Observations & Theory. Vithal Shet Tilvi. A Dissertation Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy

Transcription

1 Luminosity Function of Lyα Emitters at the Reionization Epoch: Observations & Theory by Vithal Shet Tilvi A Dissertation Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy Approved August 2011 by the Graduate Supervisory Committee: Sangeeta Malhotra, Chair James Rhoads Evan Scannapieco Patrick Young Rolf Jansen ARIZONA STATE UNIVERSITY December 2011

2 ABSTRACT Galaxies with strong Lyman-alpha emission line (also called Lyα galaxies or emitters) offer an unique probe of the epoch of reionization - one of the important phases when most of the neutral hydrogen in the universe was ionized. In addition, Lyα galaxies at high redshifts are a powerful tool to study low-mass galaxy formation. Since current observations suggest that the reionization is complete by redshift z 6, it is therefore necessary to discover galaxies at z > 6, to use their luminosity function (LF) as a probe of reionization. I found five z = 7.7 candidate Lyα galaxies with line fluxes > erg cm 2 s 1, from three different deep near-infrared (IR) narrowband (NB) imaging surveys in a volume > Mpc 3. From the spectroscopic followup of four candidate galaxies, and with the current spectroscopic sensitivity, the detection of only the brightest candidate galaxy can be ruled out at 5σ level. Moreover, these observations successfully demonstrate that the sensitivity necessary for both, the NB imaging as well as the spectroscopic followup of z 8 Lyα galaxies can be reached with the current instrumentation. While future, more sensitive spectroscopic observations are necessary, the observed Lyα LF at z = 7.7 is consistent with z = 6.6 LF, suggesting that the intergalactic medium (IGM) is relatively ionized even at z = 7.7, with neutral fraction x HI 30%. On the theoretical front, while several models of Lyα emitters have been developed, the physical nature of Lyα emitters is not yet completely known. Moreover, multi-parameter models and their complexities necessitates a simpler model. I have developed a simple, single-parameter model to populate dark mater halos with Lyα emitters. The central tenet of this model, different from many of the earlier models, is that the star-formation rate (SFR), and hence the Lyα luminosity, is proportional to the mass accretion rate rather than the total halo mass. This simple model is successful in reproducing many observable including LFs, stellar masses, SFRs, and clustering of Lyα emitters from z 3 to z 7. Finally, using this model, I find that the mass accretion, and hence the star-formation in > 30% of Lyα emitters at z 3 occur through major mergers, and this fraction increases to 50% at z 7. i

3 ACKNOWLEDGEMENTS During my graduate life at ASU, I have enjoyed the support of many people, which made this part of my life to be cherished. First and foremost, I wish to thank my advisor Sangeeta Malhotra; her constant little pressure to write observing proposals, attending conferences, and presenting my research have helped me tremendously throughout my graduate studies. She was always ready to help. Thank you for all your mentoring, encouragement, and ideas that led me through this dissertation. I am greatly indebted to James Rhoads, my unofficial advisor during my graduate studies. He has been a great mentor throughout, and always willing to listen to any queries ranging from data reduction to science. I have learnt a lot from him not just about research, but also about observing at big facilities including the Keck telescope. Evan Scannapieco, my mentor for all my theoretical work, have helped me all along. I was never hesitant to knock his door for his help. I thank Rolf Jansen, and Patrick Young, my committee members, for their help serving on my committee and for their suggestions. I also want to thank Rogier Windhorst for providing me office space during my early days of graduate studies. I gratefully thank Seth Cohen, Pascale Hibon, Steve Finkelstein, Nimish Hathi, Russell Ryan, and my other fellow students, Zhenya Zheng, Lifang Xia, Emily Mclinden, Katie Kaleida, and Hwihyum Kim for all their help. I also thank the staff of the School of Earth & Space Exploration for all their help during my studies at ASU. Financial support during my graduate studies has come from various sources, and I want to thank for this. My passion for research would not have been the same if it were not for the inspiration, and support that my parents, sisters, and friends have given me. You are the backstage performers who makes it worthwhile, fun, and successful. I cannot thank you enough for your unwavering support throughout. Finally, but more importantly, I wish to thank my wife Siddhi for her unconditional support and be my better half for last few years. You deserve more credit then I can express for your diligence, patience, and for being everything I am not. ii

4 TABLE OF CONTENTS Page LIST OF TABLES LIST OF FIGURES vii viii CHAPTER 1 INTRODUCTION Sources of Reionization Methods for Discovering the Most Distant Galaxies Continuum Selected Lyman Break Galaxies Emission Line Selection Blind Spectroscopic Emission Line Searches Probes of the Reionization Epoch Gunn-Peterson Effect: Observations of Quasars Cosmic Microwave Background cm Observations Luminosity Function of Lyα Galaxies Theoretical Understanding of Lyα Emitters Mass Accretion and Star Formation in Lyα Emitters Outline THE LUMINOSITY FUNCTION OF LYMAN ALPHA EMITTERS AT REDSHIFT Overview Introduction Data Handling Observations and NEWFIRM Filters Data Reduction Photometric Calibration Limiting Magnitudes iii

5 Chapter Page 2.4 Lyα candidate selection Constant Flux Test Contamination of the Sample Foreground Emission Line Objects Other Contaminants Monte-Carlo Simulations Lyα Luminosity Function at z= Lyα Equivalent Width Summary and Conclusions SPECTROSCOPY OF Lyα GALAXIES AT z 8: ARE WE PROBING A NEUTRAL UNIVERSE? Overview Introduction Observations and Data Reduction Lyα Candidate Selection Spectroscopic Observations NIRSPEC Observations MMIRS Observations D Spectra Extraction NIRSPEC MMIRS D Spectra and Flux Calibration Survey Sensitivity Results and Discussion Expected Number of Lyα Sources Comparison With Model Predictions Is the Universe Neutral at z = 7.7? Future Lyα Surveys iv

6 Chapter Page 3.7 Summary and Conclusions Lyα EMITTERS AT z 8: CONSTRAINTS ON THE LUMINOSITY FUNCTION FROM THE DARK AGES SURVEY Overview Introduction Survey and Data Reduction Observations and NEWFIRM Filters Data Reduction UNB Stack Quality Assessment Calibration and Catalog Generation Photometric Calibration Photometric Uncertainty Source Catalog and Survey Completeness Lyα Candidate Selection Contamination of the Sample Discussion Lyα Luminosity Function at z= Expected Number of Lyα Galaxies in Our Survey Combining with Previous Observations Comparison to Model Prediction Implications for Reionization Summary and Conclusions A PHYSICAL MODEL OF Lyα EMITTERS Overview Introduction Physical Model for Lyα Emitters Simulation & Halo Catalogs Lyman Alpha Luminosity Function v

7 Chapter Page 5.6 Results Physical Properties of LAEs Dust Mass in Our Model LAEs UV Luminosity Function of LAEs Evolution of Lyα Luminosity Function Clustering of LAEs Two-point Spatial Correlation Function Summary and Conclusions PREDICTING THE MERGER FRACTION OF LYMAN ALPHA EMIT- TERS FROM REDSHIFT z 3 to z Overview Introduction Methods Modeling Lyα Emitters Simulation and Halo Catalogs Lyα Emitter Catalogs Merger Tree Model Results and Observational Comparisons Merger Fraction at z Effect of Halo Mass Resolution and t Lyα Redshift Evolution of Merger Fraction Comparison with the Observations Uncertainties from Model Predictions and Observations Dependence of Merger Fraction on the Progenitor Mass Ratio Definition103 Merger Observability Timescale Summary and Conclusions Conclusions REFERENCES vi

8 LIST OF TABLES Table Page 2.1 Coordinates of our Lyα Candidates Lyα Searches at z > Summary of Spectroscopic Observations Summary of Observations of Lyα Galaxies at z > vii

9 LIST OF FIGURES Figure Page 1.1 Schematic of NB selection of Lyα galaxies at z = Schematic of Lyα LF test for the reionization of the IGM Normalized narrowband filter transmission curve (green line) and nightsky OH emission lines (Rousselot et al., 2000) (black line) with arbitrary flux. Here we have shown the transmission curve (at the center of the field) of only narrowband filter to demonstrate the use of very narrow region between OH lines, to search for Lyα emitters at z=7.7. Two weak OH emission lines with λ = and µm (Rousselot et al., 2000) appear in the UNB images as concentric rings beyond 12 radius, since the central wavelength of the UNB filter shifts to the blue for positions away from field center Postage stamps (50 wide) of all four Lyα candidates in combined (chisquare) optical image (left panel), UNB (middle), and J-band filter (right panel). The positions of Lyα candidates are marked with circles 16 in diameter Cumulative Lyα luminosity function of our z=7.7 candidates (filled circles). The filled points show the LF that will result if all four Lyα galaxy candidates are confirmed. The upper error bars are Poisson errors based on our sample size, while the down-arrows below each data point indicate the possibility of a lower LF if some candidates are extreme emission line galaxies at lower redshift. The open circles represent the LF from Hibon et al. (2010) while the dashed line and dotted line show Lyα LFs at z=5.7 (Ouchi et al., 2008) and z=6.5 (Kashikawa et al., 2006) respectively. The open square is the LF at z=6.96 (Iye et al., 2006) Slit configuration for the brightest candidate Lyα galaxy (circled) in the NB1063 image. The slit shares another continuum object which allows for accurate stacking of dithered frames viii

10 Figure 3.2 Top four panels: 2-D images of sky-subtracted spectra of all four objects. Page The circles enclose the expected position of the Lyα galaxies. Bottom panel: 2-D spectrum showing the positions of strong night skylines (dark means brighter) Top panel: 1-D spectrum of our brightest object. Middle panel: 1σ limiting flux density for our spectra. Note the minimum flux density at µm, the expected position of our science objects. Bottom panel: Coadded spectra of all four objects Constraints on the Lyα LF at z = 7.7. Shaded region indicates our spectroscopic constraint, while diagonal symbols show continuum selected LBGs with strong Lyα emission at z = 7 (Vanzella et al 2011). Black filled circles are the NB selected candidates from Tilvi et al (2010). The z = 6.6 Lyα LF is shown with dotted line Comparison of model Lyα LF (dashed line) with the spectroscopically confirmed galaxies at z 7. The dot-dashed line is the Lyα LF at z=7.7 in a 50% neutral IGM. We scaled our model predicted LF using a scaling factor obtained from McQuinn et al (2007). The scaled LF for 30% neutral fraction is also shown (dotted line) Normalized narrowband filter transmission curves (dahed-blue lines), and night-sky OH emission lines (Rousselot et al., 2000) (black line) with arbitrary flux. Here we have shown the transmission curve (at the center of the field) of both narrowband filters: UNB1056 & UNB1063 filter. to demonstrate the use of very narrow region between OH lines, to search for Lyα emitters at z= Postage stamps of our candidate Lyα galaxy at z = 7.7. The circles are centered on the candidate galaxy. Note an object, similar in NB flux and morphology, close to the candidate in the UNB1063, is clearly detected in all the filters ix

11 Figure 4.3 Comparison of our cumulative Lyα LF at z = 7.7 in the EGS field (filled Page circle) with different studies at similar redshifts. Filled squares indicate the Lyα LF from our combined data : EGS(this study) +COSMOS (Krug et al 2011)+Cetus (Tilvi et al 2010). Arrows indicate that these are upper limits on the observed Lyα LF at z = 7.7. The Lyα at z = 8.8, z = 7.7, z = 6.6, and z = 5.7 are represented in green, black, blue, and red color respectively. Different lines indicate LF limits from different studies: At z = 8.8, red dotted line, red long-dashed line, and red dot-dashed line show LF limits from Sobral et al (2009), Cuby et al (2007), and Willis et al( ) respectively. At z = 7.7, blue dot-dashed, and blue dotted line show Lyα LF limit from Clement et al (2011), and Hibon et al (2010), respectively. The black dotted line, black dot-dashed line, and solid line represent Lyα LF at z = 6.6 from Hu et al (2010), Kashikawa et al (2011), and Ouchi et al (2010) respectively. The green dashed line shows Lyα LF at z = 5.7 (Ouchi et al 2008). The open square is the spectroscopically confirmed Lyα galaxy at z = 6.96 (Iye et al 2006) Comparison of observed Lyα LF at z = 7.7 with model predictions. The dashed line indicates the model predicted Lyα LF (without including the evolution of IGM) at z = 7.7 (Tilvi et al 2009), while the long-dashed and dashed-dotted lines show how the LF would appear in an IGM with neutral fraction with x HI = 30%, and x HI = 50% respectively Evolution of Lyα LFs at redshifts z 3 7. The dotted lines show results from our model and the symbols with error bars are the observational data. (a) The best-fit model Lyα LF at z = 3.1 yields a SFE of 2.5%. We use this SFE to construct model Lyα LFs at z=4.5, 5.7, and 6.6 (b)-(d). The references for the data: z = 3.1 (Gronwall et al., 2007), z = 4.5 (Dawson et al., 2007), z = 5.7 (Ouchi et al., 2008), and z = 6.6 (Kashikawa et al., 2006) x

