The Dependence of the Evolution of Early-Type Galaxies on their … · 2015-07-15 · The...

The Dependence of the Evolution ofEarly–Type Galaxies on their

Environment

Dissertationzur Erlangung des Doktorgrades

der Mathematisch-Naturwissenschaftlichen Fakultatender Georg-August-Universitat zu Gottingen

vorgelegt vonAlexander Fritz

aus Wien, Osterreich

Gottingen 2006

D7Referent: Prof. Dr. K. J. FrickeKorreferent: Prof. Dr. W. LauterbornTag der mundlichen Prufung: 17. Mai 2006

To my parents

Contents

1 Introduction 11.1 Galaxy Formation Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.2 Types of Galaxies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.2.1 Morphology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.2.2 Three–Dimensional Shape . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61.2.3 Spectral Energy Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . 71.2.4 Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

1.3 Scaling Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101.3.1 The Colour-Magnitude Relation . . . . . . . . . . . . . . . . . . . . . . . . . 101.3.2 The Kormendy Relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111.3.3 The Fundamental Plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

1.4 Galaxy Evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141.4.1 Theory meets Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141.4.2 High versus Low Densities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151.4.3 Elliptical and S0 Galaxies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

1.5 Motivation and Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2 Sample Selection and Observations 212.1 Cluster Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

2.1.1 Calar Alto Observatory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222.1.2 MOSCA Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222.1.3 Abell 2390 Cluster . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232.1.4 Low-LX Clusters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242.1.5 Selection Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242.1.6 Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

2.2 Field Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312.2.1 The FORS Deep Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312.2.2 William Herschel Deep Field . . . . . . . . . . . . . . . . . . . . . . . . . . . 322.2.3 FORS Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322.2.4 Selection Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332.2.5 Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

i

3 Data Reduction 393.1 Bias Correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393.2 Cosmic Rays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403.3 Distortion Correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403.4 Flat–fielding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423.5 Sky Subtraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423.6 Wavelength Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 433.7 Template Spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

4 Photometric Analysis 474.1 Photometry of Abell 2390 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

4.1.1 Ground-based UBI Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . 494.1.2 HST Photometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

4.2 Photometry of Low-LX Clusters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 534.2.1 Low-LX Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 534.2.2 Ground-based Photometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . 544.2.3 HST Photometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

4.3 Surface Brightness Profile Fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 604.3.1 Choice of Luminosity Profile . . . . . . . . . . . . . . . . . . . . . . . . . . . 604.3.2 The Surface Brightness Models . . . . . . . . . . . . . . . . . . . . . . . . . . 624.3.3 Error Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 664.3.4 Rest–Frame Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

4.4 Morphologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 704.4.1 Visual Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 704.4.2 Quantitative Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

4.5 Luminosity Derivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 844.5.1 Apparent Magnitudes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 854.5.2 Galactic Absorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 854.5.3 K-Correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

4.6 Luminosity Distribution and Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

5 Kinematic Analysis 935.1 Galaxy Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

5.1.1 Kinematic Extraction Methods . . . . . . . . . . . . . . . . . . . . . . . . . . 935.2 The FCQ Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

5.2.1 Error evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 995.3 Velocity Dispersions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

5.3.1 Visual Inspection of σ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1045.3.2 Aperture Correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1065.3.3 Comparison between Different Absorption Features . . . . . . . . . . . . . . . 1075.3.4 Comparison between Repeat Observations . . . . . . . . . . . . . . . . . . . . 1115.3.5 Comparison between Different Extraction Procedures . . . . . . . . . . . . . 1125.3.6 Comparison between Different Stellar Templates . . . . . . . . . . . . . . . . 113

5.4 Distribution in σ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1145.4.1 Cluster Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

ii

5.4.2 Field Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1175.5 Redshift Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1185.6 Comparison between Rich Clusters . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122

6 Galaxy Scaling Relations at z∼ 0.2 1276.1 The Local Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1276.2 The Faber–Jackson Relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1296.3 The Kormendy Relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1326.4 The Fundamental Plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135

6.4.1 Elliptical versus S0 galaxies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1426.5 M/L Evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1476.6 Comparison with Previous Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152

7 Environmental Effects on Galaxy Properties 1557.1 The Faber–Jackson Relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

7.1.1 Luminosity Dependence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1567.1.2 Mass Dependence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1597.1.3 Radial Dependence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1617.1.4 A Dependence on Galaxy Colours? . . . . . . . . . . . . . . . . . . . . . . . . 166

7.2 Stellar Population Ages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1687.3 Further Discussion and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . 173

8 Summary and Outlook 175

A Stellar Templates 179

Bibliography 185

iii

Chapter 1: Introduction 1

Chapter 1

Introduction

Since the last twenty years extragalactic astro-physics and cosmology have experienced anenormous progress. The scientific knowledgeand understanding of structure formation andthe chemical evolution within the cosmologicalframework is steadily growing. Thanks to thedevelopment of high–speed parallel computer ar-chitectures, high–resolution cosmological simula-tions on various spatial scales of the Universe canbe carried out. As a powerful complement, theconstruction of new 10 m-class telescopes suchas the four ESO Very Large Telescopes (VLT,Chile) and the twin telescopes at W. M. KeckObservatory (Hawaii) provide new light on as-trophysical issues from the observational side.Even so, several topics remain puzzling. Withinthe standard paradigm of modern cosmologytoday, a hierarchical structure formation usingCold Dark Matter particles, a wealth of observa-tional evidences on large scales of >1 Mpc1 fromthe early cosmic beginning up to the Universeas seen today can be modelled very successfully.On smaller scales in the order of kpc and below,smaller galaxies with lower–masses are formedfirst at early epochs, whereas larger high–massgalaxies are assembled through merging and ac-cretion events of smaller sub–units over longertimescales up to the recent past. In this pic-ture, different evolutionary paths are predictedfor the densest environments of clusters of galax-ies and the lowest densities of isolated individual

1One parsec (pc) corresponds to a distance of approx-imately 3.26 light years or 3.086 · 1016 m.

galaxies. The environment is suggested to playan important role to quantify and characterisethe formation and evolution of galaxies as wellas their overall properties.

Clusters of galaxies offer a unique laboratoryto study the physical properties of galaxies andtheir formation and evolution. The environmentof rich galaxy clusters, where degree of richnessrefers to the overall mass density of a cluster(typically between 1014 to 1015 M), providesan ideal location to explore how the physicalprocesses of galaxies are related to varying localnumber density, interactions with other galax-ies and exposure to a hot intra–cluster medium,in which the galaxies are embedded. In par-ticular for the most massive galaxies known,the early–type galaxies, the effects of environ-ment on their formation and evolution can berevealed. Most previous studies concentrated onthe environments of galaxy clusters as they pro-vide the opportunity to observe a large numberof early–type galaxies with less amount of tele-scope time simultaneously using the technique ofmulti-object spectroscopy.

Detailed investigations of galaxies in the nearbyUniverse (z < 0.05) have revealed very com-plex formation and evolution histories of gal-axies, thereby involving merging events, burstsof star formation, and morphological diversifi-cations. Nevertheless, the global properties ofthe galaxies, which encompass luminosities, ro-tation velocities, velocity dispersions, sizes, andabsorption line-strengths, obey very tight em-

2 Chapter 1: Introduction

pirical scaling relations, such as the Fundamen-tal Plane for early–type galaxies and the Tully–Fisher relation for spiral galaxies. Such scalinglaws put strong constraints on theoretical gal-axy formation models and on the evolution ofgalaxies. Both high signal-to-noise spectroscopyand deep multi-colour photometry is required toconstruct these powerful tools.

Local early–type galaxies from the Sloan-Digital-Sky-Survey (SDSS) show a weak trend in theirkinematics and structure parameters with galaxydensity (Bernardi et al. 2003). Galaxies residingin the dense environments of cluster cores haveslightly different properties to their counterpartsin lower density regions. Such a dependence onenvironment can only be discovered with suffi-cient number statistics but was not yet studiedat higher redshift.

It was one of the basic motivations of this studyto investigate a large sample of early–type galax-ies at z ∼ 0.2, thereby covering the densest envi-ronments of galaxy clusters down to the lowestdensities of isolated field galaxies in order to testa possible dependence of their galaxy propertieson environment. A detailed analysis of the in-ternal kinematics, structure and stellar contentof these galaxies and their involved physical andchemical processes is therefore essential. Beforethe strategy of this thesis will be given in moredetail, the two basic galaxy formation scenar-ios will be opposed to each other (next section)and the main properties of galaxies in the lo-cal Universe discussed (section 1.2). Empiricalscaling relations which are also well understoodfrom the theoretical point of view, are describedin section 1.3. An application of these tools tostudies of galaxy evolution is one of the maintopics in section 1.4, where the current state oftheory and observations as well as the role of en-vironment and stellar populations of early–typegalaxies will be addressed. Finally, the motiva-tion of this study and a brief overview of thethesis will be the content of section 1.5.

1.1 Galaxy Formation Theory

Ever since galaxies have been classified intodifferent morphological types substantial effortwas undertaken to understand their formationand evolution. Although a number of mod-els have been introduced, a complete consensusbetween theoretical approach and observationalevidence on the individual mechanisms and in-volved timescales has not yet been reached. Inthe current adopted “concordance” cosmology,two main scenarios have been suggested: themonolithic collapse and the hierarchical scen-ario. Fig. 1.1 gives a brief overview of the twodifferent theories. The wealth of empirical con-straints supports parts of both these predictions.However, a full understanding of the formationof early–type galaxies as a function of morphol-ogy, redshift and environment is still pending.The truth probably is a combination of boththeories. In the following the two alternativescenarios of monolithic collapse and hierarchicalparadigm will be introduced.In the monolithic collapse which is also calledthe classical model because historically first sug-gested, galaxies form out of individual super–giant gas clouds (Eggen et al. 1962; Larson,1974). A collapse of a cool proto gas cloud (oralso several gas clumps, see Fig. 1.1) where gasis falling to the centre in radial orbits generatesrapidly old stars with low metallicity. The starscollapse quickly from a halo to a thin rotatingdisc and star formation is initially high and en-riches the disc with heavy elements. The mor-phology of a galaxy is determined by the angularmomentum of the gas cloud and the timescale oftransforming gas into stars. If the gas is turnedinto stars on longer timescales than the timescaleof the gas cloud (high angular momentum) adisc galaxy is generated. Via dissipation the gasis settled in a disc plane where star formationcontinues until the remaining gas is completelyused up. For timescales shorter than the dy-namical friction timescale of the cloud (low an-gular momentum) a spheroidal system is formed


through processes of violent relaxation (Lynden-Bell 1967). A critical point in Fig. 1.1 is whenthe first massive stars evolve to the supernova(SN) phase. The gas is expelled from the systemby the SN-driven wind and star formation canonly continue if the dark matter halo is massiveenough. If all gas is removed a spheroidal starsystem remains and the subsequent evolutiondepends on the environment. Cluster galaxiesloose their material to the intra–cluster medium(ICM), while for galaxies in low density envi-ronments a subsequent accretion of gas onto thedisc with a quiescent star formation over an ex-tended period is possible (Larson 1975; Arimoto& Yoshii 1987). In this traditional picture, thebulk of stars in early–type galaxies are createdin a short intense burst of star formation at highredshift z >∼ 2, where the most massive struc-tures form first in the primordial density fluctu-ations (top–down scenario). The modern versionof the classical scenario is not necessarily assum-ing that galaxies form by a collapse of a singlegas cloud but rather that ellipticals are gener-ated at high redshift and on shorter timescalesrelative to spirals and that they are assembledout of gas and not of preexisting stars.

By contrast, in the hierarchical scenario early–type galaxies are formed via mergers of two discgalaxies at relatively recent times (Toomre &Toomre 1972), White & Rees 1978). Mass is ac-creted in a continuous process over an extendedtime period through multiple major and minormergers as well as accretion events of smallersubunits up to the recent past. In a first stepa bulge is formed in the assembled structure,then accretion of baryonic gas sets in and finallya thin disc is build up where star formation isinduced. Thick discs could be created throughheating of the thin disc by companions. Duringa lifetime of a galaxy this process is repeated sev-eral times and thus a significant fraction of starsis formed below redshift of unity. As a result,massive galaxies should have more extended starformation histories (bottom–up scenario). How-

Figure 1.1: A sketch of galaxy formation in theClassical and Hierarchical scenarios from Bower et al.(1999). In the Classical model the spheroidal compo-nent is formed in a rapid burst of star formation athigh redshift and a subsequent passive evolution (E)or a continuous gas accretion onto the disc over along period (isolated galaxy in a low density environ-ment). In the Hierarchical model, galaxies can changebetween different morphological types depending onthe number and strength of mergers and interactions.

ever, the latter prediction is no longer supportedby recent observations (down-sizing theory).

The hierarchical assembly of galaxies is nowa-days implemented in semi–analytic simulationswithin the framework of Cold–Dark–Matter(CDM) haloes (Kauffmann 1996; Baugh et al.1996; Cole et al. 2000). In these models, envi-ronmental processes such as galaxy interactionsor starbursts at high redshift can be successfullysimulated and tested against the observations.Contrary to the classical approach, there is anongoing not variable star formation occurring atdifferent epochs. The equilibrium is regulatedbetween the inflow of gas and the ejection rateby a SN-driven wind. A replenishment of thegas for the continuous star formation processesis warranted through the haloes of the galaxies.The mergers promote star formation and rear-range the stars in spheroids and over time ac-cretion of further gas in low densities is possiblethereby changing the morphological appearance


Figure 1.2: The Hubble “tuning fork” classifica-tion for morphological galaxy types. A range fromearly–type galaxies (ellipticals and S0 galaxies) on theleft to late–type galaxies (spirals) on the right handside is shown. Ellipticals are divided according totheir axial ratio, normal and barred spirals are sub–classified according to the fractional size of the cen-tral bulge component and the size of the spiral arms.From left to right the disc (nonexistent in ellipticals)together with the amount of gas and young stars be-comes more prominent. The presence of a central baris denoted by a “B” in the lower panel. ( c© Z. Freiand J. E. Gunn, Princeton University Press 1999).

of the system.A key difference between the scenarios is gal-axy morphology. In the classical model this par-ameter is established at an early time, whereas inthe hierarchical formation galaxies have no fixedmorphology but can vary between different mor-phological types depending on the number andstrength of interactions and merger processes.This segregation between the two scenarios canbe interpreted as a division between the roles ofnature (monolithic) and nurture (hierarchical) informing the morphology of the galaxies.While some predictions of the hierarchical modelfor the evolution of individual galaxies are sup-ported by empirical evidences, other observa-tional results favour a classical galaxy forma-tion. A comparison between the theory and ob-servation in the picture of galaxy evolution to-

gether with their pros and cons is presented insection 1.4.

1.2 Types of Galaxies

Nearby galaxies can be separated into three prin-cipal categories: Ellipticals, spirals and irregulargalaxies. Spirals comprise two classes, normalspirals and barred spirals, both can be further di-vided according to the fractional size of the cen-tral bulge component. The fundamental prop-erties of ellipticals and spirals as inferred fromtheir morphological shape, spectral characteris-tics and kinematics are distinct from each other.As a third group, irregular galaxies, which are inmany cases low–mass gas–rich systems, cannotbe assimilated in either of the two classes formedby ellipticals or spirals. However, research dur-ing the last two decades yielded a more complexpicture than the above separation into three dis-tinct types of galaxies.In the following, three criteria will be employedto construct a classification scheme for galaxies.Emphasis will be given to elliptical and S0 gal-axies and their general shape. The kinematicsand dynamics of these systems are presented insection 1.2.4.

1.2.1 Morphology

In 1926, Edwin Hubble introduced his famous“tuning fork” diagram, see Fig. 1.2, through aclassification of galaxies in terms of their mor-phology (Hubble 1926). This scheme representsalso a sequence of decreasing luminous matterand increasing dark matter from early–type tolate–type galaxies.Elliptical galaxies, also know as spheroidal gal-axies, basically comprise a one–component struc-ture of a bulge and are sub–classified into E0and E7 ellipticals according to their ellipticity ε.The ellipticity is defined over the axial ratio b/a(varying from 1 to 3) as ε = 1− b/a where a andb being the major and minor axis, respectively.


Values of ε range from 0 to 0.7. The radial sur-face brightness distribution of ellipticals is verysmooth with a high central concentration andfollows closely a r1/4-law, where r denotes thegalactocentric radius, which is know as the “deVaucouleurs”–profile (de Vaucouleurs 1948).

Spirals basically consist of two components, acentral bulge characterised by a r1/4-law and anadditional disc component with an exponentialluminosity profile proportional to e−r/rd . Thecharacteristic size of the disc is given by the discscale length rd. Normal spiral galaxies are sub–classified from Sa (“early”) to Sd (“late”) ac-cording to the fractional size of the central bulgecomponent (the disc–to–bulge D/B ratio) andthe size and form of the spiral arms. The bulgefraction decreases from Sa to Sd whereas at thesame time the dominance and open loose formof spiral arm as well as the presence of gas, dustand ionised regions of young stars increases fromearly to late spirals, which can be seen from leftto right in Fig. 1.2. For barred spirals which ex-hibit a central bar, the nomenclature is changedto SBa to SBd.

Lenticular galaxies (S0) form a transition zonebetween elliptical and early–type spirals. Theyhave a bulge and a clearly visible disc with nospiral arm structure but only a little gas frac-tion and dust. Their luminosity profile alongthe major axis is well characterised by an r1/4-law, which is overlayed by an exponential disccomponent. With respect to the ellipticals thedifferences in the shape of S0s are small (cf. sec-tion 1.2.2). The ellipticity has a uniform radialdistribution along the major axis with values upto ε = 0.7. Similar to ellipticals, generally theposition angle is constant out to large radii.

Irregular galaxies (Im) and peculiar galaxies (ab-breviated with an additional “pec”) do not fit inany of these schemes. Together with the groupof very late–type Sd spirals these objects arenot displayed in Fig. 1.2. Galaxies of this classare isolated systems with an irregular or pecu-liar isophotal shape which are neither described

Figure 1.3: Schematic drawing of de Vaucouleurs3-D classification volume (de Vaucouleurs 1959).

via a de Vaucouleurs nor an exponential lumi-nosity profile. Systems which clearly show signsof interactions with objects in the close neigh-bourhood (e.g., distortions via tidal forces) oreven undergo a merging event also fall into thisregime.Dwarf galaxies are a separate class of their own.These low–mass systems are usually classifiedon basis of their absolute B-band magnitude ofMB > −18 or their total (virial) mass with alimit of Mvir < 109M (e.g., Babul & Rees1992). Sub–classes of dwarf galaxies comprisee.g., dwarf irregulars, dwarf spheroidals, bluecompact dwarf or tidal dwarf galaxies.Another morphological classification by de Vau-couleurs (de Vaucouleurs 1959) uses a revisedversion of the Hubble system of galaxies whichis shown in Table 1.1 and in Fig. 1.3. Along theHubble sequence the stage parameter T (on a nu-merical scale −5 ≤ t ≤ +10) correlates well withseveral fundamental parameters of a galaxy, suchas photometric structure, colour index or hydro-gen content et cetera. The long axis relates tothe basic physics of the galaxy, while the cross–section displays details on the dynamics. Thepresence or absence of varieties within a class(e.g., bar or ring in spirals) should have only mi-nor influence on the dynamics rather than basic


Table 1.1: The revised Hubble sequence of galaxies according to de Vaucouleurs (de Vaucouleurs 1959).

Stage t -6 -5 -4 -3 -2 -1 0 1 2 3 4Type T E− E0 E+ S0− S00 S0+ S0a Sa Sab Sb SbcStage t 5 6 7 8 9 10 11Type T Sc Scd Sd Sdm Sm Im Im+

differences in the physical properties of the gal-axy. For example, a compact elliptical galaxy isrepresented as E−, a normal elliptical describedas E0 and an cD galaxy (which is the centralgalaxy of a cluster) is denoted as E+. Irregu-lar galaxies are specified in this scheme as Imand Im+, which include compact irregular sys-tems (cIm) and dwarf irregular galaxies (dIm).Furthermore, the classification allows to distin-guish between transition stages, e.g. S0/a, whichindicates a type between lenticulars and spirals.This classification system has been used in thisthesis too, see chapter 4.4 on page 70 for details.

One striking discovery was that the fraction ofvarious types of galaxies to the overall galaxypopulation depends on the environment. Thenumber of galaxies changes as a function of meanprojected density in clusters of galaxies. Beyonda critical density of ∼1 galaxy per Mpc3 the frac-tion of elliptical and S0 galaxies increased dra-matically and exceeds that of gas–rich spiral andirregular galaxies (Dressler 1980). At the highestdensities in the centres of clusters of galaxies el-lipticals and S0s contribute the dominant part ofgalaxy population whereas in the lowest densitiesof the field isolated spirals are the most frequentgalaxy type. This is known as the morphology–density relation.

A more quantitative classification of morphologi-cal types provides the bulge–to–total ratio B/T ,which gives the contribution of the bulge com-ponent to the total luminosity of a galaxy. Thisratio decreases from early to late–type galaxies,with the highest values for bulge dominated el-lipticals and the lowest values for spiral Sd and

irregular galaxies. In the case of a non-detectablebulge B/T = 0. For a more detailed descriptionand an application of this morphological classi-fication the reader is referred to chapter 4.4.2.

1.2.2 Three–Dimensional Shape

Since the first observations in the 18 century upto the end of 1970s, elliptical galaxies were in-terpreted as galaxies without any substructure.The effects of projection prohibited insights intothe intrinsic shape of ellipticals. For a longtime, therefore, they were thought of axialsym-metric ellipsoids or triaxial systems, which areflattened by rotation and consisting entirely of(metal rich) population II stars and devoid ofcold gas. Shortly after the beginning of the Uni-verse, these old stellar systems should have beenformed via a single collapse of a huge gas cloud.In the late eighties detailed high–signal–to noiseinvestigations of the intrinsic isophote shapes oflocal elliptical galaxies were carried out, reveal-ing substantial substructure (Carter 1987; Ben-der 1988). The isophotes which represent con-tours of constant surface brightness of an ob-served galaxy can be modelled by ellipses to de-rive the variations of the ellipticity from the cen-ter to the edge. Many ellipticals do not appearto feature a perfectly elliptical shape when pro-jected onto the sky. Usually, the deviations froma perfect ellipsoid symmetric along both majoraxes are with <∼ 4% very small, described bythe parameter a4 which defines a separation intoboxy or discy isophotes. Typical values for theratio of a4 to mean radius a0 are in the range of−0.02 ≤ a4/a0 ≤ 0.04. Discy ellipticals have in


addition a parameter a6 6= 0.The shape of the isophotes is very important as itcorrelates with numerous properties of ellipticalgalaxies. For this reason, Kormendy & Bender(1996) proposed a revision of the Hubble schemefor the elliptical galaxies and suggested that theycan be divided basically into two classes accord-ing to their isophotal shape:

• Boxy E(b): Generally these are lumi-nous galaxies (MV <∼ −22.0) which aresupported by anisotropic velocity disper-sions (σ). These almost isothermal ob-jects show a slow or zero rotation rate of(vrot/σ)∗ < 1 along the major or minor axis.Deviations from perfect ellipses are nega-tive (a4 < 0) and they often feature hotX–ray gas (9.8 ≥ log (LB) ≥ 11.2 L) orradio emission (∼1020–1025 W/Hz). Theircentres often show signs of kinematically pe-culiar/decoupled cores, which is believed tobe a by–product of mainly stellar mergers.Deep HST imaging detected shallow (cuspy)core profiles (Faber et al. 1997).

• Discy E(d): Usually these are fainter gal-axies (MV >∼ −20.5) which are sustainedby ordered rotation along the minor axis(vrot/σ)∗ ∼ 0.2 − 1.5. Variations of theisophotes are specified as a4 > 0. Withinthe resolution of the Hubble Space Telescope(HST), high density power–law profiles inthe galaxy centres were found but no signsof a distinct core was revealed. Kinemati-cal peculiar cores and X–ray emission fromhot gas are quite rare (log (LB) <∼ 9.2 L).This together with the presence of a faintdisc may suggested that dissipation was anessential ingredient during their formation,which includes also mergers of two spiralswith dissipation involved.

Besides these two groups also intermediate typesexist. Elliptical galaxies with luminosities be-tween −22.0 <∼ MV <∼ −20.5 are represented inboth classes. Furthermore, many discy L∗ el-

Figure 1.4: Spectrum of an elliptical galaxy fromKennicutt (1992).

lipticals contain faint discs which usually con-tribute a few per cent but sometimes up to ap-proximately 30% of the total luminosity of thegalaxy (Rix & White 1990). Note that the se-quence from boxy to discy ellipticals is not neces-sarily a continuous sequence (Kormendy & Ben-der 1996).Being aware of these different physical proper-ties of discy and boxy ellipticals, several authorsproposed different formation scenarios. By defi-nition lenticular galaxies comprise a bulge and aclearly visible disc component. As a consequenceof the disc for most cases the disciness parametera4 takes values a4 > 0 and a6 6= 0. Therefore,from a morphological point of view, S0 galaxiesare very close to discy ellipticals. In this contextit seems obvious to consider the Hubble tuningfork from discy E(d) over S0 to spiral galaxies asa continuous sequence of decreasing bulge–to–disc ratios (Bender et al. 1992).

1.2.3 Spectral Energy Distribution

The morphological galaxy types along the Hub-ble sequence correlate strongly with the overallproperties of the Spectral Energy Distribution(SED). From early to late–types the broad bandcolours become on average bluer. Generally, the


Figure 1.5: Spectrum of an Sc spiral from Kenni-cutt (1992).

nucleus of elliptical galaxies is redder than theouter regions. This colour gradient is consideredbeing solely due to metallicity variations as for agiven age a stellar population appears to be bluerwhen the metal abundance is lower. The coloursof spirals show a very uniform radial distribu-tion in the disc. Globally, their disc componentbecomes bluer with later–types, whereas the redcentral bulges of spiral galaxies feature similarcolours to the ellipticals and S0 galaxies.

A similar view is gained from the proper-ties of integrated optical spectra at rest–framewavelengths. Early–type galaxies comprisestrong absorption lines, in particular CN, CaH+K, G–band, Mgb and the Balmer series(Hδ, Hγ Hβ), with a characteristic break intheir spectra, called the 4000A break D4000.Emission lines are rare and only weak ifpresent. The strength of emission lines, typi-cally apparent via [OII]3727, Hβ [OIII]5007 andHα(+[NII]6548/6584), increases from early tolate–type spiral galaxies. While a combinationof age and metallicity in the stellar content ofE+S0 galaxies is responsible for the depths of ab-sorption lines, the fraction of hot, high–massivestars in the spiral galaxy population correlateswith the emission line strengths. Examples of

Figure 1.6: Spectrum of a very late–type galaxy(Sm/Im) from Kennicutt (1992).

a typical elliptical, an Sc and a very late–typeSm/Im spiral taken from the Kennicutt (1992)catalog are shown in Figs. 1.4 to 1.6. Note thatthese spectra have been normalised to the fluxat 5000A.

The colour trend seen for local galaxies along theHubble sequence can be explained by looking atthe Star Formation Rate (SFR) ψ(t). If the StarFormation History (SFH) of ψ(t) for a specificgalaxy type is approximated by an exponentiallaw of ψ(t) ∝ e−t/τ , where τ is the characteristice-folding time (i.e. the timescale to turn a massM/e into stars, typically τ ≈ 1 − 5 Gyr), thenτ increases from early to late–types (Larson &Tinsley 1978). The stellar populations of early–type galaxies indicate a short intense star burstat high redshift, which is followed by a slow qui-escent evolution. During this period, the popula-tions become mainly the red, cool, very old pop-ulation II giant stars, whereas the blue, very hot,high–mass stars are rapidly destroyed in super-novae explosions of Type Ia. The smooth radialcolour distribution of spiral galaxies in the discsuggest that their star formation rate is constantindependent on the amount of cold gas. Indeed,the SFR of spiral galaxies shows a shallower de-crease or may be even a constant rate, which


yields to hot, massive and young supergiant starsof population I and therefore blue colours.However, today there is no doubt that at leastsome elliptical galaxies comprise young stellarpopulations. The situation gets even more com-plex as at higher redshifts post–starbust gal-axies which are devoid of emission lines andtherefore without any signs of star formationcould be miss-classified as S0 galaxies. These so-called E+A (or k+a) galaxies show very strongBalmer absorption lines (preferably in Hδ) andcontribute approximately 20% to the populationin clusters at z ∼ 0.5 (Poggianti et al. 1999).

1.2.4 Kinematics

With respect to the internal kinematics of gal-axies, i.e. the motion of their stars and gas, twomain classes can be distinguished. Early–typegalaxies which comprise the hot, pressure sup-ported spheroidal elliptical and lenticular galax-ies, are stabilised by the random motion of thestars as expressed in their velocity dispersion (σ).This regime is also valid for the dominant bulgesof early spiral galaxies. For this reason, thesesystems are referred as Dynamically Hot Galax-ies (DHGs, e.g. Dressler et al. 1987), where thevelocity dispersion exceeds their rotation veloc-ity (Vrot) at all radii. In contrast, for spiral gal-axy discs the movement of their stars and gas arestabilised by ordered rotation. Therefore, theyare dynamically cold systems where the sourceof their angular momentum could be the resultof tidal torques created during the process of discformation (e.g., Silk 2000). Both quantities, thevelocity dispersion and the rotation velocity, area function of galactocentric radius.Detailed studies of the kinematics of ellipticalgalaxies in measuring the velocity Vrot and dis-persion σ as a function of radius from the galaxycentre exist only for a minority of nearby gal-axies. In most cases, the central galaxy profilewith the systemic radial velocity vrad defined bythe position of lines and the profile width givingthe dispersion at the centre σ0 can be derived.

These difficulties are based on the fact that themeasurements of rotation curves in early–typegalaxies rely solely on absorption lines which areusually relatively weak. As the luminosity de-creases very rapidly from the centre, the deriva-tion of rotation curves are limited to radial dis-tances of approximately r < 10 kpc (Davies etal. 1983). Nevertheless, these measurements aresufficient to detect the maximal rotation velocityVmax which is generally located at r ∼ 2 kpc. Be-yond this characteristic radius the rotation curvedeclines slowly. Furthermore, as projection ef-fects are impossible to correct for elliptical andS0 galaxies, the true rotation velocity remainsinaccessible.The internal velocity dispersions of early–typegalaxies give insight into two basic properties,their galaxy mass and their formation. A con-siderable fraction of the kinetic energy in thesesystems is distributed in random motions of thestars, which is expressed via the (stellar) velocitydispersion. Applying the virial theorem as

M v2 = GM2(3Re)−1 (1.1)

where v2 is the mean square of velocitiesweighted by the mass of the stars and Re be-ing the radius containing half of the light (ormass) of a system, the galaxy mass M can bederived using this “chaotic” distribution of veloc-ities. Moreover, as the velocity dispersion shouldhave not been changed after the initial collapseof the galaxy, it provides information about theprocesses which played a role during the forma-tion of the stellar system. The stellar kinetic en-ergy is a measure of the binding energy per par-ticle and hence gives the rate of dissipative pro-cesses which occurred during the collapse phase(Fish 1964). For this reason, the velocity disper-sion is a key parameter which must be accountedfor in any galaxy formation theory.Faber & Jackson measured the internal veloc-ity dispersions of 25 elliptical galaxies (Faber &Jackson 1976) and discovered that the luminosi-ties of these galaxies are tightly correlated with


Figure 1.7: Colour–Magnitude Relation for Virgoand Coma cluster galaxies from Bower et al. (1992).Open and filled symbols are Virgo and Coma galaxies,respectively. Elliptical and S0 are denoted by circlesand triangles, S0/a and later types by stars. Solidlines represent the median fits to the relation and thescatter around the fit for early–type galaxies amountsto only 0.05 mag. The dashed line shows the expectedComa cluster relation predicted from the Virgo zero–point plus a relative distance modulus.

their spread in velocities according to

L ∝ σα0 , (1.2)

where L is the luminosity and σ0 is the cen-tral stellar velocity dispersion of an early–typegalaxy. In the B-band the power–law exponentin the so–called Faber–Jackson Relation (here-after FJR) yields α ≈ 4, which can be easilyexplained if galaxies are not enveloped by dark–matter haloes. Usually, the velocity dispersionsof elliptical and S0 galaxies increase with abso-lute magnitude as LB ∝ σ4

0. The central velocitydispersion of the bulges of spirals also follows thisrelation. However, the slope is not valid for allluminosities, e.g., the bright cD galaxies locatedin the centres of galaxy clusters have a σ0 whichis slightly less than the ∼ L4 curve. Due to thedependence of L and σ, small errors in the ve-locity dispersions propagate into large errors inabsolute magnitude. For this reason, both the

velocity dispersions and the luminosities have tobe measured with high accuracy. As the FJRis a strong correlation between the two observ-ables it provides a powerful test on the formationand evolution of elliptical and lenticular galax-ies. As for distant galaxies the apparent sizesdecrease, a measurement of the velocity disper-sion is not restricted to the central parts of thegalaxy but contains a larger area or maybe thewhole size of the system. For this reason, whendistant galaxies are compared to galaxies in thelocal Universe with central velocity dispersions(σ0), the measured velocity dispersions have tobe corrected for aperture size. More details onthe kinematics, in particular the derivation of ve-locity dispersions, will be discussed in chapter 5on page 93.

1.3 Scaling Relations

Since a long time it is well known that several ob-servable properties of early–type galaxies whichcharacterise the dynamical state, size or chemi-cal composition correlate with their luminosity.Typical parameters include the effective radiusRe (Fish 1964), the central velocity dispersion σ0

(Faber & Jackson 1976), the integrated colours(Bower et al. 1992) or metal absorption line in-dices such as the Mg2 index (Bender et al. 1993).Apart from the Faber–Jackson Relation whichwas already presented in the previous section,in the following, a short review of the basic em-pirical relationships together with their involvedquantities will be given.

1.3.1 The Colour-Magnitude Relation

Early–type galaxies in clusters at low–redshift(z < 0.05) form an extremely homogeneous classof galaxies. It has been known for a long timethat for early–type galaxies exists a fundamentalrelation between their colour and magnitude, re-ferred to as the colour–magnitude relation (e.g.,Sandage & Visvanathan 1978). The relation,which is often also called red–sequence, implies


a link between the mass of the stellar popula-tion and its age or metallicity. Presently, thelatter parameter is being favoured because high–redshift clusters have bluer colour–magnitude re-lations with a small scatter than in local clusters(e.g., Kodama & Arimoto 1997).

In the local Universe, early–type galaxies obeya tight colour–magnitude relation (CMR) witha homogeneity across various clusters. Brighterearly–type galaxies are redder than less luminousones. At given absolute magnitude, the disper-sion in optical and infrared colours is small withless than 0.05 mag (Bower et al. 1992). Since thecolours of the stellar populations evolve with age,the width of the CMR puts strong constraintson the age scatter at a given absolute magni-tude. For the local Coma and Virgo clusters therms scatter in age was less than 15%, which im-plies that these galaxies have most likely formedat high–redshift of z > 2. A widely supportedinterpretation suggests that early–type galaxiesare more metal rich and that the star forma-tion processes in cluster early–types have takenplace and already ceased at an early cosmic timethat the resulting effects of a possible spread intheir formation epoch are not visible throughtheir broad–band colours. Evidence for this isaccumulated by the redshift evolution of theCMR. The almost constant slope of the CMRwith redshift indicates that this scaling relationis not an age-mass sequence (e.g., van Dokkumet al. 2000). Moreover, the small scatter of theCMR in massive clusters at high–redshift outto z = 1.2 suggest that the stellar populationsof early–type galaxies are uniformly old zf > 2and quiescent (Ellis et al. 1997; Lidman et al.2004). However, there are some hints for a flat-ter colour–magnitude sequence at high–z (vanDokkum et al. 2001a) and a possible evolutionin the scatter of the colours (Holden et al. 2004).This picture gets further support by a modestevolution of the colour–gradients in cluster gal-axies from z ∼ 0.4 to z = 0 in the V RI HST filterpassbands (Saglia et al. 2000). This homogene-

ity of early–type galaxies in clusters needs to beconsidered with some considerations. The pop-ulation of early–type galaxies in distant clustersat high–redshift does not only comprise possibleprogenitors of present–day early–type galaxies.Some progenitors at z ∼ 1 might not have yetbeen accreted onto the cluster or might not bevisually detectable due to ongoing morphologicaltransformations from late–type star forming toearly–type passive galaxies. For this reason, thetightness of the CM sequence in clusters at high–z could partly represent a selection effect (theso–called progenitor bias; van Dokkum & Franx2001b). Most dramatic scenarios are rather un-likely due to the evolution of the morphology–density relation (Dressler 1980; Dressler et al.1997).Up to now, there are only a few studies whichfocus on the CMR for distant early–type fieldgalaxies (Schade et al. 1999; Bell et al. 2004).Results suggests that there is a CM sequence inthe field out to z ∼ 1, albeit with a considerablylarger scatter than in clusters.

1.3.2 The Kormendy Relation

In 1977, John Kormendy published a pioneer-ing study on the structural parameters of 35 el-liptical galaxies (Kormendy 1977). The surfacebrightness and the size of an elliptical galaxy arerelated as

〈µe〉X = α+ β logRe (1.3)

where 〈µe〉X is the surface brightness in the pass-band X in units of mag/arcsec2 and Re denotesthe effective radius, given in kiloparsecs. Re isalso referred as the radius which contains halfof the total galaxy light distribution. Becauseof the colours and colour gradients in the stel-lar populations of early–type galaxies the quan-tities of 〈µe〉 and Re depend on the observedwavelength. Kormendy measured for the zero–point α = 19.74 and for the slope β = 3.02 inthe Johnson B-band. The Kormendy Relation


(hereafter KR) holds for both types of early–type galaxies, elliptical as well as S0 galaxies.In general, brighter, i.e. more luminous galax-ies, have larger effective radii and fainter sur-face brightnesses. With a relatively large dis-persion of 0.28 mag in the B-band for normalellipticals the measurements of both structuralparameters have to be necessarily performedin a very accurate way. Subsequent studiesdemonstrated that the magnitude–size relationholds for bright early–type galaxies in differentpassbands from the optical to the near-infrared(Hoessel et al. 1987; Lubin & Sandage 2001; LaBarbera 2003). Based on Virgo cluster galax-ies, Capaccioli, Caon & D’Onofrio (1992) foundthat the ellipticals, early-type dwarfs, and bulgesof S0s and of spirals form two distinct familiesin the plane of the effective parameters: Brightearly–type galaxies follow the KR which mayhave undergone a merger phenomenon, whereasa so–called ‘ordinary’ family comprising galax-ies fainter than approximately MB = −19.3 aremore diverse and heterogenous. This effect isalso well–known as the “dichotomy” betweenearly–type galaxies. To solve this issue, Graham& Guzman 2003 suggested that the dichotomy isonly valid between bright and dwarf early-typeswith a continuous transition plane of structuralparameters between the sub–classes. Differencesin their properties and in the slope of their re-lations can be explained by a change in surfacebrightness profile slope (γ′) with their luminos-ity.

The magnitude–size relation is the projection ofthe Fundamental Plane (FP) along the velocitydispersion onto the photometric plane and a veryuseful probe to study the luminosity evolutionof early–type galaxies. As this relation does notcomprise galaxy kinematics, it can be studiedto fainter magnitudes and therefore higher red-shifts. A number of studies have utilised theKR to investigate the luminosity evolution ofE+S0 galaxies up to redshifts of one (Ziegler etal. 1999; Lubin & Sandage 2001; Treu et al.

2001b; La Barbera 2003; Holden et al. 2005).Most previous works suggested that the evolu-tion of the zero–point of the KR can be explainedby a pure Tolman signal which is superposed tothe luminosity evolution of stellar populationshaving a high formation redshift (zf > 2). TheTolman test predicts that the bolometric surfacebrightness magnitudes decrease as (1+z)4 and ithas been applied as an observational verificationthat redshifts really correspond to an expansionof the Universe, and not to an once assumed phe-nomena of light dimming (Sandage & Perelmuter1990). A drawback of the Kormendy relation isthat the samples have larger scatter than in theFP because they rely only on the photometricmeasurements and thus are more affected by se-lection biases (Ziegler et al. 1999). At z = 1.2the observed scatter of cluster early–type galax-ies is 0.76 mag around a fixed size (Holden et al.2005). Nevertheless, the amount of luminosityevolution found for the Kormendy relations arein good agreement with results obtained by theevolution of colours or from the FP of early–typegalaxies.

1.3.3 The Fundamental Plane

In the search of a “second parameter” to explainthe relatively large scatter of the correlationsbetween different parameters mentioned in thebeginning of this section, large spectrophotome-trical surveys during the 1980s discovered a re-lationship, called the Fundamental Plane. Thiscorrelation represents a refinement of the Faber–Jackson Relation and is a powerful tool to mea-sure the star formation history of early–type gal-axies at a given mass by studying the evolutionof the Fundamental Plane with redshift.In a three dimensional parameter space, definedby three observables, the effective radius Re, ef-fective surface brightness within Re, 〈µe〉, andvelocity dispersion σ, the Fundamental Plane(FP) establishes a tight correlation (Djorgovski& Davis 1987; Dressler et al. 1987) in the fol-


lowing form:

logRe = α log σ + β 〈µe〉 + γ , (1.4)

where α and β are the slopes and γ denotes theintercept. The half–light radius Re is given inkpc, σ is in km s−1 and 〈µe〉 is in units of magarcsec−2. This empirical relationship relates gal-axy structure and size (Re and 〈µe〉) to kinemat-ics (σ). As the Hubble constant enters the rela-tion via the effective radius in kpc, the value ofγ depends on H0. In the following, the physicalorigin of the FP will be outlined as well as itsconnection to the mass–to–light ratio and appli-cation to constrain the formation and evolutionof elliptical and S0 galaxies.The actual existence of the FP is a remarkablematter of fact. Any galaxy formation and evo-lution theory must account for the tightness anduniversality of the FP, which amounts along theedge-on projection to only 0.08 rms in logRe(Bernardi et al. 2003). Thanks to this small in-trinsic scatter of the local FP, the formation andevolution of E+S0 galaxies can be constrainedwith high precision.In general, FP studies presume that E+S0 gal-axies are a homologous group, i.e., that they ex-hibit a similar structure. Under this assumptionthe total galaxy massM (including dark matterif present) is proportional to its virial mass Mas (Treu et al. 2001b):

M ≡ σ2ReG−1 · C = M (1.5)

where the constant C depends on the mass, i.e.,weak homology holds. Let us further define aneffective luminosity L

logL = −0.4 〈µe〉 + 2 logRe + log 2π. (1.6)

Based on the definitions of the equations 1.5and 1.6, the (effective) mass–to–light (M/L) ra-tio can be easily deduced in terms of the FPas M/L ∝ 100.4〈µe〉σ2R−1

e . By comparing theM/L ratio to the value as predicted by the FP ata given effective radius and velocity dispersion,

Figure 1.8: Fundamental Plane relation for dy-namically hot galaxies from Bender et al. (1992).Elliptical galaxies are ordered according to their B-band luminosity, giant E (red squares), intermedi-ate E (green circles), bright dwarf E (open triangles),spiral bulges (blue stars), compact E (filled triangles,e.g., M32) and dwarf spheroidals (crosses). Note thatMB encompasses from giant to compact ellipticals ap-proximately 8 orders of magnitudes and thus is notan intrinsic property of a few specific objects.

the effective surface brightness can be eliminatedwhich yields

M/L ∝ 10−γ

2.5β σ10β−2α

5β R2−5β

5βe . (1.7)

As the scatter of the FP can be translated into ascatter inM/L the tightness of the FP implicatesa very low scatter inM/L for E+S0 galaxies, too.The evolution of the Fundamental Plane canbe expressed in terms of its zero-point γ thatvaries as a function of redshift z. So far, no dra-matic change in the slopes with redshift were re-ported. However, an accurate measurement istough as large samples are required and intrinsiceffects such as completeness corrections and fit-ting methods have to be accounted for (Kelsonet al. 2000b). If a galaxy sample at z > 0 whichobservables are characterised by a superscript iis considered, the offset of the mass-to-light ratio


from the local value can be addressed via

∆ log(M/L)i = ∆(

10β − 2α5β

)log σi

+ ∆(

2− 5β5β

)logRie −∆

(γi

2.5β

), (1.8)

where ∆ describes the difference of the quanti-ties at two different redshifts and γi is logRie −α log σi − β 〈µe〉i (Treu et al. 2001b). Under as-sumption that the FP coefficients α and β areconstant the evolution of the M/L is indepen-dent on the mass M , and by adopting that thereis no structural and dynamical evolution (i.e. theeffective radius Re and the velocity dispersion σare constant), the evolution of ∆log(M/L)i de-pends only on the evolution of the parameter γi,hence it follows

∆log(M/L)i = −∆γi

2.5β. (1.9)

By measuring γi for a number of galaxies athigher redshift and by comparing it to the inter-cept of the local Universe, the average evolutionof ∆log(M/L) can be derived as

〈∆log(M/L)〉 = −〈∆γ〉2.5β

. (1.10)

As the evolution of this effective M/L ratio cor-responds to an evolution of the dynamical (stel-lar and dark matter) M/L ratio, the applicationof the FP becomes a powerful diagnostic toolto assess the evolution of the stellar populationsin the galaxies under consideration. A practicalrealisation to observational data will be demon-strated in chapter 7.2 on page 168.

1.4 Galaxy Evolution

1.4.1 Theory meets Observations

Theoretical galaxy formation models based on ahierarchical structure growth use numerical N -body computations to simulate the gravitationalclustering of DM haloes from the initial density

perturbations up to the present–day Universe.These CDM simulations can reproduce the largescale structure which range from cluster of gal-axies and superclusters up to ∼100 Mpc, such asthe Great Wall (Geller & Huchra 1989) but alsoaccount for the voids in between. The hierarchi-cal scenario predicts that early-type galaxies ingalaxy clusters were assembled at an early epochat high redshift (z >∼ 2) because the redshift of acollapse on galaxy scales is heavily driven by thepresence of the surrounding overdensity. There-fore, galaxies forming out of the highest peaksof primordial density fluctuations which collapseat z >∼ 2 quickly merge together within groupsand their star formation gets truncated whenthey have used up their gas. However, only forthe dense environments of rich clusters, clustergalaxies experienced their last major merger atz ∼1–2 (Kauffmann 1996). Because of ongoingmerging events, the population of early-type gal-axies in lower–densities should be more diverse.For individual galaxies, the hierarchical modelpredicts that low–mass galaxies were formed firstin the early cosmos and larger massive galax-ies subsequently build up in merging or accre-tion of smaller systems at a later stage. Semi–analytical CDM models derive explicit age varia-tions for early-type cluster galaxies and E+S0s inlow-density regions. For clusters, models predictellipticals to have a mean luminosity–weightedage of 9.6 Gyrs and lenticular galaxies to beyounger by ∼1 Gyr (Baugh et al. 1996; Coleet al. 2000). Both types, ellipticals and S0 gal-axies, indicate a weak trend that fainter galax-ies (−17.5 > MB > −20.1) are older. On thecontrary, for early-type galaxies in low-densityregions the hierarchical cluster models predicta considerably broader age spread over a largerluminosity range and mean luminosity-weightedages of ∼5.5 Gyrs. In addition, brighter fieldgalaxies should feature on average younger agesand comprise more solar element abundance ra-tios than their cluster representatives (Thomas,Maraston & Bender 2002).


A wide range of formation histories is suggestedfor galaxy clusters of similar present–day rich-ness class (Kauffmann & Charlot 1998). Forthe timescale of the cluster accumulation strongradial age gradients are predicted, whereas thespread in metallicities between cluster and fieldgalaxies should be similar. This would implythat scaling relations such as the CMR and FPshould be more diverse with a larger scatter forthe field, which is not endorsed by observations(Schade et al. 1999; van der Wel 2005). How-ever, in support of the hierarchical scheme comesthe observed large increase of the merger fractionwith redshift (van Dokkum et al. 1999). Furtherempirical evidence from a dynamical perspectiveis accumulated by interacting and ongoing merg-ing events of galaxies at low redshift and thedetection of ellipticals with disturbed morpholo-gies, such as kinematically peculiar (decoupledor counter–rotating) cores, dust lanes, rippleset cetera (e.g., Franx & Illingworth 1988; Kor-mendy & Djorgovski 1989), or the excess of bluegalaxies with redshift regarding to the colour–sequence of passive elliptical galaxies (Butcher–Oemler effect, Butcher & Oemler 1984). On theother hand, models have still problems with thehigh angular momentum of spiral galaxies (Coleet al. 2000) and too low predicted [α/Fe] ra-tios for early–type galaxies (Thomas, Maraston& Bender 2002). In particular, the observationsfind that the [α/Fe] enhancements are increas-ing with velocity dispersions whereas the modelssuggest the opposite trend.

In the classical galaxy formation scenario early–type galaxies are formed dissipationally in arapid, single burst of star formation at high red-shift (z >∼ 2). After this initial period, galax-ies evolve passively without merging and yielda population of old ellipticals where their meanmetallicity scales with galaxy mass (Eggen et al.1962; Larson 1975). As already outlined in sec-tion 1.1, SN feedback and the chemical enrich-ment of the ICM are crucial ingredients in thecurrent monolithic formation scenarios. Based

on the stellar population properties of local el-lipticals, the estimated timescales required to de-velop a wind are relatively short with ≤1 Gyr(Pipino & Matteucci 2004). These models canexplain the majority of empirical constraints rel-ative to the stellar populations of early–type gal-axies, such as CMR, Mg–σ relation, FP and theincrease of the [α/Fe] ratio with galactic mass,the existence of metallicity gradients and thecharacteristic r1/4 surface brightness profile.Qualitatively the two scenarios seem to pre-dict opposite trends. The monolithic collapsemodel suggests that ellipticals form on shortertimescales than spirals and the hierarchicalscheme predicts that spirals form before ellip-ticals which continue to assemble until recenttimes.

1.4.2 High versus Low Densities

In the nearby Universe, inconsistent results havebeen acquired regarding any possible differencebetween early-type galaxies in high densities(galaxy clusters) and E+S0 galaxies in low den-sity environments (isolated field objects). Forexample, de Carvalho & Djorgovski (1992) de-rived from a subset of cluster and field early-typegalaxies taken from the “Seven Samurai” group(Faber et al. 1989) and a second sample drawnfrom Djorgovski & Davis (1987) that field ellip-ticals show a larger scatter in their propertiesindicating that they consist of younger stellarpopulations than cluster galaxies. Bernardi etal. (1998) analysed a large sample of ENEARfield and cluster galaxies and found slight zero-point changes in the Mg2 − σ relation. Theyexplain this as an age difference, with field ob-jects being younger by ∼1 Gyr. However, theyconclude that the bulk of stellar populations ofE+S0 in both environments has been formedat high redshifts (z >∼ 3). James & Mobasher(1999) investigated near-infrared (NIR) spectraof 50 ellipticals in three nearby clusters and inthe field, using the CO (2.3µm) absorption fea-ture to explore the presence of an intermediate–


age population. They detected no stronger COabsorption for the field ellipticals. Very isolatedfield ellipticals show a very homogenous popu-lation and a small range of metallicity with nosign of recent star formation (SF). In groups, el-lipticals have a wide range in metallicity, mostlyshowing evidence for an intermediate-age popu-lation, whereas in rich clusters they exhibit in-termediate properties in metallicity and CO ab-sorption. Kuntschner et al. (2002) detected ina sample of nine local early-type galaxies (fivemorphologically disturbed) in low-density envi-ronments (LDR) no strong ongoing SF. The re-sults were compared to cluster E+S0s in Fornax.The ages of the LDR galaxies are spread overa broad distribution, similar to that of FornaxS0 galaxies and being on average by 2–3 Gyrsyounger than the E+S0s in Fornax. These LDRgalaxies indicate 0.2 dex higher metallicities andsuper solar Mg/Fe ratios (in conflict with semi-analytical models), which suggests that the for-mation of E+S0 galaxies in low-densities con-tinues to z <∼ 1, whereas in clusters most starshave already been generated at z >∼ 2. Recently,Sanchez-Blazquez et al. (2003) studied 98 E+S0galaxies in the field and in clusters and foundhigher C4668 and CN2 absorption line strengthsfor the field population. They interpret this as adifference in abundance ratios arising from dif-ferent star formation histories. However, bothfield and cluster E+S0s show similar relations inMgb−σ and 〈Fe〉 − σ.

At higher redshift differences between field andcluster galaxies should become more apparent.Recent results from studies. based on the Funda-mental Plane at intermediate redshift (z ≤ 0.5),indicate no significant variations between thecluster and field early-type populations (vanDokkum et al. 2001c; Treu et al. 2001b; Rusin &Kochanek 2005). With respect to the mean ageof these populations, field galaxies seem to com-prise slightly younger stars than the cluster pop-ulation, whereas the majority of stars must haveformed at a much higher redshift of zf > 2. How-

ever, at higher redshift (z ∼ 0.7), some investiga-tions derive a significant offset between field andcluster galaxies (Treu et al. 2002; Gebhardt etal. 2003). Differences might mainly arise due tothe low number of analysed galaxies. As the se-lection criteria differ strongly among these stud-ies, the samples might also be affected by theprogenitor bias (van Dokkum & Franx 2001b).

1.4.3 Elliptical and S0 Galaxies

Over the last years a multiplicity of investiga-tions of distant rich clusters have been performed(Ellis et al. 1997; Dressler et al. 1997; Kelsonet al. 2000b; van Dokkum et al. 2000; Ziegleret al. 2001a; Wuyts et al. 2004). Most of thesestudies can be reconciled with the picture of amonolithic collapse with a high redshift forma-tion of the stellar populations of E+S0 galax-ies. Results from these distant clusters have notfound any differences in the properties of E andS0 galaxies (e.g., Kelson et al. 2000b). Recently,in a re–analysis of two high redshift clusters atz = 0.58 and z = 0.83 no environmental depen-dence of the FP residuals was detected (Wuyts etal. 2004). When looking at the residuals of theFP, and suggesting that the residuals correlatewith environment, it is difficult to distinguish ifthis effect is due to changes in velocity disper-sion, size or luminosity of the galaxies. Selec-tion effects have strong influence on the param-eters and can also mimic possible correlations.In a study of ∼9000 early-type galaxies from theSDSS (Bernardi et al. 2003), a weak correlationbetween the local density and the residuals fromthe FP was revealed, in the sense that the resid-uals in the direction of the effective radii increaseslightly as local density increases. However, theoffset is quite small and subject to selection andevolutionary effects. The open question still toaddress is, how this dependence occurs.Looking at the morphology, the formation andevolution of lenticular galaxies is different andstands in contrast to elliptical galaxies. Deepstudies of galaxies in distant rich clusters using


the WFPC2 camera onboard the Hubble SpaceTelescope revealed that S0 galaxies show a rigor-ous evolution with redshift in these dense envi-ronments (e.g., Dressler et al. 1997). AlthoughS0 galaxies form the dominant population in lo-cal rich clusters of ∼60%, at intermediate red-shift (z ∼ 0.5) spiral and disturbed galaxiescompose the major part of the luminous galax-ies, whereas S0 galaxies are less abundant (10–20%). Schade et al. (1999) studied early-typefield galaxies at intermediate redshifts (z ∼ 0.5)and detected [OII]λ3727 emission lines in about1/3 of these galaxies, which indicates ongoingstar formation. Furthermore, in about the samefraction of faint spheroidal Hubble Deep Fieldgalaxies significant variations of internal colourswere found, frequently showing objects with bluecores (Menanteau et al. 2001). The authors con-clude that at z ∼ 1 about half of the field S0galaxies show clear signs of star formation activ-ity. Some evidence for young populations in lo-cal S0 galaxies was recently found by Mehlert etal. (2003). Based on high signal–to–noise spec-troscopy of early-type galaxies in the Coma clus-ter, two families of S0 galaxies were detected, onegroup with old (∼10 Gyrs) stellar populationscomparable to ellipticals and a second one withvery young average ages (∼2 Gyrs) and weakermetallic lines.

These results seem to imply that galaxy trans-formation via interaction is an important phe-nomenon in clusters. Due to the large velocitydispersion mergers are less frequent in rich clus-ters, whereas effects such as ram–pressure strip-ping by the hot intra cluster medium or tidal in-teractions between the galaxies are more likely.A unique mechanism for the transformation intoS0 galaxies is still missing to explain the strongdecrease in the frequency of S0’s since the last5 Gyrs (z ∼ 0.5). A possible scenario is thatfield spiral galaxies falling into the cluster cen-tre experience a starburst phase, resulting in theButcher–Oemler effect. Ram–pressure strippingby the ICM (also maybe through tidal stripping)

over a short time-scale of less than one Gyr,could cause the wide-spread and rapid decline instar formation leading to post–starburst galaxiesand red passive spiral galaxies (Barnes & Hern-quist 1992). Harassment by the tidal field of thegalaxy cluster and high speed encounters havea non negligible effect on the following passiveevolution of a galaxy by removing stars from thedisc which may end up in an S0 galaxy (Mooreet al. 1996; Poggianti et al. 1999).

In terms of structural parameters, elliptical gal-axies comprise not a single homogenous groupof galaxies but encompass two different groups,discy and boxy ellipticals (see section 1.2.2 fordetails). The shape of these galaxies is veryimportant since it correlates with other physi-cal properties, such as luminosity, shape, rota-tion (axis) and core profile. Recently, the ori-gin of discy and boxy ellipticals was investigated(Naab & Burkert 2003). Equal-mass mergersresult in an anisotropic system with slow ma-jor axis rotation and a large amount of minor-axis rotation (boxy elliptical), whereas unequal-mass merger of mass ratio 3 : 1 and 4 : 1 leadto a rotationally supported system with only asmall rotation along the minor-axis (discy ellipti-cal). Generally, giant high-luminosity ellipticalspreferably contain boxy isophotes, whereas low-luminosity ellipticals comprise a discy structure,which might be a hint for different evolutionarypaths. At intermediate redshift it is impossibleto distinguish between discy and boxy galaxies.However, with respect to the large sample in thisthesis low–luminosity elliptical galaxies can beseparated from high–luminosity ones and possi-ble differences in their evolution can be explored.Results of such a comparison would give conclu-sions if the two types of ellipticals might undergodifferent formation scenarios.


Table 1.2: Overview of samples in this thesis.

Density Locus Tel./Instrument Sample 〈z〉 Nobj Nobj

[N/Mpc2] E+S0 FP

∼100–200 Rich Cluster CA 3.5/MOSCA A 2390 0.23 48 14∼100–200 Rich Cluster WHT/LDSS2 A 2218? 0.18 48 20∼50–100 Poor Cluster CA 3.5/MOSCA Cl 0849/Cl 1701/Cl 1702 0.23 27 10

∼1 Field VLT/FORS FDF/WHDF 0.40 24 21

?From Ziegler et al. (2001a).

1.5 Motivation and Overview

Scaling relations are powerful tools to analysethe formation and evolution of early–type gal-axies. A combination between kinematic andstellar population properties is an even superiorapproach to investigate in great detail the dy-namics and stellar content of these galaxy typesas well as to constrain their predicted evolutionscenarios. With increasing look–back time, the-oretical galaxy formation models suggest a vari-ety of evolutionary paths depending on the en-vironment of a galaxy. To give more insight intothe formation epoch of early–type galaxies, thestudy described in this thesis is based on threehomogenous data sets looking into different en-vironmental galaxy densities. For each sampleseveral criteria have to be fulfilled:

• High signal–to–noise spectroscopy to studythe internal kinematics and stellar popu-lations of early–type galaxies over a largerange of masses.

• Detailed morphological and structural in-formation based on HST/WFPC2 andHST/ACS images for 25% of all galaxies.

• A large coverage of various environments toreveal differences within the samples.

• Wide field–of–view to explore any radial de-pendence of the galaxy properties in clusterscomplemented by a wide range in luminosity

to distinguish between brighter and faintergalaxies.

• Large number of galaxies for each projectat a redshift of z ∼ 0.2, correspondingto a look–back time of ∼3 Gyrs, equiv-alent to 20% the age of the Universe(∼14 Gyrs according to recent achievements(e.g., Spergel et al. 2003).

Using the Calar Alto 3.5 m and the ESO/VLT8.2 m telescopes in Multi Object Spectroscopymode, spectra of 121 early–type galaxies and 28kinematic stellar templates were taken in totalover a period of time between September 1999to October 2002. Table 1.2 gives an overview ofthe samples which are the subject of this the-sis. Besides the density and environment spec-ifications, the instrumentation, median redshiftand the total number of early–type galaxies foreach data set are listed. In the last column, gal-axies within the HST field–of–view are shown,which have kinematic plus structural character-istics and enter the Fundamental Plane relation.Note that only spectroscopic confirmed clustermembers are considered. The Abell 2218 clus-ter analysis was already outlined in Ziegler et al.(2001a), but is also shown as the data will becombined at a later stage with the Abell 2390cluster to study a total of ∼100 early–type gal-axies in the densest environments of rich galaxyclusters. The basic aims of this project were thefollowing:


• Derivation of velocity dispersions and radialvelocities of the early–type galaxies.

• Construction of robust scaling relations atredshift of z ∼ 0.2. Quantifying the evolu-tion in luminosity and comparison to theo-retical evolutionary model predictions.

• To distinguish between the sub–populationsof early–type galaxies, the spherical ellipti-cal and discy S0 galaxies.

• Test of environmental effects on the galaxyproperties.

• Comparison of the observations with modelpredictions for the evolution of single-agestellar populations to constrain the forma-tion redshift of the galaxies and the simula-tions of hierarchical structure formation.

These topics will be addressed in the chapters 4to 7. Prior, chapter 2 describes the construc-tion of the individual samples for the Calar Altoand VLT spectroscopy used in this work. Detailson the telescope and instrumental configurationsand the observation strategies are given.Chapter 3 describes the spectroscopic data re-duction steps for the target galaxies and stel-lar templates. The photometric analysis is pre-sented in chapter 4. This includes the ground–based and HST photometry, analysing the sur-face brightness profile distributions for the gal-axies based on HST imaging and computationof the rest–frame properties, measurements ofstructural parameters, visual and quantitativemorphological classification and derivation of lu-minosities from the ground–based photometry.In chapter 5 the kinematic analysis is discussed.Radial velocities and internal velocity disper-sions are measured for each galaxy. A number ofsimulations of velocity dispersions are performedto ensure a reliable and accurate treatment ofuncertainties on the velocity dispersions at highredshift. Various tests for the kinematics of thegalaxies are carried out and the redshift distri-butions for the galaxy clusters and the field ob-

jects are determined. To extend the rich clustersample by a factor of two, a detailed compari-son of the overall properties of cluster galaxiesin Abell 2390 and Abell 2218 was performed.Galaxy scaling relations are the main topic inchapter 6. After defining an appropriate localreference, the Fundamental Plane and its pro-jections, the Faber–Jackson and Kormendy re-lations will be constructed. It will be demon-strated that there is mass dependent evolutionwith a faster evolution for lower–mass galaxies.Differences between the populations of ellipti-cal and lenticular galaxies will be explored viathe Fundamental Plane and the evolution of themass–to–light ratios.Possible dependences of the galaxy properties onthe environment will be addressed in chapter 7.In this part, the stellar populations of early–typegalaxies will be compared to stellar populationmodels which provide constraints on their evolu-tionary histories. The impact of these results onthe formation scenarios will be discussed.A summary of the thesis is given in chapter 8,which also comprises a brief outlook of the con-tinuation of this project.Unless otherwise stated, the concordance cos-mology for a flat universe will be assumedthroughout this thesis, with Ωm = 0.3, ΩΛ = 0.7and H0 = 70 km s−1 Mpc−1. These values arein compliance with the most recent analysis ofthe power spectrum of the Cosmic MicrowaveBackground observed by the WMAP satel-lite (Wilkinson Microwave Anisotropy Probe,Spergel et al. 2003).

Chapter 2: Sample Selection and Observations 21

Chapter 2

Sample Selection and Observations

Cluster environments provide the opportunity toobserve a larger number of early-type galaxies atthe same redshift with multi-object spectroscopysimultaneously. Therefore, the main goal for thecluster sample was to acquire spectroscopicallya large and representative sample of ∼50 early–type galaxies. At a redshift of z ∼ 0.2, the clus-ter galaxies are bright enough to observe evensub-L∗ systems with 4 m class telescopes, whilestill representing a look-back time of ∼3 Gyrs,adequate to address evolutionary questions. Inparticular, for the rich cluster Abell 2390 (sec-tion 2.1.3) it was proposed to investigate gal-axy sub-populations, like Balmer–line enhancedgalaxies or morphologically classified S0 galax-ies, with sufficient numbers that their individualFPs can be constructed. With a large sampleof early–type galaxies the evolution of the lumi-nosity in the Faber–Jackson relation (FJR) canbe explored accurately by comparing the distantgalaxies directly with the relationship of localsamples like Coma and Virgo (Davies et al. 1987;Lucey et al. 1991). The low number of observedcluster galaxies at higher redshift (z >∼ 0.4) withinstruments like the TWIN spectrograph prohib-ited the measurement of the scatter and possi-ble slope change amongst the clusters (Ziegler& Bender 1997). In targeting ∼50 cluster mem-bers, an order of about one magnitude improve-ment, the evolution within the FJR can be stud-ied for different sub-populations as well as sub-samples, such as low– and high–luminosity orlow– and high–mass galaxies. Furthermore, in

combining the A 2390 sample with a previousstudy of the rich cluster A 2218 at redshift ofz = 0.18 (Ziegler et al. 2001a), the samplewould be large enough to explore possible ra-dial dependences in the luminosity evolution ofthe FJR. Moreover, the investigation of a secondcluster at a similar redshift is essential to notonly understand possible systematic effects aris-ing from instrument characteristics but to inves-tigate the variance of cluster galaxy populationsboth within and between different rich clusters.Both clusters may well serve in the future as suit-able benchmarks for the comparison to rich, highredshift clusters, since aperture corrections areless crucial than, for example, to the Coma clus-ter.

