation in the lower part of Figure 5.11 together with our six data points. Our systems seem to have somewhat fewer globular clusters than predicted. This is partly explained by the fact that Beasley et al. do not take into account disruption processes, which might 146 chapter 5: Globular Clusters of Shell Galaxies

’474’ ’1344’

0.5

’2865’

0 ’7626’ ’5982’ ’3923’

-0.5

-1

4

3.5

3

2.5

2 10 10.5 11

Figure 5.11: Top: Ratio of blue and red clusters vs galaxy luminosity. Simulated data (Beasley et al. (2002)) and the six shell galaxies (NGC numbers). Bottom: Total number of clusters vs. luminosity for simulated data and six shell galaxies (large squares) .

reduce the number of clusters by 10%-20%. The upper part of the same Figure shows the logarithm of the ratio between the number of blue and red clusters. NGC 474 and NGC 1344, the two points with lowest luminosity, lie somewhat outside the point cloud due to their low number of red clusters. 5.7: Discussion 147

Galaxy SSP Σ-method YP/NP (Σ2) Shell Dynamical GCs (1) (2) (3) (4) (5) (6) +2.0 NGC 474 7.3−2.4 5.4 ± 1.9 - - Old NGC 1344 4.0 ± 1.0‡ - NP (5.5) 0.5 Old NGC 2865 Old plus 0.1-1.7 † - YP (10.6) - Old plus 0.5-1 +0.5 NGC 3923 2.6−0.6 - NP (10.3) 0.8 − 1.3 Old +1.9 NGC 5982 12.3−2.0 6.8 ± 1.5 YP (6.8) 0.2 Old +4.9 NGC 7626 13.9−2.4 8.0 ± 1.3 NP (1.4) - Old plus 2-5

Table 5.6: Age estimates. Column 2: SSP ages (Gyr) from Denicoló et al. (2005) except †: Hau et al. (1999) and ‡ Kuntschner et al. (2002). Column 3: Ages (Gyr) using fine-structure index Σ method (Schweizer & Seitzer (1992)). Column 4: Old (NP) and Young (YP) systems together with their Σ2 index according to Michard & Prigniel (2004). Column 5: Dynamical ages of type I shell systems using Nulsen (1989). Column 6: Approximate age indications in Gyr from GC (V-I) distribution (this paper)

5.7.3 Possible evidence for recent GC formation in NGC 7626 and NGC 2865 In this section we examine the possibility that the complex (V-I) histograms of NGC 7626 and NGC 2865 provide evidence for recent GC formation, in addition to the old GC population which gives both the red and blue peaks. In NGC 7626, Figure 5.3 shows that the brightest GCs of NGC 7626, with MI ≈ −12.5±0.2 and V-I=1.1, are more than 1.0 magnitude over-luminous with respect to the brightest GCs of the other five galaxies (all at about -11.0), and with respect to its own universal blue population. It is the most luminous galaxy in our sample(MV = −22.16), and as such the GCLF would be expected to extend to the brightest magnitudes, simply because of the larger population at the sparsely populated faint end of the GCLF. We compare the GC population of NGC 7626 with that of NGC 4472 (Rhode & Zepf (2001)), a more luminous galaxy (MV = −22.7) in the Virgo cluster. NGC 4472 has its brightest clusters in the blue (universal) peak, at R=19 (equivalent to MI = −12.5). The brightest clusters in NGC 7626 are about 0.2 mag brighter and 0.2 mag redder in (V-I) than this, in a less luminous galaxy. This group of bright clusters appears in the GCLF of NGC 7626 (Figure 5.10) as a small excess near I = 21. We can compare this population with that found by Whitmore et al. (1997) in NGC 3610, a dynamically young elliptical. Their Figure 15 illustrates the evolution of a young, metal-rich population compared with an old, metal-poor population. In time, the young, metal rich, GC population will fade in luminosity and will become redder. After three Gyr, this population has become redder than the old population but still has several GCs which are brighter than the brightest old metal poor GC. This is exactly what is visible in our colour magnitude diagram. Whitmore et al. quantify the age differences using the ∆(V − I) vs. ∆V10 diagram (see their Figure 18). This diagram represents an age sequence by plotting vertically the difference in V-I between the peaks and horizontally the magnitude difference between the 10th brightest globular cluster in the young and old populations. In NGC 7626 ∆V10 is extremely difficult to estimate, because the young population is dominated in number not just by the old metal-poor population, but by an old, red, metal-rich population as well. Replacing ∆V10 by an estimate of the difference between the magnitude of the brightest clusters 148 chapter 5: Globular Clusters of Shell Galaxies and the brightest in the blue peak, we estimate a very tentative merger age of 2 - 5 Gyr. We conclude that NGC 7626 appears to have some young GCs, probably created in a recent (2-5 Gyr old) minor merger. These young GCs are superimposed upon a much richer, bimodal, old cluster population. For all galaxies with a substantial red population, we find significantly steeper slopes for the red clusters. This is reflected in Figure 5.6, where gradients are seen in the ratio of red to blue clusters as a function of position: the red GCs are more centrally concentrated than the blue GCs. Similar effects are also seen in other early type galaxies (NGC 1407: Forbes et al. (2006); NGC 4649 Forbes et al. (2001b); NGC 1399: Dirsch et al. (2003); NGC 4636: Dirsch et al. (2005) and others). Since the converse situation is never or rarely seen, this must reflect some important difference between blue and red clusters. This difference will be related to different formation processes, combined with disruption, which will affect populations differently depending upon their orbital structure. Less radial orbits for the red GCs might explain this, but without kinematic data on large samples of GCs nothing conclusive can be said about the cause of these differences. NGC 2865 shows all signs of a recent merger event: a very luminous shell and a KDC with evidence of a recent, 0.4-1.7 Gyr, starburst (Hau et al. (1999)), and an HI disk (Schiminovich et al. (1995)). The (V-I) histogram of the GCs of NGC 2865 (Figure 5.3) is more complex than a simple bimodal structure, with a population of very blue, not particularly luminous, GCs near (V-I) = 0.7. The colour of this population is consistent with an age in the range 0.5 - 1 Gyr (Whitmore et al. (1997)), consistent with the nuclear starburst age of 0.4 - 1.7 Gyr (Hau et al. (1999)), but the luminosity of the brightest clusters is much fainter than predicted. This could be attributed to the small number of clusters in the young population, or else to different physical conditions imposing a different Globular Cluster Mass Function. An alternative hypothesis is that the structure in the colour-magnitude diagrams of the GC systems of NGC 2865 and NGC 7626 is entirely due to metallicity variations. In the case of NGC 2865 this would require a population of old clusters with (V −I) ∼ 0.7, which is too blue for an old population at any metallicity (e.g. Lee & Carney (2002)). In NGC 7626 an old population of intermediate metallicity could fill in the dip in the CM diagram, but it would have to have a very unusual luminosity function to produce the numbers of bright clusters that we see.

5.7.4 Ages and minor mergers. In this Section we compare ages derived for the stellar population of the galaxy; for the shell-forming merger events; and for the young GCs that we argue are present in NGC 2865 and NGC 7626. Stellar ages derived by comparison with Single Stellar Population (SSP) models are uncertain for a number of reasons. First, galaxies are clearly not SSPs, and merger remnants in particular will have at least three episodes of star formation, corresponding to the two progenitors and to the merger induced star formation event. Second, the degeneracy between age and metallicity, combined with uncertainty in isochrone models, and assumptions made about other parameters of the stellar population, such as the mass function, render SSP ages very uncertain (Poggianti et al. (2001b)). Third SSP ages are derived from nuclear spectra only, which can be unrepresentative of the galaxy as a whole (e.g. Proctor et al. (2005)). Nevertheless SSP 5.8: Conclusions 149 ages have been computed for four of our sample by Denicolo et al. (2005), for NGC 1344 by Kuntschner et al (2002), and for NGC 2865 a starburst age has been derived by Hau et al. (1999), and these are listed in Table 5.6. Schweizer & Seitzer (1992) determine a Fine Structure Index, based upon the galaxy morphology, and derive an empirical correlation between this and the stellar age. The Fine Structure Parameter measures the age since the galaxy as a structure was built up, and this is different from the ages of the stars in the galaxy. However it determines a combination of the age of and the magnitude of a merger event, so there is a degeneracy here as well. Schweizer & Seitzer quote Fine Structure ages for three of our sample which are listed in column 3 of Table 5.6. A similar approach was carried out by Michard & Prugniel (2004). They divided their peculiar elliptical galaxy sample into a normal (NP), reddish, sample, with no signs of a young stellar population and a bluish sample (YP) with evidence for a younger stellar population mixed with an old one. They list five of our galaxies and their results are listed in the fourth column of Table 5.6. Dynamical ages for phase-wrapped, type I shell systems such as NGC 1344, NGC 3923 and NGC 5982 can be derived from shell radii and spacings of the outer shells, where dynamical friction and tidal effects are unimportant in determining the particle distribution (Nulsen (1989)). We used Nulsen’s equation 88 to derive the dynamical ages. Mass estimates were taken from the literature (NGC 1344: Teodorescu et al. (2005); NGC 3923: Fukazawa et al. (2006); fundamental plane masses: eq. 7 of van Dokkum & Stanford (2003)). The calculated ages are shown in Table 5.6 and are of the order of several 100 Myrs. It is clear that these young ages are not reflected in the GC populations for these galaxies, implying that the shell forming event did not in these cases give rise to a substantial GC population. Our sample consists of minor merger remnants, and only if it is possible to separate the old and merger related populations, as in the case of NGC 2865, can we see a correspondence between the different age estimators. More detailed studies of the stellar populations of all of the other galaxies would be valuable, as would better theoretical estimates of the dynamical ages of the more complicated Prieur (1990) Type II and III shell systems.

