Demographics of the Cold Universe with ALMA: From Interstellar and Circumgalactic Media to Cosmic Structures (Springer Theses) 9789811649783, 9789811649790, 9811649782

Table of contents :
Supervisor’s Foreword
List of Publication
Acknowledgements
Contents
1 Introduction
1.1 Hidden, Cold Side of the Universe
1.2 Hierarchical Structure of the Universe
1.2.1 Inter-Stellar Medium Scale
1.2.2 Circum-galactic Medium Scale
1.2.3 Cosmic Structure Scale
1.3 Cosmic Star-Formation Rate Density
1.4 Scope of This Thesis
References
2 Data and Reduction
2.1 Our Dataset
2.1.1 ALMA-DUST
2.1.2 ASAGAO
2.1.3 ALMA-z6CII
2.1.4 ALMA-FAINT
2.2 Reduction
2.2.1 Flag and Calibration
2.2.2 Imaging
References
3 Interstellar Medium Scale I: Galaxy Size
3.1 Data Analysis
3.1.1 Source Detection
3.1.2 Flux and Size Measurement
3.1.3 Simulation and Correction
3.1.4 Comparison with Previous Measurements
3.1.5 Selection and Measurement Completeness
3.1.6 Optical and NIR Counterpart
3.1.7 Our Sample on the SFR-Mstar Plane
3.2 Result
3.2.1 Redshift Distribution
3.2.2 FIR Size and Luminosity Relation
3.2.3 Slope of FIR Size and Luminosity Relation
3.2.4 Spatial Offset Between FIR and UV-Optical Emission
3.2.5 Sizes in UV, Optical, and FIR
References
4 Insterstellar Medium Scale II: Galaxy Morphology
4.1 Data Analysis
4.1.1 Source Detection
4.1.2 Flux Density and Position Measurement
4.1.3 Multi-wavelength Properties of Our Sample
4.1.4 Visibility-Based Stacking
4.1.5 nFIR and Re,FIR Measurements
4.1.6 Additional Sample from Archive
4.2 Result
4.2.1 Size and Morphology in FIR
4.2.2 Size and Morphology in Optical
4.2.3 Comparison Between FIR and Optical
References
5 Circumgalactic Medium Scale: Metal-Enriched Gas Halo
5.1 Data Analysis
5.1.1 3D Position in ALMA Cube
5.1.2 ALMA Visibility-Based Stacking
5.1.3 HST/H-Band Stacking
5.2 Result
5.2.1 Discovery of [CII] Halo
5.2.2 Effect of [CII]–UV Offset
5.2.3 Radial Ratio of L[CII] to Total SFR
5.2.4 Scale Length of [CII] Halo
5.2.5 [CII] Spectrum Stacking
5.2.6 Comparison with Model
References
6 Cosmic Structure Scale: Number Density and Clustering
6.1 Data Analysis
6.1.1 Source Detection
6.1.2 Spurious Sources
6.1.3 Completeness and Flux Boosting
6.1.4 Flux Measurement
6.1.5 Mass Model
6.1.6 Survey Area
6.2 Result
6.2.1 Number Counts at 1.2 mm
6.2.2 Comparison with Previous Number Count Measurements
6.2.3 Contribution to the CIB
6.2.4 Galaxy Bias
6.2.5 Multi-wavelength Properties of the Faint ALMA Sources
References
7 Discussion
7.1 Insterstellar Medium Scale
7.1.1 Do Mergers Trigger Dusty Starbursts?
7.1.2 Size and Stellar Mass Relation
7.1.3 Evolutions of Size and Morphology
7.1.4 Compact Dusty Bulge Versus AGN
7.2 Circumgalactic Medium Scale
7.2.1 Possible Origins of [Cii] Halo
7.2.2 From Observational and Theoretical Results
7.2.3 What Made the Primordial CGM Metal-Enriched?
7.3 Cosmic Structure Scale
7.3.1 Comparison with the IR Luminosity Function 0leqz leq4
7.3.2 Comparison with the Stellar Mass Function 0leqz leq8
7.4 Cosmic Star-Formation Rate Density
References
8 Conclusion
Appendix Curriculum Vitae
Research Interests
Work Experience
Education
Awards/Prizes
Awarded Telescope Times (S. Fujimoto as a PI)
Outreach
International Conferences in the Past 5years

Citation preview

Springer Theses Recognizing Outstanding Ph.D. Research

Seiji Fujimoto

Demographics of the Cold Universe with ALMA From Interstellar and Circumgalactic Media to Cosmic Structures

Springer Theses Recognizing Outstanding Ph.D. Research

Aims and Scope The series “Springer Theses” brings together a selection of the very best Ph.D. theses from around the world and across the physical sciences. Nominated and endorsed by two recognized specialists, each published volume has been selected for its scientific excellence and the high impact of its contents for the pertinent field of research. For greater accessibility to non-specialists, the published versions include an extended introduction, as well as a foreword by the student’s supervisor explaining the special relevance of the work for the field. As a whole, the series will provide a valuable resource both for newcomers to the research fields described, and for other scientists seeking detailed background information on special questions. Finally, it provides an accredited documentation of the valuable contributions made by today’s younger generation of scientists.

Theses may be nominated for publication in this series by heads of department at internationally leading universities or institutes and should fulfill all of the following criteria • They must be written in good English. • The topic should fall within the confines of Chemistry, Physics, Earth Sciences, Engineering and related interdisciplinary fields such as Materials, Nanoscience, Chemical Engineering, Complex Systems and Biophysics. • The work reported in the thesis must represent a significant scientific advance. • If the thesis includes previously published material, permission to reproduce this must be gained from the respective copyright holder (a maximum 30% of the thesis should be a verbatim reproduction from the author’s previous publications). • They must have been examined and passed during the 12 months prior to nomination. • Each thesis should include a foreword by the supervisor outlining the significance of its content. • The theses should have a clearly defined structure including an introduction accessible to new PhD students and scientists not expert in the relevant field. Indexed by zbMATH.

More information about this series at https://link.springer.com/bookseries/8790

Seiji Fujimoto

Demographics of the Cold Universe with ALMA From Interstellar and Circumgalactic Media to Cosmic Structures Doctoral Thesis accepted by University of Tokyo, Tokyo, Japan

Author Seiji Fujimoto Cosmic DAWN Center Copenhagen, Denmark

Supervisor Prof. Masami Ouchi Department of Astronomy University of Tokyo Tokyo, Japan

ISSN 2190-5053 ISSN 2190-5061 (electronic) Springer Theses ISBN 978-981-16-4978-3 ISBN 978-981-16-4979-0 (eBook) https://doi.org/10.1007/978-981-16-4979-0 © Springer Nature Singapore Pte Ltd. 2021 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore

Supervisor’s Foreword

In the sixteenth century, Thomas Digges recognized the strange fact that the night sky was dark. Now it is known as the Olbers’ paradox, an argument that the darkness of the night sky is contradictory to the infinite and eternal static universe. After recognitions of the importance of the night sky brightness, referred to as the background light, in the cosmological context, the background light was detected in γ -ray to radio wavelengths in the second half of the twentieth century. Among the background light, the Cosmic Infrared Background (CIB) was accurately measured by Cosmic Background Explorer (COBE), and revealed that the total energy of CIB emission was as high as one of the cosmic optical (stellar) background lights. The major origin of CIB was unclear. In the 1980s, the Infrared Astronomy Satellite (IRAS) has identified a large population of local dusty starbursts whose interstellar cold dust, heated by massive stars, emits infrared-continuum radiation that contributes to the CIB. Subsequently, submillimeter observations have detected high-redshift dusty starbursts of high star-formation rates (SFRs : 1000 Me yr-1 ), a.k.a. submillimeter galaxies, with very bright (1012–14 L e ) infrared emission that is redshifted to the submillimeter band in the observed frame. While the sources of infrared emission were found in the local and high-redshift universe, only a fraction of the CIB were resolved, due to the limited sensitivities and statistics. Such observations also revealed rest-frame infrared continua and emission lines in starburst galaxies, especially [C II] 158μm, radiated from cold gas of the interstellar medium (ISM) in the local and high-redshift universe. The [C II] emission is one of the strongest emission lines of galaxies in the rest-frame infrared band, where the [C II] luminosity is as bright as 1% of the total infrared luminosity in galaxies. Although the bright [C II] line and the continuum are good probes for spatial distributions of the cold ISM, the spatial distributions of [CII] and continuum were known only in luminous infrared galaxies in the local universe until recently. The following Ph.D. thesis written by Seiji Fujimoto addresses these important issues, exploiting observational data of the state-of-the-art radio interferometer, the Atacama Large Millimeter/submillimeter Array (ALMA). Although the sensitivity of ALMA is extremely higher than those of the other existing radio facilities, the small area and frequency coverages of ALMA do not provide a statistical data set v

vi

Supervisor’s Foreword

even in large observational programs approved by the early ALMA time allocation committees. Fujimoto’s thesis uses ALMA archival data, in addition to his original data, to make the largest ALMA samples ever obtained, pioneering statistical studies of cold dust and gas in high-redshift star-forming galaxies (their SFRs down to a few Me yr-1 ) with the ALMA archive. On the basis of comprehensive analyses of the ALMA data, Fujimoto presents three major conclusions on the scales over the galaxy ISM to the cosmic structures that are summarized below in an order of scientific relevance. First, dust-continuum radiation is confined in a compact region of star-forming galaxies with a disk-like morphology of Sérsic index n ∼ 1. Although the dusty star-formation region is compact, the spatial distributions of star formation traced by ultraviolet radiation of massive stars and infrared radiation of heated dust suggest that a stellar profile of elliptical galaxy with n : 4 cannot be produced by the star-forming activities, but dynamical processes taken place after the star formation. Second, the sky surface density of dust-continuum sources of star-forming galaxies resolved by this thesis is high enough to explain the CIB. In other words, the major origin of the CIB is not dust in the inter-galactic space, but in the compact regions of star-forming galaxies. Third, the last and the most surprising result of this thesis is the first identification of a 10-kpc-sized [C II] emitting halo around high-redshift star-forming galaxies. Because carbon is not produced by the Big Bang nucleosynthesis, carbon ions emitting the [C II] line are probably remnants of early gas outflow from the high-redshift galaxies. No galaxy-formation models have so far predicted such a large halo of cold gas emitting [C II] in the circumgalactic medium. Fujimoto claims that the galaxyformation models, to date, miss important processes such as strong feedback making a significant cold gas outflow. The recognition of this thesis work led to Fujimoto taking a prestigious research position, DAWN Fellowship, in Copenhagen, Denmark, that allows him to conduct independent researches for innovative results. He is now extending his ALMA studies toward two directions as his new challenges, high-redshift galaxy halos and evolution of supermassive black holes, exploiting his experiences in the ALMA archival statistical studies presented in this thesis. This thesis is the first set of his innovative studies, and worth reading for those who aim to investigate the cold side of the universe and to carry out statistical work with big data in radio wavelengths. Tokyo, Japan February 2020

Masami Ouchi

List of Publication

Parts of this thesis have been published in the following journal articles: • S. Fujimoto, M. Ouchi, A. Ferarra, A. Pallottini, R. J. Ivison, C. Behrens, S. Gallerani, S. Arata, H. Yajima, and K. Nagamine “First Identification of 10-kpc Scale [C II] Halo around Star-Forming Galaxies at z = 5 − 7”, 2019, ApJ, 887, 107. • S. Fujimoto, M. Ouchi, K. Kohno, Y. Yamaguchi, B. Hatsukade, Y. Ueda, T. Shibuya, S. Inoue, T. Oogi, S. Toft, C. Gomez-Guijarro, T. Wang, T. Nagao, I. Tanaka, Y. Ao, H. Umehata, Y. Taniguchi, K. Nakanishi, W. Rujopakarn, R. J. Ivison, W. Wang, M. Lee, K. Tadaki, Y. Tamura, and J. S. Dunlop “ALMA 26 arcmin2 Survey of GOODS-S at One-millimeter (ASAGAO): Average Morphology of High-z Dusty Star-Forming Galaxies is an Exponential-Disk (n; 1)”, 2018, ApJ, 861, 7. • S. Fujimoto, M. Ouchi, T. Shibuya, H. Nagai, “Demonstrating a New Census of Infrared Galaxies with ALMA (DANCING—ALMA) I. FIR Size and Luminosity Relation at z = 0 − 6 Revealed with 1034 ALMA Sources”, 2017, ApJ, 850, 83. • S. Fujimoto, M. Ouchi, Y. Ono, T Shibuya, M. Ishigaki, H. Nagai, R. Momose “ALMA Census of Faint 1.2 mm Sources down to :0.02 mJy: Extragalactic Background Light and Dust-Poor High-z Galaxies”, 2016, ApJS, 222, 1.

vii

Acknowledgements

Special thanks to my thesis supervisor Masami Ouchi for his invaluable advice, encouragement, and support during the period of my Ph.D. I also deeply thank Ph.D thesis committee members: Satoru Iguchi, Ryohei Kawabe (Chair), Kentaro Motohara, Toshikazu Shigeyama, and Hidenobu Yajima for their constructive comments and suggestion. I am grateful to Kotaro Kohno and Yoichi Tamura for leading me to an exciting field of radio astronomy. I appreciate the group members including the previous members: Yoshiaki Ono, Takatoshi Shibuya, Jun Toshikawa, Ken Mawatari, Rieko Momose, Mariko Kubo, Akira Konno, Yoshiaki Naito, Masafumi Ishigaki, Hiroko Tamazawa, Yuichi Harikane, Takashi Kojima, Shiro Mukae, Yuma Sugahara, Ryo Higuchi, Haibin Zhang, Miftaful Hilmi, Shotaro Kikuchihara, Ryohei Itoh, Ryota Kakuma, Karin Shimodate, Nao Sakai, Yuki Isobe, and Takako Idomura for daily discussions, chats, taking lunch/dinner, playing tennis, and going to the hot spring that motivate me to enjoy the daily research life in Kashiwa. I thank Yuki Yamaguchi, Haruka Kusakabe, and Taichi Uyama for studying and encouraging together from undergraduate in the Department of Astronomy. I would like to express my gratitude to Hiroshi Nagai, Bunyo Hatsukade, Yuichi Matsuda, Tadayuki Kodama, Soh Ikarashi, Hideki Umehata, Izumi Takuma, Koichiro Nakanishi, Daniel Espada, Kazuya Saigo, Daisuke Iono, Fumi Egusa, Junko Ueda, Toshiki Saito, Akifumi Seko, Akio Taniguchi, and Minju Lee for giving me helpful supports, advice, and fruitful discussions as the experts of the extragalactic studies with the submillimeter/millimeter facilities including ALMA. I would also appreciate the helpful feedbacks and advice offered by Tohru Nagao, Akio Inoue, Koji Ohta, Yoshihiro Ueda, Hidenobu Yajima, Masayuki Umemura, Yoshiaki Taniguchi, Kenichi Tadaki, Takuya Hashimoto, Ryota Kawamata, Masafumi Onoue, Kyoko Onishi, and Ryosuke Goto. I am indebted to the Italian colleagues of Andrea Ferrara, Andrea Pallottini, Christoph Behrens, Simona Gallerani, and Graziano Ucci for supporting my research stay in Pisa, as well as other foreign colleagues of Sune Toft, Olivier Le Fèvre, David Elbaz, Raffaella Schneider, J. S. Dunlop, R. J. Ivison, Nick Scoville, Ian Smail, Dan Coe, and Fabian Walter for hosting me in the seminars. I would also thank all the collaborators all over the world including Tao Wang, Ichi Tanaka, Yiping Ao, Wiphu ix

x

Acknowledgements

Rujopakarn, Wei-hao Wang, Walter Gear, Carlos Góme-Guijarro, Darko Donevski, Scott Chapman, James Simpson, Clements Dave, Marcin Sawicki, Chien-Hsiu Lee, Roland Bacon, Jeremy Blaizot, and Floriane Leclercq. I appreciate the support of the staff at the ALMA Regional Center. This thesis makes use of the following ALMA data: ADS/JAO.ALMA #2011.0.00115.S, #2011.0.00232.S, #2011.0.00319.S, #2011.0.00648.S, #2011.0.00767.S, #2012.1.00076.S, #2012.1.00602.S, #2012.1.00719.S, #2012.1.00323.S, #2012.1.00536.S, #2012.1.00610.S, #2012.1.00934.S, #2012.1.00953.S, #2011.0.00097.S, #2011.0.00294.S, #2012.1.00076.S, #2012.1.00090.S, #2012.1.00245.S, #2012.1.00307.S, #2012.1.00326.S, #2012.1.00523.S, #2012.1.00756.S, #2012.1.00775.S, #2012.1.00869.S, #2012.1.00979.S, #2012.1.00983.S, #2013.1.00034.S, #2013.1.00118.S, #2013.1.00151.S, #2013.1.00173.S, #2013.1.00718.S, #2013.1.00884.S, #2013.1.00205.S, #2013.1.00208.S, #2013.1.00566.S, #2013.1.00668.S, #2013.1.00781.S, #2013.1.00999.S, #2013.1.01292.S, #2015.1.00137.S, #2015.1.00540.S, #2015.1.00664.S, #2015.1.01495.S, #2015.1.00098.S #2013.1.00815.S, #2015.1.00834.S, #2015.1.01111.S, #2015.1.01105.S, #2016.1.01240.S, #2012.1.00523.S, and #2012.1.00602.S. ALMA is a partnership of ESO (representing its member states), NSF (USA) and NINS (Japan), together with NRC (Canada) and NSC and ASIAA (Taiwan), in cooperation with the Republic of Chile. The Joint ALMA Observatory is operated by ESO, AUI/NRAO, and NAOJ. This work was supported by the World Premier International Research Center Initiative, MEXT, Japan, and KAKENHI 16J02344, (A) 23244025, (A) 15H02064, (A) 17H01110, (A) 17H01111, (A) 17H01114, and (S) 17H06130 Grant-in-Aid for Scientific Research through the Japan Society for the Promotion of Science, Graduate Research Abroad in Science Program 2017, Hayakawa Satio Fund awarded by the Astronomical Society of Japan (cycle 89, 95, 106), and ALMA Japan research Grant of NAOJ Chile Observatory (ID145, 164, 179, 197).

Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 Hidden, Cold Side of the Universe . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Hierarchical Structure of the Universe . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.1 Inter-Stellar Medium Scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.2 Circum-galactic Medium Scale . . . . . . . . . . . . . . . . . . . . . . . . 1.2.3 Cosmic Structure Scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Cosmic Star-Formation Rate Density . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4 Scope of This Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1 1 1 3 4 6 8 9 9

2 Data and Reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1 Our Dataset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.1 ALMA-DUST . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.2 ASAGAO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.3 ALMA-z6CII . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.4 ALMA-FAINT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.1 Flag and Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.2 Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

11 11 11 15 15 18 19 19 21 25

3 Interstellar Medium Scale I: Galaxy Size . . . . . . . . . . . . . . . . . . . . . . . . . 3.1 Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.1 Source Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.2 Flux and Size Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.3 Simulation and Correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.4 Comparison with Previous Measurements . . . . . . . . . . . . . . . 3.1.5 Selection and Measurement Completeness . . . . . . . . . . . . . . . 3.1.6 Optical and NIR Counterpart . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.7 Our Sample on the SFR−Mstar Plane . . . . . . . . . . . . . . . . . . . . 3.2 Result . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.1 Redshift Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

27 27 27 28 29 31 32 34 35 37 37 xi

xii

Contents

3.2.2 3.2.3 3.2.4 3.2.5 References

FIR Size and Luminosity Relation . . . . . . . . . . . . . . . . . . . . . . Slope of FIR Size and Luminosity Relation . . . . . . . . . . . . . . Spatial Offset Between FIR and UV-Optical Emission . . . . . Sizes in UV, Optical, and FIR . . . . . . . . . . . . . . . . . . . . . . . . . . .....................................................

38 45 45 47 48

4 Insterstellar Medium Scale II: Galaxy Morphology . . . . . . . . . . . . . . . . 4.1 Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.1 Source Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.2 Flux Density and Position Measurement . . . . . . . . . . . . . . . . . 4.1.3 Multi-wavelength Properties of Our Sample . . . . . . . . . . . . . . 4.1.4 Visibility-Based Stacking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.5 n FIR and Re,FIR Measurements . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.6 Additional Sample from Archive . . . . . . . . . . . . . . . . . . . . . . . 4.2 Result . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1 Size and Morphology in FIR . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.2 Size and Morphology in Optical . . . . . . . . . . . . . . . . . . . . . . . . 4.2.3 Comparison Between FIR and Optical . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

51 51 51 52 53 54 55 58 59 59 60 61 62

5 Circumgalactic Medium Scale: Metal-Enriched Gas Halo . . . . . . . . . . 5.1 Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1.1 3D Position in ALMA Cube . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1.2 ALMA Visibility-Based Stacking . . . . . . . . . . . . . . . . . . . . . . 5.1.3 HST/H-Band Stacking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Result . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.1 Discovery of [CII] Halo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.2 Effect of [CII]–UV Offset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.3 Radial Ratio of L [C I I ] to Total SFR . . . . . . . . . . . . . . . . . . . . . 5.2.4 Scale Length of [CII] Halo . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.5 [CII] Spectrum Stacking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.6 Comparison with Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

63 63 63 64 68 70 70 71 71 74 76 77 79

6 Cosmic Structure Scale: Number Density and Clustering . . . . . . . . . . 6.1 Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1.1 Source Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1.2 Spurious Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1.3 Completeness and Flux Boosting . . . . . . . . . . . . . . . . . . . . . . . 6.1.4 Flux Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1.5 Mass Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1.6 Survey Area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 Result . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2.1 Number Counts at 1.2 mm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2.2 Comparison with Previous Number Count Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

81 81 81 82 84 85 86 87 88 88 90

Contents

xiii

6.2.3 Contribution to the CIB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 6.2.4 Galaxy Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 6.2.5 Multi-wavelength Properties of the Faint ALMA Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 7 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.1 Insterstellar Medium Scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.1.1 Do Mergers Trigger Dusty Starbursts? . . . . . . . . . . . . . . . . . . 7.1.2 Size and Stellar Mass Relation . . . . . . . . . . . . . . . . . . . . . . . . . 7.1.3 Evolutions of Size and Morphology . . . . . . . . . . . . . . . . . . . . . 7.1.4 Compact Dusty Bulge Versus AGN . . . . . . . . . . . . . . . . . . . . . 7.2 Circumgalactic Medium Scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.2.1 Possible Origins of [Cii] Halo . . . . . . . . . . . . . . . . . . . . . . . . . 7.2.2 From Observational and Theoretical Results . . . . . . . . . . . . . 7.2.3 What Made the Primordial CGM Metal-Enriched? . . . . . . . . 7.3 Cosmic Structure Scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.3.1 Comparison with the IR Luminosity Function 0 ≤ z ≤ 4 ........................................... 7.3.2 Comparison with the Stellar Mass Function 0 ≤ z ≤ 8 . . . . 7.4 Cosmic Star-Formation Rate Density . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

119 119 119 124 125 127 128 129 131 132 133 133 135 138 140

8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143 Curriculum Vitae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151

Chapter 1

Introduction

1.1 Hidden, Cold Side of the Universe There are three major energies coming from the universe as the cosmic background light (Fig. 1.1): cosmic optical (COB), infrared (CIB), and microwave background light (CMB). The COB and CMB are explained by the stellar continuum of galaxies and the leftover radiation from the Big Bang, respectively. On the other hand, the origin of the CIB has not been fully understood since the CIB was first identified by the Cosmic Background Explorer (COBE ) satellite [1–5]. The universe reserves a large amount of the cold gas that forms stars. These stars produce the dust when they die via the supernova explosions, and the dust soon becomes cold through widely distributed in the cold gas. The ultra-violet (UV) emission originated from the stars is absorbed by the cold dust and re-emitted as the thermal infrared emission. Because the CIB is thought to be contributed by the cold dust emission, the CIB shows another side of the universe obscured by the cold dust (i.e., Cold Universe). It is thus essential to unveil the Cold Universe: where and how much amount of the cold dust emission is originated in the universe. Furthermore, it is also required to reveal the physical properties of the individual cold dust emission to understand what kind of roles the dusty obscured star-forming activity plays in the galaxy formation and evolution.

1.2 Hierarchical Structure of the Universe Where does the cold dust emission come from ? Galaxies are thought to be formed in their dark matter halos in the frameworks of Λ cold dark matter (ΛCDM) structure formation models. These galaxies are evolved in the interplay with the surrounding medium via gas inflow [e.g., 6] and outflow [e.g., 7] that regulate star-formation by gas-supply and feedback processes [8]. Figure 1.2 presents the numerical simulation results for the gas density in a 100×100×20 comoving Mpc box. In Fig. 1.2, while the © Springer Nature Singapore Pte Ltd. 2021 S. Fujimoto, Demographics of the Cold Universe with ALMA, Springer Theses, https://doi.org/10.1007/978-981-16-4979-0_1

1

2

1 Introduction

Fig. 1.1 Schematic Spectral Energy Distributions of the most important (by intensity) backgrounds in the universe, and their approximate brightness in nW m−2 sr−1 written in the boxes (Credit: Dole et al., A&A, 451, 417, 2006, reproduced with permission ©ESO.)

Fig. 1.2 A 100×100×20 comoving Mpc (cMpc) slice through the EAGLE simulation, illustrating the dynamic range attainable with state-of-the-art SPH simulations (Fig. 1 of [9], MNRAS, 446, 521, reproduced by permission of Oxford University Press). The intensity represents the gas density while the color indicates the gas temperatures (blue=green-red from cooler to hotter). The inset shows a region 10 cMpc and 60 ckpc on a side. The zoom in to an individual galaxy with stellar mass 3×1010 M shows the optical band stellar light

1.2 Hierarchical Structure of the Universe

3

galaxy is formed in the local gas density peak with the interstellar medium (ISM), the large amount of the gas is still widely distributed in the scales of the circumgalactic medium (CGM) and the cosmic structure. Since the cold gas and dust generally coexist through the interplay, the cold dust emission is expected to be originated in the scales of the CGM and the cosmic structure as well as the ISM. To comprehensively understand the obscured side of the galaxy formation and evolution in this hierarchical structure of the Cold Universe, it is thus important to fully resolve the cold dust emission in the ISM, the CGM, and the cosmic structure instead of focusing on a part of them.

1.2.1 Inter-Stellar Medium Scale Size and morphology of high-redshift galaxies are the key to understand the galaxy formation and evolution. In particular, the effective radius Re and the S´ersic index n [10, 11] are key quantities to evaluate the size and morphological properties. In the rest-frame ultra-violet (UV) and optical wavelength regimes, the Hubble Space Telescope (HST) has revealed the Re and n properties for the high-redshift galaxies up to z ∼ 6 − 10 [e.g., 12–21]. These HST studies show that quiescent and star-forming galaxies are represented by a spheroidal morphology with a S´ersic index in the rest-frame optical wavelength n opt ∼ 4 and an exponential-disk morphology with n opt ∼ 1, respectively. Moreover, the quiescent galaxies are more compact than the star-forming galaxies at a given stellar mass. These HST results indicate that the transformation in both size and morphology might be attributed to the evolution from star-forming to quiescent galaxy (Fig. 1.3). However, the actively star-forming regions obscured by dust cannot be unveiled with the rest-frame UV and optical studies. The stars formed in the dusty starburst may eventually dominate over the stars already present in the host galaxies due to an extreme star formation rates in these dusty star-forming galaxies. Studies of the size and morphology in the rest-frame far-infrared (FIR) wavelength are thus essential to comprehensively understand the evolutionary connections from the high-z star-forming to compact quiescent galaxies. Recent ALMA observations enable us to measure n and Re in the rest-frame FIR wavelength, n FIR and Re,FIR , due to its high sensitivity and angular resolution. For bright submillimeter galaxies (SMGs; S1mm  1 mJy), recent ALMA studies report that SMGs have the exponential-disk morphology with n FIR = 0.9 ± 0.2 [22], and Re,FIR of ∼1 − 2 kpc [e.g., 22–25] smaller than Re in the rest-frame optical wavelength Re,opt of ∼3 − 4 kpc [e.g., 25–28]. For faint SMGs (S1mm < 1 mJy), although there are several attempts to estimate Re,FIR by using deep ALMA imaging [25, 29, 30], large uncertainties still remain due to the small number statistics and observational challenges. There are two major observational challenges to measure n FIR and Re,FIR for faint SMGs. First is the sensitivity. To perform a secure profile fitting, high signal-to-noise (S/N) level is required. Extremely deep observations are thus needed. Second is the angular resolution. The rest-frame UV and optical studies report a positive correlation

4

1 Introduction

Fig. 1.3 Size-stellar mass distribution of late- (blue) and early-type (red) galaxies at 1.5 < z < 2 (Fig. 5 of [17], reproduced by permission of the AAS). The lines indicate model fits to the early- and late-type galaxies. The dashed lines represent the model fits to the galaxies at redshifts 0 < z < 0.5. The solid lines represent fits to the higher-redshift samples. The mass ranges used in the fits are indicated by the extent of the lines in the horizontal direction

between the galaxy size and luminosity [e.g., 17, 18], which suggests that faint SMGs are more compact than bright SMGs. To resolve the compact objects, we need high angular resolution.

1.2.2 Circum-galactic Medium Scale In the early universe, star-forming galaxies are known to be surrounded by the CGM-scale neutral hydrogen, which emits in the Lyα line, giving rise to Lyα haloes [Fig. 1.4; e.g., 31, 32]. After they form, from the primordial (H+He) gas, the first stellar populations in the earliest galaxies produce heavier elements. Theoretical models predict these are expelled into the CGM via powerful outflows driven by active galactic nuclei and/or supernova explosions [33, 34]. The metal-enrichment in the primordial CGM is thus a powerful probe for the outflow activities in the early galaxies. It is known that rare, massive galaxies—formed via mergers [35] and possibly cold-stream accretion [36], with an abundance of massive stars [37]—drive outflows throughout cosmic time, forming the metal-enriched CGM [38–41]. However, there has hitherto been no evidence that gas is expelled from primaeval galaxies with more modest star-formation rates (SFRs), which dominate the earliest period of cosmic star-formation history. Any expelled gas would be much more diffuse than the interstellar medium and there is thought to be insufficient energy available to drive outflows at low SFRs [42, 43].

1.2 Hierarchical Structure of the Universe

5

Fig. 1.4 (Left) Composite rest-frame continuum (top) and Lyα images of the z = 5.7 Lymanalpha emitters (LAEs) produced by the mean-combined method. (Right) Radial surface brightness profiles of composite images of rest-frame UV continuum (top) and Lyα (bottom) from the z = 5.7 LAEs (solid lines) and PSFs (dotted lines). These figures are reproduced from Fig. 2 and 3 of [31], MNRAS, 442, 110, by permission of Oxford University Press on behalf of the sponsoring Monthly Notices of the Royal Astronomical Society

ALMA has opened our views to the metal-enriched medium in the early universe, providing the secure detections such as the dust continuum and the [Cii]-158 µm line emission even up to z ∼ 7 [e.g., 44–52]. There have also been several attempts to measure the size and morphology in the rest-frame FIR continuum and the [Cii] line for such high redshift galaxies at z ∼ 5−7, where [53] report that the effective radius of the [Cii] line-emitting region is larger than that of the rest-frame UV region. However, large uncertainties still remain due to the small number statistics and observational challenges. One critical challenge is the flux recovery for the extended emission with ALMA. In the interferometric observations, it is required to sample enough coverage of the uv-visibility, especially in the short baselines, to recover the flux of the extended emission, while the general ALMA [C ii] observations for the high-z star-forming galaxies are carried out under the assumption that the [C ii]-emitting regions are distributed in the ISM scale. If the [C ii] line emission extends over the ISM scale, such extended [C ii] emission might be missed in the previous ALMA observations. Another critical challenge is sensitivity. The recent ALMA studies show that signalto-noise ratio (S/N) > 10 is needed to obtain reliable size measurement results both on the image-based and visibility-based analyses [e.g., 23, 24], while the majority

6

1 Introduction

of the previous ALMA detections of the dust continuum and the [Cii] line from z ∼ 5−7 galaxies show the S/N less than 10. If the S/N level is poor, noise fluctuations significantly affect the profile fitting results. Moreover, [22] show that the combination of the original smoothed galaxy profile and the noise fluctuations can make the morphology more clumpy. To obtain the reliable size and morphological results without the missing flux, extensively deep and good uv-sampling observations are thus required.