To look for the environmental dependence withinthe cluster, it is desirable to investigate possi-ble radial dependences in relations between theabsorption line-strengths of early–type galaxiesand stellar parameters (age and metallicity). Inparticular, the relationship between the kine-matics and stellar population (e.g., Mgb–σ re-lation) which was found to be very tight for lo-cal galaxies, can be used to probe a large num-ber of early–type galaxies in different clusters athigher redshifts. Furthermore, if, as suggestedby some local studies (Mehlert et al. 2003), el-liptical or S0 galaxies do have young populationsthese should be increasingly apparent at z = 0.2.For a sub-sample of E+S0 galaxies with struc-tural parameters based on HST/WFPC2 imag-ing, the FP could be explored. At the begin-

22 Chapter 2: Sample Selection and Observations

ning of this project, there were only few previousstudies at higher redshift (z ≥ 0.33) and nonewithin 0.05<z<0.33. Therefore, this FP investi-gation at a redshift of z = 0.228 would be veryimportant to cover this interval in redshift space.By observing two clusters at z ∼ 0.2 in detail,changes in the form of the FP with redshift canbe revealed and tested if the change in mass-to-light ratio (M/L) is confirmed. Moreover, it ispossible to look for evidence of a variation in thetilt and scatter of the FP with redshift and en-vironment.

But all these observational approaches applyonly to the special environment of dense richclusters. Therefore, the constraints are biasedtowards the densest peaks with the largest ampli-tudes of the initial density fluctuation spectrum.In general, the semi-analytic Cold Dark Matterparadigm predicts a hierarchical galaxy evolu-tion with a continuous assembling and mergingof galaxies and sub–units. Only in the dense en-vironments of rich clusters galaxies did experi-ence their last major merger at z ∼1–2. The agedistribution and thus the populations of early–type galaxies are however believed to be morediverse in lower density regions. For this reason,a project was proposed to analyse the evolution-ary status of early–type galaxies in poor clusters(section 2.1.4) at a look-back time of ≈3 Gyrswith the same quantitative methods as was ex-tensively done in rich clusters. In particular, theMg–σ relation will reveal whether there is morespread in the age/metallicity distribution thanin the rich clusters Abell 2218 and Abell 2390 atsimilar epochs. Results on the FP will give in-sights whether the mass/structure evolution ismore consistent with the monolithic or the hier-archical formation scenario.

To compare the different cluster samples witheven lower-densities of individual galaxies, twofield samples of early–type galaxies were selectedfrom the FORS Deep Field (section 2.2.1) andthe William Herschel Deep Field (section 2.2.2).Based on deep VLT/FORS spectroscopy comple-

mented with HST/ACS imaging, the kinematicstatus of these field elliptical galaxies will serveas an additional comparison sample to test envi-ronmental effects in detail.

2.1 Cluster Samples

2.1.1 Calar Alto Observatory

The Centro Astronomico Hispano Aleman(CAHA)1 at Calar Alto is located at an alti-tude of 2168 m in the Sierra de Los Filabres,Andalucia, Spain (longitude 02 32′ 48′′ west;latitude 37 13′ 16′′ north). The observatoryis operated jointly by the Max-Planck-Institutfur Astronomie, Heidelberg, and the Institutode Astrofısica de Andalucıa in Granada (CSIC)and provides three telescopes with apertures of1.23 m, 2.2 m and 3.5 m. In addition, an 1.5 mtelescope is operated under the control of theObservatory of Madrid.The CAHA observatory is a powerful counter-part to the ESO observatories at La Silla in tar-geting clusters of galaxies in the northern hemi-sphere. In particular, faint cluster galaxies atz ∼ 0.2 are still within the reach of the 3.5 m tele-scope with reasonable exposure times. However,to measure line-strengths in the observed galaxyspectra as well, observations have to be splittedinto multiple exposures in order to increase thefinal signal–to–noise (S/N) in the spectra. Thisrequires a very stable and reliable combinationbetween telescope and spectrograph instrument,which in the latter case is perfectly fulfilled withthe MOSCA spectrograph.

2.1.2 MOSCA Configuration

The focal reducer MOSCA2, mounted at theRichey-Chretien (RC) Focus of the 3.5-m tele-scope on Calar Alto, is a powerful instrument

1Local page: http:\\www.caha.es\CAHA\index.html.2The MOSCA Multi Object Spectrograph for Calar

Alto instrument had first light in September 1996.(http:\\www.mpia-hd.mpg.de\CAHA\index.html).


because it provides the possibilty of both directimaging and spectroscopy of low– and interme-diate resolution. Furthermore, due to MOSCA’slarge field–of–view (FOV) of 11′×11′, valuableexposure time can be reduced in using a Multi–Object–Mask of up to 25 slits on different objecttargets. The following chapter gives a brief sum-mary of the MOSCA configuration and capabil-ities when used in spectroscopic mode, in par-ticular for executing Multi-Object Spectroscopy(MOS) with user-supplied masks.

The design of MOSCA itself is quite simple.Depending on the observational mode, variousapertures, such as mask, long–slit or multi–slit can be chosen to be mounted in the focalplane. A collimator optic sends a parallel lightbeam which bypasses one or two analysers (filter,grism, Fabry–Perot Instrument (FPI), and/orpolarizer) and is then focussed by a camera op-tics onto the detector. A second filter wheel issituated in the convergent beam. The calibra-tion unit of MOSCA has three different spectrallamps for wavelength calibration (Ne, Ar, andHgAr) and one halogen–continuum source withvariable intensity (40%, 60%, 80%, and 100%).

In order to gain a dispersion function with highprecision for the wavelength calibration, thespectral lamps have a different wavelength sen-sibility. The Ar lamp samples strong lines withlarge intensity at ∼7200 A to ∼8500 A, the Nelamp numerous lines within a range of ∼5500 Ato ∼7000 A, whereas the combined HgAr lampis mainly sensible to the strong lines between∼3500 A to ∼4100 A. A detailed discussion onthe derivation of an accurate dispersion functionfor the wavelength calibration in presented insection 3.6.

Table 2.1 gives a brief overview of the generalproperties of MOSCA. The optical system has areduction ratio of 3.67 with an effective focal ra-tio of f/2.7. This yields an image scale of 3 pixelper arcsec and a total FOV of 11′ × 11′ (effec-tively of ∼ 10′ × 10′) with a pixel size of 15 µm.A thinned CCD (SITE # 16a) with 2048×4096

pixels is used as a detector.Based on laboratory tests, the grism green

1000 gives a wavelength coverage of λλ = 4200–6500 A, with a central wavelength of λ0 = 5300 Aand a dispersion of 5.85 A/arcsec for a typicalMOSCA slit width of 1.5′′.

Table 2.1: Overview of basic MOSCA facts.

Instrument properties:

effective focal ratio: f/2.7reduction ratio: 3.67CCD chip: SITE # 16a : 2048×4096 pixel

Image scale in focal plane:

1 mm = 5.88′′ , 1′′ = 169.7 micronmean magnification factor: 0.27551 pixel (15 micron pixel size) = 0.3208′′

FOV: 660′′× 660′′ = 11′× 11′

The reproduction scale of an optical system isdefined as the ratio between the image size andthe object size. The variation of the reproduc-tion scale along the image area is called distor-tion. This means that an image is deformed. Asquare pattern of straight lines will give an im-age as a pattern of convex curves when there isa barrel-shaped distortion and an image of con-cave curves when there is pincushion or pillow-shaped distortion. Due to the fact that MOSCAis a focal reducer an original image is displayedwith concave lines, i.e. it gets distorted in pillow-shaped form. For this reason, a special correc-tion method was applied to the MOS frames (seesection 3.3 for a further discussion).

2.1.3 Abell 2390 Cluster

With the up-coming of the MOSCA instrumentat CAHA with its multi-object capability andits big field-of-view in 1998, it was feasible andreasonable to carry out the proposed observa-tion of a large number of galaxies in the clusterA 2390. Using the TWIN spectrograph, only a


few brighter galaxies would be possible to ac-quire thereby providing only a limited picture ofthe whole early-type galaxy population of a clus-ter (e.g., Ziegler & Bender 1997). Furthermore,a similar investigation of 48 different early-typecluster galaxies in Abell 2218 at z = 0.175 us-ing LDSS2 at the WHT already proved the ex-tremely high efficiency of the multi-object spec-troscopy technique. At a cluster redshift ofz = 0.23, high S/N spectra of early-type galax-ies covering a wide wavelength range (≈2000A)and spanning a wide range of luminosity (downto MB = M∗ + 1) could still be acquired witha 4 m class telescope and reasonable exposuretimes of ∼8–10 hours. Given the definition ofthe Abell radius of RA = 1.72′

z , the 10′×10′ FOVof MOSCA matches the typical angular size of arich cluster at z ∼ 0.2 quite well. The physi-cal FOV for the A 2390 cluster at z = 0.23 is∼1.4 Abell radii (∼2.7 Mpc). Therefore, combin-ing the high quality spectra for a large fractionof cluster members and addressing a significantlook–back time, detailed tests of the stellar pop-ulation evolution are possible.Besides the MOS spectroscopy with MOSCAat the 3.5 m telescope, the study of the early-type galaxy population in A 2390 is based uponoptical photometry with the 5.1 m Hale tele-scope at Mount Palomar Observatory. In ad-dition, HST/WFPC2 images were exploited toprovide high-quality morphological informationfor a subset of the sample. In section 2.1.6 onpage 30, Table 2.2 gives a summary of the ac-quired observations for the A 2390 project.

2.1.4 Low-LX Clusters

This project is an extensive observational cam-paign to study the evolutionary status of gal-axies in 10 poor clusters in the northern hemi-sphere3 at intermediate redshifts of 0.2 < z < 0.3selected to have very low X-ray luminosities

3The cluster Cl 1444+63 turns out to be two clustersaligned along the line of sight (see section 4.2), and thusthe two clusters are considered separately.

(LX < 5 × 1043 erg/s)4, 100 times lower thanthe rich X–ray bright CNOC clusters, and pooroptical richness class. It comprises ground-basedoptical and NIR photometry, medium-resolutionmulti-object spectroscopy, and HST/F702Wimaging. From the previously analysed low-resolution MOSCA spectra (Balogh et al. 2002b)the location of cluster members with little orno ongoing star formation is known. Therefore,early-type cluster members have been selectedbased on the low-resolution spectra and theirHST morphology (see next section 2.1.5 for de-tails).HST/F702W imaging during Cycle 8 (P.I.Prof. R. L. Davies (Oxford), Proposal ID 8325)was completed in July 2000 for 8 of the 9 clus-ters. The cluster Cl 1633+57 was observed inCycle 7 by another group (see section 4.2.3 forfurther details). With an integration time of7.500 sec, these images are perfectly suited toderive the structural and photometric parame-ters of those galaxy candidates targeted for theFP study. Outside the HST field additional ∼10E+S0 galaxies in each cluster will provide a lookinto the Mg–σ and Faber–Jackson relations indetail. For these scaling relations the number ofobserved galaxies is sufficient to explore possiblevariations between the different poor clusters asa function of their velocity dispersion σ or X-rayluminosity LX .

2.1.5 Selection Criteria

A pre–selection of possible targets for spec-troscopy for all data sets was performed withinthe ESO-MIDAS (Munich Image Data Analy-sis System) environment, whereas the final con-struction of the setups for the field galaxies (sec-tion 2.2) was carried out with the FORS Instru-mental Mask Simulator (FIMS). For construct-ing the MOSCA masks, special MIDAS pro-grams, kindly provided by Prof. J. W. Fried,were utilised. Different catalogues containing

4LX between 0.1–2.4 keV, H0 = 100 km s−1 Mpc−1,see Table 2.3.


the positions, magnitudes, structural parame-ters, morphologies et cetera of the target gal-axies were created within MIDAS and used asinput for setting up of the masks and visualisingthe slits.

The aim of the A 2390 project is to investigatethe stellar populations of a large sample of early-type cluster galaxies spanning a wide range inluminosity. However, in order to enable a goodsky subtraction which requires long slit lengths(with an average length of 22′′), each mask wasconstrained to only about 20 galaxies (with typi-cal galaxy sizes of 4′′) in total. For this reason,we were very careful to select only galaxies whichwere likely to be cluster members based upontheir UBI broad-band colours. Target objectswere selected on the basis of the ground-based I-band images and a combination of defined colourregions. The selection procedure was very sim-ilar to the cluster galaxies in A 2218 (Ziegler etal. 2001a). Using the existing catalogue for thefield (Smail et al. 1998), the distribution of gal-axies on the (U − B) − (B − I) colour-colourplane was investigated, after a correction for bluefield galaxy contamination by ∼28% at I = 20m

with a deep U exposure in a high-latitude blankfield was performed. The colour-colour diagramwas compared to the locus of colours expectedfor non-evolved Spectral Energy Distributions(SEDs) representative for the spectral types ofgalaxies with E/S0, Sab, Sbc, Scd and Sdm mor-phologies in the local Universe, as they wouldbe observed at z = 0.23 (see Smail et al. 1998for further details). A region in UBI colourspace was defined which was most likely occupiedby cluster members. Galaxies falling outside acolour region, defined as 2.60 < (B − I) < 3.50and −0.2 < (U − B) < 0.40, were rejected. Thewidth of this colour range puts negligible restric-tions on the stellar populations of the selectedobjects, but eliminates the majority of back-ground galaxies. Combined with the richness ofthe cluster this approach ensured a high rate ofsuccess in targeting cluster members in our spec-

Figure 2.1: (B − I)–I colour magnitude diagramfrom the Hale imaging for bright galaxies lying in a9.7′×9.7′ (2.12 Mpc) region centered on A 2390. Thesequence of red cluster members is readily seen ex-tending down to I ∼ 21m (MB ∼ −19m). The boxencompasses galaxies matching the selection criteria,while squares denote those galaxies which were even-tually observed spectroscopically.

troscopic sample (4 non–members out of 52 tar-gets). The colour-selected sample of galaxies in-cludes 122 galaxies with 15.53m < Itot < 19.14m

within the 9.7′ × 9.7′ area covered by the Haleimages. Fig. 2.1 illustrates the selection in the(B− I)− I colour-magnitude diagram for galax-ies brighter than I = 23.5 mag within the Halefield–of–view. The box encompasses all galaxiesmatching the selection criteria, while those gal-axies which were actually observed are indicatedas squares. The magnitude limit for the selectionof spectroscopic targets was therefore adopted asItot = 19.14m.A catalogue of the colour-selected galaxy sam-ple was utilised as an input for the mask de-sign preparation. Two masks contained gal-axies with an average I-band surface bright-ness within the spectroscopic slit brighter thanµI = 21.0 mag arcsec−2. In a third mask fainter


galaxies with surface brightnesses as faint asµI = 21.5 mag arcsec−2 were included. In split-ting the galaxies between the individual masksin this way, enabled to vary the cumulative ex-posure times of the masks to ensure that asimilar signal–to–noise of about 30 per A (inthe continuum at λobs = 6300 A) was ob-tained in each galaxy spectrum. Particular carewas taken to maximise the number of galax-ies with early-type morphologies (E–S0–S0/Sa)within the HST field. For this reason, the po-sition angles of the masks were aligned accord-ingly to fulfill this requirement. With a totalof 56 selected galaxies, a sub-population of 25%will have sufficient members to construct an in-dependent FP and Mg–σ relation for differentgalaxy populations. This required number ofgalaxies was fulfilled by selecting galaxies downto V = 20.1m (0.5 mag fainter than L∗) whichshow typical average R-band surface brightness-es of 〈µR〉 ≈ 20.6 mag arcsec−2, assuming an in-tegration over a spatial region of 1.5′′. In orderto study the evolution of early-type galaxies outto large clustercentric distances, target galaxieswere selected covering the whole MOSCA field–of–view of ∼10′ × 10′ (see Fig. 4.1). Each slitwas allocated to a galaxy of interest manually.An average slit length of 22′′ was used (for afew exceptions it was set to a minimum lengthof 15 arcsec) and the position of the slit was alsochosen so that fainter companion galaxies did notobscure the sky and an accurate sky subtractionwas possible.

A first target selection and determination ofworld coordinates of the sample of poor clusterswas performed using the ESO skycat tool. In asecond step, the MOS masks for the MOSCA in-strument were constructed with special MIDASprograms. The central coordinates of the fieldof interest (in α and δ), target positions withprecision < 0.3 arcsec, the slit positions and thelength of slits were computed. The slit widthswere set to 1.5 arcsec. Afterwards these defini-tions were transformed to α and δ coordinates

of the same equinox on the plane sky (includinga rotation of the coordinates according to a givenposition angle) and for a spectral coverage of theused grism and visualised via a graphic versionof the mask. Examples of final MOSCA masksare illustrated in Fig. 2.2.

Finally, individual configuration setups for com-puting a mechanical MOS plate mask forMOSCA were created. These program files weresent to the mechanical shop at the MPIA (Hei-delberg) which manufactured the final MOSCAplates. For the construction of configuration filesfor the mechanical plates special CNC machineprograms were used. Before the observations,these masks are supplied by each observer andput into the aperture unit. A maximum of twomasks are accessible via the GUI interface butone plate can be changed during the night ifnecessary. The orientation of the mask in theMOSCA instrument is unique and thus fixed.

To align the mask and targets, relatively brightstars (V ≈ 19m) were selected and distributedover the whole MOSCA field. Their positionscan be measured precisely even on short expo-sures (∼10 sec). A small but important improve-ment in the programs was the use of three (in-stead of usually at least two) reference stars forthe mask alignment. An additional star warrantsan even higher accurate positioning and align-ment also for lower horizontal distances. Theholes for the stars have a radius of 5 arcsec (seeFigs. 2.2 and 2.3). In taking short exposuresthrough these holes and measuring the positionsof the stars, Cassegrain flange angle and tele-scope position can be corrected which results inan improvement of the mask alignment. Thistricky alignment of the MOS mask is visualisedin Fig. 2.3. For each mask and exposure, the po-sition of the galaxies and alignment stars had tobe verified to avoid loss of galaxy fluxes. Sincethe positions of holes with 10 arcsec diameter cannot be measured precisely enough on images, ad-ditional reference holes with 1 arcsec radius areslightly offset put from the large holes in the


Figure 2.2: MOSCA masks for the poor clusters Cl0849 and Cl1701/Cl1702 seen from below (i.e., theupper surface is the focal plane). The images visualise the whole MOS mask (solid line) with all slits andthe field of MOSCA is indicated as a dotted square. The position angle PA of MOSCA is defined in theusual way (i.e. zero degrees, which is normally used for direct imaging, gives direct images in the standardorientation. The values of Cassegrain flange angle have to be rotated by 90, i.e. Cl0849 has an angle ofPA=0 and Cl1701 has PA=+90. Note that for Cl1701 the slits are in East-West direction.


Figure 2.3: Mask alignment image for Cl0849. Anexposure of ∼90 sec was taken through the slits toverify that all galaxies are accurately positioned ontotheir slits and the stars positioned in their holes. Thelengths for two slits are indicated. The PSF of thisimage is 1.1 ′′ FWHM. The vertical line are six badcolumns located on the CCD chip between 1540–1551pixel.

mask. The positions of these smaller referenceholes can be measured precisely.The basic selection criteria for the poor clus-ters was different to the rich clusters. Thanksto our previous study of the poor clusters basedon MOSCA and LDSS-2 low–resolution spec-troscopy using the grism green 500 and themedium-blue grism (Balogh et al. 2002b), for atotal of 317 galaxies spectroscopic redshifts wereavailable. By using these measurements and thecatalog of 172 Low–LX cluster members, thetarget selection for the follow–up spectroscopycould be performed very efficiently.In order to fill each mask, several cataloguesof early-type candidates were constructed. Onecatalogue contained all objects (spiral and early-type galaxies) with low–resolution spectroscopy,another one held galaxies with absorption linesat cluster redshift (i.e. passive early–type spec-tra) and another list only galaxies with struc-tural information (i.e. residing in the HST field).Furthermore, galaxy catalogues suitable for fill–

up objects (at different redshifts than the clusteror without any spectral information) were cre-ated. All catalogues were loaded into the ESOskycat software simultaneously and marked withdifferent symbols (see Figs 2.4 and 2.5). Forthis reason, the final object selection could beperformed time–efficiently. As the clusters com-prise relatively few members in comparison to arich cluster and only one mask was constructedfor each cluster, particular care was taken to se-lect only cluster members which show early–typespectral features in their low–resolution spectra.Furthermore, the position angles of the maskswere chosen in such a way that the number ofgalaxies with early–type morphologies (E–S0–S0/Sa) selected within the HST field was maxi-mum. For the outermost slits located at themask edges, the coverage of the catalogues waspoor or even null. Therefore, anonymous objectshad to be selected as fill–up objects. In this situ-ation, preference was given to galaxies which fea-tured similar apparent luminosities as the clustergalaxies. However, for all setups the number ofanonymous galaxies was not larger than three.

2.1.6 Observations

The observations were carried out during twoobserving runs in four nights of September 7.–11., 1999 and five nights of July 26.–31., 2000.To achieve intermediate-resolution spectra forthe galaxies in A 2390 at z = 0.23, the grismgreen 1000 with a typical wavelength rangeof λλ = 4500–7500 A was chosen, encompass-ing important absorption lines such as Hγ, Hβ,Mgb, Fe5270 and Fe5335 at the cluster redshiftof z = 0.228. The slit widths were set to1.5′′ and the instrumental resolution around Hβand Mgb (5900 <∼ λobs <∼ 6400 A) was 5.0 AFWHM, corresponding to σinst∼ 100 km s−1

around λobs = 6360 A (see section 5.3 fora discussion). The spatial scale was 0.33′′

per pixel and seeing conditions varied between1.2′′ ≤ FWHM ≤ 1.7′′.Three masks were observed with total exposure


Figure 2.4: Hale R-band image of CL 0849(z=0.2343) used for the mask preparation. The imagesize is 8.′4×7.′0. Galaxies with low–resolution spectraare indicated as diamonds. Possible spiral galaxies orobjects without any absorption lines and/or differentredshifts are marked as squares. Early–type clustercandidates are denoted as circles. Stars used for themask alignment are also shown (# 111, # 333 and# 444). The bar corresponds to 300 kpc in projectionat the clusters redshift of z = 0.2343 and the smallbar to 10′′ (38.1 kpc).

times between 29880 sec (8.3 hrs) and 42480 sec(11.8 hrs) each (see Table 2.2). In total, 63 high-signal-to noise spectra of 55 different galaxieswere obtained, of which 15 are situated withinthe HST field. Three objects, # 1507, # 1639and # 2933, turned out to be background galax-ies at z = 0.3275, z = 0.3249 and z = 0.3981,respectively; one galaxy (# 3038) is located inthe foreground at z = 0.1798. Four other gal-axies fell into the slits by coincidence (# 2222,# 5552, # 5553 and # 6666) which are all clus-ter members. However, apart from # 6666, theirspectra are too faint (S/N <∼ 8) to determineaccurate velocity dispersions and no photome-try is available for the other objects (cf. sec-tion 5.5). Therefore, these galaxies were also dis-carded from the sample yielding a total of 48 dif-ferent early-type cluster members, proving that

Figure 2.5: INT R-band image of CL 1701(z=0.2458) and CL 1702 (z=0.2233) used for themask preparation. The image size is 11.′11 × 10.′82.Galaxies with low–resolution spectra are indicated asdiamonds. Possible spiral galaxies or objects with-out any absorption lines and/or different redshifts aremarked as squares. Early–type cluster candidates aredenoted as circles. Stars used for the mask align-ment are also shown (# 333, # 444 and # 555). Thebar corresponds to 300 kpc in projection at the clus-ters redshift of z = 0.2458 and the small bar to 10′′

(38.9 kpc).

our sample selection was highly efficient.For the Low–LX clusters, spectroscopic MOSCAobservations were carried out during six nights inApril 17.–22., 2001 and six nights during Febru-ary 4.–9., 2002. A summary of the cluster samplewith the mean redshifts, number of objects on amask and total exposure times for spectroscopyis given in Table 2.3. In total, five masks (oneper cluster) were observed. The only exceptionis the mask for the clusters with α=17h (Cl 1701and Cl 1702), which were acquired with the samemask due to resons of telescope time–efficiency.For the study within the FP, spectra of 5−7 gal-axies lying within the HST fields in a single MOSmask were obtained. Based on the whole clustersample, a combined FP sample of ∼30 early–


Table 2.2: Observations of Abell 2390

Tel. Instrument Date Band (phot) Texp

Mask / Nspec [ksec]

HALE COSMIC 09–12/06/94 U 3.00COSMIC 09–12/06/94 B 0.50COSMIC 09–12/06/94 I 0.50

HST WFPC2 10/12/94 F555W 8.40WFPC2 10/12/94 F814W 10.50

CA 3.5 MOSCA 07–10/09/99 1 / 17 29.88MOSCA 07–10/09/99 2 / 22 42.12MOSCA 26–28/07/00 3 / 24 42.48

type galaxies can be constructed, which is com-parable to that of a single rich cluster. Outsidethe HST field additional spectra of ∼10 E+S0galaxies in each cluster serve for the investiga-tion of the Mg–σ and Faber–Jackson relation. Intotal, 76 high S/N galaxy spectra of poor clustercandidates were acquired.

In this thesis the analysis concentrates on theobservations from 2001 of the three clustersCl 0849+37 (z=0.2343), Cl 1701+64 (z=0.2458)and Cl 1702+64 (z=0.2233). Note, that Cl 1701and Cl 1702 were only observed with one maskand the number of galaxies and the observingtime is not splitted for into each cluster in Ta-ble 2.3.

During the observations at Calar Alto, thealignment of the MOSCA masks was conductedwith the special program alimask. First, theCassegrain flange was changed to the requiredposition angle and the telescope moved to thecenter coordinates of the mask. Second, thealignment process with alimask was initiated.Usually, this procedure has four steps whichhave to be performed consecutively (1) exposureof the mask with the installed calibration lampto measure the positions of the reference holes,(2) short exposure of the whole field to measurethe alignment stars for improvement of telescopeposition and Cassegrain flange angle, (3) two

short exposures to measure the positions of thestars in two large holes in the mask to computeflange angle and telescope position corrections.This step may be iterated. (4) Exposure throughthe slits to verify that all objects are located intheir slits. This step is optional (see Fig. 2.3).Together with the previous settings the wholeprocedure takes about 25 minutes. However,as two observers (A. Fritz & K. Jager in 2001,A. Fritz & B. Ziegler in 2002) could worksimultanously (one with the telescope settingsand one with the mask alignment) and due toan optimised scheduling (e.g., skipping step 2when executed once for a night, changing cutsfor image display) the alignment process couldbe shortened to ≈15 minutes. In particular, theverification of the mask alignment (step 4) wasdone after starting the 2700 sec science exposureand thus valuable minutes of dark time couldbe gained. This was a non-negligible factor asa high amount of dark time was lost due tobad weather (60% in 2001, 50% in 2002) anddue to a forest fire (sixth night of 2001) and anearthquake (first night of 2002).


Table 2.3: Observations of six poor clusters

Cluster R.A. DEC. < z > σc LX1 Date Nobj Texp

(J2000) [km/s] mask [ksec]

Cl 0818+56 08 19 04 +56 54 49 0.2670 651±165 1.50 04–09/02/02 14 29.40Cl 0841+70 08 41 44 +70 46 53 0.2397 399±170 1.22 04–09/02/02 13 24.54Cl 0849+37 08 49 11 +37 31 09 0.2343 764±090 1.93 17–22/04/01 17 18.00Cl 1633+57 16 33 42 +57 14 12 0.2402 582±360 0.49 04–09/02/02 14 30.97Cl 1701+64 17 01 47 +64 20 57 0.2458 834±647 0.40 17–22/04/01 18 27.90Cl 1702+64 17 02 14 +64 19 53 0.2233 386±426 0.74 17–22/04/01

1 LX (0.1–2.4 keV) 1043h−2100 erg s−1.

2.2 Field Sample

2.2.1 The FORS Deep Field

The FORS Deep Field (FDF) is a very deep,multi–band photometric study of a sky region lo-cated near the south galactic pole. The primaryaim of the FDF project was to perform a multi–colour imaging investigation in the optical andnear-infrared with visible limiting magnitudescomparable to the Hubble Deep Fields (HDFs,Williams et al. 1996, 2000) but with a substan-tially larger field-of-view of ∼ 7′ × 7′ (about 4times the combined HDFs). Its main scientificdriver was to improve the understanding of theformation and evolution of galaxies in the youngUniverse using a combination of deep photom-etry and extensive follow-up spectroscopy (seeAppenzeller et al. 2000 for an overview).In the selection of a suitable field, the sky areahad to fulfill several selection criteria. The cut-off limit on the galactic extinction was chosen tobe E(B − V ) < 0.02m and large, nearby galax-ies, bright stars (V < 18m) within the FOV andvery bright stars (V < 5m) within 5 should beabsent in the field in order to allow reasonablylong exposure times and to avoid ghost imagesetc. The two latter constraints ruled out the op-tion to use the HDF-S region. Moreover, the fieldhad to be devoid of strong radio or X-ray sources

to avoid potentially present galaxy clusters atmedium redshifts. The field which was finallychosen is located near the south galactic polewith center coordinates of α2000=01h 06m 03.6s,δ2000=−25 45′ 46′′.The optical observations were carried out withthe Focal Reducer / Low Dispersion Spectro-graph (FORS, see Appenzeller et al. 1998or www.eso.org/instruments/fors1/ for anoverview) which was designed and constructedin a collaboration of the Universitats-SternwarteGottingen, the Landessternwarte Heidelberg andthe Universitatssternwarte Munchen. Since1999, two versions of the FORS instrument(called FORS1 and FORS2) are available,mounted on two different units of the Very LargeTelescope (VLT) which is operated by the Eu-ropean Southern Observatory (ESO) at CerroParanal, Chile. In return for the instrumen-tal contributions, the three institutes above weregiven a pool of guaranteed observing time (so–called Guaranteed Time Observations, GTOs)by ESO. A high fraction of this GTO was re-served for the FDF imaging in the optical (≈10GTO nights), the rest of the allocated GTOwas shared between several follow–up studies ofthe FDF project (e.g., the Tully–Fisher relationproject, see Ziegler et al. 2002, Bohm et al. 2004;or a study of high–redshift galaxies, see Mehlertet al. 2002; Noll et al. 2004). In the next sec-


tions 2.2.3 and 2.2.4, the instrumental setup andthe target selection procedures for the field ellip-tical study are outlined.

2.2.2 William Herschel Deep Field

The William Herschel Deep Field (WHDF, Met-calfe et al. 2001) is a deep optical and near-infrared UBRIHK photometric survey per-formed over a 7× 7 arcmin square sky area withthe William Herschel Telescope (WHT) at Ca-nary Islands, and the United Kingdom InfraredTelescope (UKIRT) at Mauna Kea and the 3.5 mtelescope on Calar Alto. The WHDF is locatedin the northern hemisphere with center coordi-nates of α2000=00h 22m 33.3s, δ2000=+00 20′ 57′′

(centre of B-band image). Compared to theFDF, the WHDF is not as deep and has muchworse seeing conditions varying between 1.2′′ ≤FWHM ≤ 1.5′′ (mean seeing 1.3′′ FWHM). Themain scientific driver for our study was to per-form spectroscopy within the WHDF to enlargethe FDF sub-samples of field spiral and early-type galaxies by a factor of two in order to gainlarge data set with statistical significance.The WHDF sample of elliptical galaxy candi-dates were drawn from the deep UBRIHK pho-tometry of the WHDF. UBRI imaging is basedon the WHT, the infrared K and Kdeep (only1.8×1.8 arcmin2) imaging was obtained with theUKIRT. After our observations in 2002, theWHDF was observed in the H-band (whole field)using the Ω Prime camera at the 3.5 m telescopeon CAHA. The 3σ magnitude limits for pointsources as derived from images are 26.8, 27.9,26.3, 25.6, 22.8, 20.5, 22.8 in the U, B, R, I,H, K, Kdeep filters (2 arcmin diameter sub-area),respectively. All magnitudes are in the Vega-system.

2.2.3 FORS Configuration

The spectroscopic observations of the sample ofFDF field ellipticals were performed simultane-ously with those of the late–type galaxies in

the FDF which were subject to the investiga-tion of the evolution of the Tully–Fisher rela-tion (Ziegler et al. 2002; Bohm et al. 2004).A full description of the instrumental setup isgiven in Bohm (2003). Here, only the setupfor the investigation of the field elliptical gal-axies is described. In addition, a second sam-ple of field early-type galaxies selected from theWHDF was observed together with a projectstudying the Tully–Fisher relation of late–typegalaxies (Bohm & Ziegler 2006) under similarconditions with the VLT at Cerro Paranal. Asthe sample selections and instrument configura-tions were very similar the two individual datasets are discussed in combination.

In order to construct a large spectroscopic sam-ple with less amount of observing time, the MultiObject Spectroscopy (MOS) Mode was chosento be the best option. The Mask eXchangeableUnit (MXU) mode of FORS2 was non existentand thus the FDF observations in 2001 were re-stricted to the usage of FORS1. For an effectivefilling of the MOS masks, the early-type galax-ies were observed simultaneously with the spiralgalaxies. Both versions of FORS instruments of-fer 19 slitlets in the MOS configuration, 9 withslit lengths of 22 arcsec, 8 with slit lengths of20 arcsec, whereas the uppermost and lowermostslits have lengths of ∼ 11 arcsec.

FORS was operated at standard resolutionsetup, i.e. for the configuration of the CCDs(read out in one–port mode) and the collimatorthe default values of the MOS mode were used.The spatial scale was 0.2 arcsec/pixel and in lowgain mode (using port A) FORS1 offers a gainof 3.51 e−/ADU and a read-out-noise (RON) ofRON=7.21 e−. The grism 600R was an optimalchoice to achieve a medium resolution within theresulting wavelength range appropriate for spi-ral galaxies at z ≤ 1.0 that covers the spectralranges of the [O II] 3727 doublet or the Hβ andthe [O III] 5007 emission lines. The default orderseparation filter GG435+81 was used. A slit po-sitioned in the center of the CCD chip covers a


wavelength range between 5200 A ≤ λ ≤ 7400 A.However, as the starting and the ending wave-length depend on the position of the slit on the Xaxis, the wavelength range can vary between twoextremes of the MOS mask 6300 A≤ λ ≤ 8500 A(extreme left hand side) or 4100 A≤ λ ≤ 6300 A(extreme right hand side).Each MOS slit can be set individually to slitwidths between 0.3 arcsec and 60 arcsec. Forall observations a fixed value of one arcsec-ond was chosen, yielding a spectral resolutionof R≈ 1200 with the 600R grism. Although asmaller slit width would have increased the in-strumental resolution, a significant loss of fluxwould have occurred for seeing conditions aboveFWHM≈ 0.8 arcsec, which roughly correspondsto the median seeing at Paranal since 1998 ac-cording to the Differential Image Motion Moni-tor (DIMM).Between October and November 2001 an up-grade of the FORS2 CCD system was installedat UT4 (Yepun) for testing purposes, which wasavailable for science use in March 20025. Thenew FORS2 upgrade detector system consistsof two 2048×4096 pixel MIT/LL CCID-20 CCDchips (15µm pixel size). The MIT CCDs pro-vide much higher response in the red wavelengthrange (>8000A), with impressively low fringeamplitudes. At default (standard) setup, thenew chip has a spatial scale of 0.25 arcsec/pixelwith a gain of 0.70 e−/ADU (in high gain100 kHz mode). The RON differs slightly be-tween the chips, for the “master” chip (“Thor”)RON=2.7 e− and the “slave” chip (“Belenos”)RON=3.0 e−. Moreover, the volume phasedholographic grism 600RI for FORS2 turned outto be even more efficient than the grism 600R,particularly at redder wavelengths. Again, thedefault order separation filter GG435+81 wasused.In order to get an S/N ∼ 16 over an area of1′′×0.25′′ at λc ≈ 6552 A in the continuum

5FORS2 upgrade: www.eso.org/projects/odt/-

Fors2/Fors2u.html

for a typical early-type galaxy with brightnessR ≤ 20.5m, the on–line ESO Exposure Time Cal-culator (www.eso.org/observing/etc/), sug-gested an integration time of ∼ 2.5 hrs with theinstrumental configuration as given above.

2.2.4 Selection Criteria

After a pre–selection of possible spectroscopicfield target objects, different galaxy catalogueswere generated within the ESO-MIDAS environ-ment and used as input for setting up and con-structing the MOS masks with the FIMS tool.The FIMS package output consisted of three filesper mask. For example, these give definitions onthe position, angle and orientation of the maskon the plane of the sky, the positions of the indi-vidual slits, coordinates of reference stars in thefield–of–view (for positioning of the mask duringthe observations with an accuracy < 0.1 arcsec),et cetera. In turn, these output files were usedfor the construction of the so–called ObservingBlocks (OBs) which act as input to the operat-ing system of the VLT telescope in the execu-tion of the observations at Paranal for the fieldsample. For a more extensive description of theoperation environment, the reader is referred tothe “FORS1+2 User’s Manual” released by ESO(www.eso.org/instruments/fors/doc/).The object selection for the FDF sub-sample wasbased on the deep UBgRI photometry of theFDF (Heidt et al. 2003). The 50% complete-ness limits for point sources as derived from theco-added images are 25.64, 27.69, 26.86, 26.68,26.37 in the U, B, g, R, I filters (Vega-system),respectively. The candidates were selected ac-cording to their spectrophotometric type, theirelongated structureless appearance, their lumi-nosity and photometric redshift. Main criterionwas the apparent brightness. The other param-eters were taken into account to put additionalconstraints on the selection. For the early-typetargets, this limit in total apparent magnitudewas R ≤ 22.0m as derived with the Mag auto al-gorithm of the Source Extractor package (Bertin


& Arnouts 1996). This constraint was set to en-sure sufficient signal-to-noise of S/N≥10 in theabsorption lines which is mandatory for a ro-bust determination of velocity dispersions andline strengths measurements. For this reason,faint early-type candidates with apparent mag-nitudes R > 20.5m were included in more thanone MOS setup and the individual spectra com-bined after data reduction.

Spectrophotometric types and estimated red-shifts were selected from the FDF photometricredshifts catalogue of more than 3800 objects inthe August 2000 release (Bender et al. 2001).Only respective candidates with an early (E/S0)model Spectral Energy Distribution (SED) wereconsidered and the photometric redshifts wererestricted to zp ≤ 0.6. Based on the photometricredshifts, the elliptical candidates were spreadout on different CCD positions among the spi-rals for each MOS setup in the observed spec-tral wavelength range such that either the Mgb-feature passband (λ0 ≈ 5170 A) or the G-band(λ0 ≈ 4300 A) was included. Both features ofinterest could not be observed for all candidatesbecause of the correlation of the slit position withthe available wavelength window (see also sec-tion 2.2.3 before). Moreover, for an additionalstar/galaxy separation criterion, targets with in-significant photometric redshift and large uncer-tainty (i.e., zphot − dzphot ≤ 0) were rejected.

In contrast to spiral galaxies which require aselection upon inclinations and position anglesfor the derivation of rotational velocities, forthe MOS targets with an early–type SED, nei-ther inclinations nor position angles have beentaken into account as selection criteria. How-ever, as a small fraction of early–type galax-ies in the photometric redshift catalogues couldbe misclassified M–dwarf stars, it was manda-tory to use structural parameters for a ro-bust star/galaxy separation. Detections of theSource Extractor on the I-band image with 0.49′′

FWHM which were most likely bonafide stellarobjects exhibited a star classification parameter

star≥ 0.9 and a≈ b≈ 2.5 pixel. Therefore, toavoid misclassifications targets with star≥ 0.9or a≈ b≤ 3 pixel have been rejected.

Although the two-dimensional distribution ofFDF objects indicated a possible cluster of gal-axies with zp ≈ 0.33 and the primary goal wasto target field elliptical galaxies, no preselec-tion against such candidates was performed. Inthe spectral analysis (see section 5.5), it wasconfirmed that 15 out of 32 observed ellipticalgalaxy candidates are most probably membersof a cluster. Based on the the radial veloc-ity measurements for these galaxies, the lowerlimit for the velocity dispersion of the clusteris σc >∼ 430 km s−1. Since the cluster centre isnot located on the FDF but only its outskirts,the true σc is probably larger. At a redshift ofz = 0.33, the Mg 5170 absorption line is unfor-tunately corrupted due to the terrestrial absorp-tion of the B band. As an accurate measurementof the internal galaxy velocity dispersions is im-possible due to this effect these cluster galaxieswere discarded from the analysis.

Apart from photometric redshifts, the target se-lection for the WHDF field early-type galaxieswas performed with the same constraints as forthe FDF field elliptical galaxies. As no pho-tometric redshifts were available, the redshiftsof field early-type candidates were estimatedusing a combination of colour-colour diagramsand apparent magnitudes. Fig. 2.6 illustratesthe target selection of the WHDF field ellipti-cal candidates in the (B − R)–(R − I) colour-colour diagram. The WHDF data and the finalWHDF field ellipticals are compared to evolu-tionary tracks for E/S0 with different formationredshifts of zf = 2 (solid line) and zf = 4 (dot-ted line) as predicted by the passive evolutionmodels of Bruzual & Charlot 1993; GISSEL96version, hereafter BC96). Note, that the BC96models have been computed for a slightly differ-ent cosmology with H0 = 60 km s−1 Mpc−1 andq0 = 0.1. In the (B−R)–(R−I) colour-colour di-agram elliptical and spiral galaxies are well sep-


Figure 2.6: (B − R)–(R − I) colour-colour diagramfor the WHDF field elliptical galaxies. The WHDFspiral candidates (squares) are displayed with the fi-nal WHDF field ellipticals (circles) and compared toevolutionary tracks for E/S0 with different formationredshifts of zf = 2 (solid line) and zf = 4 (dottedline) predicted by the BC96 models. For the BC96models a cosmology with H0 = 60 km s−1 Mpc−1 andq0 = 0.1 was adopted.

arated until z ≈ 0.5 (indicated by the arrows)and follow the predictions of the evolutionarymodels quite well. The apparent R magnitudewas used as an additional constraint to avoid aselection of high redshift elliptical galaxies withz > 1. As shown in Fig. 2.6, galaxies populatinga narrow range in the colour-colour diagram of2 >∼ (B −R) >∼ 2.9 and 0.7 >∼ (R− I) >∼ 1.1 havebeen selected as bona fide field ellipticals andsuccessfully verified. One object (ID # 14) with(R− I) = 0.79, (B −R) = 1.78 and R = 19.04m

(z = 0.1060) is an intermediate-type Sc spiralgalaxy and was therefore discarded as an early-type candidate. In fact, on the ACS images itturned out that this galaxy is a spiral with aclearly visible disc.

2.2.5 Observations

To construct a distant field elliptical galaxy sam-ple with Nobj ≈ 30 within the FDF, roughly 30 %of the reserved observing time for the Tully–Fisher project were used. However, as half ofthese galaxies turned out to be possible membersof a cluster at z = 0.33, an additional sampleof ten elliptical galaxies was selected from theWHDF. The FDF and WHDF target selectionfor the observations was done by A. Bohm andA. Fritz. An outline of the observations of thefield early-type galaxies is given in Table 2.4 forthe FDF sample and in Table 2.5 for the WHDFgalaxies.The first FDF spectroscopic observations werecarried out with FORS1 in December 1999 inVisitor Mode by C. Mollenhoff (Heidelberg) andK. Reinsch (Gottingen) with a mean DIMMseeing of 0.66 arcsec. However, only one MOSsetup was observed and two major drawbacksoccurred. At this time the FDF imaging data re-duction was not fully completed yet and the tar-get selection was therefore based on a coadded1200 sec R-band image (Point-Spread-Function(PSF) of 0.72′′ FWHM) gained during commis-sioning time of FORS1. Furthermore, for theselection only a preliminary version of the pho-tometric catalog was available. In total, 3 early-type galaxies, 13 spiral galaxies and 3 objectswithout any catalog information (without SEDand zphot) have been selected. The galaxies clas-sified as “anonymous” turned out to be early-type galaxies with R-band magnitudes of 19.1m,19.4m and 20.8m. Unfortunately, after the red-shift analysis 3 early-type galaxies were identi-fied to be members of a cluster at z = 0.33.Therefore, these galaxies were rejected for thisinvestigation.In September and October 2000 a total of 6 MOSsetups were observed by S. Noll and D. Mehlert(both Heidelberg) with FORS2. Each setup wassplitted into three single exposures of 3000 secwith a total integration time of 2.5 hours. Thetargets were selected from a deep FORS2 I-


Table 2.4: Observations of FDF early-type galaxies

Tel./Instrument Date θ range airm. DIMM E+S0 anon. Texp

[ksec]

VLT/FORS1 12/99 −30/ 0 1.08 0.66′′ 3 3 9.0∑6 -

VLT/FORS2 09/27/00 −90/−60 1.21 0.51′′ 5 1 9.009/27/00 −60/−30 1.33 0.43′′ 5 - 9.010/04/00 −30/ 0 1.40 0.81′′ 6 1 9.010/05/00 0/+30 1.36 0.80′′ 5 2 9.010/05/00 +30/+60 1.28 0.74′′ 8 - 9.010/06/00 +60/+90 1.15 0.66′′ 7 2 9.0∑

28 6

VLT/FORS1 10/12/01 0/+30 1.43 0.76′′ 6 1 9.010/14/01 +30/+60 1.07 0.89′′ 5 0 9.010/12/01 +60/+90 1.38 0.82′′ 5 3 9.0∑

10 4

band reference image consisting of 10 seeing av-eraged I-band images (between 0.47–0.50 arcsecFWHM) with a final integration time of 3000 secand a PSF of 0.49 arcsec FWHM (see Bohm 2003for the construction of the FORS1/FORS2 ref-erence images). All in all 28 E+S0 galaxies wereobserved from a list of 32 candidates with avail-able spectrophotometric information (see previ-ous section). Eight faint early-type candidateswith apparent magnitudes R > 20.5m were ob-served with two MOS setups to increase the re-spective exposure time and the S/N in the finalspectra. Two galaxies were already observed in1999. Typically, each setup contained 2–4 el-liptical candidates. Using slit widths of 1′′, withthe 600R grism a spectral resolution of R = 1160was achieved. The seeing conditions (DIMM see-ing values) were varying between 0.43′′ and 0.81′′

FWHM (median of 0.66′′ FWHM) and sufficientto meet the Nyquist theorem to allow a per-fect reconstruction of the signal from the sam-ples (see section 5.1.1). The constraint on theairmass was A ≤ 2.0 in order to limit the correc-tions of atmospheric absorption.

Early in 2001, an additional amount of GTOtime was granted by ESO for the FDF collabora-tion to compensate the high amount of time lossduring the imaging phases of the FDF which wascaused by the El Nino phenomenon. From thisGTO pool, the P.I. of the FDF collaboration,Prof. I. Appenzeller (Heidelberg), thankworthyallocated one night of dark time for the field gal-axy project of spiral and elliptical galaxies.

In October 2001 additional three different setupswere acquired with with FORS1 at the VLT inVisitor Mode by J. Heidt (Heidelberg). The tar-get selection was based on a FORS1 I-band ref-erence image with a PSF of 0.52 arcsec FWHM.In the search of the photometric redshifts cat-alogues for new field elliptical candidates withR ≤ 22m and zphot < 0.6 only two objectswere detected which had not been observed pre-viously. Again the setups were splitted into threeseparate exposures (Ttot = 9000 sec) with an in-tegration time of 2.5 hours for each setup. 15out of 29 galaxies located in the southwesterncorner of the FDF were neglected to avoid theoutskirts of the cluster at z = 0.33 (see chap-


Table 2.5: Observations of WHDF early-type galaxies

Tel./Instrument Date θ range airm. DIMM E+S0 anon. Texp

[ksec]

VLT/FORS2 07/10, 10/03/02 −90/−60 1.18 0.64′′ 1 1 9.008/07, 09/11/02 −60/−30 1.19 0.73′′ 2 - 9.008/04, 08/07/02 −30/ 0 1.51 0.92′′ 3 - 9.008/04, 10/02/02 0/+30 1.13 0.89′′ 3 1 9.009/10, 09/12/02 +30/+60 1.18 0.91′′ 2 - 9.009/12, 10/4-5/02 +60/+90 1.14 0.85′′ 2 - 9.0∑

11 1

ters 5.5 for a further discussion of this topic). Inaddition, eight bonafide field E+S0 galaxies withR > 20.5m which already were observed in 2000were selected to increase the S/N in the com-bined spectra. A total of 10 early-type galaxieswas targeted and thus except one bright ellipti-cal with R = 19.0m, all early–type candidateswere observed two or three times including thespectroscopy in 1999 and 2000. The seeing con-ditions were slightly poorer than during autumn2000, but still in the sub–arcsecond regime. Themedian DIMM seeing was 0.82′′ FWHM.

Between July and October 2002 spectroscopicobservations for 10 WHDF ellipticals were car-ried out with the FORS2 instrument attachedto the ESO/VLT at Cerro Paranal, Chile. Thisobserving time was not part of any GTO pro-gramme but additionally granted. A total ofsix different setups were observed, each splittedinto three separate exposures of 3000 sec witha total integration time of 2.5 hours. Targetswere selected through a combination of colour-colour diagrams and apparent magnitudes (seesection 2.2.4). For the MOS mask constructionwith FIMS, a deep WHDF I-band reference im-age consisting of 5 seeing averaged I-band im-ages with a final integration time of 1500 secand a PSF of 0.6 arcsec FWHM was used. Theslit width was set to 1′′, and with the 600RIgrism a spectral resolution of R = 1000 wasachieved. The DIMM seeing values were be-

tween 0.73′′ and 0.92′′ FWHM with a medianof 0.89′′ FWHM. Two additional objects, one inthe setup 0 ≤ θ ≤ +30 and one in the setup−90 ≤ θ ≤ −60, fell by coincidence into theslit of the galaxy # 810 and # 92, respectively.The second object in slit # 810 turned out to bea field S0 galaxy at z = 0.2118 and was named# 810b. The other object in slit # 92 is a back-ground spiral galaxy at z = 0.5569 and thuswas discarded. Therefore, the total sample ofWHDF elliptical galaxies comprises eleven fieldearly-type galaxies.

Chapter 3: Data Reduction 39

Chapter 3

Data Reduction

This thesis comprises a large amount of datafrom three different projects, the rich Abell 2390cluster, the three Low–LX clusters and two fieldgalaxy samples from the FDF and WHDF, allaiming to investigate the population of early–type galaxies in various environments. The re-duction of the optical ground-based multi–bandimaging data for the two cluster samples, fromwhich the apparent magnitudes of the early-typegalaxies were derived, was performed mainly byM. L. Balogh and I. R. Smail. Nevertheless,some aspects for this study had to be re–analysed(see chapter 4). For the FDF and WHDF fieldelliptical galaxies A. Bohm was kindly providingthe measurement of ground-based magnitudes.As a full review of the imaging reduction is be-yond the topic of this thesis, it therefore will notbe included. A summary of the observations andimaging for Abell 2390 is provided in the Ta-ble 2.2 on page 30 and for the three poor clustersin Table 2.3 (page 31) and Table 4.2 (page 56).All the subsequent sections will focus on thespecial aspects of the MOSCA spectra reduc-tion. As the reduction of the FDF and WHDFVLT/FORS spectra was performed in an analo-gous manner it thus will not be described hereseparately. The reduction procedure was under-taken using the eso-midas

1 environment withown fortran program routines and the iraf

2

1ESO–MIDAS, the European Southern ObservatoryMunich Image Data Analysis System is developed andmaintained by the European Southern Observatory.

2IRAF (Image Reduction and Analysis Facility) is dis-

data analysis software and followed the standardprocedure.The ESO-MIDAS package provides a variety ofalgorithms for data processing for applications oflong–slit reduction, such as wavelength calibra-tion and rebinning of two–dimensional spectra.Nevertheless, special routines had to be devel-oped, for example, to rectify the distortions ofthe spectra due to the focal reducer (see section3.3).A theoretical review of data reduction willnot be given here, see, e.g., Wagner (1992)or Volume B of the ESO-MIDAS User’sGuide, Chapter 6 (available for download atwww.eso.org/projects/esomidas/doc/ ). Fora discussion of the derivation of HST structuralparameters, see section 4.3). In the following,the individual stages of the reduction procedurewill be described in the same order as they wereapplied to the data.

3.1 Bias Correction

For each observing run, bias frames were takenat the beginning and end of the night. Allframes showed a very stable two–dimensionalstructure with spatial variations of ≤3 ADU.Thus, all bias frames from the individual nights

tributed by the National Optical Astronomy Observato-ries (NOAO) in Tucson, Arizona, which is operated bythe Association of Universities for Research in Astronomy(AURA), Inc., under cooperative agreement with the Na-tional Science Foundation.

40 Chapter 3: Data Reduction

were used to generate master biases, one foreach observing run. An averaged super biasimage was constructed of median–scaled indi-vidual bias frames. After calculating the me-dian in the overscan region of the super bias(X>2022 pixel), which is ∼271 ADU, the superbias frame was scaled to the median of each in-dividual bias image. A total of six bad columnswere cleaned by interpolating from non affectedadjacent columns. Finally, this scaled mas-ter bias frame was subtracted from each frame.Variations in the bias level were ≤5%.Several dark exposures were gained at the begin-ning and the end of one night for each observingrun. Since the dark current was very uniformacross the CCD chips and amounted to less then5 ADU per 45 minutes, an additional dark framesubtraction has been neglected.

3.2 Cosmic Rays

Although the observation of each MOS masksetup was split into at least five individual expo-sures, a cosmic ray rejection by a simply median–averaging of the frames was not adequate be-cause the CCD positions of the individual gal-axy spectra were not time–independent. If dur-ing one observation of a mask the zenithal dis-tance was significantly varying, the spatial po-sition of a spectrum differed by up to two pixel(0.66 arcsec) between the first and last exposure.Therefore, 2D images of each slit spectrum wereextracted individually from the MOS frame (af-ter bias subtraction) and reduced separately aslong–slit data. In total, for each sample a num-ber of Nexp. × Nslits science spectra per MOSsetup and Nmask×NMOS science spectra were re-duced, in numbers 735 science spectra for A 2390(3 masks) and 264 science spectra for the Low–LX clusters (one mask each). Table 3.1 gives asummary of the analysed science spectra.Cosmic ray events were removed very carefulusing the MIDAS algorithm FILTER/COSMIC toeach individual slit spectrum. The parameters of

Table 3.1: Summary of MOSCA science spectra.

Sample Exp. Slits/Obj. MOS∑

A 2390 11 16/17 17613 20/22 26013 23/24 299 735

Cl 0849 6 17/17 102Cl 1701/Cl 1702 9 18/18 162 264

this κ–σ clipping which tune the discriminationbetween physical objects and artificial artefactswere set to moderate values in order not to af-fect possible emission features from the galaxies.In addition, after the filtering process each slitframe and the map of rejected pixels were visu-ally inspected and compared to each other. Asecond cosmic ray inspection was done at thestage of the sky subtraction. However, this finalcorrection was performed by a purely manual re-moval of any contamined pixels, if artificial fea-tures were identified in a visual investigation ofthe completely reduced 1D spectra.Furthermore, between one and up to five badcolumns per slitlet were cleaned by interpolatingfrom unaffected adjacent columns.

3.3 Distortion Correction

Normally, the step of rectification is applied afterthe correction of the science spectra using flat-field exposures. However, due to the fact thatthe MOSCA focal reducer causes very strong ge-ometric distortions and an approximation of theCCD response function is impossible, the recti-fication has to be performed before the flatfieldcorrection.For the slits at the top and bottom of the CCDchip which are lying closest to the edges of theMOS field, the distortions were strongest and thecurvature of spectral features corresponds to adisplacement of up to 20–22 pixel in Y direction,


Figure 3.1: Lower part of a raw MOS frame ofmask 1 for Abell 2390 with 3000 seconds exposuretime showing the slits # 1 to # 10. The magnificationscale along the X axis corresponds to ∼ 680 pixel, andalong the Y axis to ∼ 720 pixel. The bright spectrumin the middle belongs to a star which was used forthe alignment of the MOS mask. The wavelength isincreasing parallel to the X axis from left to right.The curvature of the spectra is clearly visible for thelower slits, whereas slit # 7 in the center of the CCDchip is nearly undistorted.

whereas slits # 7 and # 8 situated in the mid-dle of the chip were least undistorted (9 pixel).In the worst case, the outermost slit # 17 wasdistorted by 35 pixel (Low-LX April 2001 run).Figs. 3.1 and 3.2 visualise the curvature of theraw MOS spectra. Note, that the slits are num-bered from bottom to top by convention.For the purpose of correcting the S–distortions(curvature) of the spectra, an IRAF programwas upgraded which fits a user-defined Legen-dre polynomial (in most cases of third order) tothe galaxy spectra in each slit to derive the pa-rameters for the curvature correction along thespatial axis. The same set of coefficients werethen applied to the science, flatfield and corre-sponding wavelength calibration frames, respec-tively. The rectification was performed by re–binning the respective spectra in Y direction toa scale of 0.033 arcsec and shifting each col-umn according to the fit either in the +Y di-rection for the slits # 1 to # 8 or −Y direction

a. original

b. corrected

Figure 3.2: S–distortion correction for the A 2390galaxy # 2453 (I = 17.27 mag) of slit # 1. The upperpanel a. shows the uncorrected 2D spectrum andin the lower panel b. the corrected 2D spectrum isplotted. The X axis corresponds to ∼ 2000 pixel, theY axis to ∼ 180 pixel. The positions of the skylines,which showed no distortion in spatial direction wereused to judge the quality of the curvature correction.

in case of slits # 9 to # 17 for mask 1 and 2 ofthe Low-LX April 2001 runs. The number ofslits varied for different MOS setups (see e.g.,Table 2.2 of Observations for A 2390), with amaximum of 24 slits for mask 3 of the A 2390Sep. 2000 run only. This procedure is shownas an example for galaxy # 2453 in Fig. 3.2,where the original, uncorrected 2D spectrum iscompared to the spectrum after the S–distortioncorrection. Tests with other rectification algo-rithms confirmed that this procedure leads toflux–conserving results to within a few percent.In addition, the routine features also a step forthe removal of cosmic ray hits (the “clean” task).However, this rejection was not applied as no cos-mics were detected.

Possible artifical tilts of the emission lines (i.e.such visible in FORS spectra) were tested fordifferent wavelength calibration frames. How-ever, as no distortions in spatial direction weredetected, no further rectification along the X axiswas necessary. For this reason, the positions ofthe skylines were applied to judge the quality ofthe curvature correction.


3.4 Flat–fielding

In this step two features are addressed, the cor-rection of the pixel-to-pixel variations due to in-trinsic sensitivity differences of the detector andto account for the illumination function of theslit. Dome flat–field (FF) exposures throughthe masks were taken by illuminating the tele-scope dome with the external (within the domeinstalled) FF lamp. With exposure times of100 sec, spectra have typically ∼4000 ADU. Forboth Low-LX runs all three lamps 2, 3 and 4 wereswitched on together, yielding in ∼40000 ADUfor an exposure time of 30 sec. Due to the rel-atively short exposure times, none of these cali-bration frames were contaminated by cosmic rayartefacts, which was additionally confirmed by acosmic ray removal test and a visual inspection.A minimum number of five dome flat–fields wereused to construct a master flatfield through acombination of median–averaged single flatfieldexposures. After the rectification, the individualflatfield slit exposures were approximated sepa-rately by a spline function of third degree to thecontinuum of each spectrum in dispersion direc-tion (X axis) to account for the CCD responsefunction. The smoothed spline fit was afterwardsapplied for normalisation of the original flatfieldframe in order to correct for pixel-to-pixel vari-ations.Several tests with different setups using the in-ternal tungsten continuum source (at 60% illu-mination) resulted in FF images with too lowcounts. In addition, there are two main prob-lems in using the tungsten lamp. Firstly, thecontinuum lamp is hot and therefore generatesa blackbody spectrum different from the nightsky, which has to be removed before it can beapplied for the flatfielding correction. Secondly,as this lamp is located relatively near the CCDchip, the slits get illuminated in a different waythan when pointed at the sky. However, after aremoval of the blackbody spectrum of the tung-sten source, the counts in the FF frames to-wards the end of the spectrum are quite low.

This might affect blue absorption indices suchas CN (λc ∼ 3830 A) or CaH+K (λc ∼ 3980 Aand λc ∼ 3950 A, respectively) for a galaxy at aredshift of z ∼ 0.1, which fall near to the blueend of the spectrum. A FF calibration exposurewith low S/N will only add additional noise tothese blue features of interest.Skyflats offer maybe the best results in correct-ing the illumination of the CCD chips. How-ever, due to varying weather conditions only afew dusk sky FF calibration frames per nightcould be taken. As a median over at least fiveskyflats is needed for a master sky FF, a flatfieldcorrection with skyflats was neglected.

3.5 Sky Subtraction

The sky background was subtracted using theFortran routine hisupoly, kindly provided byProf. R. Bender. This algorithm approximatesthe night sky by iteratively fitting each CCDcolumn separately through a κ–σ clipping al-gorithm. The order of the polynomial fit is afree parameter. For most cases a first order wasused, if the night sky was relatively bright orvaried within the slit frame, a fit of zero or-der was applied. The quality of sky subtractionwas warranted through a generated verificationframe containing all fitted and extracted nightsky lines. If this image comprised only few skylines or some artefacts were still within the skysubtracted spectrum the procedure was repeatedagain.To ensure consistent profile centers to within atleast half a pixel, corresponding to 0.17′′, theone-dimensional spectra were extracted for eachexposure separately using the Horne–algorithm(Horne 1986). Therefore, possible shifts of thespectra which may have occurred during differ-ent nights are accounted for in the final one-dimensional summed-up spectra. The Horne–procedure optimally weights the rows of the ex-tracted spectrum profile to maximise the signal-to-noise in the resulting one-dimensional spec-


trum. In addition, this algorithm removes cos-mic ray events by analysing the profile perpen-dicular to the dispersion (along the Y axis).However, the method of optimal weighting byHorne (1986) might have an effect on the veloc-ity dispersion (σ) measurements of the galaxies.For this reason, a second independent extractionmethod was performed but the impact on thevelocity dispersions is smaller than their typicalmeasurement uncertainty (see section 5.3.5 for amore detailed discussion).

3.6 Wavelength Calibration

For the wavelength calibration exposures, theHgAr and Ne lamps were switched on togetherin order to gain a sufficient number of emissionlines over a large range of λλ ≈ 5450–7450 A.After each series of two science mask observa-tions, while the telescope still pointing at thetarget object, an additional HgAr/Ne exposurewas taken. This procedure allows to account forany possible flexure or deflection in the instru-ment due to a varying telescope pointing posi-tion which might affect the wavelength calibra-tion frames if they were gained while the tele-scope was still in parking mode and pointingat zenith. Repeat observations were unaccept-able because of a too large overhead time. How-ever, the combination of HgAr/Ne lamps hasthe disadvantage of almost total lack of calibra-tion lines below 5461 A: only two stronger Hgemission lines at 4358 A and 4047 A are visi-ble. In order to overcome this inconvenience, acombination between the Ar lamp and the stan-dard MOSCA BV–filter (λ0 = 4720 A, with a≈ 785 A bandwidth over a wavelength region ofλλ ≈ 3930 − 5510 A) was used to increase theuseable wavelength lines within the blue spec-tral region. During the CAHA observations ofthe Low-LX run (April 2001), different setups ofthe Ar lamp with the BV–filter (472/78) wereobserved. Table 3.2 gives an overview of wave-length calibration setups with and without the

Table 3.2: Calibration setups using the BV–filter.