12 Figure 5.2 Accreted mass function and halo mass functions. Solid lines show accreted Page mass functions at z = 3.1 (violet), z = 4.5(blue), z = 5.7 (green) and z = 6.6(red). The dashed blue line shows the dark matter halo mass function at z = 4.5 to compare with the corresponding accreted mass function. The vertical dotted lines enclose the region of observed Lyα luminosities in LAEs The DM accretion rate of halos as a function of halo mass at z 3 7. The solid line is the least-square fit to the median mass in each M/dt bin. The slope of the lines is nearly constant over all redshifts UV LFs of LAEs at z = 3.1 and z = 5.7. The filled and open circles are the data from Ouchi et al. (2008) & Gronwall et al. (2007), respectively. The dotted line is our model predicted UV LF of LAEs Field-to-field variance of number of LAEs at z = 6.6, measured in eight subvolumes ( Mpc 3 ), plotted as a function of survey detection limit. The variance σ 2 > 30% for a typical narrowband LAE survey with Lyα detection limit > erg s The spatial distribution of our model LAEs in a slice from DM simulation at two redshifts z = 5.7 (left) and z = 4.5 (right) in a volume h 3 Mpc 3. The small (red) and big (blue) filed circles represent the positions of DM halos and model LAEs, respectively. Only LAEs with L Lyα > erg s 1 are plotted. In general, the LAEs are located in high density regions. Also note that different halos host LAEs at the two redshifts, depending on whether they are accreting or not. This gives rise to a duty cycle quite naturally xi

13 Figure 5.7 Correlation lengths of LAEs at different redshifts. Left: comparison of Page correlation lengths of our model and observed LAEs at different redshifts. Here we include LAEs with Lyα luminosity greater than the survey limit at each redshift. The filled circles are our model results, other symbols are from different observations as shown in the labels. The observed r 0 values shown here for z=4.5 & 4.86 are for the contamination corrected (the maximum value permitted) LAE sample. Our model results are slightly shifted to avoid overlap with other observational points. Right: The correlation lengths of our model LAEs with a constant Lyα luminosity cutoff (L Lyα > erg s 1 ) at all redshifts, showing that the apparent evolution of correlation length with redshift (seen in left panel) is mostly due to different luminosity detection limits in LAE surveys Merger schematic showing classification of mergers based on the progenitor mass ratio The fraction of Lyα emitters that have formed from major mergers, minor mergers, and smooth accretion at z = 3.1. The major and minor merger fractions are shown in solid, and dotted lines respectively, while the long dashed line shows the fraction of Lyα emitters that accrete mass through smooth accretion. The shaded area are the poisson errors on the major merger fraction. About 35% of the Lyα emitters at z 3 accrete their mass through major mergers. The steep decline in the major merger fraction at Log L Lyα < 42 erg s 1 is not real but results from limited simulation mass resolution (see Section 6.3) xii

14 Figure 6.3 Comparing the merger fractions between Gadget1024 and the Millenium- Page II simulation. The vertical dotted lines represent the luminosity range below which our results are not reliable. In the Millenium-II simulation, due to higher mass resolution, the vertical dotted lines shift towards the lower luminosity. Also seen is the decreased noise at the brighter luminosities in the Millenium-II simulation due to its larger simulation volume Model prediction of major merger fraction for different stellar ages of Lyα emitters at z 7. The smaller t Lyα implies younger stellar population in Lyα emitters Redshift evolution of major merger fraction (left panel), minor merger fraction (middle), and smooth accretion (right panel) from z = 3.1 to z = 6.6. We see a mild evolution of major merger fraction (left), and in smooth accretion (right). On the other hand the minor merger fraction remains constant over this redshift range Comparison of our model predicted major merger fraction with the observations. Filled circles are our model prediction while other symbols represent observations at z = 5.7 (Taniguchi et al 2009), z 5 (Pirzkal et al., 2007), z = 3.1 (Bond, 2009), and z 0.3 (Cowie et al., 2010). Error bars on model predictions indicate Poisson errors xiii

15 Chapter 1 INTRODUCTION The present day universe, filled with abundant luminous sources, formed from a smooth initial phase, which was once opaque to the electromagnetic radiation. About 370,000 years after the Big Bang, neutral hydrogen formed for the first time, and the universe became transparent to the electromagnetic radiation presently observed as the Cosmic Microwave Background (Penzias & Wilson 1965). However, the universe entered the Dark Ages until the formation of first stars and first galaxies, formed from the overdense regions. The ionizing photons with energies > 13.6 ev corresponding to a wavelength λ < 912 Å from these early sources most likely reionized the universe. While current observations suggest that the universe reionized at 6 z 15, the exact details about how this progressed remain unclear. Studies suggest that initially the ionizing photons from the first stars and galaxies ionized its surroundings, forming ionized HII bubbles ( e.g. Wyithe & Loeb 2005; Furlanetto, Zaldarriaga, & Hernquist 2004; Furlanetto & Oh 2005), and with the expansion of the universe the number of ionizing sources increased along with their mass, and luminosity. Consequently, due to increase in the size, and abundance of the ionized bubbles they overlap increasing the mean free path of the ionizing photons. Reionization is complete when enough number of ionized bubbles overlap to the extent that the regions devoid of any ionizing sources are also ionized. 1.1 Sources of Reionization Independent of how reionization occured, current observations suggest that reionization is nearly complete by z 6 (Becker et al., 2001, Fan et al 2001, 2006). Any source emitting ultra-violet (UV) ionizing photons is contributing to reionizing the universe. To ionize, and keep the intergalactic medium (IGM) ionized, the ionization rate must exceed the rate of recombination. 1

16 While the nature of ionizing sources is not completely understood, few sources including first stars (Pop III), first galaxies, and quasars are the primary candidates for ionizing the IGM. Other more exotic objects such as miniquasars, black holes, and decaying dark matter particles have also been proposed. While luminous quasars emit ionizing photons effectively, and are the primary sources of ionizing radiation at z < 4, their number density declines at higher redshift z 6 (Fan et al 2006, Dijkstra et al 2004). Other studies (Madau et al 1999) also show that quasars alone cannot provide sufficient ionizing radiation to ionize the entire universe. The other candidate, Pop III stars, are expected to be the first luminous objects to form in the universe in a metal-free environment, and are thus the first sources to start reionizing the universe. Several authors have proposed Pop III stars as a candidate for reionizing the universe. Wyithe & Loeb (2003) argue that the contribution from Pop III stars with heavy initial mass function (IMF) is necessary in order to account for the electron scattering depth inferred from Wilkinson Microwave Anisotropy Probe (WMAP) observations. Ciardi, Ferrara & White (2003) show that Pop III stars with a Salpeter IMF can produce results consistent with the Three year WMAP results. However, the initial mass function for these first stars is unknown. Moreover, currently there is no observational evidence for Pop III stars. It is expected that star-forming galaxies likely dominate the early universe, and hence the ionizing radiation. While the observed number density of star-forming galaxies could be computed relatively accurately, their contribution to the ionizing budget depends on other factors including the star-formation rate (SFR), and escape fraction of the ionizing photons (e.g. Gnedin & Ostriker 1997; Gnedin 2000) which is not very well known, especially at high redshifts. Thus, in order to estimate these values more reliably it is important to discover star-forming galaxies at z > 7. 2

17 1.2 Methods for Discovering the Most Distant Galaxies Recent improvements in the near-ir detectors has led to tremendous progress in discovering galaxies at z > 6 when reionization occured. Moreover, increasing sample of candidate galaxies even up to z 10 have been identified (e.g. Bouwens et al 2010). Currently, there are two popular techniques: continuum selected and emission line detection, for discovering galaxies at z > 6. Continuum Selected Lyman Break Galaxies The continuum selected or Lyman-break galaxies (LBG) are identified based on the fact that the flux blueward of the Lyman limit (λ< 912 Å in the restframe) is completely absorbed by the neutral gas around the galaxy. Thus, the broadband filters at λ > 912 Å in the rest-frame of the galaxy will show a higher flux compared to the flux in the filters blueward of this wavelength. This break in the flux around λ = 912 Å has been successfully used to identify candidate galaxies even up to z 10. Recently, few LBGs have been spectroscopically confirmed at z 7 (Vanzella et al 2011, Ono et al 2011). Lehnert et al (2010) spectroscopically confirmed a LBG at z = 8.6, however with a weak detection. While this method has been successful in identifying a large numbers of candidate galaxies at z < 6, it is challenging to discover z > 6 LBGs due to their faint continuua. Currently, most of the LBGs at z > 6 are identified using the Hubble Space Telescope (HST), which has a much smaller field-of-view compared to the ground based telescopes. In addition, the spectroscopic confirmation of z > 6 LBGs is also challenging since they might not necessarily have strong Lyα emission line, a primary tool for the spectroscopic confirmation of high-redshift galaxies. Emission Line Selection A complimentary approach to search for high-redshift star-forming galaxies is NB imaging, where a narrowband filter is centered on the redshifted Lyα emission line. Lyα galaxies are characterized by their strong Lyα emission line with large 3

18 Lyα equivalent widths. Partridge & Peebles (1967) predicted a large population of star-forming galaxies with strong Lyα emission line resulting from the interaction of ionizing radiation from hot young stars with the surrounding gas. However, it took about three decades to successfully discover high-redshift Lyα galaxies. The primary reason for the failure to discover these galaxies was the overestimation of the Lyα line flux, as well as insufficient survey sensitivities. Over the past decade, the number of Lyα galaxies detected from z = 0.3 to z = 7 have been rapidly increasing. Figure 1.1 shows the schematic of the NB selection of galaxies at z = 7.7. The NB imaging technique employs a combination of broadband, and a NB filter such that the NB filter is centered on the redshifted Lyα emission line. In order to increase the sensitivity of the survey, the central wavelengths of the NB filters are chosen to take advantage of the night sky window free of OH lines. Thus, an emission line galaxy will contribute a large flux in the NB filter compared to the continuum flux in the broadband filters. A combination of broadband filters blueward of the emission line is usually used to eliminate the foreground emission line galaxies. The unique advantage of NB selected galaxies over the broadband selection is that, they are relatively easier for the spectroscopic followup since these galaxies are preselected based on the strong emission line flux, unlike broadband selected LBGs. Blind Spectroscopic Emission Line Searches A parallel approach to searching high-redshift emission line galaxies is through blind spectroscopic surveys which benefit from the low sky background. In addition, spectroscopic searches taking advantage of gravitational lensing of very high-redshift galaxies by foreground galaxy clusters, have also been performed (Ellis et al 2001, Santos et al 2004, Stark et al 2007). However, these surveys suffer from smaller survey volumes compared to the NB imaging surveys. While, the number density of star-forming galaxies can be used to estimate the contribution of ionizing photons to the IGM, and hence constrain the epoch of 4

19 Figure 1.1: Schematic of NB selection of Lyα galaxies at z = 7.7. reionization, there are different approaches to study the reionization process. 1.3 Probes of the Reionization Epoch The reionization is expected to be a gradual process rather than instantaneous. Moreover, different probes of reionization probes different stages of this process. Below, we describe these probes in more detail. Gunn-Peterson Effect: Observations of Quasars Observations of Gunn-Peterson effect (Gunn & Peterson 1965)- Lyα absorption of distant quasar spectra (Becker et al 2001, Fan et al 2003) in the presence of neutral hdrogen, provides a direct probe of neutral hydrogen in the IGM at high-redshifts. For quasars at or beyond the epoch of reionization where the neutral hydrogen is expected to be in abundance, the IGM will completely absorb the flux blueward of the Lyα line (λ=1216 Å). The optical depth to the Lyα photons in an uniform IGM can be written as ( τ GP (z) = Ωm h 2 ) 1/2 ( Ωb h 2 ) (1 ) + z 3/2 ( nhi n H 5 ). (1.1)

20 Since the flux buleward of the Lyα line will be completely absorbed even for x HI 10 4 this is an ideal probe to study the end phases of reionization. Current observations of quasars ( Fan et al 2003, 2004, White et al 2003) suggest that the reionization is nearly complete by z 6, about 1 billion years after the Big Bang. However, studying reionization using quasars at z > 6 is increasingly difficult due to their low number density at higher redshifts. Cosmic Microwave Background The Cosmic Microwave Background (CMB) radiation, the oldest electromagnetic radiation that can be observed today, carries a wealth of information about the initial conditions of the early universe, since the reionization affects both the temperature anisotropy, and the CMB polarization. The polarization of the CMB photons due to Thompson scattering is sensitive to free electrons. Since reionization will produce free electrons, the result of these free electrons is imprinted on the CMB polarization which can then be used to study the epoch of reionization. Thus, CMB polarization is more suited to probe the onset of the epoch of reionization as opposed to the end of reionization as probed using quasars. While the CMB polarization can be used to probe the begining phases of the reionization, it cannot reveal the small scale details such as the overlapping structure of ionized bubbles created by individual ionizing sources. The distribution of these ionizing sources induces temperature anisotropies in the CMB (Gruzinov & Hu 1998, Knox et al.1998, McQuinn et. al. 2005). Since this anisotropy depends on the patchiness of the ionizing sources, the small scale information that can be provided by the CMB polarization is also limited by the cosmic variance (Hu & Holder 2003). The temperature anisotropy on the other hand is a useful probe of the reionzation at smaller angular scales (Gruzinov & Hu 1998, Knox et al.1998, Haiman & Knox 1999, Barkana & Loeb 2001, Santos et al. 2003, Zahn et al. 2005). 6

21 Figure 1.2: Schematic of Lyα LF test for the reionization of the IGM. 21 cm Observations The 21 cm transition in a neutral hydrogen is caused by the spin-flip transition relative to the nucleus. While the intensity of this signal is weak, the presence of large amounts of neutral hydrogen before the reionzation offers a direct probe to observe the distribution of the neutral hydrogen during the reionization era. The signatures of 21 cm observations have been explored using numerical simulations (Gnedin & Ostriker 1997), and are possible with the instruments including Square Kilometer Array 1. However, one of the challenges in observing 21 cm from neutral hydrogen is the foreground contamination from the man-made radiations. Luminosity Function of Lyα Galaxies Lyα galaxies, by selection, are star-forming galaxies with strong Lyα emission line, and observed to have smaller stellar masses compared to LBGs. Lyα emission line 1 7