1.2.3 Cosmic Structure Scale In the cosmic structure scale, the origin of the CIB is one of the biggest questions. To reveal the origin of the CIB, first we need to detect individual galaxies and evaluate their contribution to the CIB. The most advantageous regimes are the submm/mm wavelengths due to the negative k-correction. In previous blank field observations with single-dish telescopes, it has revealed that the bright dusty high-z objects of submm galaxies [SMGs, S1mm  1 mJy; 54] are not major contributors of the CIB. SCUBA and LABOCA observations have shown that the 20–40% of CIB is resolved at submm wavelength [e.g., 55–59]. Similar results have been obtained by AzTEC observations that they have resolved 10–20% of the CIB at mm wavelength [e.g., 60– 62]. These observations suggest that there should exist high-z populations different from SMGs, and that such populations are major CIB contributors. The contribution from the faint submm/mm sources can be now addressed by ALMA, owing to its capabilities of high angular resolution and sensitivity. In other words, recent ALMA observations allow us to study the faint submm/mm sources with the negligibly small uncertainties of the source confusion and blending which The first ALMA mm search for faint sources below S1mm ∼ 1.0 mJy with no lensing effect has been conducted by [63]. The authors resolve ∼80% of the CIB down to S1mm = 0.1 mJy in 20 targets residing in one blank field of the Subaru/XMMNewton Deep Survey [SXDS; 64]. However, this result would include unknown effects of cosmic variance, because of the single field observations. Subsequently, [65, 66] have claimed the CIB resolved a fraction of ∼60% that is smaller than the one of [63], utilizing multi-field deep ALMA 1mm maps of Bands 6 and 7 down to the flux limit of S1mm = 0.1 mJy, where the uncertainties from cosmic variance are probably reduced with the multi-field data. As a brief summary of these ALMA studies, only about a half of the CIB has been resolved with the present ALMA flux limit of S1mm = 0.1 mJy (Fig. 1.5), including the effects of the cosmic variance. No one still knows whether the remaining half of the CIB is resolved by the very faint (S1mm < 0.1 mJy), individual submm/mm sources, or by the faint diffuse emission, extended over the CGM and cosmic structure scales, that has been missed in the previous studies. There is another remaining issue about the origin of the CIB. Although these studies newly identify faint ALMA sources (S1mm < 1.0 mJy) and their contribution to the CIB nearly the half, these studies do not clearly answer to the question about the

1.2 Hierarchical Structure of the Universe

7

Fig. 1.5 Resolved fraction of the CIB with previous ALMA observations against the flux limit (Slimit ) at 1.2 mm. The magenta, blue, and black lines present the results of H13 [63], O14 [65], and Ca15 [66], respectively. These lines are obtained by re-fitting the Schechter functions to the data points of each literature and [62] for the bright end. The red line shows the CIB value estimated from the COBE observations [2]

connection between the faint ALMA sources and optically selected high-z galaxies. Recent observations with Herschel have revealed that typical UV-selected galaxies such as Lyman-break galaxies (LBGs) have a median total (8–1000 µm) luminosity of L IR  2.2 × 1011 L  [67–69]. The stacking analysis of Herschel and ALMA data has also shown that K -selected galaxies, including star-forming BzK galaxies (sBzKs), have total IR luminosities of L IR = (5 − 11) × 1011 L  [70]. These ranges of the IR luminosities of LBGs and sBzKs correspond to the mm flux of S1mm ∼ 0.1 − 1 mJy if we assume a modified black body with βd = 1.8, dust temperature Td = 35K, and a source redshift z = 2.5. These sources could be mm counterparts of the optically selected galaxies, because the mm flux range shown in the stacking analyses is similar to that of the faint ALMA sources. There is another approach, focusing on the dark matter halo property, to characterize the faint ALMA sources. The clustering analysis is a powerful tool for understanding the connection between various galaxy populations, since the spatial distribution of galaxies is related to the underlying distribution of dark haloes in the standard scenario of structure formation in ΛCDM universe. Clustering analyses have been carried out over the past decade for bright SMGs observed with singledish telescopes, and concluded that the clustering amplitudes, or galaxy biases, are large, and that the hosting dark halo masses of the bright SMGs are estimated to be ∼1013 M ([59, 71–74], cf. [75]). On the other hand, however, the dark halo properties of the faint ALMA sources are poorly known, except for the result of [65] who have obtained, for the first time, a meaningful constraint on a galaxy bias, bg ≤ 4, for faint ALMA sources. In this way, the faint ALMA sources are not well studied. To understand the physical origin of faint ALMA sources, one should study faint ALMA sources with a more complete data set on the basis of individual sources as well as statistics.

8

1 Introduction

Fig. 1.6 Cosmic SFR density (ρSFR ) measurements in previous optical-radio observations. The blue symbols are the SFR density measurements from the optical observations (SFRDUV ) in [76–78], referred to as “Sc05”, “Re09”, and “Bo15”, respectively. The orange symbols are the SFR density values after dust correction (SFRDcorr ) from the effects of dust extinction using the observed UV slopes β [79] and the IRX–β relationship [80]. The blue and orange regions are the uncertainties of the SFRDUV and SFRDcorr values that are presented in Fig. 18 of [78]. The light-red region and green pentagon show the SFR density measurements from the direct IR observations (SFRIR ) in [81, 82], referred to as “Le05” and “Sm09”, respectively. The SFRIR measurements have been limited at z  1

1.3 Cosmic Star-Formation Rate Density As summarized in Sect. 1.2.1–1.2.3, no one still knows where and how much amount of the star-forming activity is obscured by dust in the universe. This indicates that the current understanding of the cosmic SFR density is incomplete. Figure 1.6 shows the current results of the cosmic SFR density measurements. The light red and blue shades are estimated from the rest-frame UV galaxies identified in the optical-NIR surveys with and without the dust correction, respectively, and the light red shade measurement is usually used to discuss the total (= dust obscured and un-obscured) cosmic SFR density. However, if the previous optical-near infrared (NIR) surveys have missed a number of the dusty obscured star-formation activities in the scales of the ISM, the CGM, and the cosmic structure, the SFR density in the previous studies underestimates the total value. The direct measurements for the obscured star-formation activity with the cold dust emission are thus required in all these there layers.

1.4 Scope of This Thesis

9

1.4 Scope of This Thesis In this paper, we analyze the large dataset of multi-field deep ALMA Band 6/7 data including our original and archival data open for public by 2018 April. We obtain 1220 ALMA sources extracted from 1643 independent field maps in our analyses. With this largest ALMA dataset so far obtained, we investigate the Cold Universe in the scales of the ISM, the CGM, and the cosmic structure to unveil the total cosmic SFR density. The structure of this paper is as follows. In Chap. 2, we describe the observations and data reduction. Chapter 3 shows the statistical size measurement results for the dust emission in the ISM scale. The detailed morphology in the ISM scale is analyzed with the high-resolution ALMA data in Chap. 4. In Chap. 5, we examine the emission from the dust and metal-enriched gas in the CGM scale. Chapter 6 outlines our statistical results of the number density and the clustering in the cosmic structure scale. In Chap. 7, we discuss the total cosmic SFR density based on our results obtained in the scales of the ISM, the CGM, and the cosmic structure. A conclusion composed of the summary is presented in Chap. 8. Throughout this paper, we assume a flat universe with Ωm = 0.3, ΩΛ = 0.7, σ8 = 0.8, and H0 = 70 km s−1 Mpc−1 . We use magnitudes in the AB system (Oke & Gunn 1983).

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24.

Puget J-L, Abergel A, Bernard J-P et al (1996) A&A 308:L5 Fixsen DJ, Dwek E, Mather JC, Bennett CL, Shafer RA (1998) ApJ 508:123 Hauser MG, Arendt RG, Kelsall T et al (1998) ApJ 508:25 Hauser MG, Dwek E (2001) ARA&A 39:249 Dole H, Lagache G, Puget J-L et al (2006) A&A 451:417 Dekel A, Birnboim Y, Engel G et al (2009) Nature 457:451 Muratov AL, Kereš D, Faucher-Giguère C-A et al (2015) MNRAS 454:2691 Lilly SJ, Carollo CM, Pipino A, Renzini A, Peng Y (2013) ApJ 772:119 Schaye J, Crain RA, Bower RG et al (2015) MNRAS 446:521 Sérsic JL (1963) Boletin de la Asociacion Argentina de Astronomia La Plata Argentina 6:41 Sérsic, JL (1968) Atlas de Galaxias Australes Shen S, Mo HJ, White SDM et al (2003) MNRAS 343:978 Ferguson HC, Dickinson M, Giavalisco M et al (2004) ApJ 600:L107 Hathi NP, Malhotra S, Rhoads JE (2008) ApJ 673:686 Oesch PA, Bouwens RJ, Carollo CM et al (2010) ApJ 725:L150 Ono Y, Ouchi M, Curtis-Lake E et al (2013) ApJ 777:155 van der Wel A, Franx M, van Dokkum PG et al (2014) ApJ 788:28 Shibuya T, Ouchi M, Harikane Y (2015) ApJS 219:15 Bowler RAA, Dunlop JS, McLure RJ, McLeod DJ (2017) MNRAS 466:3612 Kawamata R, Ishigaki M, Shimasaku K, Oguri M, Ouchi M (2015) ApJ 804:103 Kawamata R, Ishigaki M, Shimasaku K et al (2018) ApJ 855:4 Hodge JA, Swinbank AM, Simpson JM et al (2016) ApJ 833:103 Hodge JA, Swinbank AM, Simpson JM et al (2015) ApJ 810:133 Hodge JA, Swinbank AM, Simpson JM et al (2015) ApJ 799:81

10

1 Introduction

25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82.

Fujimoto S, Ouchi M, Shibuya T, Nagai H (2017) ApJ 850:1 Targett TA, Dunlop JS, McLure RJ et al (2011) MNRAS 412:295 Targett TA, Dunlop JS, Cirasuolo M et al (2013) MNRAS 432:2012 Chen C-C, Smail I, Swinbank AM et al (2015) ApJ 799:194 Rujopakarn W, Dunlop JS, Rieke GH et al (2016) ApJ 833:12 González-López J, Bauer FE, Romero-Cañizales C et al (2017) A&A 597:A41 Momose R, Ouchi M, Nakajima K et al (2014) MNRAS 442:110 Leclercq F, Bacon R, Wisotzki L et al (2017) A&A 608:A8 Sravan N, Faucher-Giguère C-A, van de Voort F et al (2016) MNRAS 463:120 Turner ML, Schaye J, Crain RA et al (2017) MNRAS 471:690 Neri R, Downes D, Cox P, Walter F (2014) A&A 562:A35 Narayanan D, Turk M, Feldmann R et al (2015) Nature 525:496 Zhang Z-Y, Ivison RJ, George RD et al (2018) MNRAS 481:59 Falgarone E, Zwaan MA, Godard B et al (2017) Nature 548:430 Ivison RJ, Papadopoulos PP, Smail I et al (2011) MNRAS 412:1913 George RD, Ivison RJ, Smail I et al (2014) MNRAS 442:1877 Maiolino R, Gallerani S, Neri R et al (2012) MNRAS 425:L66 Heckman TM, Alexandroff RM, Borthakur S, Overzier R, Leitherer C (2015) ApJ 809:147 Li M, Bryan GL, Ostriker JP (2017) ApJ 841:101 Watson D, Christensen L, Knudsen KK et al (2015) Nature 519:327 Maiolino R, Carniani S, Fontana A et al (2015) MNRAS 452:54 Capak PL, Carilli C, Jones G et al (2015) Nature 522:455 Pentericci L, Carniani S, Castellano M et al (2016) ApJ 829:L11 Knudsen KK, Richard J, Kneib J-P et al (2016) MNRAS 462:L6 Matthee J, Sobral D, Boone F et al (2017) ApJ 851:145 Carniani S, Maiolino R, Amorin R et al (2018) MNRAS 478:1170 Smit R, Bouwens RJ, Carniani S et al (2018) Nature 553:178 Bowler RAA, Bourne N, Dunlop JS, McLure RJ, McLeod DJ (2018) MNRAS 481:1631 Carniani S, Maiolino R, Pallottini A et al (2017) A&A 605:A42 Lagache G, Puget J-L, Dole H (2005) ARA&A 43:727 Eales S, Lilly S, Webb T et al (2000) AJ 120:2244 Smail I, Ivison RJ, Blain AW, Kneib J-P (2002) MNRAS 331:495 Coppin K, Chapin EL, Mortier AMJ et al (2006) MNRAS 372:1621 Knudsen KK, van der Werf PP, Kneib J-P (2008) MNRAS 384:1611 Weiß A, Kovács A, Coppin K et al (2009) ApJ 707:1201 Perera TA, Chapin EL, Austermann JE et al (2008) MNRAS 391:1227 Hatsukade B, Kohno K, Aretxaga I et al (2011) MNRAS 411:102 Scott KS, Wilson GW, Aretxaga I et al (2012) MNRAS 423:575 Hatsukade B, Ohta K, Seko A, Yabe K, Akiyama M (2013) ApJ 769:L27 Furusawa H, Kosugi G, Akiyama M et al (2008) ApJS 176:1 Ono Y, Ouchi M, Kurono Y, Momose R (2014) ApJ 795:5 Carniani S, Maiolino R, De Zotti G et al (2015) A&A 584:A78 Reddy N, Dickinson M, Elbaz D et al (2012) ApJ 744:154 Lee K-S, Alberts S, Atlee D et al (2012) ApJ 758:L31 Davies LJM, Bremer MN, Stanway ER, Lehnert MD (2013) MNRAS 433:2588 Decarli R, Smail I, Walter F et al (2014) ApJ 780:115 Webb TM, Eales S, Foucaud S et al (2003) ApJ 582:6 Blain AW, Chapman SC, Smail I, Ivison R (2004) ApJ 611:725 Williams CC, Giavalisco M, Porciani C et al (2011) ApJ 733:92 Hickox RC, Wardlow JL, Smail I et al (2012) MNRAS 421:284 Miller TB, Hayward CC, Chapman SC, Behroozi PS (2015) ArXiv e-prints, arXiv:1501.04105 Schiminovich D, Ilbert O, Arnouts S et al (2005) ApJ 619:L47 Reddy NA, Steidel CC (2009) ApJ 692:778 Reddy NA, Steidel CC (2015) ApJ 803:34 Bouwens RJ, Illingworth GD, Oesch PA et al (2014) ApJ 793:115 Meurer GR, Heckman TM, Calzetti D (1999) ApJ 521:64 Le Floc’h E, Papovich C, Dole H et al (2005) ApJ 632:169 Smolˇci´c V, Schinnerer E, Zamorani G et al (2009) ApJ 690:610

Chapter 2

Data and Reduction

2.1 Our Dataset To achieve a complete study, we utilize our original data and make full use of ALMA archival data in cycles 0 − 4 that became public by 2018 April. We collect a total of 1643 continuum maps in Band 6/7. Note that our data analyses for the contents in this thesis were carried out in different time epochs, when the available datasets were different. We thus made the best dataset for each content to maximize the scientific gain at each time epoch. In Table 2.1, we summarize the dataset used in each content in this thesis. In the following subsections, we describe the details of each dataset.

