Rights / License: Research Collection In Copyright - …Research Collection Doctoral Thesis The...

Research Collection

Doctoral Thesis

The Evolution of Star-forming and Quiescent Massive Galaxiesthrough Cosmic Time

Author(s): Faisst, Andreas L.

Publication Date: 2015

Permanent Link: https://doi.org/10.3929/ethz-a-010498478

Rights / License: In Copyright - Non-Commercial Use Permitted

This page was generated automatically upon download from the ETH Zurich Research Collection. For moreinformation please consult the Terms of use.

ETH Library

https://doi.org/10.3929/ethz-a-010498478

http://rightsstatements.org/page/InC-NC/1.0/

https://www.research-collection.ethz.ch

https://www.research-collection.ethz.ch/terms-of-use

DISS. ETH NO. 22641

The Evolution of Star-forming andQuiescent Massive Galaxies

through Cosmic Time

A thesis submitted to attain the degree of

DOCTOR OF SCIENCES of ETH ZURICH

(Dr. sc. ETH Zurich)

presented by

ANDREAS LUKAS FAISST

MSc ETH Physics, ETH Zurich

born on 24.01.1988

citizen of Amlikon-Bissegg, TG

accepted on the recommendation of

Prof. Dr. Marcella Carollo, examinerProf. Dr. Claudia Scarlata, co-examiner

2015

Acknowledgements

Without the help of many people – no matter whether professional astronomers or “just”good friends – I would not have been able to realize this thesis.

First of all, my supervisor, Marcella Carollo, deserves a great “thank-you” forsupporting me throughout my 4-year Ph.D. The numerous visits to Caltech, which shemade possible, helped me to build up my future research collaborations. She taughtme scientific thinking, how to make scientific plots, how to present scientific work atconferences, and, most importantly, how to interact with different kinds of people in thecompetitive world of professional astronomy. I hope that we will stay in contact and willshare our scientific interests in the future.

I also want to thank my external collaborators in California, specifically at Caltechwhere I have spend a significant time of my Ph.D. Special thank goes to Peter Capak andNick Scoville for their hospitality and for creating a very nice and stimulating workingenvironment. They supported me in my projects and introduced me to the worlds largesttelescopes. I’m looking forward to continue working with you in the following years!

But... the most important contributors to a good working environment are the otherPh.D. students and post-docs, which whom I have spend a considerable time of mycareer. I would like to thank all of you (nota bene: not only cosmologists but alsoplanetary scientists!) for supporting me during my Ph.D. Sascha, for numerous supportingdiscussions about academic life as well as some very interesting discussions about (exo-)planets; Benny, for support on writing talks, applications, and proposals (although notaccepted); Adam, Aseem, Joanna, Katarina, Kurt, Maryam, Masato, and Will, for fruitfulscientific debates and discussions; Anna, for keeping my chocolate level high enough soI was able to work; Andrina and Elena, for watering my plants; Federica, Lia, Martina,Micol, and Antonio, for delicious Italian food; Neven and Simon, for joining the volleyballteam (keep it up!); Claudio and Sebastian, for everything that has to do with statistics,beer, and dogs; Sarah, for joining me for lunch. Last but not least, Sandro, it was a lotof fun having you as officemate! A very special thank also goes to Esther, our secretary.Thanks a lot for (re-) booking flights and hotels, organizing trips and christmas parties,charging batteries, and printing lecture notes!

However, my parents (Jacqueline and Siegfried) are the most important support. Theystood behind me in any situation of my life, in moments of distress and overflow!

Andreas Faisst

iii

Contents

Acknowledgements iii

Contents v

List of Figures viii

List of Tables x

List of Acronyms and Symbols xi

Summary xvii

Zusammenfassung xxi

1 Introduction 31.1 The Epoch of Re-ionization . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2 The cosmic star-formation rate density . . . . . . . . . . . . . . . . . . . . . 5

1.3 The main-sequence of star-forming galaxies . . . . . . . . . . . . . . . . . . 7

1.4 The emergence of quiescent galaxies . . . . . . . . . . . . . . . . . . . . . . 9

1.5 Size evolution of quiescent and star-forming galaxies . . . . . . . . . . . . . 11

1.6 Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Part I The Tools: Surveys and Instrumentations

2 Surveys and Instrumentations 17

2.1 The Cosmic Evolution Survey . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.1.1 Space based observations . . . . . . . . . . . . . . . . . . . . . . . . 18

2.1.2 Ground based observations . . . . . . . . . . . . . . . . . . . . . . . 182.1.3 Spectroscopic follow-up: zCOSMOS . . . . . . . . . . . . . . . . . . 19

2.2 Spectroscopy at Keck: DEIMOS and MOSFIRE . . . . . . . . . . . . . . . 19

Part II The Early Universe: Star-formation and Re-ionization

v

CONTENTS

3 Spectroscopic observations of LAEs at z ∼ 7.7 & implications on re-ionization 233.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233.3 Observations & analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.3.1 Candidate selection by Krug et al. (2012) . . . . . . . . . . . . . . . 25

3.3.2 MOSFIRE observations & data reduction . . . . . . . . . . . . . . . 263.3.3 Tests and Simulations: Establishing our detection limits . . . . . . . 27

3.3.4 No detection of Lyα in LAE1 and LAE2 . . . . . . . . . . . . . . . . 28

3.4 The evolution of the Lyα LF from z = 3.1 to z = 7.7 . . . . . . . . . . . . . 28

3.5 The fraction of neutral hydrogen at z ∼ 8 . . . . . . . . . . . . . . . . . . . 31

3.5.1 A model of the LAE galaxy population . . . . . . . . . . . . . . . . 31

3.5.2 Interpreting the evolution of LAEs . . . . . . . . . . . . . . . . . . . 33

3.5.3 Constraint on xHI and Lyα optical depth at z ∼ 7.7 . . . . . . . . . 33

3.6 Expected number detections of LAEs at z ∼ 8.8 in other surveys . . . . . . 36

3.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

Part III After the Cosmic Peak: Quenching of Star-formation inMassive Galaxies

4 Massive Galaxies at z . 2 in COSMOS: Size Evolution and Quenching 43

4.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

4.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434.3 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

4.3.1 UltraVISTA near-IR imaging data . . . . . . . . . . . . . . . . . . . 46

4.3.2 Photometric redshift and stellar mass catalog . . . . . . . . . . . . . 46

4.3.3 CANDELS/COSMOS near-IR imaging data . . . . . . . . . . . . . . 47

4.4 The Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

4.4.1 High- and low-mass galaxies . . . . . . . . . . . . . . . . . . . . . . . 47

4.4.2 Selection of quiescent and star-forming galaxies . . . . . . . . . . . . 49

4.5 Size measurements and calibration . . . . . . . . . . . . . . . . . . . . . . . 494.5.1 Determination of the spatially varying PSF . . . . . . . . . . . . . . 49

4.5.2 Guess-parameters for surface brightness fitting . . . . . . . . . . . . 51

4.5.3 Uncalibrated size measurements . . . . . . . . . . . . . . . . . . . . 514.5.4 Correcting for measurement biases using simulated galaxies . . . . . 52

4.5.5 Final calibration of size measurements using CANDELS . . . . . . . 53

4.5.6 Correction for internal color gradients . . . . . . . . . . . . . . . . . 55

4.5.7 Verification of accuracy of size measurement . . . . . . . . . . . . . . 55

4.6 Results: Size evolution of very massive galaxies . . . . . . . . . . . . . . . . 57

4.6.1 The stellar mass vs. size relation . . . . . . . . . . . . . . . . . . . . 574.6.2 Size evolution and indication of fast quenching of massive galaxies . 58

4.7 Predicted size evolution of massive quiescent galaxies . . . . . . . . . . . . . 61

4.7.1 The main-sequence life time τMS . . . . . . . . . . . . . . . . . . . . 63

vi

CONTENTS

4.7.2 The consumption time scale τcons . . . . . . . . . . . . . . . . . . . . 63

4.7.3 Size evolution of our model galaxies . . . . . . . . . . . . . . . . . . 65

4.8 How to quench massive galaxies . . . . . . . . . . . . . . . . . . . . . . . . . 66

4.8.1 ”Waning model”: tension with observations . . . . . . . . . . . . . . 66

4.8.2 The effect of mergers (”instantaneous quenching”) . . . . . . . . . . 66

4.9 Summary & conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

Conclusions & Future Research 75

Appendix 79

A Summary of collaborative projects 79

B Keck-I MOSFIRE spectroscopy of a z ∼ 12 candidate galaxy 83

B.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

B.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83B.3 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85B.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89B.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

C Dust attenuation in high redshift galaxies: “Diamonds in the Sky” 93

C.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

C.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93C.3 CIV Absorption as a Signpost for the SED . . . . . . . . . . . . . . . . . . 96

C.4 Galaxy Sample Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

C.5 Numerical Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98C.6 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100C.7 Summary and Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

Bibliography 105

Publications 119

vii

CONTENTS

viii

List of Figures

1.1 The three phases of re-ionization . . . . . . . . . . . . . . . . . . . . . . . . 4

1.2 Cosmic star-formation density . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.3 The main-sequence of star-forming galaxies . . . . . . . . . . . . . . . . . . 8

1.4 Fraction of quiescent galaxies as a function of redshift and stellar mass . . . 10

1.5 Size evolution of quiescent galaxies . . . . . . . . . . . . . . . . . . . . . . . 12

3.1 Slit alignment and simulations of LAEs at z ∼ 7.7 . . . . . . . . . . . . . . 26

3.2 MOSFIRE Y−band sensitivity curve . . . . . . . . . . . . . . . . . . . . . . 29

3.3 Compilation of Lyα luminosity function measurements . . . . . . . . . . . . 30

3.4 Model predictions of Lyα luminosity functions up to z ∼ 9 . . . . . . . . . . 32

3.5 Fraction of strong Lyα emitting galaxies at z ∼ 7.7 . . . . . . . . . . . . . . 34

3.6 Lyα optical depth as a function of redshift . . . . . . . . . . . . . . . . . . . 37

4.1 Selection of ultra massive galaxies at log(m/M) > 11.4 . . . . . . . . . . . 48

4.2 Selection of quiescent and star-forming ultra massive galaxies atlog(m/M) > 11.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

4.3 Calibration of ground based size measurements using space based imaging . 54

4.4 Size evolution of ultra massive galaxies at log(m/M) > 11.4 . . . . . . . . 56

4.5 m−Re evolution of star-forming and quiescent galaxies . . . . . . . . . . . 58

4.6 Visualization of the ”waning model” to evolve galaxies from star-formingto quiescent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

4.7 Consumption time scales for different redshifts and stellar masses . . . . . . 64

4.8 Difference between predicted and observed size evolution of quiescent galaxies 65

4.9 Effect of mergers on the size evolution of quiescent galaxies . . . . . . . . . 67

4.10 Gas fraction as a function of stellar mass and redshift . . . . . . . . . . . . 71

B.1 Slit alignment of UDFj-39546284 and three other bright galaxies . . . . . . 86

B.2 MOSFRE H−band sensitivity curve . . . . . . . . . . . . . . . . . . . . . . 87

B.3 Signal-to-noise map for and jack-knife sampling of UDFj-39546284 . . . . . 88

C.1 Simulated spectra from the Starburst99 library and sample selection . . . . 95

C.2 Redshift and stellar mass distribution of z = 2− 6 galaxies . . . . . . . . . 97

C.3 Example fits for τ1300A

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

C.4 Distribution of τ1300A

for the full sample z = 2 to 6 . . . . . . . . . . . . . . 99

ix

LIST OF FIGURES

C.5 Extinction curve for 2.0 < z < 6.5 . . . . . . . . . . . . . . . . . . . . . . . . 100C.6 Split extinction curves for 2 < z < 4 and 4 < z < 6.5 . . . . . . . . . . . . . 101

x

List of Tables

3.1 Large LAE surveys at 3 < z < 5 . . . . . . . . . . . . . . . . . . . . . . . . 39

3.2 LAE surveys at z ∼ 5.7, 6.6, 7.7, and 8.8 . . . . . . . . . . . . . . . . . . . . 40

4.1 Size evolution of star-forming and quiescent galaxies at z < 2 . . . . . . . . 60

4.2 MR evolution of star-forming and quiescent galaxies at z < 2 . . . . . . . . 60

B.1 Targeted objects together with UDFj-39546284 . . . . . . . . . . . . . . . . 89

C.1 Best-fit attenuation curve for 2.0 < z < 6.5 . . . . . . . . . . . . . . . . . . 102

xi

ACRONYMS AND SYMBOLS

xii

List of Acronyms and Symbols

The most commonly used acronyms throughout this work.

ACS Advanced Camera for Surveys (on board of HST)AGN active galactic nucleus

ALMA Atacama Large Millimeter/submillimeter ArrayAO adaptive opticsBH black holes

B/T bulge-to-total light ratioCANDELS Cosmic Assembly Near-infrared Deep Extragalactic Legacy SurveyCDM cold dark matterCMB cosmic microwave backgroundCOSMOS Cosmic Evolution SurveyDLA damped Lyα systemE-ELT European Extremely Large TelescopeEoR epoch of re-ionizationERS Early Release Science

eV electron Volt (= 1.6 10−19 joule)EW equivalent widthFUV far-ultraviolet

FWHM full width at half maximum (= 2.35× σ for gaussian)GALEX Galaxy Evolution ExplorerGOODS Great Observatories Origins Deep SurveyGOODS-MUSIC GOODS Multiwavelength Southern Infrared CatalogHI neutral hydrogenHII ionized hydrogenHST Hubble Space TelescopeHUDF Hubble Ultra Deep Field

HUDF09 WFC3/IR follow-up survey of the HUDFIGM inter-galactic mediumIMF initial mass functionIR infra-red

IRAC Infrared Array Camera (on board of Spitzer)

xiii


ISM inter-stellar mediumJWST James Webb Space TelescopeKMOS K-band multi-object spectrographkpc kilo-parsec

KS (-relation) Kennicutt-Schmidt (relation)LAE Lyman alpha emitterLBG Lyman break galaxyLF luminosity functionLMC Large Magellanic CloudLyα Lyman αMF mass functionMpc mega-parsecMS star-forming main sequenceMW Milky WayNB narrow-bandNIR near-infraredNUV near-ultravioletPS power spectrumPSF point-spread functionSDSS Sloan Digital Sky SurveySED spectral energy distribution

SFH star-formation history (in M yr−1)SFR star-formation rate

SFRD star-formation rate density (in M yr−1 Mpc−3)SINFONI spectrograph for integral field observations in the near infraredSKA Square Kilometer ArraySMC Small Magellanic Cloud

S/N signal-to-noise ratioSPLASH Spitzer Large Area Survey with Hyper-Suprime-Cam

sSFR specific star-formation rate (= SFR/m in yr−1)

SXDS Subaru/XMM-Newton Deep SurveyTMT Thirty Meter Telescope

UMG ultra massive galaxy (log(m/M) > 11.4)UV ultra-violet

WFC3 Wide Field Camera 3 (on board of HST)WMAP Wilkinson Microwave Anisotropy ProbeZEBRA Zurich Extragalactic Bayesian Redshift AnalyzerZENS Zurich Environmental StudyZEST Zurich Estimator of Structural TypeszCOSMOS spectroscopic follow-up of COSMOS

xiv


The most commonly used symbols throughout this work.

E(B − V ) extinction in rest-frame B − V color (in magnitudes)fesc escape fraction of ionizing photons into the IGM

m stellar mass (in M)M∗ characteristic mass of Schechter function

Re effective half-light radius of a galaxy (in kpc of not declareddifferently)

R spectral resolution, R(λ) ≡ λ/∆λΩm cosmological matter density (Ωm ∼ 0.30)

ΩΛ cosmological dark energy density (ΩΛ ∼ 0.70)xHI

fraction of neutral hydrogenXLyα fraction of strong Lyα emitting LBGs

Z Metallicity (in Z)

xv


xvi

Summary

The formation of the first galaxies started roughly 13 Gyrs ago by the accretion ofbaryons onto the deep gravitational potentials of Dark Matter (DM) halos. The ionizingphotons produced by the first stars in these young galaxies induced a phase transitionin the universe, turning its hydrogen from neutral to ionized (this is called the Epochof Re-ionization, EoR). Subsequently, the galaxies grow in size and mass by mergingand accretion of gas from their surroundings. Finally, around 7 − 10 Gyrs ago, thestar-formation of galaxies starts to cease. This causes them to turn from star-forminggalaxies into red and quiescent galaxies out of which a majority shows an ellipticalmorphology. This thesis focuses on two main stages during the evolution of galaxies.The EoR at redshifts z > 6 and the emergence of the quiescent galaxy population shortlyafter the peak of the cosmic star-formation density at z ∼ 2.

Part I - The Tools: Surveys and Instrumentations

The first part of this thesis is devoted to surveys and instrumentations which havesubstantially pushed forward the field of observational cosmology. If nothing else, theseare the backbone of this thesis. One of the most important roles plays the Comic EvolutionSurvey (COSMOS) field which covers roughly two square degrees on sky in more than 30pass-bands from the ultra-violet (UV) to the infra-red (IR). The COSMOS field is crucialto study the rarest galaxy populations that include very high redshift as well as themost massive currently observable galaxies. Using a synergy of large area ground basedimaging and small area space based imaging allows advancing to interesting territoriesin astrophysics and tackle the currently open questions. Furthermore, efficient near-IRspectrographs as the Multi-object Spectrometer for Infra-red Exploration (MOSFIRE)or the K-band multi-object spectrograph (KMOS) with first-light in 2013 enable largespectroscopic surveys at high redshifts as well as detailed studies of galaxy properties(kinematics, metallicity, etc) at lower redshifts.

Part II - The Early Universe: Star-formation and Re-ionization

Currently one of the most debated topics in modern astrophysics is how and whenthe universe was ionized as well as to characterize the very first galaxies in terms oftheir physical properties. The most promising candidates for ionization are low massstar-forming galaxies that are abundant in the early universe. At later time, quasars andactive galactic nuclei (AGNs) are likely to dominate the budget of ionizing photons. The

xvii

SUMMARY

end of re-ionization is measured from absorptions in quasar spectra to be around z ∼ 6.The time evolution of the fraction neutral hydrogen as well as its spatial distribution inthe universe before z ∼ 6, is however largely unknown. Both is important to understandthe formation of the first galaxies. Probing the physical properties of galaxies at z > 6and using them as tracers of the EoR is still a big challenge. Most importantly, thecontamination of the very high redshift (z & 7) samples with low redshift galaxies orfalse-positives is a serious problem. The recently installed near-IR spectrograph MOSFIREas well as KMOS enables us to follow-up the brightest high redshift candidates efficientlyin order to confirm our samples spectroscopically and also to characterize the rate ofcontamination.

In Part II of this thesis, we outline our spectroscopic follow-up of two strong Lyαemitting galaxies at z ∼ 7.7. These types of galaxies are commonly used as tracers ofthe cosmic re-ionization. Although the two galaxies seemed to be reliably detected inground-based narrow-band imaging, we are not able to detect their strong Lyα emissionin our much deeper observations and identify them as false-positives. This states ahigh contamination rate (by low redshift galaxies and spurious detections) in searchesof z > 7 galaxies based on ground-based narrow-band surveys. This has a ground-shakingimpact on past determinations of the neutral hydrogen fractions at z > 7 using strongLyα emitting galaxies and resolves the tensions with results from other, independentmeasurement techniques at lower redshifts. We model the evolution of the Lyα luminosityfunction and put the first constraints on the neutral hydrogen fraction at z ∼ 8 thatwe estimate to be around 50 − 70%. The advantage of our model is that it treats alsothe evolution of internal properties of galaxies (like star-formation rate and dust) that isdegenerate with the effects of re-ionization.

Part III - After the Cosmic Peak: Quenching of Star-formation in MassiveGalaxies

At lower redshifts (z < 2), the emergence of quiescent galaxies and their evolution withcosmic time is currently one of the hot topics in astrophysics. Many plausible processesthat shut-down the star-formation (“quenching”) are suggested but their dominancecompared to each other is not well understood, yet. The size evolution of star-forming andquiescent galaxies is substantially different in terms of normalization as well as steepness.What causes the size evolution of quiescent galaxies as it is, is still not fully understood,either, but must be linked to the quenching process. Moreover, different quenchingmechanism can lead to cessation of star-formation that may occur on different time scales.

In Part III of this thesis, we use the size evolution of massive log(m/M) > 11.0quiescent galaxies at z . 2 as a diagnostic tool to investigate the process and time scalesof quenching. We follow an empirical model based on the stellar mass, star-formation,and size evolution of star-forming galaxies and mimic a slow (due to gas cut off) andan instantaneous (due to mergers) quenching process. We find that the observed sizeevolution of such massive galaxies is best described by the second process.

xviii

SUMMARY LIST OF TABLES

Appendix - Collaborative Work

The characterization of young, star-forming galaxies at high redshifts (z > 3) in terms oftheir physical properties is observationally challenging. On one hand, the contaminationin high redshift galaxy samples at z > 7 by low redshift galaxies as well as spuriousdetections is a serious issue (see Chapter B based on Capak et al. 2013). On the otherhand, we are only able to observe the rest-frame UV wavelength range at these redshiftswith current instrumentations. This part of wavelength is however affected a lot bydust obscuration. Dust extinction corrections therefore become crucial at high redshifts.Commonly, an empirically derived dust extinction curve from local star-burst galaxies istherefore used at high redshifts. Different extinction curves (small Magellanic cloud orMilky Way like) with different steepness as a function of wavelength are suggested as well.In Scoville et al. (2014) (see Chapter C), we investigate the dust extinction properties of 266spectroscopically confirmed galaxies at 2 < z < 6 by deriving an empirical dust extinctioncurve. We find that our sample of galaxies is described well by the dust extinction curveof local star-burst galaxies. This derivation of the dust extinction curve provides a firmbasis for color and extinction corrections of high redshift galaxy’s photometry.

Moreover, I have contributed to projects that focus on the dependence of physicalparameters on the galaxy environment.

In Scoville et al. (2013) we derive the large scale structure on the 2 square degreeCOSMOS field using more than 150,000 galaxies out to z ∼ 3 in 127 redshift slices. Wefind ∼ 250 statistically significant over-dense structures and confirm on a larges samplethat red, quiescent, early-type galaxies preferentially live in over-dense environments.

In Carollo et al. (2015, in prep.) we study the morphological mix and galaxy sizeversus mass relation in different environments at 0.2 < z < 0.8. We use more than200 spectroscopically confirmed galaxy groups to split the population into centrals (mostmassive in group), satellites (everything but centrals), and field galaxies. Our preliminaryresults, suggest that the morphological mix (disk galaxies versus ellipticals) is constantwith radial position from the center of the groups. Also the mass-size relation is notsignificantly different in different environments.

xix

SUMMARY

xx

Zusammenfassung

Die Bildung der ersten Galaxien begann vor rund 13 Milliarden Jahre durch dieAkkretion von Baryonen in die starken Gravitationspotentiale dunkler Materie (DM).Die ionisierenden Photonen, welche durch die ersten Sterne in diesen jungen Galaxienproduziert wurden, starteten einen Phasenubergang des fruhen Universums indem sie denneutralen Wasserstoff ionisierten. Diese Epoche wird als die Epoche der Reionisation(EoR) bezeichnet. In der Zeit danach wuchsen die Galaxien in Grosse und Masse durchKollisionen mit anderen Galaxien und durch die Akkretion von Gas aus ihrer Umgebung.Schlussendlich, vor rund 7 bis 10 Milliarden Jahren, begann die Sternproduktion derGalaxien sich zu verlangsamen. Das bewirkte eine Veranderung der Galaxien; vonsternbildenden zu roten und inaktiven Galaxien, von welchen die meisten ihre Strukturverloren (und als elliptische Galaxien bezeichnet werden). Diese Doktorarbeit beschaftigtsich mit zwei wichtigen Stadien in der Entstehungsgeschichte von Galaxien: Der EoR beiRotverschiebungen von z > 6 und der Bildung von inaktiven Galaxien kurz nach dermaximalen Dichte der Sternproduktion bei z ∼ 2.

Teil I - Das Werkzeug: Surveys und Instrumente

Der erste Teil dieser Doktorarbeit beschaftigt sich mit verschiedenen Surveys undtechnischen Instrumenten, welche beigetragen haben das Feld der Astrophysikvoranzutreiben. Gerade deshalb bilden sie das Ruckgrat dieser Arbeit. Eine wichtigeRolle spielt dabei die Cosmic Evolution Survey (COSMOS), welche fast zwei Quadratgraddes Himmels in mehr als 30 photometrischen Bandern im Ultra-violetten (UV) bis hinzum Infraroten (IR) beobachtet hat. Die COSMOS ist sehr wichtig fur das Studium derseltensten Populationen von Galaxien, welche Galaxien bei sehr hoher Rotverschiebungals auch sehr massive Galaxien umfassen. Mit dem Zusammenwirken von Beobachtungenvon grossen Feldern am Himmel von Boden aus und kleinen Feldern vom Weltall aus,konnen viele der brennendsten Fragen der heutigen Astrophysik beantwortet werden.Sehr effiziente Spektrographen im nahen Infrarot, wie der Multi-object Spectrometer forInfra-red Exploration (MOSFIRE) oder der K-band Multi-Object Spectrograph (KMOS),haben es ermoglicht grosse Beobachtungskampanien bei sehr hoher Rotverschiebung alsauch sehr detaillierte Studien der Eigenschaften der Galaxien bei tieferer Rotverschiebungdurchzufuhren.

xxi

ZUSAMMENFASSUNG

Teil II - Das fruhe Universum: Sternentstehung und Reionisation

Eines der meist debattierten Gebiete in der modernen Astrophysik ist es, wie und wann dasUniversum re-ionisiert wurde, sowie welche Eigenschaften die ersten Galaxien aufweisen.Die besten Kandidaten fur die Reionisation sind weniger schwere, sternbildende Galaxien,welche im fruhen Universum sehr weit verbreitet sind. In der darauf folgenden Zeit ist eswahrscheinlich, dass sogenannte Quasare und Galaxien mit aktivem Kern (AGNs) diemeisten ionisierenden Photonen entsenden. Das Ende der Reionisation wurde durchdie Absorption in Spektren von Quasaren gemessen und wird auf z ∼ 6 geschatzt.Die Anderung des Anteils von neutralem Wasserstoff sowie auch dessen Verteilung imUniversum oberhalb von z ∼ 6 ist nach wie vor unklar. Beides ist wichtig um dieEntstehung der ersten Galaxien zu verstehen. Die physikalischen Eigenschaften derGalaxien bei z > 6 zu messen und sie bei der Untersuchung der EoR zu nutzen isteine grosse Herausforderung. Sehr dominant ist die Vermischung der Galaxienprobenbei hoher Rotverschiebung (z > 7) durch Galaxien tieferer Rotverschiebung und anderenicht physikalische Artefakte in den Messungen. Die neulich installierten SpektrographenMOSFIRE und KMOS, welche im nahen Infrarot arbeiten, erweisen sich als sehr wichtigum die Rotverschiebung dieser Galaxien zu bestatigen und auch die Vermischung mitanderen Galaxien zu quantifizieren und zu verhindern.

Im Teil II dieser Doktorarbeit beschreiben wir die spektroskopischen Beobachtungenzweier Galaxien mit sehr starker Lyα Emission bei z ∼ 7.7. Diese Art von Galaxienwerden benutzt um die Reionisation zu untersuchen. Obwohl die zwei Galaxien sehrvielversprechend durch Beobachten in Filtern mit kleiner Bandbreite vom Boden ausdetektiert wurden, konnten wir sie durch unsere sensitiveren Beobachtungen nichtbestatigen. Dies weist auf eine hohe Vermischung der ausgewahlten Galaxien mit Galaxientieferer Rotverschiebung oder anderen Artefakten hin, speziell bei Beobachtungen beiz > 7 vom Boden aus in Filtern mit kleiner Bandbreite. Dieses Resultat hat wichtigeFolgen fur die Bestimmung des neutralen Wasserstoffs durch stark Lyα emittierendeGalaxien in der EoR bei z > 7 in vorhergehenden Studien. Es bereinigt Inkonsistenzen mitanderen Resultaten von unabhangigen Messungen bei tieferer Rotverschiebung. Zudemmodellieren wir die Veranderung der Lyα Helligkeitskurve und quantifizieren zum erstenMal den Anteil an neutralem Wasserstoff bei z ∼ 8 zu 50 − 70%. Der Vorteil unseresModells ist es, dass es die Veranderung der physikalischen Eigenschaften der Galaxien(wie die Sternproduktionsrate oder Staub), welche die Effekte der Reionisation imitierenkonnten, miteinbeziehen.

Teil III - Das Beenden der Sternproduktion in massereichen Galaxien

Bei tieferer Rotverschiebung (z ∼ 2) wirft die Bildung und zeitliche Evolution voninaktiven Galaxien immer noch Fragen auf. Viele plausible Prozesse fur das Beenden derSternproduktion werden zur Zeit diskutiert, allerdings ist der genaue Anteil dieser Prozesseunklar. Wir wissen, dass die Veranderung der Grossen der Galaxien (d.h. absolut undmit der Zeit) davon abhangt, ob eine Galaxie Sterne bildet oder nicht. Was genau derZusammenhang ist, ist unklar, aber mit grosster Wahrscheinlichkeit hangt dies mit demProzess welcher die Sternproduktion stoppt zusammen. Ebenso fuhren diese verschiedenenProzesse zum Stopp der Sternproduktion auf verschiedenen Zeitskalen.

xxii

ZUSAMMENFASSUNG LIST OF TABLES

Im Teil III dieser Doktorarbeit verwenden wir die beobachtete zeitliche Veranderungder Grossen von sehr massereichen, inaktiven Galaxien (log(m/M) > 11.0) umden Prozess und die Zeitskalen zur Beendigung der Sternproduktion zu untersuchen.Wir entwerfen ein empirisches Modell basierend auf der Masse, Sternproduktionsrateund Grossenveranderung von sternbildenden Galaxien und imitieren ein langsames undsofortiges Beenden der Sternproduktion. Das Erstere wird erreicht durch das Abschneidender Galaxie vom Gaszufluss, das Zweite durch eine Galaxienkollision. Unsere Resultatezeigen, dass der zweite Prozess die Grossen von massereichen inaktiven Galaxien am bestenbeschreibt.

Anhang - Projekte in Zusammenarbeit

Die Charakterisierung der physikalischen Eigenschaften von jungen, sternbildendenGalaxien bei hoher Rotverschiebung (z > 3) ist eine Herausforderung. Zum Einen ist dieVermischung der Galaxienproben bei sehr hoher Rotverschiebung (z > 7) mit Galaxientieferer Rotverschiebung oder anderer Artefakten erheblich (Kapitel B basierend auf Capaket al. (2013)). Zum Anderen ist es uns mit den modernen technischen Ausrustung nurmoglich das UV Licht (im Ruhesystem der Galaxien) bei diesen Rotverschiebungen zubeobachten. Dieser Teil des Spektralbereiches wird jedoch stark durch Staub absorbiert.Die Korrektur dieser Absorption ist deshalb sehr wichtig bei hoher Rotverschiebung.Ublicherweise wird die empirisch gemessene Absorptionskurve von lokalen Galaxien mitextremer Sternproduktionsrate fur die Korrektur bei hoher Rotverschiebung benutzt.Andere Absorptionskurven (zum Beispiel der kleinen Magellanschen Wolke oder derMilchstrasse ahnelnd) werden jedoch nicht ausgeschlossen. In Scoville et al. (2014)(im Kapitel C) messen wir die Absorptionskurve von 266 spektroskopisch verifiziertenGalaxien bei 2 < z < 6. Unsere Resultate zeigen, dass diese Galaxien eine sehr ahnlicheAbsorptionskurve wie die lokalen Galaxien mit extremer Sternproduktionsrate aufweisen.Diese Messung kann benutzt werden um das gemessenen Licht von Galaxien bei hoherRotverschiebung fur den Staub zu korrigieren.

Des Weiteren habe ich an Projekten mitgearbeitet, welche die Abhangigkeit vonphysikalischen Eigenschaften von der Umgebung der Galaxien untersucht.

In Scoville et al. (2013) messen wir die Galaxiendichte auf grossen Skalen auf den zweiQuadratgraden der COSMOS. Dabei verwenden wir mehr als 150’000 Galaxien bis z ∼ 3,eingeteilt in 127 verschiedene Rotverschiebungen. Wir finden ∼ 250 statistisch signifikanteStrukturen und bestatigen mit unserer grossen Anzahl von Galaxien, dass rote, inaktive,elliptische Galaxien sich mehrheitlich in Umgebungen mit einer hohen Galaxiendichtebefinden.

In Carollo et al. (2015, in prep.) untersuchen wir den Mix in der Struktur vonGalaxien so wie den Zusammenhang von Masse und Grosse der Galaxien als Funktionvon verschiedenen Galaxiendichten bei 0.2 < z < 0.8. Um dies zu erreichen, benutzen wir200 spektroskopisch verifizierte Galaxiengruppen, welche wir einteilen in Zentralgalaxien(die massereichste Galaxie in der Gruppe), Satellitengalaxien (alle anderen Galaxieneiner Gruppe) und Galaxien, welche nicht zu einer Gruppe gehoren. Unsere vorlaufigenResultate zeigen, dass der Mix in der Struktur von Galaxien (scheibenformig oderelliptisch) nicht von der Position der Galaxie in der Gruppe abhangt. Zudem verandert

xxiii

ZUSAMMENFASSUNG

sich der Zusammenhang zwischen Masse und Galaxiengrosse nicht als Funktion derGalaxiendichte.

xxiv

Background and Goals

Chapter 1

Introduction

With the newest instruments mounted on today’s largest telescopes in space and on theground, we are able to push towards the highest redshifts ever to study the universe as itwas only a few 100 million years after the Big Bang. At the same time, the photometricand spectroscopic capabilities of these instruments allow the detailed study of low redshiftgalaxies as well. All in all, in the last decades we have been able to trace galaxies overalmost 99% of the cosmic look-back time from redshift z = 0 to more than z = 10.Because we are not able to see the galaxy population evolving in real time, but focus ondifferent slices in redshift corresponding to different times in the universe, we are onlyable to study the average population of galaxies. The big difficulty then is to associateprogenitors and descendant galaxies. This leads to the so called “progenitor bias” (e.g.,van Dokkum & Franx 1996) that can result in a distortion of the true, measured evolutionof galaxy properties over cosmic time.