Case Lamp Filter Texp [sec]

A HgAr – 5B HgAr BV 15C HgAr/Ne – 15D Ar BV 120E Ar – 3000

BV–filter.Different combinations between a wavelengthcalibration frame and the BV–filter were tested.Overall, wavelength calibration setups with ex-posure times of 5 sec (case A) or without us-ing the Ne lamp (case B) resulted in too fewemission lines. For case E, only a long expo-sure of the Ar lamp, too many lines were over-exposured and thus were not useable for mea-suring the line–widths. The most promising testwas a standard wavelength calibration frame of15 sec of all three HgAr/Ne lamps, augmentedby a 120 sec exposure of the Ar lamp using theBV–filter (case C+D). This test wavelength cal-ibration spectrum is shown in Fig. 3.3. Thelong exposure with the single Ar lamp was usedin order to uncover weak emission lines withinthe blue wavelength region and the BV–filterto suppress lines at the blue and especially atthe red end of the spectrum, which would oth-erwise become due to the long integration timesaturated. As a result, a couple of weaker emis-sion lines are gained below 5461 A and a hand-ful of lines within the blue region of interest be-low 4358 A. However, despite the BV–filter allof these blue lines are overexposured and there-fore “smeared–out”, making them impossible fora reliable wavelength calibration (see “blue re-gion” in Fig. 3.3). In addition, the important lineat 4047 A gets totally corrupted by these verybroad blue lines, thereby becoming nearly invis-


Figure 3.3: Wavelength calibration spectrum witha supplementary BV–filter. The dotted lines indicatecalibration lines of the standard HgAr/Ne wavelengthspectrum, whereas the ‘blue region’ marks a region ofsaturated, broad emission lines below ∼ 4358 A. Theusage of these blue lines revealed no improvementsfor the wavelength calibration. Since the displayedlines cover strengths over a broad range, the countsare given in logarithmic scale.

ible and not clearly detectable in the final wave-length calibration image. Nevertheless, severaltests of wavelength calibrations with the BV–filter were performed. None of these resulted inan improved calibration of the spectra. There-fore, an usage of the combined HgAr/Ne andAr+BV calibration frame for the wavelength cal-ibration was discarded.For the wavelength calibration the input cata-logue which is scanned by the MIDAS routinesfor the line search was thoroughly edited to en-sure that it only contained emission featureswhich are clearly detectable in the MOSCA spec-tra. This was carried out by comparing two cal-ibration spectra (one with very low starting andone with very high ending wavelengths, respec-tively) to the list of calibration lamp lines as pre-sented in the MOSCA User Manual which wasreleased by the PI of MOSCA Prof. J. W. Friedin May 20003. A total of 53 lines were includedin the final catalogue.

3http://www.mpia-hd.mpg.de/MOSCA/index.html

For the performance of the wavelength calibra-tion, two of the most crucial parameters are thetilt of the emission lines along the Y axis and thedegrees of the polynomial fits to the dispersionrelation along the X axis, respectively. Firstly,different spectra were examined and neither a tiltof emission lines as a function of exposure time orwavelength nor of slit position on the MOS maskwas found. Secondly, the fit degrees were fixedfor the entire spectrum (using the MIDAS algo-rithm REBIN/LONG), whereas the fit coefficientsare defined via a linear rebinning row-by-row.Various degrees of fits along the spatial axis werecompared for individual spectra. It turned outthat a two–dimensional dispersion relation us-ing polynomial fit functions of third and seconddegree in dispersion and in spatial direction, re-spectively, yielded the best results. As a roughguide, one should aim for an r.m.s. value of lessthan 10% in the dispersion for the calibrationprocess (e.g., Wagner 1992), which is in case ofMOSCA <0.1 A. The typical mean r.m.s. (er-ror) that was achieved with the dispersion fit was0.05-0.06 A at a dispersion of ∼1.28 A/pixel.After the wavelength calibration, individual ex-posures were summed up. Fig. 3.4 shows typi-cal final one-dimensional spectra in the A 2390sample. These fully reduced 1D spectra were re-binned to logarithmic wavelength steps in prepa-ration for the measurement of the velocity dis-persions (chapter 5).

3.7 Template Spectra

In a similar manner to the spectra of the galax-ies, the spectra of standard stars were reduced.Therefore, only a summary is presented but noexplicit description will be given here. Duringeach observing run at least two spectrophoto-metric flux standards (HZ2 and HZ4) were ob-served through an acquisition star hole in onemask. Moreover, a multiplicity of kinematictemplate stars were gained during each observa-tion period for all described projects. In total, 12


templates in Sep. 1999, ten in July 2000, six inApril 2001 and three stars in February 2002. Thestars were observed through different longslitwidths between 0.5–1.5′′. However, only thestars of the last run (Feb. 2002) had a spectralresolution which was sufficiently high to utilisethem as templates for even resolving velocity dis-persions with σgal ≈ 90 km s−1. A more detaileddescription of the data for these kinematic tem-plates with an application to the Lick/IDS sys-tem is illustrated in appen. A. For the kinematictemplates of the Feb. 2002 run, three K giantstars (SAO 32042 (K3III), SAO 80333 (K0III),SAO 98087 (K0III)) were observed through a0.5′′ longslit using the same grism green 1000

as for the galaxies. These stars had a spectralresolution at Hβ and Mgb of ∼2.2 A FWHM, cor-responding to σ∗ = 55 km s−1). A stellar tem-plate spectrum was split into several positionsalong the Y axis, which were slightly shifted fromeach other by <3×10−3′′ via a manually con-trolled zig–zag movement. To minimise the ef-fect of any possible variation (e.g., of instrumen-tal line profile, see also section 5.2) in slit width,the star spectra were averaged over a small num-ber of rows only.


Figure 3.4: Example spectra (not flux-calibrated) of early–type galaxies in the A 2390 sample. Thelower X axis represents the rest frame wavelengths, the upper one the observed wavelengths (bothin A). The ordinate gives the flux in counts (1 ADU=1.1 e−1). Prominent absorption features aremarked and the ID and the determined redshift are given above each panel.

Chapter 4: Photometric Analysis 47

Chapter 4

Photometric Analysis

Apart from the VLT/FORS and MOSCAintermediate–resolution spectroscopy, this studyuses multi–band imaging data from differentsources. The ground–based photometry pro-vides the basis for the derivation of the lu-minosities of the early–type galaxies which areutilised in the scaling laws of the Faber–Jacksonrelation. Hubble Space Telescope (HST) imag-ing with the Wide Field and Planetary Cam-era 2 (WFPC2) offers a spatial resolution of∼ 0.15 arcsec FWHM and the Advanced Cam-era for Surveys (ACS) which is operational sinceJuly 2002 even ∼ 0.05 arcsec FWHM. Both pho-tometric instruments are perfectly suited to anal-yse the surface brightness distribution of the dis-tant galaxies and to perform a detailed mor-phological classification into the sub–classes ofearly–type galaxies. Photometry in differentpassbands (U , B and I for Abell 2390 and B,V , R and I for the Low–LX clusters) for bothcluster samples was acquired from ground–basedtelescopes. Additionally, each cluster centre wasobserved with the HST/WFPC2 instrument ei-ther in the V I (A 2390) or the R (Low–LX clus-ters) filters. Thanks to the very deep multi–bandimaging of the FDF and WHDF, the basis for thederivation of the luminosities of the field early–type galaxies is more robust than most of otherdistant field early–type samples. For the FDF,photometry in four filters (B, g, R and I) is avail-able, whereas for the WHDF sample three filterpassbands (B, R and I) could be used.

In the next two sections 4.1 and 4.2, details will

be given on the ground–based and HST pho-tometry of the Abell 2390 cluster and of thethree Low-LX clusters, respectively. Besides adescription of the imaging data, the analysis willconcentrate on the derivation of absolute mag-nitudes for the galaxies. The reduction andcalibration of this large amount of photometricdata is described elsewhere (for A 2390: Smailet al. 1998; for the Low–LX clusters: Baloghet al. 2002b) and was not part of this the-sis. The measurement of structural parametersby modelling the surface brightness distributionfor all early–type galaxy candidates located onthe HST/WFPC2 and HST/ACS images is pre-sented in section 4.3. To distinguish betweenthe main classes of early–type galaxies, ellipti-cal and S0 galaxies, section 4.4 will focus on themorphological classification. Two independentapproaches will be presented, a visual type classi-fication according to the original Hubble schemeand a quantitative analysis following the de Vau-couleurs’ classification of the revised Hubble se-quence using the bulge–to–total fraction. A pos-sible correlation of the measured bulge fractionwith visual morphological Hubble type is inves-tigated and performance tests between differentfilter passbands are discussed. The analysis stepsto derive the absolute magnitudes, one basic par-ameter for the construction of the Faber–Jacksonscaling relations, will be presented in section 4.5.Results on the ground–based luminosities will beoutlined in section 4.6. In the following analysis,it is assumed that the spectroscopic redshifts of

48 Chapter 4: Photometric Analysis

Figure 4.1: I-band image of the cluster Abell 2390 at z = 0.23, taken with the 5.1-m Hale telescope onMt. Palomar. The total FOV is 9.′6 × 8.′5. North is up, east to the left. The length of the scale bar in theleft corner corresponds to 100′′, or 365 kpc at the distance of A 2390. Galaxies with spectra are indicatedby squares together with their ID number. The cD galaxy is marked with a circle. Stars used for the maskalignment are labelled as ID = 900X. An overlay of the HST/WFPC2 field-of-view is also shown.


the galaxies have already been determined. Themeasurements are presented in chapter 5.

4.1 Photometry of Abell 2390

The cluster Abell 2390 (α2000 = 21h53m34.′′6,δ2000 = +1740′10.′′9) at z = 0.228, rich-ness class 1, has a large velocity disper-sion of σ = 1100 ± 63 km s−1(Carlberget al. 1996) and a high X-ray luminos-ity, LX(0.7− 3.5 keV) = 4.7 × 1044 erg s−1

(Le Borgne et al. 1991). Carlberg et al.(1996) analysed the dynamical state of thecluster and its mass distribution and founda virial radius of Rv = 3.156h−1

100 Mpcand virial mass of Mv = 2.6 × 1015h−1

100M,which makes A 2390 more massive than Coma(Mv = 2.1 × 1015h−1

100M). At a mean interiordensity of 200ρc, A 2390 and Coma have M200-masses of 1.2 and 1.3 × 1015h−1

100M1, respec-

tively.The study of the early–type galaxy population inA 2390 is based upon Multi-Object Spectroscopy(MOS) using MOSCA at the Calar Alto 3.5-mtelescope on Calar Alto Observatory in Spain.Section 2.1.6 describes these spectroscopic ob-servations. In addition, optical photometry fromthe 5.1-m Hale telescope at Palomar Observatoryis available (section 4.1.1) and the WFPC2 im-ages taken with HST providing high-quality mor-phological information for a subset of the samplehave been exploited (section 4.1.2). Table 2.2 onpage 30 gives a summary of the photometric andspectroscopic observations for this project.Since the galaxies were distributed over thewhole field-of-view (FOV) of ∼10′ × 10′, corre-sponding to 1.53× 1.53 h−2

70 Mpc2,the evolutionof early–type galaxies out to large clustercentricdistances of ∼0.5 virial radii can be studied.

1For the virial masses a cosmology withH0 = 100 km s−1 Mpc−1, q0 = 0.1 and Λ = 0 wasassumed.

Figure 4.2: Zoom of the Hale I-band image of thecluster Abell 2390 at z = 0.23. Galaxies with spectraare denoted with squares and the cD galaxy is markedwith a circle. The HST/WFPC2 field-of-view is over-layed, with the WF chip corresponding to 114′′ alongone side and the PC chip with a size of 52′′.

4.1.1 Ground-based UBI Imaging

Abell 2390 was observed at the 5.1-m Haletelescope on Mount Palomar using COSMIC(Carnegie Observatories Spectroscopic Multislitand Imaging Camera) in the U (3000 sec), B(500 sec) and I-band (500 sec), allowing to se-lect early-type galaxies in the full field-of-view ofMOSCA due to the nearly equally large field-of-view of COSMIC of 9.7′×9.7′. Seeing conditionsranged from 1.4′′ in the U -band, 1.3′′ in the B to1.1′′ in the Cousins IC-band (Smail et al. 1998).Hereafter the Cousins IC-band is referred to asI-band. At I = 22.5 mag a completeness level of80% from a comparison with deeper field countsis warranted. All frames from the ground-basedimaging data were reduced in a standard mannerwith iraf using standard reduction packages.Fig. 4.1 shows the I-band image of the clusterAbell 2390 at z = 0.23, gained with the 5.1-m


Figure 4.3: Left: Distribution of Abell 2390 cluster members at z = 0.228 by Yee et al. (1996). In total,159 early–type galaxies (E) and 13 E+A galaxies are distributed over a strip of 6.47′×35.43′, centered onthe cluster cD galaxy (square). Only early–type galaxies (asterisks) and E+A galaxies (triangles) within aredshift interval of zclus ± 2σ are shown. Right: Comparison between Yee et al. (1996) and the cataloguedA 2390 galaxies. Spectroscopically observed galaxies of this work are indicated with open circles. Ten galaxieswith spectra which are included in both data sets are marked as solid circles.

Hale telescope on Mount Palomar. The totalFOV is 9.′6 × 8.′5, north is up and east to theleft. Galaxies with available MOSCA spectra aredenoted by squares and the central cD galaxy(no spectroscopic information) is indicated witha circle. The central 150′′ × 150′′ of the clusterare covered by the HST/WFPC2 mosaic, whichprovides a detailed morphological and structuralanalysis.

As a consistency check of the ground-based pho-tometry the photometric data was comparedwith the results for the cluster A 2390 by Yeeet al. (1996), which were derived as part of theCNOC cluster redshift survey. Through a cross-correlation, 12 galaxies with spectra were iden-tified which are included in both data sets. Theselection process for these galaxies is visualisedin Fig. 4.3. The left panel shows the distribu-tion of 159 early–type cluster members and 13E+A galaxies in A 2390 by Yee et al. (1996)over a large strip of the sky of 6.47′×35.43′ cen-

tered on the galaxy cluster. On the right handplot, a comparison between Yee et al. (1996)and the A 2390 galaxies in this study is illus-trated. Galaxies with available spectra are in-dicated with open circles, whereas the galaxiesby Yee are indicated by asterisks. In total, tenspectroscopic cluster members which were foundin both samples are marked as solid circles. Oneobject (# 5552), fell by coincidence into the slitand only a redshift was derived but no magni-tude could be measured. The galaxy (# 3038)shown with an open circle is a foreground gal-axy at z = 0.18 (cf. Table 4.1). Fig. 4.2 showsa zoom of the ground-based Hale I-band imagecontaining these 12 early–type galaxies which arein common. After the transformation of the I toGunn r observed magnitudes using r − I = 1.14based on Kinney E/S0 spectra, no significantdifference between the magnitudes was found,∆(|r − rYee|) = 0.04± 0.12 mag. Figure 4.4 andTable 4.1 give a comparison of the Gunn r mag-


nitudes for the 11 galaxies which are in com-mon. In addition, the redshift determinationsof all these objects show a very good agreement(|z−zYee| = 0.3−3±0.2−3). Note that the trans-formation from Iobs to robs for the I-band mag-nitudes are very large at a redshift of z = 0.23.For this reason, in the subsequent analysis ofthe scaling relations the magnitudes were trans-formed to Gunn r rest-frame magnitudes whichinvolves much smaller k-corrections and there-fore less uncertainties. A detailed description onthe transformation of magnitudes from observedto rest–frame wavelength range can be found insection 4.5.3.

In Fig. 4.5 a) the colour-magnitude diagram(CMD) (B − I) versus I from the Hale imagingis shown for all galaxies brighter than I = 23.5m

lying in a 9.7′ × 9.7′ (2.12 Mpc) region centeredon A 2390. Bona fide stellar objects selectedby a SExtractor based star classification par-ameter of star≥ 0.9 have been discarded. The48 early-type cluster members of A 2390 withavailable spectroscopic information are indicatedwith open squares. The red sequence of early-type cluster members is readily seen extendingdown to I ∼ 21m (MB ∼ −19m). A least-squares fit to seven ellipticals in A 2390 gives(B− I)E = −0.011 (Itot) + 3.405 which is shownby the solid line in Fig. 4.5. One elliptical E/S0galaxy (# 6666) is not included in the CMDs asthis galaxy was not selected as a spectroscopictarget but fell by coincidence into the slit dur-ing the observations. The outlier object # 2237with the bluest colour of (B − I) = 2.77 wasclassified a spiral Sa galaxy. Fig. 4.5 b) dis-plays the (U − B)–I colour magnitude relationfor the same galaxies as in Fig. 4.5 a. Again, aleast-squares fit to seven A 2390 cluster ellipti-cals yields (U −B)E = −0.020 (Itot) + 1.092.

The (B − I)–(U − B) colour–colour plane forgalaxies brighter than I = 23.5m is shown inFig. 4.6. At first glance a “clump” of red galax-ies at (B− I) ∼ 3.2 and (U −B) ∼ 0.6 is clearlyvisible. This region is occupied by all E+S0 clus-

Figure 4.4: Comparison between ground-basedGunn r magnitudes by Yee et al. (1996) and thisstudy for A 2390 cluster members (solid circles).Gunn r magnitudes were derived from the Itot mag-nitudes. The galaxy denoted with the open circle isa foreground galaxy at z = 0.18 (see text for details).

ter members of A 2390 which are indicated withopen squares. The observed galaxies are com-pared to evolutionary tracks for E/S0 galaxiesusing the BC96 models assuming a formationredshift of zf = 2. The predicted BC96 colourfor a passively evolved galaxy with an old stellarpopulation at redshift z = 0.23 suggests coloursof (B− I) = 3.06 and (U −B) = 0.57, which areshown as dashed lines in Fig. 4.6. Note that forthe computation of the BC96 models a slightlydifferent cosmology with H0 = 60 km s−1 Mpc−1

and q0 = 0.1 was adopted.

4.1.2 HST Photometry

Apart from the deep multi-colour ground-basedimaging, detailed morphological information isprovided via two HST/WFPC2 images of A 2390for a subset of the galaxies in the rich clustersample.During Cycle 4 A 2390 was observed on Decem-


Table 4.1: Comparison between ground-based Gunn r magnitudes by Yee et al. (1996) and the Gunn rmagnitudes of this study. The mean errors in Gunn r and (g− r) colour by Yee et al. (1996) are 0.05m and0.07m, respectively. The uncertainties in the redshifts are listed in units of 10−5.

ID z r IDYee rYee (g − r) zYee δzYee ∆(r − rYee)[mag] [mag] [mag] 10−5 [mag]

2120 0.2280 19.83±0.10 100873 20.15 0.78 0.2286 23 −0.3162138 0.2465 18.58±0.04 100908 18.64 0.86 0.2470 28 −0.0562180 0.2283 19.21±0.05 100995 19.23 0.91 0.2290 24 −0.0162198 0.2335 19.98±0.06 100975 20.05 0.87 0.2332 29 −0.0662438 0.2372 20.18±0.07 101197 20.09 0.85 0.2371 37 0.0942460 0.2319 18.91±0.05 101106 19.29 0.94 0.2318 25 −0.3762592 0.2300 18.60±0.05 101183 18.58 0.87 0.2304 22 0.0243038 0.1798 18.90±0.04 101992 18.87 0.78 0.1796 22 0.0343053 0.2282 19.08±0.05 101930 19.04 0.91 0.2281 28 0.0443473 0.2228 19.19±0.06 101961 19.1 0.94 0.2221 26 0.0945552 0.2231 101987 20.25 0.94 0.2233 296666 0.2240 18.82±0.11 101190 18.83 0.88 0.2244 24 −0.006

ber 10, 1994 with the HST 2 in the F555W (V555)and in the F814W (I814) filters as part of a largegravitational lensing survey (P.I. Prof. B. Fort(CNRS, Paris), Proposal ID 5352). In total, 5exposures in the I814-filter and 4 in the V555-filter each with 2100 sec resulting in a total ex-posure time of Ttot = 10.5 ks and Ttot = 8.4 ksfor the I814 and V555, respectively. These expo-sure times are deep enough to determine struc-tural parameters down to rest–frame MB ∼ 23m

(Ziegler et al. 1999). The final mosaic covers afield of approximately 2.5′×2.5′ at∼0.15′′ resolu-tion in the core of the cluster. All single standardWFPC2 pipeline reduced frames were further re-duced and afterwards combined to a final mosaicimage by I. Smail. Here only a brief overview isgiven to the reader. Further details about thegeneral reduction process and the mosaic align-

2Based on observations made with the NASA/ESAHubble Space Telescope, obtained at the Space TelescopeScience Institute, which is operated by the Associationof Universities for Research in Astronomy, Inc., underNASA contract NAS 5–26555. These observations areassociated with program # 5352.

ment are given in Smail et al. (1997). In oderto allow for hot pixel rejection, individual expo-sures were grouped in sets of four single-orbitexposures each offset by 2.′′0. After standardpipeline reduction, the images were aligned us-ing integer pixel shifts and then combined usingthe IRAF/STSDAS3 task CRREJ. In the finalmosaic the PC chip has the same linear scaleas the WF chips and the sky background levelsare equalised for all four chips. Details on thecalibration will be outlined in section 4.3. TheWFPC2 image of A 2390 has a 1σ surface bright-ness limit of µI ∼ 28 mag/pixel2, which is morethan adequate to provide high-quality morpho-logical information on the brighter cluster mem-bers. A two colour V I HST image of the clusterA 2390 is shown in Fig. 4.7. Thumbnail images ofthe 14 galaxies for which MOSCA spectra havebeen acquired are displayed in Fig. 4.12 and the

3The Space Telescope Science Data Analysis System(STSDAS) is a data processing software pipeline for cal-ibrating and analysing data from the HST using IRAF,operated by the STScI. (http:\\www.stsci.edu\resources\software hardware\stsdas).


Figure 4.5: Left: (B − I)–I colour magnitude diagram from the Hale imaging for galaxies brighter thanI = 23.5m lying in a 9.7′ × 9.7′ (2.12 Mpc) region centered on A 2390. The sequence of red cluster membersis readily seen extending down to I ∼ 21m (MB ∼ −19m). The 48 early-type cluster members of A 2390with available spectra are denoted with squares. Filled squares represent the ellipticals in A 2390 which enterthe Fundamental Plane. The solid line is a least-square fit to seven ellipticals. Right: (U − B)–I colourmagnitude diagram from the Hale photometry for bright galaxies lying in a 9.7′ × 9.7′ (2.12 Mpc) regioncentered on A 2390. Symbol notations as in the left Figure.

position of the HST field is indicated on Fig. 4.1.The cD was spectroscopically not observed andthus is not shown. Several strong and faint grav-itational lensed (cuspy and folded) arcs are visi-ble next to numerous galaxies (e.g., cD, # 2592,# 6666). Remarks on special features for individ-ual HST objects of Abell 2390 are summarisedin Table 4.9. Two arcs features (a cusp arc H3and a fold arc H5) associated with the early-typegalaxy # 6666 were analysed through lens mod-elling by Pello et al. (1999) and identified as mul-tiple images of a high-redshift source at z = 4.05,which is in rest-frame wavelength a Lyα emitterbehind the cluster A 2390. The “straight arc” as-sociated in the western part of the galaxy # 2592was analysed and can be splitted into two com-ponents, where component “A” is located in thesouthern part with zA = 1.033 and object “C”in the northern part with zC = 0.913 (Bunker,Moustakas & Davis 2000).

The analysis of structural parameters and mor-phologies as well as the derivation of rest-frameproperties for the sub-sample of E+S0 galaxiesfalling into the HST field is presented in sec-tion 4.3.

4.2 Photometry of Low-LXClusters

4.2.1 Low-LX Sample

One of the most comprehensive X-ray-selectedgalaxy cluster samples is the Vikhlinin catalogue(Vikhlinin et al. 1998). This catalog consists of203 clusters of galaxies serendipitously detectedas spatially extended X-ray emission sources inthe inner 17.′5 of 647 ROSAT Position SensitiveProportional Counter (PSPC) pointings. TheROSAT (ROentgen SATellit) PSPC has a field-of-view with a diameter of 2. For the cata-


Figure 4.6: (B−I)–(U−B) colour–colour plane fromthe Hale imaging for galaxies brighter than I = 23.5m

lying in a 9.7′ × 9.7′ (2.12 Mpc) region centered onA 2390. A 2390 E+S0 cluster members with availablespectroscopic information are represented with opensquares. The predicted BC96 colour for a passivelyevolved galaxy with old stellar populations (zf = 2)at z = 0.23 takes values (B−I) = 3.06 and (U−B) =0.57 and is indicated with the dashed lines.

logue, only high Galactic latitude pointings cov-ering 158 square degree were included. Out of203 (91%) optically confirmed galaxy clusters,73 feature spectroscopic redshifts (based on 2-3galaxies per cluster) while photometric redshiftsare given for all galaxy clusters. Galaxy clusterredshifts range from z = 0.015 to z > 0.5. TheX-ray luminosities LX derived from the observedfluxes in the 0.5–2.0 keV energy band vary from1042 to ∼5×1044erg s−1, corresponding to verypoor groups up to rich clusters respectively.The Low–LX clusters of galaxies comprising tenX-ray faint clusters in the northern hemispherewere selected from the Vikhlinin et al. (1998)cluster sample. The target clusters were re-stricted to a relatively narrow redshift range,0.22 ≤ zspec ≤ 0.29 and a mean redshift ofz = 0.25, in oder to reduce the effects of dif-ferential distance moduli and k-correction ef-

fects on the comparison between the systems.In total, using these redshift limits and the re-striction of accessibility from the northern hemi-sphere, a sample of 9 galaxy cluster were selectedwith an upper constraint on the X-ray flux ofLX < 2 − 20 × 1044 erg s−1. The X-ray lumi-nosities of these selected systems cover a range of0.15 ≤ LX ≤ 1.5×1044h−2

50 erg s−1. Based on thefollow–up spectroscopic investigation, the clusterCL 1444+63 turned out to be two separate struc-tures aligned along the line of sight (Balogh etal. 2002b). This yields to a final sample of 10X-ray faint clusters. In contrast to typical X-ray luminosities of X-ray selected clusters stud-ied with the HST so far (LX ∼5×1044 erg s−1),these galaxy clusters exhibit 1.0 dex lower X-rayluminosities.The X-ray luminosities LX were computed inthe 0.1–2.4 keV band from the observed fluxesin the 0.5–2.0 keV band and were corrected forgalactic Hi absorption and transformed to rest–frame (Balogh et al. 2002a). These luminositiesare listed in Table 4.4, and range from 0.40 to4.0 × 1043h−2

100 erg s−1 [0.1–2.4 keV]. AlthoughCL 1444+63 is treated as two separate clusters,the X-ray luminosity in Table 4.4 is that of thecombined clusters.

4.2.2 Ground-based Photometry

The ground-based imaging for the clusters isbased on three different observatories (i) CalarAlto CAFOS (2.2 m telescope) and MOSCA(3.5 m telescope) imaging mode observations,(ii) Isaac Newton telescope (INT) Wide-FieldCamera (WFC) imaging and (iii) Palomar 200-inch (P200) telescope observations at PalomarObservatory. Table 4.2 gives a summary ofthe ground-based imaging observations of all tenpoor clusters. The Calar Alto CAFOS observa-tions were carried out by R. G. Bower duringthe 28th of March 1999 and the MOSCA obser-vations by B. L. Ziegler between the 3.–6. March2000. The COSMIC imaging of the Palomar 200-inch telescope was performed by I. Smail dur-


Figure 4.7: V I colour composite HST WFPC2 image of the cluster Abell 2390, The total FOV is2.′5 × 2.′5. Galaxies with MOSCA spectra are indicated with their ID numbers. The cD galaxy isalso marked, but was not observed. The PC chip is not shown as no galaxy was selected from it.All candidates display red colours. Object # 2933 is an Sbc background galaxy at z = 0.398.


Table 4.2: Ground–based Observations of Low–LX clusters. Total exposure times are given in [ksec].

Name Instrument Date B V R I

CL 0818+56 COSMIC 26/11/98 0.60 0.30CAFOS 28/03/99 1.20

CL 0819+70 COSMIC 26/11/98 0.60 0.25CAFOS 28/03/99 0.30 0.60 1.80MOSCA 03/03/00 1.20

CL 0841+70 COSMIC 26/11/98 0.60 0.25CAFOS 28/03/99 0.30 0.60 1.80MOSCA 03/03/00 1.20

CL 0849+37 COSMIC 26/11/98 0.60 0.25CAFOS 28/03/99 0.30 0.60 1.20MOSCA 04/03/00 1.20MOSCA 05/03/00 0.60MOSCA 05/03/00 0.30

CL 1309+32 INT/WFC 19/06/98 1.80 2.40 0.60CAFOS 28/03/99 0.30 0.60 0.60MOSCA 06/03/00 5.40

CL 1444+63 INT/WFC 18/01/99 1.10 2.40 0.72CAFOS 28/03/99 0.30 0.30 0.30MOSCA 06/03/00 5.40

CL 1633+57 INT/WFC 10/02/99 1.80 0.60 1.20CAFOS 28/03/99 0.30 0.30 0.60 0.30MOSCA 04/03/00 1.20MOSCA 05/03/00 1.20

CL 1701+64/ INT/WFC 10/02/99 2.40 0.60 0.90CL 1702+64 CAFOS 28/03/99 0.30 0.30 0.30


Table 4.3: Overview of the observing conditions forground-based photometry of four Low–LX clusters.The mean seeing FWHM values for a night are givenin each column.

Name CAFOS MOSCA P200R B R

n2 / n3

CL 0819+70 1.9′′ 1.1′′

CL 0841+70 1.9′′ 1.5′′

CL 0849+37 2.5′′ ∼3.8′′ / 1.4′′ 1.7′′

CL 1633+57 1.6′′ ∼3.8′′ / 1.0′′–1.4′′

ing two nights in November 26.–27. 1998. Thegalaxy clusters CL 1701 and CL 1702 cover thesame field on the sky. For this reason, only onepointing was needed and thus are shown in com-bination in Table 4.2. Seeing conditions werenot photometric and varied between values of1.′′0 ≤ FWHMB ≤ 1.′′4 1.′′6 ≤ FWHMR ≤ 1.′′9for the B and R-band images, respectively. Ex-posures with seeing conditions worse than 2.′′5were replaced by repeat observations or by im-ages of other telescopes. Table 4.3 lists the ob-serving conditions for ground-based photometryof the CAHA B and R-band images as well asthe P200 R-band exposures of the CL 0849+37cluster. The I-band images are not included be-cause they were not used for this study. Firstresults of the Calar Alto B and R-band imagesfor four clusters of the sample, CL 0819, CL 0841,CL 0849 and CL 1633 were presented by Gaztelu(2000) and Ziegler et al. (2001b). As the R-band data from Calar Alto of the cluster CL 0849had poor seeing (2.5′′), the Palomar R image wasused instead. The remaining imaging data wasreduced and analysed by M. L. Balogh and I. R.Smail.During the imaging observations the conditionswere not photometric. Therefore, the imageswere calibrated by comparing aperture magni-tudes of several (usually 2–3) relatively isolated,

Figure 4.8: Comparison between ground-based Rmagnitudes by Gaztelu (2000) and the R magnitudesfor CL 0849 cluster members. Open circles denotemeasured R magnitudes by Gaztelu (RAG) with anaperture radius of rap = 2.65′′, solid circles repre-sent re–measured RAG magnitudes with an apertureof rap = 5.57′′ (see text for details).

early-type galaxies with the F702W photometryon the WFPC2 images (see next section 4.2.3),and converted this to standard R magnitudes,assuming RF702W − Rc = −0.2 (Fukugita etal. 1995). The photometric calibration is accu-rate to approximately 0.1m, including blue clus-ter members (but see discussion in the followingparagraph).Based on ground-based imaging from the Palo-mar 200-inch telescope and the INT, a sampleof 581 galaxies was selected for low–resolutionfollow–up spectroscopy of all 10 clusters (Baloghet al. 2002b). The sample selection was per-formed on the single INT WFC chip cover-ing 11.4′ at 0.′′33 arcsec/pixel. The PalomarCOSMIC images have a field-of-view of 13.7′

with a pixel scale of 0.′′4 arcsec/pixel. Gal-axies were selected for spectroscopic follow–upfrom the R−band images, except for two clusters(CL 1309+32 and CL 1444+63), where the selec-


tion was done on the I−band images becauseno R−band images were available. Reliable red-shifts could be derived for a total of 317 galaxies,of which 172 are cluster members (see section 5.5for more details). From this final catalog of 172Low–LX cluster members, galaxies for follow–upspectroscopy with intermediate–resolution for adetailed kinematic analysis presented here wereselected.

To check the consistency between the differentdata sets, the ground-based R photometry ofthe CL 0849 cluster galaxies was compared withthe results from Gaztelu (2000). Fig. 4.8 gives acomparison between the ground-based R magni-tudes by Gaztelu (2000) and the R magnitudesused in this work for CL 0849 cluster membersat z = 0.234. On average the differences be-tween the magnitudes of 〈∆R〉 = 0.06m are sat-isfactory for fainter galaxies with R >∼ 18.4m.However, for the brightest galaxies significant de-viations arise. Note that the work of Gazteluconcentrated only on the fainter galaxy popula-tion and thus chose a f ixed aperture radius ofrap = 2.65′′ (5 pixel) to measure the R-bandmagnitudes. Furthermore, the Palomar R bandimage was analysed with a wrong seeing valueand calibrated using the CAHA CAFOS R im-age which introduced an additional calibrationoffset of 0.05m to the magnitudes. Therefore,the calibrated magnitudes by Gaztelu have anaccuracy of 0.08m only. The measured R mag-nitudes by Gaztelu are indicated as open cir-cles. The choice of this fixed aperture size hasa dramatic effect on the brighter galaxies, show-ing fainter magnitudes with differences of about∆R ≈ 0.45m, for the brightest object even upto 0.9 mag. Due to the inconsistent values andto check the SExtractor analysis of the galax-ies in this study, the R magnitudes of the gal-axies by Gaztelu (2000) were re–measured us-ing a nominal aperture of rap = 5.57′′. The sizeof the aperture was examined for the whole setof brightest galaxies and yielded the best ap-proximation. The re–measured R magnitudes

of Gaztelu derived with an aperture diameter ofdap = 11.14′′ are represented as solid circles. Re-sults of these re–measurements were comparedto the aperture magnitudes derived with vari-able elliptical apertures using the SExtractorand GIM2D algorithms on both ground–basedand HST images (section 4.5.1). All three in-dependent approaches resulted in very similarR magnitudes. The differences were less thantheir scatter, e.g., ∆〈RSEx −RGIM2D〉 = 0.001m.For the subsequent analysis the SExtractormagnitudes were applied. These total appar-ent magnitudes showed mean uncertainties ofδR = 0.029± 0.013m (median value 0.03).

Fig. 4.9 shows the (B−R)–R CMD from the Haleimaging for all galaxies brighter than R = 22.5m

lying in a 9.7′ × 9.7′ (2.17 Mpc) region centeredon the cluster CL 0849. The 14 early-type clustermembers with available intermediate–resolutionspectroscopic information are represented withopen squares. Galaxy with ID # 54 has no(B − R) colour and is therefore not included.One galaxy (object # 3), which was previouslynot targeted in Balogh et al. (2002b), wasrevealed to be a cluster member based on itsintermediate–resolution spectra. Circles denote13 early-type cluster members, which were ver-ified by low–resolution spectroscopy (Balogh etal. 2002b). Rejecting one galaxy which is in-cluded in both setups, this yields a total sam-ple of 26 spectroscopic confirmed cluster mem-bers for CL 0849. At R ∼ 17.7m, R ∼ 19.3m

and R ∼ 20.6m the mean errors in R are 0.01m,0.03m and 0.07m and in (B − R) 0.05m, 0.08m

and 0.18m, respectively. These typical uncer-tainties are shown in the top of Fig. 4.9 as er-ror bars. A least-squares fit to three ellipticalsgives (B−R)E = −0.0667 (Rtot) + 3.6889 whichis shown by the solid line in Fig. 4.9. The CMsequence for the CL 0849 cluster is less well es-tablished than for the colour-magnitude relationof A 2390 (cp. Fig. 4.5) and shows a larger scat-ter compared to the early–type galaxies in mas-sive clusters. To verify these results, the ob-


servations can be compared to theoretical SSPmodel predictions by Kodama & Arimoto (1997,hereafter KA97). These authors calibrate theirmodels using the CM zero-point and slope of theobserved (V −K) vs. MV colour-magnitude rela-tion (CMR) for Coma cluster ellipticals (Boweret al. 1992). The population synthesis code usesas main characteristic parameters an IMF slopeof x = 1.20, a mass range 0.10 ≤ m/M ≤ 60.0,a metallicity range of 0.1−3 ≤ Z ≤ 0.05 anda burst duration of τ = 0.1 Gyr. The KA97models assume that elliptical galaxies have beenformed in a single burst event at a high formationredshift of zf ∼ 2 with a following passive evolu-tion of their stellar populations. KA97 demon-strate that for sub-L∗ galaxies the CM slopechanges are mainly caused by fainter galaxiesdue to variations in their mean stellar metallic-ity. The timescales for the loss of metals viagalactic winds is shorter for less massive gal-axies, which results on average in lower metal-licities for these systems. The scatter aroundthe CM relation is rather influenced by age ef-fects. Applying the KA97 models, the evolutionof the slope and the zero-point for any givenred sequence of galaxies as a function of redshiftcan be derived. Comparing our observed redsequence with SSP KA97 model assumptions,for a redshift of z = 0.234 a red sequence of(B−R)E = −0.0668 (Rtot)+3.6500 is predicted,which is indicated in Fig. 4.9 by the dashed line.This “theoretical” red cluster sequence is in ex-tremely good agreement with the observed redcluster galaxies spanned by the elliptical galaxiesof CL 0849. The model CMR has almost iden-tical slope and zero-point as the observed CMRdefined by the red cluster elliptical galaxies.

For the spectroscopic confirmed cluster membergalaxies of CL 0849 the fraction of blue galaxiesfb within the CM plane can be derived (Butcher& Oemler 1984). Galaxy clusters at low red-shift have a very homogeneous group of galax-ies with the cluster cores essentially devoid ofblue galaxies, whereas at z > 0.1 the blue galaxy

Figure 4.9: (B − R)–R colour magnitude diagramfrom the P200 (R-band) and CAHA (B-band) imag-ing for galaxies brighter than R = 22.5m lying ina 9.7′ × 9.7′ (2.17 Mpc) region centered on CL 0849.14 early-type cluster members with available spectraare denoted with squares. Filled squares representellipticals which enter the Fundamental Plane withthe solid line being a least-square fit to three ellipti-cals. Circles indicate 13 early-type cluster membersverified through low–resolution spectroscopy. Thedashed line is the predicted CMDs using SSP modelsby Kodama & Arimoto (1997). The vertical dottedline, which denotes MV < −20m, and the horizontaldotted line are used to define the blue galaxy fraction.

fraction increases with redshift and the numberof fb becomes significantly higher at z ∼ 0.5.Only galaxies brighter than MV < −20m in therest-frame (indicated as the vertical dotted linein Fig. 4.9) and those objects whose rest-frame(B − V ) colors are at least 0.2 mag bluer thanthe ridge line of the early-type galaxies at thatmagnitude cut are considered. For the clusterredshift of CL 0849 MV < −20m corresponds toR = 20.34m and at z = 0 ∆(B − V ) = −0.2 isapproximately the difference in colour of an ellip-tical and Sab-type spiral galaxy (Fukugita et al.1995). At z = 0.2, the ∆(B−V ) colour criteriumtranslates into (B − R)E = −0.0668 (Rtot) +


3.2500, which is indicated as the horizontal dot-ted line in Fig. 4.9. The blue galaxy fraction forCL 0849 gives fb = 0.077± 0.052. To put proba-bility limits on the galaxy fractions, the error offb was evaluated assuming Poisson statistics.

4.2.3 HST Photometry

HST/F702W (R702) imaging during Cycle 8 (P.I.Prof. R. L. Davies (Oxford), Proposal ID 8325)was acquired in July 2000 for 8 of the 9 clus-ters between 0.2 < z < 0.30 and with a meanredshift of z = 0.25. Each cluster was observedwith three single orbit exposures in the F702Wfilter with exposure times ranging from 2100 to2600 sec per orbit. Total integration times of7.500 sec warrant an accurate measurement ofthe structural and photometric parameters ofthose galaxy candidates located within the HSTfields. The three pointing positions for the ex-posures were offset by 10 pixels (Wide FieldCamera (WFC) pixel size) from one another butduring the reduction procedure the exposureswere aligned and coadded in oder to removecosmic rays artefacts and hot pixels. The im-ages were not drizzled or regridded because thisdoes not accurately preserve the noise charac-teristics of the pixels. After coadding the singleexposures, each WFC chip image was trimmedto the area of full sensitivity (see Balogh et al.2002a for details). The photometry is calibratedon the Vega system and updated zero-pointswere taken from the current instrument man-ual. The final images reach a 3σ point sourcesensitivity of R702 ∼ 25.5m, and cover a field of2.5′ × 2.5′ (or 0.57h−1

70 Mpc at z = 0.25) withan angular resolution of 0.17′′ (∼0.71h−1

70 kpc).The cluster CL 1633+57 was observed in Cycle 7(P.I. Prof. P. Rosati (ESO), Proposal ID 7374).The data of cluster CL 1633+57 were retrievedfrom the CADC4 HST archive. These data areWFPC2/F702W observations, with four expo-

4Canadian Astronomy Data Centre, which is operatedby the Herzberg Institute of Astrophysics, National Re-search Council of Canada.

sures of 1200 seconds. The total exposure timeof 4800 seconds is therefore less than that of theother nine clusters. Table 4.4 gives a summaryof the HST imaging and the X-ray properties forthe rich and poor cluster samples.For the Low–LX clusters a nominal magnitudelimit of R702 = 23.0m at z = 0.25 was adopted,which corresponds to M702 = −17.6+5 logh70 orapproximately MR ∼ −17.4 + 5 logh70 (assum-ing a ΛCDM cosmology and including small k-corrections of <∼ 0.1m), which is 4.2 magnitudesbelow M∗R ≈ −21.60m (Blanton et al. 2001).Galaxies brighter than this limit are luminousenough and have large enough sizes in oder towarrant a reliable and accurate classification.

4.3 Surface Brightness ProfileFitting

4.3.1 Choice of Luminosity Profile

A fundamental question when determining thestructural parameters of galaxies is what kind ofsurface brightness profile type is the most suit-able. Generally, derivation of total magnitudesinvolves an extrapolation of curves of growth toinfinity, thus relying on fits to the luminosityprofiles of the galaxies. For example, Caon etal. (1993) and La Barbera et al. (2002) usedSersic profiles to fit the surface brightness profileof early-type galaxies. Another approach is touse the classical de Vaucouleurs (r1/4) law withan exponential disc component (e.g., Simard etal. 2002; Tran et al. 2003).Tran et al. (2003) explored the results of threedifferent profile models (double exponential,Sersic bulge+exponential disc and an r1/4

bulge+exponential disc) using the GIM2Dpackage (Simard et al. 2002, see section 4.4.2for details) on a sample of 155 cluster galaxiesin CL 1358+62 at z = 0.33. Based on theirextensive profile tests the authors conclude thatthe de Vaucouleurs bulge with exponential discprofile is the most appropriate way to investigate


Table 4.4: HST observations of the cluster samples.

Name R.A. Dec. z Ttot LX (0.1–2.4 keV) Rlim702

[J2000] [ksec] 1044 h−250 erg s−1

Low–LX Sample of Poor Clusters

CL 0818+56 08 19 04 +56 54 49 0.2670 7.2 0.56 23.00CL 0819+70 08 19 18 +70 55 48 0.2296 6.9 0.48 22.80CL 0841+70 08 41 44 +70 46 53 0.2397 6.9 0.46 22.90CL 0849+37 08 49 11 +37 31 09 0.2343 7.8 0.73 22.85CL 1309+32 13 09 56 +32 22 14 0.2932 7.8 0.74 23.40CL 1444+63a 14 43 55 +63 45 35 0.2923 7.5 1.461 23.45CL 1444+63b 14 44 07 +63 44 59 0.3006 7.5 1.461 23.45CL 1633+57 16 33 42 +57 14 12 0.2402 4.8 0.20 –CL 1701+64 17 01 47 +64 20 57 0.2458 7.5 0.15 22.70CL 1702+64 17 02 14 +64 19 53 0.2233 7.5 0.33 22.90

High–LX Sample of Rich Clusters

A 2218 16 35 54 +66 13 00 0.175 6.5 10.9 22.10A 2390 21 53 35 +17 40 11 0.228 10.52 26.6 22.40

1CL 1444+63 is only detected as a single X-ray source; this LX presumably includes contributionfrom both clusters.

2F814W filter and Ilim814.

the structural properties. As the cluster samplesin the present study comprise only early-typegalaxies a similar approach is used, namely acombination of an r1/4 plus exponential discprofile by Saglia et al. (1997a). These authorsdemonstrated that the Sersic r1/n profiles canbe understood as a “subset” of r1/4+exponentialmodels. This algorithm also explores the effectof an efficient sky-subtraction. For example, forn > 4 large extrapolations are needed whichcould lead to large errors in the sky-subtractionof up to ±3% and, therefore, to additionaluncertainties in the absolute magnitudes andhalf-light-radii. The two-dimensional surfacebrightness distribution of the early–type fieldgalaxies was fitted with a combination of r1/n

plus exponential disc profile using the GALFITpackage (Peng et al. 2002). The Sersic exponentn was restricted to variable only within a range

between n = 1 − 4. For an internal consistency,the field galaxies in the FDF were modelledwith both algorithms and the parameterscompared to each other. It turned out, that acombination of de Vaucouleurs and exponentialdisc surface brightness profiles as used for thecluster galaxies is an equivalent approach to theGALFIT measurements and yields within theuncertainties to similar results.To test the effects on the structural parametersa second pure r1/4 profile model was appliedto fit the surface brightness distribution of thegalaxies. Results were extensively tested andtheir variations and imprints will be discussedtogether with the scaling relations in sections 6.3and 6.4.


4.3.2 The Surface Brightness Models

Structural parameters were analysed using thefitting algorithm developed by Saglia et al.(1997a, 1997b). The galaxies’ parameters arederived searching for the best combination ofseeing-convolved, sky-corrected r1/4 and expo-nential laws. This approach accounts for vari-ous types of observed luminosity profiles, i.e. itmodels the extended luminosity profiles of cDgalaxies and provides fits to the range of profileshapes of early-type galaxies (ellipticals featur-ing a flat core to S0 galaxies with the presenceof a prominent disc).

Subframes of all early–type cluster galaxies bothwith spectroscopic and HST information wereextracted and each galaxy was analysed individ-ually. In a first step stars and artifacts aroundthe galaxies were masked in order not to causeany problems in the fitting process. In thenext step the circularly averaged surface bright-ness profile of the galaxy was fitted with PSF-convolved r1/4 and exponential components,both simultaneously and separately. The indi-vidual PSFs for the profiles were computed usingthe TinyTim program by considering the Fouriertransform of the PSF as exp[(−kb)γ ] and adopt-ing γ = 1.6. The γ exponent defines the expo-nential decrease of the PSF function (hereafterreferred as γ PSF) and for γ = 2 the PSF is astandard Gaussian. The assumption of γ = 1.6was extensively tested by Saglia et al. (1997b)using γ PSFs of γ = 1.5−1.7 with a pixel size of0.8′′. The choice of a different PSF introduced asmall effect. If the true PSF is γ = 1.5, then thetotal luminosity, half-luminosity radius and scalelength of the bulge will be slightly overestimatedand the disc–to–bulge (D/B) ratio slightly un-derestimated. Opposite trends are found if thetrue γ is 1.7. Systematic differences become neg-ligible when Re > 2 FWHM.

For the r1/4 bulge profile a special form of theSersic profile (Sersic 1968), the classical de Vau-couleurs profile, in the following form was ap-

plied:

µ(R) = µe exp

−7.67

[(R

Re

)1/4

− 1

],

(4.1)where µ(R) denotes the surface brightness at Ralong the semi–major axis radius and µe the ef-fective surface brightness at the effective radiusRe, which refers to half of the total light of thegalaxy. The effective surface brightness at theeffective radius is given as µe = −2.5 × log Ie,where Ie is the intensity at the effective radiusRe. The disc profile is well represented by anexponential law and defined as:

µ(R) = µ0 exp(−Rh

), (4.2)

where µ0 is the central (face-on) surface bright-ness of the disc and h indicates the exponentialdisc scale length. For an de Vaucouleurs pro-file, the central surface brightness of the disc µ0

can be computed as µ(R = 0) ≡ µ0 = µe−8.327.This method allowed to derive the effective (half-light) radius Re (in arcsec), the total F702W-band magnitude R702 (Rtot) or total F814W-band magnitude I814 (Itot), the central surfacebrightness µ0 and the mean surface brightnesswithin Re, 〈µe〉, for the entire galaxy as well asthe luminosity and scale of the bulge (mb andRe,b) and disc (md and h) component separately,within the limitations described by Saglia et al.(1997a). An extensive description of the uncer-tainties and errors is discussed in the next sec-tion 4.3.3.In total, structural parameters could be deter-mined for 29 early-type galaxies, 14 in the richcluster A 2390 and 15 in the three poor clus-ters CL 0849, CL 1701 and CL 1702, which areall spectroscopic confirmed cluster members. Innumbers, 14 out of 15 (one is a background gal-axy) in A 2390 (see Table 4.5), 6/8 in CL 0849,7/12 in CL 1701 and 2/8 CL 1702. In case of theLow–LX clusters, all brighter cluster candidatesdown to R702 ∼ 20.75m were fitted because at


Figure 4.10: Circularly averaged I814 surface brightness profiles of four early-type galaxies of A 2390 withavailable spectral information. The calibrated surface brightness profile (squares) is shown together withthe fitted bulge (r1/4) and/or disc (exponential) profiles as a function of the 1/4 power of the distance Rfrom the galaxy center in units of arcsec. The horizontal dotted line denotes the position corresponding to1 per cent of the sky background level. The arrow indicates the position of the effective radius Re. Valuesfor the total magnitude Itot, the surface brightness of the disc µd,0 (∀µd,0 = 0→ @µd,0), the effective radiusRe and the effective radius of the bulge Re,b, and the disc-to-bulge ratio D/B are listed. Galaxies areordered with increasing D/B ratio. Below each surface brightness profile the residuals of the fits, ∆µ (R) =µobs (R)− µfit (R) (dots) and the integrated magnitude difference ∆mag (R) = magobs (< R)−magfit (< R)(full lines) are indicated. Note that ∆mag (R) < 0.06, even if large deviations of ∆µ are present.


the time of this analysis no spectra were avail-able to select galaxies based upon their redshifts.For this reason, the first number represents thefinal poor cluster members (including the low–resolution spectroscopy) and the second the to-tal number of analysed galaxies within the HSTFOV, i.e., members plus “anonymous” galaxies(see section 4.4.1). Fig. 4.10 and Fig. 4.11 showtypical examples of the surface brightness profilefits for early-type galaxies in A 2390. In mostcases, the observed profile is best described bya combination of a bulge (de Vaucouleurs law)and a disc component (exponential law), indi-cated in the figures as (sum: r1/4+exp.-law).Note that for an r1/4 profile the combination(r1/4+exp.) represents the seeing convolved r1/4

fit only. Fig. 4.10 illustrates in detail the resid-ual errors on the structural parameters for dif-ferent profile fits. For this reason, below eachsurface brightness profile the residuals of the fits,∆µ (R) = µobs (R) − µfit (R) (dots), in units ofmag/arcsec2, and the integrated magnitude dif-ference ∆mag (R) = magobs (< R)−magfit (< R)(full lines), in units of magnitudes, are shown.Note that for all cases ∆mag (R) < 0.06, even iflarge deviations of ∆µ are present. Note that ifthe surface brightness of the disc µd,0 is shown asµd,0 = 0 it has no value but is indefinite. Typi-cal residuals ∆µ814 between the observed profileand the fit as a function of R1/4 are in the or-der of 0.01m ≤ ∆µI ≤ 0.10m. Comparing thedifferences between elliptical and lenticular (S0)galaxies, a small disc component for the S0 ex-hibiting a lens-like structure is detected. For ex-ample, the elliptical galaxy # 2438 of the A 2390cluster is well represented by a pure de Vau-couleurs profile. In the case of the S0 galaxy# 2946 a combined model of an r1/4-law and anadditional disc component results in the best lu-minosity profile. In order to explore the impactof an additional exponential component to theobserved surface brightness profile, Fig. 4.10 liststhe early-type galaxies ordered with increasingD/B ratio, ranging from 0 < D/B < 0.93. The

different panels clearly show how the exponen-tial component becomes stronger as a functionof D/B ratio with a simultaneous weakening ofthe bulge component. For large D/B values ofD/B > 0.6 the exponential component domi-nates over the r1/4 component. In case of thegalaxy # 2198, the weak flattening of the sur-face brightness profile at R ≈ 1.2′′ is caused bya close neighbour. However, this effect startsat faint magnitudes µ814 ≈ 28.7 mag/arcsec2 atlarge radii R ≈ 1.2′′ and thus not influence themeasurement of structural parameters.

For some cases (e.g., # 2198, # 2438 and # 2763in the A 2390 sample) no additional disc com-ponent is detected, i.e. a pure classical de Vau-couleurs law profile (D/B = 0) results in the bestprofile for the galaxy’s light distribution. There-fore, this approach also allows for the possibilitythat dynamically hot galaxies may all have r1/4

profiles. But most of the galaxies are well de-scribed by the superposition of r1/4 bulge and ex-ponential disc profiles. Bulge+disc models pro-vide better fits than a pure r1/4 law, and theaddition of an exponential disc component im-proves the fit for many of the galaxies. The sur-face brightness at the half-light radius is a strongfunction of the chosen fitting profile and the half-light radius is dependent on the assumed surfacebrightness model. However, the product Re 〈Ie〉β(with β ' 0.8) provides a Fundamental Planeparameter which is stable for all applied profiletypes, from r1/4 to exponential laws (Saglia etal. 1993b; Kelson et al. 2000a). Under the con-straint of this coupling, galaxies can only moveparallel to the edge-on view of the FundamentalPlane.

Some previous Fundamental Plane studies de-rived their structural parameters with a purer1/4 profile. Therefore, the results were alsotested using surface brightness models based on apure de Vaucouleurs law only. Only systems fea-turing a dominant disc (Sa bulges) show some de-viations. As expected, these galaxies move onlyalong the edge-on projection of the Fundamental


Table 4.5: Galaxy properties of the A 2390 HST sub-sample. (1) galaxy ID, (2) velocity dispersion σc(aperture corrected) with errors [in km s−1], (3) extinction-corrected, Vega-based total magnitude in WFPC2F814W filter, (4) mean surface brightness magnitude 〈µe〉 in Gunn r corrected for cosmic expansion, (5)rest-frame Gunn r magnitude [in mag], (6) effective radius [in arcsec] and (7) effective radius [in kpc], (8)effective radius of the bulge [in arcsec], (9) FP parameter Re 〈Ie〉0.8, (10) disc scale-length [in arcsec], (11)disc-to-bulge ratio (D/B), (12) morphology T of the galaxy.

ID σc I814 〈µe〉r Mr Re logRe Re,b FP h D/B T[km s−1] [mag] [mag/2′′] [mag] [′′] [kpc] [′′] [′′]

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)

2106 144.5±16.8 19.18 18.28 19.93 0.282 0.013 0.250 0.033 0.20 0.36 E2120 140.3±15.3 18.88 18.22 19.63 0.314 0.060 0.202 0.154 0.34 0.46 S02138 165.4±08.7 16.97 20.60 17.72 2.264 0.918 1.324 1.800 4.48 0.36 E2180 161.2±10.6 17.80 19.49 18.55 0.928 0.530 0.580 1.195 0.88 0.60 S02198 148.9±15.6 18.59 17.92 19.34 0.313 0.058 0.313 0.155 0.00 0.00 E2237 162.2±12.3 17.66 19.94 18.41 1.217 0.648 1.195 1.385 0.73 0.81 Sa2438 119.8±15.9 18.74 19.59 19.49 0.631 0.363 0.631 0.808 0.00 0.00 E2460 258.6±10.2 18.00 18.90 18.75 0.642 0.370 0.509 0.892 1.09 0.17 E2592 192.0±11.4 17.20 20.11 17.95 1.625 0.774 1.104 1.618 2.55 0.29 E2619 229.3±14.3 18.53 18.20 19.28 0.366 0.126 0.320 0.326 0.27 0.37 E2626 195.6±12.7 18.31 18.65 19.06 0.497 0.259 0.325 0.640 0.38 0.92 S02763 300.5±12.2 17.55 18.89 18.30 0.789 0.460 0.789 1.108 0.00 0.00 E2946 131.9±15.2 18.69 19.26 19.44 0.554 0.306 0.405 0.708 0.60 0.34 S06666 211.0±13.5 17.65 18.79 18.40 0.721 0.421 0.488 1.024 1.00 0.32 E/S0

Plane. Note, that the errors of the FP parame-ters of effective radius (Re) and surface bright-ness (µe) are highly correlated. However, the rmsuncertainty in the product Re 〈Ie〉0.8 which prop-agates into the FP is only ≈4 % for all galaxiesin this study, corresponding to ∆µe = β∆ logRe

with β = 0.328 and a scatter of the latter rela-tion of 0.05 mag arcsec−2. The results of ther1/4 structural measurements are presented inTable 4.6. In chapter 6, the impacts of differentsurface brightness models on the FundamentalPlane will be discussed. In particular, Fig. 6.14on page 144 shows the Fundamental Plane forA 2218 and A 2390 constructed based on a purer1/4-law profile.

Surface brightness profiles of those galaxies sit-uated within the HST field of Abell 2390 with

available spectroscopic information are shown inFig. 4.11. The surface brightnesses were cor-rected for galactic extinction. In most cases theobserved profile is best fitted by a combinationof a bulge (de Vaucouleurs law) and a disc com-ponent (exponential law). Values for the totalmagnitude Itot, the surface brightness of the discµd,0, the effective radius Re and the effectiveradius of the bulge Re,b, and the disc-to-bulgeratio D/B are listed. The arrow in each dia-gram indicates the position of the effective ra-dius Re. Typical residuals ∆µI between the ob-served profile and the modelled fit are in the or-der of ±0.10 mag at R1/4 ≈ 1.3 arcsec. Note thatthe disc fit indicated in Fig. 4.11 represents thedisc component of the combined best fit for thecorresponding galaxies. The single galaxy which


Table 4.6: Galaxy parameters of the A 2390 HST sub-sample based on pure de Vaucouleurs surface bright-ness fits. Columns are analogous tabulated as in Table 4.5. The velocity dispersions σ with correspondingerrors [in km s−1] are not aperture corrected.

ID σ I814 〈µe〉r Mr Re logRe Re,b FP[km s−1] [mag] [mag/2′′] [mag] [′′] [kpc] [′′]

(1) (2) (3) (4) (5) (6) (7) (8) (9)

2106 131.0±16.8 19.07 18.55 19.82 0.335 0.088 0.335 0.2192120 127.2±15.3 18.75 18.51 19.50 0.382 0.145 0.382 0.3632138 150.0±08.7 17.14 20.04 17.89 1.621 0.772 1.621 1.6302180 146.2±10.6 17.67 19.79 18.42 1.133 0.617 1.133 1.3422198 135.0±15.6 18.59 17.92 19.34 0.313 0.058 0.313 0.1552237 147.1±12.3 17.12 21.10 17.87 2.667 0.989 2.667 1.8022438 108.6±15.9 18.74 19.59 19.49 0.631 0.363 0.631 0.8082460 234.5±10.2 18.01 18.86 18.76 0.628 0.361 0.628 0.8722592 174.1±11.4 17.27 19.87 18.02 1.408 0.711 1.408 1.5332619 207.9±14.3 18.40 18.55 19.15 0.456 0.222 0.456 0.5532626 177.4±12.7 18.01 19.29 18.76 0.768 0.448 0.768 1.0332763 272.5±12.2 17.55 18.89 18.30 0.789 0.460 0.789 1.1082946 119.6±15.2 18.62 19.36 19.37 0.599 0.340 0.599 0.7786666 191.3±13.5 17.65 18.74 18.40 0.701 0.408 0.701 1.000

could not be fitted (# 2933) is a late-type spiralgalaxy (Sbc) with clear signs of spiral pattern atz = 0.398. Thumbnail images for the 14 galaxiesin A 2390 are provided in Fig. 4.12. A detaileddescription of the morphological classification ofgalaxies residing in the HST field is outlined inthe section 4.4. The final classification is listed inTable 4.5 and is also given on top of each galaxysurface brightness plot in Fig. 4.11.

As an additional comparison, an isophote analy-sis based on the procedure introduced by Bender& Mollenhoff (1987) was performed. Deviationsfrom the elliptical isophotes were recorded as afunction of radius by a Fourier decompositionalgorithm. The presence and strength of the a4

coefficient, which represents the signature of dis-cyness, is in good agreement with the visuallyclassified morphologies of S0 and spiral bulges.

4.3.3 Error Evaluation

The errors of the photometric parameters of ef-fective radius and surface brightness are corre-lated and enter the FP in combination. Theerrors in Re (in arcseconds) and µe (in magni-tudes) propagate into the FP relationship in thefollowing form (Treu et al. 2001a):

δFPphot = log δRe − β δµe (4.3)

An error calculation with this formula is not anapproximation but a particularly robust compu-tation, as in three-dimensional space defined bythe FP it is the only natural way to provide anaccurate error estimation (Saglia et al. 1997a;Kelson et al. 2000a). For this reason, the errorsin the FP are only shown in the edge-on pro-jection along the short axis that separates kine-matic measurements and photometric parame-ters (see upper right panel in Fig. 6.11). Theerrors are calculated for the F814W filter with a


Figure 4.11: I814 Surface brightness profiles of A 2390 galaxies lying within the HST image with availablespectral information. The I814 surface brightness magnitude (extinction AF814W corrected) is plotted againstthe radius R1/4 (in arcsec). Filled squares show the observed profile, lines the different best models for thebulge (de Vaucouleurs law), disc (exponential law) component fit and for a combination of bulge and disccomponents (sum: r1/4+exp.-law). The arrow indicates the position of the effective radius Re. Somestructural parameters are given (see text for details).


Table 4.7: Error assessment via the parameter Q.

Name Q = 1 Q = 2 Q = 3

A 2390 6 8 0CL 0849 2 3 1CL 1701 0 7 0CL 1702 2 0 0

parameter value of β = 0.328. Since the slope βis very well defined with only a small variationwith wavelength, the error estimations changenegligibly within the observed range of β values(Pahre et al. 1998).The errors on the derived structural parametersare assigned through a combination of individualquality parameters, including the effects of skysubtraction errors, seeing and pixel sampling, av-erage galaxy surface brightness relative to thesky, S/N variations, extrapolation involved toderive Mtot, radial extent of the profile and re-duced χ2 of the fit resulting in a final global qual-ity parameter. The total quality parameter Qis defined by the maximum of the quality pa-rameters of individual errors Q = Max(Qmax,QΓ, QS/N , QSky, QδSky, QE, Qχ2) (Saglia et al.1997a). The accuracy of the quality parameter Qis given by a combination of several factors, suchas a maximal radius ratio which should be largerthan the effective radius (Qmax), the seeing con-ditions and/or possible inadequate sampling ef-fects (QΓ), the signal–to–noise (QS/N ), the dif-ference between the sky surface brightness andthe average effective surface brightness of themodels (QSky), sky subtraction errors (QδSky),the fraction of light extrapolated beyond Rmax

used in the determination of Mtot (QE) and thereduced χ2 of the fit to the luminosity profile ofthe galaxy (Qχ2). For Q = 1 the typical uncer-tainty in total magnitude is ∆Mtot = 0.05 andin effective radius ∆ log Re = 0.04 (< 10%). Avalue of Q = 2 leads to errors of ∆Mtot = 0.15and ∆ log Re = 0.1 (< 25%), respectively. A

Table 4.8: Error analysis for the FP of A 2390. Thetotal error on the combination that enters the FP,δFPphot = log δRe−β δµe (see equation 4.3), is listedas δFPphot, adopting β = 0.328. The final FP errorin the rest-frame properties is given in the columnδFPr.

ID δµe log δRe δFPphot δFPr

2106 0.078 -0.055 0.081 0.0822120 0.071 -0.020 0.044 0.0452138 0.136 0.124 0.080 0.0812180 0.049 -0.001 0.017 0.0192198 0.073 -0.020 0.044 0.0462237 0.076 0.050 0.025 0.0272438 0.027 -0.040 0.049 0.0512460 0.030 -0.019 0.029 0.0312592 0.105 0.094 0.059 0.0612619 0.054 -0.018 0.035 0.0372626 0.029 -0.030 0.040 0.0412763 0.032 -0.004 0.015 0.0162946 0.021 -0.026 0.033 0.0346666 0.028 -0.014 0.023 0.025

quality parameter of Q = 3 ensues errors of∆Mtot = 0.4 and ∆ log Re = 0.3. Table 4.7 givesan overview of the quality parameter Q mea-surements for the galaxies in A 2390, CL 0849,CL 1701 and CL 1702 which were revealed to becluster members. For the galaxy in CL 0849 (ID# 25) which has Q = 3 and for the galaxies inCL 1701 # 118, # 124, # 141 and # 149 whichhave all Q = 2, only low–resolution spectroscopyis available. For A 2390 it was found that 43%(6) of the galaxies have Q = 1, 57% (8) galaxieshave Q = 2 and no galaxy yield a quality par-ameter of Q = 3. The average error in I814 wasestimated to be ≈10 % and the error in Re of< 25 %.Furthermore, for the error evaluation variationsin the zero-point calibration of δZP = 0.03were accounted for (Saglia et al. 1997b), anuncertainty in galactic extinction δE(B−V) =0.010, as well as errors in the k-correction ofδmr = 0.02m. The combination of the uncer-


tainties on the measured parameters δµe [in mag]and log δRe [in arcsec] according to equation 4.3are listed as δFPphot. Adding the errors of galac-tic extinction correction, the k-correction andthe errors in zero-point in quadrature ensues thefinal total error of the FP in the rest-frame Gunnr-band, δFPr. A summary of the error analysisfor the galaxies in A 2390 is given in Table 4.8.