2.1.1 ALMA-DUST For the Re(FIR) study, we use the ALMA archival data in cycles 0–3 that are open for public by 2017 July. We collect 1627 ALMA continuum maps in Band 6 and 7 from the archival data in the regions of Hubble Ultra Deep Field (HUDF), Hubble Frontier Fields (HFF), Cosmic Evolution Survey (COSMOS; [1]), The Great Observatories Origins Deep Survey South (GOODS-S; [2]), and Subaru/X M M − N ewton Deep Survey (SXDS; [3]) that have rich multi-wavelength data. Tables 2.2 present the example (see electronic supplementary material Table 2.1 for the full version) of the data list for those of 1627 continuum maps. We refer the ALMA dataset of these 1627 maps as ALMA-DUST.

Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/978-981-16-4979-0_2.

© Springer Nature Singapore Pte Ltd. 2021 S. Fujimoto, Demographics of the Cold Universe with ALMA, Springer Theses, https://doi.org/10.1007/978-981-16-4979-0_2

11

12

2 Data and Reduction

Table 2.1 Our ALMA dataset and content Dataset

Section

Table

Nmap

Nsource

(1)

(2)

(3)

(4)

(5)

ALMA project ID (6)

ALMADUST

3

2.2

1627

1034

#2011.0.00097.S, #2013.1.00034.S

7.1

(1)

#2011.0.00294.S, #2013.1.00118.S #2012.1.00076.S, #2013.1.00151.S #2012.1.00090.S, #2013.1.00173.S #2012.1.00245.S, #2013.1.00718.S #2012.1.00307.S, #2013.1.00884.S #2012.1.00326.S, #2013.1.00205.S #2012.1.00523.S, #2013.1.00208.S #2012.1.00756.S, #2013.1.00566.S #2012.1.00775.S, #2013.1.00668.S #2012.1.00869.S, #2013.1.00781.S #2012.1.00979.S, #2013.1.00999.S #2012.1.00983.S, #2013.1.01292.S #2015.1.00137.S, #2015.1.00540.S #2015.1.00664.S, #2015.1.01495.S

ASAGAO

4

3.2

1

45

7.1 ALMAz6CII

5

#2015.1.00098.S, #2012.1.00173.S #2015.1.00543.S

2.4

18

18

7.2

#2013.1.00815.S, #2015.1.00834.S #2015.1.01111.S, #2015.1.01105.S #2016.1.01240.S, #2012.1.00523.S #2012.1.00602.S

ALMAFAINT

6

2.5

7.3

(2)

67

133

#2011.0.00115.S, #2011.0.00232.S #2011.0.00243.S, #2011.0.00319.S #2011.0.00648.S, #2011.0.00767.S #2012.1.00076.S, #2012.1.00602.S #2012.1.00676.S, #2012.1.00719.S #2012.1.00323.S, #2012.1.00536.S #2012.1.00610.S, #2012.1.00934.S #2012.1.00953.S

Total†

1643

1220

(1) Name of our ALMA dataset (2) Section that use the dataset (3) Table ID that summarizes the properties of the dataset. Table IDs in electronic supplementary material are shown in parenthesis (4) Number of the ALMA maps produced from the dataset (5) Number of the sources that are detected in the ALMA maps (6) ALMA project ID used in the dataset †The total number is corrected for the double count because of the existence of the overlapped data among the datasets

COSMOSLowz64_0 COSMOSLowz64_1 COSMOSLowz64_10 COSMOSLowz64_11 COSMOSLowz64_12 COSMOSLowz64_13 COSMOSLowz64_14 COSMOSLowz64_15 COSMOSLowz64_16 COSMOSLowz64_17

(2)

(1)

2011.1.00097.S 2011.1.00097.S 2011.1.00097.S 2011.1.00097.S 2011.1.00097.S 2011.1.00097.S 2011.1.00097.S 2011.1.00097.S 2011.1.00097.S 2011.1.00097.S

Target

Project ID

0.88 0.88 0.88 0.88 0.88 0.88 0.88 0.88 0.88 0.88

(3)

λobs (mm)

342 (7) 342 (7) 342 (7) 342 (7) 342 (7) 342 (7) 342 (7) 342 (7) 342 (7) 342 (7)

νobs (Band) (GHz) (4)

0.58×0.42 0.59×0.42 0.61×0.41 0.61×0.42 0.62×0.41 0.62×0.41 0.62×0.42 0.63×0.41 0.63×0.41 0.64×0.42

21 − 384 21 − 384 21 − 384 21 − 384 21 − 384 21 − 384 21 − 384 21 − 384 21 − 384 21 − 384 0.45 0.44 0.47 0.46 0.5 0.47 0.48 0.46 0.47 0.53

Beam size ( × ) (7)

Ant. Dist. (m) σf (mJy beam−1 ) (5) (6)

0.49 (SB 2) 0.49 (SB 2) 0.5 (SB 2) 0.5 (SB 2) 0.5 (SB 2) 0.5 (SB 2) 0.51 (SB 2) 0.5 (SB 2) 0.5 (SB 2) 0.51 (SB 2)

(8)

θcirc ( )

(1) ALMA project code (2) Target name in the initial ALMA observation (3) Wavelength in the observed frame (4) Frequency in the observed frame (5) Range of the antenna distances (6) One sigma noise measured after the CLEAN process (7) Synthesized beam size (weighting= “natural”) (8) Circularized beam size. Classification of the sub-dataset is presented in the parentheses sparabreak (Full version of table is shown in electronic supplementary material Table 2.1)

1 2 3 4 5 6 7 8 9 10

Map ID

Table 2.2 Examples of ALMA-DUST maps

2.1 Our Dataset 13

14

2 Data and Reduction

Table 2.3 Summary of Sub-dataset of ALMA-DUST Sub-dataset θcirc (1) (2) SB 1 SB 2 SB 3 SB 4 SB 5 SB 6 SB 7 SB 8 SB 9 SB 10

0 − 0. 2 0. 2 − 0. 4 0. 4 − 0. 6 0. 6 − 0. 8 0. 8 − 1. 0 1. 0 − 1. 2 1. 2 − 1. 4 1. 4 − 1. 6 1. 6 − 1. 8 1. 8 − 2. 0

Number of maps (3) 48 303 394 203 254 154 228 11 18 14

Notes (1): ALMA maps with the circularized beam sizes θcirc from 0 to 2. 0 with a step of 0. 2 are referred to as SB1 to SB10, respectively (2): Criterion of the θcirc range for each sub-dataset. (3): Number of the ALMA maps in each sub-dataset

Based on the angular resolutions (Table 2.3), we divide the ALMA-DUST dataset into 10 sub-datasets, because the Re(FIR) estimates might be affected by different systematics depending on the angular resolutions. Here, we regard the angular resolutions as the circularized beam size θcirc given by θcirc =

√ ab,

(2.1)

where a and b represent the full width half maximums (FWHMs) of the majorand minor-axis of the synthesized beam, respectively. We summarize the observing details for the 1627 continuum maps in electronic supplementary material Table 2.1. Note that the 1627 continuum maps have been taken in two different types of the observation modes. One is the single pointing observations, and the other is mosaic observations with several pointing. We refer the single pointing and mosaic observations as “single-field data” and “mosaic data”, respectively, that are presented in electronic supplementary material Table 2.1. Because the HUDF region is observed in two independent ALMA projects (#2013.1.00718.S and #2013.1.00173.S), we combine these two data sets with the concat task, after calculating the data weights with the statwt task based on its visibility scatters which include the effects of integration time, channel widths, and systematic temperature.

2.1 Our Dataset

15

2.1.2 ASAGAO To carry out the detail morphology measurements, we make use of the ALMA dataset taken in the high-resolution ALMA 1-mm mapping survey of ASAGAO: ALMA twenty-Six Arcmin2 survey of GOODS-S at One-millmeter (#2015.1.00098.S, PI: K. Kohno; see also [4, 5]). In September 2016, the ASAGAO survey carried out over an area of ∼26 arcmin2 in GOODS-S with a total observing time of 45 h. The baseline length takes a range of 15−3247 m with 38−45 antennas, where the array configuration was C40-6. The frequency setting adopts the two tunings centered at 1.14 and 1.18 mm, where the frequency covers the ranges of 244−248, 253−257, 259−263, and 268−272 GHz. To enhance the sensitivity and the uv-visibility for the ALMA data, we also utilize the previous ALMA 1-mm mapping data in GOODS-S [6, 7].

2.1.3 ALMA-z6CII To study the general [C ii] line properties at z ∼ 6, we make a sample of z ∼ 6 normal star-forming galaxies whose [C ii] lines are individually detected in the previous ALMA observations. The sample is drawn mainly from the literature by selecting only star-forming galaxies at z > 5 whose [Cii] line is detected (S/N  5) with ALMA. In order to focus on representative galaxies in the early universe, we limit our sample to (i) star-formation rate (SFR) < 100 M /yr, (ii) not reported as active galactic nucleus (AGNs), (iii) not giant Lyman-alpha systems such as Himiko [8] and CR7 [9], (vi) not gravitationally lensed systems behind massive galaxy clusters, and (v) full-width-half-maximums (FWHMs) of the [Cii] line emission is broader than 80 km/s. Note that the thermal noise fluctuation can produce peaky false source signals even with S/N > 5, if we examine the large volume data such as the ALMA 3D data cubes. This is the reason for (v). We also note that our sample does not include the tentative [Cii] line detections reported in the ALMA blind line survey (e.g., [10, 11]), that have not been spectroscopically confirmed yet. We identify 16 [Cii] line sources that meet the above criteria in the literature. Table 2.4 summarizes our sample and references that describe the details of the ALMA observation. In Table 2.4, the ALMA synthesized beam sizes in our sample are comparable within a factor of ∼2, indicating that the [Cii] line detections in our sample are originated from the similar physical-scale structures in the galaxies. In addition to the literature sample, we have identified new [Cii] line detections from two star-forming galaxies of WMH13 [22] and NB816-S-61269 [23] at 5.983 and z = 5.688, respectively. In Fig. 2.1, we present the spectra for these two [Cii] line detections with the velocity integrated maps. In the velocity integrated maps of WMH13 and NB816-S-61269, the [Cii] line is detected at the 5.2 and 6.0 σ levels at the peak, respectively, while the dust continuum is undetected from either

2.378358

181.403878

336.958267

150.099014

150.197356

149.971828

150.517186

150.039247

149.618760

150.089576

149.876925

150.016894

149.965404

NTTDF6345

BDF2203

COS13679

COS24108

Hz1

Hz2

Hz3

Hz4

Hz6

Hz7

Hz8

NB816−S−61269

6.122 (6.118)

−35.147529

5.985 (5.983) 5.684 (5.688)

2.207528

5.541 (5.548)

5.153 (5.148)

5.253 (5.250)

5.293 (5.290)

5.544 (5.310)

5.542 (5.546)

5.670 (5.670)

5.689 (5.690)

6.623 (6.629)

−5.493392

2.134061

2.586324

2.051850

2.3371611

1.928936

2.118142

2.478931

7.142 (7.145)

6.698 (6.701)

−7.756192

2.343517

6.808 (6.816)

6.854 (−)

2.217294

8.0 9.8 27.1 14.4

−22.8 −21.8 −21.8 −21.9

−20.4 93.3

27.0

10.2

−22.3

−22.0

6.9

5.3

−22.0 −3.6

27.0

−21.6 −21.9

15.0

−21.4

−21.7

9.9

−20.9

15.0

−2.9 16.2

−21.9 −22.0 −21.5

50.0

−22.6

13.0

(3)

EWLyα (Å)

22

16

43

30

35

32

64

47

35

27

20

18

20

20

25

27

18

8

(4)

† σcont. (µJy/beam)

0.45 × 0.42

1.15 × 0.89

0.64 × 0.54

0.40 × 0.29

0.47 × 0.38

0.67 × 0.50

0.89 × 0.51

0.77 × 0.42

0.83 × 0.53

0.75 × 0.52

0.81 × 0.75

0.85 × 0.85

1.85 × 1.05

1.25 × 0.97

1.07 × 0.74

1.08 × 0.74

0.52 × 0.45

0.50 × 0.46

(5)

Beam ( × )

2015.1.01111.S Y

2012.1.00602.S N

2013.1.00815.S N

2012.1.00523.S Y

2012.1.00523.S Y

2012.1.00523.S Y

2012.1.00523.S Y

2012.1.00523.S Y

2012.1.00523.S Y

2012.1.00523.S N‡

2012.1.00523.S Y

2015.1.01105.S N‡

2015.1.01105.S N‡

2016.1.01240.S N††

2015.1.01105.S N

2015.1.01111.S Y

F16

W15

C15 (B17)

C15 (B17)

C15 (B17)

C15 (B17)

C15 (B17)

C15 (B17)

C15 (B17)

C15 (B17)

P16

P16

C18

P16

S18 (S15)

S18 (S15)

W15

W15, J17

2013.1.00815.S N

W15, J17

(8)

Ref.

2015.1.00834.S

(7)

HST

2013.1.00815.S N

(6)

ALMA ID

Notes (1) Spectroscopic redshift determined by the [Cii] (Lyα) line emission (2) Absolute magnitudes (3) Rest-frame Lyα EW (4) One sigma noise measured by the standard deviation of the pixel values in the continuum map before primary beam correction (5) Synthesized beam size of our ALMA maps (weighting = “natural”) (6) ALMA project ID (7) “Y” (“N”) indicates the sources (not) included in the ALMA-HST sample (8) ALMA (HST) data reference (W15: [13], J17: [14], S18: [15], P16: [16], C18: [17], S15: [18], C15: [19], B17: [20], F16: [21]) † Our additional flagging and difference in the imaging parameter setting may produce different values from the data references †† Although there is the F105W data, we do not include this source in the ALMA-HST sample due to the differences in the PSF and the rest-frame wavelength from the F160W data ‡ Although there is the F160W data, we do not include these sources in the ALMA-HST sample due to the large offsets even after the astrometry correction (see text)

149.985580

34.438567

WMH13

New detection

Hz9

2.626631

150.124400

COS298703

2.266613

150.125803

COS301855

6.166 (6.176)

−4.271706

37.012319

CLM1

−22.6

(2)

(1)

6.069 (6.076)

MUV (mag)

z [CII] (z Lyα )

−4.877333

Dec. (J2000)

36.612542

R.A (J2000)

WMH5

Literature

Target

Table 2.4 ALMA-z6CII maps and sources 16 2 Data and Reduction

2.1 Our Dataset

17

Fig. 2.1 New [Cii] line detections of NB816-S-61269 (top) and WMH13 (botttom). Left: Naturalweighted 4 × 4 field image of the velocity-integrated [Cii] line intensity (moment zero) with contours at the −2σ (white), 2σ , 3σ , 4σ , and 5σ (red) levels (Fig. 1 of [12], reproduced by permission of AAS). The ALMA synthesized beams are presented at the bottom left. Right: [Cii] line spectra with an aperture radius of 0. 6. The solid curves denote the best-fit profile of the single gaussian with the best-fit values of the FWHM and the frequency peak. The yellow shades present the integrated velocity ranges for the [Cii] line intensity maps in the left panel

galaxy. In Table 2.4, we list the details of the ALMA observations for WMH13 and NB816-S-61269. To address the rest-frame UV properties, we also use final flat-field and fluxcalibrated science products of the HST Wide Field Camera 3 (WFC3) in F160W (i.e., H -band) maps from the Hubble Legacy Archive. There are 9 out of 18 sources in our sample that have been observed with the HST/H -band. The 9 and the 18 sources are referred to as the “ALMA-HST” and “ALMA-ALL” samples, respectively. The HST data references and the sources included in the ALMA-HST sample are summarized in Table 2.4. To correct the potential offsets in the HST astrometry reported in previous studies (e.g., [6, 24]), we use the Gaia Data Release 2 catalog [25] and calibrate the astrometry of the HST H -band maps as follows. First, we detect bright objects with sextractor version 2.5.0 [26] in the H -band maps. Second, we cross-match the GAIA catalog and the bright H -band objects. Finally, we evaluate spatial offsets of the the bright objects from the GAIA catalog positions. We find that the bright objects identified in the H -band maps have the spatial offsets from the GAIA catalog in the range of

18

2 Data and Reduction

∼ 0. 1 − 0. 3. Based on these offsets, we correct the astrometry of each H -band map 1. Note that we also check the bright quasars used as the phase calibrators in the ALMA observations and confirm that the astrometry of the ALMA maps shows an excellent agreement with the GAIA catalog within a milli-arcsec scale. Thus, we do not apply any astrometry corrections to our ALMA maps. By combining the literature and the additional samples, we obtain a total of 18 [Cii] line sources. These 18 [Cii] line sources are characterised with the source redshift determined by the [Cii] lines z [CII] = 5.153 − 7.142, the absolute rest-frame UV magnitudes (MUV ) in the ranges of  −22.8 to −20.4, and SFR  10 − 70 M /yr. We summarize the physical properties of z [CII] , MUV , and the Lyα equivalent-width (EWLyα ) in Table 2.4. We refer to the ALMA dataset for the total of 18 normal star-forming galaxies at z ∼ 6 as ALMA-z6CII.