5.8 Conclusions

The properties of the globular cluster systems of six shell galaxies were analysed. For NGC 2865 and NGC 7626 we observe anomalous features in the (V-I) histograms, in addition to the bimodal structure which is normal for ellipticals. The features represent excesses at intermediate colour in NGC 7626, and at very blue colours in NGC 2865, consistent with a small population of GCs, of age 2-5 Gyr and 0.5-1 Gyr respectively, possibly formed in the merger event which created the shells. In each case the young population is dominated by the much larger, old, bimodal distribution. The data for two galaxies (NGC 1344 and NGC 3923) allow the determination of their globular cluster luminosity function distances. Fitting Gaussians to their GCLF give distances moduli of 31.72 and 31.50 for NGC 1344 and NGC 3923 respectively. The properties of NGC 3923 and NGC 5982 are very similar. Their bimodal V-I distributions and radial density profiles of blue and red clusters are typical for old GC systems in ellipticals. NGC 474 and NGC 1344 show one single blue peak and very shallow red peaks in their V-I histograms. These properties are unusual for bright 150 chapter 5: Globular Clusters of Shell Galaxies ellipticals (Kundu & Whitmore (2001), Peng et al. (2006)) and may indicate less early mergers in their formation history. NGC 3923 and NGC 7626 have higher specific frequencies (respectively 5.6 and 3.9 within 50kpc) than normal for galaxies located in a low density environment. The luminosity of the X-ray halo detected in the former galaxy is probably not sufficient to explain its high SN as proposed by McLaughlin (1999). The SN of the other galaxies have values of around 2 and are typical for galaxies located in a low density environment. Although for some of these six galaxies new GCs may have been formed recently, the general properties (like V-I distributions and flattening of GC density profile) of the globular cluster systems of these shell galaxies do not deviate systematically from ’normal’ elliptical galaxies. In a future paper (Sikkema et al., in preparation) we will investigate the morphology and stellar populations of the diffuse galaxy, in particular in the shells. Bibliography

Abraham, R. G., van den Bergh, S., & Nair, P. 2003, ApJ, 588, 218 Andersen, V. 1996, AJ, 111, 1805 Andredakis, Y. C., Peletier, R. F., & Balcells, M. 1995, MNRAS, 275, 874 Aragón-Salamanca, A., Bedregal, A. G., & Merrifield, M. R. 2006, A&A, 458, 101 Ashman, K. M., Bird, C. M., & Zepf, S. E. 1994, AJ, 108, 2348 Ashman, K. M. & Zepf, S. E. 1992, ApJ, 384, 50 Athanassoula, E., Lambert, J. C., & Dehnen, W. 2005, MNRAS, 363, 496 Athanassoula, E. & Misiriotis, A. 2002, MNRAS, 330, 35 Athey, A., Bregman, J., Bregman, J., Temi, P., & Sauvage, M. 2002, ApJ, 571, 272 Balcells, M. 1997, ApJ, 486, L87+ Balcells, M. & Carter, D. 1993, A&A, 279, 376 Balcells, M., van Gorkom, J. H., Sancisi, R., & del Burgo, C. 2001, AJ, 122, 1758 Bamford, S. P., Milvang-Jensen, B., & Aragón-Salamanca, A. 2007, MNRAS, 378, L6 Bamford, S. P., Nichol, R. C., Baldry, I. K., Land, K., Lintott, C. J., Schawinski, K., Slosar, A., Szalay, A. S., Thomas, D., Torki, M., Andreescu, D., Edmondson, E. M., Miller, C. J., Murray, P., Raddick, M. J., & Vandenberg, J. 2008, ArXiv e-prints Barazza, F. D., Jablonka, P., & the EDisCS collaboration. 2009, ArXiv e-prints Barazza, F. D., Jogee, S., & Marinova, I. 2008, ApJ, 675, 1194 Battinelli, P. & Capuzzo-Dolcetta, R. 1989, ApJ, 347, 794 Baum, W. A. 1959, PASP, 71, 106 Beasley, M. A., Baugh, C. M., Forbes, D. A., Sharples, R. M., & Frenk, C. S. 2002, MNRAS, 333, 383 Bekki, K., Couch, W. J., & Shioya, Y. 2002, ApJ, 577, 651 Bell, E. F. & de Jong, R. S. 2001, ApJ, 550, 212 Bender, R., Doebereiner, S., & Moellenhoff, C. 1988, A&AS, 74, 385 Berentzen, I., Athanassoula, E., Heller, C. H., & Fricke, K. J. 2003, MNRAS, 341, 343 —. 2004, MNRAS, 347, 220 Berentzen, I., Shlosman, I., Martinez-Valpuesta, I., & Heller, C. H. 2007, ApJ, 666, 189 Bershady, M. A., Jangren, A., & Conselice, C. J. 2000, AJ, 119, 2645 Bertin, E. & Arnouts, S. 1996, A&AS, 117, 393 152 BIBLIOGRAPHY

Block, D. L., Bournaud, F., Combes, F., Puerari, I., & Buta, R. 2002, A&A, 394, L35 Bournaud, F. & Combes, F. 2002, A&A, 392, 83 Bower, R. G., Lucey, J. R., & Ellis, R. S. 1992, MNRAS, 254, 601 Bower, R. G., Terlevich, A., Kodama, T., & Caldwell, N. 1999, in Astronomical Society of the Pacific Conference Series, Vol. 163, Star Formation in Early Type Galaxies, ed. P. Carral & J. Cepa, 211–+ Braglia, F., Pierini, D., & Böhringer, H. 2007, A&A, 470, 425 Brown, R. J. N., Forbes, D. A., Kissler-Patig, M., & Brodie, J. P. 2000, MNRAS, 317, 406 Buote, D. A. & Canizares, C. R. 1998, MNRAS, 298, 811 Burstein, D., Ho, L. C., Huchra, J. P., & Macri, L. M. 2005, ApJ, 621, 246 Butcher, H. & Oemler, Jr., A. 1984, ApJ, 285, 426 Byrd, G. G., Valtonen, M. J., Valtaoja, L., & Sundelius, B. 1986, A&A, 166, 75 Caon, N., Capaccioli, M., & D’Onofrio, M. 1993, MNRAS, 265, 1013 Cappellari, M. & Copin, Y. 2003, MNRAS, 342, 345 Capuzzo-Dolcetta, R. & Donnarumma, I. 2001, MNRAS, 328, 645 Capuzzo-Dolcetta, R. & Tesseri, A. 1997, MNRAS, 292, 808 Carollo, C. M., Franx, M., Illingworth, G. D., & Forbes, D. A. 1997, ApJ, 481, 710 Carter, D. 1978, MNRAS, 182, 797 Carter, D., Allen, D. A., & Malin, D. F. 1982, Nature, 295, 126 Carter, D., Thomson, R. C., & Hau, G. K. T. 1998, MNRAS, 294, 182 Charmandaris, V. & Combes, F. 2000, in Astronomical Society of the Pacific Conference Series, Vol. 209, IAU Colloq. 174: Small Galaxy Groups, ed. M. J. Valtonen & C. Flynn, 273–+ Christlein, D. & Zabludoff, A. I. 2004, ApJ, 616, 192 Cohen, J. G., Blakeslee, J. P., & Côté, P. 2003, ApJ, 592, 866 Cohen, J. G., Blakeslee, J. P., & Ryzhov, A. 1998, ApJ, 496, 808 Colbert, J. W., Mulchaey, J. S., & Zabludoff, A. I. 2001, AJ, 121, 808 Colless, M., Dalton, G., Maddox, S., Sutherland, W., Norberg, P., Cole, S., Bland- Hawthorn, J., Bridges, T., Cannon, R., Collins, C., Couch, W., Cross, N., Deeley, K., De Propris, R., Driver, S. P., Efstathiou, G., Ellis, R. S., Frenk, C. S., Glazebrook, K., Jackson, C., Lahav, O., Lewis, I., Lumsden, S., Madgwick, D., Peacock, J. A., Peterson, B. A., Price, I., Seaborne, M., & Taylor, K. 2001, MNRAS, 328, 1039 Combes, F. 2007, in IAU Symposium, Vol. 235, IAU Symposium, ed. F. Combes & J. Palous, 19–23 Combes, F. & Sanders, R. H. 1981, A&A, 96, 164 Conselice, C. J., Bershady, M. A., & Jangren, A. 2000, ApJ, 529, 886 Conselice, C. J., Rajgor, S., & Myers, R. 2008, MNRAS, 386, 909 Cote, P., Marzke, R. O., & West, M. J. 1998, ApJ, 501, 554 Couch, W. J., Barger, A. J., Smail, I., Ellis, R. S., & Sharples, R. M. 1998, ApJ, 497, 188 Crowl, H. H. & Kenney, J. D. P. 2008, AJ, 136, 1623 Dale, D. A., Giovanelli, R., Haynes, M. P., Hardy, E., & Campusano, L. E. 1999, AJ, 118, 1468 BIBLIOGRAPHY 153 de Jong, R. S. 1996a, A&A, 313, 45 —. 1996b, A&A, 313, 377 de Jong, R. S., Seth, A. C., Bell, E. F., Brown, T. M., Bullock, J. S., Courteau, S., Dalcanton, J. J., Ferguson, H. C., Goudfrooij, P., Holfeltz, S., Purcell, C., Radburn- Smith, D., & Zucker, D. 2007, in IAU Symposium, Vol. 241, IAU Symposium, ed. A. Vazdekis & R. F. Peletier, 503–504 de Jong, R. S., Simard, L., Davies, R. L., Saglia, R. P., Burstein, D., Colless, M., McMahan, R., & Wegner, G. 2004, MNRAS, 355, 1155 de Vaucouleurs, G. 1948, Annales d’Astrophysique, 11, 247 de Vaucouleurs, G., de Vaucouleurs, A., Corwin, H. G., Buta, R. J., Paturel, G., & Fouque, P. 1991, Third Reference Catalogue of Bright Galaxies (Volume 1-3, XII, 2069 pp. 7 figs.. Springer-Verlag Berlin Heidelberg New York) Dirsch, B., Richtler, T., Geisler, D., Forte, J. C., Bassino, L. P., & Gieren, W. P. 2003, AJ, 125, 1908 Dirsch, B., Schuberth, Y., & Richtler, T. 2005, A&A, 433, 43 Draine, B. T. & Salpeter, E. E. 1979, ApJ, 231, 77 Dressler, A. 1980, ApJ, 236, 351 Dressler, A., Oemler, A. J., Couch, W. J., Smail, I., Ellis, R. S., Barger, A., Butcher, H., Poggianti, B. M., & Sharples, R. M. 1997, ApJ, 490, 577 Dupraz, C. & Combes, F. 1986, A&A, 166, 53 Ellis, R. S., Smail, I., Dressler, A., Couch, W. J., Oemler, A. J., Butcher, H., & Sharples, R. M. 1997, ApJ, 483, 582 Elmegreen, B. G. & Elmegreen, D. M. 1985, ApJ, 288, 438 Emsellem, E., Cappellari, M., Peletier, R. F., McDermid, R. M., Bacon, R., Bureau, M., Copin, Y., Davies, R. L., Krajnović, D., Kuntschner, H., Miller, B. W., & de Zeeuw, P. T. 2004, MNRAS, 352, 721 Erben, T., Schirmer, M., Dietrich, J. P., Cordes, O., Haberzettl, L., Hetterscheidt, M., Hildebrandt, H., Schmithuesen, O., Schneider, P., Simon, P., Deul, E., Hook, R. N., Kaiser, N., Radovich, M., Benoist, C., Nonino, M., Olsen, L. F., Prandoni, I., Wichmann, R., Zaggia, S., Bomans, D., Dettmar, R. J., & Miralles, J. M. 2005, Astronomische Nachrichten, 326, 432 Erwin, P. 2005, MNRAS, 364, 283 Eskridge, P. B., Frogel, J. A., Pogge, R. W., Quillen, A. C., Davies, R. L., DePoy, D. L., Houdashelt, M. L., Kuchinski, L. E., Ramírez, S. V., Sellgren, K., Terndrup, D. M., & Tiede, G. P. 2000, AJ, 119, 536 Faber, S. M., Wegner, G., Burstein, D., Davies, R. L., Dressler, A., Lynden-Bell, D., & Terlevich, R. J. 1989, ApJS, 69, 763 Fall, S. M. & Rees, M. J. 1977, MNRAS, 181, 37P Fall, S. M. & Zhang, Q. 2001, ApJ, 561, 751 Ferrari, F., Pastoriza, M. G., Macchetto, F., & Caon, N. 1999, A&AS, 136, 269 Ferreras, I., Lisker, T., Carollo, C. M., Lilly, S. J., & Mobasher, B. 2005, ApJ, 635, 243 Forbes, D. A., Beasley, M. A., Brodie, J. P., & Kissler-Patig, M. 2001a, ApJ, 563, L143 Forbes, D. A., Brodie, J. P., & Grillmair, C. J. 1997, AJ, 113, 1652 Forbes, D. A. & Forte, J. C. 2001, MNRAS, 322, 257 Forbes, D. A., Franx, M., & Illingworth, G. D. 1995a, AJ, 109, 1988 154 BIBLIOGRAPHY