4.3.4 Rest–Frame Properties

In the following, a description of the transfor-mation of the measured surface brightnesses andeffective radii on the HST images to rest-frameis given. Here, only a summary of the syntheticphotometry is presented as it will be explainedin more detail for the ground-based photometryin section 4.5.3.At the cluster redshift of A 2390, the observedF814W passband is close to rest-frame Gunnrrest. Therefore, it is more promising to use asingle k-correction term of the observed I814 toGunn r in rest-frame rather than to take a morecomplicated way in converting to RC first andthen using a colour transformation. The latterprocedure would increase the overall uncertain-ties.Analysis of the rest-frame luminosities of the gal-axies, was performed in the following way. Ac-cording to Holtzman et al. (1995), the instru-mental magnitude in a filter X is given by:

IX = −2.5 log(DN)/texp + ZP + 2.5 log(GR)(4.4)

where X denotes the observed HST filter pass-band, either the F814W or F702W filter. GRrepresents the respective gain ratios for the WFchips and ZP the zero-point for an exposuretime texp = 1 sec with ZP = 20.986 (I814),ZP = 22.581 (R702), both with considerationof the difference between “short” and “long” ex-posures of 0.05 mag (Hill et al. 1998) and anaperture correction of 0.0983 for the I814 filterand of 0.1087 for the R702 passband (Holtzmanet al. 1995), respectively.

Since no ground-based (V − I) colours are avail-able for the cluster A 2390, a transformationfrom Cousins I to Gunn r as published in theliterature (e.g., Ziegler et al. 2001a) can not beconducted. Instead, the transformation was per-formed in the following way. A spectral tem-plate for typical E/S0 galaxy by Kinney et al.(1996) was selected and was redshifted to thecluster redshift of z = 0.23 to determine the fluxthrough the I814 filter and afterwards at z = 0through the Thuan & Gunn (1976) r filter (seesection 4.5.3 for details). From this procedurethe translation I814 obs−Gunn rrest = −0.75 wasderived. The result was checked by transformingthe synthetic SEDs by Moller et al. (2001) andcomparing the results to each other. Both ap-proaches gave very similar results with the latteryielding only a slightly higher value of 0.02 mag.Typical transformations from the R702 filter toGunn r for the Low–LX clusters using the E/S0spectral template by Kinney et al. (1996) gaveR702 obs −Gunn rrest ≈ −0.25.In a next step, the rest–frame properties of thegalaxies were derived. The mean surface bright-ness within Re is defined as:

〈µr〉e = rrest+2.5 log(2π)+5 log(Re)−10 log(1+z),(4.5)

where the parameter rrest denotes rest frameGunn r magnitude. The dimming due to the ex-pansion of the Universe is corrected by the lastterm of this equation. The mean surface bright-ness 〈I〉e in units of L /pc2 is defined as

log〈I〉e = −0.4(〈µr〉e − 26.4). (4.6)

For each cluster, the angular distance d(z) ata specific cluster redshift z was calculated indi-vidually. This yields for A 2390 (z = 0.2280)d = 753.41 Mpc and for CL 0849 (z = 0.2343)d = 769.06 Mpc, CL 1701 (z = 0.2458)d = 797.04 Mpc, CL 1702 (z = 0.2233)d = 741.60 Mpc, respectively. Using the an-gular distance, the angular scale as in units ofkpc/arcsec was determined and the effective ra-


dius in kpc was computed for each galaxy as

log(Re) = log(re) + log(as), (4.7)

where re is the measured effective radius in arc-seconds assuming the concordance cosmology.For Coma (z = 0.024) the angular distance isd = 99.90 Mpc.

4.4 Morphologies

One main aim of this work was to investi-gate possible differences in the properties ofthe two major morphological sub–classes ofearly–type galaxies, elliptical and S0 galaxies.Unfortunately, the resolution provided by theground–based photometry was not sufficient fora morphological classification. However, for asub–sample of early–type galaxies the acquiredHST/WFPC2 and HST/ACS images offer apowerful alternative to perform a detailed inves-tigation of the visual structure of individual gal-axies.In Edwin Hubble’s original galaxy classificationscheme (Hubble 1926), one of the main qual-itative selection criteria was the bulge–to–disc(B/D) ratio. As a combination of bulge and discmodels measure a basic extension of the B/Dparameter, the bulge–to–total fraction (B/T ), itshould be possible to correlate the bulge frac-tion with Hubble type. However, depending onthe adopted bulge profile type (classical de Vau-couleurs law or Sersic r1/n model), the bulgefraction varies. A fundamental question still re-mains: Which profile gives the best correlationfor galaxies spanning a range of different mor-phologies in the Hubble sequence?Morphological studies of distant cluster galaxieswith HST introduced a visual classification ofgalaxies onto the revised Hubble system (Abra-ham et al. 1996b; Smail et al. 1997; van Dokkumet al. 2000). Another approach is the applicationof automated techniques which allow a measure-ment of a quantitative morphology related to butdistinct from the Hubble classification (Schade

et al. 1995; Marleau & Simard 1998; Simard etal. 2002). For the purpose of this work, bothmethods were chosen to yield the most reliableresults.Elliptical and S0 galaxies were morphologicallyclassified in three independent ways. First, a vi-sual classification was performed for a smallersub–set of elliptical and lenticular galaxies (sec-tion 4.4.1). In addition, the results of the sur-face brightness profile fitting (section 4.3) pro-vided a second verification of the visual struc-ture and appearance for the galaxies. Thirdly, ageneral quantitative analysis using an automaticsoftware algorithm (section 4.4.2) was conductedover the whole field covered by the HST imagesfor a large sample of early–type and late–typegalaxies. The information of all three indepen-dent approaches was combined to yield the bestand most reliable decision on the final morpholo-gies.

4.4.1 Visual Classification

A visual classification of the early–type galax-ies was performed by four people separately butfor different sample sizes. All galaxies were in-spected by A. Fritz who additionally classifiedsolely the field galaxies in the WHDF. For com-parison, morphologies of the galaxies in the richcluster A 2390 were obtained by I. Smail and forthe Low–LX cluster galaxies by M. L. Balogh,whereas the field galaxies in the FDF were visu-ally analysed by M. Pannella. The results werecompared to each other and yielded a very goodagreement with very similar morphologies. In allcases, differences were only between transitionstages of one galaxy class, e.g. S0 or S0/Sa. Butthere were never discrepancies between a generalgalaxy type, i.e. an elliptical was never classifiedas an S0 or vice versa.Thumbnail images of the cluster members andfield galaxies are presented in Figs. 4.12 to 4.19.Objects were morphologically classified in twoindependent ways, based on visual inspection aswell as fitting routine output. The findings from


Table 4.9: Noteworthy remarks on individual HST galaxies in Abell 2390.

ID morp comment

2120 S0 signs of isophotal twists out to outer isophote2138 E very extended low-surface brightness (LSB) environment2198 E close to a bright neighbour2237 Sa [O III] 5007 and Hβ emission, asymmetric LSB extension2460 E lensed?, very close to S02592 E next to giant arc2763 E slight twisted low SB2946 S0 small companion2933 Sbc no member, no structural analysis, weird spiral structure, no interaction

the luminosity profile fitting provided a consis-tency check, which resulted in the same clas-sification scheme except for four objects. Inthe cluster A 2390, # 2180: E (visual), S0 (fit)and # 2763: S0/Sa (visual), E (fit), in CL 0849,# A9: S0 (visual), E (fit) and for one field galaxyin the WHDF, # 810b: S0 (visual), E (fit). Ob-jects which are best described with an r1/4-laware classified as an elliptical (E). Those galaxiesfor which an additional exponential componentyields in a slightly improved fit (without a dra-matic change in structural properties) are classi-fied as E/S0. S0 galaxies are best approximatedby a combination of an r1/4-law plus an expo-nential profile.

Remarks on special features on individual HSTobjects of Abell 2390 are listed in Table 4.9.Comments on the Low–LX cluster CL 0849 canbe found in Table 4.10, on the cluster CL 1701in Table 4.11 and on the cluster CL 1702 in Ta-ble 4.12. Morphologies and notes for all field gal-axies in the FDF and in the WHDF are listed inTable 4.13 and in Table 4.14, respectively. Thefirst column of each table shows the galaxy ID,the second indicates the morphology of the ob-jects and the last one the respective comments.The central cD galaxy of a galaxy cluster (i.e.the the Brightest Cluster Galaxy BCG) is de-noted as EcD. In the third column, additional

information about special features regarding tothe galaxy is given. Some notations and ab-breviations which will be used in the followingFigures 4.12 to 4.19 and Tables 4.9 to 4.14 are:background (bg), interaction (int), merger (m),object (obj) and weird (w). Anonymous galax-ies, which are either no cluster members or with-out any spectroscopic information, are classifiedwith an additional letter “A”. For some anony-mous galaxies intermediate–resolution spectrahave been acquired but they are field galaxies,in CL 1701 the galaxies # A77 and # A101 andin CL 1702 the object # A96.


Figure 4.12: 10′′ × 10′′ HST/WFPC2 images of 14 galaxies with available spectroscopic information whichfall within the WFPC2 mosaic of A 2390. The labels give the galaxy ID and the visual morphology as listedin Table 4.9. North is up and East to the left. The colour coding ranges from white (sky-background at∼26.30 mag arcsec−2) to black (≤23.11 mag arcsec−2). Isophotal contours are indicated between the range22.00 ≤ Ci ≤ 24.50 mag arcsec−2, in increments of 0.5 magnitudes.


Figure 4.13: 10′′× 10′′ HST/WFPC2 F702W images of spectroscopic confirmed cluster galaxies of CL 0849which fall within the WFPC2 mosaic. The labels give the galaxy ID and the visual morphology as listedin Table 4.10. Galaxy # 25 is a cluster member based on its low–resolution spectrum and galaxy # 27 fallsbetween two chips. East is up and north to the right.

Figure 4.14: 10′′ × 10′′ HST/WFPC2 F702W images of 8 anonymous galaxies classified as “A” which fallwithin the WFPC2 mosaic of CL 0849. The labels give the galaxy ID and the visual morphology as listed inTable 4.10. East is up and north to the right.


Table 4.10: Noteworthy remarks on individual HST galaxies in CL 0849.

ID morp comment

# A2 SB 2 stars at top# A10 Em merger# A12 Em merger# A14 E between WF3 & WF4 chip# 24 EcD star left# 27 E between WF2 & WF3 chip# 34 Em merger, no SF in spectrum


ID morp comment

# A6 Ew isophotal twists, merger?, or bg spiral, faint# A7 S0 close to chip edge, between WF2 & WF3 chip# A17 E between WF3 & WF4 chip# 62 E small companion, LSB environment# 120 S0 LSB environment# 123 EcDm double nucleus and star close by# 124 SB0 lensed obj in lower right corner?, spiral arms, no int with # 123# 141 SBa 2. bg gal at top, no int# 149 SBm SB merger with E


ID morp comment

# A2 E star close by# A4 Sm merger, weird structure# A10 Sa close to chip edge# 81 EcD star close by# 96 S0 merger?


Figure 4.15: 10′′×10′′ HST/WFPC2 F702W images of 7 spectroscopic confirmed cluster galaxies of CL 1701which fall within the WFPC2 mosaic. The labels give the galaxy ID and the visual morphology as listed inTable 4.11. For the galaxies # 118, # 124, # 141 and # 149 cluster membership is based on low–resolutionspectra. East is up and north to the right.


Figure 4.16: 10′′ × 10′′ HST/WFPC2 F702W images of 10 anonymous galaxies classified as “A” which fallwithin the WFPC2 mosaic of CL 1701. The labels give the galaxy ID and the visual morphology as listedin Table 4.11. For the galaxies # A77 and # A101 spectra are available but they are not cluster members.East is up and north to the right.


Figure 4.17: 10′′×10′′ HST/WFPC2 F702W images of 2 spectroscopic confirmed cluster galaxies of CL 1702.Galaxy # A96 is not a cluster member based on its intermediate–resolution spectrum. The labels give thegalaxy ID and the visual morphology as listed in Table 4.12. Anonymous galaxies are classified as “A”. Eastis up and north to the right.


Figure 4.18: 5′′×5′′ HST/ACS F814W images of 12 FDF galaxies with available spectroscopic information.The labels give the galaxy ID and the visual morphology as listed in Table 4.13.


Figure 4.19: 10′′ × 10′′ HST/ACS F814W images of 10 WHDF galaxies with available spectroscopic infor-mation. South is up and east is to the left. The labels give the galaxy ID and the visual morphology aslisted in Table 4.14.


Table 4.13: Morphologies and noteworthy remarks on individual HST/ACS FDF galaxies.

ID morp comment

1161 S04285 E gal with dust lane, (merger of S0 & S with dust lane?)5011 E5908 E6307 E model too boxy, residuum at core6336 Sa Sb?, weird structure, warps6338 S no ACS image, no analysis, morph based on spec6439 S07116 E hot pixel7459 E no analysis7796 Sa weird structure, SF in upper spiral arm?, warps8372 S0 hot pixel8626 E/S0 close to edge of chip

Table 4.14: Morphologies and noteworthy remarks on individual HST/ACS WHDF galaxies.

ID morp comment

92 S0 2 bg? obj in north111 E/S0 external halo in upper north part, 2 obj in north-west158 Sa173 S0318 E close to star, star in center437 E face on, faint star in north-east, [O II] 3727 emission508 Sa weird, dust pattern in south-west?, rotation?, [O II] 3727 emission?, star in west749 E/S0 no ACS image, no analysis, morph based on spec810 E boxy isophotes, stars in center, [O II] 3727 emission810b E discy structure, faint, [O II] 3727 emission946 S0/Sa SB0?, spiral arm in north-east, int with obj in south-east?


4.4.2 Quantitative Analysis

A very popular and often used algorithm foranalysing two–dimensional surface brightnessprofiles of galaxies is the Galaxy IMage Two–Dimensional (GIM2D, see Simard et al. 2002)package which operates within the IRAF envi-ronment5. GIM2D is a software package in Sub-set Pre–Processor (SPP) language using variousprocedure calls of the IRAF operating system toperform detailed bulge and disc decompositionsof low signal–to–noise images of distant galax-ies. The program derives a maximum of up to12 different structural parameters for each inputgalaxy.A detailed morphological analysis for the Low–LX cluster sample down to R702 = 23.0m wasperformed with the GIM2D software algorithmon the HST/F702W images within our collab-oration (Balogh et al. 2002a). For the central∼0.6 Mpc of these clusters which are coveredby the HST imaging, a large fraction of disc–dominated galaxies (36%) with low B/T ∼ 0was found, whereas high X-ray–luminous galaxyclusters are characterised by galaxies with inter-mediate bulge–to–total fractions of B/T ∼ 0.4.However, only 18±5% of these Low–LX galaxypopulations were revealed to be spectroscopi-cally confirmed cluster members brighter thanR702 = 20.0m.To ensure a uniform measurement of visual ap-pearances and to quantify the results of themorphological visual classification between richand poor clusters, two verifications have beenemployed. A first consistency check of B/Tmeasurements between different filters was per-formed. In addition, the correlation betweenB/T and visual morphological type was estab-lished. In the following, these verifications arediscussed in detail.In November 2004, the author initiated a projectwith M. Pannella and Dr. R. P. Saglia (bothMPE/Munich) to investigate in detail galaxy

5GIM2D: http://www.hia-iha.nrc-cnrc.gc.ca\STAFF\lsd\gim2d\

structural parameters using various combina-tions of surface brightness profiles and differentfitting techniques. A comparison of these inde-pendent methods will reveal how the colour gra-dients depend on the computed structural pa-rameters. Structural properties of an initial sam-ple of 16 early-type galaxies in A 2218 in theHST/ACS BV I filters were analysed using acombination between a classical de Vaucouleurs(r1/4) bulge with an exponential disc compo-nent in GIM2D. In addition, the same galaxieswere investigated with the algorithm by Sagliaet al. (1997a) on the HST/WFPC2 BV I im-ages. Both approaches utilised the same lumi-nosity profile type (r1/4+exp. disc) which wasapplied to model the surface brightness distribu-tion of the galaxies (see section 4.3.2). For thisreason, the GIM2D and Saglia et al. (1997a) pa-rameters could be directly compared to the re-sults of surface brightness modelling. Througha cross–correlation of all three filter passbands,the B/T ratios for these galaxies in the clusterA 2218 were derived. In Fig. 4.20, the bulge–to–total ratios for the 16 early-type cluster gal-axies of A 2218 down to a magnitude limit ofI814 = 21.20m are shown. Note that all early-type galaxies are spectroscopically verified clus-ter members, which were selected based upon thestudy of Ziegler et al. (2001a). The E+S0 clus-ter galaxies of A 2218 cover a range in apparentmagnitudes of 15.68 < R702 < 18.68 with a me-dian value of 〈R702〉 = 17.67. The dotted linein Fig. 4.20 indicates the adopted break betweenbulge B/T ≥ 0.4 and disc B/T < 0.4 dominatedgalaxies.

From Fig. 4.20 it is clearly visible that thesegalaxies are bulge–dominated (B/T ≥ 0.4)systems with a median in the I814 filter of〈B/T814〉 = 0.69. Early-type galaxies encompasssub–classes from elliptical over S0 to Sa, split-ted into 8 E, 1 E/S0, 4 S0, 2 SB0/a and 1 Sa.Apart from one galaxy which was not includedon the WFPC2 frames (# 2738 which is an ellip-tical galaxy based on ground–based Hale I image


Figure 4.20: Bulge–to–total ratios of 16 early-type cluster galaxies brighter than R702 = 18.68m inA 2218. The dotted line denotes the adopted breakbetween bulge B/T ≥ 0.4 and disc B/T < 0.4 do-minated galaxies. All early-type cluster galaxies arebulge–dominated systems.

of A 2218), all galaxies were analysed via the FP,see discussion in section 6.4. For the V606 andB450 similar values to the I814 filter are derived〈B/T606〉 = 0.74 and 〈B/T450〉 = 0.61. Compar-ing the B/T measurements and other structuralparameters in all three filter passbands, onlyslight differences in their general distributionsare found. In particular, the results in V606 andI814 are in very good agreement and the distribu-tions do not dependent strongly on the assumedfilter. The uncertainties in the B/T measure-ments were assessed through Monte Carlo sim-ulations in the GIM2D software. Typical meanerrors in the B/T ratios are between 0.04 to 0.07with a median value of only 0.02 between allBV I filter passbands. The B/T ratios for theA 2218 galaxies are also in good agreement withthe GIM2D B/T determinations of A 2218 in theR702 passband by Balogh et al. (2002a). For gal-axies down to a magnitude limit of R702 = 22.1m

within the HST image, these authors derive amean of 〈B/T 〉 ∼ 0.60. To account for contam-ination by field galaxies, this study performeda statistical background correction, very simi-

Figure 4.21: Bulge–to–total ratios of 26 early-type cluster members brighter than R702 = 19.48m inthree Low–LX clusters. The shaded histogram givesthe distribution of early–type cluster members whichwere selected for the FP study.

lar to the method as described in section 2.1.5.Although the galaxies were not selected uponphotometric or spectroscopic membership, theB/T distribution is very similar to the one inFig. 4.20.

Fig. 4.21 displays the bulge–to–total ratiosof early–type cluster members brighter thanR702 = 19.48m in the three Low–LX clustersCL 0849, CL 1701 and CL 1702. This analysisis restricted to spectroscopic verified cluster gal-axies with available B/T measurements in theR702 filter only. One sample contains 17 clustermembers based on the low–resolution study byBalogh et al. (2002b), which are indicated by theopen B/T distribution in Fig. 4.21. A sample of9 Low–LX cluster galaxies (for the galaxy # 130no B/T could be derived), which were selectedfor an investigation within the FP, are denotedby the shaded B/T histogram. The Low–LXFP galaxies show a spread in apparent magni-tudes of 16.12 < R702 < 19.48 (〈R702〉 = 18.15).The B/T distribution of these galaxies is char-acterised by higher B/T ratios with a medianof 〈B/T 〉 = 0.71 than the low–resolution samplewith 〈B/T 〉 = 0.65. This is to be expected, as


Figure 4.22: Bulge–to–total fractions as a func-tion of visual morphological Hubble type of 16 early-type cluster members brighter than R702 = 18.68m

in A 2218. B/T is derived using a de Vaucouleursbulge plus an exponential disc. The dotted line de-notes the adopted break between bulge B/T ≥ 0.4and disc B/T < 0.4 dominated galaxies. The derivedB/T corresponds well to visual morphologies.

the FP sample was selected to have early–typemorphologies yielding on average to higher B/Tratios. The outlier galaxy # 923 has the lowestB/T ratio with B/T = 0.12, which is the sec-ondary galaxy of an ongoing merger in the cDgalaxy # 123 of the cluster CL 1701. As the B/Tvalues and their distributions in different filters(B450, V606 and I814) yield consistent results andno strong variations were found, a comparisonwith the B/T ratios for the Low–LX clusters inthe R702 filter is allowed.In comparison to the distribution of rich clustergalaxies of A 2218 in Fig. 4.20, the B/T measure-ments of the Low–LX clusters are slightly differ-ently distributed. The poor cluster galaxies showa broader distribution and more disc–dominatedgalaxies with lower B/T ratios in the R702 band.This first analysis confirms that the Low–LX FPsample is restricted to bulge–dominated galaxiesand that the same class of early–type galaxieswill be compared when analysing the FP for richand poor clusters.To test in even greater detail if the B/T distri-

Figure 4.23: Bulge–to–total fractions as a func-tion of visual morphological Hubble type of 21 early-type cluster members brighter than R702 = 19.48m

in three Low–LX clusters. B/T is derived using ade Vaucouleurs bulge plus an exponential disc. Thedotted line denotes the adopted break between bulgeB/T ≥ 0.4 and disc B/T < 0.4 dominated galaxies,the dashed line is a least-squares fit to all galaxieswith B/T > 0.2. The derived B/T corresponds wellto visual morphologies.

butions comprise the same populations of bulge–dominated galaxies and to separate between theclasses of elliptical and S0 galaxies, a correla-tion between measured B/T and visual classi-fied morphology was derived. Fig. 4.22 showsthe B/T fractions as a function of visual mor-phological type for 16 early-type cluster mem-bers brighter than R702 = 18.68m in A 2218. InFig. 4.23, B/T versus visual morphological typeis displayed for the Low–LX sample. All galaxieswere classified according to the de Vaucouleurs’classification of the revised Hubble scheme (deVaucouleurs 1959). Visual morphological typesare represented as E (−5), cD (−4), E/S0 (−3),S0 (−2), SB0/a or SB (0) and Sa (1). The dashedline is a least-squares fit to all galaxies withB/T > 0.2, T = −7.09 ·B/T + 1.63, with T be-ing the morphological Hubble type. In Figs. 4.22and 4.23, elliptical and S0s galaxies are denotedas filled and open circles, while early–type spiralgalaxies of SB0/a and Sa are indicated as trian-


gles and squares. The adopted break betweenbulge (B/T ≥ 0.4) and disc (B/T < 0.4) dom-inated galaxies is visualised by the dotted line.There is a large scatter and only a weak trendbetween the B/T and visual type for the de Vau-couleurs bulge+exponential disc. An explana-tion is that the SB0/a and Sa types have a strongbulge component and therefore a high B/T ra-tio. With the de Vaucouleurs bulge+exponentialdisc profile, for the Low–LX E+S0 galaxies aweak correlation between the B/T and and vi-sual morphological type is found, albeit with alarge scatter. The trend between de Vaucouleursbulge+exponential disc and type is not surpris-ing. First, the surface brightness distribution ofbright ellipticals and the bulges of early–type spi-rals are well–fitted by a de Vaucouleurs profile(Andredakis et al. 1995). Second, the majorityof cluster galaxies are bulge–dominated systems.The lack of a stronger correlation between B/Twith visual classified morphology in Figs. 4.22and 4.23 is partly due to a low number of spiralgalaxies in both galaxy samples. However, apartfrom the found trend, there is a large scatter inthe distribution of B/T ratio versus visual typeand it is difficult to distinguish between early–type spirals from elliptical and S0 galaxies usingthe B/T parameter alone.In the next section details on the derivation ofthe absolute magnitudes will be given.

4.5 Luminosity Derivation

Rest–frame absolute magnitudes were calculatedfrom the deep ground–based multi–band imagingdata. For each project, ground-based imagingin various filter passbands was acquired whichturned out to be especially beneficial for the fieldstudies in the FDF and WHDF. The apparentbrightness of a given field galaxy was derivedfrom the filterX which best matched the B-bandresponse curve in rest–frame. According to theredshift of the respective galaxy, this filter X waseither the B-, g-, R- or I-band (see section 4.5.3).

For the E+S0 galaxies the adopted equation forthe computation of absolute Y magnitudes was

MY = mX −AX −DMΛ(z,H0)− kY (X,T, z) , (4.8)

where MY denotes either the absolute B magni-tude (field galaxies) or absolute Gunn r magni-tude (cluster galaxies).mX is the apparent total brightness in the pass-band X as derived with the Source Extractor(see section 4.5.1).AX is the Galactic absorption in the wavelengthregime of filter X (see section 4.5.2).DMΛ(z,H0) is the distance modulus for the re-spective redshift of a cluster or field object. Asalready stated earlier, throughout this thesis theso–called concordance cosmology (e.g., Spergelet al. 2003) is adopted, i.e. a flat universewith a matter density Ωm = 0.3, a cosmolog-ical constant corresponding to an energy den-sity of ΩΛ = 0.7 and a Hubble constant ofH0 = 70 km s−1 Mpc−1. In most previous stud-ies also a non-zero cosmological constant was as-sumed. Possible deviations caused due to dif-ferent values of H0 were accounted for when acomparison with the literature was performed.kY (X,T, z) is the k–correction for the transfor-mation from filter X to the rest–frame Y -bandwhich accounts for three effects: (i) the differentresponse curves of filter X in observer’s frameand filter Y (either B or Gunn r) in rest–frame,(ii) for the redshifted and “stretched” SED inobserver’s frame and (iii) the SED type T . Forall cases, a SED type T = −5/− 2, i.e. an early-type E/S0 galaxy was adopted. The individualcorrection factors kY have been derived via syn-thetic photometry, which is described in detailin section 4.5.3.In a second independent approach, for the sub-sample of galaxies with HST structural informa-tion, the rest–frame absolute magnitudes werecomputed using the transformation procedure asdiscussed in section 4.3.4.


Table 4.15: Galactic extinction coefficients for the poor cluster fields.

Name R.A. Dec. AB BH AB SFD E(B − V ) AF702 AR AI[mag] [mag] [mag] [mag] [mag] [mag]

CL 0849+37 08 49 11 +37 31 09 0.090 0.151 0.035 0.085 0.094 0.068CL 1701+64 17 01 47 +64 20 57 0.060 0.112 0.026 0.063 0.070 0.050CL 1702+64 17 02 14 +64 19 53 0.060 0.110 0.026 0.063 0.068 0.050

4.5.1 Apparent Magnitudes

Total apparent magnitudes were derived withthe Source Extractor package (SExtractor,Bertin & Arnouts 1996). This program offersdifferent algorithms for photometry of stellar ob-jects and extended sources and a choice betweencircular or elliptical apertures. Apparent mag-nitudes measured within fixed circular aperturesare the best approach for the derivation of galaxycolours. Total apparent magnitudes are not reli-able to compute galaxy colour indices as the ap-parent diameter of the galaxies could deviate be-tween different passband filters. For the deriva-tion of absolute magnitudes it is mandatory toprevent flux losses. For this purpose, variableelliptical apertures yield the best results.

SExtractor offers three algorithms for the usageof elliptical apertures, called Mag iso, Mag bestand Mag auto. Tests showed that the lattertwo routines are consistent to within a few0.01 mag for early-type galaxies in all bandpassesand for different field and cluster environments.However, in particular for faint detections, theMag iso algorithm systematically resulted in un-derestimated brightnesses. The Mag auto auto-matic aperture routine is based on the “first mo-ment” algorithm by Kron (1980) and designed tobest reproduce the total magnitudes of extendedsources. This algorithm has been used for bothfield studies whereas for the cluster galaxies theMag best routine was applied.

4.5.2 Galactic Absorption

The correction for the extinction due to Galac-tic absorption was performed in using a specialprogram written by the author, called SFDextc.This program calculates automatically the ex-tinction coefficients for a specific E(B−V ) inputvalue based on the COBE dust maps by Schlegelet al. (1998, hereafter SFD). As a consistencycheck, results were compared to the Galactic ex-tinction values by Burstein & Heiles (1982, here-after BH) and AV = 3.1 · E(B − V ). As thedust maps provide a more reliable measurementonly the SFD extinction values were adopted.For comparison, for the Low–LX clusters alsothe AB coefficients by BH are shown (see Ta-ble 4.15). The uncertainties in the extinctionvalues are E(B−V) = 0.010m.For the cluster A 2390 an E(B − V ) = 0.110m

was assumed, resulting in extinction coefficientsfor the Johnson-Cousins filters of AU = 0.600m,AB = 0.476m and AI = 0.214m, respec-tively. For the HST/WFPC2 F814W filterA814 = 0.214m was derived.The Galactic extinction coefficients for theJohnson-Cousins and the HST F702W filterpassbands of the Low–LX cluster fields are sum-marised in Table 4.15.For the Galactic absorption at the coordinatesof the FDF, the values which are given inHeidt et al. (2003) were adopted. Thesecoefficients were derived with the formulae givenin Cardelli et al. (1989) under the assumptionof E(B − V ) = 0.018m (derived from BH)and AV = 3.1 · E(B − V ). For the individual


broadband filters used for the FDF imaging,the resulting respective absorption factors areAU = 0.087m, AB = 0.076m, Ag = 0.062m,AR = 0.041m and AI = 0.035m. For theHST/ACS F814W filter A814 = 0.035m was com-puted. At the central coordinates of the WHDFof α2000=00h 19m 59.6s, δ2000=+00 04′ 18′′

(B1950.0), an E(B − V ) = 0.025m was adopted,yielding to Galactic extinction coefficients ofAU = 0.137m, AB = 0.108m, AR = 0.067m, andAI = 0.049m, AH = 0.014m, and AK = 0.009m.The absorption coefficient for the HST/ACSF814W filter is A814 = 0.049m.

4.5.3 K-Correction

The k-correction accounts for the fact thatsources which are observed at different redshiftsare compared with each other at different rest-frame wavelengths. The transformation betweenan observed apparent magnitude in a photomet-ric broad passband X to a corresponding rest–frame magnitude in another broad-band photo-metric passband Y involves the term known asthe “k-correction”. Such a transformation is de-fined by

kY (X,T, z) = X(T, z)obs − Y (T, 0)rest (4.9)

If photometric broad-band information is lim-ited to a few filters, the different wavelengthranges covered by a passband in observer’s frameand rest–frame introduce a strong dependenceof the achievable k-correction accuracy on SEDtype. At redshift z = 0.2, e.g., the differencebetween SED types E/S0 and Sbc correspondsto a change of ∆kB = 0.4m in the transfor-mation from Bobs to Brest (e.g., Frei & Gunn1994). For this reason, even a slight misclas-sification can introduce a substantial offset inthe derived luminosity if observations are lim-ited to one or two filters only. In contrast tothis, the field galaxy projects of the FDF andWHDF greatly benefitted from the multi–bandimaging data. With the FDF photometry in B,g, R and I, and the WHDF photometry in B,

Figure 4.24: Transmission curves of the FORS U ,B, g, R and I filters which were used for the imagingof the FDF.

R and I, the filter that best matched the rest–frame B-band could be used to transform an ap-parent magnitude X into absolute magnitudesMB. For both, rich and poor early-type clustersamples, observations were carried out in the Ror I filter bands. At z = 0.2, the observed R,I and the R702, I814 HST filter passbands (cf.section 4.3.4) are very close to rest-frame Gunnr. For the clusters CL 0849 (z = 0.234), CL 1701(z = 0.246) and CL 1702 (z = 0.223) the trans-formation Robs → rrest amounts to 0.02m only.Note that these numbers are less than the cor-rection term for Galactic extinction. Therefore,the advantages of using the Gunn r-band as rest-frame passband instead of the bluer Johnson Vor B bands are the smaller k-corrections and thelower corrections of galactic absorption.In order to ensure the highest possible accu-racy in the transformation between observed–frame and rest–frame, the kY -corrections werenot taken from the literature but computedvia synthetic photometry as outlined in Bohm(2003). For a given SED of type T at redshiftz with wavelength–dependent flux F (T, λ), themagnitude in a passband X was defined as

mX(T, z) = −2.5 log∫F (T, z, λ)SX(λ) dλ∫

SX(λ) dλ


Figure 4.25: Transmission curves of the Johnson UJ ,BJ , VJ and Cousins RC , IC filters.

+ CX + Csys (4.10)

(e.g., Fritze-v. Alvensleben 1989). The par-ameter SX(λ) is the response function of filterX. CX is a constant for the calibration of therespective filter within the corresponding filtersystem (e.g, the Johnson, Cousins or Gunn filtersystem). Csys is an optional constant to calibratebetween different filter systems. The derivationof the calibration factors CX and Csys is dis-cussed below. Note that Eq. 4.10 yields rela-tive magnitudes which are appropriate for thetransformation between different filters or differ-ent systems, but is not reliable for the compu-tation of physical apparent brightnesses in theform given above. For the latter purpose, anadditional calibration constant is needed, whichcan be derived via synthetic photometry withan observed SED of a galaxy with known ap-parent brightness. However, since the syntheticphotometry only was executed to derive the k-corrections, i.e., for relative photometry, this cal-ibration was not needed.In Fig. 4.24, the FORS filters U , B, R, I andGunn g are shown. This set of filters was usedfor the imaging of the FDF in the optical regime.In Fig. 4.25, the filter response curves of thestandard Johnson-Cousins system (UJ , BJ , VJ ,RC , IC) are displayed. These curves were re-

Figure 4.26: Transmission curves of the Thuan-Gunn u, g and r filters.

trieved as ascii tables from the ESO webpages.Fig. 4.26 shows the filter response curves of theThuan-Gunn filter system (u, g, r) which wereretrieved as ascii tables from the CAHA instru-ments webpages. All these filter curves were usedfor the definition of the respective response func-tions SX .For the Johnson system, the constants CX weretaken from Allen (1973). They are defined insuch a way to yield zero colour terms for anA0V star spectrum in all passbands. To cal-ibrate the FORS filter system, the A0V spec-trum from Pickles (1998) was used and the fine–tuning of the respective factors CX was per-formed with the equations given in Sinachopou-los & van Dessel (1998). Finally, the constantCsys for transformations between the FORS andthe Johnson system was calculated from thecolour (Vfors − VJ) (ibid.). For the CousinsRC and IC filters, the factors CR and CI ofthe FORS calibration were adopted as initialvalues. The calibration factors were verifiedthrough a comparison of the (RJ − RC) and(IJ−IC) colours, assessed for various SED types,to the numbers published in Fukugita et al.(1995), which yielded consistent values within<0.03 mag.Since the FORS g filter has a slightly differentshape from the original Thuan-Gunn g filter def-


Figure 4.27: SED template used for the compu-tation of the k-corrections via synthetic photometry.This is an observed spectrum of an E/S0 galaxy withan age of 12 Gyrs (Kinney et al. 1996). For thisplot, the template has been normalised to the flux at5500 A.

inition, the following more complicated calibra-tion procedure has to be performed. The FORSg filter is broader by ∼ 90 A and has a ∼ 130 Ahigher central wavelength than the Thuan-Gunng filter. A first calibration was conducted adopt-ing the transformation by Jørgensen (1994)

ggunn − VJ = 0.503 (BJ − VJ)− 0.226. (4.11)

Based on the A0V spectrum, the colour (gfors −ggunn) then was computed, which yielded a valueof 0.10 mag and thus a difference to the Johnsonsystem of (gfors(A0V ) − VJ(A0V )) = −0.13m.Again, the colours were compared through across–correlation with the g passband definitionsgiven by Fukugita et al.As SED templates for the computation of the k-corrections, the UV/optical spectra published byKinney et al. (1996) were used. These observedtemplate spectra comprise different morphologi-cal galaxy types of elliptical, bulge, S0, Sa, Sb,and Sc and starburst galaxies and cover a wave-length range of ∼1200–9800 A with a resolutionbetween ∼8 to 10 A. For the synthetic photom-etry described here, only elliptical galaxy tem-plates with an age of 12 Gyrs were used solely.One of these templates is shown in Fig. 4.27.For the purpose of this graph, the spectrum wasnormalised to the flux at 5500 A. No separation

Figure 4.28: SED template used for the com-putation of the k-corrections via synthetic photom-etry. This spectrum was generated via chemicallyconsistent evolutionary synthesis models (Moller etal. 2001). For this plot, the template has been nor-malised to the flux at 5500 A.

between elliptical and S0 types was done be-cause both offer similar (optical and UV) spec-tral shapes, Balmer discontinuities (4000 A) andabsorption features for λ > 5000 A which areindistinguishable from their broadband opticaland near-IR colours.For internal consistency, the synthetic spectralSED templates by Moller et al. (2001) werealso applied to compute the k-corrections. Thesetemplates were generated with evolutionary syn-thesis models and the authors provide two ver-sions of each template, one including dust red-dening and one without dust. For the syntheticphotometry performed here, templates withoutdust were used solely, since no intrinsic absorp-tion has to be accounted for early-type galaxiesin this wavelength regime. The basic advantageof synthetic spectra with respect to observedspectra is of course that no noise is introducedin the photometry. Elliptical galaxy templateswith an age of 12 Gyrs were used in order to givea good reproduction of the observed colors ofpresent–day early-type galaxies. One E/S0 tem-plate is shown in Fig. 4.28, which was normalisedto the flux at 5500 A for plotting purposes. Gasemission was not incorporated in the models, butthis has a negligible effect on broad band pho-tometry for very late–type SEDs (T ≥ 8) only.To derive the flux F at wavelength λ, the SEDs


were redshifted according to

F (T, z, λ) =F0(T, λ/(1 + z))

(1 + z)(4.12)

(e.g., Contardo, Steinmetz & Fritze-v. Alvens-leben 1998), where F0 is the flux of the un–redshifted spectrum. Finally, the transforma-tion from the apparent magnitude of a spectrumof type T (E/S0) at redshift z observed with aFORS/Johnson-Cousins filter X to the JohnsonB or Gunn r magnitude mY in rest–frame wasperformed via

kY (X,T, z) = mX(T, z)−mY (T, z = 0), (4.13)

where the two terms on the right hand side werederived according to Eq. 4.10. k-corrections werecomputed with the equation given above for allsource filters (FORS and Johnson-Cousins filtersystem), redshifts (cluster redshift or redshift ofthe field galaxy) and environments covered bythe early-type galaxies. Typical deviations in thek-corrections between the templates of Kinneyet al. (1996) and Moller et al. (2001) are verysmall, e.g. ∆kr = 0.03m in the Gunn r filter, forthe SEDs of both ellipticals and S0 galaxies.To transform to Johnson B rest–frame (cf.Eq. 4.8), the following filters were used depend-ing on the redshift: Bfors for z < 0.25, gfors for0.25 ≤ z < 0.55 and Rfors for 0.55 ≤ z < 0.7(only one object). For this reason, the k-correction was much less sensitive to spectraltype than using a global transformation ofBobs → Brest, which was particularly importantat higher redshifts z >∼ 0.6 in the case of spi-ral galaxies with late–type SEDs (T ≥ 1) inthe FDF (Bohm 2003). An analogous approachwas performed for the WHDF elliptical galaxies.For field objects with z > 0.25 a transforma-tion to B rest–frame was conducted based onthe B passband and for 0.55 ≤ z < 0.75 theRC filter was utilised. Nevertheless, for internalconsistency the rest-frame magnitudes derivedwith the FORS filter systems were checked bytransforming the observed F814W-magnitudes

into rest-frame Johnson-B. The deviations be-tween magnitudes were small (differences wereless than their errors).For testing purposes, the colors of the templateswere derived purely within the Johnson-CousinsFilter system and afterwards compared to thevalues given by Fukugita et al. The deviationswere very small, with absolute colors (X − Y )within the range 0.03m ≤ |∆(X − Y )| ≤ 0.08m.

4.6 Luminosity Distributionand Errors

The distribution of the early-type galaxies inthe cluster A 2390 in absolute Gunn r-bandmagnitude is shown in Fig. 4.29. A range of−22.99 ≤ Mr ≤ −20.47 (〈Mr〉 = −21.31)is covered by the 48 E+S0 cluster members ofA 2390. The dashed arrow gives the mean value(M r = −21.48) of the distribution and the me-dian is displayed by the solid arrow. The FPsub–sample of 14 galaxies with high–resolutionstructural HST information, shows a spread−22.56 ≤ Mr ≤ −20.35 (〈Mr〉 = −21.21).Note that the central cD galaxy of A 2390 (i.e.the Brightest Cluster Galaxy BCG) was notspectroscopically observed. Including a set of 4non–cluster members with determined redshiftswhich were either identified as foreground or asbackground objects (see section 5.5 for a dis-cussion) to the cluster sample of A 2390 yieldsa distribution with a range in absolute Gunnr magnitudes of −22.99 ≤ Mr ≤ −20.16(〈Mr〉 = −21.28). These non–cluster galaxiesaffect only the fainter end of the distribution ofabsolute magnitudes.Note that the distributions of the complete clus-ter samples are taken from the ground-basedimaging, whereas the FP sub–sample of galax-ies are based on the HST magnitudes. In gen-eral, the HST photometry using the techniquesof curves of growth fitting is deeper and there-fore fainter magnitudes can be derived than ifmeasurements are restricted to a circularly aper-


Figure 4.29: Absolute Gunn r-band magnitude dis-tribution of all 52 cluster candidates in A 2390 withdetermined redshifts (dashed line), the sub–sample of48 early-type cluster members of A 2390 from whichvelocity dispersions could be derived (solid line) andthe subset of 14 E+S0 galaxies with HST/WFPC2 in-formation (shaded region). The dashed and the solidarrow give the mean (Mr = −21.48) and the me-dian (〈Mr〉 = −21.31) values of the cluster memberdistribution, respectively.

ture. For this reason, small differences betweenthe histograms in the order of ∼0.05 mag mightarise and cause a slightly different distribution.In Fig. 4.30 the distribution of the early-typegalaxies in the three Low–LX clusters CL 0849,CL 1701 and CL 1702 in absolute Gunn r-bandmagnitude is displayed. The 27 E+S0 clustermembers (combined for all Low–LX clusters),cover a range of −23.74 ≤ Mr ≤ −20.08(〈Mr〉 = −21.77). Again, the dashed ar-row indicates the mean value (M r = −21.69)of the distribution and the median is dis-played by the solid arrow. An FP sub–sampleof ten galaxies is distributed over a range of−23.88 ≤ Mr ≤ −20.68 (〈Mr〉 = −21.85).Two cD galaxies (of the clusters CL 0849 andCL 1702) are included in both samples of clustermembers and FP galaxies.A set of additional 9 non–cluster members withdetermined redshifts which revealed to be ei-ther foreground or background objects (cf. sec-

Figure 4.30: Absolute Gunn r-band magnitude dis-tribution of all 36 cluster candidates in the Low–LXclusters CL 0849, CL 1701 and CL 1702 with deter-mined redshifts (dashed line), the sub–sample of 27early-type cluster members for these three clusterswith available velocity dispersions (solid line) and thesubset of 10 E+S0 galaxies with HST/WFPC2 infor-mation (shaded region). The dashed and the solidarrow give the mean (Mr = −21.69) and the me-dian (〈Mr〉 = −21.77) values of the cluster memberdistribution, respectively.

tion 5.5) is included for the distribution of clus-ter candidates for the Low–LX galaxies (dashedline). These 36 galaxies cover a range in Gunnr magnitudes of −23.74 ≤ Mr ≤ −19.83(〈Mr〉 = −21.85).

In Fig. 4.31 the distribution of the early-typefield galaxies in the FDF and WHDF in ab-solute Johnson B-band magnitude are shown.Only the rest–frame magnitudes derived withthe FDF and WHDF photometry are shown (seealso discussion in section 4.5.3). The 23 E+S0field galaxies cover a range in absolute B-bandmagnitude of −22.78 ≤ MB ≤ −19.40(〈MB〉 = −20.60). The dashed arrow corre-sponds to the mean value (MB = −20.74),whereas the median is denoted by the solid ar-row.

Total errors on the absolute Gunn r and B-bandmagnitudes for all galaxies were computed via


Figure 4.31: Absolute B-band magnitude distribu-tion of 23 early-type field galaxies in the FDF andWHDF. Rest–frame magnitudes were derived basedon the FDF and WHDF photometry. The dashedand the solid arrow give the mean (MB = −20.74)and the median (〈MB〉 = −20.60) values of thedistribution, respectively.

the relation

δM2Y = δX2 + δk2 + δA2

X . (4.14)

where δMY is the total error in the absolute rest–frame of Gunn r passband δMr or in the absoluteJohnson B passband δMB.δX indicates the photometric error of ground-based photometry in the respective filter pass-bands, either B, g, R or I-band. For the clus-ter galaxies in A 2390 the total I-band magni-tude was adopted and for the Low–LX clustersthe total R-band magnitude was utilised. Theerrors in total I-band magnitudes δI cover arange between 0.02m ≤ δI ≤ 0.09m, with amedian uncertainty of 〈δI〉 = 0.04m. Errorsin total R-band magnitudes δR take values of0.01m ≤ δ I ≤ 0.05m, with a median of〈δI〉 = 0.03m. Thanks to the very deep imagingof the FDF and deep WHDF, the uncertaintiesδB for the field E+S0 galaxies are very smallfor all filter passbands. The median values are〈δX〉 < 0.015m in B, g, R and I, respectively.δk denotes the k-correction error. The uncer-tainty between the SED templates is very small,

a typical error of δkr = 0.03m was found. Forall morphological types (E, E/S0, S0, Sa bulge)and redshifts of the field galaxies, the error inthe k-correction is δkB ≤ 0.1m. No error due toSED type misclassification was assumed as theSEDs of both ellipticals and S0 galaxies are verysimilar (section 4.5.3).Finally, δAX in Eq. 4.14 gives the uncertainty ofthe galactic absorption correction.Contributions of the distance modulus to δMY

were assumed to be negligible. For the A 2390sample, the total error in δMr falls into therange 0.07m ≤ δMr ≤ 0.14m, with a medianof δMr = 0.09m. For the Low–LX sample,the total uncertainty in δMr covers a range be-tween 0.06m ≤ δMr ≤ 0.12m, with a medianerror of only δMr = 0.07m. For the FDF andWHDF field sample, the total error falls into therange 0.08m ≤ δMB ≤ 0.20m with a median of0.15 mag.

Chapter 5: Kinematic Analysis 93

Chapter 5

Kinematic Analysis

5.1 Galaxy Kinematics

An extracted galaxy spectrum is the superposi-tion of each individual star spectrum over thisselected region of the galaxy along the line-of-sight. Because of the random motions of thestars in early-type galaxies, the stellar spectraare slightly shifted to each other in wavelengthrange. The resulting net effect is a broadeningin the galaxy spectrum owing to these shifts. Inother words, the galaxy spectrum is the convo-lution of the star spectra with an appropriatevelocity distribution along the line-of-sight.In order to quantify this broadening, the Line-Of-Sight-Velocity-Distribution (LOSVD) F (vlos)is characterised as the fraction of stars contribut-ing to the spectrum with a line-of-sight betweenvlos and (vlos+d vlos). Under the assumption thatall the stellar spectra are identical (S(u)), withu being the stellar velocity, the galaxy spectrumis given as:

G(u) ≈∫F (vlos) · S(u− vlos) · dvlos (5.1)

This is one of the basic equations in the fieldof galaxy kinematics. To derive the kinematicsof one particular galaxy a spectrum of a stel-lar template is needed, which is required to givea good approximation of the galaxy’s spectrumand its dominant stellar population. The betterthe selection of the template star, the more accu-rate the fit to the galaxy spectrum will get andthus the more precise the kinematics of the gal-

axy under consideration. Two main propertiesof the LOSVD are its mean velocity v:

v =∫F (vlos) · vlos · dvlos (5.2)

and its velocity dispersion σ2los:

σ2los =

∫F (vlos) · (vlos − vlos)2 · dvlos (5.3)

As an initial approximation, the LOSVD of thestars is assumed to be Gaussian. However, dueto improvements of the detectors and new de-veloped methods it is possible to measure thedeviations from the Gaussian shape. In the fol-lowing paragraph, a review of the most relevantalgorithms from the literature is given.

5.1.1 Kinematic Extraction Methods

All techniques described here benefit from thecircumstance that the convolution show inEq. 5.1 becomes a multiplication in Fourier do-main. The deconvolution therefore involves di-viding the Fourier transform of the galaxy spec-tra by the Fourier transform of the stellar tem-plate. However, noise makes such a directcomputation impractical. Noise dominates theFourier transformed spectra at high frequencies,hence the ratio of the transforms at these fre-quency components will consist mainly of veryamplified noise. Generally, the noise is white,i.e. its transform is essentially a constant outto high frequencies. To derive robust results, it

94 Chapter 5: Kinematic Analysis

is therefore necessary to suppress the noise athigh frequencies through filtering. For exam-ple, “Wiener” or “optimal filtering” (Brault &White 1971) provides one of the best balancingbetween filtering insufficiently, resulting in moreresidual noise than necessary and filtering tooseverely, thereby losing important information.Fig. 5.1 illustrates the effect of Wiener filteringof the Fourier transform. The Wiener filter isconstructed by a least-square fitting to the com-puted power spectrum Pobs. The complex sig-nal and noise power spectra are approximatedby a model of the signal power P ′signal and amodel of the noise power P ′noise as Pobs(κ) =P ′signal(κ) + P ′noise(κ) which represents a goodapproximation of the observed power spectrumPobs. The modelled signal and noise componentsare used to constructed the filter, which is in-dicated as the solid line in Fig. 5.1. Workingin ‘Fourier space’ offers two great advantages.On the one hand, the LOSVD is easier to ex-tract and to analyse. On the other hand, a fil-tering of residual low-order and high-order fre-quency components in the original spectrum ispossible which arise from an imperfect contin-uum subtraction and Poisson noise. An alterna-tive approach to fit the LOSVD by a direct least-squares solution in ‘Real space’ and describingthe velocity dispersion by a set of delta-functionsin logarithmic wavelength space has been de-veloped by Rix & White (1992). To overcomethe amplification of high frequency noise in theLOSVD, a method of weighting the eigenvaluesis used. However, such computer codes (Rix &White 1992; van der Marel & Franx 1993) willnot be discussed here in greater detail.

• Classical Fourier Quotient Method(Sargent et al. 1977): The algorithm adoptsthat the LOSVD F (vlos) has a Gaussianshape. Under this assumption, a Gaussianfunction is fitted to the quotient of the gal-axy spectra and the template spectra in

Fourier space as:

F (κ) =G(κ)S(κ)

≈ a·exp[−1

22πκσN

+2πiκvN

](5.4)

with F (κ), G(κ) and S(κ) being the dis-crete Fourier transforms of F (vlos), G(v)and S(v), respectively. a is a normalisationfactor which measures the strength of thegalaxy lines with respect to stellar lines (i.e.,the mean relative line strength), σ denotesthe velocity dispersion and v the mean ra-dial velocity, which is derived through thelogarithmic redshift in the channels. Theparameter N gives the number of pixels inthe input spectra.

• Cross-Correlation Method(Simkin 1974; Tonry & Davies 1979; Franx& Illingworth 1988): Initiated by Simkin(1974) and improved by Tonry & Davies(1979), the Cross-Correlation approach isbased on the fitting of a smooth symmet-ric function (quadratic polynomial) to thecross-correlation peak defined by the func-tions G(κ) and S(κ):

C(κ) =1

NσgσtG(κ) · S∗(κ) (5.5)

where σg is the rms of the galaxy spec-tra with σ2

g = 1N

∑uG(u)2 and σt the

rms of the template spectrum, respectively.The star ∗ denotes the complex conjugation.Cross-correlating the galaxy spectrum withthe template spectrum produces a functionC(κ), which has a peak at the redshift of thegalaxy and a width related to the dispersionof the galaxy. The resulting function is thenfitted with a Gaussian in Fourier space.

• Fourier Correlation Quotient (FCQ)(Bender 1990): In contrast to the two pre-vious methods, which assume a Gaussianbroadening function, the FCQ algorithm isan improvement of the former and a gener-alisation of the problem because it provides


Figure 5.1: Illustration of the optimum (Wiener) filtering technique for the power spectrum of the galaxyWHDF # 437. The definition of the smooth power models of the signal P ′signal and the noise P ′noise areconstructed by a least-square fitting (dot-dashed line) of the computed power spectrum Pobs. The cutofffrequency sc of the measurement, which is beyond the band-limit of the basic signal, and the characteristicfrequency due to sampling sNy are also indicated. The Nyquist frequency or critical frequency sNy, is halfthe sampling rate for a signal (sNy = 0.5/∆λ). The sampling theorem is satisfied as long as the band-widthof the measurement is less than the Nyquist frequency. However as noise is broadband, an band-limitinganalog filter must be employed prior to sampling to avoid aliasing (i.e., contributions from undersampledhigh-frequencies) by the transform of the random noise signal in the spectrum.

the full information of the LOSVD function.The method is based on the deconvolutionof the peak of the template-galaxy correla-tion function with the peak of the autocor-relation function of a template star as:

F (κ) =G(κ) · S∗(κ)S(κ) · S∗(κ)

(5.6)

In comparison to other procedures, the mainadvantage of the FCQ is that the most ap-propriate function to fit the LOSVD can beperformed a posteriori rather than a pri-ori. Furthermore, the method is less sensi-tive to template mismatch due to the factthat the LOSVD is determined from the ra-tio of the template-galaxy correlation andtemplate autocorrelation functions, whichare both smooth functions, instead of thedivision of the galaxy and the stellar spec-

tra (see next section 5.2 for details).

• Unresolved Gaussian Decomposition (UGD)(Kuijken & Merrifield 1993): This tech-nique adopts that the deviations from apure Gaussian are due to the composition ofseveral Gaussians which are uniformly dis-tributed in mean velocity and velocity dis-persion. However, the Gaussians are not ho-mogeneously distributed in amplitude:

F (κ) ≈N∑κ=1

aκ · exp[−(vlos − vκ)2

2σκ

](5.7)

where aκ indicates the amplitude, vκ themean velocity and σκ the velocity dispersionof each Gaussian. The UGD method com-putes the different amplitudes aκ for eachcomponent that provide the best fit to thegalaxy spectrum in the least-square sense.


Figure 5.2: LOSVDs F (R, vlos) in the galaxyUGC 12591. The surface height corresponds to thestellar density as a function of major axis radius andline-of-sight velocity. The surface is displayed fromtwo opposite perspectives, demonstrating the symme-try between the LOSVDs from both sides of the gal-axy. It is clearly evident that the line profiles are notGaussian distributed (Kuijken & Merrifield 1993).

However, this can only be achieved undercertain physical constraints (i) The LOSVDmust be non-negative everywhere, (ii) TheLOSVD is expected to vary smoothly onsmall scales and (iii) The LOSVD is as-sumed to be non-zero over a fairly smallrange in velocity only. In addition, theRayleigh criterion is required in order to beable to define the separation between twoadjacent peaks. To avoid dips between twopeaks, the separation should be less than2σ.

Fig. 5.2 shows two different viewpointsof the LOSVDs for the local galaxyUGC 12591 (S0/Sa). This galaxy is mod-erately inclined to the line-of-sight (≈ 67),with the azimuthal component being themain contribution to the 3D velocity dis-

persion along the line-of-sight on the ma-jor axis. UGC 12591 has a large rotationvelocity vrot ≈ 500 km s−1(Giovanelli etal. 1986) and a central velocity dispersionof σ0 ≈ 780 km s−1(Kuijken & Merrifield1993). Fig. 5.2 demonstrates the symme-try between the LOSVDs from two differentperspectives of the galaxy and clearly showsthat the line profiles are not Gaussian.

• Gauss-Hermite Expansion(van der Marel & Franx 1993; Gerhard1993): The approach of this method as-sumes the line profile F (vlos) to be a Gaus-sian with free parameters of line strength a,mean radial velocity v and velocity disper-sion σ as:

F (vlos) =aα(ω)σ

, ω ≡ (vlos − v)σ

(5.8)

where α(y) defines the standard Gaussianfunction:

α(y) =1√2π

e−(1/2)y2(5.9)

The LOSVD is derived via the product ofthis normal Gaussian and a truncated ex-pansion of the Hermite polynomials in theform:

F (vlos) =[aα(ω)σ

]·

1 +N∑j=3

hjHj(ω)

(5.10)

The Hermite coefficients hj describe de-viations from the Gaussian shape of theLOSVD, h3 measures asymmetries (mostlytaking values of h3 <∼ 0) and h4 symme-tries (h4 ≥ 0) and Hj characterise the Her-mite polynomials. Results for the parame-ters of (a, v, σ, h3, ..., hN ) are found us-ing a simultaneous N + 1 parameter fit thatare based on a χ2 residual minimisation ofthe galaxy spectrum and the convolution ofthe LOSVD (broadening function) with thetemplate spectrum in the Fourier domain.


In practice, the fits are restricted to the caseN = 4, but extension to higher orders isstraightforward. Note, that although physi-cally unmotivated, negative LOSVD are al-lowed (e.g., resulting from a non-zero h3).

5.2 The FCQ Algorithm

In the next paragraphs, a general description ofthe Fourier Correlation Quotient technique willbe given. A specific application to the dataof this work with slight modifications and welladapted parameter settings is discussed in sec-tion 5.3.The absorption lines of early-type galaxies getbroadened due to a combination of intrinsic ve-locity dispersion and integration along the line-of-sight. For this reason, an analysis of thebroadening function gives more detailed insightsinto the internal kinematics than, for example,using the mean velocity integrated along the line-of-sight and the velocity dispersion. Algorithmssuch as the Fourier quotient technique (e.g., Sar-gent et al. 1977) or the correlation method (e.g.,Franx & Illingworth 1988) provide no informa-tion about the velocity dispersion along the line-of-sight.The Fourier Correlation Quotient (FCQ)method (Bender 1990) measures the kinematicalbroadening function of early-type galaxies basedon the deconvolution of correlation peaks. Inparticular, the shape of the broadening function(see also Fig. 5.7) is recovered in great detail bydeconvolving the peak of the template-galaxycorrelation function with the peak of the au-tocorrelation function of a template star. Onemajor advantage is a high insensitivity dueto template mismatching than the standard(classical) Fourier quotient method. In thefollowing a brief overview of the FCQ techniqueis presented. For a more detailed descriptionas well as different calibration tests and errortreatments the reader is referred to Bender(1990).

As a first step the input galaxy and templatespectra must have been reduced in the same way(see chapter 3) and afterwards linear rebinnedon a logarithmic wavelength scale. The rea-son for going to logarithmic space is manifold.Firstly, the LOSVD is easier to extract and toanalyse there and secondly, even more impor-tant, an unique and therefore reliable solutionis derived as a specific logarithmic wavelengthregion corresponds to a well, unique definedspace in velocity space (along X-axis). Thirdly,the logarithmic technique shows more clearly inthe power spectrum the subtle power variationsat the high frequencies. In addition, followingrequirements have to be fulfilled: (i) the ab-sorption line widths in a template star and ina typical star of the galaxy are negligible incomparison to the instrumental broadening. Ingeneral, early-type galaxies comprise mostly ofK0 III/K3 III stars which have a small intrinsicline width (≈20 km s−1). Thus, this assumptionshould not induce any problems. (ii) the con-tinuum level in the spectra has been subtractedby dividing a low-order polynomial (zero or firstorder) from the spectra. (iii) the instrumentalline profile is constrained not to vary stronglyalong the wavelength range covered by the spec-tra and can be characterised as a function L(x),with x being the location of a specific line. How-ever, variations of the instrumental line profilealong the slit are allowed, which can be easilyaccounted for by taking stellar template spectraat several positions, slightly shifted from eachother along the slit (see section 3.7).As for a template star the line width profile L(x)is identical to the instrumental profile, a tem-plate spectra can be defined as line profile L(x):

S(x) =∑m

am L(x− xm)

S(λ) =∑m

am L(λ− λm) (5.11)

where xm is again the location of the mth ab-sorption line and am indicates the line strengthof the mth absorption line. As the location x


corresponds to the wavelength range, the equa-tion can be also expressed as a function of lnλ(see equation 5.11). For reasons of simpleness,only λ is written although strictly speaking it isln (λ) as the analysis is in logarithmic space. Theline profile G(λ) of galaxy spectrum is describedby the convolution of the instrumental line pro-file with the velocity dispersion integrated alongthe line-of-sight, or also the broadening functionB(λ).

G(λ) =∫B(λ− y)

∑m

bm L(y − λm)

dy

(5.12)The parameter bm gives the line strength of themth absorption line and the redshift of a spec-trum is represented by the shift of the center ofB(λ) in the λ-domain.

Based on the theorem of the classical Fourierquotient method that a convolution is a multipli-cation in Fourier space, a transformation of thegalaxy spectrum G(λ) results in:

G(κ) =∫G(λ)eiκλdλ =

∑m

bmeiκλmB(κ)L(κ)

(5.13)with the superscript tilde denoting the trans-formed functions. The parameter i is the imag-inary unit, and the exponential form of a com-plex number z in polar coordinates is z = ρ eiκ,where ρ is defined as the modulus and κ the ar-gument. Therefore, are G(λ) and eiκλ the realand the imaginary part of the function G(κ),respectively. In an analogous way the Fouriertransformation of the template spectrum S(λ)yields the transformed template function S(κ),but without the broadening function B(κ). Areliable broadening function can only be derivedif am = bm for all m, in other words if the instru-mental line profiles of the star and the galaxy areidentical. A slight mismatching would introducewave patterns and thus fail to deduce a correctbroadening function. The wave patterns ω affect

the broadening function as

G(κ)S(κ)

= ωB(κ) (5.14)

with ω =

∑mbme

iκλm∑nane

iκλn. At this point the FCQ

method comes into play. It uses an analogousway as the classical Fourier quotient method butwith the difference of deconvolving the corre-lation peak of the template-galaxy correlationfunction (KS,G(z)) with the peak of the autocor-relation function of the star (KS,S(z)). As a con-sequence the Fourier transformations of KS,G(z)and KS,S(z) yield the same result as in equa-tion 5.14 (see Bender 1990 for details). How-ever, in contrast to the classical Fourier quotienttechnique, solely the correlation peaks are de-convolved (instead of the complete correlationfunctions). The resulting formulae are a com-bination of two terms, between the correlationpeak plus the correlation function outside thecorrelation peak. Using a minimisation of theabsorption lines of the spectra, the distance be-tween the correlation peak and the nearest sec-ondary peak can be derived. This distance isindependent of the broadening of the spectra.Assuming that the width of the broadening func-tion B(λ) and the width of instrumental line pro-file L(λ) are described by velocity dispersions σBand σL, the correlation peak will not affect theclosest secondary peak if the two peaks are sep-arated larger than approximately two times oftheir FWHM. For this case, the deconvolutioncan be restricted to a relatively narrow (wave-length) interval around the correlation peaks andthe second terms can be neglected. The result-ing correlation functions of the peaks are definedas:

KS,G,peak(κ)KS,S,peak(κ)

=∑

m ambm∑n anan

B(κ) (5.15)

In other words, the correlation functions of thepeaks are described by a constant factor C times


the broadening function B(κ):

KS,G,peak(κ)KS,S,peak(κ)

= C · B(κ) (5.16)

Therefore, the complete broadening function canbe recovered independently from star templatemismatching. Furthermore, the FCQ methodwarrants an accurate removal of the instrumen-tal broadening and a direct way of calculatingthe velocity dispersion along the line-of-sight. Itis worth mentioning that the integrated area ofthe broadening function is a measure of the ratioof mean line strength of the galaxy to the startemplate.

5.2.1 Error evaluation

Bender (1990) performed extensive tests on bothnoisy and noise-free synthetic spectra. The meanredshift, velocity dispersion and line strengthwere calculated from a Gaussian fit to the broad-ening function and the errors in these parameterscan be derived from the rms deviations from theGaussian fit. In addition, the reducing effects ofthe Wiener filter which lowers the contributionsof high frequency noise to the rms deviationsare accounted for using a scaling factor with√N/NWiener, where N is the total number of

channels in the Fourier space and NWiener is thenumber of channels not removed by the Wienerfiltering. For sufficient high signal-to-noise spec-tra (S/N > 10 per pixel), even velocity dis-persions of ≈1/2 of the instrumental resolution(σinst) can be determined reliably. Furthermore,the FCQ technique is highly insensitive on metal-licities. Using different metallicities the resultson the derived velocity dispersions differ onlyless than 2%. In contrast, the classical Fourierquotient method induces an error of ≈10% for ametallicity difference of ∆[Fe/H] = 0.9 (Laird &Levison 1985).

Figure 5.3: Hale I-band image of the A 2390 clustergalaxy # 2511. The average size of the galaxy is ∼12pixel, which corresponds to 3.4′′ along the major-axisat z = 0.23. The slit width for spectroscopy was 1.5′′

and thus covers half the size of the galaxy.