2.1.4 ALMA-FAINT To unveil the origin of the CIB, it is essential to identify the individual faint submm/mm emission. To maximize the sample number of the faint ALMA continuum sources, we collect 67 continuum maps obtained by ∼ 120 pointing of the ALMA cycle 0–2 observations in Band 6/7 from our original and the archival data that became public by 2015 June. Some examples of these 67 ALMA maps are shown in Fig. 2.2. We define sub-datasets by the mapping modes and the depths because these two conditions would make different systematics. For the mapping mode definitions, there are maps targeting field regions by single pointing observations, referred to as ‘field’ data, 4 and 62 out of which are taken from our programs and the ALMA archive, respectively. One map for a galaxy cluster taken by mapping observations, referred to as ‘cluster’ data, is from the ALMA archive. Tables 2.5 and 2.6 present the detailed properties and the summary of the subdataset, respectively, for these 67 continuum maps. To define the data depth, we divide these 66 maps of ‘field’ data into two subdatasets with high (> 60 µJy) and low (≤ 60 µJy) noise levels, which we refer to as ‘medium-deep’ and ‘deep’ data, respectively. This is because more systematic noise could be included in the short-integration data. We use the sub-dataset names of A and B for the deep and medium-deep field data, respectively. The sub-dataset name of the deep cluster data is C. We summarize the name of the sub-dataset in Table 2.6.

2.2 Reduction

19

Fig. 2.2 Examples of our ALMA maps in the ALMA-FAINT dataset. The red squares denote the faint ALMA sources detected with ≥ 5σ level. The white dotted squares indicate the original target of each ALMA observation that are detected with ≥ 5σ level. Note that we do not include these original target sources in our analyses because our purpose is to perform a blind mm source survey. The white circle presents the primary beam of each ALMA observation that is typically ∼ 10 − 13 radius at Band 6/7. The synthesized beams are illustrated with the yellow ellipses at the bottom right. The target name for each ALMA observation is shown below each image. The map ID (electronic supplementary material Table 2.2) is also listed at the upper right

2.2 Reduction 2.2.1 Flag and Calibration The data are reduced with CASA versions from 3.4 to 4.7.2. in a standard manner with the scripts used for the data reduction provided by the ALMA observatory. Exceptionally in the cycle 0 data with the patterned noise or the high noise level, we use the calibrated data provided by the ALMA observatory. Similarly, because the coordinate of a phase calibrator of #2011.0.00648.S is found to be wrong (∼ 0. 3), which causes positional offsets [40, 41], we use a re-calibrated data provided by Seko et al. for #2011.0.00648.S If we find other problems in the final images that there remain striped patterns or that the noise level is significantly higher than the calibrated images provided by the ALMA observatory, we apply additional flaggings to some baselines. Our CASA reduction has three major steps: flagging bad data, bandpass calibration, and gain calibration with flux scaling. In the first step, we remove the edge channels of spectral windows and the data of the shadowed antennas. Besides, we do not use the unreliable data including a low phase/amplitude gain or jumps. Because we find that the flaggings in the scripts are good enough for our scientific goals, we apply the same flaggings as described in the scripts without additional flaggings. In

NB101-S-2904

NB816-S-61269 1.03

rxj0806

Himiko

BDF-521

IOK-1

rxj2143

CFHQSJ23290301

3

4

5

6

7

8

9

10

250 (6)

238 (6)

232 (6)

230 (6)

259 (6)

229 (6)

291 (7)

238 (6)

222 (6)

21.0

20.8

18.8

17.7

17.0

16.6

13.0

11.4

10.3

8.5

(3)

σcont (µJy beam−1 )

0.73×0.63

0.60×0.43

1.08×0.78

0.68×0.51

0.81×0.56

0.78×0.63

0.45×0.42

0.43×0.33

0.98×0.75

0.79×0.58

(4)

1.00

1.16

1.25

1.28

0.90

1.31

0.63

1.16

1.45

1.28

(5)

Beam size ( × ) S1.2mm /Sobs

(m),(n)

(l)

(j),(k)

(a),(b)

(h),(i)

(g)

(f)

(e)

(c),(d)

(a),(b)

(6)

Ref.

2011.0.00243.S

2012.1.00610.S

2011.0.00767.S

2012.1.00719.S

2011.0.00115.S

2012.1.00610.S

2012.1.00602.S

2012.1.00088.S

2012.1.00953.S

2012.1.00719.S

2012.A.00040.S

Project ID

(1): Wavelength in the observed frame (2): Frequency in the observed frame (3): One sigma noise measured in each map before primary beam correction (4): Synthesized beam size of our ALMA maps (weighting = ‘natural’) (5): Ratio of the flux density at 1.2 mm, S1.2mm , to the one at the observed wavelength, Sobs , that is estimated with the modified blackbody spectrum of βd = 1.8, a dust temperature of Td = 35 K, and a redshift of z = 2.5 (6): Reference. (a) [27]; (b) [28]; (c) [29]; (d) [30]; (e) [31]; (f) [32]; (g)[33]; (h) [34]; (i) [35]; (j) [36]; (k) [37]; (l) [38]; (m) [39]; (n) [22] (Full version of table is shown in electronic supplementary material Table 2.2)

1.20

1.26

1.29

1.30

1.16

1.31

1.26

1.35

GRB090423

2

227 (6)

(2)

(1)

1.30

νobs (Band) (GHz)

λobs (mm)

BDF-3299

Deep data (A)

Target

1

Map ID

Table 2.5 Examples of ALMA-FAINT maps

20 2 Data and Reduction

2.2 Reduction

21

Table 2.6 Summary of sub-dataset of ALMA-FAINT Sub-dataset Number of maps (1) (2) Deep data (A) Medium-deep data (B) Cluster data (C)

41 25 1

Notes The sub-dataset names of A, B, and C are written in parenthesis. (1): ALMA maps with low (≤ 60 µJy) and high (> 60 µJy) noise levels are referred to as Deep and Medium-deep data, respectively (2): Number of the ALMA maps in each sub-dataset

the second step, we obtain the bandpass calibration in the phase and amplitude after we calibrate the phase time variation on the bandpass calibrator scan. In the last step, we estimate the time variations of the phase and amplitude with the phase calibrators. We apply the linear interpolation with the results of the phase calibrators and then transfer the calibration to the target source. The water vapor radiometer is used for the correction of the short-time phase variations. We estimate flux scaling factors, using bright quasars and solar-system objects with the flux models of ALMA Calibrator Source Catalog and ‘Bulter-JPL-Horizons 2012’, respectively. The difference of the opacity between the target sources and the flux calibrators is corrected with the system noise temperature (Tsys ) measurements. The systematic flux uncertainty is typically 10% in bands 6 and 7 (ALMA proposer’s guide1 ).

2.2.2 Imaging The reduced ALMA uv-visibility data is Fourier transformed to create “dirty” continuum maps in the CLEAN algorithm with CASA. To maximize the scientific gain, this imaging procedures are different among the contents in this thesis. We describe the different procedures in the following subsections.

2.2.2.1

ALMA-Dust

For the single-field data, we apply the clean algorithm to produce the images in the following three steps: 1) identifying bright peak pixels with the ≥ 50σd level,2 2) setting clean boxes with boxit at the peak positions given in the step 1), and 3) carrying out clean down to the 3σd depth in the CLEAN boxes with the natural weighting. We change from 50σd to 20σd , 10σd , and 5σd in the step 1) and repeat the steps from 1) to 3) for four times. We refer to the above procedure from the dirty maps to the final maps as “clean process”. We evaluate the standard deviation of 1 2

Section A7: https://almascience.nao.ac.jp/documents-and-tools/cycle-1/alma-proposers-guide. σd is the standard deviation of the pixels in the dirty map.

22

2 Data and Reduction

the pixel values in the residual map and define it as the 1σ noise level of each ALMA map, σf . For the mosaic data, we find that the large data size is too large to perform the clean process. Thus, we perform the following three steps to reduce the data size of the mosaic data: (I) We detect the sources with the peak pixels with the ≥ 5σd level in the dirty maps. (II) We select pointings whose central positions fall within 20 from the source positions given in (I). (III) We create a new dirty map only with the uv-visibility taken with these pointings given in (II). After completing these three steps, we step forward to the CLEAN process to make final maps in the same manner as the single-field data. Because our scope is the continuum emission from the dust, atomic and molecular transition lines in the FIR wavelength detected in the ALMA data cube could be the contamination in our analysis. There are two possible cases that the channels of ALMA Band 6 and 7 include the FIR lines. First case is that the initial scope of the ALMA data from the archive is studying the FIR lines. In our ALMA data, the ALMA projects of #2012.1.00523.S, #2012.1.00076.S, and #2013.1.00668.S are taken for the [C ii] 158 µm or CO lines from high-z galaxies. For those of three ALMA data, we check the data cubes and if robust line profiles are identified, we remove the line channels in the clean process. Second case is that the FIR lines fall in the data cube by chance. Among the FIR lines, the [C ii] 158 µm line is one of the brightest FIR lines in the star-forming galaxies [42]. If the source redshifts are at z ∼ 4−7, the [C ii] 158 µm line is potentially included in ALMA Band 6 and 7. In fact, two [C ii] 158 µm lines from two SMGs at z = 4.42 and z = 4.44 have been identified in an ALMA survey of submillimeter galaxies in the Extended Chandra Deep Field-South (ALESS) from 126 ALMA data cubes [43]. This indicates that the chance probability of the [C ii] 158-µm line contamination could be negligible (∼ 1−2%). Moreover, the flux densities from the [C ii] 158-µm line identified in [43] is calculated to have the contribution to the dust continuum to be only  10%. Therefore, we interpret that the second case negligibly affect our statistical results and do not take it into account through this paper. The 1627 final maps of the pointing and mosaic data achieve the synthesized beam size (FWHM) of 0. 18 × 0. 18 − 4. 08 × 1. 07 and the sensitivities of 0.01 − 1.4 mJy beam−1 before the primary beam corrections. The properties of the final maps are summarized in electronic supplementary material Table 2.1.

2.2.2.2

ASAGAO

The details of the data calibration and reduction are described in [5]. Note that the ASAGAO and the previous ALMA data in GOODS-S [6, 7] are all combined with the concat task before the following procedure. The entire map was produced by the CLEAN algorithm with the tclean task. The CLEAN boxes were set at the peak pixel positions with S/N ≥ 5 in the manual mode. For the CLEAN box, we adopt a circle with a radius of 1. 0. The CLEAN routines were proceeded down to the 2σ level. The final natural-weighted image is characterized by a synthesized beam size

2.2 Reduction

23

32

21

20

38

6

-27°45'00"

28 24

9

11

39 7

Declination

46'00"

37

27

16 1

8

10

4

5

23

14

47'00"

40

2

13 35

41

12

15

43

48'00"

36

45

25

42

44

19

49'00"

29 26

17

34

3 30

18

3h32m48s

33

31

22

42s

36s

30s

Right Ascension Fig. 2.3 The ASAGAO map with a 250-kλ taper after the primary beam correction (Fig. 6 of [5], PASJ, 70, 105, reproduced by permission of Oxford University Press). The squares represent the detected sources. This figure is refereed from Fig. 7 of [5]

of 0. 21 × 0. 17 and an rms noise level of 38 µJy/beam. We also produced another map with a uv-taper of 160 kλ whose final synthesized beam size is 0. 94 × 0. 67 and rms noise level is 87 µJy/beam. We refer to the natural-weighted (high-resolution) and the uv-tapered (low-resolution) images as “HR” and “LR” maps, respectively. We use the HR map except for source detections and positional measurements in Sects. 4.1.1 and 4.1.2. The LR map is shown in Fig. 2.3.

2.2.2.3

ALMA-z6CII

The continuum images and line cubes are produced by the CLEAN algorithm with the tclean task with a pixel scale of 0. 01. For the line cubes, the velocity channel width is re-binned to 20 km/s, where the velocity center is adjusted to the redshift determined by the Lyα or other rest-frame UV metal lines. We do not CLEAN the line cubes because the [Cii] line is faint in each 20-km/s channel. The CLEAN boxes were set at the peak pixel positions with S/N ≥ 5 in the auto mode, and the CLEAN routines were proceeded down to the 3σ level. The final natural-weighted maps

24

2 Data and Reduction

obtained from the ALMA-z6CII achieve the standard deviation of the pixel values of 8.0–64 µJy/beam and the synthesized beam sizes of 0. 45 × 0. 42–1. 85 × 1. 05, which we summarize in Table 2.4.

2.2.2.4

ALMA-FAINT

The continuum maps are made using all the line free channels of the four spectral windows. We process the maps with the auto CLEAN algorithm down to the depth of 3σ noise levels of the dirty maps, using natural weighting. The final cleaned maps achieve angular resolutions of 0. 45 × 0. 33 −1. 88 × 0. 90. The sensitivities of these ALMA maps range from 8.5 to 95.2 µJy beam−1 before the primary beam corrections. Because the observations of the cluster data were taken place in 4 different epochs, we recalculate the data weights with the statwt task in CASA based on its visibility scatters which include the effects of integration time, channel width, and systematic temperature. After applying the data weights to the datasets, we combine these two datasets with the concat task in CASA. Since the mosaic mapping observations were conducted for the cluster, the noise levels vary by positions. To estimate the noise

Fig. 2.4 False-color image of A1689 cluster (red: i 775 , green: V625 , red: B475 ). The critical lines for background sources at z = 2.5 are shown with the magenta lines (Fig. 1 of [21], reproduced by permission of AAS). The white curves are the contours indicating the 50% sensitivity of the deepest part in the mosaic map. The white thick lines and ticks indicate the 9 regions of our noise estimates

2.2 Reduction

25

levels of the cluster data, we divide the cluster’s mosaic map in 9 regions as shown in Fig. 2.4. The data of the 9 regions achieve the 1σ noise levels in the range of 38.4–40.6 µJy.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43.