1.1 The Epoch of Re-ionization

Prior to 380,000 years after the Big Bang, the universe was in an ionized state, consistingof roughly 75% hydrogen (ionized, HII) and 25% helium (mostly neutral by then). As thetemperature of the universe cooled down to ∼ 4000 K, the hydrogen turned neutral (HI)and the coupling between photons and matter broke down (around z ∼ 1000). The photonsthat are emitted at the time of decoupling lead to the Cosmic Microwave Background(CMB) that can be observed today with all temperature/matter fluctuations recordedfrom that time (e.g., Larson et al. 2011, Planck Collaboration et al. 2015). This makes theCMB to one of the most valuable probes of our models for the formation of the universe.Subsequently, as the universe cooled down further, the first stars and galaxies start toform in the deep dark matter (DM) potential wells (caused by DM density fluctuations inthe early universe) through the collapse and accretion of gas (e.g., White & Rees 1978).Because of inefficient cooling of the gas, the first stars are more than 100 times moremassive than average stars today and quickly die in super-nova (SN) explosions. Thesubsequent enrichment of the gas with elements heavier than hydrogen allow for moreefficient cooling which leads to today’s populations of stars and galaxies. The EoR is

3

CHAPTER 1. Introduction

HI

HII

Pre-

over

lap

phas

e(z

> 1

0)O

verla

p ph

ase

(6 <

z <

10)

Post

-ove

rlap

phas

e(z

< 6

)ionization front(normal galaxy)

ionization front(quasar)

Damped Lyα systems(high HI density)

more ionizing photons, increased galaxy formation:→ accelerated ionization (front)

Fig. 1.1 — Illustration of the three phases of re-ionization. (i) The “pre-overlap” phase (z > 10)in which stars in galaxies start to ionize their surrounding IGM. Because the ionizing radiationof quasars is more energetic, their ionization front are much deeper compared to normal star-forming galaxies. (ii) The “overlap” phase (6 . z . 10) in which the volume fraction of ionizedhydrogen increases rapidly due to the increasing star-formation in galaxies. (iii) The “post-overlap” phase (z < 6). At this time most of the hydrogen is ionized, except for some regionsconsisting of a high density of HI or are self-shielded from the surrounding background ofionizing photons.

then introduced by the re-ionization of HI by the ultra-violet (UV) radiation of the firstcollapsed astrophysical objects around a redshift of z ∼ 15 − 30. The re-ionization ofhydrogen is measured to be completed by z ∼ 6, except in regions of dense HI cloudswhere the re-combination rate (proportional to the density squared) is high enough to keepthem neutral or they are self-shielded from the surrounding UV radiation (see Figure 1.1).Tracing the re-ionization history of the universe is therefore crucial to understand theformation of the first stars and galaxies.

The very first direct observational measurement of the (end of the) EoR comes fromspectra of high redshift quasars (see Fan et al. 2006). Quasars are galaxies with activelyaccreting super-massive Black Holes (BHs) in their centers and emit strong radiation inthe rest-frame UV part of their spectrum. The photons blue-ward of 912 A (correspondingto the ionization barrier of hydrogen, 13.6 eV) are absorbed by HI in the Inter Galactic

4

1.2. The cosmic star-formation rate density

Medium (IGM) on the line-of-sight between us and the quasar. Because of their brightness,quasars are excellent beacons to probe the HI content of the universe as a function ofredshift and thus to trace the last part of the EoR. The increased absorption of rest-frameUV radiation at z & 6 compared to lower redshifts indicates a decrease in optical depthof UV photons at z . 6 and therefore marks the end of the EoR.

More recently, with the advent of new near-infrared (near-IR) instruments capable tosearch for very high (z > 6) redshift galaxies, new doors have opened for the study ofthe EoR. The so called Lyman Break Galaxies (LBGs, e.g., Guhathakurta et al. 1990,Steidel & Hamilton 1992) and strong Lyα emitters (LAEs, e.g., Malhotra et al. 2001, Huet al. 2004) selected by ground- and space-based photometry and followed-up with modernnear-IR spectrographs are today commonly used to characterize the properties of the IGMin the EoR. Fundamentally, the Lyα photons (produced by de-excitation of hydrogen inthese systems) are scattered in regions of neutral hydrogen in the line-of-sight. Severalstudies based on LBGs find a decrease in the fraction of strong Lyα emitting galaxiesabove z ∼ 6, consistent with the measurements of the IGM optical depth form quasars(Schenker et al. 2012a, Treu et al. 2013, Caruana et al. 2014, Vanzella et al. 2014, Schenkeret al. 2014, Pentericci et al. 2014). Another way to probe the ionization state of the IGMis through the Lyα luminosity function (LF) measured from LAEs. Changes in the LyαLF above z ∼ 6 indicated the end of the re-ionization, consistent with quasar and LBGmeasurements (on large samples at z > 5, e.g., Malhotra & Rhoads 2004, Ouchi et al.2008; 2010, Hu et al. 2010).

Although from the above measurements it is clear that re-ionization happened, itsredshift dependence and the sources of ionizing radiation in the early universe are stilldebated. In the light of recent measurements, it is most likely that stellar objects(star-forming galaxies) are the dominant driver of the re-ionization under reasonableassumptions of the escape fraction of Lyα photons from the galaxy into the IGM andHI clumping factors reflecting the recombination rate of the IGM (e.g., Robertson et al.(2015) and references therein). Also, the finding of very massive and dusty galaxies at highredshifts suggest a period of very strong star-formation and emission of a large amount ofionizing radiation. On the other hand, quasars and active galactic nuclei (AGNs) couldprovide parts of the necessary UV flux to re-ionize the universe at z ∼ 6. However,the strong decline of the quasar luminosity density at high redshifts suggests they donot contribute significantly to the cosmic re-ionization of HI (Fan et al. 2001, Dijkstraet al. 2004, Meiksin 2005, Fan et al. 2006, Masters et al. 2012). Also X-rays, primarilyemitted by BH accretion, are able to ionize HI directly and also via secondary ionizationsby photo-electrons from ionized helium. Limits from observations of the unresolved softX-ray background suggest that this mechanism is subdominant (< 50%) in ionizing HI athigh redshifts (Dijkstra et al. 2004). More exotic possibilities like sterile neutrinos or DMannihilation are also being studied, but their exact contribution and existence are debated.

1.2 The cosmic star-formation rate density

The mass growth of the very first galaxies can be described within the “hierarchicalassembly paradigm” in which DM structures of small scales merge hierarchically to build

5


DataFit z < 6 (Hopkins+06)

0 2 4 6 8 10 12

!4.0

!3.5

!3.0

!2.5

!2.0

!1.5

!1.0

!0.5

redshift

log

SF

R d

ensi

ty Part IIIof this thesis

Part IIof this thesis

Fig. 1.2 — Average cosmic SFRD (in M yr−1 Mpc−3) across 13 billion years of cosmic time.The data is compiled from several works at lower redshifts (z < 6, Hopkins & Beacom 2006; andreferences therein) and very high redshifts (z > 6, Oesch et al. 2013b; and references therein).All SFRs are corrected for dust obscuration. Mentionable are three different epochs: the EoRat z > 6, the peak of cosmic SFRD at z ∼ 2 and the time after at z . 2 where most of thequiescent galaxies emerge. The solid line shows a fit to the data at z < 6 by Hopkins & Beacom(2006) using the parametrization of Cole et al. (2001).

more massive systems. At later times galaxies evolve via galaxy-galaxy mergers and theaccretion of gas leading to an “inside-out” growth (e.g., van Dokkum et al. 2010, Patel et al.2013). The total stellar mass of a galaxy is described by its past integrated star-formationrate (SFR). The SFR (correlated to the UV light from young stars) is commonly estimatedby the fitting of empirical or synthetic galaxy model templates to the observed spectralenergy distributions (SED) from either photometry or spectroscopy. The determination ofSFRs is however not straight-forward as a substantial amount of star-formation might beobscured by dust. Thus IR observations are needed to catch the re-emitted UV radiationat redder wavelength. Because in high redshift galaxies only the rest-frame UV to opticalpart of the spectrum can be observed with the current instrumentations, it becomes moreand more difficult to measure reliable SFRs at high redshifts (z & 3).

The cosmic (average) SFR density (SFRD, SFR per comoving volume) has been foundto evolve significantly across cosmic time (Figure 1.3 shows a compilation of SFRDmeasurements from Hopkins & Beacom (2006) and Oesch et al. (2013b)). Specifically,the SFRD increases from high redshifts down to z ∼ 2 and decreases afterwards. Thepeak at z ∼ 2 argues for a phase of strong build-up of stellar mass. Linked to this, thespecific SFR (sSFR, SFR/m), which is similar to a mass doubling for a given galaxy, isof great interest and characterizes the mass growth of galaxies. Compared to today, at

6

1.3. The main-sequence of star-forming galaxies

z ∼ 2 the sSFR is a factor of ∼ 20 higher. There is growing consensus amongst differentstudies that the sSFR is closely linked to the hierarchical growth of DM structures, i.e.,the specific DM accretion rate (e.g., Bouche et al. 2010, Lilly et al. 2013, Birrer et al. 2014)

With the current near-IR multi-slit spectrographs, we are able to study the galaxiesat the peak of the cosmic SFRD in more detail. These observations show that thesegalaxies are indeed vigorously star-forming at rates of ∼ 100 M yr−1 at masses oflog(m/M) ∼ 11. Furthermore, the majority of these systems are characterized bya large gas content, ordered rotation (rotating disks as measured by high resolutionHα kinematic maps), and large (∼ 1 kpc, i.e., roughly a spatial resolution element)star-forming clumps (e.g., Forster Schreiber et al. 2011, Tacchella et al. 2014). Similarsystems are found also at higher redshifts up to z ∼ 4 (Elmegreen et al. 2007, Hodgeet al. 2012, Guo et al. 2014), however, because Hα is out of reach at these redshiftsfor current spectrographs, it is not yet possible to constrain their kinematics and todistinguish between merging galaxies and star-forming clumps residing within the galaxy.The Atacama Large Millimeter/Submillimeter Array (ALMA), when fully operating, willbe able to trace the kinematics of these galaxies via direct measurements of CO emission(Capak et al. 2015, submitted). Simulations show that such clumpy disks can be causedby high gas accretion rates, which may be clumpy by itself and similar to a “chain” ofminor mergers (e.g., Dekel et al. 2009). Specifically, higher gas densities de-stabilize thedisks (via Jeans instabilities) and lead to a fragmentation of the disks with induced starformation. In addition to the accretion of gas, minor and major mergers play a role inthe evolution of galaxies. Major mergers (merging galaxies of roughly equal mass) havea strong impact on galaxy growth (doubling their mass) and morphology (formation ofmassive elliptical galaxies). From simulation it is expected that the size increase scaleslinearly with the stellar mass increase (e.g., Hernquist et al. 1993, Naab et al. 2009, Welkeret al. 2015). On the other hand, minor merger events (merging of galaxies with lower massratios) have little effect on the mass of the galaxies. Instead, stellar mass will be depositedon the outskirts of the main galaxy and will let it grow proportional to the square of themass increase (e.g., Naab et al. 2009, Hilz et al. 2013, Welker et al. 2015). It is howeverimportant to note that the gas increase due to mergers alone would not be able to sustainthe star-formation at intermediate and high redshifts. Because of the high sSFRs, galaxieswould consume their gas in a few 100 million years. Therefore a steady accretion of gas isnecessary to keep these systems star-forming.

This thesis will focus on two of the most important stages in the life of galaxies. Thisis the phase of early mass build-up well above the peak of cosmic SFRD as well as thedeclining part at z < 2.

1.3 The main-sequence of star-forming galaxies

In addition to the change of average SFR and stellar mass over cosmic time, tight relationsat a fixed redshift are measured as well.

The existence of a very tight relation between stellar mass and SFR, the so calledstar-forming main-sequence (MS) has been observed well up to z ∼ 2 (Brinchmann et al.2004, Noeske et al. 2007, Elbaz et al. 2007, Daddi et al. 2007, Pannella et al. 2009,

7


Fig. 1.3 — The main-sequence of star-forming galaxies at z < 0.2. The red dashed line showsthe average SFR at a given stellar mass. The dashed lines blue dashed lines show the limitscontaining 95 per cent of the galaxies at a given stellar mass. Studies suggest that this main-sequence persists up to z ∼ 2 and even higher. (Credit: Brinchmann et al. (2004))

Rodighiero et al. 2011, Whitaker et al. 2012, Rodighiero et al. 2014). It is found thatits scatter (0.3 dex) and slope (SFR ∝ mα and α close to unity) are relatively constantacross cosmic time. The origin of the scatter is not fully declared, but it is suggestedto be due to stochastic changes in the SFR, for example resulting from changing gasaccretion rates. Very recent studies using deep Herschel IR data that accounts also fordust-obscured star-formation suggest that the star-forming MS persists up to z ∼ 4 withthe similar scatter and slope as at lower redshifts (e.g., Speagle et al. 2014, Schreiber et al.2014). The mass and SFR determination at these high redshifts is however challenging,and much larger galaxy samples with deep IR observations are necessary to verify theseresults.

The amount of metals in a galaxy depends on the integrated star-formation, andtherefore a coupling between stellar mass and metallicity is expected. Such a relationhas been suggested for at least 30 years (Lequeux et al. 1979) and has been thoroughlymeasured (e.g., Tremonti et al. 2004) in local galaxies using the Sloan Digital Sky Survey(SDSS York et al. 2000). These measurement show a decrease in metallicity in low-massgalaxies and a saturation at high masses at fixed redshift and SFR. A possible explanationfor this could be the star-formation driven outflow of metals in low-mass galaxies. With theadvent of more and more powerful spectrographs (especially in the near-IR, like MOSFIREor KMOS), it has been possible to probe these relations at higher redshifts (e.g., Erb et al.2006, Maiolino et al. 2008, Mannucci et al. 2010, Yuan et al. 2013, Belli et al. 2013, Maier

8

1.4. The emergence of quiescent galaxies

et al. 2014). It has been found that the mass versus metallicity relation evolves withredshift, but at the same time, different populations of galaxies in terms of their SFR arebeing probed. This lead to the idea of a mass vs. metallicity vs. SFR fundamental planewith a positive correlation between stellar mass and metallicity and a negative correlationbetween SFR and metallicity. However, its “fundamentallity” with cosmic time is stilldebated (Mannucci et al. 2010, Belli et al. 2013, Maier et al. 2014, Zahid et al. 2014).

At least below z ∼ 2, all the above relations can be explained within a “gas regulator”paradigm, in which galaxies are thought of as gas reservoirs (e.g., Bouche et al. 2010, Daveet al. 2012, Lilly et al. 2013). In such a model, their SFR is determined by the level ofgas in the reservoir, which is set by the net gas in- and out-flow and therefore is closelyconnected to the DM halo accretion rate. Likewise, the metallicity content is set by theamount of star-formation balanced by the inflow of pristine (i.e., low metallicity) gas.

1.4 The emergence of quiescent galaxies

Not all galaxies reside on the star-forming MS. Roughly 1%− 4% of star-forming galaxieslie above the relation with significantly higher (factor 5 or more) SFRs than the averagepopulation. These so called “star-burst” galaxies may result from major merger events,however, only contribute very little (∼ 10%) to the total star-formation below z = 2. Incontrast, an increasing fraction of galaxies with cosmic time lies significantly below theMS of star-forming galaxies. These galaxies are called “quiescent” or “quenched”, becausetheir star-formation has been shut down in the past by yet not completely understoodprocesses.

Quiescent galaxies are characterized by a low sSFR (on the order of < 10−2 Gyr−1)and are selected to be a factor of 10 or more below the MS of star-forming galaxies.They lack of strong emission lines and are dominated by absorption features and largeBalmer breaks at 4000 A. Also, these galaxies are on average smaller in half-light size,more compact (in terms higher Σm = m/(2πR2

e)), and redder than their star-formingcounterparts at a given redshift and stellar mass (e.g., Blanton et al. 2003, Shen et al.2003, Baldry et al. 2004). Ideally, quiescent galaxies are selected spectroscopically by theirabsorption features and non-existence of emission lines. The observation of absorptionlines is demanding in terms of telescope time. Spectroscopic surveys need to have enoughdepth and resolution in order to observe the continuum and the absorption features. Whilebeing done routinely at low redshifts, this is challenging at higher redshifts because of thedecreasing brightness of the galaxies. Over the last couple of years, several quiescentgalaxies have been spectroscopically confirmed up to z ∼ 2 (e.g., Cimatti et al. 2008,Onodera et al. 2012, Gobat et al. 2013, Whitaker et al. 2013, Onodera et al. 2014). Thisnumber drops to a hand full above z ∼ 2 (e.g., Kriek et al. 2008, Gobat et al. 2012, Belliet al. 2014b, Krogager et al. 2014). Due to their red colors because of their quiescence,color versus color diagrams provide an alternative way to select quiescent galaxies, whichis, however less reliable than direct spectroscopic observations.

The fraction of quiescent galaxies is found to increase sharply by a factor of two ormore at all masses between 1 < z < 2 (see Figure 1.5 and Hartley et al. 2013, Muzzinet al. 2013, Ilbert et al. 2013, Brennan et al. 2015). Furthermore, the fraction of quiescent

9


SFR

den

sity

(Ilb

ert+

13)

10.0 < log M < 10.510.5 < log M < 11.011.0 < log M < 11.411.4 < log M < 12.0

SelectionIR onlyIR + i+

Ilbert+13 MF

11 10 9 8 7 6 5 4 3 20.0

0.2

0.4

0.6

0.8

1.00.3 0.5 0.7 1 1.5 2 3

redshift

age of the Universe (Gyr)

quie

scen

t fra

ctio

n

Fig. 1.4 — Fraction of quiescent galaxies as a function of stellar mass and redshift. The symbolsshow the fraction for mass complete samples in UltraVISTA (near-IR selected, filled circles)as well as including Subaru i+ selected galaxies (open triangles). The lines show the fractionsderived from the Ilbert et al. (2013) stellar mass functions (color bands). The cosmic SFRdensity is shown in gray for comparison. The quiescent fraction rises rapidly by almost a factorof two for the most massive galaxies between z = 2 and z = 1. Furthermore, there is a strongdependence of the quiescent fraction on stellar mass at all redshifts.

galaxies is found to be higher for more massive galaxies as well as for galaxies living inmore dense environment at a given redshift. It has been shown that those fractions areindependent of each other, which led to a picture in which the quenching process which islikely separable in mass and environment (e.g., Dressler 1980, Balogh et al. 2004, Baldryet al. 2006, Peng et al. 2010b, Scoville et al. 2013, Kovac et al. 2014).

The origin, emergence, and evolution of quiescent galaxies is still not understood in largeparts. There are several proposed mechanisms how galaxies can be quenched, however it ischallenging to disentangle them. The mechanisms proposed include “DM halo quenching”(e.g., Croton et al. 2006, Cattaneo et al. 2008, Woo et al. 2013; 2014, Carollo et al. 2014,Schawinski et al. 2014), accelerated quenching through major mergers and AGN feedback(e.g., Toomre & Toomre 1972, Hopkins et al. 2009b, Haas et al. 2013, Schawinski et al.2014), gravitational/morphological quenching (Martig et al. 2009, Genzel et al. 2014), and(indirectly through) violent disk instability (e.g., Bournaud et al. 2007, Mandelker et al.

10

1.5. Size evolution of quiescent and star-forming galaxies

2014, Dekel & Burkert 2014).

In more detail, DM halo quenching acts on galaxies residing in halos of masses largerthan a few ×1012 M. Above this critical mass, infalling gas is heated by the high virialtemperature and prevented from accreting onto the central galaxy to form stars (e.g.,Birnboim & Dekel (2003), Keres et al. (2005), van de Voort et al. (2011)). In addition, forgalaxies that infall into massive DM halos of other galaxies or galaxy groups/clusters (socalled satellite galaxies) other processes may apply. These include strangulation (strippingof the hot gas supply, e.g., Balogh et al. (2000)) or ram pressure stripping (removalof cold gas leading to almost immediate turn off of star-formation, e.g., Gunn & Gott(1972), Abadi et al. (1999)). Second, gas-rich major mergers may drive gas towards thecentral region of the galaxies triggering a central star-burst and thus a high central stellarmass density. During the star-burst phase, the gas will be consumed quickly or drivenout via strong winds with speeds more than 1000 km s−1. Furthermore, some studiessuggest that AGNs are triggered from the accretion of gas onto the central BHs duringthis process. Third, a rather recent idea of violent disk instabilities is based on theobservation of the gas-rich, clumpy z ∼ 2 galaxies (see above). Simulations suggest themigration of these giant star-forming clumps towards the center of the galaxies leading tothe build-up of a dense central star-forming region. Again, the induced star-formationresults in a consumption of gas and powerful outflows. Close to this is the idea ofgravitational/morphological quenching. The formation of a bulge stabilize the disk andprevents it from Jeans-fragmentation and the formation of stars.

1.5 Size evolution of quiescent and star-forming galaxies

Similar to the relation between stellar mass and SFR, there exists a relation between stellarmass and galaxy half-light size (m−R relation). In addition, measurements suggest thatthe slope as well as the scatter of this relation is not significantly changing as a function ofcosmic time, however, its normalization changes analogous to the evolution of SFR acrosscosmic time.

The m−R relation of quiescent and star-forming galaxies has been measured at localredshifts (e.g., Shen et al. 2003, Cibinel et al. 2013b). It is found that quiescent galaxiesexhibit much steeper m−R relations with logarithmic slopes of ∼ 0.5− 0.6. The relationof star-forming galaxies on the other hand show slopes of ∼ 0.2 − 0.4. Furthermore, thenormalization of the relations is very different. Quiescent galaxies are found to be muchsmaller in half-light size and more compact at a fixed stellar mass and redshift than theirstar-forming counterparts. These observational facts are also seen at higher redshifts up toz ∼ 3 by several studies (e.g., Daddi et al. 2005, Trujillo et al. 2006, Franx et al. 2008, vanDokkum et al. 2008, Carollo et al. 2013a, Bruce et al. 2014, van der Wel et al. 2014) andtherefore completing the picture in which quiescent galaxies are smaller and more compactat fixed mass at all redshifts. The difference in slope of the m − R relations, however,indicates that the size difference vanishes at high masses above log(m/M) ∼ 11.5. Thementioned studies also show that the change of the normalization of the m − R relation(i.e., the average size evolution with cosmic time) is different as well for quiescent andstar-forming galaxies. For star-forming galaxies, a size evolution proportional to (1+z)−β

11


Gro

wth

fact

orbe

twee

n z

= 1

and

z =

0.2

Half Light Radius (kpc)

Gro

wth

fact

orbe

twee

n z

= 1

and

z =

0.2

Fig. 1.5 — What causes the size evolution of quiescent galaxies? Growth factors betweenz = 1 and z = 0.2 of the number density of quenched early-type galaxies at a fixed size fortwo different bins in stellar mass. The growth factor is defined as the ratio between the sizefunctions at 0.2 < z < 0.4 and 0.8 < z < 01.0. The gray bars show 1σ, 2σ, and 3σ confidenceintervals. The growth in population of small galaxies is very small below z ∼ 1 compared to thegrowth of the population of larger galaxies. This can be interpreted with additional galaxiesat later times stemming from larger star-forming galaxies (progenitor bias). (Credit: Carolloet al. (2013a))

with 0.6 . β . 1.0 is measured. Quiescent galaxies on the other hand show in general asteeper evolution with β & 1.

The physical explanation of the observed size evolution of quiescent galaxies is currentlyunder debate. As quiescent galaxies are (by definition) not forming stars, they shouldneither increase their stellar mass nor change their size. Letting the individual quiescentgalaxies grow by minor mergers is an attractive solution for this problem, because thisprocess increases the galaxy size roughly proportional to m2 without changing much themass (e.g., Naab et al. 2009, Hopkins et al. 2010a, Feldmann et al. 2010, Oser et al.2012, Belli et al. 2013). However, progenitor bias is expected to play a role as well (e.g.,Carollo et al. 2013a, Belli et al. 2014b). The idea is that newly quenched, large quiescentgalaxies (stemming from younger, therefore larger star-forming galaxies) that are addedto the quiescent populations are responsible for the growth of the average size with cosmictime. This idea is supported by recent spectroscopic measurements that show that largerquiescent galaxies are younger compared to smaller ones. Which of the two mechanismsis the dominant in shaping the size evolution of quiescent galaxies at which stellar massand redshift is still debated and the topic of the current research.

12

1.6. Goals

1.6 Goals

This thesis targets two very different epoch during the evolution of our universe. Theseare the EoR at z > 6 and the time after the peak of cosmic SFRD at z . 2. In particular,we aim to answer the following questions:

1. How does the ionization state of the universe during the EoR change and how muchdo galaxies contribute to the re-ionization?

2. How are galaxies quenched as a function of stellar mass and redshift?

3. How does quenching mechanisms relate to the size evolution of quiescent galaxies?

We tackle the first set of question in Part II of this thesis. Questions 2 and 3 areinvestigated in Part III.

13


14

Part I

The Tools: Surveys andInstrumentations

Chapter 2

Surveys and Instrumentations

2.1 The Cosmic Evolution Survey

The Cosmic Evolution Survey (COSMOS, Scoville et al. 2007b) is – with 2 deg2 (roughly16 times larger than the full moon) – one of the largest contiguous imaged field on sky. It isan equatorial field centered at α(2000) = 10h00m28.6s and δ(2000) = +0212′28.6′′. Thelocation of this field was chosen to exhibit low and uniform galactic extinction (E(B−V ) ∼0.02 mag) as well as a low galactic infrared IR background, which is essential for IRbased observations. The large area of COSMOS has clear advantages over other fields.It allows the search for and study of rare galaxy species (like AGNs and quasars at highredshifts or to most luminous and massive galaxies), galaxy groups and clusters, as wellas the characterization of the large-scale structure (LSS) and its redshift dependence.Furthermore, the continuous extension of the COSMOS data to unprecedented wavelengthregions as well as its spectroscopic follow-up observations make this field essential fortoday’s and future astrophysics.

In particular, the main science goals of COSMOS include (see also Scoville et al. 2007b)

• the assembly of the first galaxies and galaxy clusters at very high redshifts z > 4;

• the reconstruction of DM distributions through weak and strong gravitational lensingat z < 1.5 and clustering on scales up to 2× 1014 M;

• the evolution of galaxy morphology, merger rates, AGNs, intergalactic gas, andstar-formation as a function of LSS and redshift up to z ∼ 6;

• the evolution of AGNs and the dependence of black hole growth on galaxymorphology and environment.

In the following we outline very briefly ground- and space-based observations carriedout on the COSMOS field. For further reading we refer to the original COSMOS papers(Scoville et al. 2007b;a, Capak et al. 2007, Sanders et al. 2007, Lilly et al. 2007).

17

CHAPTER 2. Surveys and Instrumentations

2.1.1 Space based observations

The COSMOS field is entirely imaged by the Hubble Space Telescope’s (HST) AdvancedCamera for Surveys (ACS) in the optical at λ ∼ 8037.2 A (F814W filter pass-band). Itshigh resolution allows detailed morphological classification of galaxies in the observedoptical light. Also, COSMOS is covered in the far-UV and near-UV by the GalaxyEvolution Explorer (GALEX), in X-rays by the Chandra X-ray telescope and the X-rayMulti-Mirror (XMM-Newton) telescope as well as in the near-IR to far-IR by the SpitzerSpace Telescope (SST) and Herschel.

For a comprehensive understanding of galaxy evolution, an unbiased sample selectionis crucial. While the ACS/F814W broad-band filter is beneficial for the selection of lowredshift (z < 1) galaxies, it may miss them at higher redshifts as well as dusty, old, andquiescent galaxies. The Cosmic Assembly Near-infrared Deep Extragalactic Legacy Survey(CANDELS, Grogin et al. 2011) field that covers the central ∼ 3% of COSMOS offers avaluable extension of the wavelength range to the near-IR (1.6 µm, the F160W filter ofthe Wide Field Camera 3, WFC3, on HST) for the selection and morphological study ofhigh-z and dusty/quiescent galaxies.

2.1.2 Ground based observations

The COSMOS field is imaged at wavelengths from UV to IR by a number of groundbased instruments. The most important ground based facilities covering the UV up to IRinclude: the Very Large Telescope (VLT), Subaru, the Kitt Peak National Observatory(KPNO), the Cerro Tololo Inter-American Observatory (CTIO), the Canada FranceHawaii Telescope (CFHT), the United Kingdom Infrared Telescope (UKIRT) as well asthe Magellan telescope in Chile. Furthermore, several observations in the millimeter toradio wavelength regime are carried out, including: the Institute de RadioastronomieMillimetrique (IRAM) telescope, the Very Large Array (VLA), the Caltech Sub-millimeterObservatory (CSO), the Atacama Sub-millimeter Telescope Experiment (ASTE), theAtacama Pathfinder Experiment (APEX), and the James Clerk Maxwell Telescope(JCMT). The recent completion of the Atacama Large Millimeter/Sub-millimeter Array(ALMA) offers high resolution observations between 30 and 950 GHz and opens the doorfor observations of dust and star-formation in very high redshift galaxies via, e.g., the COand [CII] emission lines.

The extension of ground based observations to the near-IR (Y , J , H, and Ks bandsof the UltraVISTA survey, McCracken et al. 2012) covering essentially all of COSMOShas revolutionized the galaxy selection and has led to more reliable stellar mass andluminosity functions including high-z as well as dusty and quiescent galaxies at lowerredshifts. Because of the relatively good seeing and quality of these data (PSF FWHM∼ 0.8′′), a study of morphology in the observed near-IR is feasible for the brightest andmost massive galaxies if calibrated to space based data (see Chapter 4).

18

2.2. Spectroscopy at Keck: DEIMOS and MOSFIRE

2.1.3 Spectroscopic follow-up: zCOSMOS

zCOSMOS (see Lilly et al. 2007; 2009) is the spectroscopic follow-up of COSMOS (startedin 2005) using the Visible Multi-object Spectrograph (VIMOS Le Fevre et al. 2003) onthe VLT. The survey is designed to characterize the environments of COSMOS galaxiesfrom the 100 kpc scales of galaxy groups up to the 100 Mpc scale of the cosmic web.It is sensitive to mostly star-forming galaxies but also quiescent galaxies and AGNs. Itconsists of two parts containing in total ∼ 30, 000 spectroscopically confirmed galaxies intwo redshift bins at 0.1 < z < 1.2 and 1.5 < z < 2.5.

The low redshift part (“zCOSMOS-bright”) covers ∼ 20, 000 galaxies in an area of1.7 deg2. The sample is primarily I-band magnitude limited at IAB = 22.5 mag. Theobjects are carefully checked on their validity by a comparison of the selected galaxies onindependent HST and ground based imaging such that bright stars and spurious objectsare rejected. The spectroscopic setting covers a wavelength range between 5550− 9650 Ain order to cover strong spectral features near the 4000 A break. Thus, the R ∼ 600MR grism at a 1 h integration is used for observations leading to a velocity accuracy ofbetter than 100 km s−1. Note that this observational set-up is comparable to independentspectroscopic measurement at z ∼ 0.1 in terms of selection, sampling, success rate, andvelocity accuracy, making it ideal for the connection to these low-z observations. Thesuccess rate1 is better than 80%, up to 90% at 0.5 < z < 0.8.

The high redshift part (“zCOSMOS-deep”) covers ∼ 10, 000 galaxies in the central1 deg2 of COSMOS. The galaxies are primarily selected by a combination of the BzKcriteria (Daddi et al. 2004) and the ultraviolet UGR “BX” and “BM” selection (Steidelet al. 2004). Also, X-ray and sub-millimeter detected galaxies are added to the selection.The combination of these two selection methods ensures a fairly complete sample ofstar-forming galaxies in terms of dust obscuration, star-formation and stellar mass. Inaddition to this, a magnitude cut at BAB = 25 mag is introduced to ensure the feasibilityof detecting UV emission features in the spectra. The spectroscopic set-up is 4 − 5 h ofintegrations using the R ∼ 200 LR-blue grism with a wavelength range at 3600− 6800 A.The success rate is around 60%.

2.2 Spectroscopy at Keck: DEIMOS and MOSFIRE

Spectroscopy is important for confirming redshifts and to measure various physicalquantities as nuclear activity, star-formation rates, metallicity, and stellar/gas dynamics.In this section, we describe two frequently used multi-object spectrographs working inthe optical and near-IR wavelength range, both mounted on the Keck I and II telescopes,respectively, on Mauna Kea in Hawaii. Furthermore, these instruments build the basis forthe spectroscopic observations throughout this work.

The Deep Imaging Multi-object Spectrograph (DEIMOS, Faber et al. 2003), first lightin 2002, offers the optical wavelength range between 4100 A and 11, 000 A with userdefined slit widths and slit lengths of up to 16.3′. Four different blocking filters areavailable (GG400, GG455, GG495, and OG550) with spectral lengths of 5300, 3840, 2530,

1The success rate is the fraction of observed galaxies for which redshifts are obtained.

19

CHAPTER 2. Surveys and Instrumentations

and 2630 A providing resolution up to R ∼ 6000. DEIMOS (allowing up to 130 slitletsper mask) is advantageous for redshift confirmation of galaxies up to z ∼ 7 by targetingLyα emission. Furthermore, it is used for UV-absorption line studies up to z ∼ 6 andobservation of Hα at low redshifts. Information on the data reduction pipeline can befound at http://www2.keck.hawaii.edu/inst/deimos/pipeline.html.

The Multi-object Spectrometer For Infra-Red Exploration (MOSFIRE, McLean et al.2012), first light in 2013, operates in the near-IR, covering the Y , J , H, and K bands (0.97to 2.41µm). It is used at low and high redshifts: for measurements of Hαat redshifts upto z ∼ 2.5, metallicity diagnostic lines up to z ∼ 3 or Lyα for confirmation of the highestredshift candidates (up to z ∼ 11.9). In contrast to DEIMOS, MOSFIRE is equippedwith a cryogenic Configurable Slit Unit (CSU) allowing to form a mask with up to 46 slits(depending on the slit length) in less than 6 minutes during the observations. To uses theCSU efficiently, a high density of targets is necessary, as the separate slits can only beadjusted in dispersion direction and cannot be tilted each separately. The resolution inthe given wavelength range is up to R ∼ 3500 for a 0.7′′ slit width. The data reductionpipeline (to which I contributed in several ways during of my Ph.D.) can be downloadedat https://mosfire-datareductionpipeline.github.io/MosfireDRP/.

20

http://www2.keck.hawaii.edu/inst/deimos/pipeline.html

https://mosfire-datareductionpipeline.github.io/MosfireDRP/

Part II

The Early Universe:Star-formation and Re-ionization

Chapter 3

Spectroscopic observations of LAEs atz ∼ 7.7 & implications on re-ionization

The content of this chapter is based onA. L. Faisst, P. Capak, C. M. Carollo, et al. (2014)

3.1 Summary

We present spectroscopic follow-up observations on two bright Lyα emitter (LAE)candidates originally found by Krug et al. (2012) at a redshift of z ∼ 7.7 using theMulti-Object Spectrometer for Infra-Red Exploration (MOSFIRE) at Keck. We rule outany line emission at the > 5σ level for both objects, putting on solid ground a previousnull result for one of the objects. The limits inferred from the non-detections rule outthe previous claim of no or even reversed evolution between 5.7 < z < 7.7 in the Lyαluminosity function (LF) and suggest a drop in the Lyα luminosity function consistentwith that seen in Lyman Break galaxy (LBG) samples. We model the redshift evolutionof the LAE LF using the LBG UV continuum LF and the observed rest-frame equivalentwidth distribution. From the comparison of our empirical model with the observed LAEdistribution, we estimate lower limits of the neutral hydrogen fraction to be 50 - 70%at z ∼ 7.7. Together with this, we find a strong evolution in the Lyα optical depthcharacterized by (1 + z)2.2±0.5 beyond z = 6 indicative of a strong evolution of the IGM.Finally, we extrapolate the LAE LF to z ∼ 9 using our model and show that it is unlikelythat large area surveys like UltraVISTA or Euclid pick up LAEs at this redshift assumingthe current depths and area.

3.2 Introduction

Understanding when and how the universe was re-ionized is fundamental to ourunderstanding of how galaxies and large scale structure form and evolve and is sensitiveto global cosmological parameters. In particular, the fraction of neutral hydrogen, xHI ,in the intergalactic medium (IGM) is closely tied to early galaxy formation because it

23

CHAPTER 3. Spectroscopic observations of LAEs at z ∼ 7.7 & implications onre-ionization

is related to the gas accretion rate onto galaxies. From current measurements it is stillunclear when re-ionization occurred and what the sources of re-ionizing radiation are.