Forbes, D. A., Franx, M., Illingworth, G. D., & Carollo, C. M. 1996, ApJ, 467, 126 Forbes, D. A., Georgakakis, A. E., & Brodie, J. P. 2001b, MNRAS, 325, 1431 Forbes, D. A., Reitzel, D. B., & Williger, G. M. 1995b, AJ, 109, 1576 Forbes, D. A., Sánchez-Blázquez, P., Phan, A. T. T., Brodie, J. P., Strader, J., & Spitler, L. 2006, MNRAS, 366, 1230 Forbes, D. A., Strader, J., & Brodie, J. P. 2004, AJ, 127, 3394 Forbes, D. A. & Thomson, R. C. 1992, MNRAS, 254, 723 Fort, B. P., Prieur, J.-L., Carter, D., Meatheringham, S. J., & Vigroux, L. 1986, ApJ, 306, 110 Frenk, C. S. & Fall, S. M. 1982, MNRAS, 199, 565 Friedli, D. & Benz, W. 1993, A&A, 268, 65 Fritze-v. Alvensleben, U. 2004, A&A, 414, 515 Fukazawa, Y., Botoya-Nonesa, J. G., Pu, J., Ohto, A., & Kawano, N. 2006, ApJ, 636, 698 Fukugita, M., Shimasaku, K., & Ichikawa, T. 1995, PASP, 107, 945 Gadotti, D. A. 2008, MNRAS, 384, 420 Gallazzi, A., Bell, E. F., Wolf, C., Gray, M. E., Papovich, C., Barden, M., Peng, C. Y., Meisenheimer, K., Heymans, C., van Kampen, E., Gilmour, R., Balogh, M., McIntosh, D. H., Bacon, D., Barazza, F. D., Böhm, A., Caldwell, J. A. R., Häußler, B., Jahnke, K., Jogee, S., Lane, K., Robaina, A. R., Sanchez, S. F., Taylor, A., Wisotzki, L., & Zheng, X. 2009, ApJ, 690, 1883 Garcia, A. M. 1993, A&AS, 100, 47 Gascoigne, S. C. B. & Kron, G. E. 1952, PASP, 64, 196 Gebhardt, K. & Kissler-Patig, M. 1999, AJ, 118, 1526 Genzel, R., Tacconi, L. J., Eisenhauer, F., Förster Schreiber, N. M., Cimatti, A., Daddi, E., Bouché, N., Davies, R., Lehnert, M. D., Lutz, D., Nesvadba, N., Verma, A., Abuter, R., Shapiro, K., Sternberg, A., Renzini, A., Kong, X., Arimoto, N., & Mignoli, M. 2006, Nature, 442, 786 Gerin, M., Combes, F., & Athanassoula, E. 1990, A&A, 230, 37 Ghigna, S., Moore, B., Governato, F., Lake, G., Quinn, T., & Stadel, J. 1998, MNRAS, 300, 146 Giovanelli, R., Haynes, M. P., Salzer, J. J., Wegner, G., da Costa, L. N., & Freudling, W. 1995, AJ, 110, 1059 Girardi, L., Chiosi, C., Bertelli, G., & Bressan, A. 1995, A&A, 298, 87 Gnedin, O. Y. & Ostriker, J. P. 1997, ApJ, 474, 223 Gómez, P. L., Nichol, R. C., Miller, C. J., Balogh, M. L., Goto, T., Zabludoff, A. I., Romer, A. K., Bernardi, M., Sheth, R., Hopkins, A. M., Castander, F. J., Connolly, A. J., Schneider, D. P., Brinkmann, J., Lamb, D. Q., SubbaRao, M., & York, D. G. 2003, ApJ, 584, 210 González-García, A. C. & Balcells, M. 2005, MNRAS, 357, 753 Goto, T., Yamauchi, C., Fujita, Y., Okamura, S., Sekiguchi, M., Smail, I., Bernardi, M., & Gomez, P. L. 2003, MNRAS, 346, 601 Goudfrooij, P., Alonso, M. V., Maraston, C., & Minniti, D. 2001a, MNRAS, 328, 237 Goudfrooij, P., Gilmore, D., Whitmore, B. C., & Schweizer, F. 2004, ApJ, 613, L121 Goudfrooij, P., Hansen, L., Jorgensen, H. E., & Norgaard-Nielsen, H. U. 1994, A&AS, BIBLIOGRAPHY 155

105, 341 Goudfrooij, P., Mack, J., Kissler-Patig, M., Meylan, G., & Minniti, D. 2001b, MNRAS, 322, 643 Graham, A. W. & Driver, S. P. 2005, Publications of the Astronomical Society of Australia, 22, 118 Grillmair, C., Pritchet, C., & van den Bergh, S. 1986, AJ, 91, 1328 Gunn, J. E. & Gott, J. R. I. 1972, ApJ, 176, 1 Harris, W. E. 1991, ARA&A, 29, 543 Harris, W. E., Harris, G. L. H., & McLaughlin, D. E. 1998, AJ, 115, 1801 Harris, W. E., Kavelaars, J. J., Hanes, D. A., Hesser, J. E., & Pritchet, C. J. 2000, ApJ, 533, 137 Harris, W. E. & van den Bergh, S. 1981, AJ, 86, 1627 Hau, G. K. T., Balcells, M., & Carter, D. 1996, in IAU Symposium, Vol. 171, New Light on Galaxy Evolution, ed. R. Bender & R. L. Davies, 388–+ Hau, G. K. T., Carter, D., & Balcells, M. 1999, MNRAS, 306, 437 Hernquist, L. & Quinn, P. J. 1987, ApJ, 312, 1 —. 1989, ApJ, 342, 1 Hernquist, L. & Spergel, D. N. 1992, ApJ, 399, L117 Hibbard, J. E. & Sansom, A. E. 2003, AJ, 125, 667 Hilton, M., Collins, C., De Propris, R., Baldry, I. K., Baugh, C. M., Bland-Hawthorn, J., Bridges, T., Cannon, R., Cole, S., Colless, M., Couch, W. J., Dalton, G. B., Driver, S. P., Efstathiou, G., Ellis, R. S., Frenk, C. S., Glazebrook, K., Jackson, C. A., Lahav, O., Lewis, I., Lumsden, S., Maddox, S. J., Madgwick, D., Norberg, P., Peacock, J. A., Peterson, B. A., Sutherland, W., & Taylor, K. 2005, MNRAS, 363, 661 Ho, L. C., Filippenko, A. V., & Sargent, W. L. W. 1997a, ApJS, 112, 315 —. 1997b, ApJ, 487, 568 Hohl, F. 1971, ApJ, 168, 343 Holden, B. P., Illingworth, G. D., Franx, M., Blakeslee, J. P., Postman, M., Kelson, D. D., van der Wel, A., Demarco, R., Magee, D. K., Tran, K.-V., Zirm, A., Ford, H., Rosati, P., & Homeier, N. 2007, ApJ, 670, 190 Holley-Bockelmann, K., Weinberg, M., & Katz, N. 2005, MNRAS, 363, 991 Holwerda, B. W., Gonzalez, R. A., Allen, R. J., & van der Kruit, P. C. 2005, AJ, 129, 1396 Hubble, E. 1929, Proceedings of the National Academy of Science, 15, 168 Hubble, E. P. 1926, ApJ, 64, 321 Ibata, R. A., Gilmore, G., & Irwin, M. J. 1995, MNRAS, 277, 781 Jedrzejewski, R. & Schechter, P. L. 1988, ApJ, 330, L87 Jedrzejewski, R. I. 1987, MNRAS, 226, 747 Jørgensen, I., Franx, M., & Kjaergaard, P. 1992, A&AS, 95, 489 Jogee, S., Barazza, F. D., Rix, H.-W., Shlosman, I., Barden, M., Wolf, C., Davies, J., Heyer, I., Beckwith, S. V. W., Bell, E. F., Borch, A., Caldwell, J. A. R., Conselice, C. J., Dahlen, T., Häussler, B., Heymans, C., Jahnke, K., Knapen, J. H., Laine, S., Lubell, G. M., Mobasher, B., McIntosh, D. H., Meisenheimer, K., Peng, C. Y., Ravindranath, S., Sanchez, S. F., Somerville, R. S., & Wisotzki, L. 2004, ApJ, 615, 156 BIBLIOGRAPHY

L105 Kalnajs, A. J. 1972, ApJ, 175, 63 Kambas, A., Davies, J. I., Smith, R. M., Bianchi, S., & Haynes, J. A. 2000, AJ, 120, 1316 Kauffmann, G., White, S. D. M., Heckman, T. M., Ménard, B., Brinchmann, J., Char- lot, S., Tremonti, C., & Brinkmann, J. 2004, MNRAS, 353, 713 Kennicutt, Jr., R. C. 1983, AJ, 88, 483 Kessler, M. F., Steinz, J. A., Anderegg, M. E., Clavel, J., Drechsel, G., Estaria, P., Faelker, J., Riedinger, J. R., Robson, A., Taylor, B. G., & Ximénez de Ferrán, S. 1996, A&A, 315, L27 Knapen, J. H., Shlosman, I., & Peletier, R. F. 2000, ApJ, 529, 93 Koch, A., Odenkirchen, M., Grebel, E. K., & Caldwell, J. A. R. 2004, Astronomische Nachrichten, 325, 299 Kojima, M. & Noguchi, M. 1997, ApJ, 481, 132 Koopmann, R. A. & Kenney, J. D. P. 2004, ApJ, 613, 851 Kormendy, J. & Kennicutt, Jr., R. C. 2004, ARA&A, 42, 603 Kundu, A. & Whitmore, B. C. 2001, AJ, 121, 2950 Kuntschner, H., Smith, R. J., Colless, M., Davies, R. L., Kaldare, R., & Vazdekis, A. 2002, MNRAS, 337, 172 Lambas, D. G., Maddox, S. J., & Loveday, J. 1992, MNRAS, 258, 404 Larson, R. B., Tinsley, B. M., & Caldwell, C. N. 1980, ApJ, 237, 692 Lauer, T. R., Ajhar, E. A., Byun, Y.-I., Dressler, A., Faber, S. M., Grillmair, C., Kormendy, J., Richstone, D., & Tremaine, S. 1995, AJ, 110, 2622 Lauer, T. R., Faber, S. M., Gebhardt, K., Richstone, D., Tremaine, S., Ajhar, E. A., Aller, M. C., Bender, R., Dressler, A., Filippenko, A. V., Green, R., Grillmair, C. J., Ho, L. C., Kormendy, J., Magorrian, J., Pinkney, J., & Siopis, C. 2005, AJ, 129, 2138 Lauer, T. R. & Kormendy, J. 1986, ApJ, 303, L1 Lee, J.-W. & Carney, B. W. 2002, AJ, 123, 3305 Loewenstein, M., Fabian, A. C., & Nulsen, P. E. J. 1987, MNRAS, 229, 129 Lotz, J. M., Primack, J., & Madau, P. 2004, AJ, 128, 163 Loveday, J. 2002, Contemporary Physics, 43, 437 Macchetto, F., Pastoriza, M., Caon, N., Sparks, W. B., Giavalisco, M., Bender, R., & Capaccioli, M. 1996, A&AS, 120, 463 Madau, P., Ferguson, H. C., Dickinson, M. E., Giavalisco, M., Steidel, C. C., & Fruchter, A. 1996, MNRAS, 283, 1388 Maddox, S. J., Efstathiou, G., & Sutherland, W. J. 1990, MNRAS, 246, 433 Madgwick, D. S., Lahav, O., Baldry, I. K., Baugh, C. M., Bland-Hawthorn, J., Bridges, T., Cannon, R., Cole, S., Colless, M., Collins, C., Couch, W., Dalton, G., De Propris, R., Driver, S. P., Efstathiou, G., Ellis, R. S., Frenk, C. S., Glazebrook, K., Jackson, C., Lewis, I., Lumsden, S., Maddox, S., Norberg, P., Peacock, J. A., Peterson, B. A., Sutherland, W., & Taylor, K. 2002, MNRAS, 333, 133 Malin, D. F. & Carter, D. 1980, Nature, 285, 643 —. 1983, ApJ, 274, 534 Manfroid, J. & Selman, F. 2001, The Messenger, 104, 16 BIBLIOGRAPHY 157