Scoville N, Aussel H, Brusa M et al (2007) ApJS 172:1 Vanzella E, Cristiani S, Dickinson M et al (2005) A&A 434:53 Furusawa H, Kosugi G, Akiyama M et al (2008) ApJS 176:1 Ueda Y, Hatsukade B, Kohno K et al (2018) ApJ 853:24 Hatsukade B, Kohno K, Yamaguchi Y et al (2018) PASJ 70:105 Dunlop JS, McLure RJ, Biggs AD et al (2017) MNRAS 466:861 Franco M, Elbaz D, Béthermin M et al (2018) A&A 620:A152 Ouchi M, Mobasher B, Shimasaku K et al (2009) ApJ 706:1136 Matthee J, Sobral D, Santos S et al (2015) MNRAS 451:400 Aravena M, Decarli R, Walter F et al (2016) ApJ 833:71 Hayatsu NH, Matsuda Y, Umehata H et al (2017) PASJ 69:45 Fujimoto S, Ouchi M, Ferrara A et al (2019) ApJ 887:107 Willott CJ, Carilli CL, Wagg J, Wang R (2015) ApJ 807:180 Jones GC, Willott CJ, Carilli CL et al (2017) ApJ 845:175 Smit R, Bouwens RJ, Carniani S et al (2018) Nature 553:178 Pentericci L, Carniani S, Castellano M et al (2016) ApJ 829:L11 Carniani S, Maiolino R, Amorin R et al (2018) MNRAS 478:1170 Smit R, Bouwens RJ, Franx M et al (2015) ApJ 801:122 Capak PL, Carilli C, Jones G et al (2015) Nature 522:455 Barisic I, Faisst AL, Capak PL et al (2017) ApJ 845:41 Fujimoto S, Ouchi M, Ono Y et al (2016) ApJS 222:1 Willott CJ, Omont A, Bergeron J (2013) ApJ 770:13 Ouchi M, Shimasaku K, Akiyama M et al (2008) ApJS 176:301 Rujopakarn W, Dunlop JS, Rieke GH et al (2016) ApJ 833:12 Rujopakarn W, Dunlop JS, Rieke GH et al (2018) A&A 616:A1 Bertin E, Arnouts S (1996) A&A 117:393 Vanzella E, Pentericci L, Fontana A et al (2011) ApJ 730:L35 Maiolino R, Carniani S, Fontana A et al (2015) MNRAS 452:54 Tanvir NR, Levan AJ, Fruchter AS et al (2012) ApJ 754:46 Berger E, Zauderer BA, Chary R-R et al (2014) ApJ 796:96 Konno A, Ouchi M, Ono Y et al (2014) ApJ 797:16 Ouchi M, Shimasaku K, Furusawa H et al (2010) ApJ 723:869 Haberl F, Motch C, Pietsch W (1998) Astronomische Nachrichten 319:97 Ouchi M, Ellis R, Ono Y et al (2013) ApJ 778:102 Ono Y, Ouchi M, Kurono Y, Momose R (2014) ApJ 795:5 Iye M, Ota K, Kashikawa N et al (2006) Nature 443:186 Ota K, Walter F, Ohta K et al (2014) ApJ 792:34 Zampieri L, Campana S, Turolla R et al (2001) A&A 378:L5 Willott CJ, Delorme P, Reylé C et al (2010) AJ 139:906 Seko A, Ohta K, Yabe K et al (2016) ApJ 819:82 Hatsukade B, Ohta K, Yabe K et al (2015) ApJ 810:91 Stacey GJ, Geis N, Genzel R et al (1991) ApJ 373:423 Swinbank AM, Karim A, Smail I et al (2012) MNRAS 427:1066

Chapter 3

Interstellar Medium Scale I: Galaxy Size

3.1 Data Analysis In this chapter, we use the dataset of ALMA-DUST (Table 2.1) whose data is open for the public by 2017 July.

3.1.1 Source Detection With sextractor version 2.5.0 [1], we conduct source extractions for our ALMA maps before primary beam corrections, where we carry out the source extraction in the high sensitivity regions. For the single-field data, we perform the source extractions in the regions with the primary beam sensitivity greater than 50%. For the mosaic data, we use the regions where the relative sensitivity to the deepest part of the mosaic map is greater than 50%. peak We identify sources with a positive peak count (Sobs ) above the 3σf level. We peak then select only the sources with Sobs ≥ 5σf to make a 5σ -peak catalog. We obtain 1034 sources in the 5σ -peak catalog. We list these 1034 sources in Electronic supplementary material Table 3. For reliable size measurements, we select 780 sources total ) above the 10σf level from the 5σf -peak catalog having total flux densities (Sobs and obtain a 10σf -total catalog. Here we measure the total flux densities with the imfit task in CASA that carries out the fitting routine with the 2D Gaussian model on the image plane. We refer the 5σf -peak and 10σf -total catalogs as “ALL5S” and “ALL10S”, respectively. We summarize the source catalogs in Table 3.1.

Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/978-981-16-4979-0_3.

© Springer Nature Singapore Pte Ltd. 2021 S. Fujimoto, Demographics of the Cold Universe with ALMA, Springer Theses, https://doi.org/10.1007/978-981-16-4979-0_3

27

28

3 Interstellar Medium Scale I: Galaxy Size

Table 3.1 Our ALMA catalog summary Catalog Name Selection criteria (1) (2)

Source number (3)

ALL5S

Sobs ≥ 5σf

peak

1034

ALL10S OC5S

total ≥ 10σ Sobs ≥ 5σf ∩ Sobs f ALL5S ∩ optical-NIR counterpart ALL10S ∩ optical-NIR counterpart OC5S ∩ FIR-mm target obs. OC5S ∩ optical target obs.

peak

780 577

OC10S OC5S-mmT OC5S-optT

444 131 446

Notes (1) Name of our ALMA source catalog (2) Selection criteria of the ALMA sources for the catalog (see text) (3) Source number in the catalog

3.1.2 Flux and Size Measurement We estimate the flux densities as well as the sizes for our 780 objects in ALL10S. For these estimates, we use uvmultifit that is a simultaneous fitting tool for multiple objects on uv-visibility developed in Nordic ALMA Regional Center [2]. In the uv-visibility fitting, we use the flux density and the size measurements as the initial values that are obtained by the imfit task, The source positions are fixed with the imfit results. Here we model the 2-dimensional surface brightness profile in a symmetric exponential disk. This is because the typical S´ersic index n is estimated to be n = 0.9 ± 0.2 among high-z SMGs [3], which is sufficiently close to the exponential disk with n = 1.0 (See also Chap. 4). Since the uvmultifit returns the best-fit size in FWHM, the FWHM values are converted to the Re values via the relation of FWHM = 0.826 × Re in the case of n = 1 [4]. Hereafter, we refer to the Re values as our size measurements. In Fig. 3.1, we show our best-fit results with the uvmultifit task for several examples. There are two possible cases of wrong size measurements. First is the contamination of the bad visibility data. Although we perform additional flagging in the data reduction, the potential bad visibility remaining in the data might cause some systematics in the visibility fitting. Second is the complicated morphology that is hard to be modeled with the single exponential disk. To remove these potential bad visibility data and low-quality fitting results, we perform the visual inspection on the visibility amplitude plots such shown in Fig. 2 and flag the ALL10S sources. We define flag = 0, 1, and 2 as a source whose fitting result is reliable, tentative, and bad, respectively. In the following analyses in this section, we use only the results with flag = 0. Electronic supplementary material Table 3 summarizes the flux and size measurements obtained by the uvmultifit task together with the flag values. For the sources with flag = 1 and 2, we do not use their size measurements and adopt their flux density measurements alone that are obtained by the imfit task.

3.1 Data Analysis

29

Fig. 3.1 Examples of the best-fit result of the uvmultifit task for ALL10S sources in the subdatasets of SB1−SB9 on the uv-visibility (Fig. 2 of [5], reproduced by permission of AAS). The black points represent the visibility amplitudes obtained from our ALMA data. The red dots indicate our best-fit results with the exponential disk model. Because there are no ALL10S sources with flag = 0 (reliable) in SB10, we do not show the example of the best-fit result of SB10

We note that there are 254 (=1034−780) sources that are not included in ALL10S but in ALL5S. We adopt the flux densities for the rest of these 254 sources in ALL5S obtained by the imfit task, and we do not estimate their sizes.

3.1.3 Simulation and Correction We investigate the systematics in the flux density and size measurements with the uvmultifit task for the ALL10S sources by performing Monte-Carlo simulations. We conduct the Monte-Carlo simulations based on each sub-dataset (Table 2.3), because different systematics might be produced by the angular resolution of the ALMA maps in these measurements. First, we randomly select one representative ALMA map from each sub-dataset where no sources are detected. We obtain 10 representative ALMA maps. Second, we create 26, 600 artificial sources with the uniform distribution of the total flux density the circularized source size. For the total flux density, because the measurements are thought to depend on the noise level of the ALMA map, here we define the total flux define by scaling the σf value and adopt a distribution range of 5σf − 45σf . For the circularized source size, we adopt a distribution range of 0. 04 − 3. 0. We then inject the artificial sources individually into the uv-visibilities of each representative ALMA map. The artificial sources are injected at random positions within a 10 radius from the centers. Finally, we obtain the flux densities and sizes of the artificial sources in the same manner as the flux density and the size measurements for the real sources. In Fig. 3.2, we show two examples of the difference between the input and output values of the total flux density and the size.

30

3 Interstellar Medium Scale I: Galaxy Size

Fig. 3.2 Relationship between input and output values of the flux density (left) and size (right) in two sub-datasets of SB1 and SB7, as for example (Fig. 3 of [5], reproduced by permission of AAS). The black, blue, and red circles (curves) denote the values (the best-fit functions) of the samples with the output peak SNR ranges of 5−10, 10−20, and 20−30, respectively. In the left panel, the flux density of the input and output values are normalized by σf . In the right panel, the open squares present the output size measurements that are below the reliable size measurement limits (Sect. 3.1.4). We do not use the open squares for the fitting

3.1 Data Analysis

31

As shown in Fig. 3.2, we find that the the input values of both the total flux density and the size are well recovered in the output values within the 1σ uncertainty. However, we also find that the output values have a slight offset from the input values in some sub-datasets, suggesting the need for corrections for the output values. To evaluate and perform the corrections for the total flux density and size measurements in each sub-dataset, we model the input values of these measurements in Fig. 3.2 as a function of output value, Sin = A0 × Sout + A1 , Re, in = B0 × Re, out + B1 ,

(3.1) (3.2)

where Sout (Sin ) and FWHMout (FWHMin ) are the output (input) values of the flux density and the size, respectively. A0 , A1 , B0 , and B1 are the free parameters. In the model fitting, we divide our simulation results into three bins with the output peak signal to noise ratio (SNR) of 5−10, 10−20, and 20−30. In Fig. 3.2, we plot the best-fit models for these three peak SNR bins. Based on the peak SNR of each source and corresponding best-fit model results, we correct the total flux density and size measurements for the ALL10S sources. We note that we apply the best-fit model result of the peak SNR = 20 − 30 to the bright ALL10S sources with the peak SNR > 30. Electronic supplementary material Table 3, we summarize our total flux density and size measurements after the corrections.

3.1.4 Comparison with Previous Measurements We compare our flux density and size measurements with the previous ALMA results in the literature [3, 6–11] to test the potential systematics in our method. In Fig. 3.3, we present the previous ALMA results with our flux density and size measurements. For a fair comparison, we convert the size values obtained in the previous ALMA results into the Re values. Figure 3.3 shows our flux density measurements agree with the previous ALMA results in the wide flux range. We also confirm that our size measurements consistent with the majority of the previous ALMA results within the ∼ 1−2 σ errors. Although the size comparison likely shows the larger scatter than that in the flux density comparison, the large scatter suggests that the size measurement is more sensitive to the approaches of the measurements: the initial values in the fitting process, the parameter range, and the assumptions of the model profile. Therefore, we conclude that both of our flux density and size measurements are consistent with the previous ALMA results, and we do not apply any additional corrections to our measurements.

32

3 Interstellar Medium Scale I: Galaxy Size

Fig. 3.3 Flux (top) and size (bottom) comparison between our measurements and the previous ALMA results in the literature (Fig. 4 of [5], reproduced by permission of AAS). The abscissa and ordinate axes provide our measurements (Sour , Re(our) ) and the previous ALMA results (Spre. , Re(pre.) ), respectively, with black squares (H16; [3]), triangles (I15; [6]), sideways triangles (S15; [7]), diamonds, (T16; [8]), stars (B16; [9]), crosses (G17; [11]), and inverse triangles (R16; [10]). For the gravitationally lensed sources in HFF, we use the flux and size measurements before the lensing correction in the comparison

3.1.5 Selection and Measurement Completeness We investigate potential completenesses in the selection and the measurements for the ALL10S sources. In Fig. 3.4, we show the ALL10S sources in the 10 sub-datasets of SB1−SB10 on the flux density and the size plane. First, we evaluate the selection completeness. We perform the selection procedures same as those of ALL10S based on the Monte-Carlo simulation results (Sect. 3.1.3).

3.1 Data Analysis

33

Fig. 3.4 Selection and measurement completeness for the ALL10S sources in the 10 sub-datasets of SB1−SB10 (Fig. 5 of [5], reproduced by permission of AAS). The gray background scale denotes the selection completeness, which is defined with the color bar shown on the right-hand side. The dashed lines represent the limits of the reliable size measurements θmin . For the θmin estimates from Eq. 3.3, we adopt β = 0.75 and λc = 3.84 for a 2σ cutoff, and assume that S corresponds to the ratio of the total flux density to the pixel noise. The hatched areas show the regions where the measurements are not reliable. The red circles show the ALL10S sources with the flux density and the size measurements after the corrections. The red shades indicate the areas below the completeness thresholds (see text)

Figure 3.4 shows the selection completeness of the 10 representative ALMA maps in the background gray scale. We find that there are two types of the selection incompletenesses. The first incompleteness is caused by one of ALL10S criteria, total ≥ 10σf (Table 3.1). The vertical region at the faint end in Fig. 3.4 corresponds Sobs to the first incompleteness. The second incompleteness is caused by another critepeak rion of ALL10S, Sobs ≥ 5σf (Table 3.1). If a source is spatially extended, the peak flux density does not reach the criterion, especially in the high-resolution ALMA maps. In the other words, the second incompleteness is caused by the difficulty of identifying the extended sources at a given resolution. The diagonal region in each panel in Fig. 3.4 corresponds to the second incompleteness. With these two types of incompletenesses, we find that our ALMA sources are placed in the parameter space with the selection completeness of  60%. We next investigate the completeness in the size measurement. We can measure a compact source size with a small size of the synthesized beam, θbeam . Moreover, we can measure the small size accurately, if a source is bright. Therefore, we cannot obtain reliable size measurements for the compact and/or faint sources, especially in the low-resolution observations. In this context, the reliable size measurement limit with an interferometer, θmin , is calculated as θmin = β

 λ 1/4 c × θbeam , 2S2

(3.3)

where λc is related to the probability cutoff for a false size-detection [12], S is the SNR of the averaged visibilities, and β is a coefficient that typically takes the

34

3 Interstellar Medium Scale I: Galaxy Size

value in the range of 0.5 − 1.0 weakly depending on the spatial distribution of the telescopes. The hatched areas in Fig. 3.4 shows the regions in which the size values are below the reliable size measurement limit. The SB2−SB8 panels show that the source distributions overlap the hatched area, where the completeness of the size measurements is not satisfied. To select the sources that are not significantly biased by these completenesses of the selection nor the size measurement, we define the completeness threshold in each sub-dataset. The red shades in Fig. 3.4 represents the areas below the completeness thresholds. In Sect. 3.2.2, we use these completeness thresholds to verify the FIR size−luminosity relation.