The best of such current measurements come from cosmic microwave background(CMB) experiments and high-redshift quasar studies, with additional constraints fromLyman Break (LBG) and Lyα emitting (LAE) galaxy studies. WMAP (Larson et al.2011) and Planck (Tauber et al. 2010) place a ∼ 2− 3σ constraint on when re-ionizationoccurred, based on the optical depth to the CMB due to Thompson scattering of electrons.These data are usually fit by a quick re-ionization at z ∼ 10.5, but are also fully consistentwith a more gradual re-ionization with a tail ending at z ∼ 6 − 7 (Komatsu et al. 2011,Planck Collaboration et al. 2013). Direct measurements of the optical depth from quasarsindicate that the universe is neutral up to z ∼ 7.1, based on the highest redshift quasarsknown today (Fan et al. 2006, McGreer et al. 2011, Mortlock et al. 2011). Furthermore,ultraviolet (UV) continuum measurements of LBGs between z ∼ 7 − 10 (Bouwens et al.2011b, Bradley et al. 2012, Schenker et al. 2013, McLure et al. 2013) suggest that galaxieshave a difficult time re-ionizing the universe until later times unless the luminosity functionis unusually steep at the faint end, or the continuum escape fraction is high (Robertsonet al. 2013).

The fraction of strong Lyα emitters within LBG samples should give us a more direct,complementary, and unique measurement of xHI and therefore how quickly and when theuniverse is re-ionizing.

Fundamentally, Lyα photons are scattered in areas where the IGM contains moreneutral hydrogen, so the escape fraction of Lyα photons is proportional to the volumeof re-ionized hydrogen around the young galaxies. Hence the fraction of galaxies withstrong Lyα emission is related to the neutral fraction of the IGM (Haiman & Spaans1999, Malhotra & Rhoads 2004, Dijkstra et al. 2007, Malhotra & Rhoads 2006, Dijkstra& Wyithe 2010). However, it is important to note that this probe is also sensitive to theevolution of the interstellar medium (ISM) inside galaxies (like dust, see Bouwens et al.(2012), Finkelstein et al. (2012), Mallery et al. (2012)), so one must understand the effectsof galaxy evolution to probe the IGM.

The Lyα emission of LBG galaxies (selected using broad bands) is indicative ofre-ionization ending at z ∼ 6 − 7 and a neutral hydrogen fraction of ∼ 50% at z ∼ 7(Fontana et al. 2010, Stark et al. 2010, Pentericci et al. 2011, Ono et al. 2012a, Schenkeret al. 2012b, Caruana et al. 2013). In particular the fraction of strong Lyα emitters inLBG samples is found to rapidly drop beyond z > 6.5 over a range of ∆z & 1, a timescaleof only ∼ 200 Myrs (Stark et al. 2011, Curtis-Lake et al. 2012, Schenker et al. 2012b).

An alternative to LBG selection is the use of narrow-band (NB) filters to directly detectLAEs at specific redshifts (e.g., Malhotra et al. (2001), Hu et al. (2004) and referencestherein). This method allows one to directly map the Lyα LFs as a function of redshift,which can then be compared to the LBG UV continuum LFs to estimate the neutral IGMfraction.

An overall change in the Lyα LFs between 5.7 < z < 6.6 has been firmly establishedby large samples of spectroscopically confirmed LAEs (Ouchi et al. 2008; 2010, Hu et al.2010, Kashikawa et al. 2011, Malhotra & Rhoads 2004). But the source of this changecould be either an evolution in the IGM or a change in the internal ISM of the galaxies.

24

3.3. Observations & analysis

The evolution of the Lyα LF based on LAEs beyond z > 7 is far less clear. Apartfrom a few spectroscopically confirmed LAEs at z ∼ 7 (one spectroscopically confirmedout of two at z = 6.96 (Ota et al. 2008) and one spectroscopically confirmed out of threeat z = 7.22 (Shibuya et al. 2012a)), there are no confirmed LAEs at higher redshifts. Atotal sample of ∼ 15 candidate LAEs at z = 7.7 is known (Hibon et al. 2010, Tilvi et al.2010, Krug et al. 2012). Tilvi et al. (2010) and Krug et al. (2012) favor a non-evolutionof the Lyα LF between 5.7 < z < 7.7 (see also Hibon et al. (2011)), which is in tensionwith other narrow band searches for LAEs at z > 7 that only place limits on the numbercounts of LAEs (Sobral et al. 2009, Clement et al. 2012, Ota & Iye 2012, Matthee et al.2014). The reason for this tensions may be low-redshift interlopers and false detections inthe LAE samples. At z < 7, both of these are estimated to contribute less than 10–20%(see e.g., Ouchi et al. (2010)), at higher redshifts, these contribution are not known, yet,but are probably much higher (see Matthee et al. (2014) and this work). Spectroscopicfollow-up observations of high redshift candidate LAEs are therefore necessary to resolvethe tensions between the LAE and LBG results at z > 7 and to constrain the process ofre-ionization at higher redshifts.

In this chapter, we present Keck-I MOSFIRE spectroscopic follow-up of two z ∼ 7.7LAE candidates originally found by Krug et al. (2012). We then go on to compare theseresults to existing data at lower redshift and to an empirical model derived from the LBGUV continuum LF and observed equivalent width distribution. This allows us to placelimits on the neutral fraction of the IGM at z ∼ 8 and enables us to predict the LAE LFat z ∼ 8− 9. Magnitudes are given in the AB system and we assume a flat universe withΩm = 0.25, ΩΛ = 0.75, and H0 = 70 km s−1 Mpc−1.

3.3 Observations & analysis

3.3.1 Candidate selection by Krug et al. (2012)

The two targets of our study are among the brightest LAE candidates at z ∼ 7.7 (12.1and 8.6 × 10−18 erg/s/cm2, respectively, measured in UNB filters assuming negligiblecontinuum). These targets were initially selected and published by Krug et al. (2012) andthroughout this work we refer to these as LAE1 (brightest) and LAE2 (second brightest),respectively. Both LAEs were detected with an ultra narrow band (UNB) filter in theCOSMOS field (Scoville et al. 2007b) using NEWFIRM (Autry et al. 2003). Details of thedata reduction and selection are given in Krug et al. (2012), but we give a brief summaryof their results here. The effective surface area of the UNB survey is ∼ 760 arcmin2. TheUNB filter used for these candidates is centered at a wavelength of 1.056µm and has awidth of 8− 9A. This is a dark region of the spectra between bright night sky lines, andselects objects with Lyα emission at a redshift z ∼ 7.7. The UNB data were acquiredover a course of a year in three different sets of observations (February 2008, February &March 2009). This means transient objects with periods of < 1 year were rejected (seeKrug et al. (2012) and later in this section). The total usable observations add up to ∼100hours distributed over 32 nights, resulting in a limiting magnitude (defined as the 50%completeness limit) of 22.4 AB in the UNB filter. The area used to select these objects iscovered by a second UNB filter centered at 1.063µm with the same width as well as deep

25


Fig. 3.1 — SUBARU z+-band images centered on LAE1 (top) and LAE2 (bottom) overlaid withthe MOSFIRE slits configuration (left). Observed (center) and simulated 2D-spectra (right) areshown as well, both are binned to obtain R ∼ 1500. The wavelength range where the emission isexpected from the UNB observations is marked with red lines. For the simulation shown here,we assumed a rest-frame FWHM of 1.5A for the Lyα line (represented as truncated gaussian)and a spatial extent of 1′′. This simulation shows the clear detection of the line for both LAEs.

ground based broad-band data from Subaru in the optical (g, B, V, r, i, z) and UKIRTand Vista in the NIR (Y, J, H, K). This allows one to exclude continuum on the blue andred side of the potential Lyα emission line and should have eliminated low-z interlopers.Both of the candidates are not detected in any of the broad band filters as well as thesecond UNB filter. This results in rest-frame equivalent width lower limits of ∼ 7A and∼ 5A for LAE1 and LAE2, respectively.

3.3.2 MOSFIRE observations & data reduction

We observed the two LAE candidates (α = 10h00m46s.94, δ = +0208′48.84′′and α =10h00m20s.52, δ = +0218′50.04′′) with the MOSFIRE (McLean et al. 2012) spectrographon the Keck-I telescope on the nights of January 15 & 16, 2013. Each candidatewas observed with a separate mask created using the MOSFIRE Automatic GUI-basedMask Application (MAGMA1, version 1.1) and aligned using bright 2MASS stars. Theconditions were photometric on both nights, with an average seeing around 1.0′′. Theobservations were carried out in Y -band (9710 − 11250A) using the YJ grating anda 0.7′′ slit width resulting in a resolution R ∼ 3270. We used 180s exposures with16 Multiple Correlated Double Samples. The telescope was nodded by ±1.25′′ withrespect to the mask center position between exposures. The total integration times were46× 180s = 8280s = 2.3h for LAE1 and 40× 180s = 7200s = 2.0h for LAE2, respectively.

1http://www2.keck.hawaii.edu/inst/mosfire/magma.html

26

3.3. Observations & analysis

Before creating the mask, we verified that the 2MASS, COSMOS, and NEWFIRMastrometric systems agreed to within measurable errors (∼ 0.1′′). During the observationswe make sure that the masks were properly aligned by using either alignment stars and/orbright filler targets. In addition several bright sources with known fluxes and morphologiesfrom the zCOSMOS-bright spectroscopic survey (Lilly et al. 2009) were placed on the maskto verify slit losses (estimated to be 40-50%). We observed 12 and 4 of these galaxies inthe LAE1 and LAE2 masks, respectively. The comparison of the expected spatial positionfrom MAGMA to the final spatial position on the reduced 2D spectra indicates that thealignment was better than 0.2′′ during the observations.

We used the public MOSFIRE python data reduction pipeline2 for sky subtraction,wavelength calibration, and co-addition of the single exposures. The pipeline performsan A–B / B–A subtraction and co-adds the single exposures using a sigma-clipped noiseweighted mean after shifting them to a common pixel frame and masking bad pixels. Theatmospheric OH sky lines are used for wavelength calibration. The final 2D spectra have aspatial resolution of 0.18′′ per pixel and a spectral resolution of 1.09 A per pixel. Figure 3.1shows the final 2D spectra (degraded to R ∼ 1500) together with the slit positions on sky.We measured an RMS noise of 5− 10× 10−19erg/s/cm2 (4.4A resolution element) in the10545-10565A wavelength region, in good agreement with the estimated noise from thethe MOSFIRE exposure time calculator3, corrected for our estimated slit losses. Absoluteflux was measured using the white dwarf spectrophotometric standard star GD71. Thestandard star was observed during the same nights with identical settings and reduced inthe same way as the science exposures. We present the sensitivity curve for the MOSFIREY -band in Figure 3.2 together with the line fluxes of the two targets derived from UNBfilters. This shows that we would have clearly detected the two LAEs as it is furtherdiscussed below.

3.3.3 Tests and Simulations: Establishing our detection limits

Assuming the observed fluxes given in Krug et al. (2012) at 10560A and based on ourmeasured noise and our seeing of 1′′, we expect to detect the two sources at a signal-to-noiseof 12.4 and 8.2, respectively, with a line width of 200km/s (e.g., Hu et al. (2010)). Evenwith a seeing as bad as 2′′, the expected signal-to-noise is still 8.8 and 5.8, respectively. Toverify the SNR calculation and lack of detection, we simulated the expected 2D spectraby adding lines to the reduced 2D spectra. For these simulations we assumed the totalmeasured flux was distributed over a truncated gaussian with a rest-frame FWHM rangingfrom 0.5 − 3.0A (observed from stacked spectra it is ∼ 1.5A, e.g., Hu et al. (2010)). Forthe spatial extent we assume a gaussian with FWHM of 1′′ corresponding to our seeing.The following results of our simulation are not sensitive to the actual spatial extent. Wefind that the total flux of such a line would have to be less than ∼ 2−4×10−18 erg/s/cm2

for not to be visible in our data (for the range in rest-frame line FWHM). Vice versa,to miss LAE1 (LAE2) in our data, we would require a rest-frame FWHM of more than10A (7A). Figure 3.1 shows the simulated spectra rescaled to R = 1500 assuming theline fluxes measured by Krug et al. (2012), a line rest-frame FWHM of 1.5A, and spatial

2N. Konidaris, https://code.google.com/p/mosfire/3ETC version 2.3 by G. C. Rudie, http://www2.keck.hawaii.edu/inst/mosfire/etc.html

27

https://code.google.com/p/mosfire/

http://www2.keck.hawaii.edu/inst/mosfire/etc.html


extent of 1′′. This shows a clear expected detection of both Ly-α lines.

3.3.4 No detection of Lyα in LAE1 and LAE2

Our firm non-detection of line emission in the targeted LAEs yields an upper limit in Lyαline flux of 2 − 4 × 10−18 erg/s/cm2. We therefore rule these candidates out on a 7 and5σ level, respectively. This puts on solid ground a recent less significant non-detectionby Jiang et al. (2013) in 7.5 hours of LBT observation with the LUCI NIR spectrograph.Given these limits, the sources must either be a transient event with decay times of > 1year, very short periodic ( 1 year) with a large change in flux, or artifacts and/or noisespikes in the data. Considering transients, the most likely events with similar rates aresuper-luminous Supernovae (SSNe) or AGNs. Low redshift SNe are favorable because therest-frame NIR emission is decaying less rapidly than the optical (Tanaka et al. 2012).These events can account for the magnitude change measured in the UNB filters (Quimbyet al. 2007, Gezari et al. 2009, Miller et al. 2009). However, a simple calculation suggeststhat a z ∼ 0.3 SSN is visible for maximum ∼ 230 days (observed) including the riseof luminosity before its peak. However, Krug et al. (2012) searched for objects withvariability on these time scales and removed them, therefore we believe these are anunlikely source of contamination, although up to three such events could have happenedwithin 0.2 deg2 during the year of observations depending on IMF (Tanaka et al. 2012).Furthermore, short period AGNs can be excluded as a source of contamination, becauseof the amplitude of the variability which exceeds that in known AGNs (Vanden Berket al. 2004, Wilhite et al. 2008, Bauer et al. 2009). We thus conclude that the detectionsare most likely artifacts and noise. There are several reasons why this could happen.First of all, detections near the edges of an image can be caused by enhanced noise. Also,estimates of the limiting magnitude by using 50% completeness simulations and/or the useof inappropriate aperture sizes with respect to the seeing may lead to false detections. Inthe case of the Krug et al. (2012) candidates, the authors use 50% completeness simulationsto estimate their limiting magnitudes. Also, their candidates seem to lie systematicallyclose (∼ 3 arcmin) to the chip gaps between the four NEWFIRM arrays. Combinedwith the findings of Clement et al. (2012) and Jiang et al. (2013), who also find no realdetections, this raises significant questions about the reliability of the narrow band filtertechnique with NIR detectors for detecting LAEs at z > 7. Note that for z < 7, wherelarge spectroscopic follow-up studies of LAE candidates are possible, the fraction of low-zinterlopers and spurious objects is usually < 40%.

Whatever the reason, the non-detection of LAE1 and LAE2 in the MOSFIRE spectraplaces important limits on the LAE LF and implies strong evolution of it at z > 6 as itwill be discussed in the following section.

3.4 The evolution of the Lyα LF from z = 3.1 to z = 7.7

A large number of studies have looked at the Ly-α luminosity function at z < 7. Asummary of the surveys at z ∼ 3 − 5 is given in Table 3.1 and the mean data pointsadopted for this redshift range are shown in Figure 3.3, panel A. It can be seen that theLAE LF changes only slightly in this redshift range. Schechter functions fitted to the

28

3.4. The evolution of the Lyα LF from z = 3.1 to z = 7.7

LAE1

LAE2

10000 10500 11000

10−18

observed wavelength (A° )

Sen

siti

vit

y (

erg

s−

1 cm

−2 )

Fig. 3.2 — The Y -band 1σ sensitivity per 4.4A resolution element is shown. The measuredsensitivity is consistent with that of the exposure time calculator corrected for slit losses. Weshould be able to detect the two LAE candidates at several σ as shown by the red symbolsrepresenting their line fluxes as measured in the UNB filter by Krug et al. (2012).

data as a function of redshift result in less than 15% change in L∗ and φ∗, respectively(Ouchi et al. 2008). Note that in this and the following comparisons of LFs, we accountfor eventual differences in the cosmologies assumed by the authors. Furthermore, someauthors apply a correction to their Lyα luminosities to account for absorption of theLyα forest. This correction is debated as it is shown recently that the Lyα line profileis asymmetric at z ∼ 0 where IGM absorption is negligible. This suggests that Lyα isalready redshifted when escaping the galaxy and most probably make the above correctionfactor superfluous and result in an overestimation of the LAE luminosity (Scarlata et al.2014, submitted). The LFs presented in this paper are not corrected by this factor.

At 5 < z < 7 there are severals major studies (Taniguchi et al. 2005, Shimasaku et al.2006, Murayama et al. 2007, Ouchi et al. 2008, Hu et al. 2010, Ouchi et al. 2010, Kashikawaet al. 2011). All of these use the Subaru/Suprime-Cam camera with the NB812/NB921filters. Additional constraints come from Malhotra & Rhoads (2004) compiling a largesample of LAE surveys. The above studies are summarized in Table 3.2. These largestudies have significant disagreements in the derived luminosity functions with the variousstudies citing contamination rates, selection functions, and spectroscopic incompletenessas possible sources of disagreement. The Hu et al. (2010) study uses several widely spacedfields to rule out cosmic variance as the source of the discrepancy. In Figure 3.3 we combine

29


Ouchi+08 (z~3.7)

Ouchi+08 (z~3.1)

Gronwall+07 (z~3.1)

Ouchi+03 (z~4.9)

3 < z < 5

A

1042 1043

10−6

10−5

10−4

10−3

10−2

L(Lyα) (erg s−1)

n(>

L)

(M

pc−

3 )

Kashikawa+11 (z~5.7)

Hu+10 (z~5.7)

Ouchi+08 (z~5.7)

Murayama+07 (z~5.7)

Shimasaku+06 (z~5.7)

Malhotra+04 (z~5.7)

z ~ 5.7

B

1042 1043

Kashikawa+11 (z~6.6)

Ouchi+10 (z~6.6)

Hu+10 (z~6.6)

Malhotra+04 (z~6.6)

z ~ 6.6

C

1042 1043

Iye+06 (z=6.9)

Ota+10 (z~7.0)

Vanzella+11 (z=7.0)

Shibuya+12 (z=7.2)

Clement+12 (z~7.7)

Tilvi+10 (z~7.7)

Hibon+10 (z~7.7)

Krug+12 (z~7.7)(+ this work)

z ~ 7.0z ~ 7.7

D

1042 1043

z < 5.0 (median)

z ~ 5.7 (median)

z ~ 6.6 (median)

z ~ 7.0 (median)

z ~ 7.7 (limits)

3 < z < 8

E

1042 1043

Fig. 3.3 — Compilation of different studies measuring the Lyα LFs at z ∼ 3.1, 3.7, 4.9, 5.7, 6.6,7.0, and 7.7 (panels A through D). Black symbols denote single studies whereas colored symbolsrepresent their weighted medians. The error bars on the colored symbols show the standarddeviation on the median. The red symbols in panel D represent limits from single candidatesat z ∼ 7.7 from three different studies (see legend). These limits are combined and shown asred circles in panel E together with the median measurements at z < 7.7 from the other panels.The new limits at z ∼ 7.7 are consistent with an evolution of the bright end of the LAE LF atz > 6.

the various studies, and find that while the fits to the LAE LFs done by the differentauthors disagree, the data are consistent within errors, indicating counting statistics andfitting methods are the likely source of the discrepancy. We adopt weighted averages ofthe data points for the redshifts 3.1 < z < 4.9, z ∼ 5.7, and z ∼ 6.6 as indicated bythe colored symbols in Figure 3.3 panels A through C. In panel D we also show LAEdetections by Iye et al. (2006), Ota et al. (2010), Vanzella et al. (2011)4, and Shibuyaet al. (2012b) at z ∼ 7 with their weighted averages. We note that there are differencesin the normalization of the above studies which are likely linked to sample incompletenessand contamination (estimated to be less than 20% for these studies). These uncertaintiesare captured in the individual error bars, which we take into account in the final error barsof the weighted averages. At z ∼ 7.7, we combine the candidate detections from Hibonet al. (2010) and Tilvi et al. (2010) with the two remaining candidates from Krug et al.(2012) by adding up the comoving volumes of the studies. The new limits at z ∼ 7.7 areshown in Figure 3.3 panel E. The single points are shown in Figure 3.3 panel D togetherwith the limit from Clement et al. (2012) shown as gray line. Finally in Figure 3.3 panelE, we show our combined luminosity functions over the redshift range 3.1 < z < 7.7.

This clearly shows a rapid evolution in the number density of bright LAEs at 6 < z < 8.However, it is unclear whether this evolution is driven by changes in the IGM opacity, orevolution in the density of the underlying galaxy population. We will disentangle thesetwo effects in the following section.

4We note that these two galaxies are not selected by a systematic NB search. However, they could bedetected by these according to their properties (Lyα fluxes, broad-band magnitudes, and EWs) and wetherefore include them here.

30

3.5. The fraction of neutral hydrogen at z ∼ 8

3.5 The fraction of neutral hydrogen at z ∼ 8

Lyα emission is produced in young galaxies with a substantial amount of on-goingstar-formation. It is therefore the amount of UV radiation and the ISM of a galaxy whichconstrains the amount of Lyα emission. As the Lyα photons escape from the galaxy,they get scattered in areas of dense neutral hydrogen in the IGM. The amount of neutralhydrogen around galaxies sets the amount of Lyα emission that can be measured by ourtelescopes. As soon as galaxies are formed, they start to re-ionize larger and larger bubblesof neutral hydrogen around themselves and the transparency for Lyα photons is increased.By recording the amount of Lyα emission, i.e., the rest-frame equivalent width (EW0)distribution, as a function of redshift, it is therefore possible to estimate the change in thevolume fraction of neutral hydrogen, xHI , and therefore map the re-ionization process.

However, the change in the fraction of Lyα emitting galaxies also depends on thedensity of the underlying galaxy population as well as on internal (ISM) properties of thegalaxies, like star formation rate and dust content. Studies of the Lyα emission propertiesof UV-continuum selected LBGs suggest that the Lyα emission is rising with redshift ingalaxies at z = 4− 6 (Stark et al. 2010, Mallery et al. 2012, Schenker et al. 2012a), wherethe universe is thought to be fully re-ionized. In particular, Zheng et al. (2014) note thatthe EW distribution in this redshift range (4 < z < 6) is skewed to larger rest-frame EWvalues for higher redshifts. This suggests evolution of the internal properties of galaxies(e.g., dust, Bouwens et al. (2012), Finkelstein et al. (2012), Mallery et al. (2012)) enhancingthe amount of Lyα emission with increasing redshift (e.g., Treu et al. (2012)).

In order to constrain the fraction of neutral hydrogen at z ∼ 8, we have to separate theseeffects from the IGM. We therefore first model the intrinsic (i.e., without IGM absorption)Lyα LF. Later, we will compare this intrinsic LF to the observed LFs at different redshiftsand, combined with two possible implementations of the re-ionization process, constrainxHI .

3.5.1 A model of the LAE galaxy population

To separate ISM from IGM effects on the Lyα LF (see also Dijkstra & Wyithe (2012)), wefirst create an empirical model of the LAE LF based on the UV LF and the Lyα rest-frameequivalent width (EW0) distribution at z < 6, where the IGM is fully re-ionized. In brief,we assume the z = 4 − 9 UV-continuum LFs of LBGs derived by Bouwens et al. (2007;2011b) and Oesch et al. (2013b). These LFs can be well explained by assuming that theluminosity and stellar mass of a galaxy is directly related to its dark-matter halo assemblyand gas infall rate (Tacchella et al. 2013). Especially the LF at z > 7 are therefore puton more solid ground. We then convolve these UV LFs with two observed Lyα EW0

distributions of Mallery et al. (2012) (4 < z < 6) and Stark et al. (2010) (3 < z < 7) byusing a Monte Carlo sampling method to estimate a LAE LF.

We first draw random galaxies from the UV-continuum LFs. The number of galaxiesis defined by the integral of the UV luminosity function at the different redshifts. On thebright end we integrate to MUV = −30.0, above which the contribution of galaxies becomesnegligible. On the faint end, we set the integration limit to MUV = −15.0. We note,that this is ∼2 magnitudes below the Lyα luminosity which is observed at all redshifts.

31


fixed EW0

20A°

30A°

50A°

100A° (max EW)

1042 1043

10−6

10−5

10−4

10−3

10−2

3 < z < 5

(LBG z~4)

A

L(Lyα) (erg s−1)

n(>

L)

(M

pc−

3 )

1042 1043

z ~ 5.7

(LBG z~6)

B

1042 1043

z ~ 6.6

(LBG z~7)

C

1042 1043

z ~ 7.7

(LBG z~8)

D

1042 1043

z ~ 8.8

(LBG z~9)

E

Fig. 3.4 — Comparison of the measured Lyα LFs (symbols) to our empirical model combiningthe UV continuum LF with the observed rest-frame equivalent-width distribution and assumingXLyα = 1 (see text for more details). The range of LFs due to two different EW distributionsfrom literature is indicated by the shaded regions. The (intrinsic) EW distributions are thesame for all redshifts in our model. A constant EW0 of 20, 30, and 50A is shown as dot-dashed,dashed, and dotted line, respectively. The solid line denotes a fixed EW of 100A correspondingto the maximal EW with a Salpeter IMF (Mupper = 120M) and Z = 1/20 solar metallicity(Malhotra & Rhoads 2002). This comparison shows, that the LAE LF is correlated with the LyαLF derived by UV selected galaxies. We use this fact to extrapolate the LAE LF to z ∼ 8.8 asit is shown in panel E and predict an upper limit for the number of expected LAEs in differentplanned surveys. There is, however, a second order effect: The observed LAE LF is slightlychanging with respect to the model. This can be interpreted as changing properties of theIGM acting on the rest-frame equivalent width distribution of the galaxies. This can be usedto estimate the neutral hydrogen fraction of the IGM as it is outlined further in the text andFigure 3.5.

Changing MUV above this limit does not change the output of our model. This faintMUV limit however means extrapolating the observed UV-continuum LFs used from theliterature (usually going down to MUV = −18.0). So we also verified that the implicationsof our model are insensitive to changes of the faint end slopes of the UV-continuum LFsand other LF parameters between different studies (Bradley et al. 2012, McLure et al. 2013,Schenker et al. 2013). For each of the galaxies drawn from the UV-continuum LF we thenpick a random rest-frame equivalent width from the input distributions and compute thecumulative Lyα LFs. We assume no correlation between EW0 and UV-luminosity forsimplicity, although there are hints of less luminous galaxies reaching larger EW0 compareto more luminous ones (Schaerer et al. (2011) but see Nilsson et al. (2009) and Zhenget al. (2014) for a contradictory study). We also assume that every galaxy is emitting Lyα(which is then absorbed in the IGM and the EW distribution captures the ISM physics),i.e., the fraction of Lyα emission (XLyα) is 100% for our model.

Our models are shown as shaded regions in Figure 3.4, panels A through E. The pointsshow the same weighted averages as in Figure 3.3 and we find that our model is verysensitive to the assumed EW0 distribution. This is illustrated by the broad swath of theshaded region indicating the range of values obtained by the Mallery et al. (2012) and Starket al. (2010) EW0 distributions. This is not surprising, as from the comparison of the twoEW0 distributions it can be seen that Mallery et al. (2012) is missing high EW0 comparedto Stark et al. (2010) which results in a much lower Lyα LF estimate. In the following we

32


will assume the Stark et al. (2010) EW0 distribution as basis because it samples faintergalaxies which contribute to the majority of objects in our sample while Mallery et al.(2012) is restricted to UV continuum redshifts and therefore brighter galaxies. To illustratethe dependence on EW0 further, the dotted, dashed, and dash-dotted lines in Figure 3.4show constant input rest-frame equivalent widths with EW0 = 20, 30, 50A.

3.5.2 Interpreting the evolution of LAEs

We find good overall agreement between our “predicted” LAE LF and the observed valuesup to z ∼ 7. But note in Figure 3.4 the observed LAE LF moves from the bottom of thepredicted range at 3 < z < 5 to the top at z ∼ 5.7. This indicates the EW0 distributionappears to be skewing to higher values as found by Zheng et al. (2014) (compare with thelines at constant EW0 in Figure 3.4) and is likely caused by decreasing amounts of dust.In contrast, at z > 7 the LAE LF appears to return to the middle or bottom range of theshaded region predicted by our model. Assuming the (intrinsic) EW distribution does notchange, then a change in the IGM is needed to reproduce the observation. This indicatesthe IGM is becoming more opaque at z > 7, suggesting re-ionization finished at z ∼ 6−7.

3.5.3 Constraint on xHI and Lyα optical depth at z ∼ 7.7

Turning to a more qualitative analysis, we use our model to constrain the change in neutralhydrogen fraction in the IGM at z > 6.

For this, we consider two different possibilities of how we think Lyα photons getabsorbed in the IGM. The two different approaches lead to different imprints ofre-ionization in the Lyα luminosity functions. We consider (i) a “black and white” processwhere Lyα emission of a galaxy is either absorbed or not (“patchy/absorption model”)and (ii) a smooth process where the Lyα emission is attenuated by a certain degree(“smooth/attenuation model”).

The former process will decrease the number of Lyα emitting galaxies irrespective oftheir emission strength. It will lead to a “global” shift of the LAE LF. The later processwill lower the Lyα emission in all of the galaxies, preferentially removing galaxies withhigh Lyα rest-frame equivalent width. It will lead to a change in normalization and shapeof the LAE LF.

For both of these models we can constrain xHI independently. We estimate xHI forthe former by using the simulations by McQuinn et al. (2007), for the later we apply themodels by Dijkstra et al. (2011).

We note that with the current data it is not possible to (dis)prove one or the otherapproach. But we will see that both approaches will lead to the consistent results.

A patchy model of re-ionization

In this case Lyα is blocked by the neutral IGM which results in a decrease of the Lyα LFfor all luminosities. We tune our model LF to fit the observed LAE LFs at z ∼ 5.7, 6.6, and7.7 by adjusting XLyα (the total fraction of galaxies for which Lyα is not absorbed), which

33


Stark+10, XLyα = [1,1]

Stark+10, XLyα = [0.3,0.6]

Stark+10, XLyα = [0.5,0.5]

1042 1043

10−6

10−5

10−4

10−3

10−2

z ~ 5.7

(LBG z~6)

A

L(Lyα) (erg s−1)

n(>

L)

(M

pc−

3 )


Stark+10, XLyα = [0.2,0.6]

Stark+10, XLyα = [0.4,0.4]

1042 1043

z ~ 6.6

(LBG z~7)

B


Stark+10, XLyα = [0.15,0.15]

Evolution in shapeof EW distr.w.r.t. z=6 (see text)

1042 1043

z ~ 7.7

(LBG z~8)

C faint (MUV > −20.25)

bright (MUV < −20.25)

all

Sch

enk

er+

201

2

Treu

+ 2

012

Cu

rtis

−L

ake+

201

2

Ou

r li

mit

at

z =

7.7

D

4 5 6 7 80.0

0.2

0.4

0.6

0.8

redshift

XL

yα (

EW

0 >

25A

°)

Fig. 3.5 — Two methods to constrain the change in the EW distribution of Lyα emitting galaxieswith redshift by comparing our model (solid, Stark et al. (2010) EW0 distribution as basis, 100%Lyα emission) to the observe LAE LF (symbols). (i) The dashed and dotted lines show ourmodel tuned to fit the data by adjusting XLyα (dotted: overall, dashed: split in bright and faintmagnitude, see text). Panel D summarizes its evolution as a function of redshift from our work(colored symbols) compared to observations by Schenker et al. (2012b) (light gray), Curtis-Lakeet al. (2012) (dark gray) and Treu et al. (2012) (black) for galaxies with EW0 > 25A. A drop inthe fraction of Lyα emitting galaxies of a factor 4 above z = 6 is clearly visible. (ii) The dot-dashed line in panel C shows our model tuned to fit the data by skewing the EW0 distributionto lower EW0 values (i.e., adjusting its width). Both methods of modifying the EW distributionresult in consistent estimates of the lower limit of neutral hydrogen fraction at z ∼ 7.7 of 50-70%(see text).

is (at first) independent of magnitude (see Figure 3.5 panels A through C, dotted curves).We find that XLyα is almost undistinguishable between 5.7 < z < 6.6 but drops by a factorof 4 beyond z = 7 as it is shown in Figure 3.5, panel D by the filled squares. Furthermore,we follow the approach of Schenker et al. (2012b) and introduce two different values Xbright

Lyα

and XfaintLyα for simulated galaxies with MUV < −20.25AB and MUV > −20.25AB in order

to compare the fraction of Lyα emitters from our empirical model to real observations atz < 7. This is shown in Figure 3.5 panels A and B by the dashed line (we do not apply

this split at z ∼ 7.7 because of the sparse data). The values for XbrightLyα and Xfaint

Lyα for

EW0 > 25A are shown in panel D (filled and open circles, respectively). The error barsare estimated by changing the MUV cut in a range of MUV = −20.25 ± 2. Also shownare the observations by Schenker et al. (2012b) (light gray), Treu et al. (2012) (black),and Curtis-Lake et al. (2012) (dark gray) for galaxies with EW0 > 25A and the samemagnitude cut.

In general, we find a good agreement of XLyα(z) with the values observed inUV-continuum selected LBGs at z < 7. We find a significant drop of a factor of 4±1 in thefraction of Lyα emitters at z ∼ 7.7 compared to z = 6. Note that the Curtis-Lake et al.(2012) estimate of XLyα for bright galaxies is a factor of ∼ 2 higher than the estimatesfrom the other studies. Different selection and sample variance are a very likely cause forthis discrepancy. Nonetheless, their results support a strong drop of XLyα above z = 7.

This change in LF can be converted into a neutral hydrogen fraction (xHI) by using the

34


results from 186-Mpc radiative transfer simulations by McQuinn et al. (2007) as follows:their figure 4 shows the relative change of the Lyα LF as a function of neutral hydrogenfraction at z = 6.6 assuming full re-ionization at z = 6. For example xHI = 0.18, 0.38,0.53, 0.67, and 0.80 result in a re-scaling of the LF with factors of 0.76, 0.50, 0.33, 0.20, and0.05, respectively. We then assume that this re-scaling of the LF is directly proportionalto the change in the fraction of Lyα emitters, i.e., XLyα,z=7.7/XLyα,z=6 ∼ 4 (see Figure3.5, panel D, blue and red squares). Assuming the dust extinction properties at z ∼ 7.7are the same as at z = 6, we conclude that the drop in XLyα implies a neutral hydrogenfraction of at least xHI = 0.60 ± 0.07 at z ∼ 7.7. Assuming the dust content of galaxiesabove z = 6 is further decreasing and therefore extrapolating XLyα(z) from the valuesat 4 < z < 6 (see Stark et al. (2010)) implies even higher limits (xHI = 0.71 ± 0.04).Note that the small change in XLyα between z ∼ 5.7 and z ∼ 6.6 is indicative of littleneutral hydrogen. This is in line with the results by McQuinn et al. (2007) who suggestthe universe is fully ionized at these redshifts.