Maraston, C. 2005, MNRAS, 362, 799 Margoniner, V. E., de Carvalho, R. R., Gal, R. R., & Djorgovski, S. G. 2001, ApJ, 548, L143 Marinova, I. & Jogee, S. 2007, ApJ, 659, 1176 Mathews, W. G. & Brighenti, F. 2003, ApJ, 590, L5 McDermid, R. M., Emsellem, E., Shapiro, K. L., Bacon, R., Bureau, M., Cappellari, M., Davies, R. L., de Zeeuw, T., Falcón-Barroso, J., Krajnović, D., Kuntschner, H., Peletier, R. F., & Sarzi, M. 2006, MNRAS, 373, 906 McLaughlin, D. E. 1999, AJ, 117, 2398 Menéndez-Delmestre, K., Sheth, K., Schinnerer, E., Jarrett, T. H., & Scoville, N. Z. 2007, ApJ, 657, 790 Michard, R. & Prugniel, P. 2004, A&A, 423, 833 Miller, B. W., Whitmore, B. C., Schweizer, F., & Fall, S. M. 1997, AJ, 114, 2381 Monet, D. G., Levine, S. E., Canzian, B., Ables, H. D., Bird, A. R., Dahn, C. C., Guet- ter, H. H., Harris, H. C., Henden, A. A., Leggett, S. K., Levison, H. F., Luginbuhl, C. B., Martini, J., Monet, A. K. B., Munn, J. A., Pier, J. R., Rhodes, A. R., Riepe, B., Sell, S., Stone, R. C., Vrba, F. J., Walker, R. L., Westerhout, G., Brucato, R. J., Reid, I. N., Schoening, W., Hartley, M., Read, M. A., & Tritton, S. B. 2003, AJ, 125, 984 Moore, B., Katz, N., Lake, G., Dressler, A., & Oemler, A. 1996, Nature, 379, 613 Moore, B., Lake, G., Quinn, T., & Stadel, J. 1999, MNRAS, 304, 465 Moran, S. M., Loh, B. L., Ellis, R. S., Treu, T., Bundy, K., & MacArthur, L. A. 2007, ApJ, 665, 1067 Moss, C. 2006, MNRAS, 373, 167 Noguchi, M. 1987, MNRAS, 228, 635 Norman, C. A., Sellwood, J. A., & Hasan, H. 1996, ApJ, 462, 114 Nulsen, P. E. J. 1989, ApJ, 346, 690 Ostriker, J. P., Binney, J., & Saha, P. 1989, MNRAS, 241, 849 O’Sullivan, E., Forbes, D. A., & Ponman, T. J. 2001, MNRAS, 328, 461 Peebles, P. J. E. 2002, in Astronomical Society of the Pacific Conference Series, Vol. 283, A New Era in Cosmology, ed. N. Metcalfe & T. Shanks, 351–+ Pei, Y. C. & Fall, S. M. 1995, ApJ, 454, 69 Peletier, R. F., Balcells, M., Davies, R. L., Andredakis, Y., Vazdekis, A., Burkert, A., & Prada, F. 1999a, MNRAS, 310, 703 Peletier, R. F., Knapen, J. H., Shlosman, I., Pérez-Ramírez, D., Nadeau, D., Doyon, R., Rodriguez Espinosa, J. M., & Pérez García, A. M. 1999b, ApJS, 125, 363 Peletier, R. F., Valentijn, E. A., Moorwood, A. F. M., Freudling, W., Knapen, J. H., & Beckman, J. E. 1995, A&A, 300, L1+ Pellegrini, S. 1999, A&A, 343, 23 Pence, W. D. 1986, ApJ, 310, 597 Peng, E. W., Ford, H. C., & Freeman, K. C. 2004, ApJ, 602, 705 Peng, E. W., Jordán, A., Côté, P., Blakeslee, J. P., Ferrarese, L., Mei, S., West, M. J., Merritt, D., Milosavljević, M., & Tonry, J. L. 2006, ApJ, 639, 95 Pesce, E., Capuzzo-Dolcetta, R., & Vietri, M. 1992, MNRAS, 254, 466 158 BIBLIOGRAPHY

Phillips, A. C., Illingworth, G. D., MacKenty, J. W., & Franx, M. 1996, AJ, 111, 1566 Pierfederici, F. & Rampazzo, R. 2004, Astronomische Nachrichten, 325, 359 Poggianti, B. M., Bridges, T. J., Carter, D., Mobasher, B., Doi, M., Iye, M., Kashikawa, N., Komiyama, Y., Okamura, S., Sekiguchi, M., Shimasaku, K., Yagi, M., & Yasuda, N. 2001a, ApJ, 563, 118 Poggianti, B. M., Bridges, T. J., Komiyama, Y., Yagi, M., Carter, D., Mobasher, B., Okamura, S., & Kashikawa, N. 2004, ApJ, 601, 197 Poggianti, B. M., Bridges, T. J., Mobasher, B., Carter, D., Doi, M., Iye, M., Kashikawa, N., Komiyama, Y., Okamura, S., Sekiguchi, M., Shimasaku, K., Yagi, M., & Yasuda, N. 2001b, ApJ, 562, 689 Poggianti, B. M., Smail, I., Dressler, A., Couch, W. J., Barger, A. J., Butcher, H., Ellis, R. S., & Oemler, A. J. 1999, ApJ, 518, 576 Pohlen, M. & Trujillo, I. 2006, A&A, 454, 759 Porter, S. C., Raychaudhury, S., Pimbblet, K. A., & Drinkwater, M. J. 2008, MNRAS, 388, 1152 Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P. 1992, Numerical recipes in FORTRAN. The art of scientific computing (Cambridge: University Press, |c1992, 2nd ed.) Prieur, J.-L. 1988, ApJ, 326, 596 —. 1990, Status of shell galaxies. (Dynamics and Interactions of Galaxies), 72–83 Proctor, R. N., Forbes, D. A., Forestell, A., & Gebhardt, K. 2005, MNRAS, 362, 857 Puzia, T. H., Kissler-Patig, M., Brodie, J. P., & Huchra, J. P. 1999, AJ, 118, 2734 Puzia, T. H., Kissler-Patig, M., Thomas, D., Maraston, C., Saglia, R. P., Bender, R., Richtler, T., Goudfrooij, P., & Hempel, M. 2004, A&A, 415, 123 Quilis, V., Moore, B., & Bower, R. 2000, Science, 288, 1617 Quillen, A. C., Bower, G. A., & Stritzinger, M. 2000, ApJS, 128, 85 Quinn, P. J. 1984, ApJ, 279, 596 Rampazzo, R., Alexander, P., Carignan, C., Clemens, M. S., Cullen, H., Garrido, O., Marcelin, M., Sheth, K., & Trinchieri, G. 2006, MNRAS, 368, 851 Ravindranath, S., Ho, L. C., Peng, C. Y., Filippenko, A. V., & Sargent, W. L. W. 2001, AJ, 122, 653 Reda, F. M., Forbes, D. A., Beasley, M. A., O’Sullivan, E. J., & Goudfrooij, P. 2004, MNRAS, 354, 851 Reid, N., Boisson, C., & Sansom, E. A. 1994, MNRAS, 269, 713 Renzini, A. 2006, ARA&A, 44, 141 Rhode, K. L. & Zepf, S. E. 2001, AJ, 121, 210 —. 2004, AJ, 127, 302 Richter, O.-G. & Sancisi, R. 1994, A&A, 290, L9 Richtler, T. 2003, in Lecture Notes in Physics, Berlin Springer Verlag, Vol. 635, Stellar Candles for the Extragalactic Distance Scale, ed. D. Alloin & W. Gieren, 281–305 Roberts, M. S., Hogg, D. E., Bregman, J. N., Forman, W. R., & Jones, C. 1991, ApJS, 75, 751 Robin, A. C., Reylé, C., Derrière, S., & Picaud, S. 2003, A&A, 409, 523 Sadler, E. M. & Gerhard, O. E. 1985, MNRAS, 214, 177 BIBLIOGRAPHY 159

Sancisi, R., Fraternali, F., Oosterloo, T., & van der Hulst, T. 2008, A&A Rev., 15, 189 Sandage, A. & Visvanathan, N. 1978, ApJ, 223, 707 Sarzi, M., Falcón-Barroso, J., Davies, R. L., Bacon, R., Bureau, M., Cappellari, M., de Zeeuw, P. T., Emsellem, E., Fathi, K., Krajnović, D., Kuntschner, H., McDermid, R. M., & Peletier, R. F. 2006, MNRAS, 366, 1151 Schiminovich, D., van Gorkom, J. H., van der Hulst, J. M., & Malin, D. F. 1995, ApJ, 444, L77 Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500, 525 Schombert, J. M. & Wallin, J. F. 1987, AJ, 94, 300 Schombert, J. M., Wallin, J. F., & Struck-Marcell, C. 1990, AJ, 99, 497 Schweizer, F. 1980, ApJ, 237, 303 Schweizer, F. 1987, in Nearly Normal Galaxies. From the Planck Time to the Present, ed. S. M. Faber, 18–25 Schweizer, F., Miller, B. W., Whitmore, B. C., & Fall, S. M. 1996, AJ, 112, 1839 Schweizer, F. & Seitzer, P. 1992, AJ, 104, 1039 Sellwood, J. A. 1981, A&A, 99, 362 Sheth, K., Elmegreen, D. M., Elmegreen, B. G., Capak, P., Abraham, R. G., Athanas- soula, E., Ellis, R. S., Mobasher, B., Salvato, M., Schinnerer, E., Scoville, N. Z., Spalsbury, L., Strubbe, L., Carollo, M., Rich, M., & West, A. A. 2008, ApJ, 675, 1141 Sikkema, G., Peletier, R. F., Carter, D., Valentijn, E. A., & Balcells, M. 2006, A&A, 458, 53 Sirianni, M., Jee, M. J., Benítez, N., Blakeslee, J. P., Martel, A. R., Meurer, G., Clampin, M., De Marchi, G., Ford, H. C., Gilliland, R., Hartig, G. F., Illingworth, G. D., Mack, J., & McCann, W. J. 2005, PASP, 117, 1049 Smith, G. P., Treu, T., Ellis, R. S., Moran, S. M., & Dressler, A. 2005, ApJ, 620, 78 Springel, V., White, S. D. M., Jenkins, A., Frenk, C. S., Yoshida, N., Gao, L., Navarro, J., Thacker, R., Croton, D., Helly, J., Peacock, J. A., Cole, S., Thomas, P., Couch- man, H., Evrard, A., Colberg, J., & Pearce, F. 2005, Nature, 435, 629 Stetson, P. B., Vandenberg, D. A., & Bolte, M. 1996, PASP, 108, 560 Stickel, M., van der Hulst, J. M., van Gorkom, J. H., Schiminovich, D., & Carilli, C. L. 2004, A&A, 415, 95 Strader, J., Brodie, J. P., Cenarro, A. J., Beasley, M. A., & Forbes, D. A. 2005, AJ, 130, 1315 Strader, J., Brodie, J. P., & Forbes, D. A. 2004, AJ, 127, 3431 Teodorescu, A. M., Méndez, R. H., Saglia, R. P., Riffeser, A., Kudritzki, R.-P., Gerhard, O. E., & Kleyna, J. 2005, ApJ, 635, 290 Thompson, L. A. 1981, ApJ, 244, L43 Thomson, R. C. 1991, MNRAS, 253, 256 Thomson, R. C. & Wright, A. E. 1990, MNRAS, 247, 122 Tielens, A. G. G. M., McKee, C. F., Seab, C. G., & Hollenbach, D. J. 1994, ApJ, 431, 321 Tonry, J. L., Dressler, A., Blakeslee, J. P., Ajhar, E. A., Fletcher, A. B., Luppino, G. A., Metzger, M. R., & Moore, C. B. 2001, ApJ, 546, 681 Toomre, A. 1977, in Evolution of Galaxies and Stellar Populations, ed. B. M. Tinsley 160 BIBLIOGRAPHY