3.1.6 Optical and NIR Counterpart We search the optical-NIR counterparts of the ALL5S and the ALL10S sources. Table 3.2 summarizes limiting magnitudes in the H -band of the optical-NIR source catalogs of HFF, HUDF, SXDS, COSMOS, and GOODS-S that we use in our study. In recent ALMA and HST studies, it has been reported that there are two types in the offsets between the ALMA and optical-NIR source centers. The first is the intrinsic offset between the rest-frame FIR and UV-optical emission in a scale of ∼ 0. 4 [13]. The second is the astrometric uncertainty between ALMA and HST. [10] report that the sources in the HST map in GOODS-S systematically have spatial offset by Δα = −80 mas and Δδ = +260 mas from the radio emission detected in the deep VLA map whose astrometry is accurately calibrated with bright quasars in the mas scale [14, see Sect. 5.3]. We apply the corrections of Δα = −80 mas and Δδ = +260 mas [10] for the optical-NIR source centers. Since it is unknown

Table 3.2 Optical catalog summary Field Catalog reference (1) (2) SXDS GOODS-S Wide GOODS-S Deep HUDF COSMOS Wide5 1 COSMOS Deep6 2 HFF

[22] [22] [22] [22] [23] [24] [20]

Detection limit (ap.) (3) H H H H H H H

< 27.5 (0. 2) < 27.4 (0. 17) < 28.2 (0. 17) < 29.7 (0. 17) < 23.9 (2. 0)† < 26.4 (1. 0)†† 5 × 1012 L  , L FIR = 1 − 5 × 1012 L  , and L FIR < 1 × 1012 L  , respectively. The color arrows indicate the median redshifts of the three samples. Middle: The red and blue histograms present the samples of ALMA Band 6 and 7, respectively. The color arrows show the median redshifts of the two samples. Right: Redshift distribution of OC5S-mmT (solid histogram) and OC5S-optT (dashed histogram). The filled (open) arrow denotes the median redshift of OC5S-mmT (OC5S-optT)

redshift distributions originate from the same parent sample. We thus conclude that the redshift distribution of our ALMA sources are significantly affected by neither L FIR nor ALMA Band 6/7. In fact, we obtain the median redshift of z med = 2.36, which is consistent with the previous results of the blind submm/mm surveys [14, 31–33]. Second, we derive the redshift distribution of the OC5S-optT sources to test whether the OC5S-optT sources are biased due to the target selection in the initial ALMA observations. In the right panel of Fig. 3.6, we show the redshift distributions of the OC5S-mmT and OC5S-optT sources, for which we perform the KS-test again. The KS-test result indicates that we cannot rule out the possibility that these two redshift distributions arise from the same parent sample. The median redshift of the OC5S-optT sources is found to be z = 2.15 that is almost the same as z med = 2.36 obtained from the OC5S-mmT sources. The little difference of the redshift properties between the OC5S-mmT and OC5S-optT sources implies that the target selection in the initial ALMA observations does not cause significant biases in the OC5SoptT sources. Therefore, we use both OC5S-mmT and OC5S-optT in the following analyses with z med = 2.36.

3.2.2 FIR Size and Luminosity Relation Although we confirm that the redshift distributions of our ALMA sources are not significantly biased, there might be still any potential biases in the Re(FIR) and L FIR measurements for our ALMA sources. One potential bias is the existence of the opti-

3.2 Result

39

Fig. 3.7 Comparison of the FIR size−luminosity relations (Fig. 8 of [5], reproduced by permission of AAS). The filled and open circles indicate our ALMA sources above and below the completeness thresholds (Sect. 3.1.5), respectively. Top: The red and black circles represent the OC10S and ALL10S sources, respectively. For the RFIR and L FIR estimates, we assume that all of the OC10S and ALL10S sources reside at z med = 2.36. Bottom: The red and black circles denote the OC10S sources whose L FIR values are derived with the assumptions of Td = 35 K and the Td −L FIR relation (see text), respectively

cal counterparts, i.e., the difference between OC10S and ALL10S. Another potential bias is our derivation for L FIR by assuming the single dust temperature, Td = 35 K, in the modified black body (Sect. 3.1.6). Before we investigate the Re(FIR) and L FIR relation, we address the potential biases with the following two tests. First test is whether there is a difference between OC10S and ALL10S in the Re(FIR) −L FIR relation. The top panel of Fig. 3.7 shows the Re(FIR) measurements as a function of L FIR for the ALL10S (black circle) and OC10S sources (red circle). In this test, we assume all of the source redshifts at z med = 2.36 (Sect. 3.2.1) to calculate the Re(FIR) and L FIR values and to examine their typical trends. To evaluate

40

3 Interstellar Medium Scale I: Galaxy Size

the distributions of OC10S and ALL10S, a power-law function is fitted to the Re(FIR) and L FIR values. The power-law function is given by Re(FIR) = R0

L

FIR

L0

αIR

,

(3.8)

where R0 is the effective radius at a given luminosity of L 0 . We adopt the L 0 value from the best-fit Schechter parameter L ∗ at z ∼ 3. R0 and α FIR are free parameters in the fitting. Based on the ALMA sources that are unbiased neither by the measurement nor selection completenesses. the completeness thresholds (Sect. 3.1.5) is adopted in the fitting. In Fig. 3.7, our ALMA sources below and above the completeness thresholds are shown in open and filled red circles, respectively. We use our ALMA sources above the completeness thresholds alone in the following analyses in this section. The 1σ -error range of the power-law function is evaluated by the perturbation method [34]. The 1000 data sets of the Re(FIR) and L FIR for OC10S and ALL10S are generated based on the random perturbations following the Gaussian distribution whose sigma is defined by the observational errors. The best-fit power-law function is obtained for each of the 1000 data sets. The error of the power-law function is defined as the range of 68% distribution of Re(FIR) from the 1000 best-fit power-law functions. We refer to this process for evaluating the 1σ -error range as the perturbation method. In the top panel of Fig. 3.7, the 1σ -error ranges of the power-law functions of OC10S and ALL10S are shown in the black and red shades, respectively. We find that the red shade shows an agreement with the black shade. This indicates that OC10S does not have the significant difference in the Re(FIR) distribution from ALL10S. We thus conclude that the OC10S sources generally represent the ALL10S sources. Because the ALL10S sources with no optical-NIR counterparts do not allow us to examine the physical properties, we use the OC10S sources alone with the individual photometric redshifts in the following analyses in this subsection. Second test is whether our assumption of the single dust temperature (Td = 35 K) produces any potential biases in the Re(FIR) −L FIR relation. Recent studies report that there is a positive correlation between Td and L FIR among H er schel-selected galaxies at 0.1 < z < 2 (e.g., [35]). We thus re-evaluate the L FIR values with the Td −L FIR relation for OC10S, and compare the source distribution on the Re(FIR) −L FIR plane to that obtained with the original derivation with the fixed dust temperature of Td = 35 K. For the Td −L FIR relation, we model Td as a function of L FIR with a linear function, Td = C0 log(L FIR ) + C1 ,

(3.9)

where C0 and C1 are the free parameters. Note that it is unclear that H er scheland ALMA-selected galaxies have the same Td −L FIR relation. This is because the wavelength coverages of ALMA Band 6 and 7 are sensitive to the galaxies with the dust temperatures colder than H er schel-selected galaxies at a given L FIR (e.g., [21]). We thus evaluate the Td −L FIR relation with the following three steps: a) fitting the linear function to the H er schel-selected galaxies in [35], b) fixing the C0 value of the best-fit linear function obtained in a), and c) fitting the linear function with the fixed

3.2 Result

41

Fig. 3.8 FIR size−luminosity relation of our best estimates with OC10S representing ALL10S (Fig. 9 of [5], reproduced by permission of AAS). The red filled and open circles are the OC10S sources above and below the completeness thresholds (Sect. 3.1.5), respectively. The shaded region indicates 1σ uncertainty range of the best-fit power-law function that is calculated by the perturbation method. In the power-law fitting, we do not include the OC10S sources shown with the open circles. The black dashed lines denote the constant surface SFR density ΣSFR =10, 100, and 1000 M /yr/kpc2 estimated from Eq. 3.7 and the assumption of the uniform surface density with Re(FIR) . Other symbols denote the previous ALMA results in the same assignment as Fig. 3.3, except for the blue pentagons (L16; [37]). The symbol colors indicate that the types in the initial ALMA observations of the submm/mm bands selected target (black), the optical/NIR bands selected target (blue), and the blind survey (green)

C0 to the ALMA-selected galaxies whose Td and L FIR values are well determined with H er schel+ALMA bands in [36] and [21]. We can then analytically obtain the Td and L FIR values by combining this Td −L FIR relation and Eq. 3.4, because Eq. 3.4 also presents L FIR as a function of Td . In the bottom panel of Fig. 3.7, the black and red circles present the L FIR values of OC10S that are derived from the Td −L FIR relation and Td = 35 K, respectively. The 1σ -error ranges of the power-law functions for the Re(FIR) and L FIR values obtained is shown in the color shades that are evaluated by the perturbation method again. This shows that the power-law functions are consistent within the 1σ errors. We conclude that the Re(FIR) −L FIR relation is not significantly affected by our assumption of Td = 35 K. We thus adopt the L FIR values estimated by Td = 35 K in the following analyses. Figure 3.8 shows our best estimates of the Re(FIR) and L FIR values for our ALMA sources in the red circles. We find that our ALMA sources fall in the range of Re(FIR) ∼ 0.3 − 5 kpc. From Eq. 3.7 and the assumption of the uniform surface density with

42

3 Interstellar Medium Scale I: Galaxy Size

Re(FIR) , we calculate that our ALMA sources have the SFR surface densities ΣSFR in the range of ∼ 8−800 M /yr/kpc2 . In Fig. 3.8, we also show previous ALMA results of the FIR size studies with various symbols for comparison. The high-resolution (0. 16−0. 3) observations for the SMGs at z  2 are shown with the black symbols [3, 6, 7], where the FIR size for the SMGs are estimated to be Re(FIR) ∼ 0.3 − 3.0 kpc. The green symbols denote the results of the deep observations for one blank field of HUDF and three gravitational lensing clusters of HFF [10, 11]. These deep observations explore the FIR size measurements for the sources fainter than SMGs, and these faint sources are estimated to have typical FIR sizes of Re(FIR) ∼ 1.5 − 2.0 kpc. The blue symbols indicate the results for optically-selected galaxies. The stacking results for the starforming galaxies at z ∼ 2 show the typical FIR sizes of Re(FIR) ∼ 2.5 kpc [37], while the individual results for the massive (Mstar ∼ 1010.5−11.5 M ) star-forming galaxies at z ∼ 2.5 report that their rest-frame FIR sizes fall in the range of Re(FIR) = 0.3 − 4.0 kpc [8, 9]. These results of the previous studies indicate that there is a large scatter in the Re(FIR) − L FIR relation, We find that our estimates of Re(FIR) and ΣSFR are generally consistent with these previous results within the scatter. Nest, we verify the Re(FIR) − L FIR relation. We divide our sample into four redshift samples of z = 0−1, 1−2, 2−4, and 4−6. We then perform Spearman’s rank test for these redshift samples to investigate the correlation between L FIR and Re(FIR) . We identify the positive correlations between L FIR and Re(FIR) at the ∼90−98% significance levels in the four redshift samples. When we combine the redshift samples all together, the significance level increases at >99%. The power-law function of Eq. 3.8 is fitted to the Re(FIR) − L FIR relation to evaluate the slope of the positive correlation. We evaluate the α FIR value with the redshift sample of z = 2−4, because this sample has the largest source number among the four redshift samples. We obtain α FIR = 0.23 ± 0.07. With the redshift samples all together, the α FIR value becomes 0.28 ± 0.07, This indicates that α FIR falls in the range of ∼ 0.2 − 0.3 in any cases, and thus we use a fiducial value of α FIR = 0.28 ± 0.07 in the following analysis. Note that we confirm that the α FIR value is unchanged from the fiducial value within the error in the following two cases: when another profile with the S´ersic index different from n = 1 is assumed in the size measurements in Sect. 3.1.2, and when the powerlaw fitting is performed to the redshift samples all together with no completeness thresholds. We also derive the Re(FIR) − L FIR relation without the gravitationally lensed sources, the AGNs, and the potentially low-quality data in early ALMA cycles, to investigate any influences on the positive correlation by these sources. For the lensed sources, the OC10S sources are regarded as the potential lensed sources, if the optical-NIR counterparts of the OC10S sources have the photometric redshifts at z ≤ 1. This is because the lensed sources typically have the lensing pairs at z ≤1 (e.g., [21, 38]). For the AGN identifications, the OC10S sources are cross-matched with the X-ray AGNs in the literature [39–41]. If the OC10S source is selected in the initial ALMA observations because it is an AGN, we also regard the OC10S source as AGN. For the data in early ALMA cycles, we assume all of the ALMA data in cycle 0 and 1 as the potentially low-quality data. Based on the perturbation

3.2 Result

43

Fig. 3.9 Comparison of the FIR size−luminosity relations (Fig. 10 of [5], reproduced by permission of AAS). The red circles present the OC10S sources in the same assignment as Fig. 3.8. The blue, green, and black shade regions denote the 1σ uncertainty range of the best-fit Re(FIR) −L FIR relation for the OC10S sources without the AGNs (blue squares), potential lensed sources (green diamonds), and the data in ALMA cycles 0/1 (black inner circles), respectively, that are estimated by the perturbation method. The red hatched region represents the 1σ uncertainty range of the best-fit Re(FIR) −L FIR relation that is shown in Fig. 3.8. The dashed lines are the constant surface SFR density in the same assignment as Fig. 3.8

method, Fig. 3.9 shows the Re(FIR) − L FIR relations for the OC10S sources without the potential lensed sources, the AGNs, the potentially low-quality data of ALMA cycle 0 and 1 with the green, blue, and black shades, respectively. For comparison, we also present our best estimate of the Re(FIR) − L FIR relation with the red hatched region. In Fig. 3.9, we find that our best estimate of the Re(FIR) − L FIR relation agrees with all three cases within the errors. We thus conclude that the contaminations of the lensed sources, the AGNs, and the early ALMA cycles do not have the significant influence on the positive correlation in the Re(FIR) − L FIR relation. Interestingly, we also find in Fig. 3.9 that the X-ray AGNs likely have the ΣSFR values higher than the general values of the other ALMA sources. Although there remain uncertainties such in the completeness of the X-ray source catalogs and the L FIR measurements, this may indicate that the high ΣSFR values is related to the gas accreting mechanisms of the AGNs [42]. We compare our α FIR estimate with the another power-law slope obtained in the UV wavelength, α UV . [43] carried out one of the most extensive UV size studies so far by analyzing ∼ 190, 000 star-forming galaxies at z = 0−8 with deep HST images, where the α UV value was evaluated to be 0.27 ± 0.01. Interestingly, our estimate of

44

3 Interstellar Medium Scale I: Galaxy Size

Fig. 3.10 Redshift evolution of the rest-frame FIR size−luminosity relation for the OC10S sources at z = 1 − 6 (Fig. 11 of [5], reproduced by permission of AAS). The colors indicate the redshift ranges described at the bottom right of each panel. The solid lines and the shade regions present the best-fit power-law functions and the associated 1σ errors, respectively. In the power law function fitting, the α FIR value of each redshift is fixed to α FIR = 0.28 that is the best-fit value form the redshift samples all together

α FIR = 0.28 ± 0.07 shows an excellent agreement with this α UV measurement. We discuss the physical origins of the consistent α FIR and α UV values in Sect. 3.2.3. We also study the redshift evolution of the Re(FIR) − L FIR relation. Because the source number in the redshift sample of z = 0−1 is much less than that of the other samples, here we study the redshift samples of z = 1−2, 2−4, and 4−6 alone. To overcome the small statistics in the individual redshift samples, the α FIR value is fixed to evaluate the R0 values for all of the redshift samples. In Fig. 3.10, we present the best-fit functions and the associated 1σ errors in the solid lines and the shaded regions, respectively. For comparison, we show the 1σ -error shade of z = 1−2 sample in all of the redshift panels. We find that there is a decreasing trend toward high redshift in the best-fit R0 values. This redshift trend is consistent with that obtained in the rest-frame UV and optical wavelengths for star-forming galaxies (e.g., [43]). Note that the CMB effect might produce a bias in the Re(FIR) measurements towards high redshifts [44]. To examine whether this decreasing trend toward high redshift is caused by this CMB effect, we compare the Re(FIR) values between the sources observed with ALMA Band 6 and Band 7 in the redshift sample of z = 4−6 This is because the CMB effect has more significant influences on ALMA Band 6 than Band 7, especially at z > 4. We find no significant difference in the Re(FIR) values between the sources observed with ALMA Band 6 and Band 7, suggesting that the CMB effect is negligible in our results.