5.3 Velocity Dispersions

Radial velocities and velocity dispersions of thegalaxies were calculated using an updated ver-sion of the FCQ program kindly provided byProf. R. Bender. The algorithm was developedto extract the LOSVD of early-type galaxiesin the nearby Universe. In contrast to localearly-type galaxies, this study aims to derive thekinematic properties of distant early-type galax-ies. Therefore, several assumptions and mod-ifications on the FCQ parameter settings haveto be taken into account. (i) Going to higherredshift (z >∼ 0.1), the average sizes of galaxiesbecome to small to resolve (or even measure a ra-dial dependence of) the velocity dispersion alongthe line-of-sight in detail. This is demonstratedin Fig. 5.3 for the A 2390 cluster galaxy # 2511(I = 19.07m), which has an average size of ∼3′′

along the major-axis at a redshift of z = 0.2312.As a consequence of the small sizes, the Her-


Figure 5.4: Monte Carlo simulations for the MOSCA setup. Left: For a range of input velocity dispersions(σin) the simulated velocity dispersion (σsim) is displayed as a function of S/N per A. Panel (a) using thesame star as “galaxy” and template and panel (c) using a different star as template. Right: Scatter of thedistribution as a function of S/N per A. Panel (b) same star as template and galaxy, Panel (d) different staras template and galaxy. Velocity dispersions are systematically underestimated, on average by 4% (a,b) and5% (c,d). For σin ≥ 300 and S/N< 10 the σsim are systematically overestimated by ∼6%. Different linesindicate the results for different stars: Panel (a,b) 084 (solid), 116 (dot-dashed) and panel (c,d) 084 & 116(solid).


mite coefficients h3 and h4 which characterisethe deviations from the Gaussian shape of theLOSVD, cannot be recovered. Therefore, theHermite coefficients are set to zero h3,4 = 0 andonly a Gaussian function is fitted to the quotientof the galaxy spectra and the template spectrain Fourier space. (ii) Subtraction of continuumlevel. Usually, the continuum in the spectra issubtracted by dividing a low-order polynomial ofzero or first order. However, several test in fit-ting orthogonal polynomials along the dispersionof the 1D stellar templates and the galaxy spec-tra revealed that a polynomial of fourth-order isthe best approximation for both continuae. Forthis reason, the continuum level in the templateand galaxy spectra was subtracted by dividing afourth-order polynomial from the spectra.

(iii) Calibration of errors. The errors derivedwith the FCQ method are based on the S/N inthe continuum of the spectra. In particular, theuncertainties depend on the S/N, the wavelengthregion and the number of absorption lines. As foreach set of templates, the wavelength region andnumber of absorption lines varies, adjustments tothe variables which define the error terms haveto be made. Using 100 Monte Carlo simulationsfor each stellar template the errors of FCQ werecalibrated on the basis of a set of input S/N. Inparticular, for a range of values of S/N of 10, 30,50, and 80 per A and velocity dispersions σin of80, 120, 160, 220, 300, and 450 km s−1 MonteCarlo simulations for synthetic generated galaxyspectra were performed, thereby adding artificialnoise and accounting for the instrumental sam-pling and resolution. In the next step, the veloc-ity dispersions σsim were recovered by the FCQsoftware. The results of the Monte Carlo simu-lations for two different templates are shown inFig. 5.4.

For each input velocity dispersion σin the re-covered σsim as a function of S/N is dis-played. In panel (a) two examples using thesame stellar spectrum as “galaxy” and templateare shown (# 084 (K0III) and # 116 (K3III)),

and in panel (c) a different star as template(K3III) is adopted. For all values of between80 ≤ σin ≤ 300 km s−1 the velocity dispersionsare systematically underestimated, on averageby 4%. At σin ≥ 300 and S/N<10 the σsim aresystematically overestimated by ∼6%. For thesmallest values of σin = 80 km s−1 the fit val-ues are on average lower by 14%. But with in-creasing velocity dispersion the deviations dropquickly: 4% for σin ≤ 100 km s−1 for a givenS/N and only 2% for σin ≤ 100 km s−1 andS/N>10. For panel (c) the situation is similar.On average, the velocity dispersions are system-atically underestimated by ∼5%. However, forσin = 80 km s−1 σsim are underestimated by10%, for S/N=10 by 4%. Again for higher ve-locity dispersions the uncertainties are lower by4% for σin ≤ 100 km s−1 at each S/N and 3% forS/N=10. The conclusion from the simulationsis that the systematic uncertainty on small val-ues of S/N is hard to correct. For S/N≥15 theeffect becomes negligible but data with S/N<10should be rejected. In Fig. 5.4 also the relativescatter around the mean for the same configu-rations is displayed (template # 084 and # 116,panel (c); template # 084 & # 116 panel (d)).For a given S/N, the uncertainty on the veloc-ity dispersions is even small for the smallest σin

values. Only for the largest input velocity dis-persions of σin = 450 km s−1, the error is slightlylarger for lower S/N. However, as usually thesebright galaxies offer high S/N values of S/N> 40,the systematic uncertainty becomes negligible.

In general, for S/N of the same value, the er-ror estimated by FCQ were a factor 1.3 smallerthan the rms of the error values in the MonteCarlo simulation, which corresponds to σsim val-ues that are 5% too low. To account for thiseffect, this derived error parameter was fixed forall subsequent velocity dispersion measurementsto ensure a reliable error treatment.

Before feeding the FCQ algorithm, a number ofimportant input parameters (σinst, scale, row,redshift and broad) were gained in a semi–


Figure 5.5: Instrumental resolution σinst as a func-tion of wavelength as determined from the width ofHgAr/Ne lines for the MOSCA setup. Symbols repre-sent individual measurements, the solid line the bestfitting parabola.

automatic manner using a special program. Inthe following a short summary of the requiredanalysis steps for each galaxy is given:

• Measurement of the instrumental resolutionof the galaxy spectra. For this purpose, aset of 12–14 unblended, isolated and not-saturated HgAr/Ne lines in the wavelengthcalibration frames were selected and theirFWHM width were measured for ten differ-ent positions over the whole wavelength re-gion (5300 A <∼ λobs <∼ 7725 A correspond-ing to 4300 A <∼ λrest <∼ 6260 A). This wasperformed on frames in pixel space, therebycovering the same wavelength region as forthe galaxy spectra under consideration. Theresults were stable from frame to frame andfor different zenithal positions of the tele-scope as well as for different nights. Fig. 5.5shows the instrumental resolution σinst as afunction of wavelength as determined fromthe width of HgAr/Ne lines for the MOSCAsetup. Each point corresponds to a single

Table 5.1: Instrumental resolution for MOSCA.

λobs N σinst 〈σinst〉 σinst 〈σinst〉A [A FWHM] [km s−1]

5306 20 5.28±0.25 5.24 127.1±6.3 126.05998 17 5.06±0.19 5.00 106.5±2.6 105.66380 17 5.02±0.19 4.99 100.2±2.5 99.27174 17 5.75±0.21 5.71 102.3±3.8 101.57245 4 5.61±0.09 5.67 98.7±1.7 99.97439 24 6.03±0.22 5.99 103.4±3.9 102.77489 8 5.81±0.11 5.84 99.0±2.0 99.47634 19 6.48±0.19 6.47 108.3±3.2 108.07697 6 6.48±0.25 6.51 107.4±4.2 108.07723 17 6.59±0.20 6.62 108.9±3.3 109.4

measurement of an unblended HgAr/Ne linein one calibration frame. For testing pur-poses, in case of the G4300-band (G–bandhereafter), Hβ and Mgb feature, additionalmeasurements were performed within thesewavelength ranges and the mean resolutionof 3–5 calibration lines computed to verifythat during one night σinst remains at thesame resolution for each single slit observa-tion. Results of the two approaches werecompared to each other and are consistentwithin the errors. For this reason, the entireset of single measurements was used to es-tablish a relation for the resolution of thespectra as a function of wavelength. Asshown in Fig. 5.5, the measurements for theresolution (km s−1) are well described by aparabola.

All measurements in Fig. 5.5 are listed inTable 5.1. For each wavelength the mean(σinst) and median (〈σinst〉) values of the in-strumental resolution in units of A FWHMand km s−1 are given. N shows the num-ber of individual measurements, for the G-band, Hβ and Mgb band the average val-ues. Typical numbers for the instrumen-tal resolution of MOSCA spectra are be-tween 97 ≤ σinst ≤ 130 km s−1. The wave-length range around the Hβ and Mgb feature


(≈6119 A) has a σinst of ∼3.84 pixel (4.99 AFWHM) corresponding to 104 km s−1andaround the G-band (≈5290 A) σinst takesvalues of ∼4.03 pixel (5.24 A FWHM) corre-sponding to 126 km s−1. The instrumentalresolution of the FORS spectra were derivedin the same way as for the MOSCA spec-tra. For example, the galaxy WHDF # 173(z = 0.4507), has a σinst at the Mgb band of∼3.61 pixel (5.91 A FWHM) or 100 km s−1

and around the G-band ∼3.75 pixel (6.14 AFWHM), which corresponds to 126 km s−1.

• Calculation of the number of extracted rows(row) over which the two-dimensional spec-tra was summed up (see also section 5.3.2).

• Determination of the instrumental resolu-tion of the template star spectra. Thekinematic templates which were observedthrough a 0.5′′ longslit aperture have typi-cal instrumental resolutions at the Hβ andMgb feature of ∼2.3 A FWHM, correspond-ing to σ∗ = 58 km s−1and around the G-band ∼2.3 A FWHM, which corresponds toσ∗ = 71 km s−1. A complete list for all stel-lar templates together with their measuredresolution can be found in Table A.1 on page181.

• Since the slitlets of the galaxies had smallvariations in width size and were not iden-tical in width to the longslit of the stars,a simple comparison and interpretation ofthe calculated velocity dispersions would bewrong. To correct for this effect, the fol-lowing correction procedure to the stars wasapplied before using them as templates inthe FCQ algorithm to compute the velocitydispersion for each galaxy. First, the starsand galaxies were rebinned to a fixed log-arithmic wavelength scale. Afterwards thespectra of the stars are broadened onto thesame instrumental resolution (in km s−1) asthe spectra of the target galaxy using the

relation

broad =√σ2

inst Gal − σ2inst Star (5.17)

where broad denotes the factor of broaden-ing for the stellar spectra. This procedurehas the advantage that the galaxy velocitydispersion results directly and no additionalcorrection of the velocity dispersion of thetemplate star σstar is needed.

• Extraction of a spectral wavelength rangeof the broadened stars which is used as aninput for FCQ. Typical wavelength rangesfor analysing the G-band, the Mgb andthe Hβ feature cover a region betweenλλ 4142−4798 A (G4300–Fe4668), λλ 5000−5500 A ([O III]–Fe5335) and λλ 4606−5116 A(around Hβ), respectively.

• Extraction of an equivalent spectral rangein the target galaxy spectra as for the tem-plates and analysis of this region via theFCQ method. Residuals of night sky lineswithin this wavelength range are masked outin order to not affect the fitting process ofthe broadened stars to the galaxy spectra.In particular, the masking of telluric lineswas performed in setting the flux value to afixed mean value which was derived over theadjacent continuum level of the spectrum.

• Derivation of velocity dispersions (σ) andradial velocities (vrad) via the FCQ analysisby deconvolving the peak of the template-galaxy correlation function with the peak ofthe autocorrelation function of the templatestar (see previous section 5.2 for the theo-retical description). A minimum of at leastthree templates was used for determining σ,vrad and S/N for each galaxy. All individualparameter measurements were compared toeach other (see next section 5.3.1). If thesingle σ and vrad values varied only withintheir errors the mean value of the individ-ual determinations was calculated. Velocity


Figure 5.6: Visual inspection of the velocity dispersion measurement using the Mgb feature for the fieldelliptical galaxy WHDF # 437 with R = 19.50m at z = 0.260. The galaxy spectrum (solid line) is comparedto the star spectrum (dot-dashed) which is broadened to the derived σ of the galaxy. Absorption lines ofMgb Fe5270 and Fe5335 are shown, splitted into their defined feature passbands (blue and red continuumand central passband). The central and continuum bands are indicated as shaded regions. The spectralregion labelled as ‘sky’ is affected by the telluric line at λ = 6367 A. The lower X-axis displays the spectrumin logarithmic units, the upper X-axis in linear units (both observed wavelengths in A).

dispersions with too large errors resultingfrom low S/N spectra were rejected from theaveraging process. To test possible varia-tions in observing conditions and to identifyspectra with low S/N, the sum of individualspectra for each night was compared to thecombined spectra of all nights. With thismethod, exposures with very low S/N couldeasily be identified and excluded from the fi-nal summarisation procedure. For example,for the A 2390 galaxies the σ determinationsof the summarised spectra of five differentnights were compared to the final combinedspectrum of all nights. For the poor clustergalaxies, the sum of individual spectra forthree different nights and the summarisedspectrum of all three nights were comparedto each other. In general, the agreement wasvery good and for a few cases up to two sin-

gle spectra had to be rejected in the finalaveraging process. For six out of 98 clustergalaxies, no reliable measurement was pos-sible. These objects have very low signal(S/N<∼ 8 per A) and three are faint back-ground spiral galaxies.

In case of the Low–LX clusters and the twofield galaxy samples, the S/N in all galaxyspectra was sufficient to derive velocity dis-persions. A complete summary of the checkson the velocity dispersions determinationswill be presented in the next section.

5.3.1 Visual Inspection of σ

To verify the quality and reliability of the ve-locity dispersion measurements, the individualresults for each broadened star spectra have tobe compared to the target galaxy of interest.The inspection of the σ determinations was per-


Figure 5.7: Comparison of the Wiener filteredLOSVD (solid line) derived with the FCQ methodusing the Mgb feature with the broadening functionresulted from a fit by a Gauss-Hermite series (dot-dashed line) with hj = 0 for the field elliptical galaxyWHDF # 437. The broadening function gained fromthe FCQ method reproduces the LOSVD extremelywell and is hardly distinguishable from it.

formed in Fourier space, which provides a directcomparison between the LOSVD and the FCQresults. For each galaxy, following checks on σhave been executed:

• Comparison of the broadened template starspectra with the galaxy spectrum. Fig. 5.6shows an example for the broadened tem-plate star # 157 (dot-dashed line) comparedto the spectrum of the field elliptical galaxyWHDF # 437 at z = 0.2599 (solid line) withan apparent total brightness of R = 19.50m.As only a narrow wavelength range of thespectrum is displayed, the logarithmic axisln (λ) appears to be linear but actually it isa weak increasing function towards longerwavelengths. The final velocity dispersionfor this galaxy is σ = 128 ± 8 km s−1 witha signal-to-noise level of S/N = 57 per Aderived within the extracted spectral regionaround the Mgb feature. This value was av-eraged over three individual measurementsσ084 = 127±8 km s−1, σ116 = 134±8 km s−1

and σ157 = 124±10 km s−1 which are in verygood agreement as the scatter of the valuesare within their uncertainty.

• A comparison of the line-of-sight-velocity-distribution with broadening function isperformed. In particular, the Wiener fil-tered Fourier transform (LOSVD) is theoutput of the FCQ algorithm and thebroadening function is characterised by apure Gaussian. As the Gauss-Hermite-polynomial coefficients hj were set to zero,only a Gaussian was fitted to the LOSVDof the galaxy. Fig. 5.7 compares the fil-tered LOSVD to the broadening functionB(λ). The Gaussian function recovers theLOSVD extremely well and is nearly in-distinguishable from it. By contrast, thestandard Fourier quotient method becomesstrongly affected even by slight mismatchingof the template star. For example, using aK0 III star instead of a K3 III star under-estimates the velocity dispersion of the re-covered broadening function by ≈ 5% (cor-responding to 10 km s−1 for a typical gal-axy with σGal = 200 km s−1). However, thissmall mismatch is hardly detectable througha visual inspection of the template’s spec-tra (see also Fig. 3 of Bender 1990). Thewidth of the Gaussian function (broadeningfunction) in Fig. 5.7 represents the velocitydispersion of the object and the position onthe X-axis indicates the distance to the tar-get galaxy. If the position of the broadeningfunction is slightly shifted, the radial veloc-ity of the galaxy was not exactly measured.On the other hand, if the broadening func-tion cannot completely recover the width ofthe LOSVD, the absorption line might beaffected by weak sky line residuals and thetotal flux of the absorption feature cannotbe measured, which results in an uncertainvelocity dispersion of the galaxy.

• Comparison of B(λ) with the peak of thecorrelation function, see Fig. 5.8.

At a redshift of z = 0.231 the Mgb line isstrongly affected by the telluric emission line atλ = 6367 A. In particular, for the clusters A 2390


Figure 5.8: Comparison of the LOSVD (dot-dashedline) derived by a simple Fourier transformation (G)assuming the region around the Mgb band with thecorrelation function (CF, solid line) for the galaxyWHDF # 437. The wings of the correlation functionvary strongly because no filter was applied.

(z = 0.228) and the Low-LX clusters Cl 0849+37(z = 0.2343) and Cl 1702+64 (z = 0.2233) theMgb feature falls around the mean cluster red-shift, thereby causing problems when the deriva-tion of velocity dispersions is based solely on thiswavelength region. For this reason, three sep-arate wavelength regimes were analysed. Thefirst wavelength range between λλobs = 6204–6748 A was centered on the Mgb feature. In thosecases where the Mgb line was affected, either inthe central or in the adjacent continuum win-dows (see Fig. 5.6), a reliable measurement ofthe velocity dispersion could not be ensured sothat a second calculation in the range around theG4300 band between λλobs = 4964–5541 A wasperformed. In addition, for two masks of theA 2390 sample a wavelength range around theHβ feature λλobs = 5598–6248 A was analysed.

5.3.2 Aperture Correction

Early–type galaxies feature radial gradients intheir velocity dispersion, radial velocity and inthe Mg2 and Fe indices. For this reason, dependthe central measurements of these parameters onthe distance of the galaxies and the size of theaperture. The apparent sizes of distant galaxiesare so small that a slitlet of width 1–1.5′′ covers

a large fraction of the galaxy’s light distribution,typically 1–2 half-light radii Re. The derivedvelocity dispersions and line strengths are onlyluminosity–weighted average measurements, butare not equivalent to the central values of localgalaxies if a radial gradient across the galaxy ex-ists. To take this effect into account, an aperturecorrection was applied to the observed velocitydispersions based on the logarithmic gradientsas outlined by Mehlert et al. (2003), who mea-sured central velocity dispersions and absorptionline strengths in high-S/N spectra of 35 early–type Coma cluster galaxies. At a later stage,the A 2390 cluster sample will be combined withthe a sample of the cluster A 2218. Therefore, forreasons of consistency, the velocity dispersions ofthe A 2218 galaxies were re–analysed using thisaperture correction. Velocity dispersions of clus-ter and field galaxies were aperture corrected viathe relation

∆(log σ) = 0.06 · log(dClus

dComa· aClus

aComa

)(5.18)

where dX is the angular distance of the targetcluster under consideration, dClus or the refer-ence cluster dComa, and aX is the aperture radiusof the target or the local counterpart, respec-tively. For each galaxy, the aperture a duringthe observations is given by:

aClus =

√r · row · scale

π, (5.19)

where r is the slit width (for A 2218 and A 2390and the three poor clusters 1.5′′, for the FDFand WHDF ellipticals 1.0′′ was used), row de-notes the number of extracted rows over whichthe spectrum was averaged by the Horne algo-rithm, which differs from galaxy to galaxy andwas measured for each object individually andscale represents the scale in arcsec/pixel. Me-dian values of row are 4.7 pixels, correspondingto 2.8′′ for A 2218, 8.4 pixels (2.8′′) for A 2390,12.28 pixels (4.1′′) for Cl 0849, 10.89 pixels (3.6′′)for Cl 1701, 10.49 pixels (3.5′′) for Cl 1701, 6.46


pixels (1.3′′) for the FDF and 7.61 pixels (1.9′′)for the WHDF galaxies, respectively. This log-arithmic gradient is slightly steeper than theone proposed by Jørgensen, Franx & Kjærgaard(1995) and Jørgensen (1999):

∆(log σ) = 0.04 · log(dClus

dComa· aClus

aComa

)(5.20)

However, aperture corrections based on eithergradient are consistent within their errors. Tomatch the circular aperture with a radius of 1.7′′

of the Coma cluster galaxies by Jørgensen etal. (1995, 1999), the spectroscopic data wascorrected following the correction formula asgiven by equation 5.18. For the σ measure-ments this results in median corrections for thecluster A 2218 of ∆(log σA 2218) = +0.039, andthe clusters A 2390 of ∆(log σA 2390) = +0.042,Cl 0849 of ∆(log σ0849) = +0.048, Cl 1701 of∆(log σ1701) = +0.047 and for Cl 1702 of∆(log σ1702) = +0.045, and for the field galaxiesin the FDF of ∆(log σFDF) = +0.034 and in theWHDF of ∆(log σWHDF) = +0.040, respectively.

5.3.3 Comparison between DifferentAbsorption Features

For the majority of the early-type galaxies, twodifferent spectral bands could be used for thederivation of σ. However, for some clusters witha mean cluster redshift around 〈zClus〉 ≈ 0.231the Mgb absorption line is impractical dueto strong sky lines in this wavelength region.Therefore, an additional second region aroundthe G4300 index band was analysed in the caseof A 2390, Cl 0849 and Cl 1702. Moreover, forA 2390 a third wavelength range around Hβ wasused as a further verification. For the field galax-ies the situation changed for each object becausedifferent absorption passbands get affected dueto sky lines. Nevertheless, at least one σ mea-surement around the Mgb or G-band was possi-ble. In four WHDF field galaxy spectra, bothfeatures were free of telluric emission lines andtwo independent σ results could be derived.

For each galaxy, the fiducial value of the veloc-ity dispersion is based on the comparisons of thebroadened template star spectra with the galaxyspectrum, the line-of-sight-velocity-distributionwith the Gauss-Hermite (GH) polynomial fit andthe Fourier output results with the correlationfunction (cf. section 5.3.1). As a further con-sistency check, it is worthwhile to compare thevelocity dispersion measured from different ab-sorption feature passbands. For the followingtests, the aperture corrected velocity dispersionsof distant samples are used.

The reference velocity dispersion measurementsare indicated as σ1, whereas velocity dispersionsderived from other features are denoted with thecorresponding absorption line. As for the refer-ence σ values, the σ determinations of the com-parison feature were verified with the tests asoutlined in section 5.3.1. Based on these verifica-tions, velocity dispersions were classified accord-ing to a scheme from 1 to 5 based on their over-all quality and accuracy. Galaxies with a qualityparameter Qσ = 1 are very good measurementswithout artifacts or variations in their LOSVDand GH fit and free from any contamination bynight sky emission lines. Qσ = 2 are good mea-surements with slight variations in LOSVD andGH fit and Qσ = 3 corresponds to intermediatequality with some variations in LOSVD and/orslight influences of residual sky lines near the fea-ture of interest. Qσ = 4 represent bad σ deter-minations with strong variations in LOSVD andclear signs of sky lines which deplete and affectparts or the whole absorption line. In cases ofQσ = 5, no measurement was possible.

As very closely to the mean redshift of the A 2390cluster (z = 0.231) the Mgb line is strongly af-fected by a telluric emission line (λ = 6367 A),it has to be verified how strong the effect re-flects on the velocity dispersion determinationsbased on the Mgb feature. In Fig. 5.9 the ve-locity dispersions based on the Mgb feature areshown as a function of the σ reference values forthe A 2390 galaxies. A total of 36 member gal-


Figure 5.9: Comparison of velocity dispersion mea-surements derived from different absorption features.The values taken as reference are displayed on thex-axis, velocity dispersions based on the Mgb featureare shown on the y-axis. Filled symbols denote highquality σ measurements, open symbols the low qual-ity σ determinations.

axies, splitted into 28 galaxies with Qσ = 1 andQσ = 2 and 8 objects with Qσ = 3 are com-pared to each other. 11 galaxies with a qual-ity parameter of Qσ = 4 were discarded andfor one object (# 2106) no σ based on the Mgbfeature could be derived. On average, the 28galaxies with 1 ≤ Qσ ≤ 2 show a difference of∆(σ1 − σmg) = −6 ± 17 km s−1 with a median〈∆(σ1 − σmg)〉 = −2 km s−1. Both numbers areless than the typical internal uncertainty of thevelocity dispersions, which is δσ > 6 km s−1.The 8 galaxies with Qσ = 3 have an averagedifference of ∆(σ1 − σmg) = −43 ± 39 km s−1

and a median of 〈∆(σ1 − σmg)〉 = −41 km s−1.Although a careful analysis of the σ determina-tions was performed, the secondary σ determi-nations based on the Mgb absorption line yieldon average to 1% and to 27% higher σ valuesthan the reference values for Qσ = 1, 2 and forQσ = 3, respectively. A possible dependence onthe mask setup for the galaxies with quality par-

Figure 5.10: Comparison of velocity dispersion mea-surements derived from different absorption features.Upper panel: relative uncertainty of velocity disper-sions with the Mgb values as a function of velocitydispersions σ1. Lower panel: ∆σ/σ1 versus the S/Nper A. Symbol notations as in Fig. 5.9.

ameter of 1 and 2 was not detected, all threesub–samples feature the similar scatter and un-certainties. The velocity dispersions for galaxieswith intermediate quality (Qσ = 3), are clearlyoffset. A probably reason is that a weak residualsky lines in the spectra close to the absorptionline are present which affect strongly the finalvelocity dispersion measurement. In particular,all velocity dispersions with Qσ ≥ 3 show higherσ values in comparison to their reference value,which gives support to the assumption that acontamination by residual sky lines in the spec-tra is even existent for the intermediate qualitydeterminations. From this comparison it seemsalready clearly evident that the telluric emissionline around the Mgb has dramatic effect on thederived velocity dispersions. Before the conclu-sions are drawn from this test, the influence ofthe signal–to–noise ratio on the accuracy of thevelocity dispersion determinations is discussed.


The quality of the velocity dispersion measure-ment and therefore its formal uncertainty de-pends on the S/N in the spectrum, the compari-son of the galaxy velocity dispersion with the in-strumental resolution and to a lesser extent thechosen wavelength interval used for the deter-mination. Fig. 5.10 displays the formal relativeuncertainty ∆σ/σ1 as a function of velocity dis-persion (upper panel) and as a function of theS/N (lower panel) per A. ∆σ/σ1 was derived asthe relative difference in velocity dispersion val-ues of the G-band with the Mgb line, normalisedto the reference velocity dispersions σ1. As ex-pected, the deviations of velocity dispersions in-crease with decreasing S/N ratios. The relativeuncertainty for galaxies with Qσ = 3 falls intothe range 0.14 ≤ ∆σ/σ1 ≤ 0.56 with a median of〈∆σ/σ1〉 = 0.31. For galaxies with high qualitymeasurements (Qσ = 1), the relative errors coverthe range 0.02 ≤ ∆σ/σ1 ≤ 0.21 with a median of〈∆σ/σ1〉 = 0.06. Two effects of the uncertaintiescan be seen in Fig. 5.10. First, the errors of ve-locity dispersions with intermediate quality arelarge even for high S/N. This means that the σdetermination for the Qσ = 3 is not reliable anda different absorption line should be used. Sec-ond, the uncertainties of the high quality mea-surements depend only weak on the S/N, whichsuggests that the σ values are in fairly goodagreement. However, as even for the high qual-ity σ Mgb determinations a slight offset to higherσ values is found and a possible (not visual de-tectable) interference due to sky line residualscan not completely be excluded, for the major-ity of cluster galaxies in A 2390 and the two poorclusters the G-band was used for the derivationof velocity dispersions. For this reason, 45 out of47 σ1 values for the A 2390 galaxies are based onthe G4300 feature, which showed in all cases nocontamination of any residual sky lines. More-over, the G-band was appropriate feature for thepoor clusters and was solely used for the σ mea-surements of 15 and 7 cluster members in Cl 0849and Cl 1702, respectively.

Figure 5.11: Comparison of velocity dispersion mea-surements derived from different absorption features.Upper panel: relative uncertainty of velocity disper-sions with the G–band values as a function of veloc-ity dispersions σ1 of the WHDF field galaxies. Allgalaxies are high quality σ measurements. Lowerpanel: ∆σ/σ1 versus the S/N per A. Measurementsare splitted according to the S/N of the feature (Mgbcircles, G–band triangles) to visualise possible sys-tematic errors of low S/N on σ. For S/N >∼ 15,effects of systematic uncertainties become negligible.

To test if the Mgb absorption lines yield accu-rate velocity dispersion measurements and notany bias is introduced with the restriction tothe G-band absorption feature, Mgb σ determi-nations for four WHDF field early-type galaxiesare compared to the σ values derived based onthe G-band feature. Fig 5.11 shows the formalrelative uncertainty ∆σ/σ1 as a function of ve-locity dispersion (upper panel) and as a functionof the S/N (lower panel) per A for the WHDFearly–type galaxies where both measurementsare available. The lower panel is divided accord-ing to the S/N of the feature of interest, S/Nvalues from the Mgb lines are denoted as circles,whereas the S/N based on the G–band is indi-cated as triangles. Again, ∆σ/σ1 was computed


Figure 5.12: Comparison of velocity dispersion mea-surements derived from different absorption features.The values taken as reference are displayed on the x-axis, velocity dispersions based on the Hβ feature areshown on the y-axis. Symbol notations as in Fig. 5.9.

as the relative difference in velocity dispersionvalues of the Mgb with the G-band line, nor-malised to the reference velocity dispersions σ1.Both features were free from any contaminationdue to sky lines. The relative uncertainties coverthe range 0.03 ≤ ∆σ/σ1 ≤ 0.06 with a medianof 〈∆σ/σ1〉 = 0.05. The σ determinations arein very good agreement and show similar abso-lute errors between 8 to 18 km s−1. Even forthe lowest S/N ratios the formal uncertainty of5% is less than the typical individual error in theσ measurement. This limit is also the smallestformal uncertainty that can be reached with theinstrumental resolution and the template stars.Although this comparison is based on a smallnumber, no trend of increasing systematic errorson the σ values with low S/N can be found. Inaddition, velocity dispersions derived for galaxiesin the other clusters and field samples show sim-ilar results and a good agreement between σMgband σG−band. For all galaxy samples in this the-sis, regardless of their environment loci, the S/Nwas at least 13 per A. Results from the simula-

Figure 5.13: Comparison of velocity dispersion mea-surements derived from different absorption features.Upper panel: relative uncertainty of velocity disper-sions with the Hβ values as a function of velocitydispersions σ1. Lower panel: ∆σ/σ1 versus the S/Nper A. Symbol notations as in Fig. 5.9.

tions yielded that the effects of systematic errorson σ become important at S/N <∼ 10 (cf. sec-tion 5.3). As this limit is lower than the averageS/N in the spectra of low–velocity dispersion gal-axies, the formal uncertainty therefore becomesnegligible even for the lowest derived velocity dis-persions in the galaxy samples. A comparison ofthe absolute velocity dispersions for the WHDFellipticals is presented in Fig 5.14.A third comparison of velocity dispersions wasperformed for a sub–sample of 23 galaxies inA 2390 by analysing a wavelength range aroundthe Hβ feature. In contrast to metallicity-sensitive indices, the Hβ index (commonly re-ferred as an age-sensitive index) varies consider-ably among elliptical galaxies (Gonzalez 1993).In most local early-type galaxies nebular emis-sion is present at a measurable level, which read-ily can fill in the Hβ absorption, thus makingthe accurate measurement of the true absorption


a difficult business. As particularly lower orderBalmer lines (Hα and Hβ) are stronger affectedby emission from ionised gas (Osterbrock 1989),higher-order Balmer lines (Hγ and Hδ) are pre-ferred for the age determinations of galaxies athigher redshift, because lower order Balmer linespossibly get contaminated by night sky emissionlines. In addition, for galaxies at higher redshifta possible weak nebular emission might not beclearly detectable and overlayed onto the absorp-tion feature. For this reason, the agreement be-tween the reference velocity dispersions and themeasurements of velocity dispersions based onthe Hβ index are less confident and the σ valuesderived from the Hβ feature are only regardedas an additional consistency check.

In a first step, the spectra of the template starsand the galaxies were inspected if the Hβ indicesmight be subjected to weak emission. Neitherfor the stars nor for the galaxies any Hβ emis-sion was detected. However, as just mentionedabove, one must be aware of possible variationsof the Hβ index. In Fig. 5.12 the σ referencevalues for the A 2390 galaxies are compared tothe velocity dispersions derived via the Hβ ab-sorption line. Again, spectra were divided withrespect to their quality. 18 early–type spectrawere classified with a Qσ parameter of 1 and 2,five galaxies have Qσ = 3. On average, simi-lar results are found as for the velocity disper-sions based on the Mgb line, although with alarger scatter. Galaxies with quality parameterQσ = 1, 2 are shown as filled circles, objects oflower reliability with Qσ = 3 are indicated asopen circles. in Fig. 5.12. Again, the deviationsof velocity dispersions decrease with increasingS/N ratios. With respect to the reference values,the 18 galaxies with 1 ≤ Qσ ≤ 2 have a differenceof ∆(σ1−σHβ) = −22±20 km s−1 with a medianof 〈∆(σ1 − σHβ)〉 = −17 km s−1, correspondingto 8% higher σ values than the G-band. This re-sult is in fair agreement to the σ reference values.The 5 galaxies of lower quality with Qσ = 3 havea difference of ∆(σ1 − σHβ) = −34 ± 41 km s−1

and a median of 〈∆(σ1 − σHβ)〉 = −51 km s−1,which are higher in σ by ∼30%. Both σ mea-surements based on the Hβ line are larger thanthe typical internal uncertainty of the velocitydispersions.Fig. 5.13 shows the relative difference in veloc-ity dispersions derived from the Hβ feature plot-ted against the reference velocity dispersions σ1.The lower panel displays δσ/σ1 as a functionof the S/N per A. The relative uncertaintiesfor the 18 galaxies with Qσ = 1, 2 cover therange 0.02 ≤ ∆σ/σ1 ≤ 0.30 with a median of〈∆σ/σ1〉 = 0.11. For galaxies with low qualityvelocity dispersions (Qσ = 3) the relative errorsfall in the range 0.15 ≤ ∆σ/σ1 ≤ 0.52 with amedian of 〈∆σ/σ1〉 = 0.34.The agreement between the Hβ and the G-bandmeasurements is less confident because of the fol-lowing two reasons. Besides the presence of weakHβ emission superimposed over the spectrum,another effect may be small variations of the Hβline itself. The strength of the Hβ absorption lineis assumed to be sensible due to changes in the(mean) age of the underlying stellar population.Both effects can modify the shape of the absorp-tion line which has influence on the derived mea-sure of the velocity dispersion. A combination ofboth effects is therefore most likely.

5.3.4 Comparison between RepeatObservations

For a total of eight galaxies in the A 2390 sam-ple and five field galaxies in the WHDF samplewhich are included in two or more different MOSmasks, it is possible to confirm the internal re-liability of the spectral analysis. Galaxy # 2933was discarded as it is a background galaxy lo-cated behind the A 2390 cluster and no velocitydispersion based on the Mgb feature could be de-rived. A total of nine repeat observations wasacquired for the WHDF ellipticals. Two galax-ies were observed thrice (# 92 and # 111) and# 173 four times. The velocity dispersions havebeen measured for both individual setups and


Figure 5.14: Comparison of velocity dispersion mea-surements for spectra which have been observed withtwo different MOS mask setups. Solid symbols aregalaxies in A 2390, open symbols are WHDF field gal-axies. One galaxy with four measurements is denotedwith two additional large circles.

the agreement between these two determinationsis very good with a median offset in σ at the 4%level. Fig. 5.14 compares the velocity dispersionmeasurements for the eight A 2390 spectra whichwere observed with two different mask setupsand five WHDF field galaxy spectra with twoor up to four repeat observations. The A 2390and WHDF galaxies are denoted as solid dotsand open stars, respectively. In general, velocitydispersions of the secondary measurements arebased on lower S/N spectra and therefore showlarger errors. The σ values of galaxy # 173 basedon four individual spectra are shown as two sepa-rate symbols which are indicated with additionalcircles. The final measurement with the smallesterror yielded σ173 = 204±16 km s−1 (S/N = 37),which represents also the smaller deviation fromthe one–to–one line. The relative difference be-tween reference and secondary measurements ofthe A 2390 galaxies is δσ/σ1 = 0.051±0.048 with26 ≤ S/N ≤ 65. For the WHDF field ellipticalsthe relative difference between reference and sec-

ondary measurements is δσ/σ1 = 0.041 ± 0.015with 8 ≤ S/N ≤ 57. In order to increase thesignal–to–noise in the final spectra, these repeatobservations have been co–added. For example,the S/N in the final spectra of the A 2390 samplevaries between 21 and 65 with a median value ofS/N ∼ 34 per A and a mean value of S/N ∼ 37per A.

5.3.5 Comparison between DifferentExtraction Procedures

As described in section 3.5, for each exposureone-dimensional spectra were extracted sepa-rately using the Horne–algorithm (Horne 1986),which optimally weights the rows of the ex-tracted spectrum profile to maximise the signal-to-noise in the resulting one-dimensional spec-trum. Therefore, possible shifts of the spectrawhich may have occurred during different nightsare accounted for in the final one-dimensionalsummed-up spectra.However, it is possible that the method of op-timal weighting by Horne (1986) might have aneffect on the velocity dispersion measurementsof the galaxies. The reason is that the aperturecorrection assumes equal weight in the extrac-tion aperture. This means, that a certain con-tribution comes from radii outside the very gal-axy center, where the velocity dispersion is lower.With optimal weighting, rows away from the gal-axy center get lower weight, which means themeasured velocity dispersion might be higher.The following aperture correction then overcor-rects the velocity dispersions, as it assumes a sig-nificant contribution from rows away from thecenter. In order to test this effect, the veloc-ity dispersion measurements on spectra derivedusing optimally-weighted extractions and equal-weight extractions have been compared.For the following test, ten galaxies of the A 2390sample (thereby encompassing both faint andbright spectra) on different mask and slitlet con-figurations have been selected. In particular, thegalaxies of slitnumber # 01 to # 05 of mask 1 and


Figure 5.15: Comparison of velocity dispersionmeasurements on spectra derived using optimally-weighted and equal-weight extractions. Objects wereobserved on different mask and slitlet configura-tions, mask 1 (circles), mask 3 (triangles). Up-per panel: relative difference of velocity dispersions∆σ/σopt as a function of velocity dispersion usingthe method of optimally-weighted extractions. Lowerpanel: ∆σ/σopt versus the S/N per A.

the galaxies with slitnumbers # 06 and # 20 to# 23 of the mask 3 setup. All galaxies have beenreduced and analysed following the same proce-dures (wavelength calibration and velocity dis-persion measurements), except the different slitextraction procedure. Fig. 5.15 displays the rel-ative differences of the velocity dispersions usingoptimally-weighted extractions and equal-weightextractions as a function of velocity dispersionand S/N per A. The aim of this Figure is to ver-ify whether the different extraction proceduresintroduce a systematic error in the velocity dis-persions of as a function of S/N , hence a system-atic under- or overestimation has to be tested.For this purpose, the median of the relative dif-ference ∆σ/σopt is the appropriate quantity inFig. 5.15. It is demonstrated that the varia-

tions between the velocity dispersion measure-ments are small, the relative difference betweenoptimally-weighted and equal-weight extractionsis on average (σopt − σeq)/σopt = 0.022 ± 0.093with a median of ∆σ/σopt = 0.038. The effectcan be regarded as negligible, as the mean uncer-tainty in the velocity dispersion δσ for a galaxyis larger than its difference in σ. Furthermore,for the aperture corrections the full range of ex-tracted rows was not used but the rows corre-sponding to the “FWHM” of the spatial profileonly.

5.3.6 Comparison between DifferentStellar Templates

In general, a homogenous library of stellar spec-tra is a key ingredient to measure the internalkinematics of a galaxy. To be able to inves-tigate stellar population effects it should coverthe largest possible range of spectral types en-compassed by early–type galaxies, covering gi-ant stars with spectral types A, F, G and K.In order to assess possible effects of the tem-plate stars on the derivation of velocity disper-sions, a different set of templates taken from theToulouse stellar library stelib was used (A stel-lar Library for stellar population synthesis mod-els1). This library contains 249 different stellarspectra within the visible wavelength range of3200 to 9500 A with intermediate resolution of<∼3 A FWHM and a sampling of 1 A therebyspanning a wide range in metallicities (Le Borgneet al. 2003). Stars of the same spectral type,thereby covering spectral types from A0 to K5,were compared. Overall, the measurements arein good agreement (<20 km s−1) regardless ofcomparing brighter or fainter objects. Moreover,as discussed in the previous sections, the FCQmethod has a low sensitivity on template mis-matching and for S/N > 10 per pixel velocitydispersions can be measured with a high degreeof accuracy.

1stelib http:\\webast.ast.obs-mip.fr\stelib\

index.html


Figure 5.16: Distribution of velocity dispersions forthe two rich clusters A 2218 and A 2390 in compari-son to the local samples of Jørgensen, Franx & Kjær-gaard 1995 and Jørgensen 1999 (J99). The com-plete histograms comprise 48 E+S0 galaxies in A 2218(thick solid line) and 48 E+S0s in A 2390 (shaded his-togram) and 115 early-type galaxies in Coma (J99).The vertical dotted line indicates the completenesslimit of the combined distant cluster samples.

5.4 Distribution in σ

Local data samples of early-type galaxies covera broad range in velocity dispersions as thekinematics of high– and low–luminosity galax-ies can be resolved. Going to higher redshifts,a magnitude–limited sample of galaxies gets af-fected by the so-called “Malmquist bias of the2nd kind” (Malmquist 1920; Teerikorpi 1997),where the magnitude limit cuts away parts of thefainter wing of the luminosity function. Due tothis “selection effect”, the detection rate of low–luminous galaxies becomes significantly lowerand therefore the overall data set relys only onthe most luminous objects, thereby resulting ina higher average luminosity for an intermediateredshift sample than for the local counterpartsample.For this reason, it is important to check whetherthe velocity dispersions of the distant E+S0 gal-axies show a distribution different from that oflocal elliptical and S0 galaxies. If indications

Figure 5.17: Distribution of velocity dispersionsfor the combined Low–LX clusters CL 0849, CL 1701and CL 1702 in comparison to the local samples ofJørgensen, Franx & Kjærgaard 1995 and Jørgensen1999 (J99). The complete histograms comprise 15E+S0 galaxies in CL 0849, five E+S0s in CL 1701,seven E+S0s in CL 1702 and 115 E+S0s in Coma(J99). The vertical dotted line indicates the com-pleteness limit of the combined distant cluster sam-ples.

of an evolution in the galaxy kinematics are de-tected, than these would have to be taken intoaccount in the attempts to quantify the evolu-tion in luminosity within the scaling relation ofthe FP and its projections (FJR and KR).Two different large sets of local comparison sam-ples were used for this purpose. For the clusters,a sample of local early-type Coma galaxies inthe Gunn r-band was chosen, whereas for thefield galaxies a reference of Coma galaxies and adata set comprising of Virgo and Coma galaxiesin Johnson B-band was used.

5.4.1 Cluster Samples

Jørgensen, Franx & Kjærgaard 1995 andJørgensen 1999 (hereafter collectively J99) per-formed a detailed study of 115 early-type Comagalaxies in the Gunn r-band. The combinedsample is divided into subclasses of 35 E, 55S0 and 25 intermediate types (E/S0). Absolute


Figure 5.18: Left panel: Distribution of velocity dispersions for the two rich clusters A 2218 and A 2390in comparison to the local samples of Jørgensen, Franx & Kjærgaard 1995 and Jørgensen 1999 (J99). Sub-sample of 17 ellipticals (E) in A 2218 and A 2390 (shaded histogram) and 35 Coma ellipticals (J99). Rightpanel: Distribution of velocity dispersions for the two rich clusters A 2218 and A 2390 in comparison to thelocal samples of Jørgensen, Franx & Kjærgaard 1995 and Jørgensen 1999 (J99). Sub-sample of 17 lenticular(S0/Sa) galaxies in A 2218 and A 2390 (shaded histogram) and 66 Coma S0/Sa galaxies (J99). The verticaldotted line indicates the completeness limit of the combined distant cluster samples.

magnitudes cover a range down to Mr < −20.02,which corresponds according to the authors toa completeness level of 93%. In oder to matchthe local J99 sample, the parameters of the dis-tant clusters were aperture corrected (cf. sec-tion 5.3.2). In the intermediate redshift range,observations of early-type galaxies are preferablyin the R or I filters. At z = 0.2, the observed Iand I814 passbands are very close to rest-frameGunn r. Therefore, the advantages of using theGunn r-band instead of the bluer Johnson V orB bands are the smaller k-corrections and thelower galactic extinction corrections.The distribution of velocity dispersions for thecomplete cluster samples of A 2218 and A 2390is compared to the local reference in Fig. 5.16.Fig. 5.18 shows the distributions in σ for thesub-samples of ellipticals (E) and lenticular (S0)galaxies in comparison to the local J99 sam-ples of E and S0 galaxies. The binning was setto 15 km s−1 in order to give a representativeview for both data sets. Overall, both distri-butions show a similar spread in σ. However,the distant samples failed to reproduce the gal-

axies with low velocity dispersions due to theselection upon apparent magnitudes. For thisreason, the Coma galaxies were restricted toMr < −20.42 and log σ > 2.02 (105 km s−1),which represent the completeness limit of thecombined A 2218 and A 2390 data. Both thedistant clusters and the local sample are fit-ted only within this region shown by a verti-cal dotted line in Fig. 5.16 and the subsequentplots. The velocity dispersions for the galaxiesare equally distributed. In particular, the σ dis-tribution of the unrestricted 115 local galaxieshas a median value of 〈σ〉J99u = 154 km s−1

and the restricted 97 early-type galaxies a me-dian of 〈σ〉J99r = 168 km s−1. Thedistant galaxies for A 2218 and for A 2390show medians of 〈σ〉A2218 = 179 km s−1 and〈σ〉A2390 = 165 km s−1, respectively. The com-bined cluster sample of 96 galaxies indicates amedian in σ of 〈σ〉Rich = 170 km s−1, whichis in very good agreement (differs by only 1%)with the restricted local comparison data. Asthe velocity dispersion is an indicator for themass of an object and the measured σ values


exhibit similar ranges, it can also be concludedthat there are no significant differences in meangalaxy masses between the two samples.

A one-dimensional Kolmogorov-Smirnov (KS)test (Press et al. 1992) gives a maximum devia-tion between the cumulative distribution func-tions of the combined distant and the localJ99 sample of D = 0.14 with a probability ofP = 0.34 for a larger value. The KS result yieldsa fairly good agreement, although the probabil-ity value is quite high.

In Fig. 5.17 the distribution of velocity disper-sions for the complete Low–LX cluster samplesof CL 0849, CL 1701 and CL 1702 is shown.As a local comparison, again the J99 data setis applied. To account for the completenesslimit of the combined poor cluster galaxies,the Coma galaxies were again restricted toMr < −20.42 and σ > 105 km s−1. Spectro-scopically confirmed cluster members comprise15, five and seven galaxies for CL 0849, CL 1701and CL 1702, respectively. Due to the lownumber of galaxies per cluster, a separationinto elliptical and S0 is not provided for eachindividual cluster. The velocity dispersionsfor the poor clusters show a very homogenousdistribution. In particular, the σ distributionfor CL 0849, CL 1701 and CL 1702 feature me-dian values of 〈σ〉CL 0849 = 195 km s−1,〈σ〉CL 1701 = 155 km s−1 and〈σ〉CL 1702 = 164 km s−1, respectively. Thecombined cluster sample of 27 galaxies indicatesa median in σ of 〈σ〉Poor = 164 km s−1, whichis lower by 4 km s−1 than the restricted J99sample and differs only by 2% from the localcounterpart. This is in very good agreementalthough the Low-LX clusters show a lack ofgalaxies around ∼200 km s−1. In addition, theσ distribution of the Low–LX cluster galaxies issimilar to the distribution covered by the richcluster sample. Their median σ values differ by6 km s−1 or by 4%.

For a comparison of the Low–LX sample with theJ99 data set in terms of statistics, the J99 was

reduced to the same number of 27 poor clustergalaxies, thereby accounting for the original J99coverage. A KS test yields a maximum devia-tion between the cumulative distribution func-tions of the combined poor cluster and the localJ99 sample of D = 0.15 with a probability ofP = 0.93 for a larger value. This result is of lessconfidence as the distant sample contains only 27objects and depends on the selection of the localcounterpart data set. Due to the small numbersfor the sub–samples of elliptical and lenticulargalaxies, a quantitative interpretation of thesedistributions in terms of a KS test is not appro-priate.

The sub-samples of ellipticals (E) and lenticular(S0) galaxies for the two rich clusters are similardistributed to the local E and S0 galaxies of J99,which is shown in Fig. 5.18. In total, the localsample comprises 35 E, 5 E/S0, 55 S0, 6 S0/Sa.The distant sub-sample with accurate structuralparameters provided by HST comprises in total34 E+S0 cluster galaxies. The clusters of A 2218and A 2390 are divided into sub-classes of 9 E,1 E/S0, 5 S0, 3 SB0/a, 1 Sa and 1 Sab spiralbulge and 8 E, 1 E/S0, 4 S0, 1 Sa, respectively. Adetailed morphological classification is presentedin section 4.4. For the ellipticals, only morpho-logically classified galaxies as E have been in-cluded. Intermediate types of E/S0, S0/Sa andSa galaxies are accounted for in the discy sub-sample of S0/Sa types. In particular, five E/S0are within the local J99 sample and only two inthe distant sample which have a negligible ef-fect on the overall distributions if the they areconsidered as elliptical types. Moreover, the his-tograms are not biased when excluding the sub–classes of S0/Sa galaxies. There are only six lo-cal S0/a galaxies and 3 Sa(b) types in the distantclusters. This is due to the fact, that actually thebulge of Sa galaxies is measured which is domi-nated by the chaotic random motion of the stars(i.e. by σ) rather than their rotational velocity.Therefore, the notation of Sa bulge is a muchmore appropriate term and will be used in the


Figure 5.19: Distribution of velocity dispersions forthe FDF and WHDF early–type galaxies in compar-ison to the local sample of Saglia, Bender & Dressler1993 (SBD93) and Dressler et al. 1987 (7S). The com-plete histograms comprise 24 E+S0 field galaxies inthe FDF, WHDF (shaded histogram) and 39 ComaE+S0 galaxies of SBD93 (solid line) and 59 Virgo andComa E+S0 galaxies of 7S (dotted histogram).

following. The poor cluster sample comprises 9 Ebut only one S0 galaxy. Therefore, a comparisonof sub–types cannot be conducted.Median σ values of the 17 E and 17 S0 galaxiesin the two rich clusters and the 35 E and 66 S0Coma galaxies are 〈σ〉Rich E = 208 km s−1,〈σ〉Rich S0 = 170 km s−1, 〈σ〉J99 E = 170 km s−1,〈σ〉J99 S0 = 154 km s−1, respectively. Overall,there is a good agreement between the distantand local sub–classes of elliptical and lenticulargalaxies. Elliptical galaxies in the rich clustersappear to have slightly larger velocity disper-sions. However, the J99 sample comprises noelliptical galaxies with σ > 260 km s−1.

5.4.2 Field Sample

For the 24 FDF and WHDF early–type field gal-axies encompassing on average higher redshiftsover a broader range in redshift than the clus-ters, the rest-frame Gunn r-band is not an ap-propriate choice and thus the magnitudes weretransformed to rest-frame JohnsonB. For a local

reference, a well studied sample of 39 early-typegalaxies (splitted into 25 E and 14 S0s) in theComa cluster (Saglia, Bender & Dressler 1993,hereafter SBD93) of the “7 Samurai” data set(Faber et al. 1989), and the original “7 Samu-rai” sample (Dressler et al. 1987, hereafter 7S)comprising 59 early-type galaxies in the Virgoand Coma cluster were chosen. No morphologi-cal information is available for the latter.

Fig. 5.19 displays the distribution of velocitydispersions for the complete FDF and WHDFfield ellipticals compared to the local referenceof SBD93 and 7S, respectively. Overall, thereis a good agreement between the two data set.The 13 FDF ellipticals feature a median in σ of〈σ〉FDF = 178 km s−1, the 11 WHDF ellipticalsexhibit a median of 〈σ〉WHDF = 150 km s−1,whereas the combined field sample shows a me-dian in σ of 〈σ〉FDF+WHDF = 178 km s−1. Themedian velocity dispersion in the local SBD93sample is 〈σ〉SBD93 = 162 km s−1 and for the 7Ssample 〈σ〉7S = 191 km s−1.

The 24 FDF and WHDF early–type field galax-ies are divided into 10 E and 12 S0. Two galaxiesare not visible on the ACS images and thereforenot included in either group. Analogous asfor the cluster galaxies, only morphologicallyclassified field elliptical galaxies were regardedas E type (cf. section 4.4). Fig. 5.20 gives adetailed look on the sub-samples of ellipticalsand lenticular field galaxies compared to thelocal E and S0 sample of SBD93, which com-prises 25 E and 14 S0/SB0 galaxies. Both, thefield elliptical and lenticular FDF and WHDFgalaxies show a similar distribution in σ to theirlocal counterparts. Median values for the distantfield E and S0 galaxies and the E and S0s ofthe SBD93 samples are 〈σ〉Field E = 200 km s−1,〈σ〉Field S0 = 170 km s−1,〈σ〉SBD93 E = 198 km s−1,〈σ〉SBD93 S0 = 150 km s−1, respectively.The agreement between the sub–classes isgood. Moreover, differences between the FDFand WHDF ellipticals and S0s are negligible.


Figure 5.20: Left panel: Distribution of velocity dispersions for the FDF and WHDF early–type galaxiesin comparison to the local sample of Saglia, Bender & Dressler 1993 (SBD93). Sub-sample of 10 ellipticals(E) in the FDF and WHDF (shaded histogram) and 25 Coma ellipticals (SBD93). Right panel: Distributionof velocity dispersions for the FDF and WHDF early–type galaxies in comparison to the local sample ofSaglia, Bender & Dressler 1993 (SBD93). Sub-sample of 12 lenticular (S0/Sa) galaxies in the FDF andWHDF (shaded histogram) and 14 Coma S0/SB0 galaxies (SBD93).

However, for this comparison it has to be takeninto account that observations of distant fieldgalaxies at higher z tend to miss low-luminousgalaxies which lower velocity dispersions dueto limitations in luminosity. Therefore, part ofslightly different σ distributions is not attributedto deviations between the samples but due to aselection effect.

On the basis of the complete cluster and fieldsamples as well as the sub–samples for ellipticalsand lenticulars it therefore can be concluded thatthe distant early-type galaxies do not indicate anevolution in the velocity dispersions. Note thatit is particularly important when comparing theevolution of the M/L ratios for distant and localearly–type galaxies, to reduce differences in theoverall mass distribution of the samples. Thispoint will be re–addressed in section 6.5. Fromthe comparison given above it is strengthenedthat the distant cluster and field galaxies con-stitute representative early–type galaxy samplesand deviations between local and distant scalingrelations will mainly have to be attributed to anevolution in the luminosity, but not to differences

in the kinematics.

5.5 Redshift Distribution

Galaxies’ redshifts were determined by two inde-pendent methods. A first measurement was per-formed in fitting Gaussian profiles to prominentabsorption line features of a typical early-typegalaxy. In a second approach, the redshift deter-minations were refined and improved using theFCQ method.After the wavelength calibration of the 2-D spec-tra and extraction of one–dimensional spectra,which were constructed by averaging all rowswithin the FWHM of the galaxies’ profile alongthe spatial axis (Y–axis), a search of absorptionline-strengths and possible emission lines wasperformed in a semi–automatic manner. As afirst redshift estimate, either the mean redshiftof a galaxy cluster or a by–eye identification ofcharacteristic absorption features was used. Aprogram displayed the most prominent absorp-tion lines of a typical elliptical galaxy simulta-neously with the galaxy spectrum to verify the


estimated value of the redshift z. For the deriva-tion of the exact redshift, Gaussian profiles werefitted to the identified absorption lines to com-pute the central wavelength of each feature. Thecatalogue of lines which was used herein encom-passed the Balmer series, [O II] 3727, CaII H+K,CN, G4300-band, [O III] 4959, [O III] 5007, Mgband Fe5015. For a few cases with very low S/Nspectra (S/N <∼ 8), a slight filtering of the av-eraged spectra in the order of the seeing of 1.5′′

FWHM for the line search was performed to sup-press the strong sky lines which dominated thespectral continuum. Typical uncertainties werein the range of 0.0005 <∼ z <∼ 0.0015, correspond-ing to 150 to 450 km s−1.

This redshift method was extensively tested onlow–resolution spectra of the 10 Low–LX clusters(Balogh et al. 2002b). Out of 581 galaxy spectrain total, reliable redshifts could been measuredfor 317 galaxies, of which 172 are cluster mem-bers. Results were compared to two independentmethods, the routine fxcor within the iraf

environment which cross-correlates the spectrawith high S/N z = 0 galaxy spectral templatesof similar resolution and to the FCQ methodwhich was applied to spectra with higher S/N.The agreement between the three measurementswas good and generally within the uncertainty(typically ∼100 km s−1). The independent es-timates allowed to identify galaxies where onemethod failed (e.g., because of low S/N ratioaround a critical line). Spectra for which thediscrepancy between different redshift estimatescould not be resolved were always of low S/Nratio (S/N<∼5%), and were rejected from furtheranalysis.

For the second approach, the FCQ technique wasapplied. As an input for an initial estimate, thedeterminations from the Gaussian profile mea-surements were adopted. Typical uncertaintiesfor the FCQ measurements were in the order of8 to 15 km s−1, thereby improving the first com-putation by a factor of 10 or even higher. Thesefinal FCQ output values were used for the sub-

sequent analysis. Both independent approacheswere consistent within the errors for all galaxieswithout any dependency on the environment ofthe sub-sample.

In total, redshifts of 121 early-type galaxies and9 spiral galaxies could be determined. Ad-ditional spectra of 12 secondary objects wereobtained, which were covered by a MOS slitby chance during the observations. The wholeearly-type galaxy sample is splitted into 52 E+S0galaxies for the A 2390 MOSCA setup, 30 E+S0sfor the three Low–LX clusters and 39 ellipticalfield galaxy candidates for the FDF and WHDF.No galaxy spectra of the targets had too low S/Nto derive a value of z. Three of the fill–up ob-jects for the poor cluster setup were bona fidespiral galaxies which featured a clear disc struc-ture on the ground-based images. These galaxiesshowed similar apparent luminosities as to thecluster galaxies and were therefore assumed tobe likely members of the targeted cluster whichwas the primary goal if it was not possible toselect any E+S0 galaxy as a MOS target basedon the low–resolution spectra. However, unfor-tunately non of these objects turned out to be acluster member.

Figure 5.21 shows the redshift distribution ofall 51 early–type rich cluster galaxies of A 2390with determined redshifts and one foregroundgalaxy. The cluster members are indicatedwith the shaded histogram and the foregroundobject (# 3038) is an E/S0 galaxy located atz = 0.1798. From the 55 different observed gal-axies, three objects, # 1507, # 1639 and # 2933,turned out to be background spiral galaxies atz = 0.3275, z = 0.3249 and z = 0.3981, respec-tively. The spiral # 2933 shows clear signs ofspiral arms on the HST/WFPC2 image of theA 2390 cluster. As this work concentrates onE+S0 galaxies all these spirals were discardedfrom the plot. During the observations of thetarget galaxies, four other galaxies fell into theslits by coincidence (# 2222, # 5552, # 5553 and# 6666). However, apart from # 6666, their


Figure 5.21: Redshift distributions of the full sam-ple of 51 A 2390 cluster galaxies (shaded histogram)and one foreground galaxy at z = 0.1798. Threebackground objects at z ≥ 0.32 have been neglected.The mean redshift of the cluster and the 2σ scatterare indicated as dashed and dotted lines, respectively.A Gaussian of width as the 1σ deviation is overlayed.

spectra are too faint to measure accurate ve-locity dispersions and no photometry is avail-able for the first three objects. Although thesegalaxies are included for the redshift distribu-tion of A 2390 in Figure 5.21, they were rejectedfrom the final sample. The mean redshift ofthe cluster zA2390 = 0.2305 ± 0.0094 (median〈zA2390〉 = 0.2300) and the 2σ deviation de-rived from the A 2390 sample are indicated asdashed and dotted lines, respectively. The clus-ter redshift is in good agreement with the valueof zA2390 = 0.2280 derived from the CNOC studyby Yee et al. (1996). For the further analysis al-ways the CNOC redshift is adopted as this valueis based on a larger cluster sample of 217 galax-ies (within zclus±2σ), splitted into 159 E/S0, 13E+A, 31 Sa–Sbc and 14 emission line (Irr) galax-ies, and thus encompasses also the spiral galaxypopulation of A 2390. In Fig. 5.21 a Gaussian ofwidth as large as the 1σ deviation is overlayed.

Figure 5.22: Redshift distributions of the full sam-ple of 15 Cl 0849 cluster galaxies (shaded histogram)and one foreground galaxy at z=0.2090. The medianredshift of the cluster and the 2σ scatter are indicatedas dashed and dotted lines, respectively. A Gaussianof width as the 1σ deviation is overlayed.

The redshift distributions of the Low–LX gal-axy clusters are shown in the Figs. 5.22,5.23 and 5.24. Shaded histograms denotethe distributions of cluster members for 15early-type galaxies in Cl 0849, 5 E+S0 gal-axies in Cl 1701 and seven E+S0s in thecluster Cl 1702. Non cluster members areindicated as the open histogram. Again,the mean cluster redshifts are labelled withz0849 = 0.2346± 0.0030, z1701 = 0.2452± 0.0015and z1702 = 0.2228± 0.0025. The 2σ deviationsof the mean cluster redshift values are indicatedas dotted lines. In Fig. 5.24 the selection of clus-ter members for the Low–LX cluster data set isillustrated. The redshift distribution of the com-plete sample of 18 galaxies which were targetedwith a single MOS mask for the clusters Cl 1701and Cl 1702 are shown. Again, the median red-shift of both clusters are denoted as dashed lines.These values together with their 2σ deviations(dotted lines) are based on the larger samples of


Figure 5.23: Redshift distributions of the fullsample of 5 Cl 1701 and 7 Cl 1702 cluster galaxies(shaded histogram) and non cluster members (openhistogram). The median redshift of the clusters andtheir 2σ scatter are indicated as dashed and dottedlines, respectively. A Gaussian of width as the 1σdeviation is overlayed.

galaxies with low–resolution spectra which alsoaccount for cluster spiral galaxies.In Fig. 5.25 the redshift distribution of the FDFand WHDF field early–type galaxies is displayed.Spectroscopic confirmed elliptical and S0 gal-axies are shown as shaded histogram, whereasone background spiral galaxy at z = 0.5569 ismarked with an open histogram. The 13 FDFE+S0 field galaxies encompass a range in red-shifts of 0.22 ≤ zFDF ≤ 0.65 with a meanof zFDF = 0.42 ± 0.11 (〈zFDF〉 = 0.41) andthe 11 WHDF field galaxies cover a redshiftspace of 0.21 ≤ zWHDF ≤ 0.74 with a mean ofzWHDF = 0.36 ± 0.16 (〈zWHDF〉 = 0.40). Thetotal field sample of 24 E+S0 galaxies exhibitsa mean redshift of zField E = 0.39 ± 0.14 and amedian of 〈zField E〉 = 0.40.It turned out that the southwestern corner of theFDF most probably covers the outskirts of a gal-axy cluster at z = 0.33. Based on the radial ve-

Figure 5.24: Redshift distribution of the completesample of 18 galaxies which where targeted with aMOS mask for the Cl 1701 and Cl 1702 cluster. Themedian redshift of Cl 1701 and Cl 1702 are indicatedas dashed and dotted lines, respectively.

locity measurements for the elliptical galaxy can-didates in the FDF, the lower limit for the veloc-ity dispersion of the cluster is σc >∼ 430 km s−1.The cluster velocity dispersion σc is only a lowerlimit as the cluster centre is not located on theFDF but only its outskirts. Assuming a spreadin redshift space of ∆z = 0.01, which corre-sponds to ∼ 1.5 times the typical velocity disper-sion of a rich galaxy cluster (σc ≈ 1000 km s−1),it is found that a total of 13 spiral galaxies and 15early–type galaxies based on their spectroscopicredshifts are likely to be members of this clus-ter, which is also visible through a small “gap”in redshift space at z ≈ 0.3.The spatial distribution of all early-type fieldgalaxies in the FDF is shown in Fig. 5.26. Po-tential cluster members are additionally denotedby squares. Adding the possible spiral clus-ter candidates, it is suggestive that the clusterhas an elongated shape. If the center indeedis positioned to the southwestern corner of theFDF, the distribution of the galaxies will be a


Figure 5.25: Redshift distribution of the 24 fieldearly–type galaxies within the FDF and WHDF.Spectroscopic confirmed E and S0 field galaxies aredenoted with a shaded histogram. One backgroundspiral galaxy at z = 0.5569 is plotted as an openhistogram. The median redshift of for the combinedfield sample 〈z〉 = 0.4 is indicated as a dashed line.

striking example of the morphology–density rela-tion (Dressler 1980) with the early–type galaxiesmainly populating the dense, inner region andthe late–type galaxies at larger clustercentricradii. Unfortunately at a redshift of z = 0.33, theMg 5170 absorption line is strongly affected dueto the terrestrial absorption of the B band, whichmakes an accurate measurement of the internalgalaxy velocity dispersions impossible. For thisreason, the early-type galaxies which are possi-ble members of a cluster at z ≈ 0.33 will bediscarded from the further analysis.Fig. 5.27 illustrates the spatial distribution of theearly-type field galaxies in the WHDF. Apartfrom three galaxies which are located in thesouthern (upper middle) part of the image, allfield ellipticals are homogenous distributed overthe field. Based on the radial velocity measure-ments for the three galaxies it turned out thatthese galaxies exhibit a dispersion in radial veloc-

Figure 5.26: Distribution of the field early–typegalaxies within the FDF and possible members of acluster in the southwestern (lower right) corner. 15early–type cluster candidates are denoted addition-ally by squares, the 17 early–type field galaxies solelyby circles. The bar corresponds to 500 kpc in projec-tion at the clusters redshift of z = 0.33.

ities of ∆vrad ≈ 450 km s−1. Furthermore theyare not visually associated to each other throughtidal arms or interaction signatures on the ACSimages. For this reason, the three objects are arenot likely members of a group at z ≈ 0.21 butisolated early-type field galaxies.