& R. B. Larson, 401–+ Tran, H. D., Tsvetanov, Z., Ford, H. C., Davies, J., Jaffe, W., van den Bosch, F. C., & Rest, A. 2001, AJ, 121, 2928 Tremaine, S. D., Ostriker, J. P., & Spitzer, Jr., L. 1975, ApJ, 196, 407 Treu, T., Ellis, R. S., Kneib, J.-P., Dressler, A., Smail, I., Czoske, O., Oemler, A., & Natarajan, P. 2003, ApJ, 591, 53 Turnbull, A. J., Bridges, T. J., & Carter, D. 1999, MNRAS, 307, 967 Turnbull, A. J. & Collett, J. L. 1999, in Astronomical Society of the Pacific Confer- ence Series, Vol. 182, Galaxy Dynamics - A Rutgers Symposium, ed. D. R. Merritt, M. Valluri, & J. A. Sellwood, 485–+ Valentijn, E. A. 1990, Nature, 346, 153 Valentijn, E. A. & Kuijken, K. 2004, in Toward an International Virtual Observatory, ed. P. J. Quinn & K. M. Górski, 19–+ Valentijn, E. A., McFarland, J. P., Snigula, J., Begeman, K. G., Boxhoorn, D. R., Rengelink, R., Helmich, E., Heraudeau, P., Kleijn, G. V., Vermeij, R., Vriend, W.-J., Tempelaar, M. J., Deul, E., Kuijken, K., Capaccioli, M., Silvotti, R., Bender, R., Neeser, M., Saglia, R., Bertin, E., & Mellier, Y. 2007, in Astronomical Society of the Pacific Conference Series, Vol. 376, Astronomical Data Analysis Software and Systems XVI, ed. R. A. Shaw, F. Hill, & D. J. Bell, 491–+ van den Bergh, S. 1976, ApJ, 206, 883 —. 1994, AJ, 108, 2145 —. 2002, AJ, 124, 782 van der Kruit, P. C. 1979, A&AS, 38, 15 van der Marel, R. P. & Franx, M. 1993, ApJ, 407, 525 van der Wel, A. 2008, ApJ, 675, L13 van der Wel, A., Holden, B. P., Franx, M., Illingworth, G. D., Postman, M. P., Kelson, D. D., Labbé, I., Wuyts, S., Blakeslee, J. P., & Ford, H. C. 2007, ApJ, 670, 206 van Dokkum, P. G. 2001, PASP, 113, 1420 van Dokkum, P. G. & Franx, M. 1995, AJ, 110, 2027 van Dokkum, P. G., Franx, M., Förster Schreiber, N. M., Illingworth, G. D., Daddi, E., Knudsen, K. K., Labbé, I., Moorwood, A., Rix, H.-W., Röttgering, H., Rudnick, G., Trujillo, I., van der Werf, P., van der Wel, A., van Starkenburg, L., & Wuyts, S. 2004, ApJ, 611, 703 van Dokkum, P. G. & Stanford, S. A. 2003, ApJ, 585, 78 van Gorkom, J. & Schiminovich, D. 1997, in Astronomical Society of the Pacific Con- ference Series, Vol. 116, The Nature of Elliptical Galaxies; 2nd Stromlo Symposium, ed. M. Arnaboldi, G. S. Da Costa, & P. Saha, 310–+ Varela, J., Moles, M., Márquez, I., Galletta, G., Masegosa, J., & Bettoni, D. 2004, A&A, 420, 873 Vega Beltrán, J. C., Pizzella, A., Corsini, E. M., Funes, J. G., Zeilinger, W. W., Beck- man, J. E., & Bertola, F. 2001, A&A, 374, 394 Verdoes Kleijn, G., Vermeij, R., Valentijn, E., & Kuijken, K. 2007, in Astronomical Society of the Pacific Conference Series, Vol. 364, The Future of Photometric, Spec- trophotometric and Polarimetric Standardization, ed. C. Sterken, 103–+ Verdoes Kleijn, G. A., Baum, S. A., de Zeeuw, P. T., & O’Dea, C. P. 1999, AJ, 118, BIBLIOGRAPHY 161

2592 —. 2002, AJ, 123, 1334 Verdoes Kleijn, G. A. & de Zeeuw, P. T. 2005, A&A, 435, 43 Verheijen, M. A. W. 2004, in IAU Colloq. 195: Outskirts of Galaxy Clusters: Intense Life in the Suburbs, ed. A. Diaferio, 394–400 Wagner, S. J., Bender, R., & Moellenhoff, C. 1988, A&A, 195, L5 Whitmore, B. C., Miller, B. W., Schweizer, F., & Fall, S. M. 1997, AJ, 114, 1797 Whitmore, B. C. & Schweizer, F. 1995, AJ, 109, 960 Whitmore, B. C., Schweizer, F., Kundu, A., & Miller, B. W. 2002, AJ, 124, 147 Whitmore, B. C., Sparks, W. B., Lucas, R. A., Macchetto, F. D., & Biretta, J. A. 1995, ApJ, 454, L73+ Whitmore, B. C., Zhang, Q., Leitherer, C., Fall, S. M., Schweizer, F., & Miller, B. W. 1999, AJ, 118, 1551 Wilkinson, A., Prieur, J.-L., Lemoine, R., Carter, D., Malin, D., & Sparks, W. B. 2000, MNRAS, 319, 977 Williams, R. E., Blacker, B., Dickinson, M., Dixon, W. V. D., Ferguson, H. C., Fruchter, A. S., Giavalisco, M., Gilliland, R. L., Heyer, I., Katsanis, R., Levay, Z., Lucas, R. A., McElroy, D. B., Petro, L., Postman, M., Adorf, H.-M., & Hook, R. 1996, AJ, 112, 1335 Wilman, D. J., Oemler, Jr, A., Mulchaey, J. S., McGee, S. L., Balogh, M. L., & Bower, R. G. 2008, ArXiv e-prints Wolf, C., Aragon-Salamanca, A., Balogh, M., Barden, M., Bell, E. F., Gray, M. E., Peng, C. Y., Bacon, D., Barazza, F. D., Boehm, A., Caldwell, J. A. R., Gallazzi, A., Haeussler, B., Heymans, C., Jahnke, K., Jogee, S., van Kampen, E., Lane, K., McIn- tosh, D. H., Meisenheimer, K., Papovich, C., Sanchez, S. F., Taylor, A., Wisotzki, L., & Zheng, X. 2008, ArXiv e-prints Wolf, C., Meisenheimer, K., Kleinheinrich, M., Borch, A., Dye, S., Gray, M., Wisotzki, L., Bell, E. F., Rix, H.-W., Cimatti, A., Hasinger, G., & Szokoly, G. 2004, A&A, 421, 913 Wozniak, H., Friedli, D., Martinet, L., Martin, P., & Bratschi, P. 1995, A&AS, 111, 115 Zabludoff, A. I. & Mulchaey, J. S. 1998, ApJ, 496, 39 —. 2000, ApJ, 539, 136 Zamojski, M. A., Schiminovich, D., Rich, R. M., Mobasher, B., Koekemoer, A. M., Capak, P., Taniguchi, Y., Sasaki, S. S., McCracken, H. J., Mellier, Y., Bertin, E., Aussel, H., Sanders, D. B., Le Fèvre, O., Ilbert, O., Salvato, M., Thompson, D. J., Kartaltepe, J. S., Scoville, N., Barlow, T. A., Forster, K., Friedman, P. G., Martin, D. C., Morrissey, P., Neff, S. G., Seibert, M., Small, T., Wyder, T. K., Bianchi, L., Donas, J., Heckman, T. M., Lee, Y.-W., Madore, B. F., Milliard, B., Szalay, A. S., Welsh, B. Y., & Yi, S. K. 2007, ApJS, 172, 468 Zaritsky, D. & Rix, H.-W. 1997, ApJ, 477, 118 Zepf, S. E. & Ashman, K. M. 1993, MNRAS, 264, 611 Zepf, S. E., Ashman, K. M., & Geisler, D. 1995, ApJ, 443, 570 Zepf, S. E., Geisler, D., & Ashman, K. M. 1994, ApJ, 435, L117