22 is a powerful tool to study the reionization epoch, and this test has been used to constrain ionization of the IGM up to z = 6.6 (Malhotra & Rhoads 2004, Ouchi et al 2010). In a significantly neutral IGM Lyα photons will be absorbed and scattered away from the line-of-sight, causing a significant attenuation of the Lyα line flux. This will result in decrease in the observed number density of Lyα galaxies in a significantly neutral IGM, which can provide a robust test of the reionization of the IGM. In a 50% neutral IGM, we expect a factor of about three to four decrease in the Lyα LF (McQuinn et al 2007). This test was first demonstrated by Malhotra & Rhoads (2004) by comparing the Lyα LF at z = 5.7 and z = 6.5, using the Lyα galaxy sample from the Large Area Lyman Alpha (LALA) survey. The results of this survey are consistent with the IGM being < 50% neutral, in agreement with the recent results with x HI < 20% (Ouchi et al 2010). Thus, in order to use the Lyα LF test as a probe of reionization, Lyα searches at z > 6.5 are required, where we expect the IGM to be more neutral. 1.4 Theoretical Understanding of Lyα Emitters Over the past few years several theoretical models have been developed in order to explain the observed properties of Lyα emitters. However, modeling Lyα emitters is still challenging due to complex processes involved in the processing of Lyα emission in the galaxy. For example, the escape of Lyα photons, due to resonant scattering, depends on several factors including the dust geometry, inflows, and outflows (Neufeld 1991; Tasitsiomi 2005; Hansen & Oh 2006; Finkelstein et al. 2007; Dayal, Maselli & Ferrara 2010). These processes are further complicated since the production of Lyα photons depends on the escape of ionizing photons, stellar age, and SFR (e.g. Dijkstra & Wyithe 2007). The primary requirement for all of these models is to associate the Lyα emission with the mass of the galaxy or the dark matter halo. Many of the models (e.g. Salvaterra & Ferrara 2006, Le Delliou et al. 2006, Dijkstra et al. 2007b, Fernandez & Komatsu 2008) relate the Lyα luminosity to the total halo mass. In 8

23 such models, one of the difficulties that arises, is the over prediction of the number density of Lyα emitters, which should then be compensated by adopting either some duty cycle or adjusting the Lyα escape fraction. The multi-parameter models, and their complexities demands a simpler approach to understand the physical nature of Lyα emitters. 1.5 Mass Accretion and Star Formation in Lyα Emitters Mass accretion, and galaxy interactions, play vital role in the star formation history of galaxies. This is certainly true for at least some fraction of galaxies, as evident from observed interacting galaxies in the local as well as distant universe. Such interactions leave a merger-identifiable signature such as disturbed morphology, or tidal tails. While the observed sample of Lyα emitters have been increasing, and several models have been developed, we do not yet have a clear understanding of how mass accretion, and the star-formation occurs in Lyα galaxies. Recent observations have shown that many of the Lyα galaxies show irregular or distrurbed morphologies, indicative of mergers. Theoretical modeling, as well as larger sample of Lyα galaxies are necessary in order to understand physical processes responsible for mass accretion, and star-formation in Lyα emitters. 1.6 Outline In Chapter 2 we discuss the results obtained from a deep, and wide near-ir NB imaging survey in the LALA Cetus survey field, to search for z = 7.7 Lyα galaxies, to use their LF as a test of reionization. The results obtained from the spectroscopic followup observations of four candidate Lyα galaxies identified from the above survey are presented in Chapter 3. To increase the sample of Lyα galaxies at z = 7.7, we performed two additional NB imaging surveys in the Extended Groth Strip (EGS). We put constraints on the Lyα LF at z = 7.7 using these observations in Chapter 4. In Chapter 5, we present the physical model of Lyα emitters, combined with a large 9

24 dark matter simulation. In this Chapter, we compare model predicted LFs, SFRs, stellar masses, and clustering properties of Lyα emitters, with the observations from z 3 to z 7. Finally, in Chapter 6, we predict the merger fractions of Lyα emitters from z 3 to z 7. 10

25 Chapter 2 THE LUMINOSITY FUNCTION OF LYMAN ALPHA EMITTERS AT REDSHIFT Overview Lyman alpha ( Lyα ) emission lines should be attenuated in a neutral intergalactic medium (IGM). Therefore the visibility of Lyα emitters at high redshifts can serve as a valuable probe of reionization at about the 50% level. We present an imaging search for z = 7.7 Lyα emitting galaxies using an ultra-narrowband filter (filter width=9å) on the NEWFIRM imager at the Kitt Peak National Observatory. We found four candidate Lyα emitters in a survey volume of Mpc 3, with a line flux brighter than erg cm 2 s 1 (5σ in 2 aperture). We also performed a detailed Monte-Carlo simulation incorporating the instrumental effects to estimate the expected number of Lyα emitters in our survey, and found that we should expect to detect one Lyα emitter, assuming a non-evolving Lyα luminosity function (LF) between z=6.5 and z=7.7. Even if one of the present candidates is spectroscopically confirmed as a z 8 Lyα emitter, it would indicate that there is no significant evolution of the Lyα LF from z = 3.1 to z 8. While firm conclusions would need both spectroscopic confirmations and larger surveys to boost the number counts of galaxies, we successfully demonstrate the feasibility of sensitive near-infrared (1.06 µm) narrow-band searches using custom filters designed to avoid the OH emission lines that make up most of the sky background. (This Chapter has been published in ApJ 721,1853T The authors are Tilvi V., Rhoads J. E., Hibon P., Malhotra S., Wang J., Veilleux S., Swaters R., Probst R., Krug H., Finkelstein S. L., and Dickinson, M.) 11

26 2.2 Introduction Lyα emitting galaxies offer a powerful probe of both galaxy evolution and the reionization history of the universe. Lyα emission can be used as a prominent signpost for young galaxies whose continuum emission may be below usual detection thresholds. It is also a tool to study their star formation activity, and a handle for spectroscopic followup. The intergalactic medium (IGM) will obscure Lyα emission from view if the neutral fraction exceeds 50% (Furlanetto et al., 2006; McQuinn et al., 2007). Recently, Lyα emitters have been used to show that the IGM is 50% neutral at z = 6.5 (Rhoads & Malhotra, 2001; Malhotra & Rhoads, 2004; Stern et al., 2005; Kashikawa et al., 2006; Malhotra & Rhoads, 2006). This complements the Gunn- Peterson lower bound of x HI 1% at z 6.3. Completely independently, polarization of the cosmic microwave background suggests a central reionization redshift z re = 10.5 ± 1.2 (Komatsu et al., 2010). In addition to their utility as probes of reionization, Lyα emitters are valuable in understanding galaxy formation and evolution at the highest redshifts. This is especially true for low mass galaxies, as Lyα emitters are observed to have stellar masses M 10 9 M (Gawiser et al., 2006; Pirzkal et al., 2007; Finkelstein et al., 2007; Pentericci et al., 2009), appreciably below the stellar masses of Lyman break selected galaxies (LBG) (Steidel et al., 1996) at similar redshifts (e.g. Papovich et al., 2001; Shapley et al., 2001; Stark et al., 2009). Narrowband imaging is a well established technique for finding high redshift galaxies (e.g. Rhoads, 2000a; Rhoads et al., 2004, 2003; Malhotra & Rhoads, 2002, 2004; Cowie & Hu, 1998; Hu et al., 1999, 2002, 2004; Kudritzki et al., 2000; Fynbo et al., 2001; Pentericci et al., 2000; Ouchi et al., 2001, 2003, 2008; Stiavelli et al., 2001; Shimasaku et al., 2006; Kodaira et al., 2003; Ajiki et al., 2004; Taniguchi et al., 2005; Venemans et al., 2004; Kashikawa et al., 2006; Iye et al., 2006; Nilsson et al., 2007; 12

27 Finkelstein et al., 2009). The method works because Lyα emission redshifted into a narrowband filter will make the emitting galaxies appear brighter in images through that filter than in broadbands of similar wavelength. A supplemental requirement that the selected emission line galaxies be faint or undetected in filters blueward of the narrowband filter effectively weeds out lower redshift emission line objects (e.g. Malhotra & Rhoads, 2002). This has proven to be very efficient for selecting starforming galaxies up to z 7, and remains effective even when those galaxies are too faint in their continuum emission to be detected in typical broadband surveys. While large samples of Lyα emitters have been detected at z <6, both survey volumes and sample sizes are much smaller at z >6. Since the Lyα photons are resonantly scattered in neutral IGM, a decline in the observed luminosity function (LF) of Lyα emitters would suggest a change in the IGM phase, assuming the number density of newly formed galaxies remains constant at each epoch. Malhotra & Rhoads (2004) found no significant evolution of Lyα LF between z=5.7 and z=6.6, while Kashikawa et al. (2006) suggested an evolution of bright end of the Lyα LF in this redshift range. At even higher redshifts, z = 6.5 to z=7, some authors (Iye et al., 2006; Ota et al., 2008) suggest an evolution of the Lyα LF however based on a single detection. Recently, Hibon et al. (2010) found seven Lyα candidates at z=7.7 using the Wide-Field InfraRed Camera on the Canada- France-Hawai i Telescope. If these seven candidates are real and high redshift galaxies, the derived Lyα LF suggest no strong evolution from z=6.5 to z=7.7. Stark et al. (2007) found six candidate Lyα emitters at z 8 10 in a spectroscopic survey of gravitationally lensed Lyα emitters. Other searches (e.g. Parkes et al., 1994; Willis & Courbin, 2005; Cuby et al., 2007; Willis et al., 2008; Sobral et al, 2009) at redshift z 8 either had insufficient volume or sensitivity, and hence did not find any Lyα emitters. In this chapter we present a search for Lyα emitting galaxies at z =7.7, selected using a custom-made narrowband filter that avoid night sky emission lines 13

28 and therefore are able to obtain low sky backgrounds. This chapter is organized as follows. In section 2.3, we describe in detail the data and reduction. In section 2.4 we describe our selection of Lyα galaxy candidates. In section 2.5 we discuss possible sources of contamination in the sample, and our methods for minimizing such contamination. In section 2.5 we estimate the number of Lyα galaxy candidates expected in our survey using a full Monte Carlo simulation. In section 2.6 we discuss the Lyα luminosity function, and in section 2.7 we compare the Lyα equivalent widths with previous work. We summarize our conclusions in section 2.8. Throughout this work we assumed a flat ΛCDM cosmology with parameters Ω m =0.3, Ω Λ =0.7, h=0.71 where Ω m, Ω Λ, and h correspond, respectively to the matter density, dark energy density in units of the critical density, and the Hubble parameter in units of 100 km s 1 Mpc 1. All magnitudes are in AB magnitudes unless otherwise stated. 2.3 Data Handling Observations and NEWFIRM Filters We observed the Large Area Lyman Alpha survey (LALA) Cetus field (RA 02:05:20, Dec -04:53:43) (Rhoads et al., 2000b) during a six night observing run with the NOAO 1 Extremely Wide-Field Infrared Mosaic (NEWFIRM) imager (Autry et al., 2003) at the Kitt Peak National Observatory s 4m Mayall Telescope during October 1-6, We used the University of Maryland µm ultra-narrowband (UNB) filter, for a total of 28.7 hours of integration time, along with 5.3 hours integration in the broadband J-filter. Both narrow- and J-band data were obtained on each clear night of observing. NEWFIRM covers a field of view using an array of four detector chips arranged in a 2 2 mosaic, with adjacent chips separated by a gap of 35. Each chip is a pixel ALADDIN InSb array, with a pixel scale of 0.4 per pixel. The instantaneous solid angle coverage of the NEWFIRM camera is about National Optical Astronomy Observatory 14

29 The LALA Cetus field has been previously studied at shorter wavelengths, most notably by the LALA survey (Malhotra & Rhoads, 2002; Wang et al., 2009) in narrow bands with λ c 656, 660, 664, 668, and 672 nm, and λ 80Å); the NOAO Deep Wide Field Survey (NDWFS) (Jannuzi & Dey, 1999), with broadband optical B w, R, and I filters; using MMT/Megacam g, r, i and z filters (Finkelstein et al., 2007); and Chandra, with 180 ksec of ACIS-I imaging (Wang et al., 2004, 2007). In summary, we use narrowband UNB & broadband J data obtained using NEWFIRM, and previously obtained B w, R, and I -band data (NDWFS) for this study. The MMT/Megacam images cover about 55% of the area we observed with NEWFIRM, and we used these deeper optical g, r, i and z images (Finkelstein et al., 2007) to check our final Lyα candidates where possible (see section 2.3 below). The J filter on NEWFIRM follows the Tokunaga et al. (2002) filter specifications, with λ c = 1.25 µm and a FWHM of 0.16 µm. The UNB filter is an ultra narrow-band filter, similar to the DAzLE narrowband filters (Horton et al., 2004), centered at µm with a full width at half maximum (FWHM) of 8.1 Å. We used Fowler 8 sampling (non-destructive readout) in all science frames. In the UNB filter, we used single 1200 second exposures between dither positions; in the J band, two coadded 30 second frames. The NEWFIRM filter wheel places the filters in a collimated beam. As a consequence, the effective central wavelength of the narrowband filter varies with position in the field of view. Beyond a radius of 12, the central wavelength of the UNB filter shifts sufficiently to include two weak OH emission lines in the bandpass, which appear as concentric rings in the narrowband images, and which limit the survey area where the filter s maximum sensitivity (limited by only the inter-line sky background) can be achieved. Figure 2.1 shows the narrowband filter transmission curve along with night sky OH emission lines. The UNB filter is designed to avoid OH lines. 15