Note that we can estimate xHI without applying our model, by directly taking the ratioof the LAE LFs at z ∼ 5.7 and ∼ 7.7 and applying again the simulations by (McQuinnet al. 2007). This approach leads to consistent results.

Having established a lower limit on xHI , we can use the patchy model further toconstrain the Lyα optical depth. Assuming the change in XLyα above z = 6 is dueto the IGM, it can be associated to the average change of Lyα optical depth

⟨e−∆τLyα

⟩under the assumption that re-ionization is completed at z ∼ 6 (i.e., τLyα,z=5.7 = 0 and∆τLyα(z) = τLyα(z) − τLyα,z=5.7). Note, that this approach is identical to Treu et al.(2012) and we can set XLyα(z)/XLyα,z=5.7 ≡ εp(z), where εp is defined as in Treu et al.(2012) and εp,z=6 = 1 by construction. From Figure 3.5, panel D we find εp = 0.8±0.2 forz ∼ 6.6 (blue and green squares) and εp = 0.25±0.05 for z ∼ 7.7 (blue and red squares),respectively. Our z ∼ 6.6 (z ∼ 7.7) value is consistent with the z ∼ 7 (z ∼ 8) valueof 0.66±0.16 (< 0.28) found by Treu et al. (2012) (Treu et al. 2013) within errors. Wethen compute the Lyα optical depth by equating εp(z) =

⟨e−∆τLyα(z)

⟩. The final result

of ∆τLyα(z) w.r.t. z ∼ 6 is shown in Figure 3.6. Our limit at z ∼ 7.7 is important toconstraint ∆τLyα(z) as the values at z ∼ 6 and 7 are almost indistinguishable. The overallchange in optical depth as a function of redshift can be expressed by ∆τLyα(z) ∝ (1 + z)α

with α = 2.2± 0.5. Note that this exponent is a lower limit because of the upper limit inthe LAE LF at z ∼ 7.7. We find an increase in optical depth of at least 1.3 between z = 6and z ∼ 8. Our best fit model is fully consistent with the Gunn-Peterson optical depthmeasurements in quasars (Goto et al. 2011, Fan et al. 2006), however the functional formsof the estimates lead to different exponents (see Figure 3.6).

A smooth model of re-ionization

In this case there is no global scaling of the LF as before, however a steepening of theLF may occur because the EW0 distribution gets skewed to lower EW0 as the redshiftincreases beyond z = 6 (see also Zheng et al. (2014)). We represent the Stark et al. EWdistribution in the same manner as Treu et al. (2012) by using a gaussian truncated atnegative values. In contrast to the case outlined before, we now change the width of theEW0 distribution (similar to the “smooth model” in Treu et al. (2012)). As in the case

35


above, we have to take the difference in evolution between z = 6 and z = 7.7 (assumingthe IGM is fully re-ionized at z = 6). We therefore start directly with the z = 6 EW0

distribution (see Figure 3.5, panel A, dotted curve) and tune it to fit the z ∼ 7.7 limitsby changing its width (dashed-dotted line in Figure 3.5, panel C). From the final EW0

distribution at z ∼ 7.7, we compute the cumulative fraction P (> EW0) which has nowchanged w.r.t. z = 6 as we have adjusted the width of the EW0 distribution. Thisfractions can be converted into xHI by using the models by Dijkstra et al. (2011) (usingsemi-numerical simulations by Mesinger et al. (2011)) combining galactic outflow modelsand large-scale semi-numeric simulations of reionization. From our final EW0 distributionfitting the limits at z ∼ 7.7 we find P (> 100A) = 0.02 ± 0.01, P (> 75A) = 0.07 ± 0.02,and P (> 50A) = 0.20 ± 0.05 which translates, by adopting figure 5 in Pentericci et al.(2011), into upper limit neutral hydrogen fractions of xHI = 0.7 ± 0.1, 0.6 ± 0.1, and0.5±0.2, respectively. Note that xHI is more difficult to estimate for smaller EW0 cuts asP (> EW0) approaches unity for all xHI by construction (Pentericci et al. 2011). Takingthis into account, the limits we find with our second approach are consistent with theresults above.

Summary of our findings

In summary, we have looked at two different ways how re-ionization can be imprinted inthe change of Lyα LF. We have considered an absorption model resulting in a global shiftof the Lyα LF and an attenuation model resulting in a skewing of the EW0 distributionand there for a steepening of the Lyα LF. Note, that both approaches can fit the observedLAE LFs within its uncertainty and we are not able to judge which of the models is right.However, a skewing of the EW distribution is likely as it seems from the observationaldata at z ∼ 5.7 and z ∼ 6.6 that the evolution of the bright end is stronger than atthe faint end of the LAE LF. In either way, we are able to constrain xHI using bothapproaches, resulting in lower limits for the neutral hydrogen fraction between xHI = 0.53and xHI = 0.70 at z ∼ 7.7.

Finally, we stress that our results are based on the assumption that all changes in XLyα

and the EW0 distribution are caused by a change in the ionization state of the IGM atz > 6. However, and alternative explanation involves an increase of the escape fractionof ionizing photons and would lead to a drop in XLyα and thus an overestimation of xHI(Dijkstra et al. 2014). Without a changing ionization state of the IGM the escape fractionneeded to explain the observations is at odds with other studies (Wyithe et al. 2010,Kuhlen & Faucher-Giguere 2012, Robertson et al. 2013, Dijkstra et al. 2014). However,a mixture of changing xHI (∼ 0.2) and fesc (∼ 0.2 − 0.3) would be consistent with ourresults and direct escape fraction measurements.

3.6 Expected number detections of LAEs at z ∼ 8.8 in othersurveys

Given these results at z ∼ 7.7 it is important to push to higher redshifts to better constrainthe evolution of the LAE LF. Assuming that the LAE LF continues to trace the LBG LF

36

3.6. Expected number detections of LAEs at z ∼ 8.8 in other surveys

Treu+2012Treu+2013This work

∆τLyα(z) ~ (1 + z)α

α = 2.2

α = 11.0

5.5 6.5 7.5 8.5

0.0

0.5

1.0

1.5

2.0

redshift

∆τ

Ly

α

Fig. 3.6 — Change in Lyα optical depth with redshift with respect to z = 6 assuming theuniverse is fully re-ionized by then. Under this assumption, we use the Treu et al. (2012)formalism to find the mean change in Lyα optical depth with respect to z = 6 which we assumeto be proportional to the change in fraction of Lyα emitting galaxies. Our limit at z ∼ 8 isimportant to constraint τLyα(z) which we find to be best fit as (1 + z)α, α = 2.2± 0.5 (solid redline). The strong evolution of at least 1.3 beyond z = 6 is apparent and could be indicative ofa dramatic change in the properties of the IGM. Shown along with our best fit is the exponentfrom the best fit to the evolution of the Gunn-Peterson optical depth measured on Lyα, Lyβ,and Lyγ transitions in quasars (Goto et al. 2011, Fan et al. 2006).

at z > 8, we can put upper limits on the number of LAEs that should be found in plannedsurveys. The final UltraVISTA NB118 survey (McCracken et al. 2012, Milvang-Jensenet al. 2013) is able to search for potential LAE candidates at z ∼ 8.8 on 0.9 deg2 on skydown to 1.5× 10−17erg s−1 cm−2. Assuming this as limiting Lyα line flux and combinedwith our model from the LBG UV LF (optimistically assuming XLyα(z = 8.8) = 1) it isunlikely that this survey will find LAEs at this redshift (expected counts are 0.6 ± 0.3).Likewise, with the same assumptions, Euclid (Laureijs et al. 2011) is not expected to findLAEs at z > 8 with its spectroscopic configuration (1.1µm - 2µm, 3× 10−16erg/s/cm2 on

37


20,000deg2). On its proposed deep area (40deg2) a flux limit of at least 3×10−17erg/s/cm2

must be reached to find one LAE at z > 8. Other space-based spectroscopic surveys likeWISPs (Atek et al. 2010) or 3D-HST (Brammer et al. 2012) using the HST grism G141,current flux limits around 5× 10−17erg s−1 cm−2, and area of 600− 800 arcmin2 need tobe substantially (roughly 5 times) deeper to find LAEs at z ∼ 8.8. Very deep small areablind imaging surveys with instruments on 8-10m telescopes such as HAWK-I (7.5′× 7.5′)or MOSFIRE (6.1′× 6.1′) must reaching flux limits of 5× 10−18erg s−1 cm−2 in NB118 topick up one LAE at z ∼ 8.8 on a total of ∼ 10 pointings.

3.7 Conclusions

We have presented follow-up observations on two bright LAE candidates at z ∼ 7.7 usingMOSFIRE. We rule out any line emission at a level of several σ for both objects. The limitsinferred from these non-detections suggest a strong evolution of the LAE LF between 6 <z < 8, consistent with what is seen in LBG samples. We create an empirical model usingthe observed LBG UV continuum LFs and Lyα rest-frame equivalent width distributionsto understand the interplay between LAE and UV continuum selected galaxies. We findthat our model and the observed LAE LF follow each other, but note a secondary effectwhich is due to a change in the EW0 distribution of the galaxies as a function of redshift.From this differential evolution and assuming two different models on Lyα absorption,we find consistent lower limits on the neutral hydrogen fraction at z ∼ 7.7 of 50-70%.Furthermore, we find a strong evolution in the Lyα optical depth at z > 6 which can becharacterized by (1 + z)2.2±0.5. All in all, our results are indicative of a continuation ofstrong evolution in the IGM beyond z = 7.

38

3.7. ConclusionsT

ab.

3.1

—L

arg

eL

AE

surv

eys

at

3<z<

5

Red

shif

tp

hot.

can

did

ate

spec

.co

nfi

rmed

/ob

serv

edsp

ectr

.fr

act

iona

Lim

its

[AB

]A

reab

Ref

3.1

356

41/-

12%

25.3

(NB

503)

5×

0.2

Ou

chi

etal.

(2008)

3.1

160

-0%

25.4

(NB

5000)

1×

0.2

8G

ronw

all

etal.

(2007)

3.7

101

26/-

26%

24.7

(NB

570)

5×

0.2

Ou

chi

etal.

(2008)

4.9

87

-0%

26.0

(NB

711)

1×

0.1

7O

uch

iet

al.

(2003)

aF

ract

ion

of

spec

trosc

op

ically

con

firm

edgala

xie

su

sed

inth

ean

aly

sis.

bG

iven

ind

eg2.

39


Tab.

3.2

—L

AE

surv

eys

atz∼

5.7

,6.6

,7.7

,and

8.8

Red

shif

tp

hot.

can

did

ate

spec

.co

nfi

rmed

/ob

serv

edsp

ectr

.fr

act

iona

Lim

its

[AB

]T

yp

eof

lim

it,

ap

ertu

reA

reab

Ref

5.7

89

46/66

+8c

55%

26.0

(NB

816)

5σ

,2′′

ap

ertu

re1×

0.2

5K

ash

ikaw

aet

al.

(2011)

5.7

∼140

87/140d

100%

25.3

(NB

816)

5σ

,3′′

ap

ertu

re7×

0.2

Hu

etal.

(2010)

5.7

401

17/29

4%

26.0

(NB

816)

5σ

,2′′

ap

ertu

re5×

0.2

Ou

chi

etal.

(2008)

5.7

119

-0%

25.1

(NB

816)

5σ

,2′′

ap

ertu

re1×

1.9

5M

ura

yam

aet

al.

(2007)

5.7

89

28/39

+6/24

36%

26.6

(NB

816)

3σ

,2′′

ap

ertu

re1×

0.2

Sh

imasa

ku

etal.

(2006)

5.7

56e

30/35

55%

-∼

0.7

6M

alh

otr

a&

Rh

oad

s(2

004)

6.6

207

(+58)e

16/24

(+16.2

2+

1)e

13%

26.2

(NB

921)

3σ

,2′′

ap

ertu

re5×

0.2

Ou

chi

etal.

(2010)

6.6

58

42/52

+3f

74%

26.0

(NB

921)

5σ

,2′′

ap

ertu

re1×

0.2

5K

ash

ikaw

aet

al.

(2011)

6.6

∼70

30/70

100%

25.2

(NB

912)

5σ

,3′′

ap

ertu

re7×

0.2

Hu

etal.

(2010)

6.6

61g

12/23

20%

-∼

0.8

2M

alh

otr

a&

Rh

oad

s(2

004)

7.7

40/2

0%

22.4

(UN

B1056)

50%

com

pl.,

au

toap

er.

1×

0.2

Kru

get

al.

(2012)

7.7

0-

-26.0

(NB

1060)

5σ

,∼

1′′

ap

ertu

re3×

0.0

2C

lem

ent

etal.

(2012)

7.7

70/5

0%

25.2

(NB

1060)

4σ

,1.5′′

ap

ert.

(=50%

com

pl.)

1×

0.1

Hib

on

etal.

(2010)

7.7

4-

0%

22.5

(UN

B1063)

50%

com

ple

ten

es1×

0.2

Tilvi

etal.

(2010)

8.8

13h

0/5h

0%

22.2

(NBJ

)5σ

,2′′

ap

ertu

re∼

10

Matt

hee

etal.

(2014)i

aF

ract

ion

of

spec

trosc

op

ically

con

firm

edgala

xie

su

sed

inth

ean

aly

sis.

bG

iven

ind

eg2.

c20

inad

dit

ion

toS

him

asa

ku

etal.

(2006).

dP

art

of

this

sam

ple

isb

ase

don

Hu

etal.

(2004).

eB

ase

don

Kash

ikaw

aet

al.

(2006).

f28

inad

dit

ion

toK

ash

ikaw

aet

al.

(2006)

an

dT

an

igu

chi

etal.

(2005).

gT

his

sam

ple

isco

mb

ined

from

diff

eren

tst

ud

ies.

Corr

ecti

on

sfo

rfa

lse

det

ecti

on

sare

ap

plied

toth

eL

Fs.

See

Malh

otr

a&

Rh

oad

s(2

004)

for

more

info

rmati

on

.h

Incl

ud

ing

2w

ith

Jan

dK

det

ecti

on

s.i

See

als

oS

ob

ral

etal.

(2009;

2013).

40

Part III

After the Cosmic Peak:Quenching of Star-formation inMassive Galaxies

Chapter 4

Massive Galaxies at z . 2 in COSMOS:Size Evolution and Quenching

The content of this chapter is based onA. L. Faisst, C. M. Carollo, P. Capak, et al. (in prep, 2015)

4.1 Summary

We measure the size evolution of very massive (log(m/M) > 11.4) quiescent andstar-forming galaxies since z ∼ 2 using ∼ 400 galaxies from the 2-square degree surveyfield of COSMOS/UltraVISTA. We measure accurate sizes by using a position dependentpoint spread function and a calibration based on high-resolution data from the HubbleSpace Telescope. We find that, at these masses, the size evolution of star-forming andquiescent galaxies is almost indistinguishable in terms of normalization and power-lawslope. This result is used to investigate the mechanisms that lead to a shut-down thestar-formation in these massive galaxies. To this end, we predict the size evolution ofquiescent galaxies assuming two models of suppression of star-formation; the ’waningmodel’ (cut off of gas flow on galaxy leading to its consumption over > 500 Myrs) andthe ’instantaneous-quenching model’ (instantaneous gas depletion by a merger-inducedstar-burst) motivated by the high merger rates expected in these galaxies. We find that theformer over-predicts the sizes of quiescent galaxies by up to 80% at log(m/M) > 11.0. Onthe other hand, the ’instantaneous-quenching model’ predicts the sizes of quiescent galaxiesto better than 20−30% at all masses log(m/M) > 11.0 and is therefore the favored modelfor the quenching of massive galaxies. Accounting for a possible size increase during themerging event (e.g., ∆R ∝

√∆m) brings the model in even better agreement with the

observations.

4.2 Introduction

Quiescent galaxies – here defined to be galaxies that have heavily-suppressedstar-formation rates (SFRs) relative to the star-forming “main sequence” (e.g., Noeske

43

CHAPTER 4. Massive Galaxies at z . 2 in COSMOS: Size Evolution and Quenching

et al. 2007, Daddi et al. 2007) – host about half of the mass in stars in the localUniverse (Baldry et al. 2004), and have been observed in substantial numbers as earlyas z ∼ 4. Understanding the dominant processes responsible for the shut-down of theirstar-formation (often referred to as ”quenching”), as well as the connection between theseprocesses and galaxy morphology are key for understanding the emergence and evolutionof the quiescent galaxy population.

Several physical mechanisms that lead to a suppression of star-formation in galaxieshave been proposed (e.g., Croton et al. 2006, Birnboim et al. 2007, Woo et al. 2013, Hearinet al. 2013, Carollo et al. 2013b, Woo et al. 2014, Martig et al. 2009, Genzel et al. 2014,Bournaud et al. 2007, Mandelker et al. 2014, Dekel & Burkert 2014, Feldmann et al. 2011,De Lucia et al. 2012, Birnboim & Dekel 2003, Kawata & Mulchaey 2008, Cen 2014). Mostlikely, these mechanisms act simultaneously on a galaxy, and it is therefore not easy todistinguish them observationally. However, there are differences in the timescales overwhich these mechanisms work and lead to the quiescent state of a galaxy. Therefore,constraining the timescales needed to turn off star-formation for different populations ofgalaxies is an avenue to make progress in understanding their dominant quenching process.

For the purpose of this paper, we group possible mechanisms that shut-down thestar-formation into two: (i) processes that work on more than a couple of dynamicaltimescales (∼ 500 Myr) and to which we will refer to as ”waning”, and (ii) processes thatact on less than a couple of dynamical timescales (. 500 Myrs, i.e., almost instantaneous)that we will refer to as ”instantaneous quenching”. A possible physical process for theformer might be the cut-off of gas inflow onto a star-forming galaxies, which will lead toa slow starvation of the same. The latter might be induced by a merger followed by astar-burst and a fast consumption of gas on dynamical timescales (couple of 100 Myrs).

What do we know so far about this timescale and its dependence on galaxy properties?Studies focusing on galaxies in groups and clusters at masses 9.5 < log(m/M) . 11.0in the low redshift universe suggest that the transition from star-forming to quiescencetakes on the order of less than 3 Gyrs (von der Linden et al. 2010, De Lucia et al. 2012,Mok et al. 2013, Wetzel et al. 2013, Trinh et al. 2013, Cibinel et al. 2013a, Muzzin et al.2014, Hirschmann et al. 2014, Taranu et al. 2014, Schawinski et al. 2014, Peng et al. 2015).These timescales account for the time a galaxy becomes affected by the group’s potentialto when the star-formation is fully suppressed. A recent work by Tacchella et al. (2015)investigates the galaxy-internal shut-down of star-formation a function of galacto-centricdistance in individual log(m/M) ∼ 11 galaxies at z ∼ 2. Their findings suggests thesuppression of star-formation starting at the center of galaxies and slowly progressingoutwards on timescales of 3 Gyrs. Some of these works also indicate that more massivegalaxies have to shut down their star-formation on shorter timescales – and possibly byother mechanisms – than less massive galaxies.

Besides the shut down of their star-formation, the size evolution is another thing thatis curious about quiescent galaxies

Firstly, quiescent galaxies at m < M∗ (with log(M∗/M) ∼ 10.8) are characterizedby on average smaller observed half-light radii in light (Re) compared to their star-formingcounterparts at a given stellar mass and redshift. Secondly, their average size at a constantmass has increased by a factor of ∼ 3 since z = 2 (while star-forming galaxies increased

44

4.2. Introduction

their sizes by a factor of ∼ 2), which is surprising as individual quiescent galaxies shouldnot grow by itself (Toft et al. 2007, Stockton et al. 2008, Franx et al. 2008, Buitrago et al.2008, Kriek et al. 2009, Williams et al. 2010, Newman et al. 2012, Carollo et al. 2013a,van der Wel et al. 2014, Belli et al. 2014b). In order to constrain the size growth of thequiescent galaxy population over cosmic time, various studies have pointed out to look atthe size function (the distribution of sizes at a certain epoch) as a function of cosmic time.At m < M∗, studies suggest the emergence of larger quiescent galaxies at later epochs,indicative of the continuos addition of newly quenched galaxies to the large-size-end ofthe size function (Carollo et al. 2013a, Belli et al. 2014a). Above M∗, in contrast, manyindicators favor mergers, and therefore the growth of individual quiescent galaxies, for theset-up of the size evolution (Whitaker et al. 2012, Newman et al. 2012, Oser et al. 2012,Belli et al. 2014b).

In this work, for the first time, we investigate the timescale on which massive (m > M∗)star-forming galaxies shut-down their star-formation via the study of their size evolutionat z < 2. In particular, we ask the following two questions: (i) Can we set constraints onthe quenching process in m > M∗ galaxies from the average evolution of their sizes withredshift? (ii) Vice versa, can we use directly the size evolution of such galaxies (and theirstar-forming counterparts) to set constraints on the amount and properties of mergers inthese populations?

To investigate these questions, we need reliable measurements on a sample of verymassive (m > M∗) star-forming and quiescent galaxies at high redshifts. In order to findsuch galaxies, a large area on sky imaged in the near-infrared (near-IR) is crucial. Firstly,m > M∗ galaxies at high redshift become very rare with number densities of less than 10−4

per Mpc3. Secondly, quiescent galaxies are hard to detect in optical observation becauseof their old stellar populations. Current Hubble Space Telescope (HST) based surveys(e.g., the Cosmic Assembly Near-Infrared Deep Extragalactic Legacy Survey, CANDELS;Grogin et al. 2011, Koekemoer et al. 2011) lack large area as well as sufficient ancillarydata that is necessary to constrain stellar masses and SFRs of high-z galaxies from theirspectral energy distributions (SEDs). The Cosmic Evolution Survey (COSMOS, Scovilleet al. 2007b) meets all these criteria (two square-degree area and > 30 pass-band coveragefrom UV to IR). The newly acquire near-IR data of the UltraVISTA survey on COSMOS(McCracken et al. 2012) allows the selection of star-forming and quiescent galaxies out tohigh redshifts (z ∼ 4, e.g., Ilbert et al. 2013). Unfortunately, these near-IR observationsare ground based and therefore the measurement of reliable galaxy sizes is hampered bythe low resolution of this data (in terms of point spread function, PSF). The approachis therefore to calibrate these measurements with spaced based observations – the 3%of COSMOS that are included in the CANDELS/COSMOS legacy survey and thus areimaged by the HST in the near-IR.

We use this approach to measure the average size evolution of ∼ 400 star-formingand quiescent Ultra Massive Galaxies (UMGs, log(m/M) > 11.4) in the redshift range0.2 < z < 2.5. Together with determinations of the SRF vs. mass relation as a functionof cosmic time, we subsequently use our measurements of the size evolution to constrainthe timescale and process of quenching in massive (log(m/M) > 11.0) galaxies at z < 2.

This chapter is organized as follows. In Section 4.3, we outline the different data used forthe selection (Section 4.4) of star-forming and quiescent UMGs on which we measure in the

45


following the size evolution with cosmic time. In Section 4.5, we outline the measurementof galaxy sizes in detail as well as the calibration of the same using CANDELS data. Thefinal measurements are presented in Section 4.6. In Section 4.7, we present our model topredict the size evolution of quiescent galaxies. We compare our model in Section 4.8 withobservations and discuss possible modifications. Finally, we summarize and conclude inSection 4.9.

All magnitude are given in the AB system (Gunn et al. 1986) and the stellar masses(m) are scaled to a Chabrier (2003) initial mass function (IMF). Further we assume a flatcosmology with ΩΛ = 0.7, Ωm = 0.3, and H0 = 70 km s−1 Mpc−1.

4.3 Data

4.3.1 UltraVISTA near-IR imaging data

As mentioned in the previous section, near-IR data on a large area is crucial for thestudy of massive galaxies at high redshifts. Therefore, the backbone of this work is theUltraVISTA survey carried out on the 4.1 meter Visible and Infrared Survey Telescopefor Astronomy (VISTA) located at the Paranal observatory in Chile. This survey covers1.5 deg2 of the COSMOS field in the near-infrared bands Y , J , H, and Ks. Specifically,we use the (unpublished) UltraVISTA data-release (DR) 2 imaging data. Compared toDR1, this release has an improvement in H-band by up to 1 magnitude in the ultra-deepstripes (covering roughly 50% of the field) and ∼ 0.2 magnitudes on the deep stripes.The typical exposure times per pixel are between 53 and 82 hours, leading to 5 − σsensitivities of 25.4AB, 25.1AB, 24.7AB, and 24.8AB in Y , J , H, and Ks band within 2′′

aperture. The reduction of the imaging data is similar to DR1 (see McCracken et al. 2012)and is briefly outlined in the following: the data was taken in three complete observingseasons between December 2009 and May 2012. The individual science frames are visuallyinspected to remove bad frames (e.g., due to loss of auto-guiding). Each frame is skysubtracted before stacking which leads to a very flat combined image with a very smallvariation in background flux. The combined frames have an average H-band seeing of0.75′′±0.10′′. The final photometric calibration is done by using non-saturated stars fromthe Two Micron All Sky Survey (2MASS; Skrutskie et al. (2006)) sample leading to anabsolute photometric error of less than 0.2 magnitudes.

4.3.2 Photometric redshift and stellar mass catalog

Our galaxy selection (see below) is based on the public COSMOS/UltraVISTA catalog inwhich galaxies are selected from a combined Y JHKs image (Ilbert et al. 2013). This hasadvantages compared to purely optical selected catalogs as it more sensitive to galaxieswith red colors, e.g., dusty star-forming galaxies or quiescent galaxies with old stellarpopulations. The catalog comprises photometric redshifts, stellar masses, and otherphysical quantities derived from SED fitting on > 30 pass-bands from UV to IR (PSFhomogenized) for more than 250, 000 galaxies on COSMOS (see e.g., Capak et al. 2007,Ilbert et al. 2013). The photometric redshifts in that catalog are derived using Le Phare

(Arnouts et al. 2002, Ilbert et al. 2006) employing different templates including a range of

46

4.4. The Sample

galaxy types from elliptical to young and star-forming. These redshifts have been verifiedto have a precision of σ∆z/(1+z) = 0.01 up to z = 3 by comparison to a sample of morethan ∼ 10.000 spectroscopically confirmed star-forming and quiescent galaxies. Physicalquantities (mass, SFR, etc) are fitted by Le Phare at fixed photometric redshift usinga library of synthetic composite stellar population models based on Bruzual & Charlot(2003). These models include different dust extinctions (following a Calzetti et al. (2000)dust extinction law), metallicities, and star formation histories (following exponentiallydeclining τ models). Also, emission line templates are included. The emission line fluxis derived from the observed UV light using empirical relations. All these parametershave been verified by a number of other fitting routines including ZEBRA (Feldmann et al.2008) and its upgraded version ZEBRA+ (Oesch et al. 2010b, Carollo et al. 2013a). Thetypical uncertainties in masses are on the order of 0.3 dex. All quantities are computedfor a Chabrier (2003) IMF. Furthermore, all masses are given in active stars in order tocompare it to various relations in the literature. We stress that we have verified that theresults of this paper do not change if using masses from the integral of the star-formationhistories.

4.3.3 CANDELS/COSMOS near-IR imaging data

To calibrate the sizes measured on the ground based UltraVISTA imaging data, we makeuse of the overlap between UltraVISTA and HST based CANDELS/COSMOS survey(Grogin et al. 2011, Koekemoer et al. 2011). The latter covers 0.06deg2 on sky (roughly1/25th of the total UltraVISTA field) in the WFC3/IR F160W pass-band, similar to theUltraVISTA H-band, however at a much high resolution (more than 8 times smaller PSF).We use of the latest publicly available data release of the COSMOS/F160W mosaic (byFebruary 2013) with a total exposure time of 3200s and a sensitivity of 26.9 AB (5σ for apoint source).

4.4 The Sample

In the following, we describe the selection of massive galaxies at log(m/M) > 11.4building our main galaxy sample as well as less massive galaxies (10.0 < log(m/M) <11.4) that we use for the calibration of the ground-based size measurements. Furthermore,we split this sample into quiescent and star-forming galaxies.

4.4.1 High- and low-mass galaxies

The selection of the high- and low-mass galaxy sample is based on the near-IRCOSMOS/UltraVISTA photometric catalog (as described above), which allows for theselection of dusty star-forming, and quiescent galaxies.

We select a total sample of 403 massive galaxies satisfying log(m/M) > 11.4 and0.2 < zphot < 2.5 (green hatched region in Figure 4.1). We have verified these galaxiesvisually to be real (i.e., not artifacts nor stars). The exact value of this mass limit has beenchosen to correspond to the 90% completeness limit at a H-band magnitude of 21.5 AB

47


Mass completeness

star−formingquiescent

Star−formingQuiescent

UM

Gs

Lo

w m

ass

sam

ple

0.0 0.5 1.0 1.5 2.0 2.59.0

9.5

10.0

10.5

11.0

11.5

12.0

redshift

log

M

Fig. 4.1 — Sample selection. Our sample of UMGs (log(m/M) > 11.4) at 0.2 < z < 2.5 isshown with large symbols. The sample is split into quiescent (red circles) and star-forming (bluesquares) according to their location on the rest-frame (NUV − r) vs. (r− J) diagram (see alsoFigure 4.2). Other galaxies with log(m/M) < 11.4 and H < 21.5AB are shown in gray. Thedark red line shows the 90% mass completeness for star-forming (solid) and quiescent (dashed)galaxies as described in the text. The low mass sample therefore consists of the galaxies shownin gray scale which are in the dark red hatched region.

at z < 2.5, which allows us provide reliable size measurements for these galaxies (seeSection 4.5). For the estimation of the mass completeness we have used the identicalmethod as described in Pozzetti et al. (2010). With this mass cut, we select the mostmassive observable galaxies with a number density less than 10−4 Mpc−3 and 10−5 Mpc−3

at z ∼ 0.5 and z ∼ 2. These galaxies may be the progenitors of today’s most massivegalaxies, assuming these most massive galaxies keep their ranking through cosmic time.This is verified by more sophisticated methods of progenitor selections, including theselection of galaxies at a constant galaxy number density (Marchesini et al. 2014), orusing semi-empirical models that take into account galaxy mergers (Behroozi et al. 2013).

The (mass complete) low-mass galaxy control/calibration sample is selected in a similarway to have 10.0 < log(m/M) < 11.4 and H < 21.5 AB. The mass completenesslimit at H = 21.5AB as a function of redshift is shown in Figure 4.1 by the red line(solid for star-forming and dashed for quiescent galaxies). The low mass control sample(9000 galaxies in total) is consequently selected to be above the combined completenesslimit of the star-forming and quiescent galaxies and satisfies three stellar mass bins of10.0 < log(m/M) < 10.5, 10.5 < log(m/M) < 11.0, and 11.0 < log(m/M) < 11.4 with

48

4.5. Size measurements and calibration

the corresponding redshift ranges 0.2 < z < 0.45, 0.2 < z < 0.75, and 0.2 < z < 1.25.

4.4.2 Selection of quiescent and star-forming galaxies

We split our sample into quiescent and star-forming galaxies by making use of therest-frame (NUV − r) versus (r − J) color diagnostics (see Williams et al. 2009, Ilbertet al. 2013, Carollo et al. 2013a). In Figure 4.2 we show the rest-frame (NUV − r) −(r − J) diagram for six different redshift bins with our main sample of massive galaxies.The black line in each panel divides the quiescent (upper left) from the star-forming (lowerright) galaxy population. The UMGs are shown with large symbols color coded by theirspecific star-formation rate (sSFR ≡ SFR/m, the inverse of the mass doubling timescale). All the other galaxies in the same redshift bin and H < 21.5AB are shown ingray scale. We find that the color-color diagram efficiently isolates quiescent galaxies withlog(sSFR/Gyr) ∼ −1 to −2 (depending on redshift, as expected). We have verified thatother selections of quiescent and star-forming galaxies (e.g., by sSFR) do not change theresults of this paper.

4.5 Size measurements and calibration

As we have already discussed in the introduction to this paper, we are investigating thequenching process in massive galaxies via the average size evolution of star-forming andquiescent galaxies. To this end, reliable size measurements are crucial. We denote with”size” the observed semi-major axis half-light radius, Re. While we benefit from thelarge area of the COSMOS/UltraVISTA survey to select very massive galaxies, its poorresolution and PSF hampers the accurate measurement of galaxy structure parameters.

In this section, we lead in detail through (i) the determination of a spatially varyingPSF, (ii) the basic measurement of galaxy sizes, and (iii) our 2-step size-calibrationprocedure using simulated galaxies and the HST based CANDELS imaging. Finally, weoutline how we correct for the band-shifting across redshift in our sample.

4.5.1 Determination of the spatially varying PSF

Galaxy sizes are measured by the use of GALFIT, which takes into account the effectof PSF (Peng et al. 2010a). Therefore, the understanding of the PSF size (full widthat half maximum, FWHM), shape, and spatial variation is crucial. We represent the2-dimensional PSF at a given position (x, y) by a Moffat profile (Moffat 1969):

F (x, y) =β − 1

πα2

[1 +

((x− µx)2 + (y − µy)2

α2

)]−β, (4.1)

where µx, µy, α, and β are free fitting parameters. The FWHM of a PSF in thisparametrization is given by

FWHM(α, β) = 2α√

21/β − 1. (4.2)

49


−5 −2 1

log sSFR [Gyr−1]

0.2 < z < 0.5

quiescent

star−forming

0

1

2

3

4

5

6

70.5 < z < 0.75 0.75 < z < 1

1 < z < 1.5

quiescent

star−forming

0.0 0.5 1.0 1.5 2.00

1

2

3

4

5

6

7

(r−J)rest

(NU

V−

r)re

st

1.5 < z < 2

0.0 0.5 1.0 1.5 2.0

(r−J)rest

2 < z < 2.5

0.0 0.5 1.0 1.5 2.0

(r−J)rest

Fig. 4.2 — Selection of quiescent (upper left of solid line) and star-forming (lower right of solidline) UMGs on the rest-frame (NUV − r) vs. (r− J) diagram for 6 different redshift bins. TheUMGs are color coded by their sSFR. This shows that the color-color cut efficiently separatesquiescent galaxies with log(sSFR/Gyr) ∼ −1 to −2. The gray background shows less massivegalaxies with H < 21.5AB in the same redshift bins.

This has been shown to be a good approximation for ground based PSFs and has theadvantage over a pure Gaussian as it represents better the wings of the PSF (e.g., Trujilloet al. 2001). In order to create a spatially comprehensive PSF map, we select unsaturatedstars between 16 AB and 21 AB from the HST based COSMOS/ACS IF814W -bandcatalog (Leauthaud et al. 2007). We select them according to their SExtractor stellarityparameter (larger than 0.9) and using diagnostic diagrams as color vs. color and magnitudevs. size. Furthermore, we inspect the stars visually and make sure that there are no closecompanion stars (or galaxies) visible on the ACS images.