Nederlandse samenvatting

Context

n de wetenschap zijn de ontstaansvragen meestal de moeilijkste te beantwoorden vragen. I Denk maar aan de volgende nog openstaande vragen: ’hoe ontstaat zelfbewustzijn’, ’hoe is het leven ontstaan’ en ’hoe is het heelal ontstaan’. De titel en het thema van dit proef- schrift heeft ook te maken met een actuele ’hoe’ vraag in de sterrenkunde namelijk: ’hoe zijn de verschillende typen sterrenstelsels ontstaan die we zoal kunnen observeren’. Een be- langrijke stap in de classificatie van sterrenstelsels werd gemaakt door Edwin Hubble (1889 - 1953). Hij stelde een schema op van de meest voorkomende typen sterrenstelsels die hij observeerde. Figuur 1 toont zijn schema. Links staan de zogeheten elliptische stelsels, rechts staan spiraalstelsels die zelf weer opgedeeld zijn in twee takken: spiralen met centrale ’balken’ (onderste tak) en zonder balken (bovenste tak). Precies in het middelpunt van de drie takken ligt een soort tussenvorm van spiraalstelsels en elliptische stelsels: de zogeheten S0 stelsels. De elliptische stelsels, links, zijn gesorteerd naar ’ellipticiteit’, wat afplatting betekent. Van links naar rechts neemt de afplatting toe voor de elliptische stelsels. Deze stelsels vertonen vrijwel geen structuur en hebben in het echt een bolvormige of rugbybal- achtige driedimensionale vorm. Ze bevatten heel weinig tot geen gas en er vindt daardoor (vrijwel) geen stervorming plaats. De bewegingen van de sterren zijn niet geordend: alle sterren bewegen kris kras door het hele stelsel: daardoor heeft het dan ook de drie dimen- sionale vorm. De spiraalstelsels, rechts, zijn daarentegen platte schijven met een centrale verdikking. De spiraalstelsels zijn in het schema gesorteerd op grootte van deze centrale verdikking: van links naar rechts gezien wordt deze steeds kleiner. Spiraalstelsels bevatten veel gas en stervorming is daardoor een normaal verschijnsel in deze stelsels. De eerder genoemde sortering blijkt samen te vallen met de mate van stervorming: de stelsels met de kleinste centrale verdikking (rechts dus) hebben de meeste stervorming. De banen van sterren in spiraalstelsels zijn geordend: ze bewegen allemaal in dezelfde richting in de schijf. Alle sterren roteren dus rondom het middelpunt van het spiraalstelsel. Onze zon bevindt zich ook in zo’n spiraalstelsel, een balkspiraalstelsel (onderste tak), en hij doet ongeveer 200 miljoen jaar over één rondje rondom het middelpunt, samen met 100 miljard andere sterren. De tussenvorm: S0, vertoont wel een roterende schijf en een centrale verdikking, echter er vindt hier weinig tot geen stervorming plaats door gebrek aan gas. Wat niet te zien is in Figuur 1 is dat S0 stelsels ook balken kunnen bevatten. Ieder sterrenstelsel bevat ook bolho- pen. Bolhopen zijn opeenhopingen van soms wel een paar honderdduizend sterren. Figuur 164 Nederlandse samenvatting

2 toont zo’n bolhoop. Ons eigen melkwegstelsel bevat zo’n 150 bolhopen. Deze bolhopen roteren mee met de sterren in de schijf, maar hebben zelf hun eigen baanbeweging, die vaak loodrecht op de schijf staat. Elliptische stelsels bevatten soms wel duizenden bolhopen. Sterrenstelsels zelf zijn ook weer gegroepeerd in kleine groepjes of hele grote groepen: de clusters. In de clusters bevinden zich veel meer elliptische stelsels dan spiraalstelsels. Buiten de clusters, in de ’lege’ gebieden, is het precies andersom. Kennis over de geschiedenis van het heelal is noodzakelijk om deze discrepantie te begrijpen. Daarover nu meer. Het was diezelfde Edwin Hubble die de uitdijing van het heelal ontdekte. Een ontdekking die tot de belangrijkste in de sterrenkunde behoort en die grote implicaties heeft voor ons beeld van het ontstaan van het heelal. Hubble ontdekte dat alle sterrenstelsels zich van ons af bewegen. Hij was zo slim om hieruit niet te concluderen dat de aarde daarom een zeer bevoorrechte positie in het heelal inneemt (alle stelsels bewegen zich immers van ons af). Doordat hij van meerdere stelsels de snelheid kon meten en tevens de afstand, zag hij dat, als bijvoorbeeld een stelsel zich twee keer zo snel van ons af beweegt, ook de afstand twee keer zo groot was: hij concludeerde hieruit dat alle sterrenstelsels zich van elkaar af bewegen. Dit is te vergelijken met een rijzende rozijnencake in de oven. Als iedere rozijn een sterrenstelsel voorstelt , gebeurt in de cake precies hetzelfde: rozijnen op twee keer zo grote afstand van elkaar bewegen ook precies twee keer zo snel van elkaar vandaan.

Goed, we leven dus kennelijk in een uitdijend heelal. Maar dat betekent dus, dat vroeger alle sterrenstelsels zich veel dichter bij elkaar bevonden. Door terug te rekenen en rekening te houden met de remmende werking van gravitatie, kunnen we zelfs berekenen hoe oud het heelal is. Tegenwoordig denkt men dat het heelal zo’n 14 miljard jaar oud is. Op de helft van de leeftijd van het heelal, 7 miljard jaar geleden dus, stonden alle stelsels twee keer zo dicht bij elkaar. Nog veel eerder stonden ze nog veel dichter bij elkaar, zodat de wederzijdse effecten van gravitatie veel groter waren. Door de omstandigheden tijdens de begintijd van het heelal te simuleren, en door met de grootste en beste telescopen heel ver en dus heel ver terug te kijken in de tijd, is onder astronomen op dit moment een ontstaansmodel voor de verschillende typen sterrenstelsels populair dat als volgt werkt: de eerste stelsels ontstaan uit botsingen van kleinere klompjes ’bouwstenen’, die heel veel gas bevatten. Hierbij zijn waarschijnlijk ook grote aantallen bolhopen gevormd, die bij ieder sterrenstelsel zijn terug te vinden. De meeste hedendaagse elliptische stelsels zijn ontstaan in de hoogste dichtheidsgebieden, wat nu de clusters zijn. De spiralen ontstaan gemiddeld wat later en meest in de meer rustige gebieden. Doordat hier weinig grote botsingen plaatsvinden konden zo de geordende schijven ontstaan. Soms botsten deze schijven en ontstond daaruit een nieuwe elliptisch stelsel, alsmede nieuwe bolhopen. Dit soort botsingen komen nu, door het sterk uitgedijde heelal, veel minder voor. In dit model zijn S0 stelsels geëvolueerde spiraalstelsels die hun gas verloren hebben om diverse redenen. De grote meerderheid van de elliptische stelsels is dus al op zeer vroege leeftijd gevormd (10 miljard jaar geleden en eerder), en bestaan voornamelijk uit oude sterren. Spiralen zijn meer te vinden in de legere gebieden en bevatten, door hun voortgaande stervorming nog veel jonge sterren. Deze verschillende eigenschappen vertaalt zich in een belangrijke waarneming: kleurverschillen. Oude sterren zien er rood uit, terwijl jonge sterren juist blauw zijn. Een elliptisch stelsel, dat bijna geheel uit oude sterren bestaat, ziet er dan ook veel roder uit dan een spiraalstelsel. Het is belangrijk dit kleuronderscheid te onthouden voor de rest van onderstaand verhaal. De tussenvormen, de S0s, zien er door hun gebrek aan stervorming ook roder uit dan Nederlandse samenvatting 165

Figuur 1: Hubble Classificatie Schema met links elliptische stelsels en recht spiraalstel- sels die opgedeeld zijn spiralen met balken (onderste tak) en zonder balken (bovenste tak). Op het snijpunt ligt een S0 stelsel, een tussenvorm tussen elliptische stelsels en spiraalstelsels. spiralen. Toch tonen de S0s, in tegenstelling tot de elliptische stelsels, een sterke evolutie in recente tijden (laatste paar miljard jaar). Dit werd aangetoond door Dressler in 1980. Hij vergeleek de aantallen S0s in verschillende clusters tijdens de laatste paar miljard jaar. Hij vond dat er maar weinig S0s waren in clusters op 4 miljard lichtjaar, terwijl in meer nabije clusters de aantallen S0s veel groter zijn. Deze evolutie heeft dan waarschijnlijk ook met de dichtheid (omgeving) van sterrenstelsels te maken, die veel hoger is in clusters. Na deze introductie kunnen we nu het thema van dit proefschrift behandelen. De ti- tel (en thema) van dit proefschrift luidt: “De invloed van de omgeving op de Evolutie van Sterrenstelsels”. Als we over de omgeving van sterrenstelsels praten, moeten eerst omgeving definiëren: hier karakteriseren we de omgeving met behulp van de sterrenstelseldichtheid. Ook is het van belang te weten wanneer we naar die omgeving kijken: vroeger lagen stelsels immers veel dichter bij elkaar, en was de dichtheid veel hoger dan heden ten dage. In dit proefschrift bestuderen we alleen relatief nabije sterrenstelsels en clusters: we kijke ten hoogste 1 miljard jaar terug in de tijd. Dit proefschrift is ingedeeld in twee gedeelten. Het eerste gedeelte bestaat uit de hoofdstuk- ken 2 en 3. Hiervoor hebben we een aaneengesloten gebiedje aan de hemel ter grootte van 166 Nederlandse samenvatting

Figuur 2: De bolhoop M92

16 volle manen waargenomen. Het gebiedje bevat enkele clusters die zich alle op een afstand van ongeveer 1 miljard lichtjaar bevinden, dus 1 miljard jaar terug in de tijd. De tijd dus waarvan we verwachten dat er zich veel S0s vormen, de grote meerderheid van de elliptische sterrenstelsels hebbben zich hier al lang gevormd. We analyseren een groot aantal eigen- schappen van alle heldere sterrenstelsels in het gebiedje en kijken hoe de eigenschappen van sterrenstelsels veranderen als functie van dichtheid. We vinden dat spiraalstelsels de sterkste variaties vertonen en vinden mogelijk de locatie waar spiraalstelsels in S0s veranderen. Hoofdstukken 4 en 5 vormen het tweede gedeelte. Hierbij gaat het om onderzoek naar zes afzonderlijke elliptische stelsels in ’lege’ gebieden, waarvan al bekend was dat ze bepaalde afwijkingen vertonen in hun lichtverdeling t.o.v. normale elliptische stelsels. De afwijkingen zijn waarschijnlijk restproducten van veel kleinere sterrenstelseltjes die gebotst hebben met de heel veel grotere elliptische stelsels. Door die afwijkingen nauwkeurig te onderzoeken kan informatie worden verkregen over of dit bots-scenario inderdaad klopt en zo ja, over hoe die kleinere stelseltjes er vroeger dan wel uitzagen. De zes stelsels staan op slechts enkele tientallen miljoenen lichtjaren (zeer nabij dus). Verder hebben we ook onderzoek gedaan naar de bolhopen in deze stelsels.