5.6 Comparison between RichClusters

To investigate possible environmental effects inthe galaxy properties and differences in the stel-lar populations of elliptical and lenticular galax-ies, a large sample of more than 50 early–typegalaxies is required. In a recent study, Ziegleret al. (2001a) performed a detailed analysis of asample of 48 early-type galaxies in the rich clus-ter A 2218 at a redshift of z = 0.18. If this work


Figure 5.27: Distribution of the field early–typegalaxies within the WHDF. The bar corresponds to600 kpc in projection at a redshift of z = 0.33. Thethree galaxies located in the southern (upper mid-dle) part of the WHDF are not members of a group.North is down and east is to the left.

could be combined with the sample of N = 48 inA 2390, it would offer the possibility to explorethe evolution of ∼100 early-type galaxies over alarge luminosity range and wide field-of-view atsimilar cosmic epochs.Before the individual galaxies of the two clus-ters can be combined into one large sample, ithas to be proven that both data sets are charac-terised by very similar properties and that pos-sible differences within the sub-samples were ac-counted for. For this reason, the cluster galax-ies were compared with respect to their luminos-ity, colour, size and velocity dispersion distribu-tions. Both clusters feature similar global clusterproperties (e.g., richness class, X-ray luminos-ity). Furthermore, the sample selection and allobservations, especially the spectroscopy, havebeen carried out in very similar manner. There-fore, it is not expected that significant differencesbetween the galaxy samples exist.

In the subsequent analysis the cluster galaxiesare investigated over the same dynamical rangein their properties. Therefore, the two cD gal-axies of the cluster A 2218 were excluded as nocD galaxy is included in the A 2390 sub-sample.A comparison of the cluster properties in ab-solute rest-frame Gunn r magnitude, effectiveradius, (B − I) rest-frame colour and velocitydispersion (aperture corrected) is presented inFig. 5.28. For each set of parameters, the meanvalues with the resp. ±1σ scatter are indicatedas overlayed Gaussian curves. Overall, the galax-ies are similarly distributed in all plots. For thegalaxies in A 2218 the range of absolute Gunn rmagnitudes is −20.50 ≥ Mr ≥ −23.42, for theobjects in A 2390 −20.47 ≥ Mr ≥ −22.99 (up-per left panel in Fig. 5.28) The median valuefor A 2390 is 〈Mr〉 = −21.31m, 0.17 mag fainterthan the median luminosity for the galaxies inA 2218. For the size distribution only the 32objects within the HST fields are considered.In the distributions of galaxy size (upper rightpanel) small (but not significant) differences arevisible. The sizes of the A 2218 galaxies covera range between 0.22 and 0.82 in logRe (kpc),with a median of log 〈Re〉 = 0.46. The A 2390sample contains more galaxies with smaller effec-tive radii 0.01 ≤ logRe ≤ 0.92, with a mean oflogRe = 0.38±0.27 (median log 〈Re〉 = 0.37), re-sulting in a broader distribution than for A 2218.These galaxies are low-luminosity galaxies. Thisissue will be addressed in more detail in theforthcoming section 6.4. The rest-frame colourdistributions show no significant differences be-tween the clusters. The galaxies’ colours inA 2390 cover a range of 1.86 < (B − I) < 2.51,in A 2218 1.95 < (B − I) < 2.44. For the A 2390galaxies a median value of 〈(B − I)〉 = 2.29 isderived, for the A 2218 objects 〈(B− I)〉 = 2.28,respectively. Both are in very good agreementwith the typical colour of (B − I) = 2.27 forellipticals at z = 0, given by Fukugita et al.(1995). The velocity dispersions for the galax-ies are equally distributed (lower right panel),


with a median value of 165 km s−1 and a meanof log σ = 2.238±0.11 for A 2390 and 178 km s−1

(log σ = 2.253 ± 0.12) for A 2218. As the veloc-ity dispersion is an indicator for the mass of anobject and the measured σ values exhibit similarranges, it can also be concluded that there are nosignificant differences in mean galaxy masses be-tween the two samples (see section 6.5 for a fur-ther discussion). In addition, the scale lengths(h), disc-to-bulge ratios (D/B) and the surfacebrightnesses of the member galaxies were com-pared to each other and again negligible differ-ences between the distributions were found.As a conclusion no significant offset in the dis-tribution of any galaxy parameters between thetwo samples was detected. The comparison as-serted that properties of the galaxies within thetwo clusters are very homogeneous and thereforea combination of the two data sets is adequateand warranted, which results in a final sample of96 early-type galaxies.The combined sample allows an extensive in-vestigation of the evolutionary status of galax-ies in rich clusters at z ∼ 0.2. A sub-samplewith accurate structural parameters provided byHST comprises 34 E+S0 galaxies, splitted into17 ellipticals (E), 2 E/S0, 9 S0, 3 SB0/a, 2 Sabulges and 1 Sab spiral bulge that can be in-vestigated in the FP. With this large sample,possible radial and environmental dependencescan be explored in detail for the galaxy proper-ties from the cluster centre to the outskirts us-ing the Faber-Jackson relation and for differentsub-populations (E/S0 and S0/Sa bulges). Forthe subsequent analysis, the sample of early-typegalaxies in A 2218 and A 2390 is referred to as therich cluster sample.


Figure 5.28: Comparison of galaxy properties. Thin lines and hashed areas represent the distribution forthe members in A 2390, solid thick lines the characteristics of the galaxies in A 2218. Gaussian fits showingthe mean values with ±1σ scatter are overlayed. Top left: Absolute Gunn r magnitudes for the wholesample. Top right: Size distribution of the HST sub-sample. Bottom left: (B − I) rest-frame colours ofmember galaxies. Bottom right: Distribution of velocity dispersions.

Chapter 6: Galaxy Scaling Relations at z∼ 0.2 127

Chapter 6

Galaxy Scaling Relations at z∼ 0.2

After the descriptions of the sample construc-tion and the derivations of photometric proper-ties, structural parameters and the measurementof velocity dispersions, the analysis and resultsof the distant scaling relations of early-type gal-axies will be presented.

In a first step, a sample of nearby early-type gal-axies has to be chosen which acts as a local com-parison reference (section 6.1). Afterwards thescaling relations of the projections of the Funda-mental Plane, the Faber–Jackson (section 6.2)and Kormendy relation (section 6.3) of the dis-tant cluster and field early-type galaxies will becompared to the local ones in the present-dayuniverse. In a next step, the Fundamental Plane(FP) is outlined in section 6.4. After an investi-gation of the whole early–type galaxy populationin the FP, the discussion concentrates on possi-ble differences between the morphological typesof early-type galaxies, elliptical and S0 galaxies.To rule out systematic errors in the interpreta-tion of the evolution within the FP, the massranges of the galaxies are analysed and the evo-lution of the mean mass–to–light (M/L) ratiosare explored (section 6.5). A comparison withprevious studies of the FP is given in section 6.6.Detailed tests on possible environmental effectson the properties of E+S0 galaxies will be de-scribed in the next chapter 7.

Some aspects of the analysis in this chapter havebeen presented in Fritz et al. (2005a, 2005b) andZiegler et al. (2005).

6.1 The Local Reference

To derive the kinematic and/or the spectropho-tometric evolution of early-type galaxies at inter-mediate redshift, it is crucial to carefully choosea sample of early-type galaxies at low redshiftwhich can be utilised as a reference for the pur-pose of comparison. As already discussed inchapter 1, a large number of local samples havebeen constructed based on the Virgo and Comaclusters. Recent results on the SDSS Data Re-lease 1 database encompass maybe one of thelargest sample of early-type galaxies located atdifferent redshifts and in various environments.However, unfortunately at the time of this study,the SDSS sample was not yet electronically avail-able.The Coma cluster at z = 0.024 is one of thebest studied local rich clusters. The cluster red-shift of z = 0.024 corresponds to a look–backtime of only ∼0.3 Gyrs, or 2% of the age of theUniverse, and therefore galaxies within this limitcan be considered to represent the same cosmicepoch, i.e., the present-day Universe. A numberof previous works in the literature (Saglia et al.1993b; van Dokkum & Franx 1996; Jørgensenet al. 1999; Kelson et al. 2000b; Treu et al.2001b; Ziegler et al. 2001a; Holden et al. 2005)used this cluster for the purpose of comparison ofthe Fundamental Plane at z > 0.1 and it there-fore provides a reliable and widely accepted lo-cal reference when addressing evolutionary ques-tions. To achieve a direct comparability with

128 Chapter 6: Galaxy Scaling Relations at z∼ 0.2

the results of these authors, the Coma data byJørgensen 1999 and Jørgensen et al. 1995 (here-after collectively J99) will be used as a local com-parison sample for the following analysis of theearly-type cluster galaxies. These authors per-formed a detailed study of a large number ofearly-type Coma galaxies in Gunn v, g and r-bands and Johnson U and B filters. The com-bined sample comprises 115 early-type galaxies,divided into subclasses of 35 E, 55 S0 and 25intermediate types (E/S0). Note that J99 didnot find any differences between elliptical andS0 or Sa bulge galaxies. For this reason, thelocal early-type galaxies are not splitted intodifferent sub-classes based on their morpholo-gies. Absolute magnitudes cover a range downto Mr < −19.58 and the sample is 93 % com-plete at absolute magnitudes Mr < −20.02, cor-responding to r = 15.08m. In oder to matchthe local J99 sample, the parameters of the dis-tant clusters were aperture corrected (cf. sec-tion 5.3.2). In the intermediate redshift range,observations of early-type galaxies are preferablyin the R or I filters. At z = 0.2, the observed Iand I814 passbands are very close to rest-frameGunn r. For this reason, the advantages of usingthe Gunn r-band instead of the bluer Johnson Vor B bands are the smaller k–corrections and thelower Galactic extinction corrections.

In all subsequent figures, large symbols denotethe distant early-type galaxies, either of thetwo rich clusters Abell 2218 and Abell 2390 orthe three Low–LX clusters Cl 0849, Cl 1701 andCl 1702, whereas small boxes represent the lo-cal reference sample. Morphologically classifiedlenticular galaxies (S0) or bulges of early-typespiral galaxies within the HST field of the clus-ters are indicated by open symbols, elliptical gal-axies are denoted by filled symbols. Under as-sumption of a non changing slope, the residualsfrom the local relation are computed and derivea mean evolution for the distant galaxies. Fora comparison of the slope of the local and dis-tant galaxies, the bisector method is used (e.g.,

Figure 6.1: The local Faber–Jackson relation inGunn r for 115 early-type galaxies in the Coma clus-ter from Jørgensen (1999, J99) and Jørgensen et al.1995, J99). The individual objects of the J99 sampleare denoted by the small squares. The dot-dashedline is a bisector fit to the local FJR within the se-lection boundaries (dashed lines) as defined by thedistant cluster samples, together with the 1 σ errors.

Ziegler et al. 2001a), which is a combination oftwo least-square fits with the dependent and in-dependent variables interchanged. The errors onthe bisector fits were evaluated through a boot-strap re–sampling of the data 100 times.The early–type field galaxies in the FDF andWHDF cover a broader redshift range than theclusters and on average higher redshifts. A trans-formation of the measured magnitudes to theGunn r rest–frame band is not appropriate asit would involve large k–corrections (cf. sec-tion 4.5.3). For this reason, as a local refer-ence the well studied Coma sample in the John-son B-band by Saglia, Bender & Dressler (1993,hereafter SBD93) comprising 39 early-type clus-ter galaxies (splitted into 25 E and 14 S0s) waschosen. This data compilation is a sub–set of theoriginal “7 Samurai” sample (Faber et al. 1989;Dressler et al. 1987) which contains 59 early-type galaxies in the Virgo and Coma cluster (cf.


Figure 6.2: Faber–Jackson relation in Gunn r for 96early-type galaxies in A 2390 and A 2218, comparedto the local Coma sample of J99. The solid lines showthe ±1 σ errors of the 100 iteration bootstrap bisectorfits to the local FJR within selection boundaries, thehashed area indicates the bisector fits to the distantsample (within ±1 σ).

section 5.4.2). As for the latter no morphologi-cal information is available the Coma study bySBD93 was used as a local reference.

6.2 The Faber–Jackson Rela-tion

A first test of the formation and evolution of el-liptical and lenticular galaxies allows the scalingrelation between galaxy luminosity and velocitydispersion, the so-called Faber–Jackson relation.Fig. 6.1 illustrates the local Faber–Jackson re-lation (FJR) in Gunn r-band for 115 early–typegalaxies in the Coma cluster by J99. The distantearly–type galaxy samples in this study wereall selected upon apparent magnitudes and aretherefore limited in their absolute magnitudes.For this reason, the local FJR has to be re-stricted to the range in magnitudes and veloc-ity dispersions covered by the distant samples.

The distant galaxies of the combined A 2218 andA 2390 clusters have velocity dispersions down to105 km s−1 and the absolute magnitudes rangedown to Mr ≤ −20.42. To match the selectionboundaries for the distant sample represented bythe combined A 2218 and A 2390 data, the Comagalaxies by J99 were restricted to Mr < −20.42and log σ > 2.02, which reduced the numberof early-type galaxies in the Coma cluster to77. These selection boundaries are indicated asthe dashed lines in Fig. 6.1. In case of the lo-cal Kormendy relation (KR), galaxies which donot meet these criteria are omitted and shownas open squares (see section 6.3). For the subse-quent analysis of the FJR, both the distant clus-ters and the local sample are fitted only withinthe region shown by horizontal and/or verticaldashed lines.A bootstrap bisector fit to the reduced Comareference sample yields

Mr = −(6.82± 0.90) log σ− (5.90± 2.01). (6.1)

The observed ±1σ scatter is σr = 0.57m. Thebisector fit together with the average ±1 σ scat-ter of the local FJR are denoted in Fig. 6.1 asthe dot-dashed line and the dotted lines, respec-tively. For purposes of comparison, a linear χ2-fitto the reduced Coma data set gives

Mr = −(5.19± 0.54) log σ − (9.43± 1.21) (6.2)

and a flatter slope than the bisector method.The Faber–Jackson relation (FJR) in the Gunnr-band for 96 distant early-type galaxies in theclusters A 2218 and A 2390, compared to the lo-cal Coma sample of J99 is shown in Fig. 6.2. Inorder to consider the whole cluster data set forthe Faber–Jackson scaling relation, the luminosi-ties of the distant galaxies as derived from theground–based photometry with SExtractor areapplied. Very similar trends are found when thesample is restricted to the space–based magni-tudes. For the construction of the FundamentalPlane, accurate measurements of the sizes of thegalaxies are required which can only be achieved


via the high–resolution images of the HST (cf.section 6.4). In the parameter space of the FJRdiagram, the distant galaxies in the two rich clus-ters encompass a range in absolute magnitude of−20.42 ≥ Mr ≥ −23.42 and in velocity disper-sion of 105 km s−1≤ σ ≤ 381 km s−1. In Fig. 6.2,the ±1 σ scatter of the bisector fits to the localFJR is indicated as solid lines. Fits to the re-spective distant combined cluster sample (within±1 σ) are shown as the hashed area. Becauseof the younger mean ages of the stellar popula-tions, the distant cluster galaxies are on aver-age brighter than their local counterparts for agiven velocity dispersion. If a formation redshiftof zform = 2 for the cluster galaxies is assumed,the predictions of single-burst passive evolutionmodels by BC96 (Salpeter IMF with x = 1.35,mass range 0.10 < M < 125.0 and burst dura-tion of 1 Gyr) suggest an increase of their Gunn rbrightness by 0.21±0.05 mag at z = 0.2. Assum-ing that the local slope holds valid for the distantgalaxies, the mean residuals from the local FJRare analysed, i.e. the difference in luminosity foreach galaxy from the local FJR fit. The ±1σscatter of these offsets for the distant galaxies isσr = 0.61m, which is only by 0.04m or 7% largerthan the dispersion observed locally.Table 6.1 lists the derived mean and median lu-minosity evolution of the early–type cluster gal-axies for various samples in the FJR. The finaltotal error in the FJR evolution results from alinear sum of the total error in absolute magni-tude δMr (cf. Eq. 4.14) and the error as intro-duced by the velocity dispersions as

σMr = δMr + a log δσerr, (6.3)

where δσerr denotes the mean error in the ve-locity dispersion and a the bisector slope of theA 2218 and A 2390 sub-samples. Typical uncer-tainties of the velocity dispersions can be foundin Table 4.5. The error in the mean luminos-ity evolution σMr in Eq. 6.3 is given as the ±1σuncertainty in the column 4 (σMr) of Table 6.1.This mean error in the measured evolution willbe adopted in all subsequent derivations of the

Table 6.1: Evolution of the Faber–Jackson relationin Gunn r derived for various cluster samples. N isthe number of galaxies and ∆Mr indicates the meanluminosity evolution. The fourth column denotes the±1σ deviation in the mean luminosity evolution andthe last column ∆ 〈Mr〉 gives the median evolution.

Sample N ∆M r σMr ∆ 〈Mr〉[mag] [mag]

(1) (2) (3) (4) (5)

A 2218 48 −0.31 0.15 −0.29A 2390 48 −0.32 0.29 −0.35A 2218+A 2390 96 −0.32 0.22 −0.35Cl 0849 15 −0.30 0.19 −0.36Cl 1701 5 −0.42 0.19 −0.20Cl 1702 7 −0.57 0.20 −0.51Low–LX a 26 −0.39 0.19 −0.36Low–LX b 27 −0.44 0.19 −0.38

a as A 2218+A 2390: logσ > 2.02, Mr < −20.42.b with # 923: logσ > 2.02, Mr < −20.08.

amount of luminosity evolution and is listed inthe Tables as σMr . For the total sample of 96galaxies in the FJR a luminosity evolution of−0.32 ± 0.22m is detected. Similar offsets arededuced for each cluster separately. The early-type galaxies in A 2218 show an evolution of−0.31 ± 0.15m, whereas the E+S0 galaxies inA 2390 are brighter by −0.32 ± 0.29m. Both re-sults agree with the BC96 model predictions.Under assumption of a variable slope, Fig. 6.2compares the FJR of the distant clusters A 2218and A 2390 to the local Coma reference. To in-vestigate the significance of a possible slope de-crease a bootstrap bisector fit was performed,which are 100 iterations of repeated bisector fitswith random rejection of data points betweeneach iteration. In this method, the error can bederived from the scatter of the individual slopesof the fits. A bootstrap bisector fit to the distantcluster sample of A 2218 and A 2390 yields

Mr = −(5.23±0.43) log σ− (9.78±0.96), (6.4)


Figure 6.3: Faber–Jackson relation in Gunn r for27 early-type galaxies in the poor clusters Cl 0849,Cl 1701 and Cl 1702, compared to the local Comasample of J99. The linear bisector fit to the Comagalaxies was restricted to the range encompassed bythe distant rich cluster galaxies (Mr < −20.42 andlog σ > 2.02). The area bounded by dotted lines in-dicates the mean ±1σ of the local FJR. The solid lineshows the mean evolution of the poor cluster sampleassuming the local slope.

which is illustrated as the hashed area in Fig. 6.2.In this first analysis, a slope change is significanton the 2σ level based on the uncertainty of thelocal Coma slope (see Eq. 6.1).To verify this result with a different fitting tech-nique, a linear χ2-fit to the combined rich clustersample was computed as

Mr = −(3.76±0.45) log σ−(13.09±1.00). (6.5)

In this approach, a slope change would be sig-nificant on the 3σ level again using the errorestimate of the local Coma slope in Eq. 6.2.Both fitting methods give evidence that a mass–dependence evolution drives the change of theslope in the FJR. To quantify this in more detail,the rich cluster sample was splitted with respectto velocity dispersion into two equally large sub–samples and analysed separately, which will be

Figure 6.4: Faber–Jackson relation in Gunn r for27 early-type galaxies in the poor clusters Cl 0849,Cl 1701 and Cl 1702, compared to the local Comasample of J99. The dotted lines show the ±1 σ errorsof the 100 iteration bootstrap bisector fits to the localFJR within selection boundaries, the solid lines indi-cate the bisector fits to the distant sample (within±1 σ).

addressed in section 7.1.2.The Faber-Jackson Relation considers the totaldistant sample of 96 early-type cluster galaxies.For this reason, this scaling relation offers thepossibility to perform a variety of tests in divid-ing the sample with respect to luminosity, ve-locity dispersion (which is a representative par-ameter of the underlying mass of a galaxy) anddistance from the cluster centre. These extensiveanalyses will be discussed in the next chapter 7.Fig. 6.3 displays the Faber–Jackson relation inGunn r for the three Low–LX clusters Cl 0849,Cl 1701 and Cl 1702. The combined poor clustersample comprises in total 27 early–type galaxies,divided into 15 galaxies in Cl 0849, 5 in Cl 1701and 7 in Cl 1702. These numbers are also shownin brackets next to the cluster name in Fig. 6.3.Again, the linear bisector fit to the Coma galax-ies was restricted to the range encompassed bythe distant rich cluster galaxies (Mr < −20.42


and log σ > 2.02). Under assumption of a non–changing local slope with redshift, the meanresiduals from the local FJR of the poor clus-ter sample can be investigated and their meanevolution measured. The offsets of the distantLow–LX galaxies have an observed ±1σ scatterof σr = 0.65m. The dispersion is 14% larger thanthe scatter of the local Coma cluster and similarto the dispersion as for the rich clusters. A differ-ence in the scatter between the distant samplesis therefore rather unlikely. Restricting the localComa galaxies to log (σ) > 2.02, Mr < −20.08,for the total sample of 27 galaxies in the FJR anaverage luminosity evolution of −0.44 ± 0.19m

is derived. However, this result depends on theassumed selection boundaries. If the faintest gal-axy (ID # 923) is excluded and the Coma sampleis restricted to log (σ) > 2.02, Mr < −20.42, anevolution of −0.39±0.19m for the Low–LX clus-ter galaxies is found. In Table 6.1 a summary ofthe measured mean and median luminosity evo-lution of the Low–LX clusters is given.Accounting for the possibility that the distantpoor cluster sample and the local reference havedifferent slopes, the FJR is analysed using bi-sector fits in Fig. 6.4. To explore a change inthe slope, a bootstrap bisector fit to the distantcluster sample gives

Mr = −(5.58± 0.71) log σ− (9.08± 1.61). (6.6)

The ±1 σ errors of this fit are denoted as thesolid lines in Fig. 6.4. This approach impliesthat a slope change would be significant on the2σ level, which suggests again a stronger evolu-tion for lower–mass galaxies. A further discus-sion on a possible mass–dependence will be thetopic of section 7.1.2. In the following sectionthe Kormendy relation will be discussed.

6.3 The Kormendy Relation

To study the change of surface brightness evo-lution for the FP sample at a fixed size, themagnitude–size relation was constructed. The

Figure 6.5: The local Kormendy relation in theGunn r-band for 115 early-type galaxies in the Comacluster from Jørgensen (1999, J99) and Jørgensen etal. 1995, J99). The individual objects of the J99sample are denoted by the small solid squares andthe open squares are excluded low–mass galaxies tomatch the distant cluster sample. The dot-dashedline is a bootstrap bisector fit to the local KR withinthe selection boundaries as defined by the distantcluster samples, together with the 1 σ errors (dashedlines). The error bar of µe and log (Re) is indicatedin the upper right corner.

Kormendy relation (KR) represents the projec-tion of the Fundamental Plane along the veloc-ity dispersion onto the photometric plane. Asonly the structural parameters are used, sampleshave larger scatter in the KR than in the FP butresults should complement and endorse findingsobtained with the Fundamental Plane.Similar as for the FJR, the J99 Coma samplewas restricted to the range in absolute mag-nitude and velocity dispersion encompassed bythe combined rich cluster galaxies in A 2218 andA 2390 with lower limits of Mr = −20.42 andlog σ = 2.02. Galaxies in the local Kormendy re-lation which do not meet these criteria are omit-ted and shown as open squares in Fig. 6.5. Onlygalaxies denoted as solid symbols are utilised for


Figure 6.6: Kormendy relation in the Gunn r-bandfor A 2390 and A 2218, compared to the Coma sampleof J99. The area bounded by solid lines indicatesthe ±1 σ errors of the bisector fits to the local KRrelation and the hashed region shows the bisector fitsto the distant sample within ±1 σ. A typical errorbar (mean error) is indicated in the upper right corner(see text for details).

the investigation of the KR. From Fig. 6.5 it canbe seen that these neglected objects are predomi-nantly low–surface brightness galaxies which areabsent in the distant sample due to limitations inluminosity. If these faint galaxies were includedfor the KR analysis, the derived luminosity off-set of the distant sample would be to large andtherefore the evolution too strong.The Kormendy relation in Gunn r for the gal-axies in A 2390 and A 2218 is shown in Fig. 6.6.The structural parameters of effective radius Re

and mean surface brightness within Re, 〈µe〉, arebased on curve of growth fits and derived usinga combination of exponential disc plus r1/4-lawsurface brightness models. The errors in log Re

and 〈µe〉 are correlated. In the upper right cor-ner an average error which enters the KR inthe coupled form log Re − 0.328 〈µe〉 = 0.08is displayed, assuming a typical uncertainty of

Figure 6.7: Kormendy relation in the Gunn r-bandfor A 2390 and A 2218, compared to the Coma sampleof J99. The KR was constructed using solely r1/4-law structural parameters. The dot-dashed line isa bisector fit to the local KR within the selectionboundaries together with the mean 1 σ errors (dashedlines). The solid line shows the mean evolution of therich cluster sample assuming a constant local slope.

∆ 〈µe〉 = 0.05 and ∆ logRe = 0.1. In a similarmanner as for the FJR, the luminosity evolutionfor the distant galaxies in the KR is computed.Within the selection boundaries, a bisector fit tothe Coma sample gives

〈µe〉 = (3.01±0.23) logRe+(18.41±0.10), (6.7)

with an observed scatter of σKR = 0.48m in 〈µe〉.The ±1 σ scatter of the bisector fits to the lo-cal KR is shown as solid lines in Fig. 6.6. Forthe distant clusters of of A 2218 and A 2390 thebisector method yields

〈µe〉 = (3.03±0.17) logRe+(18.01±0.09), (6.8)

which suggests that there is no significant slopechange in comparison to the local reference. Theaverage scatter of the bisector fits within ±1 σto the rich clusters are indicated as the hashedarea in Fig. 6.6.


Figure 6.8: Kormendy relation in the Gunn r-bandfor the Low–LX clusters Cl 0849, Cl 1701 and Cl 1702,compared to the Coma sample of J99. The areabounded by dotted lines indicates the ±1 σ errorsof the bisector fits to the local KR relation and thesolid lines shows the bisector fits to the distant sam-ple within ±1 σ. A typical error bar (mean error) isindicated in the upper right corner.

In order to examine the offsets of the distantcluster sample from the local KR again a con-stant slope with look–back time is assumed. Thescatter of the observed offsets for the A 2218and A 2390 galaxies is σKR = 0.39m. Compar-ing the galaxies within their observed magnituderange 17.92 ≤ 〈µe〉 ≤ 22.89 with the reducedComa KR, a larger luminosity evolution thanwith the FJR is found. In particular, in the rest-frame Gunn r the 34 cluster galaxies are on av-erage brighter by −0.39 ± 0.27m (median value−0.46) compared to Coma. The two bright cDgalaxies in A 2218 at log Re ≈ 1.45 kpc andlog Re ≈ 1.75 kpc follow the extension of theKR as defined by the distant cluster galaxies.The KR depends on the assumed surface bright-ness profiles for the derivation of structural pa-rameters of half-light radius Re and mean surfacebrightness 〈µe〉, measured within the effective ra-

Figure 6.9: Kormendy relation in the Gunn r-bandfor the Low–LX clusters Cl 0849, Cl 1701 and Cl 1702,compared to the Coma sample of J99. The areabounded by dotted lines indicates the ±1 σ errorsof the bisector fit to the local KR relation and thesolid lines shows the bisector fit to the distant sam-ple within ±1 σ. A typical error bar (mean error) isindicated in the upper right corner.

dius. To test the effects on the KR, Fig. 6.7 illus-trates the KR which was constructed using 〈µe〉and Re derived by a pure r1/4 luminosity profile.The variations on the structural parameters aresmall and the galaxies move only along the av-erage inclined error bar of logRe shown in theupper right corner in the KR figures. In general,the same conclusion can be drawn if the de Vau-couleurs structural parameters are used in theKR. The E+S0 galaxies of both clusters show anoffset of −0.38± 0.27m, with a median of −0.42.Note that, if r1/4-law structural parameters areused to establish the KR, individual objects aredifferently distributed along the KR, whereas thederived average offsets from the local KR and themeasured luminosity evolution are very similar.For purposes of comparison, a bisector fit to thedistant KR of the rich clusters constructed with


de Vaucouleurs structural parameters results in

〈µe〉 = (3.09±0.17) logRe+(17.99±0.09). (6.9)

Neither a significant zero-point offset nor a slopechange is found compared to the KR of thesame sample but based on structural parameterswhich were deduced from a combination of bulgeand exponential disc surface brightness profilefitting functions.In Fig. 6.9, the KR for the three Low–LX clus-ters Cl 0849, Cl 1701 and Cl 1702 is displayed.A total of ten spectroscopically confirmed early–type cluster galaxies enters the KR and the FP,which are splitted into five, three and two galax-ies for the clusters Cl 0849, Cl 1701 and Cl 1702,respectively. A linear bisector fit to the Low–LXKR sample in Fig. 6.8 gives

〈µe〉 = (2.24± 0.42) logRe + (18.35± 0.31),(6.10)

which corresponds to a 2σ flatter slope for thepoor cluster galaxies. However, this result is atvery low confidence level as the Low–LX samplelacks galaxies with small sizes of logRe < 0.3.As a result of this effect, the slope of the distantKR decreases.To estimate the amount of luminosity evolution,the mean residuals from the local KR of the poorcluster sample by fixing the local slope are inves-tigated. With respect to the Coma reference, theE+S0 galaxies in poor clusters offer an observedscatter of σKR = 0.53m and brighter luminositiesby −0.67 ± 0.23m. Furthermore, Figs. 6.8 and6.9 show that the cD galaxies in Low–LX clus-ters are fainter and exhibit on average smallersizes which implies that they are a less dominantgalaxy population of the clusters. These galax-ies are more comparable to their cluster membergalaxies and not a distinct class of galaxies astheir counterparts in rich galaxy clusters. Thistopic will be further addressed in the next sectionas the third parameter, the velocity dispersion,provides additional conclusions.

6.4 The Fundamental Plane

Via an investigation of the Fundamental Plane,tight constrains on the formation history and thedynamical status of early–type galaxies are pos-sible. Three observables, characterising the size(effective radius Re), the luminosity (effectivesurface brightness within Re, 〈µe〉) and the kine-matics (velocity dispersion σ) of a galaxy form anempirical relationship, the Fundamental Plane,which provides information about the evolutionand formation mechanisms of early–type galax-ies. Fig. 6.10 shows two projections of the localFundamental Plane relation in the Gunn r-bandfor 99 early-type galaxies in the Coma clusterfrom J99. The left hand panel shows the face-on view of the FP, the right hand panel the FPedge-on. The distribution of galaxies is affectedby a selection effect, the completeness limit inthe absolute magnitude of the Coma sample(Mr = −20.02m), which is indicated as the thedashed line in Fig. 6.10. The upper dotted linedenotes the so-called ‘exclusion zone’ found fornearby galaxies by Bender et al. (1992), whichis a region not occupied by dynamically hot stel-lar systems but not caused by selection effects.In terms of structural parameters, this limit im-plies Reσ

7.460 〈I〉−2.72

e ≤ constant, where σ0 is thecentral velocity dispersion. As this borderlineincorporates all three physical properties, boththe density and the total mass of the progenitorsof early-type galaxies are mutually constrained,which define an upper limit on the amount ofinvolved dissipation for any given mass. In theedge-on view of Fig. 6.10, a principal componentfit to the local FP relation is indicated by thethick solid line, which was derived via a princi-pal component analysis (e.g., Ziegler et al. 2005).The correlation between 〈I〉e (which is the meansurface brightness in units of L /pc2, cf. Eq. 4.6on page 69), σ and Re is displayed in both fig-ures. Similar to the scaling relations presentedbefore, the FP was restricted to the range inmagnitudes and velocity dispersions covered bythe distant samples. This is particularly impor-


Figure 6.10: The local Fundamental Plane relation in the Gunn r-band for 99 early-type galaxies in theComa cluster from Jørgensen (1999, J99) and Jørgensen et al. 1995, J99). Symbol notations as in Fig. 6.1.Left: FP face-on, Right: FP edge-on. The thick solid line is a principal component fit to the local FPrelation. The arrows in both figures display the directions in which the basic parameters of 〈I〉e, σ and Re

drive the FP.

tant when the amount of luminosity evolution ofa distant sample with respect to a local refer-ence is derived and will be further discussed insection 6.5.

The Fundamental Plane in rest-frame Gunn rfor the distant clusters A 2218 (z = 0.175) andA 2390 (z = 0.228) is illustrated in Fig. 6.11.The figure also shows the FP for the local Comasamples of J99 (z = 0.024). All mean surfacebrightness magnitudes have been corrected forthe dimming due to the expansion of the Uni-verse. Errors are provided in the short edge-onFP projection (upper right panel of Fig. 6.11),with kinematic and photometric properties onseparate axes. At a first glance, the followingresults can be deduced. Firstly, the FP for theindividual galaxy populations as well as for thecombined sample show a well-defined, tight re-lation with a small scatter. Both intermediateredshift clusters show a similar behaviour withinand along all projections of the FP. There is noevidence for an increasing scatter with redshift.Bernardi et al. (2003) found for the SDSS sample

an rms scatter of σ = 0.08, measured as the rmsscatter of the residuals in log Re. The distantcluster galaxies have an rms scatter of σ = 0.113,which is not significantly higher than the localvalue. This topic will be re–addressed in moredetail for different sub–samples later on in thissection. Secondly, only a moderate evolution ofthe FP and hence the stellar populations of thegalaxies with redshift is seen. Assuming there isno evolution in the structure of the galaxies, i.e.,at a fixed Re and σ, the average brightening ofthe cluster galaxies can be determined.

In Table 6.2 the results for the FP in rest-frame Gunn r are presented. The mean andmedian zero-point offsets from the local ComaFP, their 1σ scatter and the derived luminosityevolution are shown. For the combined sampleof 34 early-type cluster galaxies an ZP offset of∆γ = 0.10±0.06 compared to the local Coma ref-erence is deduced, which corresponds, accordingto ∆µe ≡ ∆γ/β with β = −0.32, to a modest lu-minosity evolution of ∆µe = −0.31±0.18m. Theresults are consistent with the picture of simple


Figure 6.11: Fundamental Plane for A 2390 at z = 0.23 (circles) and A 2218 at z = 0.18 (triangles) inrest-frame Gunn r-band, compared to the Coma galaxies of J99 (small squares). Filled symbols denoteellipticals, open symbols S0 galaxies and Sa bulges. Upper panel, left: Face-on view of the FP. The dottedline indicates the so-called ‘exclusion zone’ (Bender et al. 1992), the dashed line the luminosity limit for thecompleteness of the Coma sample MrT = −20.02m. Upper panel, right: FP edge-on, along short axis. Thesolid line represents the principal component fit for the local Coma sample. Lower panel: Edge-on FP. Onthe right side a zoom of the edge-on FP with a separation into different morphologies is shown.


Table 6.2: Evolution of the FP in Gunn r as derived for various samples. N shows the number of galaxiesand ∆ γ indicates the mean FP zero-point offset. In the fourth and fifth column, the median FP zero-pointevolution ∆ 〈γ〉 and the median evolution in the FP ∆ 〈µe〉 [in mag] are listed. The last column gives the±1σ scatter of the mean offsets.

Sample N ∆ γ ∆ 〈γ〉 ∆ 〈µe〉 σγ

A 2218 20 0.032 0.026 −0.079 0.108A 2390 14 0.140 0.158 −0.482 0.088A 2218+A 2390 34 0.076 0.100 −0.305 0.113Ea 17 0.032 0.008 −0.024 0.122S0a 17 0.120 0.144 −0.439 0.085E (pure r1/4)b 17 0.032 −0.010 0.030 0.125S0 (pure r1/4)b 17 0.115 0.134 −0.409 0.079low lum.c 17 0.085 0.100 −0.305 0.114E 8 0.070 0.074 −0.226 0.127S0 9 0.098 0.122 −0.372 0.107high lum. 17 0.068 0.093 −0.284 0.114E 9 −0.001 −0.020 0.061 0.114S0 8 0.145 0.168 −0.512 0.045low massd 17 0.129 0.153 −0.466 0.106E 6 0.114 0.158 −0.482 0.150S0 11 0.138 0.153 −0.466 0.080high mass 17 0.023 0.010 −0.030 0.095E 11 −0.012 −0.015 0.046 0.081S0 6 0.087 0.100 −0.305 0.091

a based on r1/4+exp. disc surface brightness profiles.b r1/4 surface brightness profile only.c lower-luminosity: Mr > −21.493, higher-luminosity: Mr < −21.493.d less-massive: logσ < 2.283, more-massive: logσ > 2.283.

passive evolution models by BC96. Assuming aformation redshift of zform = 2 for all stars, thesemodels predict a brightening by ∆mr ≈ 0.20m.For the individual clusters, different results areobtained. A zero-point offset of 0.16 ± 0.06 forA 2390 alone is found, meaning that the early-type galaxies in A 2390 are on average more lu-minous by ∆µe = −0.49± 0.18m than the localComa sample of J99. For the E+S0 in A 2218an offset of ∆γ = 0.03± 0.06 is detected, corre-sponding to a brightening of the stellar popula-tions by ∆µe = −0.09 ± 0.18m. This variation

for the clusters may be caused by a combinationof two effects: (i) possible effect due to cosmicvariance and (ii) sample selection. For exam-ple, Jørgensen et al. (1999) have analysed twodifferent clusters (A 665 and A 2218) both at aredshift of z = 0.18 and found a slight differentevolution of ∆µe ∼ −0.15m. This may give evi-dence that not all rich clusters have the same FPzero-point offset and that cosmic variance mustbe accounted for (see section 6.6 for further de-tails on this topic). Another reason may arisefrom the sample selection. In case of A 2390


particularly fainter galaxies have been selectedin order to gain additional insights on the low-mass end of the FP at z ∼ 0.2. For this reason,the A 2390 FP sub-sample comprises more low-luminosity galaxies than that of A 2218, whichresults on average in a stronger luminosity evo-lution for the A 2390 cluster. Table 6.2 gives acomparison between the two sub-samples of lessand more-massive galaxies and between faint andbright galaxies. If the total sample is dividedinto halves regarding to velocity dispersion at∼192 km s−1(corresponding to logσ < 2.283), astronger evolution of ∆µe = −0.47 ± 0.24m forthe low-mass galaxies with respect to their moremassive counterparts ∆µe = −0.03 ± 0.21m isderived. Similar results are found when a cut inluminosity at Mr = −21.493 is made, althoughwith a larger scatter. The difference between thelower-luminosity and higher-luminosity galaxiesin the sample of the distant rich cluster is alsovisible in the size distribution of the two samples,see Fig. 5.28 on page 125 and in the Kormendyrelation (cf. Fig. 6.6). The cluster A 2390 com-prises more galaxies with small sizes and threegalaxies even with logRe ≤ 0.05.

Looking at the thickness of the FP with re-spect to luminosity, an increasing scatter forthe fainter galaxies within the rich cluster sam-ple is not found. Both lower-luminosity galax-ies with Mr > −21.493 and higher-luminosityones exhibit a rms scatter of the residuals inlog Re of 0.114. Jørgensen et al. (1996), di-vided their local sample with respect to luminos-ity into faint (Mr > −23.16m) and bright galax-ies and reported on a larger rms scatter for thelower-luminosity galaxies of 0.086, whereas thebright galaxies have 0.047. These authors ar-gued that the larger dispersion may be a resultof the presence of discs or larger variations in thestellar populations of lower-luminosity galaxiessince these objects showed a larger range in theMg2 index. On the other hand, slight differencesin the scatter between lower-mass and higher-mass cluster galaxies in A 2218 and A 2390 are

found. Lower-mass galaxies with logσ < 2.283feature a rms scatter of 0.106, but higher-massgalaxies show only 0.095. The deviations arenot significant, but nevertheless could be a hintthat the lower-mass galaxies, which also exhibita stronger luminosity evolution, have more di-verse stellar populations or different star forma-tion histories than their more-massive counter-parts.

Fig. 6.12 displays the Fundamental Plane in rest-frame Gunn r for the distant Low–LX clus-ters Cl 0849 (z = 0.234), Cl 1701 (z = 0.246)and Cl 1702 (z = 0.223). Again, the distantFP is compared to the local Coma FP of J99(z = 0.024). The mean surface brightness mag-nitudes have been corrected for the dimming dueto the expansion of the Universe and the errorsare shown in the short edge-on FP projection(upper right panel of Fig. 6.12). Similar to therich clusters, the FP for the combined poor clus-ter sample is a well-defined relation, with no sig-nificant differences or trend between the individ-ual galaxy populations of a specific cluster. Allthree intermediate redshift clusters show a sim-ilar behaviour within and along all projectionsof the FP. The combined poor cluster samplehas an rms scatter of σ = 0.161, which is twiceas large as the local Coma rms scatter. Divid-ing the sample according to luminosity, smalldifferences between lower-luminosity and higher-luminosity galaxies are detected. Fainter galax-ies (Mr > −21.855) offer a rms scatter of 0.194,whereas brighter galaxies have 0.143. Similar tothe rich cluster galaxies slight differences arisewhen the sample is separated with respect tomass. Lower-mass galaxies with logσ < 2.295exhibit a rms scatter of 0.173 and higher-massgalaxies have 0.126. This increasing scatter isalso visible in the left panel of Fig. 6.20, wherethe M/L ratios are plotted as a function of ve-locity dispersion. Table 6.3 lists the mean andmedian zero-point offsets from the local ComaFP, the measured luminosity evolution and theobserved 1σ scatter for various Low–LX clus-


Figure 6.12: Fundamental Plane for the early-type galaxies in the poor clusters Cl 0849, Cl 1701 and Cl 1702(large symbols) in rest-frame Gunn r-band, compared to the Coma galaxies of J99 (small squares). Filledsymbols denote ellipticals, open symbols S0 galaxies. Upper panel, left: Face-on view of the FP. The dottedline indicates the so-called ‘exclusion zone’ (Bender et al. 1992), the dashed line the luminosity limit for thecompleteness of the Coma sample MrT = −20.02m. Upper panel, right: FP edge-on, along short axis. Thesolid line represents the principal component fit for the local Coma sample. Lower panel: Edge-on FP. Onthe right side a zoom of the edge-on FP with a separation into different morphologies is shown.


Table 6.3: Evolution of the FP in Gunn r as derived for the Low–LX cluster samples in Cl 0849, Cl 1701and Cl 1702. N shows the number of galaxies and ∆ γ indicates the mean FP zero-point offset. In the fourthand fifth column, the median FP zero-point evolution ∆ 〈γ〉 and the median evolution in the FP ∆ 〈µe〉 [inmag] are listed. The last column gives the ±1σ scatter of the mean offsets.


Cl 0849 5 0.105 −0.002 0.006 0.193Cl 1701 3 0.116 0.040 −0.122 0.154Cl 1702 2 0.169 0.292 −0.889 0.174Low–LX 10 0.121 0.046 −0.141 0.161low lum.a 5 0.120 0.015 −0.045 0.194high lum. 5 0.123 0.046 −0.141 0.143low massb 5 0.193 0.292 −0.889 0.173high mass 5 0.050 0.040 −0.122 0.126

a lower-luminosity: Mr > −21.855, higher-luminosity: Mr < −21.855.b less-massive: logσ < 2.295, more-massive: logσ > 2.295.

ter samples. The combined sample of 10 early-type cluster galaxies exhibits an ZP offset of∆γ = 0.12 ± 0.05 compared to the local refer-ence, which corresponds to a modest luminosityevolution of ∆µe = −0.37 ± 0.15m. This mildevolution is in compliance with a passive evo-lution for the bulk of the stellar populations inthese galaxies. Dividing the sample with respectto velocity dispersion, the lower-massive galax-ies (log σ < 2.295) show a stronger evolutionwith ∆µe = −0.59±0.49m than the higher-massones with ∆µe = −0.15± 0.13m. Although thiscomparison is based on low number statistics, itconfirms the results of the FJR, with a larger off-set and therefore larger average brightening forlower–mass galaxies with respect to the Comareference. Splitting the data set according to lu-minosity at Mr = −21.855, the sub–samples in-dicate a similar scatter and evolution in luminos-ity but no significant differences. A continuationof this topic will be demonstrated in section 7.1.2based on the whole poor cluster sample.

To complete the analysis of early-type galaxies indifferent environments, the Fundamental Planefor two field galaxy samples was constructed.

Fig. 6.13 shows the FP for the 21 early-type fieldgalaxies in the FDF and WHDF in the rest-frameJohnson B-band, compared to the 39 early-typegalaxies in the Coma cluster by SBD93. Thelocal reference is indicated with small squares,whereas the distant field galaxies are shown asthe large circles and stars. Filled symbols de-note ellipticals, open symbols S0 galaxies andSa bulges. The 21 FDF and WHDF early–typefield galaxies are morphologically classified into9 E and 12 S0. Five ellipticals and six S0s origi-nate from the FDF and four Es and six S0s arelocated in the WHDF. Two galaxies are not visi-ble on the ACS images and for one elliptical gal-axy in the FDF no structural parameters couldbe derived. The distant field galaxies cover aredshift range of 0.21 ≤ zField E ≤ 0.74 witha median of 〈zField E〉 = 0.40. The two out-liers in the FP, # 6336 (log Re = 0.46) and# 508 (log Re = 1.20), reveal on the ACS im-ages a significant disc component and early spi-ral morphology. Galaxy # 6336 has an extraor-dinary strong HδF absorption, which points toa young luminosity-weighted average model ageunder assumption of a single stellar population.


Figure 6.13: Edge-on view of the FundamentalPlane for the early-type field galaxies in the FDF andWHDF, compared to the Coma galaxies of SBD93(small squares) in rest-frame Johnson B-band. Filledsymbols denote ellipticals, open symbols S0 galaxiesand Sa bulges. The distant field lenticular galaxiesshow a stronger evolution and larger scatter than theellipticals.

On the ACS images, # 508 shows a weird struc-ture which might indicate that this object is anongoing merger. Both galaxies were classifiedas Sa bulges hence these early-type spirals offerat least a low level of star formation which in-creases their total luminosity and displaces theirposition beyond the tight plane as established byearly–type galaxies.Rejecting the two outliers, for the total sam-ple of 19 early-type field galaxies in the FDFand WHDF an average luminosity evolutionof ∆MB = −0.53 ± 0.13m with a median of∆〈MB〉 = −0.39 is found. Similar results aremeasured for 17 cluster ellipticals at 〈z〉 ∼ 0.4with similar look-back times (∼5 Gyrs). Thesegalaxies show an average brightening of theirstellar populations of ∆MB = −0.44 ± 0.18m,compared to their local counterparts (Ziegler etal. 2005). Overall, the distribution along the FPand the measured luminosity evolution is similar

for cluster and field galaxies at 〈z〉 ∼ 0.4 andwithin the cluster sample no differences betweenellipticals and S0 galaxies were found. However,the sample of distant cluster galaxies lacks gal-axies with small sizes. On the other hand, thefield galaxy sample comprises no large ellipti-cals or cD galaxies. A comparison of the sizesof these two data sets is therefore not adequateand would cause misleading results. For this rea-son, the cluster galaxies at 〈z〉 ∼ 0.4 will not becompared to the field sample to test possible dif-ferences in their stellar populations. In Table 6.4a summary of the results for the FP of the early-type field galaxies in Johnson B-band is given.The thickness of the FP seems to vary betweenthe individual field galaxy samples. For theFDF an average rms scatter of 0.136 is detected,whereas the WHDF show a larger 1σ disper-sion of 0.289. This effect is partly caused bythree high redshift objects 0.4 ≤ z ≤ 0.74 inthe WHDF, which are all discy galaxies. Di-viding the field sample with respect to veloc-ity dispersion at logσ = 2.275 or luminosityat MB = −21.404, no strong variations in theamount of evolution between the sub–samplesare detected. However, the comparisons relyon the assumed cutoffs in σ and MB becauseoutliers, such as early Sa spiral galaxies, canstrongly influence the derived luminosity evolu-tion. In the next section, possible differences be-tween the populations of elliptical and lenticulargalaxies will be addressed.

6.4.1 Elliptical versus S0 galaxies

Local investigations based on the FundamentalPlane have not revealed significant differences inthe zero-point, slope and/or scatter between el-liptical and lenticular galaxies (Bender, Burstein& Faber 1992; Saglia et al. 1993b; Jørgensen etal. 1996). Moreover, they behave very similar asone single group of galaxies with respect to theirM/L ratios and within the FP. Going to higherredshifts, differences could be more significant ifrecent star formation activity plays a role for S0


Table 6.4: Evolution of the FP in Johnson B-band as derived for the early-type field galaxies in the FDFand WHDF. N shows the number of galaxies and ∆ γ indicates the mean FP zero-point offset. In the fourthand fifth column, the median FP zero-point evolution ∆ 〈γ〉 and the median evolution in the FP ∆ 〈µe〉 [inmag] are listed. The last column gives the ±1σ scatter of the mean offsets.


FDF 11 0.066 0.075 −0.235 0.136WHDF 10 0.199 0.214 −0.667 0.289FDF+WHDF 21 0.130 0.089 −0.279 0.227FDF+WHDFa 19 0.170 0.126 −0.393 0.197E 9 0.075 0.075 −0.230 0.121S0 12 0.171 0.211 −0.643 0.281S0a 10 0.255 0.241 −0.735 0.219S0b 9 0.214 0.211 −0.643 0.185low lum.c 11 0.056 0.082 −0.255 0.180high lum. 10 0.211 0.165 −0.515 0.254high lum.d 9 0.164 0.126 −0.393 0.219low masse 11 0.099 0.089 −0.279 0.221high mass 10 0.163 0.126 −0.393 0.240

a Omitting # 6336 and # 508, both Sa bulges.b Omitting # 6336 and # 508 and # 111.c lower-luminosity: MB > −21.404, higher-luminosity: MB < −21.404.d Omitting # 111 at z = 0.74.e less-massive: logσ < 2.275, more-massive: logσ > 2.275.

galaxies. However, previous studies of samplesat higher redshift did not find any differencesbetween elliptical and S0 galaxies (van Dokkum& Franx 1996; Kelson et al. 2000b). For ex-ample, within the large sample of 30 early-typegalaxies in CL 1358+62 at z = 0.33 of Kelson etal. (2000b), the 11 ellipticals displayed identi-cal zero-points as the 13 (non–E+A) S0 galaxieswith no hint for an offset between these groupsat all. Moreover, the difference in the slope of∼14 % detected between the S0 and elliptical gal-axies is not significant.

For the case of S0 galaxies, two main questionsare still a matter of debate. How many are a pri-ori S0 galaxies? Which and how many are theresult of galaxy transformations and account forthe dominant S0 population in rich clusters to-

day? For scenarios which suggest a transforma-tion of star-forming spiral galaxies into passiveS0 systems (Dressler et al. 1997), these objectswould only be classified as lenticulars after theirmorphology has been changed (∼5 Gyrs). How-ever, after this relatively long time–scale, theirstar formation (SF) could already have ceasedwhich would make these galaxies hard to detectvia their SF. E+A galaxies, which are charac-terised by strong Hδ absorption lines, could rep-resent galaxies in an intermediate stage of sucha transformation where the merging process oftwo individual galaxies just has stopped and thespheroidal remnant is indistinguishable from aregular E/S0 galaxy, but SF is still present duethe result of a starburst which ended within thelast 1.5–2 Gyr (Barger et al. 1996). Recently,


Figure 6.14: Edge-on view of the Fundamental Plane for A 2390 and A 2218 in rest-frame Gunn r. Thedistant FP is compared to the FP of the local Coma sample of J99, indicated by the principal component fit(thick solid line). Left: FP constructed using a combination of an r1/4 and exponential disc profile. Right:Similar to the left plot except that r1/4-law parameters were used.

Yang et al. (2004) performed a study of E+Agalaxies in the local Universe and argued thatthese objects only show an offset in their surfacebrightnesses but not in their total magnitudes.If such galaxies, which contribute about 20% tothe galaxy population in clusters at z ∼ 0.5, arealso within the galaxy samples investigated here,they would have no offset in their absolute lumi-nosity and therefore would be indistinguishablefrom other non–E+A galaxies in the FJR whichis based on total magnitude measurements. ButE+A galaxies could be identified within the FPdue to their high surface brightness magnitudeswhich makes them stand apart from the FP ofearly–type galaxies.

Figure 6.14 illustrates the edge-on view of theFP for 34 E+S0 galaxies of the clusters A 2218and A 2390 in rest-frame Gunn r-band. The FPconstructed for the distant clusters at intermedi-ate redshift is compared to the FP for the localComa sample of J99. The combined sample rep-resents the most extensive investigation of early-type galaxies in clusters at z∼0.2. The Fig. 6.14

has been splitted in order to visualise the smallvariations if different luminosity profile fits areused for deriving structural parameters. The leftfigure displays the FP constructed using a com-bination of an r1/4-law+exponential disc profile,the right figure is based on pure r1/4-law fits.The variations in the structural parameters onlyaffect the galaxies to move along the edge-on pro-jection of the FP plane, thereby maintaining thetightness of the plane. Both the r1/4 FP andr1/n FP have the same scatter and slope withintheir errors. A second comparison is given bythe FP parameter Re 〈Ie〉0.8 in Fig. 6.15, wherethe structural parameters of Re and 〈Ie〉 were de-rived from a combination of an r1/4+exponentialdisc profile and by a pure r1/4-law model. As theerrors in Re and 〈Ie〉 are correlated and this cor-relation is almost parallel to the FP, the prod-uct Re 〈Ie〉0.8 can be established with high accu-racy. The individual measurements of Fig. 6.15are listed in Table 4.5 and Table 4.6 (column 9).In general, the agreement is good for both ellip-tical and S0 galaxies.


Figure 6.15: FP parameter Re 〈Ie〉0.8 derivedfrom a pure r1/4-law model compared to theRe 〈Ie〉0.8 constructed using a combination of an r1/4-law+exponential disc profile. The product Re 〈Ie〉0.8holds for both elliptical and S0 galaxies and displaysa scatter of only 4%.

In the following, possible differences between el-liptical and lenticular galaxies will be consid-ered in more detail. As a local reference theComa FP coefficients by Jørgensen et al. (1996)for the Gunn r-band α = 1.24 ± 0.07 andβ = −0.82±0.02 were adopted. For the JohnsonB-band α = 1.25 ± 0.06 and β = −0.80 ± 0.02were assumed which are slightly different but stillwithin the scatter found for the Coma FP pa-rameters. The uncertainties in the coefficientsof the local FP were assessed by the authorsthrough a bootstrap fitting technique. Gener-ally, the choice of the fitting technique, the selec-tion criteria and the measurement errors whichare correlated can lead to systematic uncertain-ties in the FP coefficients in the order of±0.1 dex(Jørgensen et al. 1996). Nevertheless, as theFP coefficients α and β depend only weakly onthe wavelength and were fixed for the local andthe distant sample, a significant change within aphotometric band as a function of redshift canbe ruled out.

A morphological analysis of the HST images re-

vealed that the A 2390 sub-sample splits nearlyequally into elliptical (8) and lenticular (S0) gal-axies (6). Out of the 20 cluster members in theHST field of A 2218, nine and eleven were iden-tified as ellipticals and S0s, respectively. Theearly-type sample of the Low–LX clusters wasmorphologically classified as nine ellipticals andone S0 galaxy. As a comparison of sub–classescannot be performed with these numbers, thefollowing analysis concentrates on the rich clus-ters only. Both ellipticals and lenticular galaxiesare uniformly distributed along the surface of theFP plane. An edge-on projection can thereforebe taken for a robust comparison of their stellarpopulations. Table 6.2 lists the derived evolu-tion for the E and S0 galaxies. Fixing the slopeof the local Coma reference, the zero-point offsetfor the ellipticals in the sample is

〈∆γcE (z = 0.2)〉 = 0.01± 0.07. (6.11)

Restricting the sample to the elliptical galax-ies a negligible zero-point deviation with re-spect to the local FP is derived, which corre-sponds to an insignificant luminosity evolutionof ∆ME

r = −0.02 ± 0.21m which is within their1σ scatter. On the other hand, the ZP offset forthe S0 galaxies yields

〈∆γcS0 (z = 0.2)〉 = 0.14± 0.06. (6.12)

This zero-point offset for lenticular galaxies cor-responds to an evolution ∆MS0

r = −0.44±0.18m

with respect to the local counterparts. In com-parison to the ellipticals, the difference in evo-lution between E and S0 galaxies is significanton the 2σ level. For the combined sample of 34E+S0 cluster galaxies a median offset of

〈∆γcE+S0 (z = 0.2)〉 = 0.10± 0.06, (6.13)

is derived, which corresponds to a brightening ofthe stellar populations by −0.31±0.18 mag.The errors on the zero-points were individuallyderived as

δZP2 = δFP2r + δBS2 (6.14)


Figure 6.16: Evolution of the field FP in rest-frame Johnson B-band. The early-type field galaxies inthe FDF and WHDF are binned into three different redshift slices, each compared to the Coma galaxies ofSBD93 (small squares) and shown along the short axis of the edge-on view. Filled symbols denote ellipticals,open symbols S0 galaxies and Sa bulges and the brackets the number of the respective morphological class.For illustration reasons, only the mean error is shown in the left bottom. The offset of the distant fieldgalaxies from the local FP increases with redshift, whereas the scatter appears to increase primarily for S0galaxies with look–back time.

where δFPr denotes the total error which entersthe FP in the rest-frame Gunn r-band (cf. sec-tion 4.3.3) and δBS is the uncertainty computedthrough an iterative bootstrap re–sampling ofthe data points 100 times. On average, thelenticular galaxies in the sample show a largerluminosity evolution than the ellipticals. Inboth interpretations of Fig. 6.14 the S0 galax-ies are predominantly located below the ellip-ticals and may indicate a different evolution-ary trend between the stellar populations of el-liptical and S0 galaxies. Dividing the samplewith respect to velocity dispersion into a lowmass (log (σ) < 2.283) and a high mass sub-sample (log (σ) > 2.283), a different evolutionis found. The lower-mass galaxies are on aver-age with −0.47±0.24m more luminous than theirmore massive counterparts with −0.03 ± 0.21m

at z ∼ 0.2. This is also seen for the different sub-groups of elliptical and S0 galaxies, but with lesssignificance (see Table 6.2). To investigate dif-ferences between elliptical and lenticular galax-

ies in clusters even further, the next section 6.5concentrates on comparisons of M/L ratios andmasses of these galaxy types. Before this dis-cussion, possible variations in the morphologicalsub–classes of field galaxies will be explored.Looking at differences between the morphologi-cal types of field early-type galaxies, the S0 gal-axies display a stronger evolution than the ellip-tical galaxies. Assuming that the slope of the lo-cal reference holds valid for the distant galaxies,the 9 field ellipticals show in edge-on projectionof the FP a zero-point offset

〈∆γfE (z = 0.4)〉 = 0.08± 0.06, (6.15)

which corresponds to a brightening in their stel-lar populations of ∆ME

B = −0.23 ± 0.18m. Bycomparison, the 12 field lenticulars exhibit anoffset in the zero-point with respect to the localComa galaxies

〈∆γfS0(z = 0.4)〉 = 0.21± 0.09, (6.16)

which corresponds an average evolution of∆MS0

B = −0.64 ± 0.27m. Errors in the zero-


points of elliptical and S0 galaxies were analo-gously computed as in Eq. 6.14 with the only ex-ception that the FP error in the B-band δFPBwas utilised. Limiting the S0 galaxies to red-shifts z < 0.7 and rejecting the two Sa bulges,the same results are derived (see Table 6.4). Inaddition, the field ellipticals obey a tight FP witha small 1σ scatter of 0.121, whereas the S0 typeshave a larger dispersion of 0.219 (omitting thetwo Sa bulges). This gives further evidence thatthe discy galaxies have a larger range of differentstellar populations. Even more astonishing is thefact that three elliptical galaxies in the WHDFsample feature weak [O II] 3727 emission in theirspectra. In particular, the galaxies # 810, # 437and # 810b exhibit [O II] 3727 = −3.73± 0.47 A,−0.35 ± 0.37 A and −6.62 ± 1.87 A. Generally,a line emission is evidence for star formationactivity which increases the average luminosityand thereby causing the galaxy to be offset fromthe local FP relation. In contrast to what is ex-pected, these ellipticals indicate no offset alongthe edge-on projection of the FP but a tight re-lation. One explanation could be that the starformation rates are too low to result in an visibleeffect within the FP. Although the evidence forthe galaxy # 437 is rather poor due to its mea-surement uncertainty, it could be that the lowstar formation rates of the galaxies are either therelicts of a former starburst or the weak imprintsof galaxy–galaxy interactions which have takenplace at an earlier epoch. This suggests thatgalaxy–galaxy interactions which trigger the starformation activity are a common phenomenonamong field galaxies.

The evolution of the FP for the early-type FDFand WHDF field galaxies in the rest-frame John-son B-band is illustrated in Fig. 6.16. Early-typegalaxies were binned in redshift space to inves-tigate their location within and along the edge-on FP as a function of redshift and to test theeffects of possible outliers. Ellipticals are repre-sented as filled, S0s and Sa bulges as open sym-bols. For each morphological type the number of

galaxies within a redshift bin is indicated in thebrackets. Small squares denote the Coma galax-ies by SBD93 and the straight line is a princi-pal component fit to the local sample (Ziegler etal. 2005). Fig. 6.16 clearly shows the evolutionof the FP for the distant galaxies with respectto the local Coma FP. The offset of the fieldFP increases with redshift but the scatter ap-pears to be mainly amplified for lenticular galax-ies. The two galaxies with a “positive” evolutionin the last two panels are the Sa bulges # 6336and # 508 and show a clear disc on the ACSimages. These objects can mimic on average aweaker evolution in luminosity for the whole gal-axy population (cf. Table 6.4). Regardless ofthe Sa bulges, the S0s exhibit a larger scatter athigher redshift 0.26 < z < 0.75 than the ellipti-cals which suggests that the stellar populationsof E and S0 types are different and that lentic-ulars are a more heterogeneous group. An ex-planation could be that these S0 galaxies resem-ble post–starburst galaxies as no strong emissionlines were detected in their spectra.

6.5 M/L Evolution

The evolution of the Fundamental Plane can becharacterised in terms of its zero-point. In turn,the zero-point of the FP is related to the meanM/L ratio. Thus, if the FP zero-point for asample of early-type galaxies changes as a func-tion of redshift z, this implies an evolution ofthe mean M/L ratio and hence an evolution inthe luminosity-weighted stellar population prop-erties of the galaxies under consideration.For the present-day zero-point of the FP a boot-strap bisector fit to the early-type Coma galax-ies in rest-frame Gunn r-band was performed.This yielded a slope of 1.048± 0.038 and a zero-point of −0.029 ± 0.010. The early-type galax-ies of A 2390 show a small shift with respect tothe Coma galaxies (see Fig. 6.11). As the veloc-ity dispersions and effective radii of the galax-ies in A 2390 span similar ranges like their local


counterparts (apart from the Coma cD galax-ies with Re >∼ 30 kpc) and the surface bright-nesses are slightly increased with respect to thelocal sample, at least a fraction of the shift inthe M/L ratio is due to luminosity evolution.However, when deriving an exact amount of lu-minosity evolution, one has to be careful as ad-ditional combined effects also contribute to thetotal derived luminosity offset. Some studiesfound different FP slopes at intermediate red-shift compared to the local slope, but they suf-fer from low number statistics (van Dokkum &Franx 1996; Kelson et al. 1997; van Dokkum etal. 1998). However, more recent investigationsbased on larger samples reported similarly onlya modest change in the FP slope (Kelson et al.2000b; Treu et al. 2001b). Thus, the lack ofa strong slope change gives evidence against thehypothesis of the FP being solely an age-mass re-lation. The Fundamental Plane is mainly a rela-tion between the M/L ratio and galaxy mass M .Therefore, when comparing small samples withsignificantly different ranges in galaxy mass, anoffset in the M/L ratio can heavily depend onthe adopted FP slope. The M/L offset will be acombination of three items: (i) the difference inthe slope of the FP, (ii) the mean mass range ofthe sample and (iii) the ‘true’ offset. By reducingthe differences in the mass distribution, the M/Lresults become insensitive to the adopted slopeof the FP and any derived evolution for the dis-tant galaxies is valid for a given range of galaxymasses (Kelson et al. 2000b). For this reason,the local Coma sample was restricted to a sim-ilar mass function as for the galaxies in A 2218and A 2390 and this reference was applied forthe entire analyses within the FP and the M/Levolution.