30 UNB filter OH lines UNB filter Transmission (%) OH Lines (arbitrary flux) Wavelength (µm) Figure 2.1: Normalized narrowband filter transmission curve (green line) and nightsky OH emission lines (Rousselot et al., 2000) (black line) with arbitrary flux. Here we have shown the transmission curve (at the center of the field) of only narrowband filter to demonstrate the use of very narrow region between OH lines, to search for Lyα emitters at z=7.7. Two weak OH emission lines with λ = and µm (Rousselot et al., 2000) appear in the UNB images as concentric rings beyond 12 radius, since the central wavelength of the UNB filter shifts to the blue for positions away from field center. Data Reduction We reduced UNB & J-band data using a combination of standard IRAF 2 tasks, predominantly from the MSCRED (Valdes, 1998) and NFEXTERN 3 (Dickinson & Valdes, 2009) packages, along with custom IDL 4 reduction procedures. To remove OH rings from UNB data, we created a radial profile for each individual exposure, smoothed over a small radius interval dr, and subtracted this 2 The Image Reduction and Analysis Facility (IRAF) is distributed by the NOAO, which is operated by the Association of Universities for Research in Astronomy, Inc. (AURA) under the cooperative agreement with the National Science Foundation. 3 An external IRAF package for NEWFIRM data reduction 4 Interactive Data Language 16

31 profile from the exposure. We then performed sky subtraction by median averaging two OH ring-subtracted frames that were taken immediately before and two frames after the science frame in consideration. We then performed cosmic ray rejection on sky-subtracted frames, using the algorithm of Rhoads (2000a). Prior to flat-fielding performed using dome-flats, we created a bad pixel mask for each science frame by combining the cosmic ray flagged pixels with a static bad pixel mask for the detector. We then replaced all bad pixels with zero (which is the background level in these skysubtracted images) prior to any resampling of the frames. We adjusted the World Coordinate System of individual frames by matching the point sources to the 2MASS point source catalog using IRAF task msccmatch. We then combined the four chips (i.e four extensions) of each science exposure into a single simple image using the mscimage task in IRAF, which interpolates the data onto a common pixel grid. Here we used sinc17 for the interpolation. Using mscstack in IRAF, we then stacked all of the narrowband exposures into a single, final narrowband stack. Pixels flagged as bad are omitted from the weighted averages in this step. The average FWHM of our final narrowband stack was In addition to this stacked image, we also generated individual night stacks, which were later used to identify glitches in Lyα candidate selection. For broadband J-filter data reduction, we followed essentially the same procedure, modified by omission of the OH ring subtraction which is rendered unnecessary by the absence of noticeable OH rings in the much broader J bandpass. We now assess the accuracy of sky subtraction method, the distribution of noise, and the uncertainty in the astrometric calibration of the UNB and J-band stacks. To evaluate the sky subtraction, and to understand the noise distribution we constructed sky background, and background noise maps using SExtractor (Bertin & Arnouts, 1996). The sky subtraction is sufficiently uniform throughout the image except in the corners i.e. beyond the OH lines affected regions. The noise, due to sky brightness, is also consistent with the expected Poisson noise distribution from sky photons. 17

32 To evaluate the uncertainty in the astrometric calibration, we compared the world coordinates of the sources in the UNB stack (obtained using SExtractor) and the corresponding object coordinates from the 2MASS catalog. We found that the uncertainty in the astrometric calibration is very small, and independent of the position in the UNB image. The rms of the matched coordinates of UNB and 2MASS is about 0.2 and 0.3 arcseconds corresponding to RA and DEC respectively. We obtained reduced stacks of deep optical broadband data in B w, R, and I filters, previously observed by the NOAO Deep Wide Field Survey (NDWFS). At the end, we have one deep UNB stack, along with five single-night UNB stacks, four broadband stacks in J, B w, R, and I filters, and four deep stacks in g, r, i and z (Finkelstein et al., 2009). All the stacks were then geometrically matched for ease of comparison. Photometric Calibration We performed photometric calibration of UNB & J-band (J NF ) data by comparing unsaturated point sources, extracted using SExtractor, with 2MASS stars. From 2MASS catalog we selected only those stars that had J-band (J 2M ) magnitudes between 13.8 & 16.8 AB mag 5, and errors less than 0.1 magnitude. Since four quadrants of the UNB stack had slightly different zero-points, we scaled three quadrants, selected geometrically, to the fourth quadrant, which was closest to the mean zeropoint, by multiplying each quadrant with suitable scaling factors so as to make zero-point uniform throughout the image. We then obtained zero-points for UNB and J NF by minimizing the difference between UNB & J 2M, and between J NF & J 2M respectively. This left 0.09 rms mag between J 2M & J NF magnitudes, and 0.07 rms mag between UNB & J 2M magnitudes. The photometric calibration was based on about 30 & 80 2MASS stars for narrow-band and J-band respectively. So the accuracy of the photometric zero points is about ±0.02 mag in both J and UNB filters. 5 Since J 2M magnitudes are in Vega, we adopted the following conversion between Vega and AB magnitudes : J AB=J 2M mag 18

33 In addition to the error we have already estimated, there is some uncertainty arising due to different filter widths, and differing central wavelengths of the 2MASS and UNB filter. To estimate this uncertainty we constructed observed spectral energy distribution (SED) of stars that were common to both, the UNB image and g, r, i, z, J, H, & K images. From each SED linearly interpolated flux at central wavelengths of the UNB filter and J-filter were measured. From these SEDs we found the median offset between the UNB and J band to be < 0.1 mag. This residual colorterm uncertainty in the photometric zero points is smaller than the photometric flux uncertainty in any of our Lyα candidates. Before we proceed to calculate the limiting magnitudes, we estimate the sky brightness between the OH lines in the UNB image. To estimate this sky value, we construct the UNB stack in the same way as described in Section 2.2 but omitting the OH ring subtraction, and sky subtraction. In addition, we subtracted dark current counts from each raw frame. We estimate the average sky brightness in the UNB image by selecting 30 random regions avoiding astronomical objects and OH rings. This gives us the sky brightness, between the OH lines, of about 21.2 mag arcsec 2 equivalent to 162 photons s 1 m 2 arcsec 2 µm 1. This sky brightness is much fainter than the J-band sky brightness which is about 16.1 mag arcsec 2 equivalent to photons s 1 m 2 arcsec 2 µm 1 (Maihara et al., 1993). However, more careful analysis are needed to estimate the interline sky brightness in the UNB images. Limiting Magnitudes To obtain limiting magnitudes of stacked images, we performed a series of artificial source simulations. In each, we introduced 400 artificial point sources in an 0.1 magnitude bin of flux in the final stacked image. The positions were chosen randomly, but constrained to avoid places close to bright stars and already existing sources. We then ran SExtractor, with the same parameters as were used for the real source detection (see Section 2.4), to calculate the fraction of recovered artificial sources. We ran 20 such simulations in each 0.1 magnitude bin from UNB = 21 to 24 mag. 19

34 The 50% completeness level is UNB= 22.5 mag, which corresponds to an emission line flux of erg cm 2 s 1. The very narrow bandpass results in a relatively bright continuum limit (compared to more conventional narrowband filters with 1% to 1.5% bandpass), but the conversion between narrowband magnitude and line flux is extremely favorable, so that our line flux limits are competitive with any narrowband search in the literature. The 50% completeness for other filters B w, R, I, and J NF correspond to 26.3, 25.4, 25.0, and 23.5 mag respectively. 2.4 Lyα candidate selection We identified sources in the stacked narrowband image using SExtractor. To measure their fluxes at other wavelengths, we first combined the broadband optical images B w, R, & I into a single chi-squared image (Szalay et al., 1999) constructed using Swarp 6 (Bertin et al., 2002). A chi-square image is constructed using the probability distribution of sky pixels in each of the images to be combined, and extracting the pixels that are dominated by object flux. We then used SExtractor in dual-image mode in order to measure object fluxes in both, the broad J filter and the combined optical chi-squared image. In dual image mode, a detection image (UNB in this case) is used to identify the pixels associated with each object, while the fluxes are measured from a distinct photometry image. To identify Lyα candidates in our survey, we used the combined optical image, UNB image, and J-band image. Each Lyα candidate had to satisfy all the following criteria: (a) 5σ significant detection in the UNB filter, (b) 3σ significant narrowband excess (compared to the J band image), (c) flux density ratio f ν (UNB) /f ν (J) > 2, (d) non-detection (< 2σ) in the combined chi-square optical image, (e) consistent with constant flux from night to night (see Section 2.3), and (f) non-detection in individual optical images. 6 Swarp is a software program designed to resample and combine FITS images. 20

35 Criteria a c ensure real emission line sources. Criterion d eliminates most lowredshift sources, e eliminates time variable sources and other glitches, and criterion f eliminates LBGs at z 4 which might show up more clearly in the R or I band than in the χ 2 image. We also used deeper optical images (Finkelstein et al., 2007) in the overlapping field between MMT/Megacam and NEWFIRM for criterion f. The criteria follow the successful searches for Lyα galaxies at lower redshifts of z=4.5 and 5.7, which have 70 80% spectroscopic success rate (Rhoads & Malhotra, 2001; Rhoads et al., 2003; Dawson et al., 2004, 2007; Wang et al., 2009). Constant Flux Test In our constant flux test (criterion e above), we looked at the variation of flux of each Lyα candidate over five nights. We reject any source having individual night stack fluxes close to zero or showing flux variations above a certain chi-square value. To do this, we generated light-curves of each candidate using individual night stacks of UNB, and then determined the χ 2 of the data with respect to the best-fitting constant flux. Since we had five nights of data, we selected only those candidate that had a chi-square <5. This is in addition to requiring s/n > 5, which guards against peaks in the sky noise entering the candidate list. We also eliminated all the sources that were very close to the chip boundaries. Combining these criteria with the set of criteria from Section 2.3, we had six Lyα emitter candidates. To increase the reliability of these candidates, we finally selected four candidates after independent visual inspection by four of the authors. Figure 2.2 shows postage stamps of all four Lyα candidates. The candidates are clearly visible in the UNB images (middle panel), while undetected in the combined optical (left panel), and J band images (right panel). We provide the coordinates of our Lyα candidates in Table

36 Combined optical Narrowband J-band #1 #2 #3 #4 Figure 2.2: Postage stamps (50 wide) of all four Lyα candidates in combined (chisquare) optical image (left panel), UNB (middle), and J-band filter (right panel). The positions of Lyα candidates are marked with circles 16 in diameter. 22

37 Table 2.1: Coordinates of our Lyα Candidates. RA(J2000) DEC (J2000) LAE 1 02:04: :53:00.8 LAE 2 02:04: :00:11.5 LAE 3 02:04: :46:43.8 LAE 4 02:05: :05: Contamination of the Sample While we have carefully selected Lyα candidates based on photometric and geometric criteria, it is possible that our Lyα candidates can be contaminated by sources that include transient objects such as supernova, cool stars (L & T dwarfs), foreground emission line sources, and electronic noise in the detector. We now discuss the possible contribution of sources that can contaminate our Lyα candidate sample. Foreground Emission Line Objects Our Lyα candidate selection can include foreground emission line sources including [OII] emitters (λ = 3727Å) at z=1.85, [OIII] (λ = 5007Å) emitters at z=1.12, and Hα(λ = 6563Å) emitters at z=0.62, if they have strong emission line flux but faint continuum emission. We now estimate the number of foreground emitters that can pass our Lyα candidate selection criteria. In our UNB stack, the 50% completeness limit corresponds to a flux of erg cm 2 s 1. Therefore the minimum luminosities required by the foreground emission line sources to be detected in our survey are erg s 1, erg s 1, and erg s 1 for [OII], [OIII], and Hα emitters respectively. Given the depth of our combined optical image, we can calculate the minimum observer frame equivalent width (EW min ) that would be required for an emission line object to be a Lyα emitter candidate. We calculated the observer frame EW using the following relation (Rhoads & Malhotra, 2001): EW min [ ] fnb 1 f bb λ nb = 23 [ ] 5σnb 1 2σ bb λ nb (2.1)

38 where f nb and f bb are the fluxes in UNB and combined optical image respectively, λ nb is the UNB filter width, and σ nb and σ bb are the uncertainties in flux measurements in UNB and combined optical image respectively. (The implicit approximation that the continuum contributes negligibly to the narrowband flux, is well justified for our 9Å bandpass.) With 5σ nb = erg cm 2 s 1 Hz 1 and 2σ bb = erg cm 2 s 1 Hz 1, we found that the foreground emission line sources would require EW min 460Å to contaminate our Lyα candidate sample. Foreground [OII] ]and [OIII] emitters: Unfortunately, the equivalent width distribution of [OII] emitters has not been directly measured at z=1.85. However, several authors (Teplitz et al., 2003; Kakazu et al., 2007; Straughn et al., 2009) have studied [OII] emitters at z <1.5. Here we use [OII] EW distribution, obtained by Straughn et al. (2009) at z 1 in GOODS-south field, with the assumption that there is no significant evolution of the [OII] LF from z=1 to z=1.85. In our Lyα candidate selection, emission line sources with I AB fainter than 25.9 magnitude, and with EW obs > 460Å can contaminate our sample. We determined which sources from Straughn et al. (2009) would have passed these criteria if redshifted to z = 1.85, and scaled the result by the ratio of volumes between the two surveys. We find that less than one (0.1) [OII] emitter is expected to contaminate our Lyα candidate sample. To be conservative, even if we relax the above magnitude cut by 0.5 mag to account for any color correction, and lower the EW obs > 200Å, we find that less than 0.3 [OII] emitters should be expected to contaminate our sample. We apply a similar methodology to estimate the contamination from foreground [OIII] emitters (Kakazu et al., 2007; Hu et al., 2009; Straughn et al., 2009, 2010) at z 1.1 in our NEWFIRM data using the [OIII] emission line sources at z =0.5 in Straughn et al. (2009). We found that less than two (1.7) [OIII] emitters can be misidentified as Lyα emitters in our survey. In addition to the above estimate, we used a recent sample of emission line galaxies obtained from HST WFC3 early release science data (Straughn et al., 2010). This sample of [OIII] emitters is 24