Eerste Gedeelte: cluster onderzoek

Het aaneengesloten gebiedje ter grootte van 16 volle manen was al eerder waargenomen in een ander wetenschappelijk project: de 2dF survey. De 2dF survey had als doel om van vele heldere sterrenstelsels en clusters de afstand (roodverschuiving) te bepalen door middel van multi-fiber waarnemingen. Bij dit soort waarnemingen bepaalt men voor heel veel sterrenstelsels in hetzelfde beeldveld in één keer de roodverschuiving. Een nadeel met deze snelle methode is dat dit alleen goed werkt voor heldere stelsels. In ieder geval weten we zo dus wel dat de clusters en de sterrenstelsels in ons gebiedje Nederlandse samenvatting 167 alle op ongeveer 1 miljard lichtjaar liggen. Handig is ook dat we sterrenstelsels die veel dichterbij of verder weg liggen, weg kunnen laten, zodat onze resultaten niet beïnvloedt worden door deze ’verstorende’ stelsels. De waarnemingen voor dit gedeelte werden gedaan met een groothoek camera die op een relatief kleine (2.2 meter) telescoop is gemonteerd. Eind 2009 zal een veel grotere groothoekcamera in bedrijf komen: de OmegaCAM. Onze waarnemingen werden gebruikt als testdata, met als doel hoe we straks OmegaCAM optimaal kunnen gebruiken voor nog veel grootschaliger clusteronderzoek. In Groningen is de afgelopen jaren gewerkt aan computer programma’s, die straks OmegaCAM data gaan verwerken. Onze gegevens werden gebruikt om deze computer-programma’s te testen, zodat OmegaCAM gebruikers straks sneller aan de slag kunnen. De waarnemingen werden gedaan in twee golflengte-gebieden, het optische en rode ge- deelte van het spectrum. Dit kan bereikt worden door filters in het lichtpad van de telescoop te plaatsen. Door in beide golflengte-gebieden de lichtkracht te bepalen van sterrenstelsels kunnen we de kleur meten. Zoals in de introductie vermeld is, hebben elliptische en S0 stel- sels meestal rode kleuren, terwijl spiraalstelsels meest blauw zijn. Dit is te zien in Figuur 3: een kleur-helderheids diagram: vertikaal is de kleur uitgezet, d.w.z. van onder naar boven worden de objecten steeds roder. Horizontaal is de helderheid uitgezet (van links naar rechts neemt de helderheid af). Het diagram bevat alle stelsels in onze survey, de stelsels op 1 miljard lichtjaar (blauw), maar ook de stelsels die dichterbij (groen) en verder weg (rood) liggen, alsmede de vele stelsels waarvan we de afstand niet weten (zwart) . De elliptische stelsels bevinden zich in een vrij nauwe band, tussen de aflopende horizontale lijnen, terwijl de spiraalstelsels daar onder liggen en dus veel blauwer van kleur zijn. Kleur is dus een goede maat om sterrenstelsels te classificeren. Echter er bestaan ook spiraalstelsels met een rode kleur of elliptische stelsels met een blauwe kleur. Daarom moeten we betere methoden zien te vinden om stelsels te classificeren. Het blote oog is nog altijd de beste manier om sterrenstelsels te classificeren volgens Figuur 1. Echter, dit wordt lastig als er duizenden of honderdduizenden sterrenstelsels te classificeren zijn, wat straks met OmegaCAM zeker gaat gebeuren. Daarom hebben we eigenlijk een automatische classificatie manier nodig. We hebben heel veel automatische manieren getest en het blijkt dat de beste manier is om gebruik te maken van een parameter genaamd: Sérsic index. Deze index is eigenlijk niets anders dan een getalletje die de vorm van de lichtverdeling (profiel) van een sterrenstelsel beschrijft. Figuur 4 toont twee van zulke profielen. Elliptische stelsels hebben stijle, gekromde profielen (rechts in Fig. 4) , terwijl vlakkere rechtere profielen de schijfstelsels (spiralen en S0s) goed beschrijven (links in Fig. 4). Als we deze manier gebruiken en vergelijken met classificaties met het blote oog, blijkt dat we ongeveer 90% van de spiralen goed te classificeren. De S0s moeten natuurlijk nog wel steeds gescheiden worden. Om S0s van spiralen te onderscheiden blijken asymmetrie, kleur of stervorming de beste manier voor automatische classificatie. Deze methoden hebben echter nog steeds grote foutenmarges. Een betere methode is dus gewenst. Aangezien wij hier niet zo heel veel stelsels hebben, maakten we toch gebruik van het oog om spiraalstelsels van de andere klassen te onderscheiden. Om S0s van elliptische stelsels te onderscheiden gebruikten we de Sérsic index. Na deze classificatie kunnen we kijken of we een bijdrage kunnen leveren aan de ’hoe’ vraag ten aanzien van het ontstaan van S0 stelsels. Daartoe kijken we naar de eigenschap- pen van de verschillende klassen sterrenstelsels als functie van dichtheid. Deze dichtheid moeten we wel eerst berekenen. Hiertoe maakten we gebruik van de 2dF stelsels, waarvan we de afstand (roodverschuiving) kennen en berekenden voor ieder punt in ons gebied de 168 Nederlandse samenvatting

3

2

1

14 16 18 20

Figuur 3: Kleur-Helderheid diagram van alle sterrenstelsels in het geobserveerde ge- bied. Blauwe vierkantjes zijn sterrenstelsels op 1 miljard lichtjaar. Groene sterrretjes en rode driehoeken zijn respectievelijk voor en achtergrond sterrenstelsels. Zwarte kruisjes corresponderen met sterrenstelsels waarvan de afstand niet bekend is (dit zijn er vele!)

(drie-dimensionale) dichtheid van sterrenstelsels. Vervolgens definieerden we vier dichtheid- regimes: het hoogste dichtheids-regime zijn de binnengebieden van clusters, waar de dicht- heid het grootst is, het laagste regime ligt in vrij lege gebieden. De andere gebieden liggen er tussen in. De regimes werden zo gekozen dat ieder regime ongeveer evenveel stelsels bevat. Vervolgens berekenden we voor ieder dichtheids-regime de gemiddelde waarden van de parameters voor de drie klassen sterrenstelsels elliptisch, spiraal en S0. Drie parame- ters blijken sterk te variëren met dichtheid: asymmetrie, kleur en stervorming. Asymmetrie blijkt alleen voor spiraalstelsels te variëren. De asymmetrie wordt steeds lager naarmate we naar dichtere gebieden toegaan. Stervorming blijkt voor alle klassen op dezelfde manier te variëren, met een geleidelijke afname van stervorming van laag naar hoge dichtheden. De variatie in kleur tenslotte, vertoont voor spiraalstelsels een vrij opvallend effect: deze blijft in lege gebieden vrij blauw en constant, om opeens in de buitendelen van clusters sterk rood te worden. Deze resultaten brengen twee vragen naar voren: definieert het gebied waar de plotselinge verroding van spiraalstelsels optreedt misschien de regio waar spiraalstelsels in S0s transformeren? Zijn deze rode spiralen misschien een tussenvorm van S0s? Met betrekking tot de eerste vraag: de verroding van de spiraalstelsels kan verklaard worden doordat in deze gebieden een groot deel van het gas in spiraalstelsels verloren gaat door drie cluster specifieke processen: de eerste heet ’voorverwerking’. In de buitendelen van de cluster ondergaat het spiraalstelsel vaak nabije interacties: hierbij stroomt waarschijnlijk gas naar de centrale verdikking en veroorzaakt daar stervorming. Dit kan ook een extra populatie van balk sterrenstelsels kunnen veroorzaken (zie volgende Sectie). Het tweede Nederlandse samenvatting 169

Figuur 4: Twee panelen met morfologische informatie. Horizontaal is steeds de straal van het stelsel gemeten vanaf het centrum. Bovenste raam: lichtprofiel, midden: posi- tiehoek, onder: ellipticiteit. Links een balkspiraal: spiraal heeft een bijna recht profiel, de balk verraadt zich door gelijktijdige grote variaties in positiehoek en ellipticiteit op straal = 5 boogseconden. Dit is de buitengrens van de balk. Rechts een elliptisch stelsel, het lichtprofiel is hier duidelijk gekromd. 170 Nederlandse samenvatting proces heeft te maken met de uiterst heet gas aanwezig in de het centrum van de meeste clusters. Dit hete gas veroorzaakt dat het gas van de snel bewegende spiraalstelsels in de clusters gestript wordt. Het derde process gaat vrij langzaam en wordt ’uitsterving’ genoemd. In lege gebieden worden sterrenstelsels constant gevoed met extern, invallend, koel gas. Zo blijft de stervorming steeds op peil. In clustergebieden is dit koele gas niet aanwezig, waardoor de voorraad koel gas in het spiraalstelsel langzaam uitput. Voor een antwoord op de tweede vraag hebben we de gemiddelde morfologische eigenschappen ver- geleken van gewone, blauwe, spiralen, rode spiralen en normale, rode, S0s. Het blijkt dat voor alle eigenschappen, de rode spiralen gemiddeld tussen de normale spiralen en S0s in liggen. We concluderen dan ook dat rode spiralen een tussenvorm tussen spiralen en S0s zouden kunnen zijn.

Bars In de introductie merkten we op dat er ook spiraalstelsels en S0s zijn, die balken bevatten. Er zijn verschillende modellen die verklaren hoe balken ontstaan en verdwijnen. Waarschijn- lijk zijn kleine verstoringen of externe invloeden verantwoordelijk voor het ontstaan van balken. Simulaties laten zien dat koel gas, aanwezig in het spiraalstelsel, de balk kan laten groeien. Echter na verloop van tijd wordt de balk instabiel, valt uiteen en zien we weer een ’normaal’ spiraalstelsel. Daarna zou er weer een nieuwe balk kunnen ontstaan, zodat er in de loop der tijd, verschillende balk episodes bestaan in het leven van een stelsel, waar- bij de gemiddelde balkgrootte steeds toeneemt. Omdat onze waarnemingen verschillende dichtheidsgebieden bevatten, waar de externe invloeden verschilld van aard zijn, kunnen we misschien aanwijzingen vinden over balkvorming. Daartoe moeten we natuurlijk eerst de balken kunnen detecteren. De balken hebben we ontdekt met een beproefde methode. De structuur van een balk kan omschreven worden als één met een hoge mate van ellipticiteit, zie Figuur 1. Om de balken te detecteren fitten we eerst ellipsen als functie van straal aan de lichtverdeling van het 2 dimensionale plaatje van een stelsel. De balk zal zich verraden door grote variaties in de ellipticiteiten. Figuur 4 (links), laat het resultaat zien van het ellipsen fitten aan een balkspiraalstelsel. We zien dat de ellipticiteit op een zekere straal een maximum bereikt. Op het uiteinde van balk neemt de ellipticiteit plots weer af, wat gepaard gaat met gelijktijdige verandering in positiehoek. Deze detecties van balken werd steeds m.b.v. het blote oog geverifieerd. Detectie van balken op afstanden van 1 miljard lichtjaar, zoals het geval is voor onze dataset, ligt op de grens wat mogelijk is met een telescoop op aarde vanwege de verstorende werking van de atmosfeer. Daardoor konden we dan ook alleen de grootste balken detecteren. Door dit nu voor alle spiraalstelsels en S0s te doen, kunnen we het percentage balken onder spiraalstelsels en S0s in de lage en hoge dichtheids gebieden bepalen. Het blijkt dat het percentage balken significant hoger ligt in hogere dichtheidsgebieden. De balken zijn gemiddeld ook groter in deze gebieden. Dit kan verklaard worden doordat er bij clusters meer interacties optreden, welke zullen balkvorming stimuleren in gasrijke spiraalstelsels. Een ander resultaat dat we vonden is dat er bijna geen balken zijn te vinden in S0s. Als S0s inderdaad voortkomen uit spiraalstelsels (het feit dat we überhaupt balken zien in S0s is daar al een aanwijzing voor), betekent dat, dat deze op een of andere manier moeten uitdoven of verdwijnen. We vermoeden dat de vele snelle interacties in de clusters voor deze uitdoving zorgen in de loop der tijd, nadat al het gas in het oorspronkelijke spiraalstelsel is verdwenen (zie vorige Sectie). Nederlandse samenvatting 171