A comparison of the mass distributions of thedistant early-type galaxies of A 2390 and A 2218with the local Coma galaxies within the log σ −logRe plane is shown in Fig. 6.17. The distribu-tions of galaxy masses for the Low–LX E+S0galaxies are illustrated in Fig. 6.18. Regions

Figure 6.17: The log σ − log Re plane for theA 2390 and A 2218 early-type galaxies, compared tothe Coma galaxies of J99. Symbol notations as inFig. 6.11. Dashed lines indicate contours of constantmass 5G−1 σ2Re = 1010, 1011 and 1012M (Benderet al. 1992). The early-type galaxy masses of A 2390and A 2218 at z ∼ 0.2 are similarly distributed as theearly-type Coma sample.

of constant mass, ranging from 1012M over1011M down to 1010M, derived with the re-lation M = 5σ2Re/G in units of M (Bender,Burstein & Faber 1992) where G is the gravi-tational constant, are indicated as dashed lines.For the absolute magnitude in Gunn r of theSun, Mr, = 4.87m was adopted, which was de-rived from MV, = 5.72 and (V − R) = 0.52(Schaifers et al. 1981) and the transformationr = R + 0.41 + 0.21(V − R) (Kent 1985). Bothsamples exhibit similar ranges in mass (σ) andsize (Re), assuring that a possible M/L evolu-tion will not be driven mainly by any differ-ences between the galaxy mass ranges of thesamples. The restrictions to the local Comareference as defined for the rich cluster sam-ple were also adopted for the early-type galaxiesin the poor clusters (cf. section 6.2). To testthe effects of different limits in velocity disper-


Figure 6.18: The log σ − log Re plane for the theCl 0849, Cl 1701 and Cl 1702 early-type galaxies, com-pared to the Coma galaxies of J99. Symbol notationsas in Fig. 6.12. Dashed lines indicate contours ofconstant mass 5G−1 σ2Re = 1010, 1011 and 1012M(Bender et al. 1992). The early-type galaxy massesof poor cluster galaxies between 0.22 ≤ z ≤ 0.24 aresimilarly distributed as the early-type Coma sample.

sion on the amount of derived luminosity evo-lution for the Low–LX galaxies, the selectionboundaries of the Coma cluster were modified.The largest effect was detected for a limit oflog σJ99 > 2.10 (corresponding to a ∼22 km s−1

higher cutoff value), which yields a change in theamount of derived luminosity evolution of lessthan 0.0033m. This difference is negligible andassures that the assumption to use one definitionof selection boundaries is also valid for the poorcluster sample.

Adopting a constant slope with redshift, themedian zero-point offset for the combined sam-ple of 34 E+S0 rich cluster galaxies yielded∆γE+S0 , z=0.2 = 0.10± 0.06. Alternatively usinga variable slope in the FP for the distant clustersample, negligible changes in the FP parametersα and β compared to the locally defined param-eters by the Coma data are found. In particular,

a bootstrap bisector fit to all 34 E+S0 galaxiesgives a slightly steeper slope of 1.152±0.047 anda zero-point of −0.153 ± 0.018, with a 3σ confi-dence level for a slope change. This comparisonand the results based on the M/L ratios maygive some evidence for a mass-dependent evolu-tion which is stronger for low-mass galaxies.

Fig. 6.19 shows the observed mass-to-light M/Lratio of the distant clusters A 2218 and A 2390as a function of velocity dispersion σ (left) andversus mass M (right). The Coma sample waslimited to galaxies with log σ > 2.02 (indi-cated by the vertical dotted lines in Fig. 6.19)in order to match the area of parameter spacecovered by the distant galaxies of A 2390 andA 2218. An analysis based on bootstrap bi-sector fits to M/L = a σm revealed differ-ent M/L slopes for the distant (hashed area)and the local (solid lines) samples. Theslope difference between A 2218 and Coma is∆mA2218 = 0.27 ± 0.17 and an offset in thezero-point of ∆aA2218 = 0.040 ± 0.027. ForA 2390 ∆mA2390 = 0.11 ± 0.21 is detectedand ∆aA2390 = 0.157 ± 0.019. The com-bined sample of distant clusters has a slopedifference of ∆mz=0.2 = 0.36 ± 0.17 and∆az=0.2 = 0.036 ± 0.024. With a 2σ confi-dence different slopes for both are measured,the intermediate clusters and the Coma cluster(Coma slope value is m = 0.59 ± 0.15). How-ever, a systematic zero-point offset of the dis-tant M/L relation is detected, which is not inagreement with the proposed change due to pas-sive evolution. Adopting a formation redshiftof zform = 2 for all stars in early-type galaxies,the BC96 models predict ∆a = 0.12. Ellipti-cals and S0 galaxies seem to have different M/Lslope changes, with a steeper slope for the ellipti-cal galaxies. Ellipticals: ∆mE = 0.54± 0.30 and∆aE = 0.041 ± 0.037; S0s: ∆mS0 = 0.06 ± 0.23and ∆aS0 = 0.141± 0.026.

The M/L ratios as a function of mass Mare shown in the right plot of Fig. 6.19. Ina similar manner as for the velocity disper-


Figure 6.19: The M/L ratio in Gunn r for A 2390 and A 2218. Left: M/L ratio as a function of velocitydispersion σ, Right: M/L ratio as a function of mass in solar units. Symbol notations as in Fig. 6.11.Solid lines show the 100 iteration bootstrap bisector fits to the local Coma sample of J99 within selectionboundaries of the distant samples (1 σ errors). The hashed area denotes the bisector fits within 1 σ errorsto the combined distant sample of A 2390 and A 2218.

sions, the two samples are analysed in termsof bootstrap bisector fits. For the reducedComa cluster alone M/L ∝ M0.59±0.07 is found.With respect to Coma, the slope difference forA 2218 is ∆mA2218 = 0.025 ± 0.060 with azero-point offset of ∆aA2218 = 0.048 ± 0.029.For A 2390 a similar slope change is found∆mA2390 = 0.057 ± 0.063, but a larger offset of∆aA2390 = 0.165 ± 0.028. The two distant clus-ters as a whole have ∆mz=0.2 = 0.071 ± 0.039and ∆az=0.2 = 0.037± 0.021. Dividing the sam-ple into elliptical and S0 galaxies gives no signifi-cant slope changes, for Es: ∆mE = 0.049±0.056and for S0s: ∆mS0 = 0.087 ± 0.089. However,the S0s exhibit a larger offset in the zero-point of∆aS0 = 0.138±0.030, compared to the ellipticals∆aE = 0.050± 0.034.Analogous to the cluster galaxies, the M/L ra-tios for the poor clusters Cl 0849, Cl 1701 andCl 1702 were computed. Fig. 6.20 displays theM/L ratios as a function of velocity disper-sion σ (left panel) and as a function of massin solar units (right panel). Unfortunately, an

analysis for the morphological types of ellip-ticals and lenticulars cannot be performed asthe latter class comprises only one galaxy. Un-der assumption of a non–changing local slopewith redshift, the mean residuals from the lo-cal logM/L− log σ relation of the Low–LX clus-ter galaxies are investigated and their mean evo-lution in the M/L ratio is derived. The M/Lratios of the distant poor galaxies are on aver-age offset by ∆M/L = 0.138 ± 0.049 comparedto the Coma M/L ratios. In the right part ofFig. 6.20, the M/L ratios are shown as a func-tion of galaxy mass M . The low–mass clustergalaxies occupy a larger range in the M/L ratiosthan the rich cluster galaxies in Fig. 6.19. In ad-dition a larger offset to the local logM/L−logMrelation with ∆logM/L = 0.216 ± 0.059 for thepoor sample than for the rich cluster sample with∆logM/L = 0.109±0.021 is found, where for thelatter also the Coma slope was assumed.In both illustrations of Fig. 6.20 the M/L ratiosof the Low–LX E+S0 galaxies feature on averagea larger scatter with σM/L Poor = 0.29 than their


Figure 6.20: The M/L ratio in Gunn r for the poor clusters Cl 0849, Cl 1701 and Cl 1702. Left: M/L ratioas a function of velocity dispersion σ, Right: M/L ratio as a function of mass in solar units. Symbol notationsas in Fig. 6.12. The linear bisector fit to the Coma galaxies was restricted to the range encompassed by thedistant rich cluster galaxies. The area bounded by dotted lines indicates the mean ±1σ of the local M/Lrelation. The solid line shows the mean evolution of the poor cluster sample assuming the local slope.

counterparts in rich clusters σM/L Rich = 0.23and their M/L are slightly more offset from thegalaxies in Coma compared to the M/L ratiosin rich clusters. For a comparison, the reducedlocal reference sample has a dispersion in theM/L ratios of σM/L Coma = 0.14. In both, theFP (Fig. 6.12) and the logM/L− logM diagram(Fig. 6.20), it seems that there are two differentgroups of cluster galaxies, one class which fol-lows the local relation and another one which isclearly offset from the Coma reference. An ex-planation could be that the Low–LX E+S0 gal-axies have larger variations in their stellar pop-ulations or that star formation processes workdifferently in these galaxies. Such effects shouldalso be detectable in an analysis of the forma-tion redshift for the galaxies which gives alsoclues on the global star formation histories inthese systems (cf. section 7.2). Furthermore,in case of the bright cD galaxies, differences be-tween the Low–LX E+S0 galaxies and the richcluster members are found. Two of three cen-tral cluster galaxies in the poor clusters (# 24

with Re ≈ 14 kpc in CL 0849 and # 81 withRe ≈ 16 kpc in CL 1702) have higher velocity dis-persions with ∆σ ≈ 118 km s−1, but at the sametime smaller half-light radii of ∆Re ≈ 2.6 kpcthan their counterparts in the A 2218 cluster.The third cD galaxy of the cluster CL 1701(# 123) at σ = 244 km s−1 and Re = 8.8 kpchas a double nucleus and cannot be regarded asa typical cD galaxy as it is an ongoing mergerof two galaxies. Another galaxy which shows anoffset in the log σ− log Re diagram is the object# 62 (Re = 15.5 kpc) which is a large ellipticalgalaxy but not massive enough to be a centralcluster galaxy as it has a low velocity dispersionof ∼155 km s−1. Together with the findings inthe KR, the signs of interaction processes givefurther evidence, that the cD galaxies in low–mass clusters have longer formation timescalesup to the recent past and therefore represent adifferent galaxy population compared to the cDgalaxies in rich galaxy clusters.Morphological transformation from late-typespiral to S0 galaxies as observed from redshift of


z = 0.55 to the present day Universe (Dressleret al. 1997) can have significant effects on theevolution of colours and luminosities of early-type galaxies with redshift. If early-type galaxieswere transformed from late-type galaxies in therecent past, the youngest progenitors of present-day early-type galaxies will drop out from anobserved sample of early-type galaxies at highredshift. The samples at high and low redshiftdiffer and the high-z data set may be biasedtowards the oldest progenitors of present-dayearly-type galaxies. The correction for this so-called ‘progenitor bias’ (van Dokkum & Franx2001b) increases as a function of redshift andcan be quoted in terms of the M/L ratio as∆ln 〈M/LB〉 ≈ 0.2 × z. It has a small (<∼20%)but non-negligible effect on the measured evo-lution of the mean the M/L ratios. For theresults in the FP and on the M/L ratios atz = 0.2 the effect is insignificant, the true evo-lution in the M/L ratio is underestimated by∆ln 〈M/LB〉 ≈ 0.04 and was therefore not con-sidered. This effect comes into play at a red-shift of z >∼ 0.4 and has dramatic effect at highredshifts of about z ≥ 0.8. In the analysis ofthe field early–type galaxies and the derivationof the formation epochs in the following chap-ter, the correction of the progenitor bias will beaccounted for.

As a conclusion from the results of the scaling re-lations so far, for the whole sample of 34 E+S0cluster galaxies in A 2218 and A 2390 and 10early–type galaxies in the Low–LX clusters onaverage only a moderate amount of luminosityevolution is derived. From these findings it canbe inferred that at a look-back time of ∼2.8 Gyrsmost early–type galaxies in the rich clusters con-sist of an old stellar population with the bulkof the stars formed at a high formation redshiftof about zform ≥ 2. The elliptical galaxies inthe poor clusters however, suggest a trend for astronger evolution which favours a lower forma-tion epoch. A detailed discussion on the stellarpopulation ages will be presented in section 7.2.

In the following, the findings for the early–typecluster galaxies will be compared to previous re-sults in the literature.

6.6 Comparison with PreviousStudies

In the sections above it has been shown thatthe scaling relations of the Faber–Jackson andthe Fundamental Plane show a stronger luminos-ity evolution for lower–mass systems. To ver-ify these findings the results will now be com-pared to previous studies. The redshift intervalof 0.05 < z <∼ 0.2 is less well explored than thehigher redshift range as the main focus of cur-rent interest is on redshifts of z ≥ 0.5 and be-yond. There is only one study by Jørgensen etal. (1999) which analysed two galaxy clusters atz ∼ 0.2.Jørgensen et al. (1999, hereafter JFHD) ob-tained ground–based photometry of 31 and 61E+S0 galaxies in the clusters A 665 (z = 0.181)and A 2218 (z = 0.176). Follow–up spectroscopyin A 665 and A 2218 was performed for 6 and9 early–type galaxies, respectively. For a sub–set of objects also archival HST imaging is avail-able to be able to construct the FP. In Fig. 6.21the A 665 and A 2218 data of JFHD are shownin comparison to the FP sample of the rich(A 2390, A 2218, circles) and Low–LX clusters(Cl 0849, Cl 1701 and Cl 1702, squares) for theedge-on projection of the FP. In order to be con-sistent with this work, the absolute magnitudesof JFHD have been re–computed to the concor-dance cosmology. Furthermore, two galaxies inthe JFHD sample are based on ground–basedimaging only but included in the FP. For thisreason, for internal consistency between theirdata, JFHD calibrated their HST images to theground–based IC photometry and therefore ap-plied a zero–point offset to their HST photome-try. For the analysis here, these offsets for A 665of 0.065 mag and for A 2218 of 0.016 mag weretaken back and the two galaxies of A 2218 with-


Figure 6.21: Comparison between the cluster sam-ples and the distant FP samples of Jørgensen et al.(1999). The local Coma FP is indicated by the princi-pal component fit. Open symbols denote the galaxiesin A 2390, A 2218 and in Cl 0849, Cl 1701 and Cl 1702,solid symbols the galaxies in A 665 (pentagons) andA 2218 (triangles).

out HST photometry were discarded.Both JFHD samples comprise on average moremassive galaxies than the FP samples in thisstudy and have a larger zero–point offset in theFP and a slightly stronger evolution. The six gal-axies in A 665 show a median offset to Coma of〈µe〉 = −0.49m and the seven galaxies in A 2218are offset by 〈µe〉 = −0.40m. However, the scat-ter is quite large for both clusters and there areonly two objects which are less massive. More-over, the massive galaxies in the JFHD samplesare also more luminous 〈µe〉 = 19.27m comparedto our FP rich cluster sample 〈µe〉 = 20.04m.Both, the rich and poor cluster galaxies primar-ily populate the less–massive end of the FP. AsJFHD targeted due to instrument limitations thebrightest cluster galaxies their small sample issubject to a potential selection effect. A biasto more massive galaxies is detected which con-tain mainly old quiescent populations and thus

a modest evolution is measured for the FP. Al-though JFHD favour a high formation redshiftof zform > 1.7 for their sample, they argue thatduring the last 4 Gyrs possible star formationactivity has taken place in their cluster galaxies.

Chapter 7: Environmental Effects on Galaxy Properties 155

Chapter 7

Environmental Effects on GalaxyProperties

Most local studies of early-type galaxies didnot reveal any dependence of their propertieson a specific environment. Various approacheshave been undertaken, such as the construc-tion of scaling relations to deduce the luminos-ity evolution (Bender et al. 1992; Bernardi etal. 2003), the measurement of absorption line-strengths to determine the mean ages and metal-licities (Kuntschner 2000; Thomas et al. 2005)or the analyses of luminosity functions for thefield and cluster environments (De Propris et al.2003). On the other hand, some evidence foryoung populations in local S0 galaxies was foundby stellar population studies of the Coma clus-ter. More than 40% of the lower–luminous S0galaxies (MB >∼ −19m) have undergone star for-mation in their central regions during the last∼5 Gyrs, while such activity is absent in theellipticals (Poggianti et al. 2001). In a recentstudy on the Coma cluster two classes of S0 gal-axies were identified, one group is composed ofold ∼10 Gyrs stellar populations similar to ellip-ticals and a second one has very young averageages of ∼2 Gyrs (Mehlert et al. 2003). Howeverit is still unclear, whether these different galaxytypes are specific to the Coma cluster environ-ment or a more general property of all galaxyclusters. So far, there is only one investigationat higher redshift which concentrates on disen-tangling possible differences of early-type clustergalaxies at z ∼ 0.4 (Moran et al. 2005). For the

FP of the rich cluster Cl 0024+1654 an increasedscatter and a radial trend of the M/L ratios wasfound, with the oldest galaxies located in thecluster core and younger ones in the outer pe-riphery regions. Discrepancies between the sub-populations of elliptical and S0 galaxies were notdetected. However, this work is restricted to thespecial environment of rich galaxy clusters andup to now no studies have tested possible envi-ronmental trends of the properties of early–typegalaxies by comparing a range of different envi-ronments at z = 0.2. One reason for this lack ofknowledge is the requirement of a large sampleto explore the existence even of small differences.In section 7.1 the dependence of the cluster gal-axy properties within the Faber-Jackson Rela-tion is analysed. The rich and poor clustersamples are divided with respect to luminos-ity (section 7.1.1), velocity dispersion, which isrepresentative for the mass of a galaxy, (sec-tion 7.1.2) and distance from the cluster centre(section 7.1.3). A possible connection betweenthe colours of galaxies with their projected clus-tercentric radii will be established in section 7.1.4and the colours will be compared to the luminos-ity evolution as derived from the Faber-Jacksonrelation. For the comparisons of the variousproperties, the two rich cluster samples and thethree poor cluster samples have been combinedto increase the statistics. As illustrated in sec-tion 5.6, a combination of the data sets is ade-

156 Chapter 7: Environmental Effects on Galaxy Properties

quate and justified because the properties of thegalaxies within the two rich clusters are very sim-ilarly distributed. The combined galaxy popula-tions of the Low–LX clusters form also a homo-geneous group with respect to their overall pa-rameters such as luminosities, colours, velocitydispersions and sizes, which are comparable tothose of a rich galaxy cluster. Afterwards dif-ferences in the stellar populations of early-typeelliptical and S0 galaxies for the cluster and fieldenvironments are explored and constraints ontheir formation redshift presented (section 7.2).

7.1 The Faber–Jackson Rela-tion

The parameter space of the rich cluster sam-ple of early-type galaxies covers the ranges−20.42 ≥ Mr ≥ −23.42 in absolute magnitudeand 105 km s−1≤ σ ≤ 381 km s−1 in velocity dis-persion with median values of 〈Mr〉 = −21.49and 〈σ〉 = 170 km s−1. For a given galaxy withcorresponding values for Mr and σ, the meanresiduals from the local FJR, i.e. the offsets fromthe local FJR fit can be calculated by the follow-ing

∆Mr = 6.82 log σ + 5.90 +Mr. (7.1)

As previously mentioned (see also Eq. 6.1 onpage 129), distant early-type galaxies are likelyto comprise stellar populations which are on av-erage younger than those of early-types in thelocal Universe. Therefore, the luminosities ofthe distant E+S0 galaxies should be higher fora given mass (or a given σ). This was alreadydemonstrated in Fig. 6.2 with an overall bright-ening of the distant cluster galaxies by∼0.3 mag-nitudes in the Gunn r-band.

7.1.1 Luminosity Dependence

Dividing our sample with respect to luminosityat Mr = −21.49 in two groups of equal num-ber with each sub–sample containing 48 galax-ies, the offsets of the FJR from the local FJR

Figure 7.1: Offsets of the Faber–Jackson relation inGunn r for the early-type cluster galaxies in A 2390and A 2218 from the local FJR by the J99 Coma sam-ple as a function of luminosity Mr. Filled symbols de-note brighter galaxies with Mr < −21.49, open sym-bols fainter galaxies with Mr > −21.49. The dashedline indicates a zero offset.

are explored. Fig. 7.1 shows the offsets of theFJR in Gunn r for the early-type galaxies inthe clusters A 2390 and A 2218 from the localFJR as defined by the J99 Coma sample as afunction of luminosity Mr. Brighter galaxieswith Mr < −21.49 are indicated as solid sym-bols, fainter objects with Mr > −21.49 as opensymbols. Subdividing the total sample with re-spect to their luminosity Mr at Mr = −21.49(0.2 mag brighter than L∗r) into two sub-samplesof equal number, the brighter cluster galaxies(Mr < −21.49) show on average an evolutionwith ∆M r = −0.39± 0.13m. Fainter objects inthe rich cluster sample (Mr > −21.49) are on av-erage more luminous by ∆M r = −0.25± 0.18m.Table 7.1 gives a comparison between the twogroups of early-type galaxies with different lumi-nosities. The difference in mean luminosity be-tween lower-luminous and higher-luminous gal-axies is within their 1σ error and therefore neg-


Figure 7.2: Offsets of the Faber–Jackson relationin Gunn r for the morphologically classified ellipticaland S0 galaxies in A 2390 and A 2218 from the localFJR by the J99 Coma sample as a function of Mr.Ellipticals are indicated as solid circles, S0s as opencircles. The vertical line denotes the border betweenbrighter and fainter galaxies at Mr = −21.49 and thedashed line a zero offset. Lenticulars show on averagea stronger evolution than ellipticals.

ligible.For the offsets from the local FJR in Fig. 7.1, theerrors of the offsets on the absolute magnitudeswere computed as

δ(∆FJR)2 = δM2r clus + δM2

r J99. (7.2)

The first term on the right hand side correspondsto the total error in absolute r magnitude asgiven in Eq. 4.14 on page 91 and the second termare the errors in absolute r magnitude for theComa J99 data. Individual errors for the J99data were adopted as published by the authorswith mean errors of δMr = 0.09m on the pho-tometry.The galaxy samples with fainter and brighterluminosities exhibit different mean velocity dis-persions (log σ = 2.18, and 2.31, respec-tively). For this reason, a two-dimensional

Table 7.1: Evolution of the Faber–Jackson relationin Gunn r derived for various rich cluster samples.N is the number of galaxies and ∆Mr indicates themean luminosity evolution. The fourth column de-notes the ±1σ deviation in the mean luminosity evo-lution and the last column ∆ 〈Mr〉 gives the medianevolution.


(1) (2) (3) (4) (5)

A 2218 48 −0.31 0.15 −0.29A 2390 48 −0.32 0.29 −0.35A 2218+A 2390 96 −0.32 0.22 −0.35low La 48 −0.25 0.18 −0.29high L 48 −0.39 0.13 −0.44low massb 48 −0.62 0.34 −0.63high mass 48 −0.02 0.16 −0.01inc 48 −0.26 0.20 −0.24out 48 −0.35 0.20 −0.40

a fainter: Mr > −21.49, brighter: Mr < −21.49.b less-massive: logσ < 2.231,more-massive: logσ > 2.231.c in: only core sample R < 455.6 kpc,out: only outer region sample R > 455.6 kpc.

Kolmogorov-Smirnov (KS) test is not the ap-propriate statistical method for a comparison.Instead, bootstrap-bisector fits to the individ-ual sub–samples of lower and higher luminosi-ties are performed. A bootstrap bisector fit tothe lower-luminosity cluster galaxies restrictedto Mr > −21.49 gives

Mr = −(2.65±0.46) log σ−(15.21±1.00), (7.3)

whereas the same fitting method to the higher-luminosity ones with Mr < −21.49 yields

Mr = −(2.85±0.53) log σ−(15.47±1.23). (7.4)

Fainter and brighter galaxies have slightly dif-ferent slopes but the difference is not statisti-cally significant. An insignificant slope change


Table 7.2: Evolution of the Faber–Jackson relationin Gunn r derived for various Low–LX samples. N isthe number of galaxies and ∆Mr indicates the meanluminosity evolution. (4) denotes the ±1σ deviationin the mean luminosity evolution and the last column∆ 〈Mr〉 gives the median evolution.


(1) (2) (3) (4) (5)

Cl 0849 15 −0.30 0.19 −0.36Cl 1701 5 −0.42 0.19 −0.20Cl 1702 7 −0.57 0.20 −0.51Low–LX 27 −0.44 0.19 −0.38low La 13 −0.23 0.22 −0.36high L 14 −0.55 0.13 −0.26low massb 13 −0.69 0.73 −0.71high mass 14 −0.12 0.15 −0.07inc 13 −0.30 0.17 −0.15out 14 −0.48 0.18 −0.45

a fainter: Mr > −21.77, brighter: Mr < −21.77.b less-massive: logσ < 2.215,more-massive: logσ > 2.215.c in: only core sample R < 474.4 kpc,out: only outer region sample R > 474.4 kpc.

of ∆a = 0.2 ± 0.5 between the sub–samples isfound.An analysis of the FP for the rich cluster gal-axies in the previous chapter 6, revealed thatthe lenticular galaxies show a larger offset fromthe local FP relation and a stronger evolutionthan the elliptical galaxies. To verify this re-sult with the FJR, the FJR offsets of early-typecluster galaxies in A 2390 and A 2218 from thelocal FJR are displayed in Fig. 7.2, similar toFig. 7.1 but now divided into the morphologicaltypes of elliptical and S0 galaxies. For this anal-ysis, the total rich cluster sample is restricted tothe 33 E+S0 galaxies which enter the FP andwhere a morphological classification based uponHST images is available. The bright cD gal-

Figure 7.3: Offsets of the Faber–Jackson relation inGunn r for the early-type cluster galaxies in the Low–LX clusters CL 0849, CL 1701 and CL 1702 from thelocal FJR by the J99 Coma sample as a function ofluminosity Mr. Filled symbols denote brighter gal-axies Mr < −21.77, open symbols fainter galaxiesMr > −21.77. The dashed line corresponds to a zeroevolution.

axy of A 2218 was excluded as no counterpartof this galaxy type was observed in A 2390. InFig. 7.2 the sample of ellipticals and S0 galaxiesis shown as solid and open symbols. The higher-luminosity Es (Mr < −21.49) are brighter by〈∆M r E〉 = −0.14 with a scatter σMr = 0.11m,lower-luminosity Es (Mr > −21.49) show an evo-lution of 〈∆M r E〉 = −0.23 with σMr = 0.18m.The brighter S0s are offset to the local FJR re-lation by 〈∆M r S0〉 = −0.66 with σMr = 0.12m,fainter S0s by 〈∆M r S0〉 = −0.45 with a scatterof σMr = 0.13m. These results confirm the anal-ysis of the FP where also the lenticular galax-ies are more luminous than the ellipticals. Bothsub–samples of brighter and fainter S0s show astronger evolution than the Es. An interesting is-sue would be also if there is a difference betweenthe sub–groups of faint and bright ellipticals andS0 galaxies. Unfortunately the sub–samples are


too small to test this in greater detail.In order to investigate the offsets of the FJR ofthe Low–LX clusters from the local FJR withrespect to luminosity, the poor cluster sample isdivided at Mr = −21.77 in two groups with thehigh-luminosity sub-sample containing 14 galax-ies and the low-luminosity counterpart compris-ing 13 objects. Fig. 7.3 shows the offsets ofthe FJR in Gunn r for the early-type galaxiesin the Low–LX clusters CL 0849, CL 1701 andCL 1702 from the local FJR by the J99 Comasample as a function of luminosity Mr. Brightergalaxies with Mr < −21.77 are represented assolid symbols, fainter galaxies with Mr > −21.77are denoted as open symbols. Brighter clus-ter galaxies indicate on average an evolution of∆M r = −0.55 ± 0.13m, fainter objects in thepoor cluster sample (Mr > −21.77) are on aver-age more luminous by ∆M r = −0.23 ± 0.22m.At a first glance, these values suggest an impres-sion of a larger offset for brighter galaxies. How-ever, the scatter of the fainter galaxies is alsolarger than for the brighter ones and the me-dian offsets are similar for both sub-classes. Ta-ble 7.2 compares the individual values for the twogroups of early-type galaxies with different Mr.As the difference in median luminosity betweenlower-luminosity and higher-luminosity galaxiesis within their 1σ error, the differences in thesub-samples are negligible.The Low–LX early-type cluster galaxies were vi-sually classified as nine elliptical galaxies and oneS0 galaxy (cf. section 4.4). As the number ofgalaxies in each sub–class is unbalanced but alsotoo low, a reliable comparison and further anal-ysis as for the rich cluster galaxies cannot beperformed.

7.1.2 Mass Dependence

As the distant galaxies cover a broad range in ve-locity dispersions (2.02 ≤ log σ ≤ 2.58) down to105 km s−1, it is also possible to explore an evo-lution in the slope of the FJR. The Coma sam-ple has a steeper slope than that of the distant

clusters, with a slope difference of ∆a = 1.5±0.4(see Fig. 6.2). This gives some evidence for a dif-ference between massive and less-massive E+S0galaxies. In Fig. 7.4 the offsets of the FJR for theearly-type cluster galaxies in A 2390 and A 2218from the local FJR were computed and plottedas a function of σ. As σ is a measure for the stel-lar mass of a galaxy, a dependence on the gal-axy mass can be tested. Subdividing the totalsample with respect to their velocity dispersionat σ < 170 km s−1(logσ = 2.231) into two sub-samples of equal number, the lower-mass objects(logσ < 2.231) show on average a stronger evolu-tion with ∆M r = −0.62± 0.34m. More massivegalaxies (logσ > 2.231) are on average brighterby ∆M r = −0.02 ± 0.16m. This suggests thatthe luminosity evolution correlates with the gal-axy mass rather than the luminosity of the gal-axies. Table 7.1 gives a comparison between thetwo groups of early-type galaxies. A bisector fitas illustrated in Fig. 7.4 to the offsets for bothFJR samples yields

∆Mr = 4.64± 0.56 log σ − 10.74± 1.33. (7.5)

While galaxies with σ > 170 km s−1 show neg-ligible offsets in Fig. 7.4, the less-massive early-types show a brightening by up to 1.4 magni-tudes. For the offsets from the local FJR inFig. 7.4, the errors in the FJR diagram were de-rived as

δ(∆FJR)2 = δM2r + 6.822 δlog(σ)2. (7.6)

Again, the first term on the right hand side cor-responds to Eq. 4.14 on page 91 and the secondterm is the propagated error of σ based on theJ99 Coma slope.Galaxies with lower and higher mass exhibit dif-ferent mean velocity dispersions (logσ = 2.15,and 2.34, respectively). For this reason, a two-dimensional KS test is not the appropriate sta-tistical method for a comparison. Similar tothe previous section, bootstrap-bisector fits tothe sub-samples of different mass are computed.For the low-mass galaxies (logσ < 2.231) this


Figure 7.4: Offsets of the Faber–Jackson relation inGunn r for the early-type cluster galaxies in A 2390and A 2218 from the local FJR by the J99 Coma sam-ple as a function of σ (mass). Filled symbols denotemore-massive galaxies with logσ > 170 km/s, opensymbols less-massive galaxies with logσ < 170 km/s.The dashed line corresponds to a zero evolution, thesolid line is a bisector fit to all data. For matters ofclarity, individual error bars are shown for the more-massive objects only.

method yields

Mr = −(8.11±1.04) log σ− (3.74±2.25), (7.7)

whereas for the more-massive (logσ > 2.231)counterparts the bisector fitting routine resultsin

Mr = −(5.57± 1.53) log σ− (8.83± 3.58). (7.8)

Low-mass galaxies and high-mass galaxies havedifferent slopes with a slope difference of∆a = 2.5 ± 1.5, which is significant on the 2σlevel based on the uncertainty of the slope forthe high-mass objects. Note that this results de-pends on the defined selection boundaries uponthe Coma reference as the local slope and zero-point were adopted for the derivation of the evo-lution of the distant sub-samples. Nevertheless,

Figure 7.5: Offsets of the Faber–Jackson relation inGunn r for the early-type cluster galaxies in the Low–LX clusters CL 0849, CL 1701 and CL 1702 from thelocal FJR by the J99 Coma sample as a function of σ(mass). Filled symbols denote more-massive galaxieswith logσ > 164 km/s, open symbols less-massivegalaxies with logσ < 164 km/s. The dashed linedenotes a zero offset, the solid line is a bisector fit tothe whole sample.

it remains the trend that less–massive galaxiesfeature a larger luminosity offset which couldbe a possible hint for a faster evolution of thelow-mass galaxies compared to the more mas-sive counterparts. Another possibility is thatthe lower–mass galaxies have on average youngerstellar populations or a broad range in star for-mation histories than galaxies with higher σ.Within the spectra, no signs of strong star forma-tion activity was detected which argues againstyounger population ages. However, differencesmight also occur in the colours or in the metal-licity of the systems with a broader distributionof metal sensitive absorption line indices in thegalaxy spectra.

To explore if a mass-dependent luminosity evolu-tion is also evident in the lower–mass galaxy clus-ters, the sample was divided into two sub–groups


with respect to velocity dispersion at 164 km s−1

(logσ = 2.215). Fig. 7.5 shows the FJR offsetsfor the 13 lower-mass and 14 more-massive E+S0galaxies in the Low–LX clusters as a functionof velocity dispersion and Table 7.2 summarisesthe findings for both groups. The solid line inthe Figure is a bisector to the offsets of the com-bined sample

∆Mr = 3.69± 0.92 log σ − 8.72± 2.07. (7.9)

Galaxies with lower σ exhibit a stronger bright-ening of ∆M r = −0.69±0.73m than their high σcounterparts with ∆M r = −0.12±0.15m. How-ever, the scatter in the evolution, in particularfor the less-massive objects, is quite large. Thereason for this effect comes from a low number ofgalaxies and on average higher uncertainties inthe velocity dispersions which enter the compu-tation of the errors in combination, see Eq. 6.3.Regardless of the scatter, the difference betweenthe two sub-samples is significant. E+S0 gal-axies with large σ show a negligible evolution of〈∆Mr〉 = −0.07m, all low-σ objects feature a lu-minosity evolution with values up to ≈1.7 mag-nitudes and a median of 〈∆Mr〉 = −0.71m. Sim-ilar as for the rich cluster galaxies, bootstrap-bisector fits are executed to assess the signifi-cance of a slope change for the sub-samples ofdifferent mass. A free bisector fit to the less-massive E+S0 galaxies in the low-mass clusters(logσ = 2.215) yields

Mr = −(16.01± 4.63) log σ + (13.02± 9.88),(7.10)

whereas for the more-massive (logσ > 2.215)ones

Mr = −(5.94±3.04) log σ−(8.10±7.22) (7.11)

is found. This relation is in agreement with thefit parameters in Eq. 7.8 as derived for the moremassive sub-sample in rich clusters. The tworelations of Eqs. 7.10 and 7.11 suggest a slopedifference of ∆a = 10.1 ± 4.6, which has a 2σlevel significance assuming the uncertainty of theslope for the less–massive galaxies.

In summary, the E+S0 galaxies in the low–mass clusters show a similar trend with a mass–dependent luminosity evolution as measured forthe early–type galaxies in more–massive clusters.

7.1.3 Radial Dependence

An adjuvant test of possible environmental ef-fects on the properties of galaxies is looking atthe radial distribution of the galaxies within aspecific galaxy cluster. By combining the red-shift measurements and the morphological typeclassification the distribution of cluster membersas a function of projected cluster radius can beexplored. Unfortunately, as the HST field is re-stricted to the central ∼0.6 Mpc of a galaxy clus-ter the change of the morphological mix of gal-axies cannot be investigated. In this central re-gion of a cluster the gravitational potential ofa cluster is steepest and most of the physicalenvironmental mechanisms are simultaneous atwork. Additionally projection effects with gal-axies that are located in the outer regions play arole and have to be accounted for. Although pro-cesses such as tidal stripping and tidal triggeringwhich are only relevant in the cores of clustersare expected to be most dominant, a distinctioninto the various mechanisms which are likely tooperate is not possible.The projected radial distribution of spectro-scopic target galaxies for Abell 2390, CL 0849CL 1701 and CL 1702 is shown in Fig. 7.6.Early–type cluster members which were verifiedthrough spectroscopy are indicated as open cir-cles, non member galaxies as asterisks. Morpho-logically classified ellipticals and S0 galaxies (in-cluding Sa bulges) are represented as solid cir-cles and stars, respectively. As expected, no ra-dial trend for any morphological galaxy type isdetectable within the central parts of the clus-ters. To analyse a possible change of the galax-ies’ properties as function of distance from thecluster centre, the MOSCA spectroscopy was de-signed and carried out to cover a comprehensiverange of radial distances. However, this study is


Figure 7.6: Distribution of early–type cluster galaxies. Spectroscopic confirmed cluster members are denotedas open circles, non member galaxies as asterisks. Morphologically classified ellipticals and S0 galaxies(including Sa bulges) are represented as solid circles and stars, respectively. Upper panel left: Abell 2390 atz = 0.228, right: CL 0849 at z = 0.234, Lower panel left: CL 1701 at z = 0.246, right: CL 1702 at z = 0.223.


not comparable to wide field surveys out to largecluster radii of >∼2 virial radii (e.g., Gerken et al.2004).In order to investigate any possible dependenceon clustercentric radius within the large data set,the projected distance of each galaxy from therespective brightest cluster galaxy (BCG) wasmeasured and the samples subdivided into tworadial bins of equal size so that they compriseequal numbers of cluster members.The average projected radius for the rich clustergalaxies in the core region is 244 kpc comparedto 724 kpc for galaxies in the outer bin. For theLow–LX clusters the average projected radii are128 kpc and 740 kpc for galaxies in the centraland outer districts, respectively. Based on therelation between the virial mass of a cluster andits projected velocity dispersion (Girardi et al.1998), the virial radius Rv can be derived as

Rv ∼ 0.0035(1 + z)−1.5σp h−1, (7.12)

where the Rv is in units of Mpc and theprojected velocity dispersion σp is given inkm s−1. For A 2218, σp = 1370+160

−120 km s−1isadopted (Le Borgne et al. 1992) and forA 2390 Rv = 3.156h−1

100 Mpc is used (Carl-berg et al. 1996). For our assumed cosmol-ogy, this yields to virial radii for the clustersA 2218 and A 2390 of Rv = 3.765 ± 0.440 Mpc(1268 ± 148′′) and Rv = 2.209 ± 0.165 Mpc(605 ± 45′′), respectively. Adopting the clus-ter velocity dispersions as listed in Table 2.3 onpage 31 from the low–resolution sample (Baloghet al. 2002b), virial radii for the Low–LXclusters were derived. This results in virialradii for the poor clusters CL 0849, CL 1701 andCL 1702 of Rv = 1.950± 0.230 Mpc (523± 62′′),Rv = 2.099 ± 1.629 Mpc (543 ± 421′′) andRv = 0.999 ± 1.102 Mpc (278 ± 307′′), respec-tively.Fig. 7.7 displays the FJR offsets for the early–type rich cluster galaxies as a function of dis-tance from the cluster centre R, with the averageprojected radius R given in units of kpc. Each

Figure 7.7: Offsets of the Faber–Jackson relation inGunn r for the early-type cluster galaxies in A 2390and A 2218 from the local FJR by the J99 Coma sam-ple as a function of distance from the cluster cen-tre R in units of kpc. The cluster centre is definedby the central cD galaxy (BCG). Filled symbols de-note galaxies in the core region with R < 455.6 kpc,open symbols are galaxies in the outer districts withR > 455.6 kpc. The dashed line indicates a zero off-set, the solid line is a χ2-fit to the complete sample.

cluster centre is defined by the brightest clustergalaxy (BCG), the central cD elliptical galaxy.The cluster galaxies were divided into two radialbins of the core region (R < 455.6 kpc) and theouter districts (R > 455.6 kpc), each compris-ing 48 galaxies. Applying a linear χ2-fit to thedistant cluster sample in Fig. 7.7 of

∆Mr = −(0.0003± 0.0002)(R

kpc

)−0.2015± 0.1214 , (7.13)

a slight gradient of the FJR offsets with increas-ing clustercentric radius R is found.To test the significance of this gradient for thetwo sub–samples of inner and outer radial bin,the bootstrap bisector technique is performed. Abisector to the galaxies lying in the core region


Figure 7.8: Offsets of the Faber–Jackson relationin Gunn r for the early-type cluster galaxies in theLow–LX clusters CL 0849, CL 1701 and CL 1702 fromthe local FJR by the J99 Coma sample as a functionof distance from the cluster centre R in units of kpc.The cluster centre is defined by the central cD galaxy(BCG). Filled symbols denote galaxies in the coreregion with R < 474.4 kpc, open symbols are galaxiesin the outer districts with R > 474.4 kpc. The dashedline indicates a zero offset, the solid line is a χ2-fit tothe complete sample.

with R < 455.6 kpc gives

Mr = −(5.29±1.15) log σ−(9.60±2.59), (7.14)

whereas for the galaxies in the outer districtsR > 455.6 kpc the method yields

Mr = −(5.39±0.51) log σ−(9.48±1.13). (7.15)

The slope difference of ∆a = 0.1 ± 1.2 betweengalaxies in the central and outer bins is statis-tically insignificant as the scatter is larger thanthe detected offset.The mean and median evolution of the E+S0galaxies for the central and the outer regionare shown in Table 7.1. As the galaxies in thecore and outer region feature similar mean veloc-ity dispersions (logσ = 2.26, and 2.23, respec-tively), the two–dimensional KS test provides a

good measure of consistency. Applying the KStest results in a probability of P = 0.46 that thetwo distributions are similar. The slope for thegalaxies in the core region is similar to the slopederived for the galaxies in the outskirts. Look-ing at the residuals from the local FJR, there isa slight trend with clustercentric radius, but notsignificantly (within ±1σ). On average, galaxiesin the outskirts of the cluster indicate a larger lu-minosity offset of −0.35±0.20m than in the coreregion −0.26 ± 0.20m. However, the offset hasonly a low significance and its not sure whetherthis gradient is real or maybe just a projectioneffect.The ∆Mr −R distribution of the early-type gal-axies in the low–massive clusters is shown inFig. 7.8. Cluster members were divided into tworadial bins of a central core (R < 474.4 kpc) andan outer region (R > 474.4 kpc), comprising 13and 14 galaxies, respectively. The cluster centrewas adopted to be the location of the BCG. Asimilar trend is seen as for the rich cluster sam-ple, albeit with a larger scatter. A linear χ2-fitto the Low–LX members gives

∆Mr = −(0.0004± 0.0003)(R

kpc

)−(0.2071± 0.1951) , (7.16)

which is indicated as the solid line in Fig. 7.8. Onaverage, the residuals from the local FJR indi-cate a slight trend with increasing projected ra-dius, although it is within the ±1σ uncertainty.Table 7.2 gives a comparison between the coresample and the galaxies in the outskirts for thelow massive clusters. Galaxies far away from thecluster centre have an offset of −0.48 ± 0.18m,whereas in the central part only an brighteningof −0.30± 0.17m is measured. This difference iscomparable to the result as derived for the envi-ronments of rich clusters. A further comparisoninvolving the KS method is not possible as thesub-samples contain less than 20 galaxies wherethe KS test has no statistical significance.Fig. 7.9 displays the offsets of the FJR of theearly-type galaxies in the rich clusters from the


Figure 7.9: Offsets of the Faber–Jackson relation inGunn r for the early-type cluster galaxies in A 2390and A 2218 from the local FJR by the J99 Coma sam-ple as a function of distance from the cluster centreR, normalised by the virial radius Rv. The clustercentre is defined by the central cD galaxy (BCG).Filled symbols denote galaxies in the core region withR/Rv < 0.16, open symbols are galaxies in the outerdistricts with R/Rv > 0.16. The dashed line indi-cates a zero offset, the solid line is a χ2-fit to thecomplete sample.

local FJR as a function of distance from the clus-ter centre R, normalised to the virial radius Rv.Again, the central cD galaxy was assumed to bethe cluster centre. Galaxies in the core region(R/Rv < 0.16) are represented as filled symbols,galaxies in the outer districts (R/Rv > 0.16) asopen symbols. The linear χ2-fit to the distantgalaxies in in Fig. 7.9 is

∆Mr = −(0.65± 0.53)(R

Rv

)− (0.20± 0.12).

(7.17)In both Figs. 7.7 and 7.9, the errors of the off-sets on the absolute magnitudes in the FJR dia-grams were computed as given in Eq. 7.2. As aconsequence of the normalisation with the virialradius, the trend of the FJR offsets with increas-

Figure 7.10: Offsets of the Faber–Jackson relationin Gunn r for the early-type cluster galaxies in theLow–LX clusters CL 0849, CL 1701 and CL 1702 fromthe local FJR by the J99 Coma sample as a functionof distance from the cluster centre R, normalised bythe virial radius Rv. The cluster centre is definedby the central cD galaxy (BCG). Filled symbols de-note galaxies in the core region with R/Rv < 0.25,open symbols are galaxies in the outer districts withR/Rv > 0.25. The dashed line indicates a zero offset,the solid line is a χ2-fit to the complete sample.

ing radius R detected in Fig. 7.9 is even moredominant. In numbers, galaxies in the outskirtsof the cluster shown in this projection a slightsmaller luminosity offset of 〈∆Mr〉 = −0.34m

than their counterparts in the central region of〈∆Mr〉 = −0.40m. However, both numbers areconsistent within their 1σ deviation. The clustergalaxies cover a radial range out to (R/Rv) ∼ 0.5(1.1 Mpc), which is in case of A 2390 roughly thebeginning of the transition region of the clus-ter that is far from the cluster centre but stillwithin the virial radius of the cluster. Therefore,still many physical processes are simultaneouslyat work which could account for this slight ob-served trend. Star formation induced throughtidal triggering or tidal halo stripping are proba-


bly the most dominant mechanisms at work. Inthe transition region these two processes becomeless likely and effects such as ram pressure strip-ping, which is the removal of the galactic gas bypressure exerted by the intercluster medium, or aslow decrease of star formation activity (“stran-gulation”) come into play. In addition, harass-ment (i.e. high speed interactions between gal-axies in the gravitational potential of the cluster)is effective over almost all clustercentric regions,from the center to the outskirts of a galaxy clus-ter. The rich cluster sample does not reach theouter districts of the clusters such as the com-plete transition or periphery zone, where a pos-sible radial gradient with luminosity would bemore apparent. To test the origin of the radialtrends in even greater detail, an analysis of thegalaxy colours as a function of projected radiiwill be performed in section 7.1.4.Looking at the offsets of the FJR for the poorcluster galaxies ∆Mr as a function of R/Rv inFig. 7.10, a steeper gradient with clustercentricradius compared to the rich counterparts isfound. Galaxies in the outskirts (R/Rv > 0.25),and in the core region (R/Rv < 0.25) ex-hibit an evolution of 〈∆Mr〉 = −0.51m and〈∆Mr〉 = −0.07m, respectively. Assuming a lin-ear χ2-fit to the whole sample of

∆Mr = −(0.62± 0.42)(R

Rv

)− (0.21± 0.17),

(7.18)gives a similar relation as for the rich clustergalaxies but a somewhat steeper slope. How-ever, as the virial radii of the Low–LX clustersare smaller their members are distributed over alarger range of distances out to ∼1 Rv. As a con-sequence, a stronger radial gradient is detected.Nevertheless, two issues have to be treated withcaution. First, the low number of galaxies in theouter districts beyond ∼0.5 R/Rv used in theanalysis of radial distribution can mimic a higheroffset. Second, an explanation for a stronger ra-dial trend may be the result of different pro-cesses which occur at larger clustercentric dis-

tances than for the rich clusters. The detectednet effect might also include mechanisms suchas strangulation and ram pressure stripping. Inaddition, the truncation of star formation couldalso operate differently than in rich clusters asthe intercluster medium for lower-mass clustersis less effective and exhibits lower densities thatfor more massive clusters.

7.1.4 A Dependence on GalaxyColours?

In a previous study of the cluster A 2390, Abra-ham et al. (1996a) examined a wide field outto ∼5 Mpc using moderate S/N spectroscopyand ground-based imaging to study the infallingfield galaxy populations. A radial gradient in the(g − r) colour for galaxies in the outer districts(periphery) of 1.1 to 3.2 Mpc was found, wherethe colours become predominantly bluer with in-creasing projected radius. In contrast, galax-ies in the central regions <∼0.43 Mpc show red,evolved populations with old ages of>∼8 Gyrs, as-suming the Bruzual & Charlot (1993; GISSEL93version) stellar population models.It is therefore interesting to test if a radial trendwith the colours of the cluster galaxies existsalso in this study. For this reason, the rest-frame (B − I) colours for the two rich galaxysamples were used which show a very homoge-nous distribution, with galaxies in A 2390 en-compassing a range 1.86 < (B − I) < 2.51and in A 2218 1.95 < (B − I) < 2.44. TheE+S0 galaxies in A 2390 exhibit a median colourof 〈(B − I)〉 = 2.29, whereas for the galax-ies in A 2218 a median of 〈(B − I)〉 = 2.28 isfound. (cf. section 5.6 and Fig. 5.28 on page125). Fig. 7.11 shows the rest–frame (B − I)colour as a function of distance R from the clus-ter centre for 95 early-type galaxies. One spi-ral (Sa) galaxy outlier with very blue colour of(B − I) = −1.17 is not displayed. The errors onthe rest–frame colours were derived via the linearrelation δ(B− I) = δB+ δI + δk, where δB andδI are the errors in the respective ground-based


Figure 7.11: Rest–frame (B−I) colour as a functionof clustercentric distance (in units of Mpc) for theearly-type galaxies in A 2390 and A 2218. The centralcD galaxy was assumed to define the cluster centreand the solid line is a χ2-fit to the combined data.A slight radial trend of (B − I) galaxy colour withradius is seen.

filters and δk the uncertainty in the k-correction.A linear χ2-fit to 94 galaxies with (B − I) > 1.9gives

(B − I) = −(0.076± 0.043)(R

kpc

)+(2.302± 0.024) , (7.19)

which is indicated as the solid line in Fig. 7.11.The slight colour gradient with increasing pro-jected radius is significant on the 2σ level andis also found within the sub–samples of A 2390and A 2218. Overall, this trend of (B−I) colourwith R is weak but already detected within thecentral 1.1 Mpc, which corresponds to half of theclusters virial radii.Looking at the dispersion in the colours withincreasing clustercentric distance, for the outerregions a larger scatter is found. Galaxieswith (B − I) > 2.0 located in central partswithin (R/Rv) ∼ 0.2, which corresponds to

Figure 7.12: Offsets of the Faber–Jackson relationin Gunn r for 95 early-type galaxies in A 2390 andA 2218 from the local FJR as a function of rest–frame(B−I) colour. Filled symbols denote redder galaxieswith (B − I) > 2.287, open symbols indicate bluergalaxies with (B − I) < 2.287. A zero evolution isshown by the dashed line.

a radial projected distance of 0.44 Mpc fromthe cluster centre of A 2390 and 0.75 Mpc inthe case of A 2218, show a colour distributionin (B − I) of 2.03 < (B − I) < 2.44 (me-dian of 〈(B − I)〉 = 2.301) with a scatter ofσ(B − I) = 0.095. Cluster members far awayfrom the centre are distributed over a slightlybroader range in colours 2.00 < (B − I) < 2.51(〈(B − I)〉 = 2.286) with an approximately 35%larger scatter of σ(B − I) = 0.128.The ∆Mr − (B − I) diagram is displayed inFig. 7.12. A clear trend of the FJR offsets withrest–frame (B − I) colour is seen. Bluer galax-ies defined as (B − I) < 2.287 (open symbols)show an evolution of −0.50± 0.23m, whereas forthe redder galaxies with (B − I) > 2.287 (filledsymbols) an insignificant offset of −0.14± 0.18m

is found. In Table 7.3 a detailed comparison ofthe evolution as derived from the FJR for dif-ferent sub-samples is given. In particular it is


Table 7.3: Evolution of the Faber–Jackson relationin Gunn r as a function of rest–frame (B − I) colourfor the rich clusters A 2218 and A 2390. N is thenumber of galaxies and ∆Mr indicates the mean lu-minosity evolution. The fourth column denotes the±1σ deviation in the mean luminosity evolution andthe last column ∆ 〈Mr〉 gives the median evolution.


(1) (2) (3) (4) (5)

(B − I) > 2.287 48 −0.14 0.18 −0.13E 10 0.09 0.17 0.14S0 11 −0.28 0.12 −0.45(B − I) < 2.287 48 −0.50 0.23 −0.45E 6 −0.41 0.23 −0.30S0 6 −0.84 0.24 −0.72

interesting, if the morphological types of ellipti-cal and S0 galaxies show a trend with galaxycolour. Regardless of the colour cutoff, bothgroups of S0 galaxies exhibit a stronger evolu-tion, which is at least by a factor of two largerthan ellipticals. In case of the redder galaxies,the ellipticals even have on average an insignifi-cant (“positive”) evolution. These galaxies formmost likely the evolved, passive stellar popula-tions with old ages of >∼8 Gyrs that populatethe red–sequence in the colour magnitude rela-tion of galaxy clusters. Partly, this result is ex-pected as bluer stellar populations are on aver-age younger and therefore also show an excess intheir luminosities. This analysis showed that thewhole population of blue galaxies are the domi-nant contribution to the measured FJR offsets.

The results on the colour gradients give furtherevidence that the luminosity is the main driverfor the evolution of the distant early-type galax-ies in the FJR but the offsets are not caused dueto differences in the kinematics of the local anddistant scaling relations (see also the discussionon the σ distributions in section 5.4 on page 114).

7.2 Stellar Population Ages

The observed evolution of the M/L ratio as de-rived from the FP depends on the age of thegalaxies’ stellar population. In general, the lu-minosity of a young stellar population becomesrapidly fainter when the massive and bright starswhich have a short lifetime disappear. For an oldpopulation comprising mainly low mass stars thedimming of the luminosity is on a more gradualevolutionary path. As a consequence of this, astellar population formed at lower redshift willevolve faster than one generated at high red-shift. In the following, the observed evolution ofthe M/L ratio will be compared to simple stel-lar population models of a single burst formed atredshift zform. The models have been generatedfollowing description of the analytic models byvan Dokkum & Franx (2001b).The luminosity evolution of a single-age stellarpopulation can be described by a power law as

L ∝ 1(t− tform)κ

, (7.20)

where tform is the stellar formation time whichcorresponds to a formation redshift of zform (e.g.,Tinsley 1980). The coefficient κ depends onthe initial mass function (IMF), the metallicityand the chosen passband in which the luminos-ity is measured. In the models, a normal IMFwith Salpeter (1955) slope, solar metallicity andκB = 0.96 and κr = 0.79 for the rest–frame Band Gunn r-band was adopted (BC96). Modelpredictions are shown in Fig. 7.13 for the richand poor clusters in the rest–frame Gunn r-bandand in Fig. 7.14 for the field elliptical and S0 gal-axies in the rest–frame B-band. Note that thepredicted evolution of the models is independentof H0 as the age dependence of the M/L ratiois a power law. Two different model tracks for asingle stellar population with formation redshiftzform = 1 (lower line) and zform = 2 (upper line)are indicated as the dotted lines in Figs. 7.13,7.14 and 7.15.To assess the evolution of the stellar populations


Figure 7.13: The average evolution of the M/L ra-tio of the elliptical and S0 galaxies in the rich clus-ters, ellipticals in the poor clusters and cluster gal-axies taken from van Dokkum & Stanford (2003),in the rest–frame Gunn r-band. The small dot isthe average M/L ratio of the Coma cluster galax-ies. Dotted lines are model tracks for a single stel-lar population with formation redshift 1 (lower line)and 2 (upper line). The solid line represents the sin-gle burst evolution of a stellar population formed atthe mean cluster formation redshift of zform = 3 formasses M > 1011M. In high–mass clusters S0sevolve faster than ellipticals, whereas elliptical galax-ies in low-mass clusters evolve as fast as S0 galaxiesin rich clusters.

in the early–type galaxies, the evolution of theFP zero-point γ for the distant galaxies can bedirectly converted into an evolution of the aver-age M/L ratio as

〈∆log(M/L)〉 = −〈∆γ〉2.5β

, (7.21)

where the average evolution of ∆log(M/L) wasderived via the average zero-point offset of thedistant FP 〈∆γ〉 from the local Coma FP (cf.section 1.3.3 on page 12).Fig. 7.13 displays the evolution of the M/L de-rived as the average offset from the local FP inthe rest–frame Gunn r-band for the ellipticals

and S0 galaxies in the rich clusters (large filledand open circles) and the ellipticals in the poorclusters (large stars). The cluster galaxies arecompared to the cluster sample by van Dokkum& Stanford (2003). These authors analysed acompilation of massive galaxies (M > 1011M)in rich clusters taken from the literature andfound an average formation redshift of zform = 3.The average M/L ratio of the cluster galaxies inComa is indicated as the small dot at z = 0.024.Cluster galaxies are in compliance with a slowevolution of the bulk of their stars and a highformation redshift of zform >∼ 2.5 which is de-noted with the single burst evolution of a stellarpopulation formed at zform = 3 (solid line). Theaverage offset from the local FP 〈∆γ〉 was de-rived from the average stellar populations of onemorphological type (elliptical and S0 galaxies)and the scatter is similar to the average uncer-tainty for the massive galaxies in a single cluster.

Lenticular galaxies in the rich clusters show alarge evolution in their M/L ratios and arein agreement with a low formation redshift1 ≤ zform ≤ 2 and a fast evolution. In contrast,elliptical galaxies in rich cluster exhibit smalleroffsets from the local FP which suggests a higherformation redshift zform ≥ 2. Elliptical galax-ies in poor cluster indicate an opposite trend towhat is found for the ellipticals in higher-massclusters. These galaxies appear to evolve as fastas S0 galaxies in rich clusters.

In a recent study of 27 field early–type galaxiesbetween 0.6 < z < 1.15 a mass–dependentevolution was found (van der Wel 2005). Thissample of high–redshift galaxies showed that theevolution of low–mass field galaxies with a char-acteristic masses of M < 2 × 1011M is fasterthan their more massive counterparts. The 21field early–type galaxies in this thesis are locatedat a lower redshift range of 0.2 < z < 0.75.But it is interesting, if already at a look–backtime of ∼5 Gyrs a possible evidence for amass–dependent evolution or at least a trendwith galaxy mass could be detected. For this


Figure 7.14: The average evolution of the M/L ra-tio of the field elliptical and S0 galaxies in the FDFand WHDF and the early–type cluster galaxies takenfrom van Dokkum & Stanford (2003), in the rest–frame B-band. Dotted lines are model tracks fora single stellar population with formation redshift 1(lower line) and 2 (upper line). The model tracks areforced to go through the small dot which representsthe average M/L ratio of the field galaxies at z = 0.02by Faber et al. (1989). The solid line is the bestfitting formation redshift for the evolution of mas-sive cluster galaxies with masses M > 2 × 1011M.Massive field galaxies with M > 2 × 1011M showa similar slow evolution as cluster galaxies, whereasless–massive field galaxies with M < 2 × 1011Mevolve faster, similar to less–massive cluster galaxies.

reason, the average evolution of the M/L ratioof the field elliptical and S0 galaxies in the FDFand WHDF was derived. Fig. 7.14 displays theaverage evolution of the M/L ratios for the fieldgalaxies in the rest–frame B-band comparedto the early–type cluster galaxies taken fromvan Dokkum & Stanford (2003). These authorsdeduced for a compilation of massive early–typegalaxies with M > 2 × 1011M in rich galaxyclusters a formation redshift of zform ≈ 3, whichis indicated as the solid line. All calculatedmodel tracks are forced to go through the aver-age M/L ratio of the field galaxies at z = 0.02

Figure 7.15: The evolution of the M/L ratio ofthe field elliptical and S0 galaxies in the FDF andWHDF and the field early-type galaxies at 〈z〉 = 1taken from van der Wel (2005), in the rest–frame B-band. Dotted lines are model tracks for a single stel-lar population with formation redshift 1 (lower line)and 2 (upper line). The model tracks are forced to gothrough the small dot which represents the averageM/L ratio of the field galaxies at z = 0.02 by Faberet al. (1989). The solid line is the best fitting forma-tion redshift for the evolution of cluster galaxies withmasses M > 2 × 1011M. The scatter in the M/Lratios is large but there is a clear trend for field S0galaxies to evolve faster. Field ellipticals appear toform two separate groups with respect to M/L ratios.

by Faber et al. (1989), shown as the smalldot in Figs. 7.14 and 7.15. The field FDF andWHDF sample was splitted according to thecharacteristic mass at M = 2 × 1011M, whichgives 13 lower–mass and 8 higher–mass galaxies.The average evolution of the higher–mass fieldgalaxies is 〈∆ ln(M/LB)〉 = −0.31 ± 0.06 ata redshift of 〈z〉 = 0.44. Lower–mass galaxiesindicate on average a stronger evolution of〈∆ ln(M/LB)〉 = −0.54 ± 0.07 at 〈z〉 = 0.38.The slight deviation in the redshift is solelydue to the not uniform number of galaxies ineach sub–sample. Massive field galaxies are incompliance with a slow evolution as derived


for the massive cluster galaxies and a highformation redshift for the bulk of the stars ofzform ≈ 3. In contrast, less–massive field galaxiesin the FDF and WHDF with M < 2 × 1011Mevolve faster and follow an evolutionary trackbetween 1 <∼ zform <∼ 2 which is similar to theless–massive cluster galaxies of van Dokkum &Stanford (2003), which is indicated as the opensquare in Fig. 7.14. This result gives evidencethat already at a look–back time of ∼5 Gyrsdifferences in the measured M/L evolutionfor the FDF and WHDF field galaxies can berevealed.

To translate the evolution in the M/L ra-tios into a specific formation redshift zform forthe field galaxies, the evolution of the M/Lratios for the individual field early–type galaxiesare investigated. Fig. 7.15 illustrates the theoffset from the local FP in the rest–frameB-band for elliptical and S0 galaxies in theFDF and WHDF. The comparison sample of27 field early-type galaxies at high redshift0.6 < z < 1.15 by van der Wel (2005) is alsoshown. Again, model tracks for a single-agestellar population with formation redshift 1(lower line) and 2 (upper line) are displayed asdotted lines. All model tracks were constrainedto cross on their way to zero redshift the averageM/L ratio of the field galaxies at z = 0.02by Faber et al. (1989), indicated as the smallblack dot. The evolution of cluster galaxieswith masses M > 2 × 1011M can be bestapproximated assuming a formation redshift ofzform ≈ 3, which is denoted as the solid line inFig. 7.15. Although the scatter in the offsetsfrom the local FP is large, a clear trend for afaster evolution of field S0 galaxies is found. Onthe other hand, field elliptical galaxies appear toform two separate groups with respect to M/Lratios. To assess the formation redshift zform,galaxies with ∆ ln (M/LB) > 0.2 were neglectedand the two field galaxy samples combined.An unrestricted linear χ2-fit to 31 low–mass

early-type field galaxies with M < 2 × 1011Mof this study and by van der Wel (2005) gives

∆ ln(M/LB) = −(1.69± 0.26) z − (0.08± 0.22).(7.22)

This would correspond to a formation redshift ofzform = 1.7± 0.9 with an observed scatter in the∆ ln(M/LB)−z relation of of σ(zform) = 1.4. Forthe higher–mass sample of eleven galaxies withM > 2× 1011M a linear χ2-fit yields

∆ ln(M/LB) = −(1.46± 0.59) z − (0.03± 0.51).(7.23)

For the massive early-type field galaxies, theevolution of the M/L predicts a mean starformation epoch of zform = 3.2 ± 1.0. The 1σobserved scatter of the ∆ ln(M/LB)− z relationfor the massive field galaxies is σ(zform) = 1.9.In comparison to the lower-mass counterparts,a higher formation redshift is significant on the2σ level based on the uncertainty of the high-mass galaxies. The lack of field galaxies withmasses of M > 2 × 1011M introduces a largeruncertainty on the zform value. By comparingthe derived formation epoch of the high–masscluster galaxies with numbers for cluster gal-axies in the literature a good agreement isfound. Based on a compilation of clusters outto redshift of z = 1.27, van Dokkum & Stanford(2003) deduce zform = 2.6+0.9

−0.4. This study givesan upper limit for the formation redshift ofmassive cluster galaxies of zform = 3.5 (1σ). Theresult obtained for the massive field galaxieshere is within this upper limit, which suggeststhat these field galaxies have old passivelyevolving stellar populations similar to theirmassive counterparts in clusters. Besides thescatter of the individual galaxies, lower-massfield galaxies have a stronger evolution in theirM/L ratio. The mean stellar populations agesof the less-massive field galaxies are younger,with lower formation redshift of 1 < zform ≤ 2.This suggests a mass–dependent luminosity evo-lution with a stronger evolution for lower–massgalaxies with M < 2× 1011M.