39 closer in redshift, with median z = 1.1, to our foreground [OIII] interloper redshift of z = 1.12, thus minimizing the error in our [OIII] estimate due to possible evolution in the LF of [OIII] emitters. Using this recent sample, we found that about one [OIII] emitter is expected to contaminate our Lyα candidate sample. Foreground Hα emitters: As mentioned earlier, Hα emitters at z=0.62 can contaminate our Lyα candidate sample. Several authors (Tresse et al., 2002; Straughn et al., 2009) have studied Hα emitters at similar redshift. Tresse et al. (2002) (see their Figure 6) have plotted the Hα luminosity vs the continuum B-band magnitude of Hα emitters. To pass our selection criteria, an Hα emitter would require a luminosity greater than erg s 1, and flux density f Bw < erg cm 2 s 1 Hz 1 which corresponds to M AB = mag. Any source brighter than M AB = mag would be detected in the B w image, and hence rejected from Lyα candidate list. From figure 6 (Tresse et al., 2002), we expect to find no sources that can pass this selection criteria. In addition, we used Hα emitters at z =0.27 (Straughn et al., 2009), and found that less than one (0.4) Hα emitters are expected to contaminate our Lyα candidate sample. Other Contaminants Transient objects : We rule out the possibility of contamination of our Lyα candidates by transient objects such as supernovae, because these objects would appear in both UNB and J band stacks. Both UNB and J data were obtained on each clear night of the run. L and T Dwarfs : Following Hibon et al. (2010) we determined the expected number of L/T dwarfs in our survey. From the spectral type vs. absolute magnitude relations given by figure 9 in Tinney et al. (2003), we infer that we could detect L dwarfs at a distance of 400 to 1300 pc and T dwarfs at a distance of 150 to 600 pc, from the coolest to the warmest spectral types. 25

40 Our field is located at a high galactic latitude, so that we would be able to detect L/T dwarfs well beyond the Galactic disk scale height. However, only a Galactic disk scale height of 350 pc is applicable to the population of L/T dwarfs (Ryan et al., 2005). We derive then a sampled volume of 750 pc 3. Considering a volume density of L/T dwarfs of a few 10 3 pc 3, we expect no more than one L/T dwarf in our field. While we expect about one L/T dwarf in our survey, we further investigate if any of the observed L/T dwarf pass our selection criteria. To do this we selected about 160 observed spectra of L/T dwarf 7 (Golimowski et al., 2004; Knapp et al., 2004; Chiu et al., 2006), and calculated the flux transmitted through the UNB and J-band filter. We found that none of the L/T dwarf has sufficient narrowband excess to pass our selection criteria. Therefore it is unlikely that our Lyα candidate sample is contaminated by L/T dwarf. Noise Spikes: Noise in the detector can cause random flux increase in the UNB filter. To avoid contamination from such noise spikes, we constructed light curves of each candidate using individual night stacks i.e. we selected candidates only if their flux was constant over all nights. This method of candidate selection based on the constant flux in the individual night stacks also eliminates the possible contamination from persistence. Contribution from false detection: Finally, we performed a false detection test to estimate the number of false detection that can pass our Lyα selection criteria. To do this we multiplied the UNB stack by -1 and repeated the exact same procedure as the real Lyα candidate selection (see section 2.4). We did not get any false detection passing our selection criteria. 7 sleggett/ltdata.html 26

41 2.6 Monte-Carlo Simulations Based on the above estimates less than two [OIII] emitters are expected to be misidentified as Lyα emitters in our survey. To estimate the number of sources that should be detected in our survey for a given Lyα LF, we performed detailed Monte-Carlo simulations. This is needed, since the width of the filter is comparable to or slightly smaller than the expected line width in these galaxies, so many of the sources will not be detected at their real line fluxes. In these simulations, we used the z = 6.6 Lyα LF derived by Kashikawa et al. (2006). First, we generated one million random galaxies distributed according to the observed Lyα LF at z=6.6 (Kashikawa et al., 2006). Each of these galaxies was assigned a Lyα luminosity in the range < L Lyα < erg s 1. Here we assumed that the Lyα LF does not evolve from z=6.6 to z=7.7. Each galaxy was then assigned a random redshift z L < z < z H where z L and z H correspond to the minimum and maximum wavelengths where the transmission of the UNB filter drops to zero. Next, to each galaxy we assigned a flux F=L Lyα /4π d 2 L where d L is the luminosity distance. We distribute this flux in wavelength using an asymmetric Lyα line profile drawn from the z = 5.7 spectra of Rhoads et al. (2003). The flux transmitted through the UNB filter was then determined as f trans = f λ T λ dλ (where T λ is the filter transmission and f λ the flux density of the emission line). This accounts for the loss of the Lyα flux that results from a filter whose width is comparable to the line width (and not much greater as would be the case for a 1% filter). We then created a histogram of magnitudes after converting the convolved flux to magnitudes calculated using the following relation: ( ) ftrans mag AB = 2.5 Log 10, (2.2) f 0 and f 0 = 3.6 kjy c (1.06µ) 2 27 T λ dλ erg s 1 cm 2. (2.3)

42 Lastly, to include the instrumental effects, we multiplied the number of galaxies in each magnitude bin by the corresponding recovery fraction obtained from our artificial source simulations in our UNB image(see section 2.3). We then converted each magnitude bin to a Lyα luminosity bin, and counted the number of detected galaxies in each luminosity bin. We repeated this simulation ten times, and taking an average, we found that about one Lyα emitter should be expected in our survey. It should be noted that we assumed a non-evolving Lyα LF from z=6.6 to z=7.7, and that every Lyα emitter has the same asymmetric Lyα line profile. While we expect about one Lyα emitter in our survey there are large uncertainties mainly due to the Poisson noise, and field to field variation or cosmic variance. Tilvi et al. (2009) have estimated field to field variation of Lyα emitters to be 30% for a volume and flux limited Lyα survey with a survey volume Mpc 3. We expect a larger field to field variation for smaller survey volumes. We also estimated the cosmic variance expected in our survey using the cosmic variance calculator (Trenti & Stiavelli, 2008). For our survey we should expect a cosmic variance of about 58% assuming an intrinsic number of Lyα sources at z = 7.7 in agreement with a non-evolving Lyα LF from z = 6.6 (Kashikawa et al., 2006) to z = 7.7. On the other hand our candidate counts are quite consistent with the luminosity function at z=5.7 (Ouchi et al., 2009). 2.7 Lyα Luminosity Function at z=7.7 Using a large sample of Lyα candidates, Ouchi et al. (2008) found no significant evolution of Lyα LF between z=3.1 and z=5.7. The evolution of the Lyα LF between z =5.7 and z = 6.5 is not conclusive. For example, Malhotra & Rhoads (2004) found no significant evolution of Lyα LF between z=5.7 and z = 6.5, while Kashikawa et al. (2006) suggest an evolution of bright end of the LF in this redshift range. On the theoretical front, several models (Thommes & Meisenheimer, 2005; Furlanetto et al., 2005; Le Delliou et al., 2006; Dijkstra et al., 2007; Kobayashi et al., 2007; McQuinn et al., 2007; Dayal et al., 2008; Nagamine et al., 2008; Samui 28

43 10-3 z=7.7 Tilvi et al (This study) z=7.7 Hibon et al 09 z=6.96 Iye et al N(>L) Mpc z=6.5 Kashikawa et al 06 z=5.7 Ouchi et al Lyα Luminosity (erg s -1 ) Figure 2.3: Cumulative Lyα luminosity function of our z=7.7 candidates (filled circles). The filled points show the LF that will result if all four Lyα galaxy candidates are confirmed. The upper error bars are Poisson errors based on our sample size, while the down-arrows below each data point indicate the possibility of a lower LF if some candidates are extreme emission line galaxies at lower redshift. The open circles represent the LF from Hibon et al. (2010) while the dashed line and dotted line show Lyα LFs at z=5.7 (Ouchi et al., 2008) and z=6.5 (Kashikawa et al., 2006) respectively. The open square is the LF at z=6.96 (Iye et al., 2006). et al., 2009; Tilvi et al., 2009) have been developed to predict redshift evolution of the Lyα LF. While several models (e.g. Samui et al., 2009; Tilvi et al., 2009) predict no significant evolution of Lyα LF at z 7, the predictions differ greatly among different models. These differences among the models can be attributed to differing input assumptions, which in turn stem from our imperfect understanding of the physical nature of Lyα galaxies, and from the small samples currently available at high redshift. 29

44 Table 2.2: Lyα Searches at z >7. z Survey Detection limits No. of LAE Ref. volume (Mpc 3 ) erg s 1 candidates This study Hibon et al Stark et al arcmin Parkes et al Willis et al Cuby et al Sobral et al Willis et al 2008 At z > 6.5, there are only a few searches for Lyα emitters. Iye et al. (2006) found one spectroscopically confirmed LAE at z=6.96, and currently there are no spectroscopically confirmed LAEs at z > 7. However, there are few photometric searches (Parkes et al., 1994; Willis & Courbin, 2005; Cuby et al., 2007; Hibon et al., 2010) for Lyα galaxies, and constraints on Lyα LF at z > 7. Table 2.2 shows details of different Lyα searches at z >7. After careful selection of Lyα candidates and eliminating possible sources of contamination, we have found four Lyα emitter candidates in a survey area of arcmin 2, with a limiting flux of erg cm 2 s 1. The fluxes of these four candidates are 1.1, 0.91, 0.84 and 0.72 in units of erg cm 2 s 1. Fig. 2.3 shows the resulting cumulative Lyα luminosity function. Solid filled circles show the Lyα LF derived from our candidates, while open circles represent Lyα LF from Hibon et al. (2010). Arrows indicate that this is the upper limit on the Lyα LF, and upper error bars are the Poisson errors. The dotted and dashed lines show Lyα LFs from Ouchi et al. (2008) and Kashikawa et al. (2006) respectively. The open square is the Lyα LF at z=6.96 (Iye et al., 2006). If all of our Lyα candidates are z=7.7 galaxies, the LF derived from our sample shows moderate evolution compared to LF at z=6.5 (Kashikawa et al., 2006). On the other hand, conservatively if only one of the candidates is a z = 7.7 galaxy, then the Lyα LF does not show any evolution compared to the z = 6.6 Lyα LF. Hibon et al. (2010) conclude that the observed Lyα LF at z = 7.7 does not 30

45 evolve significantly compared to Lyα LF at z=6.5 (Kashikawa et al., 2006), if they consider that all of their candidates are real. Finally, while our Lyα LF lies above the LF obtained by Hibon et al. (2010), the counts are consistent with the number of star-forming galaxies in the HUDF with inferred Lyα line fluxes > erg cm 2 s 1 (Finkelstein et al., 2009b), and also consistent with the Lyα LF at z=5.7 (Ouchi et al., 2008). As described in Section 2.5, all surveys for Lyα emitters at z > 6 suffer from cosmic variance. We do expect to see field-to-field variation in number counts even at the same redshift. Therefore it is important to get statistics from more than one field for each redshift. The field-to-field variation is expected to be stronger for brighter sources. Therefore the higher redshift surveys, which are more sensitivity limited, are hit the hardest. 2.8 Lyα Equivalent Width Several studies have found numerous Lyα emitters having large rest-frame equivalent widths, EW rest > 240Å (Malhotra & Rhoads, 2002; Shimasaku et al., 2006; Dawson et al., 2007; Gronwall et al., 2007; Ouchi et al., 2008). These exceed theoretical predictions for normal star forming galaxies. Since the J-band filter does not include the Lyα line, we have used the following relation to calculate the rest-frame Lyα EWs for our four Lyα candidates: EW rest = f NB 1 f λ,bb (1 + z). (2.4) Here f NB and f λ,bb are the UNB line flux (erg cm 2 s 1 ), and J-band flux (erg s 1 cm 2 Å 1 ) respectively. Since none of the four candidates are detected in J-band, we used J-band limiting magnitude to calculate a lower limit on the Lyα EWs. We note that the Lyα EW will depend on the exact redshift, shape, and precise position of the Lyα line in the UNB filter. However, for simplicity and because we only put lower limits on EWs, we assume that the UNB filter encloses all the Lyα line flux in calculating EWs. 31

46 For our Lyα candidates, with line flux estimates from 7 to erg cm 2 s 1, and our broad band limit J NF 23.5 mag, we find Lyα EW rest 3Å. This EW limit is considerably smaller than the EW rest > 9Å obtained by Hibon et al. (2010) for their Lyα candidates at z=7.7. This difference arises due to the smaller bandwidth of our UNB filter, and our somewhat shallower J band imaging. Deep J-band observations will help in getting either measurements or stricter lower limits on the line EWs, but will also be observationally challenging. 2.9 Summary and Conclusions We have performed a deep, wide field search for z =7.7 Lyα emitters on the NEWFIRM camera at the KPNO 4m Mayall telescope. We used an ultra-narrowband filter with width 9Å and central wavelength of 1.063µm, yielding high sensitivity to narrow emission lines. After careful selection of candidates by eliminating possible sources of contamination, we detected four candidate Lyα emitters with line flux > erg cm 2 s 1 in a comoving volume of Mpc 3. While we have carefully selected these four Lyα candidates, we note that the number of Lyα candidates is more than the expected number obtained by using the z=6.6 luminosity function of Kashikawa et al. 2006, though quite consistent with the z=5.7 luminosity function of Ouchi et al. (2008). Hence, our results would allow for a modest increase in the Lyα LF from z = 6.5 to z 8. Spectroscopic confirmation of more than two candidates would show that such an increase is in fact required. However, more surveys are needed to account for the uncertainty due to cosmic variance. To use the Lyα luminosity functions as a test of reionization, we need to be able to detect variations in L, the characteristic luminosity, of factors of three or four. This will require larger samples, spectroscopic confirmations, and a measure of field-to-field variation. It is therefore premature to draw any conclusions about reionization from the current sample. It is, however, encouraging that we are able to reach the sensitivity and volume required to detect multiple candidates robustly. 32