Figuur 5: De schillen alsmede een stofband en wolk zijn hier duidelijk zichtbaar

Tweede Gedeelte: zes elliptische stelsels Waarnemingen Voor dit gedeelte werden waarnemingen gebruikt, die gemaakt werden door de Hubble Space Telescope (HST). Deze telescoop is de ruimte ingebracht om zo de verstorende invloed van de atmosfeer te omzeilen. Daardoor kunnen we tot wel twintig keer nauwkeuriger kijken, erg belangrijk bij dit soort onderzoek. De waarnemingen werden gedaan in dezelfde golflengte- gebieden als in het vorige hoofdstuk, het optische en rode gedeelte van het spectrum, zodat we weer de kleur kunnen meten. De zes afzonderlijke elliptische stelsels met de bepaalde afwijkingen in hun lichtverdeling, worden in het engels ook wel “shell galaxies”, schilstelsel, genoemd : de afwijkingen lijken op schillen (zie het voorbeeld in Figuur 5). Er zijn twee theorieën die de schillen kunnen verklaren. Beide theorieën doen verschillende voorspellingen over de kleuren en vormen van de schillen. De eerste theorie (botsingstheorie) beweert dat de schillen het resultaat zijn van een botsing van het al bestaande elliptische stelsel en een klein stelseltje (5% van de massa van het grotere elliptische stelsel). De schillen zijn hier restproducten van het kleine stelseltje en kunnen allerlei vormen hebben afhankelijk van wat voor type het kleine stelseltje (veel gas of weinig gas) is en hoe het precies wordt ingevangen (onder welke hoek, met welke snelheid). Computersimulaties laten een heel scala van zulke vormen zien die we kunnen vergelijken met de waarnemingen. De schillen kunnen allerlei kleuren hebben. Zoals we eerder opmerkten zijn elliptische stelsels rood van kleur. Als het kleine (spiraal-)stelseltje veel jonge sterren bevat (blauwe kleur) zullen de schillen dan blauwer zijn dan het hoofdstelsel. De tweede theorie (interactie theorie) zegt dat de schillen opgewekt worden in het ellip- 172 Nederlandse samenvatting tische stelsel, doordat een ander stelsel er vlak bij langs vloog. Hier vindt dus geen botsing plaats. De tweede theorie heeft als vooraanname dat er altijd al een bepaalde rotatie aanwe- zig moet zijn in het elliptische hoofdstelsel. Omdat de schillen deel zijn van het hoofdstelsel zullen de kleuren van de schillen gelijk zijn die van het hoofdstelsel. De vorm van de schillen is afhankelijk van de hoek af waaronder we het hoofdstelsel zien. Daarbij zijn twee extreme mogelijkheden: of zien steeds korte boogjes of we zien hele grote bogen die het hele stelsel omvatten. Weer geldt hier dat de mogelijke vormen volgen uit computersimulaties die we weer kunnen vergelijken met de waarnemingen. Ook de aanwezigheid van stof kan enige aanwijzingen geven of een bepaalde theorie juist is. Bij de botsingtheorie zullen er af en toe stelseltjes worden ingevangen die veel gas bevatten. Het is dan zeer waarschijnlijk dat dit gas naar de kern van het hoofdstelsel stroomt en aldaar stervorming veroorzaakt. Stervorming gaat altijd gepaard met stofvorming. Als we dus veel stof waarnemen in het centrum is dit een extra steun voor de botsingstheorie.

Analyse en Resultaten Om de positie van de schillen nauwkeurig te bepalen, hebben we eerst een model gemaakt van de gelijkmatige lichtverdeling van het stelsel en dat afgetrokken van het originele Hubble plaatje. Onregelmatigheden worden zo goed zichtbaar in het plaatje. Figuur 5 vertoont één zo’n stelsel na aftrek van het model. De locatie en de vormen van de schillen en stof zijn nu duidelijk zichtbaar. Door de vormen van de schillen van de zes stelsels te vergelijken met de simulaties voor de beide modellen, blijkt dat in de meeste stelsels de vormen het best overeen komen met het botsingsmodel. In de andere gevallen kunnen beide modellen de vormen goed beschrijven. Vervolgens hebben we voor iedere voldoende heldere schil nauwkeurig de kleur bepaald op een nieuwe, veel nauwkeuriger, manier. Eerst maakten we op de schillocaties een locaal model voor het stelsel te maken om dat vervolgens weer van het plaatje af te trekken. Voor sommige stelsels vinden we duidelijke kleurverschillen met het hoofdstelsel. Dit duidt er dus op dat hier de botsingstheorie beter van toepassing is. De andere gevallen geven weer geen uitsluitsel: beide modellen kunnen het goed verklaren. Alle zes stelsels vertonen stofwolkjes in hun centrum. Bij normale elliptische stelsels zien we gemiddeld de helft van de gevallen zulk soort stofwolken. Als onze stelsels normaal zijn, zou daaruit volgen dat de kans om stof in alle zes stelsels te vinden, minder dan 5% is. Hieruit concluderen we dat onze stelsels meer stof bevatten dan dat je zou verwachten. Een extra aanwijzing voor de juistheid van de botsingstheorie. Alles samen nemend komen we dus tot de conclusie dat de botsingstheorie de schillen het beste beschrijft. Hoewel in een enkel geval de interactie theorie ook zou kunnen optreden. Verder blijkt dus dat elliptische stelsels in lege gebieden nog steeds onderhevig aan een langzame geleidelijke evolutie die veroorzaakt wordt door de omgeving.

Bolhopen in de zes schilstelsels. Bolhopen (zie Figuur 2) zijn belangrijk voor het onderzoek naar de vorming van sterren- stelsels. Ons eigen melkwegstesel bevat zo’n 150 bolhopen. Elliptische stelsels, zoals onze schilstelsels, bevatten soms wel duizenden bolhopen. Bolhopen zijn meestal heel oud, 11 tot 13 miljard jaar. Een decennium geleden werd ontdekt dat elliptische stelsels, in tegen- stelling tot spiraalstelsels, twee populaties van bolhopen bezitten. De populaties kenmerken zich door kleurverschillen: één populatie met blauwe kleuren en ’eén populatie met rode Nederlandse samenvatting 173 kleuren. De blauwe populatie kleuren wordt ook in spiraalstelsels gevonden en is kennelijk een universele populatie die in alle sterrenstelsel wordt gevonden. De rode populatie is dus een extra populatie in elliptische stelsels. Onmiddellijk werden drie verschillende theorieën bedacht om de extra populatie te ver- klaren. Mogelijk zijn ze alle drie in meer of mindere mate van toepassing. De 1e theorie beweert dat de extra populaties ontstaan zijn tijdens een botsing van twee spiraalstelsels. Hierbij zouden dan veel nieuwe bolhopen ontstaan die naar verloop van tijd roder zijn dan de universele populatie. De rodere kleur komt omdat de bolhopen later in het heelal zijn ontstaan en daardoor veel meer zwaardere elementen (metalen in astronomisch jargon) be- vat dan oudere bolhopen (Intermezzo: metalen ontstaan bij het eind van het leven van een ster, als zo’n ster zwaar genoeg is, ontploft deze in een zogeheten explosie, waarbij de metalen ontstaan en in het heelal worden geblazen. In de loop der tijd komen er in het heelal dus steeds meer metalen bij.). Deze rode extra populatie, zal eerst nog extreem blauw zijn, in het geval dat de botsing zeer recentelijk heeft plaats gevonden. Zulk soort botsende stelsels worden inderdaad waargenomen en inderdaad zien we dan een blauwe populatie bolhopen. Het is echter lastig om te zien hoe dit 10 miljard jaar geleden ging, toen de meeste elliptische stelsels werden gevormd. Er bestonden toen waarschijnlijk nog maar weinig complete spiraalstelsels en de botsingen waren van een ander karakter hadden dan in het huidige heelal. De 2e theorie zegt dat de extra populatie ontstaat na het ontstaan van een elliptisch stelsel. De vorming van een elliptisch stelsel gaat gepaard gaat met een enorme golf van stervorming en vorming van bolhopen (1e populatie). De 2e theorie zegt dat de eerste golf van stervorming een tweede golf van stervorming veroorzaakt, waarbij dan de 2e populatie bolhopen ontstaat. De rode kleur van de bolhopen van de 2e populatie wordt weer door de extra metalen veroorzaakt, die tijdens de eerste golf van stervorming het heelal werden ingeblazen. De 3e theorie, tenslotte, heeft te maken met massa’s van sterrenstelsels, zegt niets over hoe bolhopen ontstaan, maar kan wel de bolhoop populaties verklaren. Observaties laten zien dat hoe groter en zwaarder een sterrenstelsel is, hoe meer metalen het bevat. Hun bolhopen zullen wat roder zijn. Kleine, lichte stelsels hebben weinig metalen: hun bolhopen hebben ook minder metalen en zullen wat blauwer zijn. Doordat grote stelsels geregeld kleine stelseltjes invangen (denk aan onze schilstelsels), zal er uiteindelijk een mix van bolhopen ontstaan die lijkt op wat we waarnemen. In de ’rest-plaatjes’ zoals Figuur 5 zijn de bolhopen vrij gemakkelijk te vinden. Echter op de miljoenen lichtjaren afstand waarop onze stelsels staan, zien we deze bolhopen slechts nog als lichtpuntjes: ze zien er uit als zwakke sterretjes. Ze zijn van sterren te onderscheiden door hun kleur, helderheid en positie. Gebruikmakend van de hoge resolutie van de HST, analyseerden we de bolhopen van de zes schilstelsels, om aanwijzingen voor mogelijke nieuwe of blauwe bolhopen te vinden. Dit zou ook weer meer steun leveren aan de botsings theorie die in de vorige paragrafen werd genoemd. Door naar de kleuren te kijken vonden we in vier van de zes stelsels een dubbele populatie, die waarschijnlijk al heel oud is. In twee van die vier stelsels vinden we aanwijzingen voor jonge bolhopen. In één stelsel is er echt een ’overschot’ aan blauwe bolhopen. In de andere vinden we een paar hele heldere blauwe bolhopen. Onze waarnemingen met de HST bieden helaas niet het bewijs of dit echt jonge objecten zijn, daartoe moeten we eerst spectra nemen van deze objecten en dat is dus iets voor de toekomst. 174 Acknowledgements Acknowledgements

There are many people and organisations I would like to thank for their help in making this thesis. First of all I would like to thank my promotor Edwin Valentijn for providing me with this position. Originally, Edwin arranged a Ph.D. which was meant to be two years of work at ESO, Munich and another two years in Groningen. However, after two years at ESO, almost all my time was spent at technical issues and almost no astronomical research was done. Edwin then provided me with a renewed contract here in Groningen. Equally important is my promotor Reynier Peletier. His many contacts, astronom- ical knowledge and quick way of seeing things and giving the right directions, were of major importance to finish this Ph.D. Both my promotores are to be thanked for their patience and trust in me in the last one and a half year. Without the help of the astro-wise group, I would probably still be busy working on the WFI data. Danny thanks for your dry humour, nice conversations and candies. Kor also for your associ- ation program and humour. Ewout thanks for your help, relaxed way of dealing with technical problems and speed-skating. Other helpful (former) members include John McFarland, Michiel Tempelaar, Willem-Jan Vriend and Ronald Vermey. I would like to thank the ’computer’ people, Eite, Martin en Wim, for always being helpful and giving useful directions about how to use the Kapteyn computer system. Teffie, thank you for your friendliness, nice drinking experiences, music taste and humour. I would like thank Peter for helping me making the front-page of this book. During all these years I have shared many office-rooms at this institute. I want to thank the following (former) office mates who created a relaxed office-atmosphere and with who I had nice conversations with. In chronological order these are Kambiz Fathi, Alicia Berciano Alba, Elif Kutdemir, Michael Pohlen, Laura Sales, Seyit Hocuk and Muhammedabdul Latife. I should of course have to thank the people from the secretariat Jacky, Hennie and Gineke. Finally, I want to thank my parents and brother for their support these last twelve years. Really finally, I want to thank Mother Nature, for providing me with the necessary inspiration to stimulate my brain. Gert Sikkema - February 2008