For each of these more than 3000 stars, we extract a 10′′ × 10′′ image stamp fromthe UltraVISTA H-band mosaic on which we will fit the PSF. We notice small shiftsof the center of the stars between ACS and UltraVISTA data of a few tenths of arcseconds (likely caused by small differences in the coordinate systems, the large differences

50


in the PSF size, and differences in the resolution of the images) which we correct for.We then fit the selected stars according to the above parametrization F (x, y|µx, µy, α, β).The accuracy and robustness of the fitting method was verified by generating stars withrandom FWHM between 0.2′′ < FWHM < 1.2′′, add noise taken from real backgroundimages, and fit them in the same way as the real data. This test shows that we are ableto recover the FWHM with an accuracy of better than 0.05′′. As a last cut, we requireless than 5% difference between the model and data in the enclosed flux up to 1.5 timesthe PSF FWHM. We end up with ∼ 800 PSF models across COSMOS/UltraVISTA. ThePSFs show variations in their FWHM between 0.65′′ and 0.80′′. We assign to each galaxyan average PSF model created from the stars within 6′ and use this for GALFIT.

4.5.2 Guess-parameters for surface brightness fitting

In this section, we describe the determination of the initial values which are fed intoGALFIT. In order to have consistency between the initial values and the actual images onwhich we run GALFIT, we do not use the values given in the public COSMOS/UltraVISTAcatalog, but we re-run Source Extractor (SExtractor, version 2.5.0, Bertin & Arnouts(1996)) on the DR2 UltraVISTA H-band images. We run SExtractor with two differentvalues of the DEBLEND MINCONT for a better de-blending of galaxies next to brightergalaxies or stars. The SExtractor input parameters are tuned manually in order tooptimize the source extraction. We mask (using Weightwatcher; Marmo & Bertin 2008)each star identified on the HST based COSMOS/ACS IF814W -band images by a circlewith a maximal radius rσ at which its flux decays to the background flux level. Thismaximal radius (which depends on the magnitude of the star) is determined by fittingrσ as a function of magnitude for a couple of different stars in a broad magnituderange. Furthermore, we match our catalog with the public UltraVISTA catalog andcompare the measured magnitudes, which we find to be in excellent agreement. Finally,we extract each of our galaxies from our SExtractor catalog to use the measuredgalaxy position (X IMAGE and Y IMAGE ), magnitude (MAG AUTO), half-light radius(FLUX RADIUS ), axis ratio (ratio of A IMAGE and B IMAGE ), and position angle(THETA IMAGE ) as initial parameters for GALFIT.

4.5.3 Uncalibrated size measurements

We use GALFIT to fit single Sersic profiles (parametrized by the half-light radius Re, Sersicindex n, total magnitude Mtot, axis ratio b/a ,and position angle θ) to the observed surfacebrightness of our galaxies. As described in the previous section, we use the SExtractor

values measured on the DR2 COSMOS/UltraVISTA images as initial parameters. Forthe Sersic index, which is not known a-priori, we assume n = 2 (and let it vary between0 < n < 8 during the fitting process). The size of the image cutout on which GALFIT

is run is variable between 71 × 71 and 301 × 301 pixels. The size is set to optimize theestimate of local sky background and to minimize the running time of GALFIT and definedsuch that the cutout contains three times more sky pixels than pixels attributed to galaxydetections. Companion galaxies on the image cutout are fit simultaneously with the maingalaxy if they are brighter than 25AB in H-band. All other detections of fainter objects are

51


masked out and not taken into account in the χ2 minimization. To access the stability ofthe fits, we run GALFIT in two different configurations: In the first configuration (referredto as “VARPOS”) we let GALFIT fit the center of the galaxy within ±10 pixels of theSExtractor input. In the second configuration (referred to as “FIXPOS”) we fix thegalaxy position to its initial SExtractor value.

We select good fits (either from the FIXPOS or VARPOS run) by comparing the resultsfrom the two configurations. We require that (i) Re > 0.1 px, (ii) the fitted position differby less than

√2/2 times the PSF FWHM from the SExtractor input (iii) the Re of the

two configurations agree better than 50%, and (iv) the total magnitude does not differ bymore than 0.5 from the SExtractor total magnitude. Roughly 70% of our galaxies satisfythese criteria and are used in the following for accessing the size evolution as a functionof cosmic time.

4.5.4 Correcting for measurement biases using simulated galaxies

The measurement of galaxy structure is prone to many biases (Carollo et al. 2013a, Cibinelet al. 2013b, Cameron & Driver 2007). Small and compact galaxies are affected by the PSF(leading to an over-estimation of Re); large and extended galaxies suffer surface brightnessdimming in the outskirts (leading in under-estimation of Re). Although GALFIT does takeinto account the effects of PSF and therefore partially cures these problems but has itslimits. It is therefore important to investigate possible biases and correct for them by usingsimulated galaxies. In the following, we outline this first step in our 2-step calibrationprocess in more detail.

Simulating galaxies

We use GALFIT to create ∼1.5 million model galaxies on a grid in (Re,Mtot, n, b/a)inparameter space: 0.2 < n < 10, 15 mag < Mtot < 26 mag, 0.2 < b/a < 1, and0.5 < Re < 15 pixels (corresponding to 0.075′′ < Re < 2.250′′). The model galaxiesare subsequently convolved with a PSF, equipped with Poisson noise and added ontorealistic sky backgrounds. For the latter, we account for the fact that the sky backgroundnoise (σsky) varies across the COSMOS/UltraVISTA field by a factor 2 or more (mainlybetween the deep and ultra-deep stripes). We compute σsky automatically in rectanglesof ∼ 0.1 × 0.1 degrees across the field. For this end, we use the SExtractor catalog(see Section 4.5.2) to mask out all detections and fit σsky to the remaining non-maskedpixels by assuming a Gaussian noise distribution. In order to make sure to remove allthe light of galaxies and stars, we increase their semi-major and semi-minor axis as givenby SExtractor by a factor of 10. We verify this procedure by manually measuring σsky

at random positions. To take into account the variations in PSF and σsky we simulategalaxies in four different representations which we will be interpolated in the end. Weuse two bracketing PSFs (FWHM = 0.65′′ and 0.85′′) as well as two bracketing σsky

(5.5×10−6 and 2.0×10−5 counts/s). On each of these model galaxies we run SExtractor

and GALFIT in the same manner as for the real galaxies (as described in Section 4.5.3)to obtain (Re,Mtot, n, b/a)out. This allows us to derive a correction function and discusspossible measurement biases as outlined below.

52


Correction function

We obtain a correction function, S(Re,Mtot, n, b/a), in an identical fashion as in Carolloet al. (2013a) and we refer the reader to this paper for additional details. We construct Ssuch that it returns a 4 dimensional median correction vector (∆Re,∆Mtot,∆n,∆b/a)for each point in measured (Re,Mtot, n, b/a)meas parameter space. The mediancorrection vector is constructed as the difference between the median of the 50 closest(Re,Mtot, n, b/a)out (with respect to (Re,Mtot, n, b/a)meas) and the median of their truevalues (Re,Mtot, n, b/a)in. We obtain this correction vector for each combination of PSFand σsky. The final correction vector is then obtained by an interpolation of the grid atthe PSF and σsky attributed to the galaxy for which the correction is computed.

Because of our imposed magnitude cut of bright H = 21.5 AB, the correction in size(usually over-estimated) are on the order of less than 20%. The simulations also show thatthe detection rate of galaxies is 100% in the worst case up to half-light sizes of at least3′′ at H = 21.5 AB, corresponding to a surface brightness limit of ∼ 25.2 mag arcsec−2.This size corresponds to ∼ 25 kpc (∼ 20 kpc) at z ∼ 2 (z ∼ 0.5).

The correction function allows an assessment of detection limits and a first correctionfor measurement biases. However, the simulated galaxies are ideal cases. The overlapbetween UltraVISTA and CANDELS is ideal to do a more thorough calibration of our sizemeasurement.

4.5.5 Final calibration of size measurements using CANDELS

The second step of our calibration process consists of the comparison of ourmeasured (and corrected with S) sizes with HST based structural measurements onCOSMOS/CANDELS, which has an overlap of 3% with the central part of COSMOS.Because of the 2.5 times higher resolution and 4 times smaller PSF of the HST images, weconsider the HST based size measurements to reflect the true galaxy sizes. We first measurethe sizes of galaxies on the publicly available CANDELS F160W mosaic as these closestmatch the UltraVISTA H-band data. For this end, we use SExtractor in order to extractthe sources and to get the initial parameters for GALFIT in the same manner as describedabove for the UltraVISTA based measurements. Subsequently, we run GALFIT for theextracted sources in the two configurations FIXPOS and VARPOS thereby applying thesame selection criteria for good fits as described in Section 4.5.3. Furthermore, we applythe a correction function S as done before but with PSF and σsky matching those ofthe COSMOS/CANDELS images. In turn, we find corrections less than 5% for galaxiesat H < 21.5 AB. As a further check, we compare the size measurements to the publicavailable COSMOS/CANDELS size catalog by van der Wel et al. (2012) and find excellentagreement.

The comparison between the HST-based (Rcandels) and ground-based (Rultravista) galaxysizes and their calibration is shown in Figure Figure 4.3. Shown are galaxies with Sersicindices n < 2.5 (blue) and n > 2.5 (orange) measured on the ground-based images in twomagnitude bins at H < 21.5AB (top and bottom row). Looking at the empty histograms(showing the log-ratio of the sizes) on the right panels we see an under-estimation ofRultravista by a factor 3 and more which we find to happen preferentially for galaxies smaller

53


16 < H < 20

0.1 < n < 2.52.5 < n < 8

0.1

1

10

Rca

nd

els (

arcs

ec)

not calib. (n<2.5)not calib. (n>2.5)

calib. (n<2.5)calib. (n>2.5)

20 < H < 21.5

0.1 1 10

0.1

1

10

Rca

nd

els (

arcs

ec)

Rultravista (arcsec)

−1.0 −0.5 0.0 0.5 1.0

log Rcandels/Rultravista

Fig. 4.3 — Final calibration of sizes (in arc seconds) using the CANDELS imaging data. Shownare two different magnitude bins below H = 21.5AB. The left panels show the comparison ofsemi-major half-light radii measured on CANDELS and ground-based UltraVISTA images aftercorrection of systematic biases and calibration. Orange and blue points show galaxies with0.1 < n < 2.5 and 2.5 < n < 8.0, the gray regions show 20%, 50%, and 100% discrepancies, andthe black lines show a running median with scatter (dashed black lines). Furthermore, the dashedhorizontal (vertical) line shows the CANDELS (UltraVISTA) PSF size. The right panels showthe normalized histograms of log(Rcandels/Rultravista) for uncalibrated (empty) and calibrated(filled/hatched) UltraVISTA size measurements. Without calibrations, the UltraVISTA sizesare substantially under-estimated (tail towards large log(Rcandels/Rultravista)). This is true inparticular for compact (n > 2.5) galaxies. The histograms indicate an uncertainty for calibratedsizes of ∼50%. The gray regions show 20%, 50%, and 100% discrepancies.

than the (UltraVISTA) PSF radius (∼ 0.3′′) and with large Sersic n (i.e., compact lightdistribution). Furthermore, an over-estimation of galaxy sizes preferentially happening forlarge galaxies (Re > 2′′) with small Sersic n.

We calibrate our ground-based size measurements by constructing a calibration function

54


C(Re,Mtot, n, b/a) in the same way as described in Section 4.5.4. Going back to Figure 4.3,the measurements with applied calibration function are shown in the filled and hatchedhistograms on the right two panels (for different magnitudes and n). Furthermore, theleft panels show the 1-to-1 comparison of the size measurements with a running medianwith 1σ scatter (dashed). The comparison of the fully calibrated sizes with the HST basedsize measurements show that we are able to recover Re on UltraVISTA to an accuracy ofbetter than 50% (1σ scatter).

4.5.6 Correction for internal color gradients

Mostly negative color gradients are ubiquitously measured in star-forming galaxies upto at least z ∼ 3, whereas this effect is much less strong in quiescent galaxies (Cassataet al. 2010, Szomoru et al. 2011, Wuyts et al. 2012, Cibinel et al. 2013a, Vulcani et al.2014, Bond et al. 2011, Pastrav et al. 2013, Bond et al. 2014, Hemmati et al. 2014). Theobserved color gradients are caused by different stellar populations and dust attributed toinside-out growth of galaxies and therefore depend on galaxy age, stellar mass, redshift,and star-formation activity. Such color gradients cause the observed size to change as afunction of wavelengths. Vice versa, at a fixed observed wavelength the observed size ofgalaxies changes as a function of redshift because as the rest-frame wavelengths shifts.Not to counteract the effect of color gradients may introduce artificial effects in the sizeevolution across redshift.

Several studies have constrained this effect using observations at different wavelengthsfor different types of galaxies and stellar masses at various redshifts (e.g., Kelvin et al.2012, van der Wel et al. 2014, Lange et al. 2015). Typical gradients are on the orderof ∆ logR/∆ log λ = 0.1 − 0.3 depending on data quality, resolution, and redshift atlog(m/M) = 10 leading to variations in size of 10 − 50% over a wavelength range ofrest-frame 0.5− 1.0µm1.

To correct our size measurements for the effect of color gradients, we use theparametrization by van der Wel et al. (2014) (see their equations 1 and 2) who measure∆ logR/∆ log λ ∼ 0.1 − 0.3 between 0.2 < z < 2.0 at log(m/M) = 10.0 for theirgalaxies by comparing measured sizes on HST-based optical (F814W, 0.8µm) and near-IR(F160W, 1.6µm) images. For a star-forming galaxy at z = 0.5 (z = 1.5) of a masslog(m/M) = 11.0, the corrections are on the order of 50% (10%). For a quiescentgalaxies at the same redshift the correction is on the order of 20% (5%). We note, that thisstudy covers the same redshift range as we do here and provides good constraints on themass and redshift dependence of this effect. Therefore we think that this parametrizationof correction of color gradients is the most beneficial to use. The difference to otherparametrization is, as listed above, minor and do not change the results of this paper.

4.5.7 Verification of accuracy of size measurement

Because our measurements at log(m/M) > 11.4 are unique so far, we cannot directlycheck whether these are reasonable.

1This is essentially the rest-frame wavelength range which we are covering with the redshift range inthis study

55


spec UMGs (quiescent)gray: lower masses

UMG: this work

SF QU (A)1.0

10.0

11 10 9 8 7 6 5 4 3 2

cosmic time (Gyr)

Re

(kp

c)

UMGs(log M > 11.4)

1.0

10.0

11 10 9 8 7 6 5 4 3 2

cosmic time (Gyr)

Re

(kp

c)

11 < log M < 11.4

UltraVista low mass

SF QU

VDW+14 (late−type)VDW+14 (early−type)

MC+13 (late−type)MC+13 (early−type)

(B)

11 10 9 8 7 6 5 4 3 2

cosmic time (Gyr)

10.5 < log M < 11

UltraVista low mass

SF QU


MC+13 (late−type)MC+13 (early−type)

(C)

0.2 0.5 1 1.5 2 3

1.0

10.0

redshift

Re

(kp

c)

10 < log M < 10.5

UltraVista low mass

SF QU


(D)

0.2 0.5 1 1.5 2 3

redshift

Fig. 4.4 — Size evolution as a function of redshift of star-forming and quiescent galaxies atdifferent masses. Panel (A) – For quiescent (open, red) and star-forming (filled, red) UMGsat log(m/M) > 11.4 with fits (dashed, solid red lines). Spectroscopically confirmed quies-cent UMGs from the literature (Krogager et al. 2014, Belli et al. 2014a, Onodera et al. 2014)are shown as black circles (median in two redshift bins with scatter). Clearly, star-formingUMGs are systematically offset to larger observed sizes at all redshifts. However, this offset issmaller compared to lower masses (shown in gray from literature). Panels (B) through (D)– Comparison of our size measurements (color points, up to redshift where mass complete) withmeasurements in the literature in gray for 11.0 < log(m/M) < 11.4, 10.5 < log(m/M) < 11.0,and 10.0 < log(m/M) < 10.5, respectively. The lines (dashed: quiescent, solid: star-forming)are fits to the size evolution by van der Wel et al. (2014). The filled (open) symbols showmeasurements of star-forming (quiescent) galaxies from Carollo et al. (2013a) (diamonds) andvan der Wel et al. (2014) (triangles). The good agreement to our measurements shows that ourmeasurement are not biased.

In the following, we use our (fully calibrated and mass complete) low-mass controlsamples at 10.0 < log(m/M) < 11.4 (see Section 4.4.1) to investigate possible systematics

56

4.6. Results: Size evolution of very massive galaxies

in our size measurement.

In panels B through D of Figure 4.4 we compare our measured size evolution of quiescent(open, color) and star-forming (filled, color) galaxies to measurements taken from theliterature (gray lines and symbols; van der Wel et al. 2014, Carollo et al. 2013a). Thelatter are based on high-resolution HST imaging and corrected for color gradients in thesame way as we do here. We find a very good agreement with our measurements.

In panel A we compare our final size evolution at log(m/M) > 11.4 to spectroscopicallyconfirmed quiescent galaxies at the same stellar mass in two redshift bins from theliterature as black circles (open black points; Krogager et al. 2014, Belli et al. 2014a,Onodera et al. 2014). These galaxies reside well within the 1 − 2σ scatter of ourmeasurements (indicated by the thin error bar), although at the lower end. This canbe explained the higher success rate of spectroscopic surveys for compact galaxies withhigh surface brightness.

Concluding, we do not expect any severe systematic biases of our measurements.

4.6 Results: Size evolution of very massive galaxies

4.6.1 The stellar mass vs. size relation

The relation between stellar mass and size (MR relation) has been measured so far onstatistically large samples at log(m/M) < 11.0. Our measurement on a large sampleof galaxies at log(m/M) > 11.4 enable us for the first time to provide an additionaldata point at high masses. In Figure 4.5, we show the MR relation in three redshiftbins measured over two orders of magnitudes in stellar mass. Shown are our data atlog(m/M) > 11.4 (large filled dots) for quiescent (red) and star-forming (blue) galaxiesas well as measurement at lower masses. The latter include the 3D-HST survey (cloudof thin points van der Wel et al. 2014), spectroscopically confirmed quiescent galaxies atz > 1 (crosses, asterisks, and pluses Krogager et al. 2014, Belli et al. 2014a, Onodera et al.2014), and galaxies at z < 0.1 with measurements in g-band from the Galaxy and MassAssembly survey (small points with error bars Driver et al. 2011, Lange et al. 2015). Thelarge blue and red symbols show the median size of star-forming and quiescent galaxies indifferent stellar mass bins. The lines show the corresponding log-linear fits (Re(m) ∝ mα,see Table 4.2) to the medians with errors from bootstrapping.

The MR relation of quiescent galaxies is much steeper than for star-forming galaxies.The average sizes of quiescent and star-forming galaxies are comparable at log(m/M) ∼11.5 independent of redshift. The logarithmic slope of the MR relation (〈αqu〉 ∼ 0.6 forquiescent and 〈αsf〉 ∼ 0.2 for star-forming galaxies) does not evolve significantly withcosmic time. Also, it is very consistent with the measurements in the local universe(z < 0.1; Shen et al. 2003, Lange et al. 2015) finding values between α ∼ 0.35− 0.60 andα ∼ 0.15− 0.25 for quiescent and star-forming galaxies, respectively (see also Figure 4.5).We note that van der Wel et al. (2014) finds consistent logarithmic slopes (α ∼ 0.2, nochange with redshift) for star-forming galaxies, however slightly higher values for quiescentgalaxies (α ∼ 0.75, no evolution with redshift, where we find 0.60). The reason for thisslight discrepancy might be that their study misses the very massive quiescent galaxies

57


at log(m/M) > 11.5. Removing these and re-fitting the MR relation of the remaininggalaxies indeed results in slightly steeper slopes.

The constant slope of the MR relation is indicative of a constant relation between thegrowth of galaxies in size (e.g., due to accretion) and stellar mass over cosmic time

0.5 < z < 1.0

This workall 3D−HSTLange+14 (z=0)

α = 0.32 ± 0.1 (star−forming)α = 0.55 ± 0.07 (quiescent)

10.0 10.5 11.0 11.5 12.0

1.0

10.0

log M

Re

(kp

c)

1.0 < z < 1.5

Onodera+14Krogager+14Belli+14


10.0 10.5 11.0 11.5 12.0

log M

1.5 < z < 2.0 median relation


10.0 10.5 11.0 11.5 12.0

log M

Fig. 4.5 — Mass versus size relation for quiescent (red) and star-forming (blue) galaxies forthe combined sample of 3D-HST at log(m/M) . 11.5 (thin dots) and our sample of massivegalaxies with log(m/M) > 11.4 (large points). Also included are spectroscopically confirmedquiescent galaxies from Onodera et al. (2014; pluses), Krogager et al. (2014; crosses), and Belliet al. (2014a; asterisks) at various masses as well as z < 0.1 galaxies from Lange et al. (2015;small dots with error bars). The medians in mass bins (large points with error bars) are fitlinearly in log-space for three redshift bins using a least-square method (red and blue pointswith error bars indicating error on the median). The logarithmic slope α (Re ∝ mα) is indicatedfor star-forming and quiescent galaxies.

4.6.2 Size evolution and indication of fast quenching of massive galaxies

In Figure 4.4 (panel A) we show the final median size evolution with cosmic time of ourmassive log(m/M) > 11.4 quiescent (red, open) and star-forming (red, filled) galaxies.These are compared to literature measurements at lower masses (Carollo et al. 2013a, vander Wel et al. 2014; gray lines and symbols) and spectroscopically confirmed quiescentgalaxies at z > 1 (Krogager et al. 2014, Belli et al. 2014a, Onodera et al. 2014; medianin two redshift bins, black open points). The dashed and solid line show fits to thesize evolution of quiescent and star-forming galaxies, respectively, parametrized as Re =B×(1+z)−β (the fitted parameters are listed in Table 4.1). The shape of the size evolutionfunctions of quiescent and star-forming galaxies is very similar in contrast to lower masseswhere quiescent galaxies show a faster size increase with cosmic time than star-forminggalaxies. Also, star-forming galaxies at these masses are only ∼ 20% larger on average ata fixed redshift and stellar mass, whereas at lower masses the different can be as large asa factor of two (see gray lines).

The similar sizes of star-forming and quiescent galaxies at very high masses areindicative of a fast process that leads to the quenching of such galaxies.

58

4.6. Results: Size evolution of very massive galaxies

In particular, the transition only takes a couple of million years, assuming star-forminggalaxies keep their size and stellar mass during the transition to quiescence (i.e., galax-ies move horizontally on the size vs. redshift diagram). We will further explore this inSection 4.8.

Finally, we note that a recent study by Peng et al. (2015) suggests that the bulk ofstar-forming z ∼ 0 galaxies at log(m/M) < 11 are being quenched within ∼ 4 Gyrs.Thus, we expect the m − Re relation of star-forming galaxies at z ∼ 0.5 to be similar tothe relation of the quiescent galaxies at z ∼ 0 if there would have been no morphologicaltransformation during the quenching at stellar masses of log(m/M) < 11.0. Clearly, thisis not the case as seen in the left panel of Figure 4.5 showing that star-forming galaxies∼ 4 Gyrs ago are significantly (up to factor two) larger at log(m/M) < 11 than localquiescent galaxies.

This is indicative of a structural change in log(m/M) < 11.0 galaxies and might bealso be an effect of disk-fading (see e.g., Carollo et al. 2014).

59


Tab. 4.1 — Qualitative comparison of the size evolution of star-forming and quiescent galaxiesat log(m/M) > 11.4 (this work) and lower stellar masses from 3D-HST (van Dokkum et al.2011, van der Wel et al. 2014). We adopt the parametrization Re(z) = B × (1 + z)−β .

Mass range star-forming quiescent

log(m/M) logBsf βsf logBqu βqu

> 11.4 1.30± 0.02 1.18± 0.15 1.25± 0.25 1.22± 0.2011.0− 11.4 1.05± 0.08 0.80± 0.18 1.05± 0.09 1.32± 0.2110.5− 11.0 0.90± 0.05 0.72± 0.09 0.75± 0.04 1.24± 0.0810.0− 10.5 0.74± 0.03 0.52± 0.08 0.47± 0.02 1.01± 0.06

Tab. 4.2 — Power-law slope (Re(m) ∝ mα) of the stellar mass vs. size relation at z > 0.5 (thiswork including 3D-HST and spectroscopically confirmed quiescent galaxies) as well as integratedover cosmic time at lower redshifts (see references).

redshift range star-forming quiescent Referenceαsf αqu

z ∼ 0 0.14 to 0.39 0.56 (1)z < 0.1 0.19± 0.02 0.41± 0.06 (2)

0.5 < z < 1.0 0.30± 0.10 0.55± 0.05 This work1.0 < z < 1.5 0.22± 0.08 0.62± 0.09 This work1.5 < z < 2.0 0.14± 0.06 0.59± 0.15 This work

z < 3 0.22± 0.05 0.75± 0.05 (3)

(1) Shen et al. (2003). For star-forming galaxies they fit α = 0.14 atlog(m/M) < 10.6 and α = 0.39 at log(m/M) > 10.6.(2) Lange et al. (2015)(3) van der Wel et al. (2014). Report no significant change in slope over 0 < z < 3

60

4.7. Predicted size evolution of massive quiescent galaxies

4.7 Predicted size evolution of massive quiescent galaxies

As we have mentioned above, quiescent and star-forming galaxies at log(m/M) > 11.0show a very similar size evolution. This similarity suggests an instantaneous quenchingprocess instead of a waning. In this section, we will investigate this further by modelingthe size evolution of quiescent galaxies by letting star-forming galaxies with an initial massand redshift evolve to their quiescent state.

The assumption of our model is that galaxies – as long as they are forming stars – evolveon the star-forming main-sequence (MS) spanned by stellar mass m and SFR. We extendthe MS here to a third parameter which is the observed size Re of galaxies, and thereforethe ”new” MS forms a plane in 3-dimensional space. The time at which the process thatultimately suppresses the star-formation starts to work is defined in a probabilistic way viathe (mass) quenching probability introduced in Peng et al. (2010b) and therefore occurs ata characteristic mass mMS. Our null-hypothesis is that galaxies are subject to waning, i.e.,that the gas flow onto their disks is cut-off at a particular time, leading to a consumption ofthe remaining gas within a times-scale of τcons. Note that we do not include rejuvenationof galaxies, thus, once they run out of gas we consider them as quiescent (or quenched).In the following, we refer to this null-hypothesis model as the “waning model”. We willsee that this is not able to explain the observations and we will thus modify it in the latersections.

The main ingredients of the ”waning model” are the following:

• The MS of star-forming galaxies at z < 2. For this we use the parametrization bySchreiber et al. (2014), compiled from recent deep Herschel observations. Note thatthe use of other parametrizations of the MS do not change the results and conclusionsof this work.

• The size evolution of quiescent and star-forming galaxies as a function of cosmictime and stellar mass (Re(m, z)). This we have derived in this paper for UMGs atlog(m/M) > 11.4 and complement it with the measurements of van der Wel et al.(2014) for less massive galaxies (see Figure 4.4).

• The gas fraction of star-forming galaxies at a given stellar mass and redshift(fgas(m, z)). The description of a parametrization for the gas fraction from acompilation of literature values is outlined in section 4.9.

In the following, we summarize the main steps of our ”waning model” in more detail.These points are visualized in Figure 4.6. More details and the derivations of thesequantities will by outlined in the following sections.

1. We start with a galaxy observed at an initial redshift zobs with a given initial stellarmass, mobs. The SFR is computed such that the galaxy is located on the star-formingmain-sequence. It will then grow over a time τMS in stellar mass according toaccretion and merging (thereby staying on the star-forming MS) and also change itsobserved size according to the observed relation of Re(m, z) of star-forming galaxies.

61


10.0 10.5 11.0 11.5 12.0−0.5

0.0

0.5

1.0

1.5

2.0

2.5

log M

log

SFR

2.5 2.0 1.5 1.0 0.5 0.0redshift

10.0 10.5 11.0 11.5 12.0−0.5

0.0

0.5

1.0

1.5

2.0

2.5

log M

log

SFR

initial redshift

0.51.0

1.52.0

initial mass

10.25 10.75

11.25 11.50

2.5 2.0 1.5 1.0 0.5 0.0redshift

initial redshift

0.51.0

1.52.0

initial mass

10.25 10.75

11.25 11.50

star-forming main sequence

at zobs

τcons

2. cut-off inflow when Pq(m,z) = 75% and consumption of gas

1. mass increasesize growth on MS

3. Shut-down of star-formation. Joining

red sequence

red sequence

zobs

star-forming main sequence

red sequence

zobs

τMS

Vis

ualiz

atio

nM

odel

stel

lar m

ass

11.5011.25

10.75

10.252.01.51.00.5 quenched in

the future quenched in the future

τMSτcons

“Waning” Model(consumption of gas on > 500 Myr timescale)

1. mass increase on MS2. gas consumption on MS3. shut-down of SF

1. mass increase on MS2. gas consumption on MS3. shut-down of SF

Fig. 4.6 — Visualization of the ”waning model” in which the star-formation of galaxies isshut-down by the cut off of gas inflow and consumption of the remaining gas on the order of> 500 Myrs. The top panels show a schematic visualization, while the lower panels show theactual ”waning model” for galaxies with different initial redshifts (different symbols) and stellarmasses (sizes of the symbols). The ”waning model” consists of three separate parts that areindicated in the boxes and dashed lines with different colors. For a more detailed description ofthis model, we direct the reader to the corresponding text in Section 4.7.

2. After a time staying on the main sequence (the “main-sequence lifetime”, τMS), thegalaxy reaches a mass at which its probability to get quenched (Pq(m, z)) becomessignificant. In the following we set Pq,lim(m, z) = 75% (which is arbitrary but is notcrucial for our results). The mass at which this happens is log(mMS/M) ∼ 11.0.

3. At that time, we cut off the gas supply onto the galaxy. The galaxy then consumesits gas during a time τcons after which the star-formation is completely suppressed(no rejuvenation). The rate at which the galaxy consumes its gas is defined by thegas fraction and the SFR of the galaxy as it stays on the star-forming main-sequenceduring the time of consumption.

4. During this gas consumption phase we assume a size increase according to theRe(m, z) relation for star-forming main-sequence galaxies.

5. Finally, the gas is fully used up and the galaxy joins the red sequence of quiescent

62


galaxies at a final redshift zqu, final mass mqu, and final size Re,qu.

The following sections describe the different quantities in more detail.

4.7.1 The main-sequence life time τMS

We assume a quenching rate according to “mass-quenching” (Peng et al. 2010b) and definethe probability to get quenched Pq as a function of the mass ratio m/M∗, where M∗ isthe characteristic turnover stellar mass in the Schechter formalism,

Pq(m, z) = 1− exp[−m/M∗(z)]. (4.3)

In the following, we use log(M∗/M) = 10.85 for all redshifts and Pq therefore onlydepends on the galaxy’s stellar mass. Note, that the probability may also depend on theenvironment in which a galaxy is residing in via the “environment/satellite-quenching”process. In the following, we assume that this mechanism is sub-dominant in our stellarmass range (log(m/M) > 10.0) with respect to mass-quenching2. The characterizationof environment for our galaxies is beyond the scope of this work. With τMS , we denote thetime a galaxy (on the star-forming main sequence starting at an observed redshift zobs)needs to reach a certain stellar mass where Pq(m, z) = 75% (corresponding to roughlylog(m/M) ∼ 11.0). The mass evolution m(z) is derived self-consistently along thestar-forming main-sequence according to sSFR(z) as well as the specific merger rates(Peng et al. 2010b).

Typical timescales τMS for a galaxy with log(m/M) = 11.25 are 0.4 Gyr at zobs = 2 and2 Gyr at zobs = 0.5. For less massive galaxies, e.g., log(m/M) = 10.25, it is τMS ∼ 1 Gyrat zobs = 2 and more than the look-back time for galaxies at zobs = 0.5. Those willtherefore quench in the future and will not be considered here. For galaxies with masseslarger than log(m/M) ∼ 11.0, we find τMS = 0 as expected, since Pq(m, z) > 75% atthese masses.

4.7.2 The consumption time scale τcons

The time that is needed to move our galaxies in the ”waning model” from star-forming toquiescent is defined by the gas consumption time τcons, which depends on the SFR of thegalaxies as well as on their gas fraction. In Appendix section 4.9, we derive an empiricalparametrization of the gas fraction (mgas/(m+mgas)) as a function of redshift and stellarmass fgas(m, z) from various measurements in the literature. We then solve iteratively forthe consumption time of the galaxy in the following way. We start off with a galaxy onthe star-forming main sequence with a certain stellar mass minit at a redshift zinit andcompute the gas fraction fgas,init(minit, zinit) (and therefore mgas,init) and, assuming thegalaxy is on the main sequence, the star-formation rate SFRinit. In each time step, thegalaxy increases its stellar mass by ∆t×SFRinit, where ∆t is the length of each time step.As we assume that the galaxy stays on the main sequence, we increase its SFR according

2This is a good approximation from the analysis of Peng et al. (2010b), showing that mass-quenchingis the dominant above log(m/M) > 10.0 (10.5) at z ∼ 2 (z ∼ 0).

63


logM = 10.25logM = 10.75logM = 11.25logM = 11.5

for galaxies on the MS

12 11 10 9 8 7 6 5 4 3 20

1

2

3

40.2 0.3 0.5 0.7 1 1.5 2 3

redshift

age of the Universe (Gyr)

con

sum

pti

on

tim

e τ

con

s (G

yr)

Fig. 4.7 — Consumption time scale τcons for our ”waning model” as a function of redshift for agalaxy staying on the star-forming main-sequence during its consumption of remaining gas afterthe start of quenching. Shown are four different stellar mass bins. The hatched error bands takeinto account the uncertainties in fgas as described in the text. The gray shade denote the regionwhere τcons is longer than the lock-back time. Galaxies in that region will therefore consumetheir gas in the future.

to its mass increase in the following time step. This is repeated until all of the gas is usedup, which then defines τcons. The final mass of the galaxy after the consumption of allthe gas is accordingly m(τcons) = mMS +mgas, where mMS is the stellar mass the galaxyhad before the cut off of gas and mgas is the gas mass. The consumption time scale asa function of redshift for different galaxy masses is shown in Figure 4.7 with solid linesincluding the uncertainty from fgas as the hatched region. For a typical galaxy at z ∼ 2and stellar mass log(m/M) = 11.0 it is τcons ∼ 500 Myrs. Because of the decrease insSFR with cosmic time, we find longer consumption times at later times. For example, atypical galaxy at z ∼ 0.5 with stellar mass log(m/M) = 11.0 has τcons ∼ 2 Gyrs.

64


redshift when quenched

0.5 1 2

dashed: error from Re(z)solid: error from fgas(m,z)

11.0 11.2 11.4 11.6−20%

0%

20%

40%

60%

80%

100%

log M (final mass when quenched)

dis

crep

ancy

mo

del

vs.

ob

serv

atio

n

(κ)

final stellar mass

logM = 11.0 logM = 11.3 logM = 11.6

dashed: error from Re(z)solid: error from fgas(m,z)

0.0 0.4 0.8 1.2 1.6 2.0 2.4−20%

0%

20%

40%

60%

80%

100%

dis

crep

ancy

mo

del

vs.

ob

serv

atio

n

(κ)

redshift when quenched (zqu)

Fig. 4.8 — Discrepancy parameter κ (comparing the model galaxy sizes with the observed; seeEquation 4.4) for our ”waning model” shown as a function of final mass and redshift (left: asa function of final mass; right: as a function of final redshift). The errors on the points includeuncertainties in τcons due to uncertainties in fgas (solid error bars) as well as in the fitting ofRe(m, z) (dashed error bars). The ”waning model” over-predicts severely (up to 80%) the sizesof quiescent galaxies in the mass range 11.0 < log(m/M) < 11.6. The over-prediction is lesssevere at lower redshifts as well as at higher masses where τcons reaches the dynamical timescaleof the galaxy.