47 Chapter 3 SPECTROSCOPY OF Lyα GALAXIES AT z 8: ARE WE PROBING A NEUTRAL UNIVERSE? 3.1 Overview The evolution of Lyα luminosity function (LF) at high redshifts provides an unique tool to probe the cosmological reionization, an important phase change in the history of the universe when most of the neutral hydrogen was reionized. Here we present the results of our spectroscopic followup of four candidate Lyα galaxies at z = 7.7, selected from a deep near-infrared narrowband (NB) imaging survey, with Lyα line luminosity > erg cm 2 s 1. All sources were observed with near-infrared (IR) spectrograph NIRSPEC, to a 5σ depth of erg cm 2 s 1. These observations disfavor the brightest source being real at 3.5 sigma, with less stringent constraints on others. By coadding the spectra of all four sources we can achieve a formal 5σ sensitivity of erg cm 2 s 1, we still reach no detection. This may be because not all sources are real or because the redshift alignment is lost. More sensitive observations are needed to confirm or rule out the reality of these candidates. Our current spectroscopic constraint suggest that the IGM is relatively ionized, with x HI 30% even up to z = 7.7. Moreover, it is encouraging that with the current instrumentation we are able to reach the necessary sensitivity in the spectroscopic observations of z 8 galaxies. We also show that the volume, and the survey depth necessary for future Lyα surveys to demonstrate or rule out any evolution in the Lyα LF by a factor of three corresponding to 50% neutral IGM, is within the reach of the current ground based instruments. (This Chapter to be submitted for publication.) 3.2 Introduction Discovering galaxies at z > 7 is essential to our understanding of the first galaxies, as well as to probe the epoch of reionization, the period when the universe transitioned 33

48 from the neutral to the ionized phase. While currently, many candidate Lyα galaxies (e.g. Hibon et al 2010, Tilvi et al 2010, Krug et al 2011), and candidate Lyman-break galaxies (e.g. Bouwens et al 2010,Oesch et al 2010 ) have been identified, only handful galaxies have been spectroscopically confirmed at z > 7 (Lehnert et al 2010, Vanzella et al 2011, Ono et al 2011). Thus, our current understanding of the observed early universe is primarily based on these few confirmed objects, and mostly unconfirmed galaxies at z > 7, and therefore spectroscopic observations are needed for any robust conclusions. While there are different approaches to study the reionization history of the universe, they probe different stages of reionization. For example, WMAP (Komatsu et al. 2010) and quasar (Fan et al 2004) observations are most sensitive to the early and late phases of the reionization, respectively. On the other hand, Lyα galaxies, which have been recently used to study the reionization (Rhoads & Malhotra 2001; Malhotra & Rhoads 2004; Stern et al. 2005; Kashikawa et al. 2006; Malhotra & Rhoads 2006, Ouchi et al 2010), are ideal to study the central stages of reionization when the IGM is, 50% neutral (Furlanetto et al. 2006; McQuinn et al. 2007). The neutral IGM will obscure the Lyα emission from view thereby causing an apparent deficit of observed number density of Lyα galaxies in an neutral IGM. From observations of quasars and Lyα galaxies, our current understanding of the universe at z < 6.6 is that the IGM is < 20% neutral. Therefore, in order to use Lyα LF test effectively it is even more important to discover Lyα galaxies at z > 7 where the IGM > 50% neutral and hence a significant decline in the observed Lyα luminosity function is expected. Currently, there are only two NB selected Lyα galaxies spectroscopically confirmed at z 7 (Iye et al 2006, Hibon et al 2011). This lack is likely due to smaller volumes probed by current high-redshift surveys. The abundance of atmospheric lines in the near-ir worsens this problem. In addition, this lack may be some indication that the IGM is relatively more neutral at z 7, thereby attenuating the Lyα 34

49 emission line flux. There are however, spectroscopic confirmations of few continuum selected LBGs at z > 7. Lehnert et al (2010) confirmed one LBG at z = 8.6, however with relatively weak detection, and two LBGs at z = 7.01 (Vanzella et al 2010). Fontana et al (2010) performed spectroscopic followup of 7 LBGs and possibly confirmed only one galaxy with Lyα emission at z = 6.97 with low signal-to-noise. Ono et al (2011) confirmed one LBG at z = 7.2. Based on these observations, it is possible to estimate the neutral fraction of the IGM, however with large uncertainty due to smaller sample size. Dijkstra et al (2011), using a semi-numerical simulation with empirically calibrated models of galactic outflows, argue that the detection of z = 8.6 LBG is consistent with > 60% neutral IGM by volume. On the other hand, using a set of cosmological simulations with different set of x HI, Dayal et al (2011) find that the detection of the same galaxy requires that the IGM to be < 20% neutral at z = 8.6. Fontana et al (2010), based on their spectroscopic observations, conclude that the lack of Lyα line is possibly due to the IGM being more neutral at z > 7. However, they do not rule out the possibility of their sample being contaminated with low-redshift objects. Thus, currently there is no consensus on the neutral fraction of the IGM at z > 7. To further investigate the nature of the IGM at z > 7, we performed spectroscopic observations of four candidate Lyα galaxies at z = 7.7 identified from a very deep near-ir narrowband imaging survey (Tilvi et al 2010). An unique advantage of Lyα galaxies is that they are easier for the spectroscopic followup since they are selected based on the strong emission line flux, and currently the Lyα emission line remains the primary tool for spectroscopic confirmations of such high-redshift galaxies. In Section 3.3 we describe the candidate Lyα galaxy selection, spectroscopic followup observations and data reduction. In Section 3.4 we present our results and compare our observations with the model predictions. We summarize and present our findings in Section

50 Table 3.1: Summary of Spectroscopic Observations. NB flux Total integration (10 17 erg cm 2 s 1 ) time (hr) 1 LAE LAE LAE LAE LAE NIRSPEC observations 2 MMIRS observations 3.3 Observations and Data Reduction Lyα Candidate Selection The Lyα candidates selected for the spectroscopic followup were identified from a deep near-infrared (IR) narrowband (NB) imaging survey in the LALA Cetus survey field (Rhoads et al 2000b). These candidates were selected using the selection criteria that has proven to have 70 80% spectroscopic success rate at z < 6 (e.g. Rhoads & Malhotra 2001; Rhoads et al. 2003; Dawson et al. 2004, 2007; Wang et al. 2009). More details about the candidate selection, identification of low-redshift foreground emission line galaxies, and possible contamination from foreground galaxies are described in Tilvi et al (2010). Here we summarize our survey, and the candidate selection process. The near-ir NB imaging survey was carried out using the NOAO Extremely Wide-Field Infrared Mosaic (NEWFIRM) imager (Autry et al 2003) on the Kitt Peak 4m Mayall telescope, during 2008 October 1-6. The Lyα candidates were carefully selected eliminating possible contamination from the foreground emission line galaxies, and spurious sources including electronic noise, and transient sources. From this deep and wide survey, we found four candidate Lyα galaxies at z=7.7, with Lyα flux > erg cm 2 s 1. The NB fluxes of individual candidate are listed in Table

51 Spectroscopic Observations We performed near-ir spectroscopic observations of candidate Lyα galaxies at z = 7.7 using two different instruments, the NIRSPEC (McLean 1998) on the Keck- II at Mauna Kea, and the MMIRS on the Magellan telescope at Las Campanas. NIRSPEC Observations We observed all four candidate galaxies using the NIRSPEC during August 03, and September 17, We obtained the data using long-slit ( ) low resolution mode, NIRSPEC-N1, covering µm wavelength range with a spatial resolution of per pixel. For each target the total integration time was about 2 to 3 hrs where each individual exposure was 15 to 20 minute long with single coadd. In order to optimize the sky-subtraction, each object was dithered along the slit with an offset of about ±3. In Table 3.1 we summarize the total integration time for each of the target. Since our targets were faint enough to be invisible in the individual frames, we followed invisible data acquisition method in order to make sure that the science target was aligned in the slit center. In this acquisition mode, the telescope is first centered on a bright star near the target, and then an appropriate offset is applied in order to center the object on the slit. In addition, we aligned the slit on our target such that the slit also shared another brighter continuum source (see Figure 3.1) which allows us to accurately stack individual dithered frames. MMIRS Observations In addition to the NIRSPEC observations, we performed spectroscopic observations of our brightest candidate Lyα galaxy using the MMIRS instrument in the multislit mode during October 26-27, However, due to bad weather conditions, we obtained useful data only for a single night. 37

52 Figure 3.1: Slit configuration for the brightest candidate Lyα galaxy (circled) in the NB1063 image. The slit shares another continuum object which allows for accurate stacking of dithered frames. The MMIRS detector is a pixels HdCdTe Hawaii-2 array with the slit mask field of (McLeod et al 2004). In addition to our brightest Lyα candidate galaxy, we observed foreground emission line galaxies using remaining slitlets on the mask. Four bright stars were included for the alignment along with addition continuum sources for flux calibration. Each exposure was 300 sec long, with ramp-up configuration which allows to remove cosmic rays effectively. 2-D Spectra Extraction NIRSPEC The 2D spectra were reduced using the NIRSPEC reduction package (developed by G. Becker; private communication). In this reduction, the spectra were first flat-fielded 38

53 Figure 3.2: Top four panels: 2-D images of sky-subtracted spectra of all four objects. The circles enclose the expected position of the Lyα galaxies. Bottom panel: 2-D spectrum showing the positions of strong night skylines (dark means brighter). 39

54 & corrected for dark current, and then sky-subtracted using optimal sky-subtraction technique (Kelson 2003). This technique attains sky-subtraction by pixel subsampling of the raw (distortion uncorrected) spectra thereby improving sky-subtraction significantly. The sky-subtracted spectra are then corrected for both, x and y distortion using the IRAF 1 tasks xdistcor and ydistcor in the wmkonspec package 2 developed for the NIRSPEC data reduction. Cosmic rays were identified using the IRAF task crmedian, and the affected pixels were replaced by average counts calculated from neighboring pixels. Each individual spectra were then average combined together using IRAF task imcombine, by aligning the continuum source that shared the slit. The top four panels in Figure 3.2 shows the sky-subtracted spectra of all four objects, with expected position shown by a circles. Note the expected position of our Lyα galaxy, is free from the OH lines (bottom panel). The UNB filter was custom-designed to avoid bright nightsky lines. MMIRS For MMIRS multislit data, each exposure was dark subtracted using a master dark created from multiple dark exposures with the same configuration as that of the science frames. Since each science frame is observed in the ramp-up configuration, we used a custom designed task mmf ixen, to collapse the multi-extension data cube into a single extension file. This task also performs cosmic ray rejection taking the advantage of the ramp-up configuration. We then flat-field all science frames using a master flat created from multiple flat frames. Since each frame contains multiple object spectra, we cut only Lyα candidate spectrum using the IRAF task imcopy. The sky was then subtracted by first creating a median filtered science image, and then subtracting this median filtered image from 1 The Image Reduction and Analysis Facility (IRAF) is distributed by the NOAO, which is operated by the Association of Universities for Research in Astronomy, Inc. (AURA) under the cooperative agreement with the National Science Foundation

55 the science image itself. We used an IRAF task crmedian to further clean images from comic rays. All the cleaned science frames are then stacked together using the IRAF task imcombine to create a stacked 2-D spectrum of our brightest candidate Lyα galaxy. 1-D Spectra and Flux Calibration For the NIRSPEC data, we extracted 1-D spectra from 2-D spectra by summing the counts in an aperture with a width of 1. The wavelength calibration was computed using the bright nightsky lines. The flux calibration was performed using the spectrophotometric standard stars observations. To do this, we first chose an appropriate Pickles model spectrum, and scaled it to make its V mag same as the observed V mag of the standard star. The sensitivity function was then created by dividing the scaled Pickles spectrum by the observed spectrum of the standard star (counts/s). Finally, to get the flux calibrated spectrum, the sensitivity function was multiplied by the 1-D science spectra (counts/s). Figure 3.3 (upper panel) shows the extracted 1-D spectrum for the brightest Lyα candidate galaxy. The Lyα emission line redshifted to z = 7.7 is expected to be at 1.063µm. We followed a similar procedure as above to extract 1-D spectra from the MMIRS 2-D spectra. In the absence of standard star observations, the flux calibration was performed by scaling the spectra of a continuum source to its J-band magnitude. Survey Sensitivity To understand the noise properties, and flux limit of our spectroscopic observations, we now determine the limiting flux as a function of wavelength. Figure 3.3 (middle panel) shows 1σ (square root of variance) limiting flux density for our brightest object. At 1.063µm, the expected position of our science objects, the limiting flux density is nearly minimum. Thus, in 1 aperture and assuming 10Å line width, we find 5σ limiting line flux erg cm 2 s 1. This flux limit is comparable to the sensitivity limit obtained by Stark et al (2007). For remaining three candidates the 41