Figure 7.16: M/LB ratio as a function of mass M in solar units as derived from the FP for the early-type field galaxies in the FDF and WHDF, divided into two different redshift bins. Filled symbols denoteellipticals, open symbols S0 galaxies and Sa bulges. In brackets the respective number of a morphologicalclass is listed. Errors for the distant sample are shown in both panels. The nearby 7S Coma cluster sampleis indicated with small dots (dashed line is a χ2-fit to Coma) and the high-redshift field E+S0 galaxies atz ∼ 1 by van der Wel et al. (2005) are shown as triangles (solid line is a χ2-fit). All data points have beencorrected for maximum progenitor bias ∆ ln(M/LB) ≈ 0.2 z. Left: M/LB −M for the whole sample of 21field E+S0 galaxies 0 < z < 0.75. Right: M/LB−M for 15 higher-redshift field E+S0s with 0.26 < z < 0.75.Ellipticals follow the local relation whereas the S0 galaxies have on average lower M/L ratios, a larger scatterand match the intermediate–mass range of the M/LB −M relation of the z ∼ 1 field E+S0 galaxies.

To test this mass–dependent evolution even fur-ther, in Fig. 7.16 the M/LB ratio of the early-type field galaxies is shown as a function ofmass M . The galaxies were divided accordingto their redshift into two different bins. The leftpanel shows the M/LB −M for all field galax-ies, whereas in the right panel the field objectsare restricted to a sub-sample of 15 high-redshiftgalaxies. As a comparison, the nearby 7S Comacluster sample (small dots) and the high-redshiftfield E+S0 galaxies at z ∼ 1 of van der Welet al. (2005) (triangles) are indicated togetherwith linear χ2-fits to both samples. All datapoints have been corrected for maximum pro-genitor bias ∆ ln(M/LB) ≈ 0.2 z. Ellipticals, inparticular below z <∼ 0.26, follow the local re-lation whereas the S0 galaxies have on averagelower M/L ratios and match the intermediate–

mass range of the z ∼ 1 field E+S0 galaxies.Lenticular galaxies show a larger scatter and thetwo outliers in the FP (Sa bulges # 6336 and# 508) are clearly offset from the rest of the sam-ple. The difference between field elliptical andS0 galaxies found in the FP can be explained bya difference in their formation epochs. Ellipti-cals, especially massive ones, were generated atearlier times and follow rather moderate quies-cent evolutionary tracks. On the contrary, lentic-ular and Sa bulges show more diverse stellarpopulations which are younger and have beenformed at recent times with formation redshifts1 < zform ≤ 2.


7.3 Further Discussion andConclusions

The analysis of the distant scaling relations ofthe FJR, KR and FP for distant early–type gal-axies at z ∼ 0.2 is in several ways consistentwith previous studies. Early–type galaxies inall environments obey tight relationships similarto the local Universe. Assuming the local slopeholds valid with increasing redshift, on averagea modest evolution in the Gunn r and John-son B-band luminosities of the galaxies is de-rived. There is no significant increase of scatterfor the distant FJR. However, elliptical galax-ies in poor clusters display a slightly larger scat-ter in the FP. In contrast to the local counter-parts, the distant FJR have a shallower slope atthe 2σ significance level, which implies a mass–dependent luminosity evolution. Lower-mass gal-axies in all cluster environments are on averagebrighter by∼ 0.6 mag in rest–frame Gunn r-bandat z ≈ 0.2 than locally, whereas the distantearly-types with masses of M ≈ 3 × 1011Mdisplay only a mild evolution and have similarabsolute magnitudes as their present–day coun-terparts.The sub–classes of early–type galaxies, ellipti-cal and S0 galaxies, appear not to represent ahomogenous group but follow different evolu-tionary tracks. Lenticular galaxies in high–massclusters evolve faster than ellipticals, whereas el-liptical galaxies in low-mass clusters evolve asfast as S0 galaxies in rich clusters.Assuming a typical galaxy in our sample witha size of Re = 0.34 kpc (which correspondsto 0.6′′) and a measured velocity dispersionof σ = 200 km s−1 gives a galaxy mass ofM ≈ 1.0 × 1011M. This represents a typi-cal mass of an early-type galaxy within themass range covered by the cluster samples atz ≈ 0.2. For galaxy masses lower than thislimit, the M/L evolution is faster than for gal-axies with masses of M > 2 × 1011M. Thistrend is clearly visible for all environments, from

the highest mass densities in rich clusters overlower-mass densities in poor clusters down tothe field environment. For the less–massive fieldgalaxies a faster evolution of the M/L ratiosof ∆ ln(M/LB) = −(1.69 ± 0.26) z is found,whereas the more–massive ones evolve sloweras ∆ ln(M/LB) = −(1.46 ± 0.59) z. This evo-lution in M/L ratios can be translated in toa formation redshift for the lower–mass galax-ies (M < 2 × 1011M) of zform = 1.7 ± 0.9,whereas for massive galaxies (M > 2×1011M)zform = 3.2 ± 1.0 is derived. The FDF andWHDF field early–type galaxies are consistentwith the mass dependence of the M/L ratios asfound in a recent study of E and S0 field galax-ies up to redshifts of z = 1 (van der Wel 2005).These authors derive from a comparison withlow–redshift data for the whole field galaxy sam-ple ∆ ln(M/LB) = −(1.75± 0.16) z whereas gal-axies with masses M > 2×1011M show a M/Levolution of ∆ ln(M/LB) = −(1.20± 0.17) z.

These results suggests that the evolution corre-lates with the galaxy mass rather than the lumi-nosity of the galaxies. The lack of age differencefound for field and cluster early-type galaxiesand the dependence of the evolution on galaxymass suggests that environmental effects play arather minor role in the formation of early–typegalaxies. More important to the evolutionaryhistory of a galaxy are its internal properties,such as mass, size, velocity dispersion or chemi-cal composition.

The evolution of the M/L ratio in all environ-ments favours a down–sizing formation scenario(Kodama et al. 2004). In this picture, massivegalaxies are predominantly dominated by redold stellar populations which evolve on longertimescales and passively. Less–massive systems,however, exhibit more extended star formationhistories. In the down–sizing approach, themass of galaxies hosting star formation processesdecreases with increasing age of the Universe.Both, the mass assembly on the one handand the star formation on the other hand are


accelerated in massive stellar systems located inthe high density environments. In contrast, inless–massive (smaller) systems, these physicalprocesses work on longer timescales. Supportfor this formation picture is endorsed by severalother observational studies, such as the lackof star formation in massive galaxies at lowerredshifts z ≤ 1 (De Lucia et al. 2004), a fossilrecord of star formation in early-type galaxiesdetected in the local Universe (Thomas et al.2005) or the evidence for a mass-dependentevolution in the Tully–Fisher scaling relation ofspiral galaxies up to z = 1 (Ziegler et al. 2002;Bohm et al. 2004).

It is difficult to anticipate on specific improve-ments of our knowledge of galaxy formation andevolution for the upcoming future. More obser-vational evidence is necessary to strengthen theresults of this thesis and other studies. Withthe continuous rapid increase of computationalpower, a detailed numerical modelling on higher–spatial resolution scales will be possible to puttighter constraints on the involved physical pro-cesses as well as the stellar populations proper-ties and chemical compositions of galaxies.In the near future, the project described in thisthesis is going to be continued and the size of thesample will be enlarged by additional high qual-ity spectra of a further ∼ 50 distant early-typegalaxies. Thanks to the intermediate–resolutionof the observed galaxy spectra, absorption line-strengths can be measured. By comparing ages,metallicities and chemical element abundance ra-tios to the calibrated library of local early-typegalaxies of the Lick/IDS index system, deeper in-sights on the involved physical mechanisms andstellar population properties of early-type galax-ies will be provided. A short outline of this planwill be presented in the outlook at the end of thefollowing summary.

Chapter 8: Summary and Outlook 175

Chapter 8

Summary and Outlook

Using high signal–to–noise intermediate–resolu-tion spectra taken with the ESO Very LargeTelescope (VLT) and Calar Alto (CAHA) 3.5 mtelescope complemented by deep ground-basedand Hubble Space Telescope (HST) imaging,a study of early–type galaxies at redshifts0.2 ≤ z ≤ 0.4 has been performed. An exten-sive sample of 121 early–type galaxies in vari-ous densities ranging from galaxy clusters withhighest richness class, rich clusters, over poorrichness class, poor Low–LX clusters, down tothe isolated field population was constructed toinvestigate the evolution over the last ∼ 3 Gyrs,corresponding to 20% of the age of the Universe.The motivation of the project was to look howthe physical properties of early–type galaxies de-pend on their environment. The evolution of gal-axies in luminosity, size, mass and their stellarpopulations were intercompared to test possibleenvironmental effects and to constrain the the-oretical formation and evolution model predic-tions.

Target objects were selected based on multi–band ground-based imaging data. For the early–type galaxies in the rich cluster Abell 2390, UBIbroad-band colours were used, whereas early–type candidates for three poor clusters were se-lected by a combination of a spectroscopic red-shift catalog and ground-based BV RI photo-metry. These Low–LX clusters display 1.0 dexlower X-ray luminosities than typical rich clus-ters and are, as a difference in LX correspondsdirectly to a difference in mass via the correlation

LX ∝ M2, therefore considered to be low–massgalaxy clusters. Early–type field galaxies wereselected over a redshift range of 0.2 < z < 0.75from the FORS Deep Field (FDF) and WilliamHerschel Deep Field (WHDF), two multi–bandimaging studies located in the southern andnorthern hemisphere respectively, with opticallimiting magnitudes comparable to the HubbleDeep Fields and each covering a field-of-view of∼ 7′ × 7′.Spectroscopic candidates were selected accord-ing to their elongated structureless appearance,their luminosity, photometric redshift and an ad-ditional constraint of spectrophotometric typefor the field targets. Main criterion was the to-tal apparent I or R-band brightness with a faintlimit of I ≤ 20 mag for the cluster galaxies andR ≤ 22 mag in case of the early–type field galax-ies. To study the evolution of early–type galaxiesout to large clustercentric distances of ∼1 virialradii, the spectroscopic targets of the cluster en-vironments were distributed over a wide field-of-view of ∼10′× 10′, which corresponds to a phys-ical field-of-view of ∼1.4 Abell radii (∼2.7 Mpc)at z = 0.23.In total, 142 high signal–to–noise spectra of121 different early–type galaxies and additional28 stellar template spectra were taken betweenSept. 1999 to Oct. 2002 in Multi Object Spec-troscopy mode with the MOSCA spectrographat the CAHA 3.5 m telescope and the FORS 1+2instruments mounted on the VLT. The seeingconditions ranged between 1.2 to 1.7 arcseconds

176 Chapter 8: Summary and Outlook

for the CAHA observations and 0.4 to 1.0 arc-seconds for the VLT Paranal observations.

Spectroscopic redshifts could be derived for 121early-type galaxies and 9 spiral galaxies and 12secondary objects. No galaxy spectra of the tar-gets had too low signal–to–noise to determine aredshift.

The multi–band photometry data provided thepossibility to measure accurate luminosities ofthe galaxies. Apparent brightnesses for the clus-ter galaxies were transformed to the rest–frameGunn r-band. For the field galaxies in the FDFand WHDF apparent magnitudes were derivedin either of the B, g, R or I passband which,depending on the redshift of the galaxy, bestmatched the rest–frame B-band. k-correctionswere computed via synthetic photometry with anobserved spectral template and a spectral temp-late from chemically consistent evolutionary syn-thesis models. Both independent approachesyielded within their uncertainties very similar re-sults. For the cluster galaxies the uncertaintiesin the k-corrections were δkr ≤ 0.05 mag and incase of the field objects δkB ≤ 0.1 mag over thecomplete redshift range.

Structural parameters for 1/4 of all early–typegalaxies were analysed on the HST/WFPC2(Wide Field and Planetary Camera 2) andHST/ACS (Advanced Camera for Surveys) im-ages using the surface brightness profile fittingalgorithm developed by Saglia et al. (1997a),which searchs for the best combination of seeing-convolved, sky-corrected r1/4 and exponentiallaws. The total HST/F702W or F814W-bandmagnitude, effective (half-light) radius, centralsurface brightness and the mean surface bright-ness within the effective radius, were derived forthe entire galaxy as well as for its bulge and disccomponent separately. This approach accountsfor various types of observed surface brightnessprofiles, such as elliptical galaxies with a flatcore, discy S0 galaxies or the extended profilesof central cD galaxies. The two-dimensionalsurface brightness distribution of the early–type

field galaxies was fitted with a combination ofr1/n plus exponential disc profiles using theGALFIT package. Both algorithms gave withintheir errors consistent results. The complete setof structural parameters was subject to a robustand careful error treatment.

Elliptical and lenticular (S0) galaxies weremorphologically classified in three inde-pendent methods using the high–resolutionHST/WFPC2 and HST/ACS images. A visualinspection was executed for a sub–set of ellip-tical and S0 galaxies. Results from the surfacebrightness profile fitting provided a secondverification and a third quantitative analysiswas conducted using the bulge–to–total (B/T )fractions of the galaxies. A weak correlationof the measured bulge fraction with visualmorphological Hubble type was found.

Radial velocities and velocity dispersions (σ) ofthe early–type galaxies were computed from theG4300-band, Hβ or Mgb absorption lines usingthe Fourier Correlation Quotient (FCQ) method(Bender 1990). To derive the kinematic proper-ties of the distant galaxies several modificationson the parameter settings were implemented,such as the subtraction of the continuum levelin the spectra or the instrumental resolution. Toensure a reliable error treatment for the veloc-ity dispersions and radial velocity measurements,Monte Carlo simulations for a set of differentstellar templates over a range of input veloc-ity dispersions and S/N ratios were performed.Synthetic galaxy spectra with artificial noise andfor a specific instrumental setup and resolutionwere generated and the velocity dispersions re-covered by the FCQ software.

Velocity dispersions based on high signal–to–noise (〈S/N〉 >∼ 30 per A) spectra could be de-rived for 110 early–type galaxies regardless oftheir environment. For six galaxies with verylow signal (S/N <∼ 8 per A), no σ measure-ment was possible. A visual inspection of the σvalues was conducted for each galaxy spectrumto assess the overall quality and reliability of

Chapter 8: Summary and Outlook 177

the measurement. Extensive consistency checkswere performed by comparing σ determinationsfrom different absorption lines, repeat observa-tions, different extraction procedures and stellarlibraries.

The total sample analysed in this study com-prises 147 early–type galaxies. 48 early–typegalaxies are members of the rich cluster A 2390(z = 0.23), 27 galaxies are associated to one ofthe three Low–LX cluster CL 0849, CL 1701 andCL 1702 (〈z〉 = 0.23), and 24 are field galaxiesin the FDF and WHDF (〈z〉 = 0.4). In addi-tion, 48 early–type galaxies of a previous studyof the rich A 2218 cluster were combined withthe A 2390 data to study a total of 96 early–typegalaxies in the densest environments.

In the further analysis, the scaling relations ofthe Faber–Jackson (FJR) and Kormendy rela-tion (KR) as well as the Fundamental Plane(FP) for distant early–type galaxies at z ∼ 0.2obey tight relationships similar to the local Uni-verse. Assuming the local slope holds valid withincreasing redshift, on average a modest evo-lution in the Gunn r and Johnson B-band lu-minosities of the galaxies is derived, regardlessof their environment locus. There is no signif-icant increase of scatter for the distant FJR,whereas elliptical galaxies in poor clusters indi-cate a slightly larger scatter in the FP. In con-trast to the local counterparts, the distant FJRhave shallower slopes at the 2σ significance level,which implies a mass–dependent luminosity evo-lution. Lower–mass galaxies in all cluster envi-ronments are on average brighter by ∼ 0.6 mag inrest–frame Gunn r-band at z ≈ 0.2 than locally,whereas the distant early-types with masses ofM ≈ 3× 1011M display only a mild evolutionand have similar absolute magnitudes as theirpresent–day counterparts. Evidence for this re-sult is even found for the early–type field galax-ies. Their mass–to–light ratios (M/L) are con-sistent with the trend detected within a recentstudy of E and S0 field galaxies in the literature.

For early–type cluster galaxies, a slight radial de-

pendence is found. Galaxies in the outer districtsof clusters at ∼0.25 virial radii (∼0.46 Mpc)show a stronger brightening by ∼0.2 mag thantheir counterparts in the central cluster regions.

A difference between the morphological sub-classes of early–type galaxies, elliptical and S0galaxies, was revealed. Lenticular galaxies inrich clusters exhibit a stronger luminosity evo-lution compared to the local Coma FP than theellipticals. The stellar populations of these sys-tems could be more diverse or comprise morecomplex star formation histories. Support ofthis assumption is accumulated by an analysis ofthe rest–frame galaxy colours. The bluer galaxypopulation ((B − I) < 2.287) shows larger FJRresiduals from the local FJR, whereas the reddergalaxies of the red–sequence of early–type clus-ter galaxies are less offset. Interestingly, the S0galaxies in both colour ranges show a stronger lu-minosity evolution (brighter by ∼0.7 mag for thebluer and ∼0.45 mag for the redder group), butthe blue ellipticals have on average only a mildbrightening of ∼0.3 mag and the redder even a“positive” evolution, i.e. they are fainter thanthe local reference.

The lack of an age difference found for field andcluster early-type galaxies and the dependence ofthe evolution on galaxy mass suggests that en-vironmental effects are not the dominant factorswhich drive the formation of early–type galax-ies. Internal properties of the galaxies, such asmass, size, velocity dispersion or chemical com-position, are the main contributors to the evolu-tionary history of elliptical and lenticular galax-ies.

The evolution of the M/L ratio in all environ-ments favours a down–sizing formation scenario(Kodama et al. 2004). Massive galaxies arepredominantly dominated by red old passivelyevolving stellar populations, whereas less mas-sive systems have more extended star formationhistories. The mass of galaxies hosting star for-mation processes thereby decreases with the ageof the Universe. Both mass assembly and star

178 Chapter 8: Summary and Outlook

formation are accelerated in massive stellar sys-tems located in the high density environments,whereas the processes work on longer timescalesin less–massive (smaller) systems.

The results of this thesis therefore strengthen theneed for more realistic implementations of stellarpopulation properties into the numerical codesof N -body simulations. In particular, the fun-damental physics of the star formation processesand their involved timescales still are not wellunderstood. Progress in the near future seemspromising, looking at the most recent analyticalapproaches and their improvements. Recently adetailed simulation of the formation of individualearly–type galaxies in a fully cosmological con-text was achieved (Meza et al. 2003). Neverthe-less, crucial mechanisms such as star formationprocesses were still treated based on simplifiedassumptions.

It is hard to predict whether a more realisticmodelling of the stellar content of the galaxieswill lead to a consistency between theory andobservations. Some modifications on shorterscales of the hierarchical scenario would prob-ably be needed to reach a better consensus.Further research in the upcoming years is goingto be very important for our understandingof the formation and evolution of early–typegalaxies to achieve a complete picture of thefundamental processes of structure growth fromthe early universe up to the present.

The project on distant early–type galaxies de-scribed in this thesis will be continued in thenext years by the author. Spectroscopy of theintegrated stellar light of a galaxy is a powerfultool for deriving the age, metallicity and grav-ity or temperature of their stellar population.To quantify these intrinsic properties, spectralindices have been introduced which characterisethe strength of a specific absorption or emissionline. Thanks to the intermediate–resolution ofthe observed early–type galaxy spectra, absorp-

tion line-strengths can be measured and com-pared to the Lick/IDS index system. Light-averaged ages, metallicities and chemical ele-ment abundance ratios can be derived and di-rectly confronted to the predictions of stellarpopulation models. In combination with the lu-minosity evolution measured via the scaling re-lations, results will give strong constraints onthe formation redshift and the involved physicalmechanisms of early–type galaxies.To extent the sample even more, observations of∼50 early–type galaxies in three Low–LX clus-ters at 0.2 < z < 3.0 have been acquired usingMOSCA spectra. This data will allow to enlargethe statistical significance of the results for thepoor clusters presented in this thesis.Furthermore, a joint research project togetherwith the MPE in Munich to study the colourgradients in cluster galaxies for different environ-ments was already initiated by the author. A de-tailed analysis of HST/WFPC2 and HST/ACSstructural parameters in different filter pass-bands and with independent algorithms for mod-elling the galaxy surface brightness profiles willreveal how the colour gradients and the shapeof galaxies depend on the computed structuralparameters.A combination of all spectroscopic data and theHST images will make up one of the largestkinematic samples of early–type galaxies at red-shift z = 0.2. Together with the analysis of thestellar populations and chemical abundances ofthese galaxies deeper observational insights willbe achieved into how early–type galaxies, in par-ticular S0 galaxies, form and evolve with red-shift.

Appendix A: Stellar Templates 179

Appendix A

Stellar Templates

During the observing runs to acquire inter-mediate–resolution MOSCA spectra for theearly–type galaxies in the rich and poor clus-ters, a total of 28 stellar spectra of 15 differentLick/IDS index standard stars with the same in-strument configuration have been observed. Thestars were taken from the Lick/IDS star librarylist by Worthey et al. (1994). Table A.1 showsthe observations for all 28 Lick/IDS index stan-dard stars from CAHA splitted into differentobserving runs. In total, 12 Lick/IDS stan-dards were observed in September 1999, 7 in July2000, 6 in April 2001, and finally 3 in February2002. The reduction of these stellar templatesis discussed in section 3.7 on page 44. A setof nine kinematic templates which exhibit thehighest instrumental resolution were utilised forthe derivation of the galaxies’ velocity disper-sions. To test possible systematic effects withinthe measurements and/or systematic offsets aris-ing from different observing runs, a comparisonbetween repeat observations of Lick/IDS starsacquired at CAHA was performed.

Absorption line-strengths for 15 different kine-matic template stars were measured and cal-ibrated to the Lick/IDS system following themethod as described in Faber et al. (1985).In order to test whether possible offsets existbetween the measurements of Lick/IDS absorp-tion line indices for different observing runs, theresults of the line-strengths for 10 stars whichare included in both runs of Sept. 1999 andJuly 2000 are compared to each other. Fig-

Figure A.1: Relative differences of the Lick/IDS ab-sorption line indices of Mgb, Hβ, Fe5270 and Fe5335for 10 kinematic stellar templates with spectral typesfrom A0V to M6III of the CAHA observing runs 1999and 2000. See also Table A.2.

ure A.1 shows a comparison of the relative differ-ences between the Lick/IDS line indices of Mgb,Hβ, Fe5270 and Fe5335 for 10 stars with dif-ferent spectral types (A0V to M6III). The rela-tive difference of line-strength indices from 1999and 2000 ∆(Index1999−Index2000)/Index2000 areplotted as a function of the derived line-strengthindex of the run in 2000 Index2000. A typi-cal Lick/IDS error of 0.2 A was adopted. Theagreement between all templates is good, onlynegligible deviations are detected. The mean

180 Appendix A: Stellar Templates

and median (in brackets) values of the relativedifferences are ∆(Mgb)/Mgb 2000 = 0.06 ± 0.18(0.02), ∆(Hβ)/Hβ2000 = −0.06 ± 0.33 (−0.07),∆(Fe5270)/Fe52702000 = 0.10 ± 0.21 (0.03) and∆(Fe5335)/Fe53352000 = −0.08 ± 0.13 (−0.07).There is no systematic offset neither a changewith spectral type found. Table A.2 gives a sum-mary of the derived relative differences for theLick/IDS indices of Figure A.1. For the starSAO 004401 (A0V) no Fe5270 and Fe5335 in-dices could be measured for the 2000 run andtherefore this star is not included in the compar-ison of the Fe-index diagrams. For purposes ofcomparison, the published Lick/IDS index mea-surements by Worthey et al. (1994) are listedin the Tables A.3, A.4 and A.5. In general, theagreement between all indices is good and onlynegligible deviations are found. In addition, thistests demonstrates the accuracy and reliabilityof the data reduction and the wavelength cal-ibration for the stellar template spectra as nosystematic errors originating from possible er-rors in flat-fielding, sky background subtractionor wavelength calibration were detected.


Table A.1: Observed stellar templates at Calar Alto. The exposure times were Ttot=60 sec.

Name ID spec type run slit width vrad σinst FWHMHβ & Mg G-band

[arcsec] [km s−1] [A] [km s−1] [A] [km s−1]

SAO 004401 048 A0V 09/99 1.5 +74.5 5.0 126 4.6 139SAO 005056 052 K4V 09/99 1.5 +07.2 5.2 131 4.6 139SAO 038597 100 G0V 09/99 1.5 +27.4 5.2 131 4.7 142SAO 055946 050 K7III 09/99 1.5 +41.8 5.1 129 4.5 136SAO 056928 054 G5 09/99 1.5 −91.1 – – – –SAO 077592 022 G8IIIv 09/99 1.5 +02.8 4.9 124 4.4 133SAO 093189 098 M6III 09/99 1.5 +67.3 5.2 131 4.6 139SAO 110065 011 K3IIIbBa0 09/99 1.5 +02.3 5.2 131 4.5 136SAO 112958 055 K0IIIbCN-2 09/99 1.5 +104.4 5.0 126 4.5 136SAO 123140 116 G8III 09/99 1.5 +43.0 5.3 134 4.6 139SAO 124799 087 K3IIIb 09/99 1.5 +25.1 5.3 134 4.7 142SAO 130649 024 F9V 09/99 1.5 +89.8 4.9 124 4.2 127

mean CAsep99: 5.1 129 4.5 136

SAO 055946 072 K7III 07/00 1.5 −12.1 5.2 131 4.5 136SAO 093189 077 M6III 07/00 1.5 – 5.0 126 4.6 140SAO 110065 070 K3IIIbBa0 07/00 1.5 −28.5 5.3 134 4.6 140SAO 112958 118 K0IIIbCN-2 07/00 1.5 +62.3 5.1 129 4.7 143SAO 123140 102 G8III 07/00 1.5 −85.6 5.2 131 4.4 133SAO 124799 141 K3IIIb 07/00 1.5 −28.8 5.3 134 4.7 143SAO 130649 164 F9V 07/00 1.5 +84.3 5.3 134 4.7 143

mean CAjul00: 5.2 131 4.6 140

SAO 032042 024 K3III 04/01 1.0 −06.2 3.5 88 3.0 91SAO 032042 019 K3III 04/01 1.0 −06.5 3.6 91 3.1 94SAO 098087 006 K0III 04/01 1.0 +63.7 3.4 86 3.2 97SAO 123140 032 G8III 04/01 1.0 +07.5 3.5 88 3.3 100SAO 123140 020 G8III 04/01 1.0 – 3.4 86 3.2 97SAO 124799 022 K3IIIb 04/01 1.0 −22.6 3.5 88 3.1 94

mean CAapr01: 3.5 88 3.2 97

SAO 032042 116 K3III 02/02 0.5 +32.4 2.2 55 2.3 70SAO 080333 084 K0III 02/02 0.5 +35.3 2.2 55 2.5 76SAO 098087 157 K0III 02/02 0.5 −24.1 2.5 64 2.2 67

mean CAfeb02: 2.3 58 2.3 71


Table A.2: Relative differences of Lick/IDS indices for the stellar templates of the observations in 1999 and2000 as show in Fig. A.1.

Name ID spec type Mgb ∆σ (Mgb) Hβ ∆σ (Hβ) Fe5270 ∆σ (Fe5270) Fe5335 ∆σ (Fe5335)[A] [A] [A] [A] [A] [A] [A] [A]

SAO 004401 048 A0V 0.23 0.50 7.36 0.09 – – – –SAO 005056 052 K4V 6.97 -0.15 -0.55 -0.81 2.72 0.58 4.30 -0.12SAO 055946 050 K7III 4.27 -0.01 0.74 -0.07 3.59 0.03 3.61 0.04SAO 077592 022 G8IIIv 2.65 0.01 2.20 -0.19 2.97 -0.18 3.10 -0.12SAO 093189 098 M6III 10.79 0.18 3.35 0.45 2.66 0.02 1.33 -0.12SAO 110065 011 K3IIIbBa0 3.39 0.05 0.91 -0.07 3.47 0.03 3.17 0.00SAO 112958 055 K0IIIbCN-2 2.83 -0.00 0.90 0.30 1.95 0.14 1.70 0.06SAO 123140 116 G8III 2.33 -0.09 1.74 -0.01 2.40 0.17 3.43 -0.36SAO 124799 087 K3IIIb 4.42 0.02 1.10 -0.10 3.53 0.02 3.15 -0.01SAO 130649 024 F9V 1.62 0.13 3.15 -0.18 0.81 0.14 0.78 -0.07

Table A.3: Lick/IDS standards stars for the indices CN1, CN2, Ca4227, G4300, Fe4383, Ca4455 and Fe4531(Worthey et al. 1994).

SAO R.A. DEC. V Spec Type CN1 CN2 Ca4227 G4300 Fe4383 Ca4455 Fe4531[mag] [mag] [mag] [A] [A] [A] [A] [A]

110065 01 38 49.559 +05 14 07.13 4.44 K3 IIIbBa01 0.208 0.260 2.660 6.590 8.060 2.530 4.590055946 02 48 25.535 +34 51 19.20 4.53 K7 III 0.114 0.192 4.380 6.020 8.815 3.140 5.885005056 03 54 48.885 +76 01 33.96 9.99 K 4 V -0.056 -0.008 4.595 5.280 8.775 3.115 5.190093189 02 52 59.608 +18 07 48.82 5.91 M6 III -0.134 -0.108 6.030 3.590 -0.105 4.175 4.065004401 01 22 03.732 +70 43 13.05 6.49 A0 Vnn -0.231 -0.142 -0.090 -2.583 -0.998 -0.201 0.337123140 18 04 54.790 +08 43 33.79 4.64 G8 III 0.170 0.197 0.510 6.430 5.470 1.690 3.530112958 05 34 09.415 +09 15 54.81 4.09 K0 IIIbCN2 0.027 0.037 0.720 6.690 4.770 1.455 3.150077592 05 45 56.744 +24 33 09.24 4.86 G8 IIIv 0.204 0.243 0.840 5.760 6.510 1.920 3.750124799 19 31 38.752 +07 16 16.89 4.45 K3 IIIb 0.244 0.302 1.890 6.550 8.010 2.300 4.260130649 03 37 49.156 −03 22 29.10 6.68 F9 V -0.069 -0.044 0.477 3.270 1.827 0.417 1.387056928 03 59 53.188 +35 09 17.28 8.50 G5 ZZZ -0.046 -0.004 1.830 5.570 4.260 0.750 1.810038597 03 05 26.709 +49 25 26.63 4.05 G0 V -0.060 -0.042 0.625 4.265 2.950 1.050 2.400110065 01 38 49.559 +05 14 07.13 4.44 K3 IIIbBa01 0.208 0.260 2.660 6.590 8.060 2.530 4.590055946 02 48 25.535 +34 51 19.20 4.53 K7 III 0.114 0.192 4.380 6.020 8.815 3.140 5.885004401 01 22 03.732 +70 43 13.05 6.49 A0 Vnn -0.231 -0.142 -0.090 -2.583 -0.998 -0.201 0.337112958 05 34 09.415 +09 15 54.81 4.09 K0 IIIbCN2 0.027 0.037 0.720 6.690 4.770 1.455 3.150077592 05 45 56.744 +24 33 09.24 4.86 G8 IIIv 0.204 0.243 0.840 5.760 6.510 1.920 3.750124799 19 31 38.752 +07 16 16.89 4.45 K3 IIIb 0.244 0.302 1.890 6.550 8.010 2.300 4.260130649 03 37 49.156 −03 22 29.10 6.68 F9 V -0.069 -0.044 0.477 3.270 1.827 0.417 1.387080333 08 37 13.995 +20 11 08.03 6.39 K0 III 0.256 0.283 0.645 6.700 6.112 1.772 2.840098087 08 41 50.747 +18 20 21.83 3.94 K0 III-IIIb 0.209 0.230 1.046 7.020 6.806 1.948 3.218032042 19 49 22.412 +52 51 37.60 5.03 K3 III-CN1 0.387 0.456 2.300 6.700 9.110 2.640 5.090


Table A.4: Same as A.3, but for the Lick/IDS indices Fe4668, Hβ, Fe5015, Mg1, Mg2, Mgb, Fe5270, Fe5335,Fe5406, Fe5709, Fe5782 and Na5895 (Worthey et al. 1994).

SAO Fe4668 Hβ Fe5015 Mg1 Mg2 Mgb Fe5270 Fe5335 Fe5406 Fe5709 Fe5782 Na5895[A] [A] [A] [mag] [mag] [A] [A] [A] [A] [A] [A] [A]

110065 5.880 0.870 6.510 0.172 0.315 3.730 3.880 3.530 2.590 1.450 1.310 3.820055946 7.655 0.600 7.205 0.237 0.436 4.050 4.140 4.110 3.350 1.290 1.370 4.290005056 2.145 -0.009 5.365 0.296 0.504 6.440 4.770 4.260 2.565 0.830 0.975 7.160093189 24.705 5.375 28.775 -0.108 0.408 15.235 4.740 1.630 0.680 -0.750 -1.360 4.970004401 -0.786 8.118 0.179 -0.008 0.008 0.258 0.190 0.181 0.123 -0.016 -0.031 0.816123140 5.820 1.810 6.260 0.037 0.139 2.310 3.030 2.410 1.620 1.250 0.760 2.060112958 3.505 1.000 4.140 0.072 0.174 2.810 2.300 1.740 1.135 0.955 0.470 1.430077592 7.500 1.860 6.520 0.089 0.216 3.030 3.380 2.920 1.990 1.280 1.030 2.820124799 8.720 1.020 6.440 0.184 0.346 4.800 3.810 3.500 2.480 1.400 1.180 3.610130649 -0.133 2.100 2.717 0.002 0.082 1.663 1.128 0.988 0.353 0.207 0.143 1.118056928 -0.245 0.550 1.900 0.080 0.247 4.625 2.200 1.685 0.895 0.325 -0.155 2.255038597 2.650 3.125 4.545 0.017 0.115 2.165 1.605 1.345 0.850 0.480 0.220 1.370110065 5.880 0.870 6.510 0.172 0.315 3.730 3.880 3.530 2.590 1.450 1.310 3.820055946 7.655 0.600 7.205 0.237 0.436 4.050 4.140 4.110 3.350 1.290 1.370 4.290004401 -0.786 8.118 0.179 -0.008 0.008 0.258 0.190 0.181 0.123 -0.016 -0.031 0.816112958 3.505 1.000 4.140 0.072 0.174 2.810 2.300 1.740 1.135 0.955 0.470 1.430077592 7.500 1.860 6.520 0.089 0.216 3.030 3.380 2.920 1.990 1.280 1.030 2.820124799 8.720 1.020 6.440 0.184 0.346 4.800 3.810 3.500 2.480 1.400 1.180 3.610130649 -0.133 2.100 2.717 0.002 0.082 1.663 1.128 0.988 0.353 0.207 0.143 1.118080333 6.584 1.770 6.650 0.041 0.168 2.550 3.190 2.730 1.962 1.014 1.138 2.620098087 7.220 1.330 5.986 0.088 0.211 3.160 3.380 2.610 1.834 1.148 1.008 2.180032042 11.520 1.170 7.860 0.215 0.382 4.620 4.350 4.100 2.880 1.540 1.560 5.500

Table A.5: Same as A.3, but for the Lick/IDS indices TiO1, TiO2, HδA, HγA, HδF and HγF . Values foreffective temperature Teff , gravity log(g), the metallicity [Fe/H] and a note if the star is used as a LickStandard (LS) are also displayed (Worthey et al. 1994).

SAO TiO1 TiO2 HδA HγA HδF HγF Teff log(g) [Fe/H] com.

[mag] [mag] [A] [A] [A] [A] K [cm/s2]

110065 0.027 0.055 -6.153 -10.051 -1.558 -2.930 4133 1.20 -0.11 LS055946 0.104 0.250 -5.844 -10.124 -1.325 -2.605 3905 1.10 -0.18005056 0.014 0.029 -5.090 -11.303 -0.775 -3.700 4357 4.61 9.99093189 0.480 0.883 -1.677 -5.633 -1.300 -1.860 3250 0.30 9.99004401 0.010 -0.003 10.678 10.866 7.790 8.130 8455 4.10 9.99123140 0.013 0.011 -4.688 -7.274 -1.040 -1.893 4957 2.00 -0.05 LS112958 0.020 0.005 -3.059 -6.807 -0.490 -2.190 4751 2.90 -0.55077592 0.020 0.016 -4.096 -7.298 -0.518 -1.883 4751 2.20 0.03 LS124799 0.018 0.032 -7.114 -9.705 -1.568 -3.290 4428 2.45 0.22 LS130649 0.005 0.002 1.642 -0.629 1.630 0.980 5780 4.27 -0.85056928 0.012 -0.009 -2.168 -6.597 0.455 -2.460 4862 4.77 -1.61038597 0.015 -0.002 0.760 -1.872 1.540 0.750 5984 4.10 0.09110065 0.027 0.055 -6.153 -10.051 -1.558 -2.930 4133 1.20 -0.11 LS055946 0.104 0.250 -5.844 -10.124 -1.325 -2.605 3905 1.10 -0.18004401 0.010 -0.003 10.678 10.866 7.790 8.130 8455 4.10 9.99112958 0.020 0.005 -3.059 -6.807 -0.490 -2.190 4751 2.90 -0.55077592 0.020 0.016 -4.096 -7.298 -0.518 -1.883 4751 2.20 0.03 LS124799 0.018 0.032 -7.114 -9.705 -1.568 -3.290 4428 2.45 0.22 LS130649 0.005 0.002 1.642 -0.629 1.630 0.980 5780 4.27 -0.85080333 0.009 0.007 -4.672 -8.752 -1.250 -2.380 4965 2.35 0.16098087 0.014 0.019 -5.643 -9.092 -1.380 -3.070 4651 2.30 -0.04032042 0.028 0.052 -8.338 -11.455 -2.044 -3.790 4355 2.15 0.42 LS

Bibliography 185

Bibliography

Abraham, R. G., Smecker-Hane, T. A., Hutch-ings, J. B., et al. 1996a, ApJ, 471, 694

Abraham, R. G.; van den Bergh, S., Glazebrook,K., et al. 1996b, ApJS, 107, 1

Allen, C. W. 1973, “Astrophysical Quantities”,3rd Edition, London

Andredakis, Y. C., Peletier, R. F., Balcells, M.1995, MNRAS, 275, 874

Appenzeller, I., Fricke, K. J., Furtig, W., et al.1998, The Messenger, 94, 1

Appenzeller, I., Bender, R., Bohm, A., et al.,2000, The Messenger, 100, 44

Arimoto, A., & Yoshii, Y. 1987, A&A, 173, 23

Babul, A., Rees, M. J. 1992, MNRAS, 255, 346

Balogh, M. L., Smail, I., Bower, R. G., Ziegler,B. L., Smith, G. P., Davies, R. L., Gaztelu, A.,Kneib, J.-P., & Ebeling, H. 2002a, ApJ, 566, 123

Balogh M., Bower, R. G., Smail, I., Ziegler,B. L., Davies, R. L., Gaztelu, A., & Fritz, A.2002b, MNRAS, 337, 256

Barger, A. J., Aragon-Salamanca, A., Ellis,R. S., Couch, W. J., Smail, I., & Sharples, R. M.1996, MNRAS, 279, 1

Barnes, J. E., & Hernquist, L. 1992, ARA&A 30,705

Baugh, C. M., Cole, S., & Frenk, C. S. 1996,MNRAS, 283, 1361

Bell, E. F., Wolf, C., Meisenheimer, K., et al.2004, ApJ, 608, 752

Bender, R., & Mollenhoff, C. 1987, A&A, 177,71

Bender, R. 1988, A&A, 193, L7

Bender, R. 1990, A&A, 229, 441

Bender, R., Burstein, D., & Faber, S. M. 1992,ApJ, 399, 462

Bender, R., Burstein, D., & Faber, S. M. 1993,ApJ, 411, 153

Bender, R., Appenzeller, I., Bohm, A., et al.2001, in “Deep Fields”, Cristiani, S., Renzini,A., & Williams, R. E. (eds.), ESO astrophysicssymposia, Springer, p. 96

Bernardi, M., Renzini, A., da Costa, L. N., et al.1998, ApJ, 508, L143

Bernardi, M., Sheth, R. K., Annis, J., et al. 2003,AJ, 125, 1866

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

Blanton, M. R., Dalcanton, J., Eisenstein, D., etal. 2001, AJ, 121, 2358

Bower, R. G., Lucey, J. R., & Ellis, R. S. 1992,MNRAS, 254, 601

Bower, R. G., Terlevich, A., Kodama, T., &Caldwell, N. 1999, in “Star Formation in EarlyType Galaxies”, Carral, P., & Cepa, J. (eds.),ASP Conference Series, Vol. 163, p. 211

Bohm, A. 2003, PhD thesis, Universitat Gottin-gen

Bohm, A., Ziegler, B. L., Saglia, R. P., et al.2004, A&A, 420, 9

186 Bibliography

Bohm, A., & Ziegler, B. L. 2006, ApJL, sub.

Brault, J. W., & White, O. R. 1971, A&A, 13,169

Bruzual, G. A., & Charlot, S. 1993, ApJ, 405,538 (GISSEL96 version, BC96)

Bunker, A. J., Moustakas, L. A., & Davis, M.2000, ApJ, 531, 95

Burstein, D., & Heiles, C. 1982, AJ, 87, 1165(BH)

Butcher, H., & Oemler, A., Jr. 1984, ApJ, 285,426

Caon, N., Capaccioli, M., & D’Onofrio, M. 1993,MNRAS, 265, 1013

Capaccioli, M., Caon, N., & D’Onofrio, M. 1992,MNRAS, 259, 323

Cardelli, J. A., Clayton, G. C., & Mathis, J. S.1989, ApJ, 345, 245

Carlberg, R. G., Yee, H. K. C., Ellingson, E., etal. 1996, ApJ, 462, 32

Carter, D. 1987, ApJ, 312, 514

Cole, S., Lacey, C. G., Baugh, C. M., & Frenk,C. S. 2000, MNRAS, 319, 168

Contardo, G., Steinmetz, M., & Fritze-v. Al-vensleben, U. 1998, ApJ, 507, 497

Davies, R. L., Efstathiou, G., Fall, S. M., Illing-worth, G., & Schechter, P. L. 1983, ApJ, 266,41

Davies, R. L., Burstein, D., Dressler, A., et al.1987 ApJS, 64, 581

de Carvalho, R. R., & Djorgovski, S. 1992, ApJ,389, L49

De Lucia, G., Poggianti, B. M., Aragon-Salamanca, A., et al. 2004, ApJ, 610, L77

De Propris, R., Colless, M., Driver, S. P., et al.2003, MNRAS, 342, 725

de Vaucouleurs, G. 1948, Annalesd’Astrophysique, 11, 247

de Vaucouleurs, G. 1959, Handbuch der Physik,53, 275

Djorgovski, S., & Davis, M. 1987, ApJ, 313, 59

Dressler, A. 1980, ApJ, 236, 351

Dressler, A., Lynden-Bell, D., Burstein, D.,Davies, R. L., Faber, S. M., Terlevich, R., &Wegner, G. 1987, ApJ, 313, 42 (7S)

Dressler, A., Oemler Jr., A., Couch, W. J.,Smail, I., Ellis, R. S., Barger, A., Butcher, H.,Poggianti, B. M., & Sharples, R. M. 1997, ApJ,490, 577

Eggen, O. J., Lynden-Bell, D., & Sandage, A. R.1962, ApJ, 136, 748

Ellis, R. S., Smail, I., Dressler, A., et al. 1997,ApJ, 483, 582

Faber, S. M., & Jackson, R. E. 1976, ApJ, 204,668

Faber, S. M., Friel, E. D., Burstein, D., &Gaskell, C. M. 1985, ApJS, 57, 711

Faber, S. M., Wegner, G., Burstein, D., Davies,R. L., Dressler, A., Lynden-Bell, D., & Terlevich,R. J. 1989, ApJS, 69, 763 (7S)

Faber, S. M., Tremaine, S., Ajhar, E. A., et al.1997, AJ, 114, 1771

Franx, M., & Illingworth, G. D. 1988, ApJ, 327,L55

Frei, Z., & Gunn, J. E. 1994, AJ, 108, 1476

Fish, R. A. 1964, ApJ, 139, 284

Fritz, A., Ziegler, B. L., Bower, R. G., Smail,I., & Davies, R. L. 2005a, MNRAS, 358, 233

Bibliography 187

Fritz, A., Ziegler, B. L., Bohm, A., Balogh,M. L., Bower, R. G., Smail, I., & Davies, R. L.2005b, in “The Fabulous Destiny of Galaxies:Bridging Past and Present”, Marseille, in press

Fritze-v. Alvensleben, U. 1989, PhD thesis, Uni-versitat Gottingen

Fukugita, M., Shimasaku, K., & Ichikawa, T.1995, PASP, 107, 945

Graham, A. W., & Guzman, R. 2003, AJ, 125,2936

Gaztelu, A. 2000, Master of Science Thesis, Uni-versity of Durham

Gebhardt, K., Faber, S. M., Koo, D. C., et al.2003, ApJ, 597, 239

Geller, M. J., & Huchra J. P. 1989, Science, 246,879

Gerhard, O. E. 1993, MNRAS, 265, 213

Gerken, B., Ziegler, B., Balogh, M., Gilbank, D.,Fritz, A., & Jager, K. 2004, A&A, 421, 59

Giovanelli, R., Haynes, M. P., Rubin, V. C., &Ford, W. K., Jr. 1986, ApJ, 301, L7

Girardi, M., Giuricin, G., Mardirossian, F.,Mezzetti, M., & Boschin, W. 1998, ApJ, 505, 74

Gonzalez, J. J. 1993, PhD thesis, University ofCalifornia, Santa Cruz

Heidt, J., Appenzeller, I., Gabasch, A., et al.2003, A&A, 398, 49

Hill, R. J., Ferrarese, L., Stetson, P. B., et al.,1998, ApJ, 496, 648

Hoessel, J. G., Oegerle, W. R., & Schneider,D. P. 1987, AJ, 94, 1111

Holden, B. P., Stanford, S. A., Eisenhardt, P., &Dickinson, M. 2004, AJ, 127, 2484

Holden, B. P., Blakeslee, J. P., Postman, M., etal. 2005, ApJ, 626, 809

Holtzman, J. A., Burrows, C. J., Casertano, S.,et al. 1995, PASP, 107, 1065

Horne, K. 1986, PASP, 98, 609

Hubble, E. 1926, ApJ, 64, 321

James, P. A., & Mobasher, B. 1999, MNRAS,306, 199

Jørgensen, I. 1994, PASP, 106, 967

Jørgensen, I., Franx, M., & Kjærgaard, P. 1995,MNRAS, 273, 1097 (J99)

Jørgensen, I., Franx, M., & Kjærgaard, P. 1996,MNRAS, 280, 167

Jørgensen, I. 1999, MNRAS, 306, 607 (J99)

Jørgensen, I., Franx, M., Hjorth, J., vanDokkum, P. G. 1999, MNRAS, 308, 833 (JFHD)

Kauffmann, G., 1996, MNRAS, 281, 487

Kauffmann, G., & Charlot, S. 1998, MNRAS,297L, 23

Kelson, D. D., van Dokkum, P. G., Franx, M.,Illingworth, G. D., Fabricant, D. 1997, ApJ, 478,L13

Kelson, D. D., Illingworth, G. D., van Dokkum,P. G., & Franx, M., 2000a, ApJ, 531, 137

Kelson, D. D., Illingworth, G. D., van Dokkum,P. G., & Franx, M., 2000b, ApJ, 531, 184

Kennicutt, R. C. 1992, ApJS, 79, 255

Kent, S. M. 1985, PASP, 97, 165

Kinney, A. L., Calzetti, D., Bohlin, R. C., Mc-Quade, K., Storchi-Bergmann, T., & Schmitt,H. R. 1996, ApJ, 467, 38

Kodama, T., & Arimoto, N. 1997, A&A, 320, 41(KA97)

Kodama, T., Yamada, T., Akiyama, M., et al.2004, MNRAS, 350, 1005

188 Bibliography

Kormendy, J., 1977, ApJ, 218, 333

Kormendy, J., & Djorgovski, S. 1989, ARA&A,27, 235

Kormendy, J., & Bender, R. 1996, ApJ, 464,L119

Kron, R. G. 1980, ApJS, 43, 305

Kuijken, K., & Merrifield, M. R. 1993, MNRAS,264, 712

Kuntschner, H. 2000, MNRAS, 315, 184

Kuntschner, H., Smith, R. J., Colless, M.,Davies, R. L., Kaldare, R., & Vazdekis, A. 2002MNRAS, 337, 172

La Barbera, F., Busarello, G., Merluzzi, P., Mas-sarotti, M., & Capaccioli, M. 2002, ApJ, 571, 790

La Barbera, F., Busarello, G., Merluzzi, P., Mas-sarotti, M., & Capaccioli, M. 2003, ApJ, 595, 127

Laird, J. B., & Levison, H. F. 1985, AJ, 90, 2652

Larson, R. B. 1974, MNRAS, 166, 585

Larson, R. B., 1975, MNRAS, 173, 671

Larson, R. B., & Tinsley, B. M. 1978, ApJ, 219,46

Le Borgne, J.-F., Mathez, G., Mellier, Y., et al.1991, A&AS, 88, 133

Le Borgne, J.-F., Pello R., & Sanahuja, B. 1992,A&AS, 95, 87

Le Borgne, J.-F., Bruzual, G., Pello R., et al.2003, A&A 402, 433

Lidman, C., Rosati, P., Demarco, R., Nonino,M., Mainieri, V., Stanford, S. A., & Toft, S.2004, A&A 416, 829

Lubin, L. M., & Sandage, A. 2001, AJ, 122, 1084

Lucey, J. R., Guzman, R., Carter, D., & Ter-levich, R. J. 1991, MNRAS, 253, 584,

Lynden-Bell, D. 1967, MNRAS, 136, 101

Malmquist, K. G. 1920, Medd. Lund Astron.Obs., Ser. II, No. 22

Marleau, F. R., & Simard, L. 1998, ApJ, 507,585

Mehlert, D., Noll, S., Appenzeller, I., et al. 2002,A&A, 393, 809

Mehlert, D., Thomas, D., Saglia, R. P., Bender,R., & Wegner, G. 2003, A&A, 407, 423

Menanteau, F., Abraham, R. G., & Ellis, R. S.2001, MNRAS, 322, 1

Metcalfe, N., Shanks, T., Campos, A., Mc-Cracken, H. J., & Fong, R. 2001, MNRAS, 323,795

Meza, A., Navarro, J. F., Steinmetz, M., & Eke,V. R. 2003, ApJ, 590, 619

Moore, B., Katz, N., Lake, G., Dressler, A., &Oemler, A. 1996, Nature 379, 613

Moller, C. S., Fritze-v.Alvensleben, U., Fricke,K. J., & Calzetti, D. 2001, Ap&SS, 276, 799

Moran, S. M., Ellis, R. S., Treu, T., Smail, I.,Dressler, A., Coil, A. L., & Smith, G. P. 2005,ApJ, 634, 977

Naab, T., & Burkert, A. 2003, ApJ, 597, 893

Noll, S., Mehlert, D., Appenzeller, I., et al. 2004,A&A, 418, 885

Osterbrock, D. E. 1989, in “Astrophysics ofGaseous Nebulae and Active Galactic Nuclei”(Mill Valley, CA, University Science Books)

Pahre, M. A., Djorgovski, S. G., & de Carvalho,R. R. 1998, AJ, 116, 1591

Pello, R., Kneib, J. P., Le Borgne, J.-F., et al.1999, A&A, 346, 359

Peng, C. Y., Ho, L. C., Impey, C. D., & Rix,H.-W. 2002, AJ, 124, 266

Bibliography 189

Pickles, A. J. 1998, PASP, 110, 863

Pipino, A., & Matteucci, F. 2004 MNRAS, 347,968

Poggianti, B. M., Smail, I., Dressler, A., et al.1999, ApJ, 518, 576

Poggianti, B. M., Bridges, T. J., Mobasher, B.,et al. 2001, ApJ, 562, 689

Press, W. H., Teukolsky, S. A., Vetterling,W. T., & Flannery, B. P. 1992, “NumericalRecipes in Fortran 77”, Cambridge UniversityPress

Rix, H.-W., & White, S. D. M. 1990, ApJ, 362,52

Rix, H.-W., & White, S. D. M. 1992, MNRAS,254, 389

Rusin, D., & Kochanek, C. S. 2005, ApJ, 623,666

Saglia R. P., Bertin G., Bertola F., et al. 1993a,ApJ 567, 572

Saglia, R. P., Bender, R., & Dressler, A. 1993b,A&A, 279, 75 (SBD93)

Saglia, R. P., Bertschinger, E., Baggley, G., etal. 1997a, ApJS, 109, 79

Saglia, R. P., Burstein, D., Baggley, G., et al.1997b, MNRAS, 292, 499

Saglia, R. P., Maraston, C., Greggio, L., Bender,R., & Ziegler, B. 2000, A&A, 360, 911

Salpeter, E. E. 1955, ApJ, 121, 161

Sanchez-Blazquez, P., Gorgas, J., Cardiel, N.,Cenarro, J., & Gonzalez, J. J. 2003, ApJ, 590,L91

Sandage, A., & Visvanathan, N. 1978, ApJ, 225,742

Sandage, A., & Perelmuter, J.-M. 1990, ApJ,350, 481

Sargent, W. L. W., Schechter, P. L., Boksenberg,A., & Shortridge, K. 1977, ApJ, 212, 326

Schade, D., Lilly, S. J., Crampton, D., Hammer,F., Le Fevre, O., & Tresse, L. 1995, ApJ, 451, L1

Schade, D., Lilly, S. J., Crampton, D., et al.1999, ApJ, 525, 31

Schaifers, K., Vogt, H. H. (eds.) 1981, Landolt–Bornstein series, Vol. VI/2a, Springer

Schlegel, D. J., Finkbeiner, D. P., & Davis, M.1998, ApJ, 500, 525 (SFD)

Sersic, J. L. 1968, Atlas de Galaxias Aus-trales (Cordoba, Argentina: Observatorio Astro-nomico, Univ. Nac. Cordoba)

Silk, J. 2000, MNRAS, 324, 313

Simard, L., Willmer, C. N., Vogt, N. P., et al.2002, ApJS, 142, 1

Simkin, S. M. 1974, A&A, 31, 129

Sinachopoulos, D., & van Dessel, E. 1998,A&AS, 130, 290

Smail, I., Dressler, A., Couch, W. J., et al. 1997,ApJS, 110, 213

Smail, I., Edge, A. C., Ellis, R. S., & Blandford,R. D. 1998, MNRAS, 293, 124

Spergel, D. N., Verde, L., Peiris, H. V., et al.2003, ApJS, 148, 175

Teerikorpi, P. 1997, ARA&A, 35, 101

Thomas, D., Maraston, C., & Bender, R. 2002,RvMA, 15, 219

Thomas, D., Maraston, C., Bender, R., deOliveira, C. M. 2005, ApJ, 621, 673

Thuan, T. X., & Gunn, J. E. 1976 PASP, 88, 543

Tinsley, B. M. 1980, Fundamentals of CosmicPhysics, Vol. 5, p. 287

190 Bibliography

Tonry, J., & Davis, M. 1979, AJ, 84, 1511

Toomre, A., & Toomre, J. 1972, ApJ, 178, 623

Tran, K. H., Simard, L., Illingworth, G., &Franx, M. 2003, ApJ, 590, 238

Treu, T., Stiavelli, M., Møller, P., Casertano, S.,& Bertin, G. 2001a MNRAS, 326, 221

Treu, T., Stiavelli, M., Bertin, G., Casertano, S.,& Møller, P. 2001b MNRAS, 326, 237

Treu, T., Stiavelli, M., Casertano, S., Møller, P.,& Bertin, G. 2002, ApJ, 564, L13

van der Marel, R. P., & Franx, M. 1993, ApJ,407, 525

van der Wel, A., Franx, M., van Dokkum, P. G.,Rix, H.-W., Illingworth, G. D., & Rosati, P.2005, ApJ, 631, 145

van Dokkum, P. G., & Franx, M. 1996, MNRAS,281, 985

van Dokkum, P. G., Franx, M., Kelson, D. D.,& Illingworth, G. D. 1998, ApJ, 504, L17

van Dokkum, P. G., Franx, M., Fabricant, D.,Kelson, D. D., & Illingworth, G. D. 1999, ApJ,520, L95

van Dokkum, P. G., Franx, M., Fabricant, D.,Illingworth, G. D., & Kelson, D. D. 2000, ApJ,541, 95

van Dokkum, P. G., Stanford, S. A., Holden,B. P., Eisenhardt, P. R., Dickinson, M., & El-ston, R. 2001a, ApJ 552, L101.

van Dokkum, P. G., & Franx, M. 2001b, ApJ,553, 90

van Dokkum, P. G., Franx, M., Kelson, D. D. &Illingworth, G. D. 2001c, ApJ, 553, L39

van Dokkum, P. G., & Stanford, S. A. 2003, ApJ,585, 78

Vikhlinin, A., McNamara, B. R., Forman, W.,Jones, C., Quintana, H., & Hornstrup, A. 1998,ApJ, 502, 558

Wagner, R. M., 1992, in “Astronomical CCD ob-serving and reduction techniques”, ASP Conf.Series Vol. 23, Howell., S. B. (ed.), AstronomicalSociety of the Pacific, p. 160

White, S. D. M., & Rees, M. J. 1978, MNRAS,183, 341

Williams, R. E., Blacker, B., Dickinson, M., etal. 1996, AJ, 112, 1335

Williams, R. E., Baum, S., Bergeron, L., et al.2000, AJ, 120, 2735

Worthey, G., Faber, S. M., Gonzalez, J. J., &Burstein, D. 1994, ApJS, 94, 687

Wuyts, S., van Dokkum, P. G., Kelson, D. D.,Franx, M., & Illingworth, G. D. 2004, ApJ, 605,677

Yang, Y., Zabludoff, A. I., Zaritsky, D., Lauer,T. R., & Mihos, J. C. 2004, ApJ, 607, 258

Yee, H. K. C., Ellingson, E., Abraham, R. G., etal. 1996, ApJS, 102, 289

Ziegler, B. L., & Bender, R. 1997, MNRAS, 291,527

Ziegler, B. L., Saglia, R. P., Bender, R., Belloni,P., Greggio, L., & Seitz, S. 1999, A&A, 346, 13

Ziegler, B. L., Bower, R. G., Smail, I., Davies,R. L., & Lee, D., 2001a, MNRAS, 325, 1571

Ziegler, B. L., Fricke, K. J., Balogh, M. L.,Bower, R. G., Gaztelu, A., Smail, I., & Davies,R. L. 2001b, in “Gas and Galaxy Evolution”, Hi-bbard, J. E., Rupen, M., & van Gorkom, J. H.(eds.), ASP Conf. Proceedings, Vol. 240, p. 619

Ziegler, B. L., Bohm, A., Fricke, K. J., et al.2002, ApJ, 564, L69

Bibliography 191

Ziegler, B. L., Thomas, D., Bohm, A., Bender,R., Fritz, A., & Maraston, C. 2005, A&A, 433,519

192 Bibliography

193

Acknowledgements

First of all I want to thank Prof. Dr. K. J. Frickefor his permanent interest in my work, hiscontinuous support and for giving me the op-portunity to acquire the doctoral level in oneof the most spellbinding fields of modern ex-tragalactic astronomy by using spectroscopicdata from the state-of-the-art telescopes, theVLT and CA 3.5 m. My further thanks go toProf. Dr. W. Lauterborn, who kindly agreedto become the secondary referee of my the-sis. In particular, I am deeply grateful toDr. B. L. Ziegler, the leader of our Junior Re-search Group (sponsored by the VolkswagenFoundation) for his support and encouragement,many helpful and interesting discussions, andalso for the thorough proof-reading of the draftgiving many useful comments on it.Furthermore, I wish to thank the other membersof the disputation committee, Prof. Dr. K. Bahr,Prof. Dr. U. Christensen, Prof. Dr. F. Kneer andProf. Dr. W. Kollatschny.Thanks also to Dr. A. Bohm, Dr. C. Hallidayand Dr. P. Papaderos for their proof-reading ofsome chapters.I am indebted a lot to my main collaborators,Prof. Dr. M. L. Balogh (Waterloo), Prof. Dr.R. G. Bower (Durham), Prof. Dr. R. L. Davies(Oxford), Prof. Dr. I. R. Smail (Durham), andDr. D. Thomas (Oxford) for their support andstimulating discussions on various topics.I am also grateful to the P.I. of the FORS DeepField project, Prof. Dr. I. Appenzeller, for hiscontinuous support of the field elliptical galaxyproject. Prof. Dr. R. Bender and Priv. Doz. Dr.R. P. Saglia (Munchen) are thanked for kindlyproviding the FCQ program and the surfacebrightness fitting routine, respectively.Thanks a lot to the (former) members of ourJunior Research Group, in particular to Drs.A. Bohm, K. Jager and I. Berentzen, for all thehelpful and interesting discussions, helps in var-ious aspects and the nice working atmosphere.

I am also grateful to Priv. Doz. Dr. J. Heidt,Drs. D. Mehlert and S. Noll (all Heidelberg), whoperformed the main parts of the spectroscopy forthe FDF project during the Guaranteed TimeObservations in 2000 and 2001. I also thank Drs.K. Reinsch and C. Mollenhoff (Heidelberg) forthe pilot observations in Dec. 1999.Several other people from foreign instituteshave contributed to my understanding of theproject in various ways and thus I want tothank are Drs. A. Diaferio (Torino), A. W. Gra-ham (RSAA), H. Kuntschner (ESO), S. Mieske(ESO), U. Thiele (CAHA), and T. Treu (UCLA).

A continuous strong assistance during the courseof this period has been the support of my par-ents, Susanne and Karl, my grandmother Agnes,my aunt Eva, uncle Willi as well as my cousinThomas. Without their big endeavours, theirlove and understanding it would have been a lotharder and therefore I am deeply thankful to allof them. I am also indebted to my parents forgiving me the possibility to carry out my studies.Last but not least, thanks a lot to Andi, Asmus,Gregor and Markus, for enriching my life withtheir great friendship and for having cheered meup so many times.

Financial support by the Volkswagen Founda-tion (I/76 520) and the Deutsche Forschungsge-meinschaft (ZI 663/1-1, ZI 663/2-1, ZI 663/3-1,ZI 663/5-1) are gratefully acknowledged. I amfurthermore grateful for travel founds to theJENAM 2002 conference in Porto and severalAG annual meetings (Munchen, Berlin, Freiburgand Prague). The Calar Alto and ESO VLT staffare thanked for efficient observational support.This research has made use of the NASA’s As-trophysics Data System and NASA/IPAC Ex-tragalactic Database (NED). This thesis hasbeen created on computers of the Universitats–Sternwarte Gottingen, Institut fur Astrophysikand the Gesellschaft fur WissenschaftlicheDatenverarbeitung Gottingen (GWDG).

Curriculum Vitae

Personliche Daten:

Name: Alexander FritzAdresse: Christophorusweg 12/102

37075 GottingenGeburtsdatum: 23. August 1976Geburtsort: Wien, OsterreichNationalitat: osterreichischFamilienstand: ledig

Ausbildung:

Jan. 2001 Beginn der Doktorarbeit an der Universitats–Sternwarte GottingenThe Dependence of the Evolution of Early–Type Galaxies on their EnvironmentBetreuer: Prof. Dr. K. J. Fricke

Sep. 2000 Verleihung des akademischen Grades ”Magister Rerum Naturalium“Sep. 2000 Diplomarbeit am Institut fur Astronomie, Universitat Wien, Osterreich

Bestimmung Extragalaktischer Entfernungen mit FlachenhelligkeitsfluktuationenBetreuer: Prof. Dr. W. W. Zeilinger

Sep. 2000 2. Diplomprufung Astronomie mit Auszeichnung (Note: ”sehr gut“)Dez. 1997 1. Diplomprufung Astronomie1994 – 2000 Studium Astronomie & Geschichte an der Universitat Wien, OsterreichJuni 1994 Matura mit Auszeichnung am Realgymnasium, 1150 Wien1986 – 1994 Privates Realgymnasium, 1150 Wien, Osterreich1982 – 1986 Offentliche Volksschule, 1230 Wien, Osterreich

Beruflicher Werdegang:

Seit 2001 Wissenschaftlicher Mitarbeiter an der Universitats–Sternwarte Gottingen undInstitut fur Astrophysik Gottingen;Kollaborationen: Munchen, Potsdam, Oxford, Durham (UK), Waterloo (CA);Mehrere Beobachtungsreisen zum Calar Alto Observatorium (Spanien);Vortrage auf internationalen und nationalen Tagungen:Alpbach (Osterreich), Munchen, Porto (PT), Heidelberg, Freiburg, Prag (CZ)

Okt. 2005 Offentliche Fuhrungen am Tag der offenen Tur des Physik–Neubaus, GottingenSep. 2004 Offentliche Fuhrungen am 2. internationalen Tag der Astronomie

”Lange Nacht der Sterne“ an der Universitats–Sternwarte Gottingen2002 Betreuung der Praktikantin Sarah BuhlerMai 2000 Teilnahme an der ”Science Week @ Austria“ fur das Inst. fur Astronomie,

Betreuer fur Sonnenteleskop, zwei Spektrographen sowie offentliche Fuhrungen1998 – 2000 Tutor im Astronomischen Praktikum fur Fortgeschrittene

The Dependence of the Evolution of Early-Type Galaxies on their … · 2015-07-15 · The...

Documents

Transcript of The Dependence of the Evolution of Early-Type Galaxies on their … · 2015-07-15 · The...