4.7.3 Size evolution of our model galaxies

We assume that while a galaxy is forming stars it remains on the 3-dimensional MS planespanned by m, SFR, and Re. The size change is therefore tracked by the Re(m, z) relationof star-forming galaxies. In the following, we denote quantities predicted by the ”waningmodel” with a ”hat” notation: Re,qu, which is the size at (mqu, zqu).

65


4.8 How to quench massive galaxies

4.8.1 ”Waning model”: tension with observations

We now compare the predicted size evolution of quiescent galaxies from our ”waningmodel” with observations. In order to quantify the differences, we introduce a so-called“discrepancy” parameter κ defined as

κ =Re,quRe,qu

, (4.4)

where Re,qu is the galaxy size output by our model at (mqu, zqu) and Re,qu is thegalaxy size of observed quiescent galaxies at the same mass and redshift, i.e., (mqu =mqu, zqu = zqu). We show κ qualitatively in Figure 4.8 in per-cent (0% means a perfectmatch between model and observation) for different values of final3 redshifts and stellarmasses. On the upper panel, we show κ as a function of mass in three final redshift bins(zqu = 0.5, blue; zqu = 1.0, green; zqu = 2.0, orange). On the bottom panel, we show viceversa κ as a function of redshift for three different mass bins (log(mqu/M) = 11.0, blue;log(mqu/M) = 11.3, green; log(mqu/M) = 11.6, orange). The solid errors visualize theuncertainties in the gas fraction fgas(m,z). The dashed error bars show the uncertaintiesin the size evolution Re(m, z). We find that at all final masses and redshifts our ”waningmodel” overshoots the measured galaxy sizes. The magnitude of over-estimation dependsstrongly on mass and it becomes especially severe at lower masses. At all redshifts, forstellar masses above log(m/M) ∼ 11.5 the discrepancies are on the order of 20%. Atlower masses (down to log(m/M) = 11.0), however, they are up to 80% depending onredshift. At these masses, κ is in general a factor of 1.5 larger at zqu = 1.0 compared tozqu = 2.0.

Summarizing, within our ”waning model” we are not able to reproduce the observed sizeevolution of quiescent galaxies. Our model (assuming the suppression of star-formationhappening after a long time) over-estimates the sizes by 20% to 80%. In the following,we modify this model to investigate the effect of mergers (possibly playing a role at thehighest masses).

4.8.2 The effect of mergers (”instantaneous quenching”)

In the picture of hierarchical galaxy growth, major mergers are expected to happen withincreasing rate at increasing redshift (e.g., Hopkins et al. 2010b, Rodriguez-Gomez et al.2015). Major mergers are usually followed by a short (on the order of a dynamical time,i.e., ∼ 100 Myrs) star-burst phase that can result in a quick consumption of gas andtherefore accelerated quenching (e.g., Mihos & Hernquist 1994).

Motivated by this, we investigate the effect of mergers on the size evolution of massivequiescent galaxies. We modify the steps 3 and 4 of our ”waning model” and include theeffect of a merger with a subsequent star-burst that consumes all the gas instantaneously4.

3“Final” means when galaxy is quenched in our model.4Note that this is an extreme version of the ”waning model” by setting τcons = 0.

66

4.8. How to quench massive galaxies

logM = 11.3 model quenched galaxies (1:4 Merger)

no size increaseRe ! m0.5

Re ! m1.0

observed quenched galaxies

14 13 12 11 10 9 8 7 6 5 4 3 2 1

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

0 0.2 0.5 1 1.5 2 3

redshift

cosmic time (Gyr)

log

Re

(k

pc)

logM = 11.6 model quenched galaxies (1:4 Merger)


Re ! m1.0


14 13 12 11 10 9 8 7 6 5 4 3 2 1

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

0 0.2 0.5 1 1.5 2 3

redshift

cosmic time (Gyr)lo

g R

e (

kp

c)

logM = 11.3 Model quenched galaxies (1:1 Merger)


Re ! m1.0


14 13 12 11 10 9 8 7 6 5 4 3 2 1

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

0 0.2 0.5 1 1.5 2 3

redshift

cosmic time (Gyr)

log

Re

(k

pc)



Re ! m1.0


14 13 12 11 10 9 8 7 6 5 4 3 2 1

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

0 0.2 0.5 1 1.5 2 3

redshift

cosmic time (Gyr)

log

Re

(k

pc)

logM

= 1

1.6

logM

= 1

1.3



Re ! m1.0


14 13 12 11 10 9 8 7 6 5 4 3 2 1

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

0 0.2 0.5 1 1.5 2 3

redshift

cosmic time (Gyr)

log

Re

(k

pc) logM

= 1

1.6

logM

= 1

1.3



Re ! m1.0


14 13 12 11 10 9 8 7 6 5 4 3 2 1

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

0 0.2 0.5 1 1.5 2 3

redshift

cosmic time (Gyr)

log

Re

(k

pc)

Fig. 4.9 — The predicted size evolution of quiescent galaxies for our ”instantaneous-quenchingmodel” (black) compared to the observed (red). We show the effect of a 1 : 1 major merger (left)and a 1 : 4 merger (right) for two different final masses at log(m/M) = 11.3 and log(m/M) =11.6 (separated for visual purposes). Compared to the ”waning model”, a instantaneous, mergerdriven quenching process is able the reproduce the observed size evolution to better than 20−30%at all masses log(m/M) > 11. Furthermore, including a size increase during the mergingevent (dashed: Re ∝ m0.5; dotted: Re ∝ m1.0) reduces the discrepancy between model andobservations substantially.

Because the suppression of star-formation happens immediately after the merging event,we refer to this model as “instantaneous-quenching model”. We implement a 1 : 1 as wellas a 1 : 4 merger at the time τMS. We assume that both merging galaxies are star-formingand have pre-merger gas fractions according to their stellar mass and redshift when themerging event happens. The total mass of the galaxy after the merging event (i.e., the finalmass when the galaxy is fully quenched) is therefore mqu = (1+fmerger)mMS +mgas,MS,1 +mgas,MS,2, where fmerger is the mass fraction of the merger and the subscripts “1” and “2”denote the gas masses of the two galaxies prior to the merging event. Because of the lowgas fractions at these masses, it is roughly mqu ∼ (1 + fmerger)mMS.

The solid black lines in Figure 4.9 show the resulting size evolution of quiescentgalaxies from our instantaneous-quenching model in two final stellar mass bins comparedto observations (solid red line) for the 1 : 1 merger (left) and 1 : 4 merger (right). Inthe case of the 1 : 1 merger, the model now under-estimates the observed galaxy sizes ofquiescent objects. In the case of a 1 : 4 merger, the under-estimation is less severe, aswell as at lower masses (log(m/M) = 11.3). Importantly, the ”instantaneous-quenchingmodel” reproduces the size evolution of quiescent galaxies to better than 20− 30% at allmasses log(m/M) > 11. It is thus doing much better than the ”waning model”.

However, nature is not as simple as that and we might expect the size of the galaxy to

67


change during a merger event. Theoretical arguments and numerical simulations suggesta size increase for major mergers of δRe ∝ δmγm with γm = 1, δRe = Re,post/Re,pre, andδm = mpost/mpre (Hernquist et al. 1993, Naab et al. 2009, Hilz et al. 2013). This can alsobe shown to first order from the virial theorem

Re,postRe,pre

=

(mpost

mpre

)(σpreσpost

)2

, (4.5)

if assuming that the velocity dispersion (σ) changes little during the merger event (assuggested by e.g., Hopkins et al. 2009a, Oser et al. 2012). The subscripts denote pre- andpost-merger phase.

The dotted black lines in Figure 4.9 show the case for γm = 1. We find again asubstantial over-prediction of the galaxy sizes for the 1 : 1 major merger. The size increasefor the 1 : 4 merger is less steep and therefore results in only a slight over-estimate.The dashed black lines show our model with γm = 0.5 that fits very well the observedsize redshift relation for log(m/M) = 11.6. In the case of perfect virial equilibrium(Equation 4.5), this would result in an increase of the velocity dispersion of 20% for the1 : 1 merger and 5% for the 1 : 4 merger. Simulations suggests a change in σ due tomergers on the order of 10 − 20% or less between 0 < z < 2 (e.g., Hopkins et al. 2009a).Given that our 1 : 1 merger case is an upper limit, the increase in σ for γm = 0.5 isbearable. Also, recent cosmological hydrodynamical simulations suggest 0.0 < γm < 1.0for disk major mergers (Welker et al. 2015). Furthermore, the merger might trigger astar-burst in the center of the galaxy resulting in a compact stellar core after the merginghappened, which again would result in smaller observed galaxy sizes.

4.9 Summary & conclusions

We use the size evolution of massive star-forming and quiescent galaxies as an independentdiagnostic tool to investigate the process of quenching at log(m/M) > 11 and z . 2. Tothis end, we measure for the first time the half-light size evolution of a statistical largesample of very massive star-forming and quiescent galaxies at log(m/M) & 11.4 on the2-square degree survey field of COSMOS/UltraVISTA. We find that the size evolution ofboth populations of galaxies at log(m/M) > 11.4 is very similar and is consistent withthe extrapolation of the mass versus size relation from lower masses.

We use this result and predict the size evolution of quiescent galaxies for two bracketingmodels for the suppression of star-formation:

• The ”waning model” in which the gas inflow onto a galaxy is cut off such that thegalaxy consumes its remaining gas slowly on the order of 500 Myr and more afterwhich the star-formation is suppressed.

• The ”instantaneous-quenching model” in which a merger-induced star-burst depletesthe gas on the spot and therefore leads to an instantaneous quenching of the galaxy.

Our main findings are the following:

68

4.9. Summary & conclusions

• The ”waning model” over-estimates the sizes of quiescent galaxies by up to 80% atlog(m/M) > 11.0. The over prediction is less severe (< 30%) at very high masseslog(m/M) > 11.5 where the consumption time of the remaining gas is short, onthe order of a couple of dynamical times (< 500 Myrs) gas. In order to make thismodel work, a substantial morphological transformation of the galaxy would needto take place.

• With the ”instantaneous-quenching model”, however, we are able to predict sizeevolution of quiescent galaxies to better than 20− 30% at all masses log(m/M) >11.0. In this case, our model under-predict the sizes of real quiescent galaxies.Assuming a size growth during the merging event proportional to

√m further reduces

the discrepancy between model and observations.

We conclude that a fast (not more than a couple of dynamical timescales) process isfavored for the shut-down of star-forming massive galaxies at z < 2 in order to reconcilethe size evolution of quiescent galaxies. Given the increased merger rate at high massesand redshifts, a major merger event followed by a star-burst and a quick depletion of gasis a plausible process.

Extrapolating to lower masses, our ’instantaneous-quenching model’ starts toover-predict the sizes and it is therefore very possible that a different process is at work.A process similar to ’waning’ (for example cause by stripping of a galaxy’s hot halo whilefalling into a cluster) followed by a substantial (20-80%) size decrease of in observed lightmight be a plausible process. Indeed, this can be achieved by a post-quenching evolutionwith disk-fading, which becomes more important for galaxies at lower masses that exhibita dominant disk component (e.g., Carollo et al. 2014).

69


APPENDIX: The gas fraction fgas(m, z)

In order to derive the consumption time scale τcons, we need the gas fraction in galaxies ata certain stellar mass and redshift. To this end, we fit an empirical relation fgas(m, z) fromdifferent studies in the literature. These include PHIBSS at z ∼ 1 − 1.5 (Tacconi et al.2013) and COLDGASS at z ∼ 0 (Saintonge et al. 2011) as well as data from lensed andother star-forming galaxies from Dessauges-Zavadsky et al. (2014) and references therein.

The result is shown in Figure 4.10 for four different bins in stellar mass. The PHIBSSand COLDGASS data is shown in black, the other measurements are shown in gray.We also show in color fgas derived from our galaxies (UMGs and lower mass controlsample) using the Kennicutt-Schmidt relation (KS relation, Kennicutt 1998, Schmidt1959), relating Σgas ∝ ΣN

SFR, where we take N = 1.31 (Krumholz et al. 2012). Notethat these derivations are not used for the fitting of the parametrization for fgas(m, z).

We derive fgas(m, z) and its uncertainty (95% CLs) by fitting the observed data asfollows. We first perform a linear fit forced through the COLDGASS data point at z = 0in order to determine the slope. The error on the slope is derived from the systematicerror of the fit and the uncertainty of the data points by bootstrapping which we bothadd in quadrature. In a second fit, we fix the slope to the one determined before and fitfor the intercept including error. The resulting uncertainty region as shown in Figure 4.10as hatched region is then derived by the unification of the errors of the two fits. To get acontinuous function for fgas, we interpolate between the four stellar mass bins.

We compared our fit to the recent work by Genzel et al. (2015). We find that theirfgas(m, z) parametrization (see their table 4) has a slightly steeper redshift dependenceresulting in 10 − 30% larger gas fractions at the highest redshifts. We have verified thatour results do not change if using the Genzel et al. (2015) parametrization for fgas(m, z).

70

4.9. Summary & conclusions

10.0 < logM < 10.5

UltraVISTA via KS relation

best linear fit + error0.0

0.1

0.2

0.3

0.4

0.5

0.6

f gas

10.5 < logM < 11.0


best linear fit + error

11.0 < logM < 11.4



0.0 0.5 1.0 1.5 2.0 2.5

0.0

0.1

0.2

0.3

0.4

0.5

0.6

redshift

f gas

logM > 11.4



literature data:

PHIBSSCOLDGASSother SFGs

0.0 0.5 1.0 1.5 2.0 2.5

redshift

Fig. 4.10 — Gas fraction as a function of redshift in the four different stellar mass bins (dif-ferent panels). The black and gray symbols show literature values measured from individ-ual galaxies (black squares: PHIBSS at z ∼ 1 − 1.5, Tacconi et al. (2013); black diamonds:COLDGASS at z ∼ 0, Saintonge et al. (2011); gray points: lensed and other star-forming galax-ies from Dessauges-Zavadsky et al. (2014) and references therein). The hatched region is a fit(fgas(m, z)) to the measured individual galaxies including uncertainty as described in the text.For comparison, the color symbols show fgas derived from the Kennicutt-Schmidt relation forthe UltraVISTA galaxies. Note that these are not used in the fitting.

71

CONCLUSIONS & FUTURE RESEARCH

72

Conclusions & Future Research


The main results and conclusions of this thesis can be shortly summarized as follows

• Our spectroscopic follow-up of LAE candidates at ∼ 7.7 makes aware of an increasingfraction of contaminants in such samples at z > 7. The Lyα LF we derive from ourobservations at z ∼ 8 indicates a decrease of a factor of 4 in Lyα transmittance sincez = 6 and we estimate a neutral hydrogen fraction of 50−70% at z ∼ 8. These valuesare consistent with galaxies being the main drivers of re-ionization (for reasonableassumptions of escape fraction of ionizing photons and clumping factors).

• The size evolution of massive log(m/M) > 11.0 quiescent galaxies are explainedby a quenching mechanism that includes a merger event with subsequent star-burstand instantaneous consumption of gas leading to instantaneous quenching. Therebywe suggest that galaxy increase their size proportional to m0.5.

I am excited to move to Caltech with my post-doctoral fellowship to develop the nextsteps in these very interesting topics of astrophysics; the growth of stellar mass in differentgalactic structures at the re-ionization process at z > 6.

In more detail, my future work at Caltech will focus on

(i) the measurement of the escape fraction of ionizing radiation that is still one of themajor unknowns in understanding the process of re-ionization, and,

(ii) the derivation of stellar mass functions using the largest available galaxy samples forthe first time at z > 4, crucial for the understanding of the stellar mass growth inthe early universe.

The fraction (fesc) of ionization radiation that is able to escape from a galaxy and isavailable for re-ionization depends on the internal properties and stellar physics (supernovawinds, stellar migration, runaway stars) of a galaxy but also the close surrounding IGM.This makes it to one of the most unconstraint parameters and subject to heavy debates.Over the last decades, several studies have been carried out with the goal to measurethe escape fraction of ionizing radiation and its evolution with redshift and other physicalparameters via the detection of escaping Lyman continuum (LyC) radiation (e,g., Leithereret al. 1995, Steidel et al. 2001, Iwata et al. 2009, Siana et al. 2010, Nestor et al. 2011,Vanzella et al. 2012, Mostardi et al. 2013, Siana et al. 2015). Conclusive results aredifficult to obtain, because the inferred escape fractions strongly depend on the selectionof the galaxies. For example different color selection are effectively selecting different fesc

(see e.g., Cooke et al. 2014). Furthermore, most galaxy selections at different redshiftsare sensitive to different evolutionary stages of the galaxies (different age, SFR, stellarmass). Last but not least, to increase the observing efficiency, high redshift galaxiesare usually selected to be within large over-densities and thus reside in a different IGM.Consistent samples with redshift are important to constrain the redshift evolution of fesc

and to provide a sensible extrapolation to z > 4 where the IGM becomes optically thick toionizing photons and therefore no direct measurement of the escape fraction can be carriedout (Inoue et al. 2014). Also many observational issues hamper the reliable detection ofthe LyC and therefore our constraint on fesc. Most of the reported LyC detections aremost likely due to contamination in redshift space or in the line-of-sight (e.g., Siana et al.

75


2015). Averaging over all the measurements with the caveat of the issues mentioned above,escape fractions on the order of 10% are commonly found, but values between 0% and 80%are found as well for single galaxies.

In order to understand the build up of stellar mass in the early universe, reliable stellarmass functions (i.e., number of galaxies per stellar mass per comoving volume) are crucial.The measurement of stellar mass becomes more and more difficult with increasing redshiftbecause of the lack of deep near-IR to IR coverage within current surveys. This part ofthe galaxy spectrum is however essential as it includes many features that are sensitiveto the stellar populations. At z > 4 the Spitzer Space Telescope (SST) can provide thisinformation. Current surveys using SST lack sufficient depth and area to reliably estimatethe star-forming MS and the mass functions at z > 4 over a large mass range fromlog(m/M) < 10 up to log(m/M) > 11.0 (e.g., Labbe et al. 2010, Gonzalez et al. 2011,Duncan et al. 2014). Furthermore, pencil-beam surveys are prone to cosmic variance (i.e.,clustering) effects that can vary the number density in stellar mass functions by up 30%.This is especially severe at high masses where the number density of objects is decreasingrapidly.

The recently approved and now on-going observations of the Spitzer Large Area Surveywith Hyper-Suprime-Cam (SPLASH, PI: P. Capak) to which I have contributed to as aCoI during my Ph.D, will be able to answer the above questions. SPLASH combinesthe deepest (25.5 AB at 5σ) IR imaging at 3.6µm and 4.5µm by the Spitzer SpaceTelescope ever taken over a contiguous area of two times 1.8 square-degrees with evendeeper (27 − 28 AB at 5σ) optical and near-IR (from 0.4µm to 1µm, narrow-band andbroad-band filters) imaging taken by the Hyper-Suprime-Cam (HSC) on the 8.2-meterSubaru telescope. It therefore provides the necessary depth of IR observations as well asthe large area on sky in order to beat down the cosmic variance as good as possible andto improve the statistics on the most massive log(m/M) > 11 galaxies. Furthermore,the HSC provides deep imaging in narrow-band filters that make the measurement of theLyC radiation possible for a large and complete sample of galaxies (through individualmeasurements or stacking).

In the far future, several new observation facilities will come into play (JWST, TMT,ELT). These will revolutionize mainly the spectroscopy at high redshifts and will increasethe spectroscopically confirmed samples of high-z galaxies and our understanding of galaxyformation.

76

Appendix

Chapter A

Summary of collaborative projects

In this chapter, we give a brief summary of the collaborative projects that I havecontributed to during the course of my Ph.D. I am the second author of the first two,highlighting the major contribution to them. Therefore, I will present these two works inmore detail in the next two chapters.

1. Keck-I MOSFIRE Spectroscopy of the z ∼ 12 Candidate Galaxy UDFj-39546284

(P. Capak, A. L. Faisst, J. D. Vieira, et al. 2013)

The observational selection and the knowledge of its limitations and uncertainties iscrucial to study properly the EoR. The spectroscopic confirmation of very high redshift(z > 7) galaxies very difficult, mainly because of the faintness of these sources (magnitudesof H = 26.5 in AB and above). Commonly, the redshift of these galaxies is determined viathe detection of the Lyα emission line. However, as a matter of fact, their Lyα emissionis strongly decreased because of the increasing amount of HI. Another possibility is to usethe Atacama Large Millimeter/submillimeter Array (ALMA) to target the CO emissionin these galaxies at z > 4.5 (e.g., Capak et al. 2015, in prep) or to target the CIII line atλ = 1909 A in z ∼ 6−7 galaxies using this and next generation of (near-)IR spectrographs(e.g., Stark et al. 2014). As there are only tentative confirmations of galaxies above z ∼ 8,it is generally difficult to understand the contamination of low redshift interlopers inthese samples. In particular, the galaxy UDFj-39546284 first selected by Bouwens et al.(2011b) was a strong candidate for the first galaxy to be detected at a redshift of z ∼ 10.0.Further investigations by Ellis et al. (2013) using deeper HST broad band photometryand Brammer et al. (2013) using HST/IR grism spectroscopy increased the redshift toz ∼ 11.9, however, the non-negligible chance for it to be at z ∼ 2 was never ruled outBouwens et al. (2013).

In Chapter B (based on Capak et al. 2013), we outline our spectroscopic follow-upof UDFj-39546284 using the near-IR spectrograph MOSFIRE. We can reproduce theobservations of an emission line within our much longer exposure as found in earlier works,however, even much deeper spectroscopy is needed to firmly conclude the nature of thisline and therefore its redshift.

79

CHAPTER A. Summary of collaborative projects

2. Dust Attenuation in high Redshift Galaxies – “Diamonds in the Sky”

(N. Z. Scoville, A. L. Faisst, P. Capak, et al. 2014)

The measurement of the physical properties of high redshift galaxies (z > 3) is basedon the rest-frame UV luminosity that is prone to dust absorption. This is especiallyan issue for measurements of SFRs, because other, more reliable, tracers accounting fordust obscured star-formation in the (near-) IR are out of reach of current instruments.The dust extinction curve is therefore crucial for reliable measurements. For example, theevolution of SFR with redshift is important to quantify whether galaxies are responsible forthe re-ionization of the universe. Commonly, the dust extinction curve of local starburstgalaxies is extrapolated and used at higher redshift (Calzetti et al. 2000). Only recentlywith ALMA, it has been possible to probe the dust content of single high redshift (z & 6)by observation of CII (e.g., Schaerer et al. 2015). This studies confirm a Calzetti like dustextinction curve on small sample sizes, however, other studies favor a relation closer tothe one observed in the small Magellanic cloud (e.g., Oesch et al. 2013a).

In Chapter C (based on Scoville et al. 2014), we outline our study on the dust extinctioncurve making use of 266 galaxies between 2 < z < 6. The advantage of our approach is tobreak the dust-age-metallicity degeneracy by constraining the age of the stellar populationsand thus model the intrinsic UV continuum emission from absorption line spectra and themeasurement of the Balmer break at 4000 A. Our measurements favor a Calzetti-like dustextinction curve together with a strong 2175 A bump feature that is also present in thedust extinction curve of the Milky Way and large Magellanic cloud and is possibly dueto graphite or polycyclic aromatic hydrocarbons (PAHs). Last but not least, the dustextinction curve obtained form this project provides a firm basis for color and extinctioncorrections of the photometry of high redshift galaxies.

3. Evolution of Galaxies and their Environments at z = 0.1− 3 in COSMOS

(N. Z. Scoville, S. Arnouts, . . . , A. L. Faisst, et al. 2013)

In this work, we investigate the relation between galaxy morphology and quiescentfraction and large scale environment out to z ∼ 3 in COSMOS using extremely accuratephotometric redshifts for more than 150,000 near-IR selected galaxies. The near-IR(Ks-band) selection here is crucial to select a complete sample of quiescent galaxies aswell as to include red, dust extinct galaxies. We use two techniques in order to estimatethe environmental densities within 127 redshift slices: adaptive smoothing and Voronoitessellation. We find approximately 250 statistically significant over-dense structures outto z = 3.0 with shapes varying from elongated filamentary structures to more circularlysymmetric concentrations. We also compare the densities derived for COSMOS withthose based on semi-analytic predictions for a ΛCDM simulation and find excellent overallagreement between the mean densities as a function of redshift and the range of densities.The galaxy properties (stellar mass, SEDs, and SFRs) are strongly correlated withenvironmental density and redshift, particularly at z < 1.0− 1.2. Classifying the spectraltype of each galaxy using the rest-frame (b− i) color (derived from SED fitting), we find astrong correlation of early type galaxies (E-Sa) with high density environments, while thedegree of environmental segregation varies systematically with redshift out to z ∼ 1.3. Inthe highest density regions, 80% of the galaxies are early types at z = 0.2 compared to only

80

20% at z = 1.5. The SFRs and the star formation timescales exhibit clear environmentalcorrelations. At z > 0.8, the star formation rate density is uniformly distributed over allenvironmental density percentiles, while at lower redshifts the dominant contribution isshifted to galaxies in lower density environments.

4. Dependence of Morphological Mix, Quiescence, and Size on Environmentat 0.2 < z < 0.8 in zCOSMOS(C. M. Carollo, A. L. Faisst, et al. 2015, in prep)

In this work in preparation, we primarily study the dependence of the morphological mix(late-type versus early-type) and the stellar mass versus galaxy size relation on differenttypes of environments at 0.2 < z < 0.8. Our sample consists of spectroscopically confirmedgalaxy groups at 0.2 < z < 0.8 and a halo mass above log(m/M) = 12.5 from the Knobelet al. (2012) group catalog based on zCOSMOS (Lilly et al. 2007; 2009). Star-formationsand stellar masses are derived from SED fitting to the COSMOS broad-band photometryand the splitting in quiescent and star-forming galaxies is based on a cut in specific SFRat 10−11 yr−1 at z ∼ 0, evolving with redshift. The morphological classification in early-and late-types is based on the HST/ACS F814W (I-band) images and derived using anautomated classification by an improved version of the Zurich Estimator of StructuralTypes (ZEST, see Scarlata et al. 2007). Our new version (ZEST+) is presented in Carollo etal. (2015, in prep). The observed galaxy half-light radii (based on F814W) are also derivedby ZEST+ and corrected for systematic biases and PSF effects using simulations. Thepreliminary results of this work show very similar trends to what is found at z ∼ 0 (Carolloet al. 2014). We find that the morphological mix of the quiescent galaxy population doesnot change as a function of halo-centric distance, while the fraction of quiescent galaxiesincreases toward the center. We also find no significant differences in the stellar massversus size relation of quiescent and star-forming galaxies as a function of environment(central versus satellite galaxies), except at very high masses (log(m/M) > 11.0) wherequiescent centrals are marginally smaller in size. We are quantifying the significance androbustness of this result before submitting this paper, as well as investigate the size versusstellar age relation as a function of environment.

81

CHAPTER A. Summary of collaborative projects

82

Chapter B

Keck-I MOSFIRE spectroscopy of az ∼ 12 candidate galaxy

The content of this chapter is based onP. Capak, A. L. Faisst, J. D. Vieira, et al. (2013)

B.1 Summary

We report the results of deep (4.6h) H band spectroscopy of the well studied z ∼ 12H-band dropout galaxy candidate UDFj-39546284 with MOSFIRE on Keck-I. These datareach a sensitivity of 5 − 10 × 10−19 erg s−1 cm−2 per 4.4 A resolution element betweensky lines. Previous papers have argued this source could either be a large equivalent widthline emitting galaxy at 2 < z < 3.5 or a luminous galaxy at z ∼ 12. We find a 2.2σ peakassociated with a line candidate in deep Hubble-Space-Telescope Wide-Field-Camera 3Infrared grism observations, but at a lower flux than what was expected. After consideringseveral possibilities we conclude these data can not conclusively confirm or reject theprevious line detection, and significantly deeper spectroscopic observations are required.We also search for low-redshift emission lines in ten other 7 < z < 10 z, Y , and J-dropoutcandidates in our mask and find no significant detections.

B.2 Introduction

Deep photometric surveys with the Hubble Space Telescope (HST) are revolutionizing ourknowledge of the star forming galaxy population at the epoch of hydrogen re-ionization,only a few hundred million years after the Big Bang. Hundreds of candidate z ∼ 7 − 8galaxies have now been discovered using the Lyman break galaxy (LBG or ’dropout’)technique using newly available HST Wide-Field-Camera-3-Infrared (WFC3-IR) data(Bouwens et al. 2010, Oesch et al. 2010a, McLure et al. 2013, Ellis et al. 2013). Recently,this technique has been pushed to z ∼ 12 by using hundreds of HST orbits to probe tounprecedented depths in the Hubble-Ultra-Deep-Field 2012 (HUDF12) (Ellis et al. 2013).

83

CHAPTER B. Keck-I MOSFIRE spectroscopy of a z ∼ 12 candidate galaxy

The LBG or ’dropout’ technique relies on absorption by intervening neutral hydrogenbelow the Lyman limit at 912 A and Lyα at 1216 A to create a strong spectral breakthat differentiates high and low redshift galaxies using only broad-band photometry. Thistechnique was first introduced in the 1980’s and 1990’s (Cowie 1988, Steidel et al. 1996)and broadly adopted as the main technique for finding candidate distant sources once itwas shown to be effective spectroscopically on large samples at z ∼ 3 − 4 (Steidel et al.1999; 2002) and then deployed at ever higher redshifts (Iwata et al. 2003, Ouchi et al.2004, Bouwens et al. 2007). The current frontier is to use the near-infrared Wide FieldCamera 3 (WFC3/IR) Hubble Ultra Deep Field data for ’J and H -dropout’ galaxies atredshifts z ∼ 8 − 12 and has led to a handful of tentative detections, most noticeablythe source UDFj-39546284, a candidate z ∼ 11.9 galaxy (Ellis et al. 2013) previouslyclaimed to be at z ∼ 10.3 (Bouwens et al. 2011a, Oesch et al. 2012). If confirmed to be atsuch high redshifts, even this single galaxy would be of paramount importance to probethe physics of the earliest phases of galaxy formation, and to quantify how star forminggalaxies contribute to the re-ionization of the Universe (Robertson et al. 2010, Bouwenset al. 2011a, Oesch et al. 2012).

However, spectroscopic redshifts are fundamental to ascertain the true redshifts ofcandidate high-z LBGs. At z > 4 application of the dropout technique has lead tonotably diverse results (Iwata et al. 2003, Ouchi et al. 2004, Bouwens et al. 2007, van derBurg et al. 2010) with only limited samples being spectroscopically confirmed (Vanzellaet al. 2009, Stark et al. 2010, Mallery et al. 2012), and the vast majority of spectroscopyfailing to detect anything significant. For example, Stark et al. (2010) and Vanzella et al.(2009), two of the largest spectroscopic samples at z > 4 to date, confirm less than halfof their targeted objects. This is particularly problematic at z > 6 where exotic objectscan contaminate the dropout selection (Capak et al. 2011).

A growing body of evidence suggests extreme line emitters and unusual evolved galaxiesat z ∼ 2 are a contaminant in z > 7 LBG selections (Atek et al. 2011, Capak et al. 2011,Hayes et al. 2012). In published spectroscopic studies of z > 7 galaxies the vast majorityof results are null, with a large fraction of detected objects placed at z < 3 (Vanzellaet al. 2011, Capak et al. 2011, Caruana et al. 2012, Hayes et al. 2012, Ono et al. 2012b,Caruana et al. 2012, Bunker et al. 2013). The null results, combined with the type ofcontamination is worrying because it is from a poorly understood population of objectsand so is difficult to include in the simulations required to quantify LBG selection criteria(Capak et al. 2011). In the near term this highlights the need for deep spectroscopic studiesat 6 < z < 8 where current spectrographs can hope to confirm existing high-z candidatesand characterize contaminating populations. In the longer run the James Web SpaceTelescope (JWST) and Thirty-meter class ground based telescopes will be necessary toobtain the requisite spectroscopic samples at z > 8 and faint fluxes needed to understandthese populations.

UDFj-39546284 has been well studied by many authors and is only detected in theF160W H band filter with HST WFC3-IR, even though the deepest currently possibledata exists at both bluer and redder wavelengths. It was first reported by Bouwens et al.(2011a) who originally claimed it was at z ∼ 10. Ellis et al. (2013) recently improved thedepth of the HUDF F160W and the F105W images by 0.2 and 0.5 magnitudes respectivelyand added deep F140W imaging which overlaps the blue half of F160W. This new data

84

B.3. Data

(HUDF12) favors UDFj-39546284 being at z = 11.9, but still allows the possibility it couldbe a strong line emitter at 2 < z < 3.5. Subsequently, Brammer et al. (2013) reported a2.7σ line detection at 15990 ± 40A with a flux of 3.5 ± 1.3 × 10−18 erg s−1 cm−2 basedon 40,500 seconds of HST WFC3-IR R ∼ 140 grism spectroscopy. However this detectionrequired significant modeling of the data to remove overlapping spectra and backgroundartifacts. Subsequent re-analysis of the deeper imaging and spectroscopic data by Bouwenset al. (2013) now lead them to the conclusion that the galaxy is more likely at z ∼ 2 thanz ∼ 10−12 based on the argument that the galaxy would be unusually luminous if actuallyat z ∼ 12.

If UDFj-39546284 is at low redshift, the photometric constraints indicate it must be ayoung blue galaxy with strong line emission that has a line flux of ∼ 3×10−18 erg s−1 cm−2

for a typical dwarf galaxy velocity dispersion (Ellis et al. 2013, Brammer et al. 2013,Bouwens et al. 2013). Furthermore, the line reported by Brammer et al. (2013) falls ina region largely free of sky lines and so can be confirmed by ground based spectroscopywhich has significantly higher sensitivity and spectral resolution.

In this paper we present R∼ 3630 H-band spectroscopy of UDFj-39546284 with theMOSFIRE multi-object infrared spectrograph (McLean et al. 2010; 2012) on the Keck-Itelescope, reaching 2 − 3× the sensitivity of the WFC3 grism spectroscopy between skylines. This new instrument saw its first-light in April 2012, and provides a substantialboost in sensitivity relative to previous facilities for studies of very faint distant galaxies.Its high-multiplexing of up to 46 adjustable slitlets over a field of view of 6′×6′ enables thesimultaneous acquisition of scores of individual sources. In addition to UDFj-39546284 wegive a summary of the results from other objects on our slit mask in Table B.1.

We adopt a cosmological model with ΩΛ = 0.7, ΩM = 0.3, and h = 0.7 and magnitudesin the AB system.

B.3 Data

Data were collected on the nights of Jan 15 & 16, 2013 using the MOSFIRE instrumenton the Keck-I telescope. Conditions were photometric on both nights, with a medianseeing of 1.2′′ . The instrument was configured with the H band grating, 0.7′′ slit widths,180s exposures, and 16 Multiple Correlated Double Samples (MCDS). The telescope wasnodded by ±1.25′′ between observations with 44 exposures taken on Jan 15, and 48 onJan 16, yielding a total exposure time of 4.6h on the mask.