56 4 F λ ( erg s -1 cm -2 A -1 ) σ Limiting Flux (10-18 erg s -1 cm -1 A -1 ) F λ ( erg s -1 cm -2 A -1 ) λ (µm) Figure 3.3: Top panel: 1-D spectrum of our brightest object. Middle panel: 1σ limiting flux density for our spectra. Note the minimum flux density at µm, the expected position of our science objects. Bottom panel: Coadded spectra of all four objects. 42

57 sensitivity limit correspond to 0.98, 1.5, and 0.8 in units of erg cm 2 s 1, higher than the flux estimated from NB imaging. From the MMIRS spectra, for the brightest candidate, we found this limit to be about erg cm 2 s 1. Since all four candidate galaxies are observed using NIRSPEC, and due to its better sensitivity limit, we focus our analysis and discussion based on the NIRSPEC observations alone. 3.4 Results and Discussion From our current spectroscopic sensitivity limit, we rule out the detection of only the brightest candidate at 5σ level. Except for our brightest candidate, we can not rule out the detection of remaining three candidates since our spectroscopic survey limit is brighter compared to the corresponding NB fluxes of these candidates. While it is possible that the brightest NB selected candidate is either a spurious source or a transient object that passed our candidate selection criteria, for robust conclusions on the remaining three candidates, future, more sensitive spectroscopic observations are needed. The coadded spectra of all four objects allowed us to reach line luminosity erg cm 2 s 1. However, in the coadded spectra, we did not detect any candidate at this flux level. Thus, while non-detection of our brightest candidate galaxy suggest that it is challenging to discover galaxies at such high-redshifts, it is still encouraging that the current spectroscopic observations can reach the required sensitivity to discover galaxies at z 8. It is also possible that the non-detection of the brightest candidate might be an indication of relatively more neutral IGM. Before we discuss the implications of our results on the reionization, we put limits on the observed Lyα LF at z = 7.7 and compare it with other observations at similar redshifts. In Figure 3.4 we show cumulative Lyα LF limit and compare it with previous studies. The shaded region is the Lyα luminosity limit obtained from this spectroscopic observations, while the dotted line indicates the z = 5.7 Lyα LF from Ouchi et al (2008). We compare our spectroscopic limit with z = 5.7 Lyα LF since 43

58 10-3 Flux at z=7.7 (10-17 erg s -1 cm -2 ) z=6.6 N(>L) Mpc z=7.7 Tilvi et al 10 z=7.01 Vanzella et al 11 z=6.96 Iye et al 06 z= z=5.7 Ouchi et al 08 z=6.6 Ouchi et al Log Lyα Luminosity (erg s -1 ) 10-6 Figure 3.4: Constraints on the Lyα LF at z = 7.7. Shaded region indicates our spectroscopic constraint, while diagonal symbols show continuum selected LBGs with strong Lyα emission at z = 7 (Vanzella et al 2011). Black filled circles are the NB selected candidates from Tilvi et al (2010). The z = 6.6 Lyα LF is shown with dotted line. it has been shown that the Lyα LF does not evolve up to redshift z = 5.7, and that the reionization is complete at z 6. Our spectroscopic limit is consistent with the observed LF at z = 5.7. In Figure 3.4 we also show the data points (diamond symbols) from spectroscopically confirmed LBGs at z = 7.01 (Vanzella et al 2011). These LBGs should formally pass all the selection criteria of Lyα galaxy selection since their rest-frame equivalent widths are > 50Å. The Lyα line fluxes for these two LBGs are 1.62 and erg cm 2 s 1, comparable to our 5σ line limiting flux. These data points should be treated as lower limit since in the continuum selected sample it is possible that the Lyα galaxies with very faint continuum (very high Lyα equivalent width) can be missed. Thus, within the observational uncertainties, our spectroscopic limit 44

59 on the Lyα LF is consistent with no evolution in the Lyα LF. However, future spectroscopic observations of remaining candidates are needed for robust determination of the Lyα LF. In the following sections we estimate the expected number of Lyα sources in our survey, compare our observations with the model predictions, and discuss the implications of these observations on the reionization history of the universe. Expected Number of Lyα Sources In order to understand the likelihood of our candidates, in Tilvi et al (2010) we had performed a detailed Monte-Carlo simulation, to estimate the number of Lyα galaxies expected in our NB imaging survey. This simulation included the instrumental effects including the filter transmission curve, survey incompleteness, and asymmetric Lyα line profile. The expected number is based on a non-evolving Lyα LF at z = 6.6 (Kashikawa et al ). While we had found four candidate Lyα galaxies at z = 7.7, we expect about one Lyα galaxy based on our simulation results. However, at such high-redshifts, field-to-field variations due to clustering are expected to be large 30 60% (Tilvi et al 2009, Trenti & Stiavelli 2008), and adds another uncertainty to the expected number of Lyα sources. Comparison With Model Predictions While we expect one Lyα galaxy in our NB survey, the non-detection of the brghtest candidate may also imply that the observed number density of Lyα galaxies decreases due to the IGM becoming increasingly more neutral. There is however, additional uncertainty when comparing the LF between two redshifts, which is due to the redshift evolution in the intrinsic number density of Lyα galaxies. Thus, in order to isolate the effect of evolution of intrinsic number density of Lyα galaxies from the IGM effect, we now compare our expected number of sources with the model predictions, without IGM correction. 45

60 Tilvi et al (2009) developed a simple physical model of Lyα emitters, which is successful in reproducing many observed properties including Lyα LF, star-formation rate, stellar masses, and clustering properties of Lyα emitters from z 3 to z 7. The central idea of this model is that the star-formation rate, and hence the Lyα luminosity, is proportional to the mass accretion rate rather than the total halo mass. We use this model to predict the Lyα LF at z = 7.7, and compare it with our spectroscopic limit. Since this model does not incorporate the effect of IGM, any evolution in the Lyα LF can be purely attributed to the redshift evolution of Lyα galaxies. In other words, a significant decline in the observed Lyα LF would indicate a significantly neutral IGM. In Figure 3.5 we compare the model predicted Lyα LF at z = 7.7 with the spectroscopic observations at similar redshifts. Since our current Lyα LF limit is larger than the predicted LF at z = 7.7, this suggest that the non-detection of even two of our brightest candidate will make the LF consistent with z = 5.7 LF (Ouchi et al 2008). 3.5 Is the Universe Neutral at z = 7.7? To understand, and quantify the evolution of neutral fraction of IGM at z > 7 we compare the model predicted Lyα LF, with and without the effect of IGM, with the observations. In Figure 3.5 we scale our model predicted LF by the scaling factors obtained from McQuinn et al (2007; see their Figure 3.4). We show predicted Lyα LF at x HI = 30% (dashed line) and for x HI = 50%. From recent observations it is now known that the reionization is complete by z 6 with x HI 1%. On the other hand, from electron scattering from microwave background radiation which is sensitive to free electrons, the instantaneous redshift of reionization z 11 (Komatsu et al 2010). However, reionization is likely to be a gradual process and might have started as early as z Recently, using a large sample of Lyα galaxies, Ouchi et al (2010) estimated that x HI < 20% at z = 6.6, consistent with earlier studies (e.g. Rhoads & Malhotra 46

61 10-3 Flux at z=7.7 (10-17 erg s -1 cm -2 ) z=7.7 model prediction constraint from this study 10-3 x HI =30% 10-4 x HI =50% 10-4 N(>L) Mpc z=7.01 Vanzella et al 11 z=6.96 Iye et al Lyα Luminosity (erg s -1 ) Figure 3.5: Comparison of model Lyα LF (dashed line) with the spectroscopically confirmed galaxies at z 7. The dot-dashed line is the Lyα LF at z=7.7 in a 50% neutral IGM. We scaled our model predicted LF using a scaling factor obtained from McQuinn et al (2007). The scaled LF for 30% neutral fraction is also shown (dotted line). 2001; Malhotra & Rhoads 2004; Stern et al. 2005; Kashikawa et al. 2006; Malhotra & Rhoads 2006). At slightly higher redshift, Fontana et al (2010) performed followup spectroscopy of seven LBGs, using FORS2 instrument on the VLT. They detected a weak emission line, possibly a Lyα line redshifted to z = They suggest that the lack of Lyα emission in most of the LBGs are either due to them being lowredshift emission line galaxies or that the IGM is more neutral at z > 6.5, quenching the Lyα emission line flux. Vanzella et al (2011) confirmed two LBGs at z = 7.01 using FORS2 instrument on the VLT. These LBGs show large Lyα equivalent width 50Å, and Lyα emission line flux erg cm 2 s 1. We treat these LBGs as Lyα galaxies 47

62 due to their large Lyα EWs. These observations (diamond symbol) are consistent with the expected Lyα LF for x HI 50%. At higher redshift z = 7.7, Sobral et al (2009) found two bright candidate Lyα galaxy, both of which were ruled out as high-redshift galaxies following the spectroscopic observations. They reached a limiting magnitude > erg s 1, much brighter than our current observations limit. Stark et al (2007) performed spectroscopic survey of gravitationally lensed galaxies at 8.7 < z < 10.2, using NIRSPEC instrument on the Keck telescope. They found that two of the five candidates are likely to be z 10 Lyα emitters. At slightly higher redshift, using recently discovered LBG at z=8.6 (Lehnert et al 2010), Dijkstra et al (2011) estimated the neutral fraction of the IGM, using semi-numerical simulations with empirically calibrated models of galactic outflows. They estimate that at this redshift, the IGM is highly neutral with x HI > 60%. On the other hand, using the same object, Dayal et al (2011) argue that the detection of this object at this redshift implies that x HI < 20%. They also suggest that there should be additional ionizing sources clustered around this discovered galaxy. The observed Lyα LF limit from our spectroscopic observations is consistent with no evolution in the Lyα LF from z = 5.7 to z = 7.7. Thus, the picture that emerges from current observations is that the reionization is complete by z 6 with x HI 1%, and at z=6.6, the IGM is still largely ionized with x HI 20%. At z = 7.7, further, more sensitive observations are needed to put stricter constraints on the Lyα LF. 3.6 Future Lyα Surveys While currently it is challenging to discover such high-redshift galaxies from the ground, the ground-based telescopes offer an advantage of much larger survey volumes, which is critical for discovering larger sample of galaxies needed for statistically robust conclusion. It is also very encouraging that the number of candidate Lyα galaxies selected via NB imaging has been growing (Krug et al 2011, Tilvi et al in 48

63 prep). In this section we discuss a strategy for future NB imaging surveys at z > 7. In particular we estimate the volume, and the survey depth needed to put robust constraints on the state of the IGM at z > 7. In Figure 3.5 we show Lyα LF at z=7.7 as it would appear in an 50% neutral IGM (dotted line). Thus, in a 50% neutral IGM we expect a decline in the Lyα LF by a factor of about three at z=7.7. Therefore, in order to either demonstrate or rule out an evolution by a factor of three, future NB imaging surveys are required to sample a volume Mpc 3, and reach survey depth equivalent to an emission line flux of about erg cm 2 s 1 equivalent to erg s 1 at z=7.7. A NB filter ( 35Å wide ) custom-designed to include all the emission line flux, and at the same time narrow enough to avoid the night sky lines thereby decreasing the sky level, should allow us to probe this volume and the survey depth with a modest telescope time of about 6-7 nights with near-ir instruments including the NEWFIRM. Such surveys in multiple survey fields, to minimize the effect of cosmic variance, should put definitive constraints on the state of the IGM at z Summary and Conclusions In order to study the evolution of Lyα LF, and to constrain the neutral fraction of the IGM at z = 7.7 we performed a near-ir spectroscopic followup of four z = 7.7 candidate Lyα galaxies identified from a deep and wide near-ir NB imaging survey in the LALA Cetus field, in a survey volume of about Mpc 3. Using the NIRSPEC in the long-slit mode, and MMIRS in the multi-slit mode, we rule out the detection of our brightest candidate galaxy at about 5σ significance level corresponding to line flux of erg cm 2 s 1. Future, deeper spectroscopic observations are needed for robust conclusions about the remaining three candidate galaxies. By combining spectra of all four objects we reached a 5σ sensitivity limit of erg cm 2 s 1. While we did not detect any object at this line flux limit, it is promising that with the current instrumentation we can reach the sensitivity necessary for the spectroscopic observations of z 8 galaxies. 49

64 Based on our simulations, while we expect about one Lyα galaxy in our survey volume, field-to-field variations likely affects the observed number of galaxies. Our current spectroscopic limit is consistent with the observed Lyα LF at z = 5.7. Comparison of observations with the model predictions suggest that the IGM is mostly ionized even at z = 7.7, with x HI 30%. While it is challenging to discover galaxies at z > 7 from the ground, the ground based telescopes, as opposed to the space telescopes, have a distinct advantage of surveying much larger volume. We propose that, in order to increase the sample of Lyα galaxies at z > 7, and put robust constraints on the evolution of the IGM at this redshifts, a modest investment of telescope time of 6-7 nights should allow us to reach the necessary survey depth. Such 3-4 surveys in different fields will further minimize the effect of cosmic variance on the observed number density of Lyα emitters, thereby allowing us to rule out or confirm the evolution of IGM at z=