We used bright 2MASS stars for alignment, but noted a significant ∼ 1′′ offset betweenthe 2MASS and HUDF12 astrometry which was corrected before generating the mask.We verified the alignment stars and galaxies in the HUDF12 were on the same astrometricsystem by comparing the astrometry of the alignment stars and galaxies on the HUDF12images and the surrounding HST-ACS images which covered a wider area. Based onrepeated alignment exposures taken every 1-2 hours during both nights the mask wasaligned to better than 0.1′′ during the observations. Finally, to verify the mask wasproperly aligned, relatively bright objects were placed on slits around the mask and theirflux checked against the expectations from photometry (See Figure B.1).

The data were reduced using the MOSFIRE python reduction package which subtracts

85


CDFS21724

B

CDFS21970

C

CDFS23549

D

UDFj-39546284

A

A

B

C

D

WFC3/IR - F160W ACS - z850 ACS - z850 ACS - z850

z = 1.2

z = 1.3

z = 1.2

z = 11.9?

Fig. B.1 — Image cutouts and the 2d MOSFIRE spectra around UDFj-39546284 (A) and threeother bright objects are shown. The MOSFIRE slit positions are marked in yellow and theobjects highlighted with a cyan circle. Analysis of the bright compact-object spectra CDFS21724(B) indicates we are losing no more than 10% of the flux due to mask mis-alignment. The strongline visible in CDFS21970 (C) is Hα at z = 1.3089. A summary of these and other objects inthe mask is given in Table B.1.

nodded pairs of images, then shifts and co-adds the exposures using a sigma-clippednoise weighted mean. The Argon and Neon arc lamps along with sky lines were usedfor wavelength calibration. We generated a combined image of all 92 (4.6h) exposures aswell as combinations including only 3/4 of the data (3.45h) to test the robustness of thereductions, noise estimates, and for temporal variations .

Flux calibration was accomplished by taking spectra of the white dwarfspectrophotometric standard star GD71. The standard star was observed with a pair ofexposures using identical setting to the science observations and reduced in the same wayas the science data. The well detected spectra of the compact, 0.36′′ FWHM in the HSTF160W images, z = 1.22 galaxy CDFS21724 was used to verify our spectrophotometriccalibration, slit loss, and noise estimates by comparing this spectra to the ISAAC Hband flux from the MUSYC catalog (Cardamone et al. 2010), the GOODS-S EarlyRelease Science (ERS) F160W flux, and the noise estimates from the MOSFIRE exposuretime calculator. Based on the CDFS21724 spectra additional slit losses due to objectextent and mask mis-alignment are estimated to be 5 − 10% greater than estimated forthe standard star. The noise measured in CDFS21724 between 15920 − 15950 A is8.5 × 10−19 erg s−1 cm−2 per 4.4 A resolution element consistent with the estimate of8.1 × 10−19 erg s−1 cm−2 from the exposure time calculator once the slit losses of 48%due to poor seeing, and 10% due to object extension and mask alignment are taken intoaccount. Two other bright objects CDFS21970 and CDFS23549 were also placed on themask to verify the throughput and resulted in the expected signal-to-noise, but were notused for quantitative analysis because of a strong emission line (CDFS21970) and complex

86

B.3. Data

15000 15500 16000 16500 17000 17500 18000Wavelength (Å)

10-18

10-17

Sens

itivi

ty (

erg

s-1 c

m-2

)

Fig. B.2 — The measured 1σ sensitivity per 4.4A resolution element at the position of UDFj-39546284 is plotted. The measured sensitivity is consistent with that predicted by the instrumentexposure time calculator. Note that we should be able to detect a ∼ 3×10−18 erg s−1 cm−2lineimplied by the photometry at several sigma between sky lines.

morphology (CDFS23549) which made a comparison to UDFj-39546284 more difficult.The estimated sensitivity per resolution element is plotted in Figure B.2 and results forthe alignment galaxies and high-z spectra are given in Table B.1.

To find potential emission lines we used SExtractor to automatically search for groupsof 9 pixels above 1σ with no smoothing yielding a > 3σ net detection, as well as visuallyinspecting the spectra at the expected position of UDFj-39546284. We find no robustline detections, but do find a 2.2σ peak at 15985.5 ± 4.4 A with a flux of 1.4 ± 0.6 ×10−18 erg s−1 cm−2 in a 6-pixel diameter aperture (Figure B.3), consistent at the1.5σlevel with the wavelength and flux reported in Brammer et al. (2013). We also measuredthe line flux using an optimal extraction assuming a 1.2′′ gaussian FWHM and found thesame result. This corresponds to 70 ± 30% of the measured broad band flux, and wouldcorrespond to an observed frame equivalent width of ∼ 6400 A.

The peak is present in all of the reductions which include only 3/4 of the data (atreduced significance, see Figure B.3) and measurement of the flux at the positions of theexpected negative images due to the dithering return a consistent negative flux, increasingthe reliability of the detection. No known detector artifacts fall on this part of the detector,and no decaying latent images from alignment or previous observations are visible in theearly frames of each observation or in the un-dithered stack of the data. A 1.5σ negative”spot” corresponding to a region of known bad pixels at one dither position is observed just

87


Fig. B.3 — Top: A flux and Signal-to-Noise (SNR) map of the region around the potentialdetection reported by Brammer et al. (2013). The region allowed by Brammer et al. (2013)is marked by the red brackets, and the potential 2.2σ detection in our data is marked by ared arrow. The region used to subtract the sky from the negative spot below the detection isindicated with a yellow arrow, note the positive fluctuation at this position. Bottom: Signal-to-noise maps of the area around the possible line detection with a quarter of the data removedin each reduction. Note the possible line detection shown in Figure B.3 is present in all 4reductions with a significance of 1.3-2.6 σ consistent with the expected noise. In contrast, thenegative spot below the detection varies significantly, indicating it is due to noise. The regionsused to subtract the sky from the negative spot below the detection is indicated with a yellowarrow.

below the line. The bad pixels have been masked, but as a result the data at this positiononly comes from one dither position. This negative spot is likely a sky-subtraction artifactdue to a corresponding positive spot visible at the other dither position where the sky wasdetermined. This is not an astrophysical object because the expected dual negative, singlepositive pattern is not observed. Furthermore, this region varies in intensity in the 3/4reductions, indicating it is either a noise fluctuation, or a transient effect in the detector.The region does not produce the adjacent positive detection observed for UDFj-39546284in other masks taken on the same nights.

Despite the positive evidence, this detection should be considered tenuous because 291other > 2σ peaks are found between sky lines, 37 of which occur in the full stack and allfour 3/4 time stacks, and 4 of which have two negative detections at the two expecteddither positions. This places the probability of a chance co-incidence between a noise peak

88

B.4. Discussion

Tab. B.1 — Targeted objects

ID RA DEC HdAB Comments

CDFS21724b 03h32m35.636s −27d43m10.16s 20.49 low-z, Continuum SNR= 9.6CDFS21970b 03h32m35.972s −27d48m50.40s 21.04 low-z, Continuum SNR= 1.75, strong Hα lineCDFS23549b 03h32m38.107s −27d44m32.59s 20.41 low-z, Continuum SNR= 9.2UDFj-39546284a 03h32m39.54s −27d46m28.4s 29.3 z ∼ 12 J drop, no detectionUDF12-4106-7304a 03h32m41.06s −27d47m30.4s 29.7 z ∼ 10 J drop, no detectionUDF12-3947-8076a 03h32m39.47s −27d48m07.6s 29.0 z ∼ 10 J drop, no detectionUDFj-43696407a 03h32m43.69s −27d46m40.7s 29.5 z ∼ 10 J drop, no detectionUDFj-35427336a 03h32m35.42s −27d47m33.6s 29.6 z ∼ 10 J drop, no detectionUDFy-33436598a 03h32m33.43s −27d46m59.8s 29.4 z ∼ 8 Y drop, no detectionUDF12-3858-6150c 03h32m38.58s −27d46m15.0s 29.9 z ∼ 8 Y drop, no detectionUDF12-3939-7040c 03h32m39.39s −27d47m04.0s 28.9 z ∼ 8 Y drop, no detectionUDF12-4057-6436c 03h32m40.57s −27d46m43.6s 28.7 z ∼ 7 z drop, no detectionUDF12-3817-7327c 03h32m38.17s −27d47m32.7s 30.3 z ∼ 7 z drop, no detectionUDF12-3853-7519c 03h32m38.53s −27d47m51.9s 29.7 z ∼ 7 z drop, no detection

a Ellis et al. (2013)b Cardamone et al. (2010)c Bouwens et al. (2011b), Schenker et al. (2013)d WFC3-IR F160W for all source names starting with UDF, and ISAAC H for those starting with CDFS.

and the reported WFC3 grism observations at between 1% and 86% depending on wetherwe assume there are 4 or 291 > 2σ peaks. In addition the peak is near a sky line increasingthe chance of residuals.

Besides UDFj-39546284, we observed 4 J dropout, 3 Y dropout, and 3 z dropout galaxiestaken from Ellis et al. (2013) and Schenker et al. (2013) with details noted in Table B.1.The goal of including these sources was to search for strong emission lines if the sourceswere actually at z < 3 since all are expected to have Lyα blue ward of our spectroscopicrange. We find no lines at a sensitivity within 10% of that indicated in Figure B.2 placinguseful limits for future spectroscopic observations seeking Ly-α in the Y and J bands.

B.4 Discussion

Considering the possibility that UDFj-39546284 is a z << 10 galaxy, the HST observationsimply a line flux of 3−4.7×10−18 erg s−1 cm−2. If we combine our 2.2σ flux measurementwith the 2.7σ result of Brammer et al. (2013) it results in a 3.5σ detection at 15985.5±4.4 Awith a flux of 1.8±0.5×10−18 erg s−1 cm−2, at the lower-end of what is required to explainthe photometry. If we take this combined result at face value along with the photometricanalysis of Bouwens et al. (2013) it indicates UDFj-39546284 is likely at z ∼ 2.19, andwe are seeing the [OIII] 5007 A line. But, at the low significance of our data and that ofBrammer et al. (2013) confirmation bias due to co-incident noise peaks is problematic andsignificant unknown systematic effects could be present in both reductions. Furthermore,for typical [OIII]/Hβ line ratios of 7.3 in strong line emitting galaxies (Kakazu et al. 2007)one would expect a > 2σ detection in F140W filter imaging if this source were at z ∼ 2.19and this is not observed. In contrast to Brammer et al. (2013) we find even the extremeobject they find with a [OIII]/Hβ line ratio of 11.4 should also result in a ∼ 2σ detection

89


assuming the Ellis et al. (2013) photometric limits. The discrepancy is due either todifferences in the depths quoted between Brammer et al. (2013) and Ellis et al. (2013)which differ by ∼ 0.4 magnitudes and/or to how the z = 1.606 extreme object photometryand spectra were compared to the UDFj-39546284 limits. In particular, we note thespectral energy distribution shown in Brammer et al. (2013) is not directly comparableto the plotted UDFj-39546284 limits because the filters in the two observations havesignificantly different bandwidths and the spectral energy distribution is dominated bylines which create significant non-linear flux changes as they redshift through the filterbandpasses. An alternative low redshift explanation is that we are detecting one of the[OII]3727 A doublet lines at z = 3.29, however this is less likely since the [OII]line istypically significantly weaker than the [OII]5007 A line. So more data are required beforecoming to any firm conclusions.

We estimate ∼ 20h of integration in good conditions with Keck-I-MOSFIRE or anequivalent instrument would confirm the tentative detection at 15985.5 A at > 5σ, butstill be insufficient to detect other lines for typical [OIII]/Hβ line ratios in high-equivalentwidth line emitting galaxies (Kakazu et al. 2007). If the preliminary detection is the strong[OIII]5006.8 A line at z = 2.19 the 4958.9A line would be behind a bright sky line and Hβwould be too faint to detect. If instead we are seeing the less likely 3727 A [OII]doubletat z = 3.29 the second line of the doublet would be hidden behind the nearby sky line.

One could also integrate for a similar amount of time in the K band, and attempt todetect the Hα line if the source is at z = 2.19 or both of the [OIII]lines if the source is atz = 3.29. Assuming our preliminary detection is real, all three of these lines fall betweensky lines and should be detectable at > 5σ for typical line ratios. This may be a moreproductive approach since it can differentiate between z = 2.19, 3.29, and 12.14.

Spectroscopic confirmation with ALMA would be expensive. Assuming this sourceis at z ∼ 12, is un-obscured, and all of the continuum flux is due to star formation,the expected CII 158µm luminosity would be ∼ 50µJy assuming 1% of the bolometricluminosity is emerging in this line. With full ALMA this would require ∼ 72 hours ofintegration in one receiver tuning to detect at 5σ. However, a non-detection would notconfirm a low-redshift because of the intrinsic scatter in bolometric luminosity to CII lineratio.

If UDFj-39546284 is at high redshift we expect no detection in the longer spectra and itis unlikely that any conclusive spectra can be obtained with current instrumentation. Thetypical equivalent width of Lyα in 4 < z < 6 galaxies is ∼ 20 A with a tail out to ∼ 80 Aand is expected to decrease at z ∼ 6−12 due to the increasing opacity of the inter-galacticmedium (IGM) (Stark et al. 2011, Mallery et al. 2012). Assuming UDFj-39546284 is atz ∼ 12, our flux limit would require a Ly-α equivalent width of ≥ 200 A to yield a 2.2σdetection. This large an equivalent width is occasionally observed in the z ∼ 4−6 universe,but less likely at z ∼ 12 where the inter-galactic medium is expected to be more absorptive.For a normal equivalent width of ∼ 20 A, ∼ 460h of observation with Keck-I-MOSFIREwould be required to detect this galaxy, and even an adaptive optics assisted thirty-meterclass telescope would require ∼ 16h of integration.

90

B.5. Conclusions

B.5 Conclusions

Using the Keck-I MOSFIRE infrared spectrograph we recover the 2.7σ line detectionreported in Brammer et al. (2013) at the 2.2σ level. Combining these results we find3.5σ line a 15985.5± 4.4 A with an estimated flux of 1.8± 0.5× 10−18 erg s−1 cm−2, butthe likelihood that this is a chance coincidence of noise peaks is non-negligible, so furtherobservations are needed. If confirmed, the detection indicates UDFj-39546284 is actuallya low redshift line emitter at z ∼ 2.19 or 3.29, so both H and K band spectra shouldbe obtained to uniquely determine its redshift. If the source is at z ∼ 12 it is unlikelyground based 8-10m telescopes or ALMA could confirm it and additional data will yielda non-detection. The difficulty in interpreting this source, and the high likelihood that itis actually at low-redshift highlights the need for spectroscopic studies at 6 < z < 8 tounderstand possible sources of contamination at even higher redshifts.

91


92

Chapter C

Dust attenuation in high redshiftgalaxies: “Diamonds in the Sky”

The content of this chapter is based onN. Z. Scoville, A. L. Faisst, P. Capak, et al. (2014)

C.1 Summary

We use observed optical to near infrared spectral energy distributions (SEDs) of 266galaxies in the COSMOS survey to derive the wavelength dependence of the dustattenuation at high redshift. All of the galaxies have spectroscopic redshifts in therange z = 2 to 6.5. The presence of the CIV absorption feature, indicating that therest-frame UV-optical SED is dominated by OB stars, is used to select objects for whichthe intrinsic, unattenuated spectrum has a well-established shape. Comparison of thisintrinsic spectrum with the observed broadband photometric SED then permits derivationof the wavelength dependence of the dust attenuation. The derived dust attenuationcurve is similar in overall shape to the Calzetti curve for local starburst galaxies. Wealso see the 2175 A bump feature which is present in the Milky Way and LMC extinctioncurves but not seen in the Calzetti curve. The bump feature is commonly attributed tographite or PAHs. No significant dependence is seen with redshift between sub-samplesat z = 2− 4 and z = 4− 6.5. The “extinction” curve obtained here provides a firm basisfor color and extinction corrections of high redshift galaxy photometry.

C.2 Introduction

Dust profoundly affects the light emitted from galaxies − requiring large corrections forextinction in the UV and optical. This causes a severe degeneracy in the derived ages ofhigh redshift galaxies and hence results in a major uncertainty in their derived properties.The extinction has a major effect on the observability of the galaxies and on their

93

CHAPTER C. Dust attenuation in high redshift galaxies: “Diamonds in the Sky”

derived stellar masses, star formation rates and luminosities. At early epochs there is anincreasing fraction of active star forming galaxies compared to the low redshift local galaxypopulation. The observed properties of these star forming galaxies are strongly dependenton the dust extinction, since many of the surveys probe the rest frame UV, where the dustextinction is higher than at visible wavelengths. The amount of dust extinction is almostalways inferred from the differential extinction or reddening. Translating this reddeninginto an estimate of the total extinction at any wavelength then requires knowledge of thedust extinction curve as a function of wavelength.

At low redshift the extinction curve has been determined along many lines-of-sightin the Galaxy (cf. Savage & Mathis 1979) and in the lower metallicity Large MagellanicCloud (LMC) and Small Magellanic Cloud (SMC) (Prevot et al. 1984, Fitzpatrick 1986).The mean extinction curves for each of these galaxies are quite different: the SMC curvehas the steepest power-law form and no 2175 A bump, whereas the LMC and the MilkyWay extinction curves both show the 2175 A feature and the Milky Way has the flattestoverall wavelength dependence (Gallerani et al. 2010). The SMC has the lowest metallicityof the three galaxies and this probably results in a different grain composition and/orsize distribution. These large differences underscore the critical importance of empiricaldetermination of the reddening curves at high redshifts. An effective extinction curve forthe integrated light from galaxies was obtained by Calzetti et al. (1994; 1996) and this iscommonly used for analysis of high redshift galaxies.

Direct determination of the differential extinction curve in high-z galaxies has beendifficult using just photometry since one needs to break the age/reddening degeneracyin the observed galaxies and this can’t be done without prior assumption of a form forthe extinction curve. Kriek & Conroy (2013) fit a grid of SED templates and power-lawattenuation curves to the photometry of a sample of galaxies at z = 0.5 to 2. They usedthe broad- and intermediate-band photometry to classify the intrinsic SED type of eachgalaxy and then correlated the extinction curve slope and the 2175 A bump strength withgalaxy type. A similar procedure, fitting a library of SED templates, was employed byBuat (2013) to analyze a sample of galaxies at z = 0.95− 2.2. Their favored attenuationcurve was similar to that in the LMC super-shell region.

The approach we advocate here relies upon the intrinsic or un-extincted spectrum of thestellar population being known or identified a priori from high resolution spectroscopicfeatures. Then, the differential extinction can be easily derived by comparison of theobserved spectral energy distributions (obtained from broadband photometry) with thatof the intrinsic spectrum. In fact, there is a UV spectral signature which clearly identifiesthe nature of the intrinsic spectrum − this is the CIV absorption feature at λ = 1549 A.The CIV absorption indicates the presence of type O and supergiant B stars and whenthe CIV absorption is strong, such stars are dominating the UV continuum emission.

The CIV absorption feature in spectra of high redshift galaxies in the COSMOS surveythus selects objects for which the broadband photometry can be used to determine thedifferential extinction curve. For such galaxies, the CIV absorption implies that theUV-optical spectral energy distribution (SED) is dominated by OB stars and the overallslope of the intrinsic (un-extincted) continuum SED is known.

We use a sample of 266 galaxies at z = 2 to 6.5 for which rest-frame UV spectroscopy

94

C.2. Introduction

shows the CIV absorption to analyze the wavelength dependence of the dust attenuation.Starburst (SB) model simulations in Section C.3 define the age range (up to ∼50 Myr) ofstellar populations for which the CIV absorption can be used as a signpost for the overallSED. In Section C.4 we describe the galaxy sample used for this study and then outlinethe numerical technique in Section C.5. The derived attenuation curve is presented inSection C.6 and discussed in Section C.7.

CIV 1549 Å!Balmer break!

Fig. C.1 — Simulated spectra for constant star formation rate (SFR) starbursts of duration 20to 500 Myr obtained using Starburst99 with a Kroupa IMF and solar metallicity. The strongCIV absorption at λ = 1549 A seen in these spectra requires significant SF within the previous10 Myr since this feature is produced in O-type stars. The overall shape of the SED is relativelyinvariant for the first ∼ 50 Myr; after that time, the Balmer break at λ = 3646 A becomesmore pronounced and thus causes the broadband SED to change in shape. Any objects witha Balmer break larger than 25%, corresponding to spectra with ages exceeding that of themodel shown with 100 Myr age were removed from our sample. Here we use the combination− strong CIV absorption plus the absence of a strong Balmer/4000 A break – to select sourcesto illuminate the dust attenuation.

95


C.3 CIV Absorption as a Signpost for the SED

In Figure C.1 spectra of starburst stellar populations are shown for constant star formationrates extending from 20 to 500 Myr. These spectra were generated using the Starburst99simulation tool (Leitherer et al. 1999) with solar metallicity and a Kroupa initial massfunction for the stars. The strongest spectral feature in these plots is the CIV absorptionat λ = 1549 A − this absorption is produced in the atmospheres and outflow winds ofO stars, and is present as long as O-type stars dominate the rest frame UV emission.In an instantaneous starburst, the feature will diminish rapidly since the typical lifetimeof those stars is ≤ 10 Myr. Hence, this feature is a robust signpost of on-going OB starformation. Then, the UV-optical continuum, dominated by OB stars, has a quite constantSED shape.

At longer wavelengths, the dominant spectral feature is the Balmer break at λ = 3646 A.As the stellar populations ages, or as the fraction of intermediate mass stars grows, thisfeature becomes more pronounced, causing the optical SED shape to evolve – specificallybecoming redder at later epochs.

At early epochs (. 20−100 Myr) the spectra are very constant in both broadband shapeand spectral features (see Figure C.1) and such sources may be used as a background sourceto probe the wavelength dependence of the foreground dust attenuation. The signatures ofappropriate galaxies to select are: 1) the detection of CIV absorption and 2) the absenceof a strong Balmer/4000 A break feature. The CIV absorption feature at λ = 1549 A isalso seen in type 2 AGN sources, but these can be recognized using AGN signatures suchas X-ray and strong radio emission or nuclear point sources. These discriminants againstAGN may not catch all AGN at the highest redshifts given the limited sensitivity of X-rayand radio data. The CIV absorption is very strong and hence easily detected. For aKroupa IMF, the CIV equivalent width is ∼ 9.5 A at early times; it can therefore beeasily seen down to 1/10 solar metallicity. This was confirmed by additional starburstsimulations not shown here.

C.4 Galaxy Sample Selection

Our source sample is drawn from the COSMOS survey field, with the primary selectionbeing those objects with spectroscopic coverage from our Keck DEIMOS survey (Scoville,PI). This survey includes ∼ 2300 galaxies at z = 2 to 6.5 down to IAB = 25 mag. From thissample, we then selected only those objects with the CIV absorption feature and at least6 broadband continuum photometric detections (average is 10.3 photometric bands foreach source). The redshifts were determined from the CIV absorption plus other emissionlines and we required at least two lines for a reliable spectroscopic redshift. As noted inSection C.3, we then removed from further consideration any objects exhibiting a Balmerbreak larger than 25% in flux. The final fitting included 266 objects at z = 2 to 6.5 andthen subsamples of 135 and 132 objects at z = 2− 4 and 4− 6.51. The redshift and stellar

1The source list is available in an ASCII table from the COSMOS data archive at IPAC/IRSA:http://irsa.ipac.caltech.edu/data/COSMOS/tables/extinction/extinction_curve_source_list.

dat.

96

http://irsa.ipac.caltech.edu/data/COSMOS/tables/extinction/extinction_curve_source_list.dat

http://irsa.ipac.caltech.edu/data/COSMOS/tables/extinction/extinction_curve_source_list.dat

C.5. Numerical Solution

Fig. C.2 — Left: The spectroscopic redshift distribution of galaxies in our sample. Right:The stellar mass distribution of galaxies in our sample. The masses were derived from SEDtemplate fitting to the photometry with a constant SFH, a Chabrier IMF (similar to Kroupa)and a metallicity range from Z = 0.005, 0.001, 0.01.

mass distributions of sources are shown in Figure C.2.

The COSMOS photometry used here includes the 34 band photometry used for theCOSMOS photometric redshifts (Ilbert et al. 2013) and the near infrared deep photometryfrom UltraVista (McCracken et al. 2012) and Spitzer SPLASH (Steinhardt et al. 2014).For the attenuation curve analysis here, we use 14 broad bands (uCFHT , BSubaru, VSubaru,gSubaru, rSubaru, iSubaru, zSubaru, ICFHT , IRAC1, IRAC2, YUltra−V ista, JUltra−V ista,HUltra−V ista, KUltra−V ista) which have the highest signal-to-noise ratios (see Table 1 -Ilbert et al. 2009). At least 6 bands were required in the rest wavelength range betweenLyα (1216 A) and 10000 A. In addition, we required at least one detection on eachside of the Balmer/4000 A break. For rest wavelengths short of Lyα there is significantintergalactic medium HI absorption at the higher redshifts. At these wavelengths wecorrected the photometry to brighter values using the standard Madau correction (Madau1995). This correction has a very large dispersion (see Madau 1995) and we found thatit clearly overcorrected the photometry, as indicated from the fact that the correctedfluxes short of Lyα exceeded those of adjacent bands just longward of Lyα . Reducingthe Madau correction by 50% avoids this. The derived dust attenuation at the one pointshort of 1216 A should therefore be viewed with some uncertainty since it hinges on thevalues of the IGM HI correction. Sample SEDs for two such objects at z = 2.01 and 4.93are shown on the left in Figure C.3 together with the constant rate starburst spectra for50 and 500 Myr duration for reference.

97


Fig. C.3 — Left: The rest frame SEDs are shown for two sources at z = 2.01 and 4.93. Theconstant rate SB spectra are shown for ages of 50 and 500 Myr for comparison. The dash anddotted lines correspond to 1216 A and 4000 A. Right: The best-fit SEDs using the derivedwavelength dependence of the attenuation (normalized to the reference wavelength λ = 1300 A)and an absolute value of the optical depth τ

1300A= 1.7 (coincidentally in both instances).

Although the best-fit curve extends to 950 A, these plots limit the spectral range of the curveto wavelengths for which each source has photometry.

C.5 Numerical Solution

We assume that the dust attenuation has a wavelength dependence common to all sourcesin each redshift sample, but that the absolute value of the dust optical depth at afiducial wavelength is freely varying between sources. Thus, the wavelength dependenceis characterized by a step function (τ(λi)) with discrete wavelength bins spanning restframe λ = 950 and 10000 A. This form of fitting has the virtue of making noassumption or analytic parametrization of the extinction. In addition, there is an absolute

98

C.5. Numerical Solution

Fig. C.4 — The distribution of derived τ1300A

is shown for the full sample at z = 2 to 6.5.

source-dependent scale factor for the optical depth at rest frame λ = 1300 A (τ1300A

).Thus,

τλ(source) = τ1300A

(source) τ(λi). (C.1)

For the normalized opacity function (τ(λi)), we use 9 bins spaced logarithmically inwavelength. If the rest-frame specific luminosity of the unobscured starburst is denotedas lν(SB), then the model we fit to the observed SEDs is given by

sν = e−τλ(source) L0(source) lν(SB). (C.2)

The underlying source SED (lν(SB)) is taken to be the 50 Myr starburst spectrumshown in Figure C.1 based on our deselection of sources with larger Balmer Break, asmeasured from the best-fit stellar population model to the photometry. We did experimentwith using the 20 and 100 Myr SED and they did not improve the resultant best fit χ2.

For the z = 2 to 6.5 sample of 266 galaxies, the number of free parameters to be solvedfor in minimized χ2 is: 266 (L0(source)) + 266 (τ

1300A(source)) + 9 (wavelength bins) =

541 parameters. For this minimization we used the Levenberg-Marquardt least-squaresminimization routine (MPFIT in IDL). We adopted uncertainties of 10% for each SEDdata point; this was done to give all SED points similar weight in the fitting. Thus, weavoid overweighting a single band which has high signal-to-noise ratio but which does notprovide much wavelength leverage. This ensures that the IRAC data (3.6µm and 5.8µm)constrain the fit at long wavelengths. The least-squares fitting was done in two passes:first using the complete sample of galaxies and then removing the few objects (∼ 10)which had T1300 > 5; these latter sources are likely to have low signal-to-noise ratio at

99


Fig. C.5 — The wavelength dependence of the attenuation is shown for the full sample spanning z= 2 to 6.5 with uncertainties determined from the Levenberg-Marquardt routine. Also shown arethe extinction curves for the Milky Way (Seaton 1979), LMC (Fitzpatrick 1986), SMC (Prevotet al. 1984) and local starburst galaxies (Calzetti et al. 2000). The wavelength dependencederived here for galaxies at z > 2 is similar to that of Calzetti curve derived for starburst galaxiesbut also includes the 2175 A bump feature as seen in the Milky Way and LMC extinction curves.

the short wavelengths. The derived results were not significantly different after the secondpass except that the χ2 was improved slightly (≤ 10%). Figure C.4 shows the distributionof derived τ

1300Aof the final solution.

C.6 Results

The derived spectral dependence of the normalized dust attenuation is shown in Figure C.5for the full sample at z = 2 to 6.5. The uncertainties on the attenuation curve shown inthe figure are from the Levenberg-Marquardt algorithm. The χ2 for the fits in the threeredshift ranges are given in Table C.1 are 3.7, 1.9 and 6.6. For the fitting, the uncertaintieson each flux measurement were set to 10%, thus the derived fits typically disperse only13− 25% from the observed fluxes on all of the individual sources.

Also shown are the extinction curves for the Milky Way, LMC and SMC plus thedust attenuation curve for starburst galaxies from Calzetti et al. (2000). All curves

100

C.7. Summary and Comments

Fig. C.6 — The wavelength dependence of the attenuation is shown for sub-samples at z = 2−4(Left) and z = 4− 6.5 (Right). Given the larger uncertainties in the z = 4− 6.5 fit, we do notthink the difference in the 2175 A bump feature is significant. The overall broadband wavelengthdependencies are more consistent with the MW, LMC and SB curves and not with the SMCcurve.

were normalized to the their values at 1300 A. The overall wavelength dependence ofthe attenuation curve for z = 2 to 6.5 is remarkably similar to that of the low redshiftattenuation curve for the SB galaxies. It is also very similar to the extinction curves for theMilky Way and LMC, but not the SMC which has a much stronger wavelength dependencein the UV. Table C.1 lists the numeric values for the derived attenuation curves.

The high redshift curve (Figure C.5, Table C.1) also clearly shows the presence ofthe 2175 A bump feature seen in the Milky Way and the LMC extinction curves butnot in the starbursts or the SMC. This feature is often attributed to graphite (hencediamonds in our title) grains and its absence in the SMC extinction curve is probablyrelated to the low metallicity of SMC, resulting in different grain properties. The absenceof the 2175 A bump feature in the Calzetti curve (derived for local starburst galaxies) iscertainly significant in Calzetti et al. (1994) (see figure 17 there).

In order to probe the redshift evolution of the attenuation, we also obtained solutionsfor sub-samples of the galaxies at z = 2 − 4 and z = 4 − 6.5. The results are shown inFigure C.6. Their attenuation curves do not appear significantly different from that shownin Figure C.5 for the full sample. The z = 4− 6.5 curve exhibits a somewhat larger widthfor the 2175 A bump feature but we don’t consider this significant given the uncertainties.

C.7 Summary and Comments

Our technique uses galaxies selected spectroscopically with strong CIV absorption andphotometrically with a small Balmer/4000 A break feature to yield a sample of galaxiesfor which the un-extincted UV-optical SEDs are dominated by OB stars. These SEDs

101


Tab. C.1 — Best-fit Attenuation Curve

λ (A) z = 2− 6.5 z = 2− 4 z = 4− 6.5

950 1.386±0.060 1.406±0.053 1.431±0.0261250 1.025±0.041 1.028±0.035 0.999±0.0241500 0.915±0.035 0.904±0.029 1.003±0.0251800 0.862±0.033 0.850±0.026 0.924±0.0302250 0.817±0.030 0.826±0.025 0.865±0.0432800 0.678±0.023 0.706±0.019 0.716±0.0613900 0.501±0.015 0.530±0.012 0.509±0.0736350 0.266±0.016 0.363±0.011 0.255±0.09210000 0.158±0.017 0.181±0.014 0.127±0.096

# of galaxiesa 266 135 132# of phot. bandsb 2747 1614 1133χ2 3.68 1.92 6.63

Note: As described in the text, the least-squares fitting solved for the wavelength dependence as a stepfunction so that each wavelength point is independent, and smoothness of continuity is not enforcedexcept if warranted by the data. The attenuation curves were each normalized to unity at λ = 1330 A.a Number of galaxies in the final samples for the fit. The samples were slightly larger on the first passfitting; after the first pass 5-15 galaxies with τ

1300A> 5 were removed for the second pass fitting since

their short wavelength points would have low signal-to-noise ratio. Removing these sources reduced theχ2 but did not significantly change the shape of the attenuation curve.b Total number of photometric band fluxes used in fitting.

have a nearly constant shape since their emission over the entire wavelength is from asingle class of stars. Our sample of galaxies taken from the large COSMOS survey field;it consists of 266 sources after removing sources with incomplete wavelength coverage.

The derived attenuation is very similar to the Calzetti et al. (2000) curve obtainedfor 40 low redshift starburst galaxies. The curve shown in Figure C.5 clearly shows the2175 A bump feature; this was not present in the Calzetti curve, but is seen in both theMilky Way and LMC. Our curve is tabulated in Table C.1 for use in correcting colors andextinctions from high redshift galaxy surveys − both star forming and non-star forminggalaxies.

It is important to remember that the curves shown here for MW, LMC and SMCwere all obtained from line-of-sight extinction measurements of stars. In contrast, thestarburst curve (Calzetti et al. 2000) and that derived here for high redshift are “effectiveattenuation” curves − they are obtained from the integrated galaxy light. The formerextinction curves include the effects of both dust absorption and scattering while theeffective attenuation curves may contain a lesser contribution from dust scattering (sincesome photons are scattered back into our line-of-sight). The attenuation curves also includethe very important effect of partial covering of the galactic continuum by dust in theline-of-sight. This is of course not relevant to extinction measurements of individual stars.

Two other instances where the high redshift attenuation/extinction curve has beenanalyzed are for GRB sources and QSOs. Elıasdottir et al. (2009) obtained an extinctioncurve for GRB 070802 at z = 2.45 which was similar to the LMC extinction curve shownin Figure C.5, including the 2175 A bump at high signal-to-noise ratio. Perley et al. (2011)also clearly detected the 2175 A feature in GRB 080607 at z = 3.03.

102

C.7. Summary and Comments

In contrast to the GRBs and the starburst galaxies analyzed here, the high redshiftQSOs show quite different extinction curves. They generally resemble the SMC extinctioncurve (Richards et al. 2003, Reichard et al. 2003, Hopkins et al. 2004), not the LMC, MilkyWay or starburst curves. This said, Gallerani et al. (2010) analyzed the dust extinctionin a sample of 39 quasars at z = 3.9 to 6.4 and found a substantially flatter extinctioncurve at λ < 1300 A than the SMC curve. In fact, their extinction curve is flatter thanthe Calzetti curve at λ < 2500 A, implying that it is also flatter than that obtained hereat high redshift. They see no evidence of the 2175 A bump feature, although in many oftheir quasars this feature might be difficult to detect due to the broad QSO emission lines.In the quasars, the obscuration occurs from three distinct regions: the inner region withgas only seen in the X-rays, the mid-IR very hot dust within a few parsec of the AGNand the host galaxy dust (Elvis 2012). In the first two regions, the dust abundance andcomposition is likely to be very different.

An alternative, future application of the technique developed here would be possible ifhigh quality spectra became available over the UV-opt wavelength range on a small numberof bright sources. Then, accurate measures of the equivalent width of the CIV spectralfeature and the Balmer break strength could be used select more precisely the source SEDtemplates for each individual source.

103


104

Bibliography

Abadi, M. G., Moore, B., & Bower, R. G. 1999, MNRAS, 308, 947

Arnouts, S., et al. 2002, MNRAS, 329, 355

Atek, H., et al. 2010, ApJ, 723, 104

—. 2011, ApJ, 743, 121

Autry, R. G., et al. 2003, in Society of Photo-Optical Instrumentation Engineers (SPIE)Conference Series, Vol. 4841, Society of Photo-Optical Instrumentation Engineers(SPIE) Conference Series, ed. M. Iye & A. F. M. Moorwood, 525–539

Baldry, I. K., Balogh, M. L., Bower, R. G., Glazebrook, K., Nichol, R. C., Bamford, S. P.,& Budavari, T. 2006, MNRAS, 373, 469

Baldry, I. K., Glazebrook, K., Brinkmann, J., Ivezic, Z., Lupton, R. H., Nichol, R. C., &Szalay, A. S. 2004, ApJ, 600, 681

Balogh, M. L., Baldry, I. K., Nichol, R., Miller, C., Bower, R., & Glazebrook, K. 2004,ApJ, 615, L101

Balogh, M. L., Navarro, J. F., & Morris, S. L. 2000, ApJ, 540, 113

Bauer, A., Baltay, C., Coppi, P., Ellman, N., Jerke, J., Rabinowitz, D., & Scalzo, R. 2009,ApJ, 696, 1241

Behroozi, P. S., Marchesini, D., Wechsler, R. H., Muzzin, A., Papovich, C., & Stefanon,M. 2013, ApJ, 777, L10

Belli, S., Jones, T., Ellis, R. S., & Richard, J. 2013, ApJ, 772, 141

Belli, S., Newman, A. B., & Ellis, R. S. 2014a, ArXiv e-prints (1409.7088)

Belli, S., Newman, A. B., Ellis, R. S., & Konidaris, N. P. 2014b, ApJ, 788, L29

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

Birnboim, Y., & Dekel, A. 2003, MNRAS, 345, 349

Birnboim, Y., Dekel, A., & Neistein, E. 2007, MNRAS, 380, 339

Birrer, S., Lilly, S., Amara, A., Paranjape, A., & Refregier, A. 2014, ApJ, 793, 12

105

BIBLIOGRAPHY

Blanton, M. R., et al. 2003, ApJ, 594, 186

Bond, N. A., Gawiser, E., & Koekemoer, A. M. 2011, ApJ, 729, 48

Bond, N. A., et al. 2014, ArXiv e-prints (1403.7463)

Bouche, N., et al. 2010, ApJ, 718, 1001

Bournaud, F., Elmegreen, B. G., & Elmegreen, D. M. 2007, ApJ, 670, 237

Bouwens, R. J., Illingworth, G. D., Franx, M., & Ford, H. 2007, ApJ, 670, 928

Bouwens, R. J., et al. 2010, ApJ, 709, L133

—. 2011a, Nature, 469, 504

—. 2011b, ApJ, 737, 90

—. 2012, ApJ, 754, 83

—. 2013, ApJ, 765, L16

Bradley, L. D., et al. 2012, ApJ, 760, 108

Brammer, G. B., van Dokkum, P. G., Illingworth, G. D., Bouwens, R. J., Labbe, I., Franx,M., Momcheva, I., & Oesch, P. A. 2013, ApJ, 765, L2

Brammer, G. B., et al. 2012, ApJS, 200, 13

Brennan, R., et al. 2015, ArXiv e-prints (1501.06840)

Brinchmann, J., Charlot, S., White, S. D. M., Tremonti, C., Kauffmann, G., Heckman,T., & Brinkmann, J. 2004, MNRAS, 351, 1151

Bruce, V. A., et al. 2014, MNRAS, 444, 1660

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

Buat, V. 2013, Earth, Planets, and Space, 65, 1095

Buitrago, F., Trujillo, I., Conselice, C. J., Bouwens, R. J., Dickinson, M., & Yan, H. 2008,ApJ, 687, L61

Bunker, A. J., Caruana, J., Wilkins, S. M., Stanway, E. R., Lorenzoni, S., Lacy, M.,Jarvis, M. J., & Hickey, S. 2013, MNRAS, 430, 3314

Calzetti, D., Armus, L., Bohlin, R. C., Kinney, A. L., Koornneef, J., & Storchi-Bergmann,T. 2000, ApJ, 533, 682

Calzetti, D., Kinney, A. L., & Storchi-Bergmann, T. 1994, ApJ, 429, 582

—. 1996, ApJ, 458, 132

Cameron, E., & Driver, S. P. 2007, MNRAS, 377, 523

106

BIBLIOGRAPHY

Capak, P., Faisst, A., Vieira, J. D., Tacchella, S., Carollo, M., & Scoville, N. Z. 2013, ApJ,773, L14

Capak, P., et al. 2007, ApJS, 172, 99

—. 2011, ApJ, 730, 68

Cardamone, C. N., et al. 2010, ApJS, 189, 270

Carollo, C. M., et al. 2013a, ApJ, 773, 112

—. 2013b, ApJ, 776, 71

—. 2014, ArXiv e-prints (1402.1172)

Caruana, J., Bunker, A. J., Wilkins, S. M., Stanway, E. R., Lacy, M., Jarvis, M. J.,Lorenzoni, S., & Hickey, S. 2012, MNRAS, 427, 3055

Caruana, J., Bunker, A. J., Wilkins, S. M., Stanway, E. R., Lorenzoni, S., Jarvis, M. J.,& Ebert, H. 2014, MNRAS, 443, 2831

Caruana, J., Bunker, A. J., Wilkins, S. M., Stanway, E. R., Lorenzoni, S., Jarvis, M. J.,& Elbert, H. 2013, arXiv:1311.0057 (1311.0057)

Cassata, P., et al. 2010, ApJ, 714, L79

Cattaneo, A., Dekel, A., Faber, S. M., & Guiderdoni, B. 2008, MNRAS, 389, 567

Cen, R. 2014, ApJ, 781, 38

Chabrier, S. 2003, Archives de Pediatrie, 10, 168

Cibinel, A., et al. 2013a, ApJ, 777, 116

—. 2013b, ApJ, 776, 72

Cimatti, A., et al. 2008, A&A, 482, 21

Clement, B., et al. 2012, A&A, 538, A66

Cole, S., et al. 2001, MNRAS, 326, 255

Cooke, J., Ryan-Weber, E. V., Garel, T., & Dıaz, C. G. 2014, MNRAS, 441, 837

Cowie, L. L. 1988, in Post-Recombination Universe, ed. N. Kaiser & A. N. Lasenby, 1–18

Croton, D. J., et al. 2006, MNRAS, 365, 11

Curtis-Lake, E., et al. 2012, MNRAS, 422, 1425

Daddi, E., Cimatti, A., Renzini, A., Fontana, A., Mignoli, M., Pozzetti, L., Tozzi, P., &Zamorani, G. 2004, ApJ, 617, 746

Daddi, E., et al. 2005, ApJ, 626, 680

107

BIBLIOGRAPHY

—. 2007, ApJ, 670, 156

Dave, R., Finlator, K., & Oppenheimer, B. D. 2012, MNRAS, 421, 98

De Lucia, G., Weinmann, S., Poggianti, B. M., Aragon-Salamanca, A., & Zaritsky, D.2012, MNRAS, 423, 1277

Dekel, A., & Burkert, A. 2014, MNRAS, 438, 1870

Dekel, A., Sari, R., & Ceverino, D. 2009, ApJ, 703, 785

Dessauges-Zavadsky, M., et al. 2014, ArXiv e-prints (1408.0816)

Dijkstra, M., Haiman, Z., & Loeb, A. 2004, ApJ, 613, 646

Dijkstra, M., Mesinger, A., & Wyithe, J. S. B. 2011, MNRAS, 414, 2139

Dijkstra, M., & Wyithe, J. S. B. 2010, MNRAS, 408, 352

—. 2012, MNRAS, 419, 3181

Dijkstra, M., Wyithe, J. S. B., & Haiman, Z. 2007, MNRAS, 379, 253

Dijkstra, M., Wyithe, S., Haiman, Z., Mesinger, A., & Pentericci, L. 2014, MNRAS, 440,3309

Dressler, A. 1980, ApJ, 236, 351

Driver, S. P., et al. 2011, MNRAS, 413, 971

Duncan, K., et al. 2014, MNRAS, 444, 2960

Elbaz, D., et al. 2007, A&A, 468, 33

Elıasdottir, A., et al. 2009, ApJ, 697, 1725

Ellis, R. S., et al. 2013, ApJ, 763, L7

Elmegreen, D. M., Elmegreen, B. G., Ravindranath, S., & Coe, D. A. 2007, ApJ, 658, 763

Elvis, M. 2012, Journal of Physics Conference Series, 372, 012032

Erb, D. K., Shapley, A. E., Pettini, M., Steidel, C. C., Reddy, N. A., & Adelberger, K. L.2006, ApJ, 644, 813

Faber, S. M., et al. 2003, in Society of Photo-Optical Instrumentation Engineers (SPIE)Conference Series, Vol. 4841, Instrument Design and Performance for Optical/InfraredGround-based Telescopes, ed. M. Iye & A. F. M. Moorwood, 1657–1669

Fan, X., et al. 2001, AJ, 122, 2833

—. 2006, AJ, 132, 117

Feldmann, R., Carollo, C. M., & Mayer, L. 2011, ApJ, 736, 88

108

BIBLIOGRAPHY

Feldmann, R., Carollo, C. M., Mayer, L., Renzini, A., Lake, G., Quinn, T., Stinson, G. S.,& Yepes, G. 2010, ApJ, 709, 218

Feldmann, R., Carollo, C. M., Porciani, C., Lilly, S. J., & Oesch, P. 2008, ArXiv e-prints(0801.3275)

Finkelstein, S. L., et al. 2012, ApJ, 756, 164

Fitzpatrick, E. L. 1986, AJ, 92, 1068

Fontana, A., et al. 2010, ApJ, 725, L205

Forster Schreiber, N. M., et al. 2011, ApJ, 739, 45

Franx, M., van Dokkum, P. G., Schreiber, N. M. F., Wuyts, S., Labbe, I., & Toft, S. 2008,ApJ, 688, 770

Gallerani, S., et al. 2010, A&A, 523, A85

Genzel, R., et al. 2014, ApJ, 785, 75

—. 2015, ApJ, 800, 20

Gezari, S., et al. 2009, ApJ, 690, 1313

Gobat, R., et al. 2012, ApJ, 759, L44

—. 2013, ApJ, 776, 9

Gonzalez, V., Labbe, I., Bouwens, R. J., Illingworth, G., Franx, M., & Kriek, M. 2011,ApJ, 735, L34

Goto, T., Utsumi, Y., Hattori, T., Miyazaki, S., & Yamauchi, C. 2011, MNRAS, 415, L1

Grogin, N. A., et al. 2011, ApJS, 197, 35

Gronwall, C., et al. 2007, ApJ, 667, 79

Guhathakurta, P., Tyson, J. A., & Majewski, S. R. 1990, ApJ, 357, L9

Gunn, J. E., & Gott, III, J. R. 1972, ApJ, 176, 1

Gunn, J. E., Hoessel, J. G., & Oke, J. B. 1986, ApJL, 306, 30

Guo, Y., et al. 2014, ArXiv e-prints (1410.7398)

Haas, M. R., Schaye, J., Booth, C. M., Dalla Vecchia, C., Springel, V., Theuns, T., &Wiersma, R. P. C. 2013, MNRAS, 435, 2931

Haiman, Z., & Spaans, M. 1999, ApJ, 518, 138

Hartley, W. G., et al. 2013, MNRAS, 431, 3045

Hayes, M., Laporte, N., Pello, R., Schaerer, D., & Le Borgne, J.-F. 2012, MNRAS, 425,L19

109

BIBLIOGRAPHY

Hearin, A. P., Zentner, A. R., Berlind, A. A., & Newman, J. A. 2013, MNRAS, 433, 659

Hemmati, S., et al. 2014, ArXiv e-prints (1409.4791)

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

Hibon, P., Malhotra, S., Rhoads, J., & Willott, C. 2011, ApJ, 741, 101

Hibon, P., et al. 2010, A&A, 515, A97

Hilz, M., Naab, T., & Ostriker, J. P. 2013, MNRAS, 429, 2924

Hirschmann, M., De Lucia, G., Wilman, D., Weinmann, S., Iovino, A., Cucciati, O.,Zibetti, S., & Villalobos, A. 2014, ArXiv e-prints (1407.5621)

Hodge, J. A., Carilli, C. L., Walter, F., de Blok, W. J. G., Riechers, D., Daddi, E., &Lentati, L. 2012, ApJ, 760, 11

Hopkins, A. M., & Beacom, J. F. 2006, ApJ, 651, 142

Hopkins, P. F., Hernquist, L., Cox, T. J., Keres, D., & Wuyts, S. 2009a, ApJ, 691, 1424

Hopkins, P. F., et al. 2004, AJ, 128, 1112

—. 2009b, MNRAS, 397, 802

—. 2010a, ApJ, 715, 202

—. 2010b, ApJ, 724, 915

Hu, E. M., Cowie, L. L., Barger, A. J., Capak, P., Kakazu, Y., & Trouille, L. 2010, ApJ,725, 394

Hu, E. M., Cowie, L. L., Capak, P., McMahon, R. G., Hayashino, T., & Komiyama, Y.2004, AJ, 127, 563

Ilbert, O., et al. 2006, A&A, 457, 841

—. 2009, ApJ, 690, 1236

—. 2013, A&A, 556, A55

Inoue, A. K., Shimizu, I., Iwata, I., & Tanaka, M. 2014, MNRAS, 442, 1805

Iwata, I., Ohta, K., Tamura, N., Ando, M., Wada, S., Watanabe, C., Akiyama, M., &Aoki, K. 2003, PASJ, 55, 415

Iwata, I., et al. 2009, ApJ, 692, 1287

Iye, M., et al. 2006, Nature, 443, 186

Jiang, L., Bian, F., Fan, X., Krug, H. B., McGreer, I. D., Stark, D. P., Clement, B., &Egami, E. 2013, ApJ, 771, L6

110

BIBLIOGRAPHY

Kakazu, Y., Cowie, L. L., & Hu, E. M. 2007, ApJ, 668, 853

Kashikawa, N., et al. 2006, ApJ, 648, 7

—. 2011, ApJ, 734, 119

Kawata, D., & Mulchaey, J. S. 2008, ApJ, 672, L103

Kelvin, L. S., et al. 2012, MNRAS, 421, 1007

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

Keres, D., Katz, N., Weinberg, D. H., & Dave, R. 2005, MNRAS, 363, 2

Knobel, C., et al. 2012, ApJ, 753, 121

Koekemoer, A. M., et al. 2011, ApJS, 197, 36

Komatsu, E., et al. 2011, ApJS, 192, 18

Kovac, K., et al. 2014, MNRAS, 438, 717

Kriek, M., & Conroy, C. 2013, ApJ, 775, L16

Kriek, M., van der Wel, A., van Dokkum, P. G., Franx, M., & Illingworth, G. D. 2008,ApJ, 682, 896

Kriek, M., van Dokkum, P. G., Franx, M., Illingworth, G. D., & Magee, D. K. 2009, ApJ,705, L71

Krogager, J.-K., Zirm, A. W., Toft, S., Man, A., & Brammer, G. 2014, ApJ, 797, 17

Krug, H. B., et al. 2012, ApJ, 745, 122

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

Kuhlen, M., & Faucher-Giguere, C.-A. 2012, MNRAS, 423, 862

Labbe, I., et al. 2010, ApJ, 716, L103

Lange, R., et al. 2015, MNRAS, 447, 2603

Larson, D., et al. 2011, ApJS, 192, 16

Laureijs, R., et al. 2011, arXiv:1110.3193 (1110.3193)

Le Fevre, O., et al. 2003, in Society of Photo-Optical Instrumentation Engineers (SPIE)Conference Series, Vol. 4841, Instrument Design and Performance for Optical/InfraredGround-based Telescopes, ed. M. Iye & A. F. M. Moorwood, 1670–1681

Leauthaud, A., et al. 2007, ApJS, 172, 219

Leitherer, C., Ferguson, H. C., Heckman, T. M., & Lowenthal, J. D. 1995, ApJ, 454, L19

Leitherer, C., et al. 1999, ApJS, 123, 3

111

BIBLIOGRAPHY

Lequeux, J., Peimbert, M., Rayo, J. F., Serrano, A., & Torres-Peimbert, S. 1979, A&A,80, 155

Lilly, S. J., Carollo, C. M., Pipino, A., Renzini, A., & Peng, Y. 2013, ApJ, 772, 119

Lilly, S. J., et al. 2007, ApJS, 172, 70

—. 2009, ApJS, 184, 218

Madau, P. 1995, ApJ, 441, 18

Maier, C., Ziegler, B. L., Lilly, S. J., Contini, T., Perez Montero, E., Lamareille, F.,Bolzonella, M., & Le Floc’h, E. 2014, ArXiv e-prints (1410.7389)

Maiolino, R., et al. 2008, A&A, 488, 463

Malhotra, S., Rhoads, J., Dey, A., Stern, D., & Spinrad, H. 2001, in Astronomical Societyof the Pacific Conference Series, Vol. 240, Gas and Galaxy Evolution, ed. J. E. Hibbard,M. Rupen, & J. H. van Gorkom, 97

Malhotra, S., & Rhoads, J. E. 2002, ApJ, 565, L71

—. 2004, ApJ, 617, L5

—. 2006, ApJ, 647, L95

Mallery, R. P., et al. 2012, ApJ, 760, 128

Mandelker, N., Dekel, A., Ceverino, D., Tweed, D., Moody, C. E., & Primack, J. 2014,MNRAS, 443, 3675

Mannucci, F., Cresci, G., Maiolino, R., Marconi, A., & Gnerucci, A. 2010, MNRAS, 408,2115

Marchesini, D., et al. 2014, ArXiv e-prints (1402.0003)

Marmo, C., & Bertin, E. 2008, in Astronomical Society of the Pacific Conference Series,Vol. 394, Astronomical Data Analysis Software and Systems XVII, ed. R. W. Argyle,P. S. Bunclark, & J. R. Lewis, 619

Martig, M., Bournaud, F., Teyssier, R., & Dekel, A. 2009, ApJ, 707, 250

Masters, D., et al. 2012, ApJ, 755, 169

Matthee, J. J. A., et al. 2014, MNRAS, 440, 2375

McCracken, H. J., et al. 2012, A&A, 544, A156

McGreer, I. D., Mesinger, A., & Fan, X. 2011, MNRAS, 415, 3237

McLean, I. S., et al. 2010, in Society of Photo-Optical Instrumentation Engineers (SPIE)Conference Series, Vol. 7735, Society of Photo-Optical Instrumentation Engineers(SPIE) Conference Series, 1

112

BIBLIOGRAPHY

McLean, I. S., et al. 2012, in Society of Photo-Optical Instrumentation Engineers (SPIE)Conference Series, Vol. 8446, Society of Photo-Optical Instrumentation Engineers(SPIE) Conference Series, 0

McLure, R. J., et al. 2013, MNRAS, 432, 2696

McQuinn, M., Hernquist, L., Zaldarriaga, M., & Dutta, S. 2007, MNRAS, 381, 75

Meiksin, A. 2005, MNRAS, 356, 596

Mesinger, A., Furlanetto, S., & Cen, R. 2011, MNRAS, 411, 955

Mihos, C., & Hernquist, L. 1994, ArXiv Astrophysics e-prints (astro-ph/9405039)

Miller, A. A., et al. 2009, ApJ, 690, 1303

Milvang-Jensen, B., et al. 2013, A&A, 560, A94

Moffat, A. F. J. 1969, A&A, 3, 455

Mok, A., et al. 2013, MNRAS, 431, 1090

Mortlock, D. J., et al. 2011, Nature, 474, 616

Mostardi, R. E., Shapley, A. E., Nestor, D. B., Steidel, C. C., Reddy, N. A., & Trainor,R. F. 2013, ApJ, 779, 65

Murayama, T., et al. 2007, ApJS, 172, 523

Muzzin, A., et al. 2013, ApJ, 777, 18

—. 2014, ArXiv e-prints (1402.7077)

Naab, T., Johansson, P. H., & Ostriker, J. P. 2009, ApJ, 699, L178

Nestor, D. B., Shapley, A. E., Steidel, C. C., & Siana, B. 2011, ApJ, 736, 18

Newman, A. B., Ellis, R. S., Bundy, K., & Treu, T. 2012, ApJ, 746, 162

Nilsson, K. K., Moller-Nilsson, O., Møller, P., Fynbo, J. P. U., & Shapley, A. E. 2009,MNRAS, 400, 232

Noeske, K. G., et al. 2007, ApJ, 660, L43

Oesch, P. A., et al. 2010a, ApJ, 709, L21

—. 2010b, ApJ, 714, L47

—. 2012, ApJ, 745, 110

—. 2013a, ApJ, 772, 136

—. 2013b, ApJ, 773, 75

Ono, Y., et al. 2012a, ApJ, 744, 83

113

BIBLIOGRAPHY

—. 2012b, ApJ, 744, 83

Onodera, M., et al. 2012, ApJ, 755, 26

—. 2014, ArXiv e-prints (1411.5023)

Oser, L., Naab, T., Ostriker, J. P., & Johansson, P. H. 2012, ApJ, 744, 63

Ota, K., & Iye, M. 2012, MNRAS, 423, 444

Ota, K., et al. 2008, ApJ, 677, 12

—. 2010, ApJ, 722, 803

Ouchi, M., et al. 2003, ApJ, 582, 60

—. 2004, ApJ, 611, 660

—. 2008, ApJS, 176, 301

—. 2010, ApJ, 723, 869

Pannella, M., et al. 2009, ApJ, 698, L116

Pastrav, B. A., Popescu, C. C., Tuffs, R. J., & Sansom, A. E. 2013, A&A, 553, A80

Patel, S. G., et al. 2013, ApJ, 766, 15

Peng, C. Y., Ho, L. C., Impey, C. D., & Rix, H.-W. 2010a, AJ, 139, 2097

Peng, Y., Maiolino, R., & Cochrane, R. 2015, ArXiv e-prints (1505.03143)

Peng, Y.-j., et al. 2010b, ApJ, 721, 193

Pentericci, L., et al. 2011, ApJ, 743, 132

—. 2014, ApJ, 793, 113

Perley, D. A., et al. 2011, AJ, 141, 36

Planck Collaboration et al. 2013, arXiv:1303.5062 (1303.5062)

—. 2015, ArXiv e-prints (1502.01589)

Pozzetti, L., et al. 2010, A&A, 523, A13

Prevot, M. L., Lequeux, J., Prevot, L., Maurice, E., & Rocca-Volmerange, B. 1984, A&A,132, 389

Quimby, R. M., Aldering, G., Wheeler, J. C., Hoflich, P., Akerlof, C. W., & Rykoff, E. S.2007, ApJ, 668, L99

Reichard, T. A., et al. 2003, AJ, 126, 2594

Richards, G. T., et al. 2003, AJ, 126, 1131

114

BIBLIOGRAPHY

Robertson, B. E., Ellis, R. S., Dunlop, J. S., McLure, R. J., & Stark, D. P. 2010, Nature,468, 49

Robertson, B. E., Ellis, R. S., Furlanetto, S. R., & Dunlop, J. S. 2015, ArXiv e-prints(1502.02024)

Robertson, B. E., et al. 2013, ApJ, 768, 71

Rodighiero, G., et al. 2011, ApJ, 739, L40

—. 2014, MNRAS, 443, 19

Rodriguez-Gomez, V., et al. 2015, ArXiv e-prints (1502.01339)

Saintonge, A., et al. 2011, MNRAS, 415, 32

Sanders, D. B., et al. 2007, ApJS, 172, 86

Savage, B. D., & Mathis, J. S. 1979, ARA&A, 17, 73

Scarlata, C., et al. 2007, ApJS, 172, 406

Scarlata et al. 2014, submitted

Schaerer, D., Boone, F., Zamojski, M., Staguhn, J., Dessauges-Zavadsky, M., Finkelstein,S., & Combes, F. 2015, A&A, 574, A19

Schaerer, D., de Barros, S., & Stark, D. P. 2011, A&A, 536, A72

Schawinski, K., et al. 2014, MNRAS, 440, 889

Schenker, M. A., Ellis, R. S., Konidaris, N. P., & Stark, D. P. 2014, ApJ, 795, 20

Schenker, M. A., Stark, D. P., Ellis, R. S., Robertson, B. E., Dunlop, J. S., McLure, R. J.,Kneib, J.-P., & Richard, J. 2012a, ApJ, 744, 179

—. 2012b, ApJ, 744, 179

Schenker, M. A., et al. 2013, ApJ, 768, 196

Schmidt, M. 1959, ApJ, 129, 243

Schreiber, C., et al. 2014, ArXiv e-prints (1409.5433)

Scoville, N., Faisst, A., Capak, P., Kakazu, Y., Li, G., & Steinhardt, C. 2014, ArXive-prints (1412.8219)

Scoville, N., et al. 2007a, ApJS, 172, 38

—. 2007b, ApJS, 172, 1

—. 2013, ApJS, 206, 3

Seaton, M. J. 1979, MNRAS, 187, 785

115

BIBLIOGRAPHY

Shen, S., Mo, H. J., White, S. D. M., Blanton, M. R., Kauffmann, G., Voges, W.,Brinkmann, J., & Csabai, I. 2003, MNRAS, 343, 978

Shibuya, T., Kashikawa, N., Ota, K., Iye, M., Ouchi, M., Furusawa, H., Shimasaku, K.,& Hattori, T. 2012a, ApJ, 752, 114

—. 2012b, ApJ, 752, 114

Shimasaku, K., et al. 2006, PASJ, 58, 313

Siana, B., et al. 2010, ApJ, 723, 241

—. 2015, ArXiv e-prints (1502.06978)

Skrutskie, M. F., et al. 2006, AJ, 131, 1163

Sobral, D., Smail, I., Best, P. N., Geach, J. E., Matsuda, Y., Stott, J. P., Cirasuolo, M.,& Kurk, J. 2013, MNRAS, 428, 1128

Sobral, D., et al. 2009, MNRAS, 398, L68

Speagle, J. S., Steinhardt, C. L., Capak, P. L., & Silverman, J. D. 2014, ApJS, 214, 15

Stark, D. P., Ellis, R. S., Chiu, K., Ouchi, M., & Bunker, A. 2010, MNRAS, 408, 1628

Stark, D. P., Ellis, R. S., & Ouchi, M. 2011, ApJ, 728, L2

Stark, D. P., et al. 2014, ArXiv e-prints (1408.3649)

Steidel, C. C., Adelberger, K. L., Giavalisco, M., Dickinson, M., & Pettini, M. 1999, ApJ,519, 1

Steidel, C. C., Giavalisco, M., Pettini, M., Dickinson, M., & Adelberger, K. L. 1996, ApJ,462, L17

Steidel, C. C., & Hamilton, D. 1992, AJ, 104, 941

Steidel, C. C., Hunt, M. P., Shapley, A. E., Adelberger, K. L., Pettini, M., Dickinson, M.,& Giavalisco, M. 2002, ApJ, 576, 653

Steidel, C. C., Pettini, M., & Adelberger, K. L. 2001, ApJ, 546, 665

Steidel, C. C., Shapley, A. E., Pettini, M., Adelberger, K. L., Erb, D. K., Reddy, N. A.,& Hunt, M. P. 2004, ApJ, 604, 534

Steinhardt, C. L., et al. 2014, ApJ, 791, L25

Stockton, A., McGrath, E., Canalizo, G., Iye, M., & Maihara, T. 2008, ApJ, 672, 146

Szomoru, D., Franx, M., Bouwens, R. J., van Dokkum, P. G., Labbe, I., Illingworth, G. D.,& Trenti, M. 2011, ApJ, 735, L22

Tacchella et al. 2015, Science, submitted

116

BIBLIOGRAPHY

Tacchella, S., Trenti, M., & Carollo, C. M. 2013, ApJ, 768, L37

Tacchella, S., et al. 2014, ArXiv e-prints (1411.7034)

Tacconi, L. J., et al. 2013, ApJ, 768, 74

Tanaka, M., Moriya, T. J., Yoshida, N., & Nomoto, K. 2012, MNRAS, 422, 2675

Taniguchi, Y., et al. 2005, PASJ, 57, 165

Taranu, D. S., Hudson, M. J., Balogh, M. L., Smith, R. J., Power, C., Oman, K. A., &Krane, B. 2014, MNRAS, 440, 1934

Tauber, J. A., et al. 2010, A&A, 520, A1

Tilvi, V., et al. 2010, ApJ, 721, 1853

Toft, S., et al. 2007, ApJ, 671, 285

Toomre, A., & Toomre, J. 1972, ApJ, 178, 623

Tremonti, C. A., et al. 2004, ApJ, 613, 898

Treu, T., Schmidt, K. B., Trenti, M., Bradley, L. D., & Stiavelli, M. 2013, ApJ, 775, L29

Treu, T., Trenti, M., Stiavelli, M., Auger, M. W., & Bradley, L. D. 2012, ApJ, 747, 27

Trinh, C. Q., Barton, E. J., Bullock, J. S., Cooper, M. C., Zentner, A. R., & Wechsler,R. H. 2013, MNRAS, 436, 635

Trujillo, I., Aguerri, J. A. L., Cepa, J., & Gutierrez, C. M. 2001, MNRAS, 328, 977

Trujillo, I., et al. 2006, ApJ, 650, 18

van de Voort, F., Schaye, J., Booth, C. M., Haas, M. R., & Dalla Vecchia, C. 2011,MNRAS, 414, 2458

van der Burg, R. F. J., Hildebrandt, H., & Erben, T. 2010, A&A, 523, A74

van der Wel, A., et al. 2012, ApJS, 203, 24

—. 2014, ApJ, 788, 28

van Dokkum, P. G., & Franx, M. 1996, MNRAS, 281, 985

van Dokkum, P. G., et al. 2008, ApJ, 677, L5

—. 2010, ApJ, 709, 1018

—. 2011, ApJ, 743, L15

Vanden Berk, D. E., et al. 2004, ApJ, 601, 692

Vanzella, E., et al. 2009, ApJ, 695, 1163

117

BIBLIOGRAPHY

—. 2011, ApJ, 730, L35

—. 2012, ApJ, 751, 70

—. 2014, A&A, 569, A78

von der Linden, A., Wild, V., Kauffmann, G., White, S. D. M., & Weinmann, S. 2010,MNRAS, 404, 1231

Vulcani, B., et al. 2014, MNRAS, 441, 1340

Welker, C., Dubois, Y., Devriendt, J., Pichon, C., Kaviraj, S., & Peirani, S. 2015, ArXive-prints (1502.05053)

Wetzel, A. R., Tinker, J. L., Conroy, C., & van den Bosch, F. C. 2013, MNRAS, 432, 336

Whitaker, K. E., van Dokkum, P. G., Brammer, G., & Franx, M. 2012, ApJ, 754, L29

Whitaker, K. E., et al. 2013, ApJ, 770, L39

White, S. D. M., & Rees, M. J. 1978, MNRAS, 183, 341

Wilhite, B. C., Brunner, R. J., Grier, C. J., Schneider, D. P., & vanden Berk, D. E. 2008,MNRAS, 383, 1232

Williams, R. J., Quadri, R. F., Franx, M., van Dokkum, P., & Labbe, I. 2009, ApJ, 691,1879

Williams, R. J., Quadri, R. F., Franx, M., van Dokkum, P., Toft, S., Kriek, M., & Labbe,I. 2010, ApJ, 713, 738

Woo, J., Dekel, A., Faber, S. M., & Koo, D. C. 2014, ArXiv e-prints (1406.5372)

Woo, J., et al. 2013, MNRAS, 428, 3306

Wuyts, S., et al. 2012, ApJ, 753, 114

Wyithe, J. S. B., Hopkins, A. M., Kistler, M. D., Yuksel, H., & Beacom, J. F. 2010,MNRAS, 401, 2561

York, D. G., et al. 2000, AJ, 120, 1579

Yuan, T.-T., Kewley, L. J., & Richard, J. 2013, ApJ, 763, 9

Zahid, H. J., et al. 2014, ApJ, 792, 75

Zheng, Z.-Y., Wang, J.-X., Malhotra, S., Rhoads, J. E., Finkelstein, S. L., & Finkelstein,K. 2014, MNRAS, 439, 1101

118

Publications

Peer-reviewed publications

N. Z. Scoville, A. L. Faisst, P. Capak, et al. Dust Attenuation in High RedshiftGalaxies – Diamonds in the Sky. Accepted for publication in The Astrophysical Journal,arXiv:1412.8219 , December 2014

A. L. Faisst, P. Capak, C. M. Carollo, C. Scarlata & N. Z. Scoville. SpectroscopicObservations of Lyα Emitters at z ∼ 7.7 and Implications on Re-ionization. TheAstrophysical Journal, 788, 87, June 2014

P. Capak, A. L. Faisst, J. D. Vieira, et al. Keck-I MOSFIRE Spectroscopy of thez ∼ 12 Candidate Galaxy UDFj-39546284. The Astrophysical Journal Letters, 773, 14,August 2013

N. Z. Scoville, S. Arnouts, . . . , A. L. Faisst, et al. Evolution of Galaixes and TheirEnvironments at z = 0.1 − 3 in COSMOS. The Astrophysical Journal Supplement, 206,3, May 2013

Articles in preparation or to be submitted

A. L. Faisst, C. M. Carollo, P. Capak, et al. Ultra Massive Glaxies at z < 2 inCOSMOS: Size Evolution and Quenching. To be submitted to The Astrophysical Journal

C. M. Carollo, A. L. Faisst, et al. Dependence of Morphological Mix, Quiescence,and Size on Environment at 0.2 < z < 0.8 in zCOSMOS. To be submitted to TheAstrophysical Journal

119

PUBLICATIONS

Selected Press Articles

• Hongger, Sehen, was man mit blossem Auge nicht sieht, January 15, 2015http://hoengger.ch/wp-content/uploads/2014/07/150115endfassung.pdf

• Polykum, ETH Zurich, Eintrittskarte zu den Sternen, November 10, 2014http://www.vseth.ethz.ch/sites/default/files/polykum-1415-3_small.pdf

• Limmattaler Zeitung, Die Sonne ist ein Stern, January 6, 2010http://www.limmattalerzeitung.ch/panorama/vermischtes/

die-sonne-ist-ein-stern-5598302/

120

http://hoengger.ch/wp-content/uploads/2014/07/150115endfassung.pdf

http://www.vseth.ethz.ch/sites/default/files/polykum-1415-3_small.pdf

http://www.limmattalerzeitung.ch/panorama/vermischtes/die-sonne-ist-ein-stern-5598302/

http://www.limmattalerzeitung.ch/panorama/vermischtes/die-sonne-ist-ein-stern-5598302/

Rights / License: Research Collection In Copyright - …Research Collection Doctoral Thesis The...

Documents

Transcript of Rights / License: Research Collection In Copyright - …Research Collection Doctoral Thesis The...