65 Chapter 4 Lyα EMITTERS AT z 8: CONSTRAINTS ON THE LUMINOSITY FUNCTION FROM THE DARK AGES SURVEY 4.1 Overview Lyα luminosity function (LF) provides a direct probe of the cosmological reionization since Lyα line flux is attenuated by neutral hydrogen consequently decreasing the observed number density of Lyα galaxies in an neutral intergalactic medium (IGM). Here we present the results of our two wide near-infrared narrowband imaging survey to search for z = 7.7 Lyα galaxies in the Extended Groth Strip (EGS) using two custom-designed ultra-narrowband filters. We found one candidate Lyα galaxy (in only one filter) at z = 7.7 in a survey volume of about Mpc 3, consistent with our expected number of sources estimated from a detailed Monte- Carlo simulation. We reached a 5σ limiting flux > erg cm 2 s 1 in a 2 aperture diameter. While the detection of one candidate galaxy is consistent with no evolution in the Lyα LF from z = 5.7 to z = 7.7, we combined this observation with previous surveys performed using same instruments, to put much tighter constraint on the Lyα LF at z = 7.7. Using these combined observations we found that there is a mild evolution in the Lyα LF from z = 5.7 to z = 7.7, however consistent with the observed Lyα LF at z = 6.6. Comparison of this LF with the model predictions suggests that the IGM is still largely ionized even at z = 7.7 with neutral fraction x HI 30%. Moreover, in order to demonstrate an evolution in the Lyα LF due to 50% neutral IGM, would require none of the candidates to be real z = 7.7 galaxies, and thus warrants sensitive followup spectroscopic observations. (This Chapter to be submitted for publication). 4.2 Introduction Observations of the earliest galaxies provide clues to the formation of the primeval galaxies, and the reionization history of the universe. Lyα emission from star- 51

66 forming galaxies is an important tool to discover very high-redshift, z 7 (Iye et al 2006) galaxies when the universe is merely < 800 Myr. In addition, evolution of Lyα luminosity function (LF) is a valuable probe of the epoch of reionization, the period during which most of the neutral hydrogen in the universe was ionized. Current observations suggest that the reionization is nearly complete at z 6 with lower bound on neutral fraction x HI 1%. On the other hand, the constraints from the WMAP suggest that the early stages of reionization occured somewhere around z 11 (Komatsu et al., 2010). Several studies (e.g. Rhoads & Malhotra, 2001; Malhotra & Rhoads, 2004; Stern et al., 2005; Kashikawa et al., 2006; Malhotra & Rhoads, 2006), using Lyα LF have shown that the IGM < 50% neutral at z = 6.5 1, consistent with the recent observations indicating x HI 20% (Ouchi et al 2010). The Lyα LF works as a good indicator of the evolution of the IGM, and hence the reionization epoch because the neutral IGM will obscure Lyα emission from view if the neutral fraction exceeds 50% (Furlanetto et al., 2006; McQuinn et al., 2007). This will result in an apparent deficit in the observed number density of Lyα emitters in an neutral IGM. Apart from using Lyα galaxies as a probe of reionization, they are also a valuable tool to understand low mass galaxy formation since Lyα galaxies observed at high redshifts have smaller stellar masses with M 10 9 M (Gawiser et al., 2006; Pirzkal et al., 2007; Finkelstein et al., 2007; Pentericci et al., 2009), and smaller sizes (Malhotra et al 2011). These observations will in turn provide strong constraints on the galaxy formation models(e.g. Tilvi et al 2009). Lyα searches using NB imaging have been very successful in discovering upto z 7 galaxies (e.g. Cowie & Hu, 1998; Hu et al., 1999; Kudritzki et al., 2000; Pentericci et al., 2000; Fynbo et al., 2001; Stiavelli et al., 2001; Ouchi et al., 2001; Hu et al., 2002; Rhoads et al., 2000b; Malhotra & Rhoads, 2002; Ouchi et al., 2003; Rhoads et al., 2003; Kodaira et al., 2003; Rhoads et al., 2004; Malhotra & Rhoads, 1 Age of the universe at z=7.7, 6,6, and 5.7 corresponds to 686 Myr, 840 Myr, and 1014 Myr respectively. 52

67 2004; Hu et al., 2004; Ajiki et al., 2004; Venemans et al., 2004; Taniguchi et al., 2005; Shimasaku et al., 2006; Kashikawa et al., 2006; Iye et al., 2006; Nilsson et al., 2007; Ouchi et al., 2008; Finkelstein et al., 2009). With the advent of wide and sensitive near-ir detectors (e.g. NEWFIRM) these searches have now been extended to even higher redshifts, z > 7 ( Hibon et al 2010, Tilvi et al 2010, Krug et al 2011,Clement et al 2011). Current observations of Lyα galaxies are consistent with a mild evolution with an upper bound of < 50% in the Lyα LF from z 5.7 to z = 6.6 (Malhotra & Rhoads 04, Ouchi et al 2010, Kashikawa et al 2011) suggesting that the IGM is predominantly ionized even up to z = 6.6. At slightly higher redshifts, from z = 6.5 to z=7, some authors (Iye et al., 2006; Ota et al., 2008) suggest an evolution of the Lyα LF however based on a single detection. Some of this decline in the observed Lyα LF can be attributed to the evolution in the intrinsic number density of galaxies (e.g. Tilvi et al 2009, Ouchi et al. 2009b; Bouwens et al.2010a; Castellano et al. 2010). Therefore, in order to probe the neutral IGM at > 50% level it is necessary to probe even higher redshifts where we expect to see a more dramatic evolution in the Lyα LF. Recently, several NB imaging surveys have been carried out to search z > 7 Lyα galaxies. Hibon et al. (2010) found seven Lyα candidates at z=7.7 using the Wide-Field InfraRed Camera on the Canada- France-Hawai i Telescope. Tilvi et al (2010) & Krug et al (2011), each found four candidate Lyα emitting galaxies at z = 7.7 using the NEWFIRM instrument on the Kitt Peak 4m telescope, in the LALA Cetus and the COSMOS field respectively. Recently, Clément et al. (2011) performed NB imaging to search z = 7.7 Lyα galaxies, and found no candidates in a survey volume of Mpc 3. Other searches (e.g. Parkes et al., 1994; Willis & Courbin, 2005; Cuby et al., 2007; Willis et al., 2008; Sobral et al, 2009) at redshift z 8 either had insufficient volume or sensitivity, and hence did not find any Lyα emitters. While none of the candidates at z = 7.7 are spectroscopically confirmed yet, 53

68 these ambitious searches suggest that with the current instrumentation it is possible to identify candidate galaxies even up to z 8, and morevover, Lyα searches using NB imaging, as against LBGs, has a distinct advantage for the spectroscopic followup, compared to the broadband selected, since these galaxies are selected based on the strong emission line. Based on our previous studies (e.g. Tilvi et al 2010, Krug et al 2011), we found that in order to minimize the effect of cosmic variance (Trenti et al 2008, Tilvi et al 2009), we need, not only multiple survey fields but also much larger survey volume. In this paper we present the results of our two NB imaging surveys to search for z=7.7 Lyα galaxies in the EGS field, using two custom-made ultra-narrowband filters with central wavelengths designed to avoid night sky emission lines in order to lower the sky background. We also combine these observations with previously studied three additional fields: LALA Cetus survey field (Tilvi et al 2010), and two surveys in the COSMOS field (Krug et al 2011), to put much stronger constraint on the Lyα LF at z = 7.7. This chapter is organized as follows. In section 4.3, we describe in detail the data and reduction. In section 4.4 and section 4.5 we describe our selection of Lyα galaxy candidates. In section 4.6 we discuss possible sources of contamination in the sample, and our methods for minimizing such contamination. In section 4.6 we estimate the number of Lyα galaxy candidates expected in our survey using a full Monte Carlo simulation. In section 4.7 we discuss the Lyα luminosity function, and compare our observations with model predictions in section 4.8. We discuss the implications of observations on the reionization. We summarize our conclusions in section Throughout this chapter we assumed a flat ΛCDM cosmology with parameters Ω m =0.3, Ω Λ =0.7, h=0.71 where Ω m, Ω Λ, and h correspond, respectively to the matter density, dark energy density in units of the critical density, and the Hubble parameter in units of 100 km s 1 Mpc 1. All magnitudes are in AB magnitudes unless otherwise stated. 54

69 4.3 Survey and Data Reduction Observations and NEWFIRM Filters As part of the Dark Ages Survey (DAS) We performed narrowband imaging in the Extended Groth Strip (EGS; RA +14:18:01 Dec +52:36:37 ) during 2008 and 2009 using the NOAO 2 Extremely Wide-Field Infrared Mosaic (NEWFIRM) imager (Autry et al., 2003) at the Kitt Peak National Observatory s 4m Mayall Telescope. The EGS field was chosen for the availability of the ancillary multiwavelength public data in the broadband optical as well as infrared data. In particular, deep optical g, r, i, z are publicly available from the CFHTLS, and J broadband is available from the WIRDS archive. We obtained about 30 and 24 hours of total integration time using the University of Maryland & µm ultra-narrowband (UNB) filter respectively, on each clear night. NEWFIRM covers a field of view using an array of four detector chips arranged in a 2 2 mosaic, with adjacent chips separated by a gap of 35. Each chip is a pixel ALADDIN InSb array, with a pixel scale of 0.4 per pixel. The instantaneous solid angle coverage of the NEWFIRM camera is about 745. The UNB filters are ultra narrow-band filters, similar to the DAzLE narrowband filters (Horton et al., 2004), centered at 1.056, and µm with a full width at half maximum (FWHM) of 7.4 & 8.1 Å respectively. In each UNB filter, we used a single 1200 second exposures between dither positions, with Fowler 8 sampling (non-destructive readout). While the central wavelengths of the UNB filters are designed to avoid the night sky OH lines, beyond a radius of 12 from the center, the effective central wavelengths shifts sufficiently to include two weak OH emission lines in the bandpass, which appear as concentric rings in the narrowband images. 2 National Optical Astronomy Observatory 55

70 UNB filter Transmission (%) Lyα Redshift UNB filter OH lines OH Lines (arbitrary flux) Wavelength (µm) Figure 4.1: Normalized narrowband filter transmission curves (dahed-blue lines), and night-sky OH emission lines (Rousselot et al., 2000) (black line) with arbitrary flux. Here we have shown the transmission curve (at the center of the field) of both narrowband filters: UNB1056 & UNB1063 filter. to demonstrate the use of very narrow region between OH lines, to search for Lyα emitters at z=7.7. Data Reduction We processed the UNB data using custom designed IDL 3 reduction procedures, and the nfextern 4 (Dickinson & Valdes, 2009) IRAF 5 package. We followed an identical data reduction procedure as described in Tilvi et al (2010). Here, we briefly describe this procedure. First we perform sky subtraction using the median averaged frame created using two frames before and two frames after the frame into consideration. The sky subtracted frame is then flat-fielded using the master dome flat created for that night. We performed OH ring subtraction by creating a radial profile for each individual 3 Interactive Data Language 4 An external IRAF package for NEWFIRM data reduction 5 The Image Reduction and Analysis Facility (IRAF) is distributed by the NOAO, which is operated by the Association of Universities for Research in Astronomy, Inc. (AURA) under the cooperative agreement with the National Science Foundation. 56

71 exposure, smoothed over a small radius interval dr, and subtracted this profile from the exposure. The bad pixels masks for individual frames were created by combing static bad pixel mask and the cosmic-rays affected pixels which were identified using the algorithm of Rhoads (2000a). All the bad pixels were replaced with zero counts which is the background level in these sky-subtracted images. We adjusted the World Coordinate System of each individual frame using the IRAF task msccmatch, and matching the point sources to the 2MASS stars. Before stacking all the frames using the task mscstack, we combined all the four chips into a single mosaic using the IRAF task mscimage which interpolates (using sinc17), the data onto a common pixel grid. Since we had two epoch data for both filters, we created one master stack for each filter as well as separate stack for each epoch data. The stack for separate stack were used to investigate the reliability of the Lyα candidate. The median FWHM of our final narrowband stacks were 1.3 and 1.2 for UNB1056, and UNB1063 respectively. The broadband optical and J-band data were already reduced, and were obtained from public CFHTLS archieve. UNB Stack Quality Assessment In this section we asses the astrometric uncertainty, and the noise distribution in our UNB stacks. The astrometric uncertainty arises due to the errors when matching the UNB point sources with the 2MASS point sources. We used SExtractor (Bertin & Arnouts, 1996) to identify the point sources, and their WCS coordinates, and compared these coordinates with the corresponding sources in the 2MASS catalog. We found that the rms of the matched coordinates between the UNB and the 2MASS point sources is about 0.3 and 0.2 arcsec for RA and Dec respectively. To understand the distribution of noise, and the accuracy of our sky subtraction in the UNB stacks, we constructed sky background maps using SExtractor. We found that except in the corners where the sky is affected by the OH rings, the sky subtraction is uniform throughout the image. 57

72 Figure 4.2: Postage stamps of our candidate Lyα galaxy at z = 7.7. The circles are centered on the candidate galaxy. Note an object, similar in NB flux and morphology, close to the candidate in the UNB1063, is clearly detected in all the filters. We now have deep stacks of broadband ugriz, and J-band obtained from CFHTLS and WIRDS surveys respectively, and two UNB stacks (1.056 and 1.063). All these images were geometrically aligned using IRAF tasks geomap and geotran. 4.4 Calibration and Catalog Generation Photometric Calibration The photometric calibration of UNB stacks was done by comparing unsaturated point sources, extracted using SExtractor, with 2MASS stars. To minimize the photometric errors, only those stars in the 2MASS catalog with J-magnitude between 13.8 & 16.8 AB mag 6, and errors less than 0.1 magnitude were selected for the calibration. The zeropoint was then determined by minimizing the difference between UNB and J 2M magnitudes. We used about 30 point sources which left a residual rms of about 0.12 mag between UNB and J 2M. Photometric Uncertainty There is an additional uncertainty in the photometric zeropoint due to differing central wavelengths between the 2MASS and UNB filters. In order to estimate this uncertainty we constructed spectral energy distribution (SED) of stars using Pickles spectra for different types of stars, and measured the expected difference between the magnitudes at the central wavelength of the J band and UNB filters. We found 6 Since J 2M magnitudes are in Vega, we adopted the following conversion between Vega and AB magnitudes : J AB=J 2M mag 58