3.2 Result

45

3.2.3 Slope of FIR Size and Luminosity Relation Here we discuss physical origins of our best-fit power-law slope of α FIR = 0.28 ± 0.07 obtained in Sect. 3.2.2. In the rest-frame UV and optical studies, the star-forming and compact quiescent galaxies show different power-law slopes in the size−mass relation (e.g., [45–48]). The star-forming galaxies show the power-law slope of ∼ 0.2 − 0.3, which is a good agreement with predictions from the disk formation models [45–47]. On the other hand, the compact quiescent galaxies show a relatively steep power-law slope of ∼ 0.5 − 0.7, where is thought to be caused by repeated mergers[45, 48]. Interestingly, our best estimate of α FIR is consistent with the case of the star-forming galaxies. This indicates that the origin of the slope α FIR might be also explained by the disk formation. In fact, recent ALMA studies show that the dusty star-forming galaxies have the disk-like morphologies in the rest-frame FIR wavelength [3, 9]. The CO observations also support that dusty star-forming galaxies have the gas disks with the rotating kinematics (e.g., [49]). These results suggest that the rest-FIR emission in our ALMA sources is taking place in the disk formation process that causes the slope of α FIR = 0.28 ± 0.07. It is worth mentioning that no ALMA sources fall in the regime of ΣSFR > 800 M /yr/kpc2 . [7] estimate the Eddington limit of the SMGs to be ∼ 1000 M /yr/kpc2 based on the balance between the radiation pressure from the star-formation and the self-gravitation, which is almost consistent with the maximum ΣSFR values among our ALMA sources. Although it is unlikely that all of our ALMA sources reach the Eddington limit, this Eddington limit may also contribute to the slope of the Re(FIR) − L FIR relation. We note that recent local (U)LIRG studies report the negative α FIR slope in the Re(FIR) − L FIR relation [50]. Given the low gas fractions in the local (U)LIRGs (e.g., [51]), the strong gas compression, such as major mergers, might be essential to give arise dusty starbursts in the local Universe. The different α FIR slopes is presumably provided by the different mechanisms for triggering the dusty starbursts between local and high redshift Universe.

3.2.4 Spatial Offset Between FIR and UV-Optical Emission We evaluate the spatial offsets between the rest-frame UV-optical and FIR emission from our ALMA sources. Here we use OC5S sources that have the optical-NIR counterparts in the 3D-HST images, because a high resolution in the rest-frame UVoptical map is important for the spatial offset measurements. By cross-matching with the 3D-HST source catalog of [24], we identify 136 OC5S sources whose optical-NIR counterparts are detected in the deep 3D-HST maps. Figure 3.11 shows the spatial offsets between the source centersz in the HST/H band and ALMA maps for the 136 OC5S sources. The individual and average spatial offsets are presented in the black circles and the red cross, respectively. We find

46

3 Interstellar Medium Scale I: Galaxy Size

Fig. 3.11 Spatial offsets between source centers of ALMA and HST/H -band counterparts (Fig. 14 of [5], reproduced by permission of AAS). The black circles show the 136 OC5S sources that are identified in the 3D-HST regions. The typical error scale is presented at the bottom right. The red cross indicates the average offset, (ΔR.A,ΔDec) = (−0. 02,0. 00). The dashed circle and the shade region denote the median and 16−84th percentiles of the data distribution, respectively

that the average spatial offset has almost no shift from the center (ΔR.A = −0. 02, ΔDec=0. 00), which falls well within the typical error scale of the individual offsets (∼ 0. 1). On other hand, we also find that some individual plots having the large scatter with a maximum spatial offset from the center of 0. 92. There are two possibilities for this large scatter. First is astrometry uncertainty. Recent ALMA studies report the astrometry uncertainty of the HST maps that cause a systematic offset of ∼ 0. 25 between the ALMA and HST images in GOODS-S [14]. In fact, the median value of the individual scatters is estimated to be 0. 24. Although no similar astrometry uncertainty in other fields might be suggested by the fact that the good agreement between the average offset and the center in Fig. 3.11, there still remain the possibility that the part of the individual scatters is caused by the astrometry uncertainty. Second is the intrinsic offset between the rest-frame UV-optical and FIR emission. In Fig. 3.11, the spatial offsets are larger than the potential astrometry uncertainty of 0. 25 in about half of the sources. The existence of the intrinsic offsets beyond the astrometry uncertainty is also supported by recent ALMA studies, where highredshift SMGs typically have the spatial offset of 0. 4 ± 0. 05 with the HST and ALMA observations [13]. In the second case, one interpretation is that the spatial offset between the restframe FIR and UV-optical emission is caused by the heavy dust obscuration. The dusty and intensely star-forming area produce the rest-frame FIR emission, while the moderately star-forming area with a small amount of dust contributes to the rest-frame UV-optical emission/ Another interpretation is that some of the opticalNIR counterparts are physically unrelated to the ALMA sources. We evaluate the probability of the misidentification due to the chance projection based on the spatial

3.2 Result

47

Fig. 3.12 Left: Redshift evolution of the average values of Re(FIR) (red cycles), Re(UV) (blue shade) and Re(opt.) (green shade) obtain in this work and the literature (see Sect. 3.2.5). The average Re(UV) and Re(opt.) values are estimated from the results of [43]. Right: Rest-frame FIR and UV-optical sizes in the individual galaxies. The filled and open circles present the Re(Opt.) and Re(UV) values, respectively, in the abscissa axis. The red symbols show the OC10S sources whose Re(Opt.) and/or Re(UV) values are estimated with flag = 0 (reliable) in [53]. The black symbols are given in the results of T16 and B16. The figure is reproduced from Fig. 12 of [5] by permission of AAS

offset range of 0. 25−0. 92 whose lower and upper limits are defined by the astrometry uncertainty and the maximum spatial offset, respectively. Following the calculation method same as [52] and assuming the number density of the H -band sources given in [24], we obtain the P-value of the chance projection to be ∼ 1% − 13%. Then, the expected number of the misidentification is estimated to be ∼ 3 by integrating the P-values, indicating that the misidentification negligibly contributes to our statistical results. Therefore, we conclude that the majority of the large spatial offset beyond the potential astrometry uncertainty of the HST map is caused by the intrinsic offset between the rest-frame FIR and UV-optical emission causes.

3.2.5 Sizes in UV, Optical, and FIR To study the multi-wavelength properties of the dusty starbursts, we compare the Re(FIR) values with the Re(UV) and Re(Opt.) values. We perform two approaches for this comparison: the statistical and individual approaches. First in the statistical approach, we compare the average Re values in the restframe UV-optical and FIR wavelengths as a function of redshift. Here we adopt our best-fit relations of the Re(FIR) − L FIR relation at z = 1−2, 2−4, and 4−6, in the range of L FIR = 1012 −1013 L  for the average Re(FIR) values. In the left panel of Fig. 3.12, we show the average Re(FIR) values as a function of redshift with the 1σ errors. For comparison, we also present the average values of Re(UV) and Re(Opt.) that are estimated from the result of [43]. The average values of Re(UV) and Re(Opt.) are

48

3 Interstellar Medium Scale I: Galaxy Size

obtained by fitting a function of Bz (1 + z)βz , where Bz and βz are free parameters. To perform a fair comparison, we adopt the fitting results for star-forming galaxies in the range of Mstar = 1010.5 − 1011 M that is comparable to our ALMA sources (Sect. 3.1.7). We find that the average Re(UV) and Re(Opt.) values are always higher than those of Re(FIR) in the wide redshift range at z = 1 − 6. Second in the individual approach, we directly compare Re between the rest-frame UV-optical and FIR wavelengths for the individual sources. To obtain the UV-optical results, the OC10S sources are cross-matched with the 3D-HST sources whose Re values are measured with the HST/J - and H -bands in [53]. Here we focus on the 3D-HST sources with the measurement flag = 0 (reliable) in [53]. We identify 53 sources by this cross-matching, and the individual comparison results are shown in the right panel of Fig. 3.12. Note that we regard the Re measurements in the H -band at z = 1 − 3 as Re(Opt.) . while the Re measurements in the J -band at z = 2.5 − 6 as Re(UV) . In the right panel of Fig. 3.12, we find that most of our ALMA sources fall in the regime of Re(FIR)  Re(UV) (Re(Opt.) ) within the errors. The small Re(FIR) trend is consistent with the previous results of T16 and B16 whose results are also presented in the right panel of Fig. 3.12. Although a few ALMA sources have Re(UV) or Re(Opt.) exceptionally smaller than Re(FIR) beyond the errors, the small Re(UV) or Re(Opt.) values are presumably caused by the misidentification for the opticalNIR counterparts or the significant dust obscuration. We conclude that Re(FIR) is generally smaller than Re(UV) and Re(Opt.) , which is consistent with the result from the first statistical approach. Both results of the statistical and individual approaches indicate that the Re(FIR) values are smaller than the Re(UV) and Re(Opt.) values. This trend would suggest that the dusty starbursts take place in the compact regions.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18.

Bertin E, Arnouts S (1996) A&A 117:393 Martí-Vidal I, Vlemmings WHT, Muller S, Casey S (2014) A&A 563:A136 Hodge JA, Swinbank AM, Simpson JM et al (2016) ApJ 833:103 MacArthur LA, Courteau S, Holtzman JA (2003) ApJ 582:689 Fujimoto S, Ouchi M, Shibuya T, Nagai H (2017) ApJ 850:1 Fujimoto S, Ouchi M, Shibuya T, Nagai H (2015) ApJ 810:133 Fujimoto S, Ouchi M, Shibuya T, Nagai H (2015) ApJ 799:81 Tadaki K-I, Genzel R, Kodama T et al (2017) ApJ 834:135 Barro G, Kriek M, Pérez-González PG et al (2016) ApJ 827:L32 Rujopakarn W, Dunlop JS, Rieke GH et al (2016) ApJ 833:12 González-López J, Bauer FE, Romero-Cañizales C et al (2017) A&A 597:A41 Martí-Vidal I, Pérez-Torres MA, Lobanov AP (2012) A&A 541:A135 Chen C-C, Smail I, Swinbank AM et al (2015) ApJ 799:194 Dunlop JS, McLure RJ, Biggs AD et al (2017) MNRAS 466:861 Chapin EL, Pope A, Scott D et al (2009) MNRAS 398:1793 Planck Collaboration, Abergel A, Ade PAR et al (2011) A&A 536:A21 Kovács A, Chapman SC, Dowell CD et al (2006) ApJ 650:592 Coppin K, Halpern M, Scott D et al (2008) MNRAS 384:1597

References 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53.

da Cunha E, Groves B, Walter F et al (2013) ApJ 766:13 Castellano M, Amorín R, Merlin E et al (2016) A&A 590:A31 Castellano M, Amorín R, Merlin E et al (2017) ApJ 839:58 Santini P, Ferguson HC, Fontana A et al (2015) ApJ 801:97 Ilbert O, McCracken HJ, Le Fèvre O et al (2013) A&A 556:A55 Momcheva IG, Brammer GB, van Dokkum PG et al (2016) ApJS 225:27 Chabrier G (2003) PASP 115:763 Salpeter EE (1955) ApJ 121:161 Murphy EJ, Condon JJ, Schinnerer E et al (2011) ApJ 737:67 Rodighiero G, Daddi E, Baronchelli I et al (2011) ApJ 739:L40 Speagle JS, Steinhardt CL, Capak PL, Silverman JD (2014) ApJS 214:15 da Cunha E, Walter F, Smail IR et al (2015) ApJ 806:110 Chapman SC, Blain AW, Smail I, Ivison RJ (2005) ApJ 622:772 Yun MS, Scott KS, Guo Y et al (2012) MNRAS 420:957 Simpson JM, Swinbank AM, Smail I et al (2014) ApJ 788:125 Curran PA (2014) ArXiv e-prints, arXiv:1411.3816 Symeonidis M, Vaccari M, Berta S et al (2013) MNRAS 431:2317 Swinbank AM, Simpson JM, Smail I et al (2014) MNRAS 438:1267 Lindroos L, Knudsen KK, Fan L et al (2016) MNRAS 462:1192 Sonnenfeld A, Chan JHH, Shu Y, et al (2017) ArXiv e-prints, arXiv:1704.01585 Ueda Y, Watson MG, Stewart IM et al (2008) ApJS 179:124 Xue YQ, Luo B, Brandt WN et al (2011) ApJS 195:10 Civano F, Elvis M, Brusa M et al (2012) ApJS 201:30 Izumi T, Kawakatu N, Kohno K (2016) ApJ 827:81 Shibuya T, Ouchi M, Harikane Y (2015) ApJS 219:15 Zhang Z-Y, Papadopoulos PP, Ivison RJ et al (2016) R Soc Open Sci 3:160025 Shen S, Mo HJ, White SDM et al (2003) MNRAS 343:978 Huang K-H, Ferguson HC, Ravindranath S, Su J (2013) ApJ 765:68 van der Wel A, Franx M, van Dokkum PG et al (2014) ApJ 788:28 van Dokkum PG, Nelson EJ, Franx M et al (2015) ApJ 813:23 Hodge JA, Carilli CL, Walter F et al (2012) ApJ 760:11 Lutz D, Berta S, Contursi A et al (2016) A&A 591:A136 Casey CM, Narayanan D, Cooray A (2014) Phys Rep 541:45 Downes AJB, Peacock JA, Savage A, Carrie DR (1986) MNRAS 218:31 van der Wel A, Bell EF, Häussler B et al (2012) ApJS 203:24

49

Chapter 4

Insterstellar Medium Scale II: Galaxy Morphology

4.1 Data Analysis In this chapter, we utilize the high-resolution (∼0. 19) ASAGAO dataset (Table 2.1).

4.1.1 Source Detection The source extraction is performed with our ASAGAO map before primary beam correction. To improve the surface brightness sensitivity, here we use the LR map, because the dust emission from the high-z star-forming galaxies is generally resolved with the angular resolutions similar to the HR map in previous ALMA high-resolution studies (e.g., [1, 2]). We then select peaks whose pixel values exceed the 3.5σ levels and obtain 631 source candidates. We perform the following two steps to identify reliable ALMA sources from the 631 source candidates. In the first step, we search possible optical and NIR counterparts for the 631 source candidates by utilizing the ZFOURGE catalog [3] and select those have the counterparts within a radius of 0. 5. This search radius is enlarged up to the half-light radius of the ALMA source, if the ALMA source is spatially extended. Given the recent ALMA reports of the systematic astrometry offset of the HST maps in GOODS-S [4, 5], a correction of −0. 086 in RA and +0. 282 in Dec is applied to the source centers in the ZFOURGE catalog with respect to the ALMA image before the above procedure, which is well-calibrated by stars in the Gaia Data Release 1 catalog [6] within the ASAGAO map. Out of the 631 source candidates, we identify 87 objects that have nearby possible optical and NIR counterparts. In the

Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/978-981-16-4979-0_4.

© Springer Nature Singapore Pte Ltd. 2021 S. Fujimoto, Demographics of the Cold Universe with ALMA, Springer Theses, https://doi.org/10.1007/978-981-16-4979-0_4

51

52

4 Insterstellar Medium Scale II: Galaxy Morphology

second step, we use magphys [7] to perform the multi-wavelength spectral energy distribution (SED) fitting to the possible optical-NIR counterparts and estimate their expected flux densities at ALMA Band 6. We regard the possible optical-NIR conterpart as a reliable optical-NIR counterpart of the ALMA source, if its expected flux density is consistent within 3σ levels with the total flux density measurement for the ALMA source. The procedures of the total flux density measurement and the SED fitting are described in Sects. 4.1.2 and 4.1.3. We find that 42 out of 87 sources have the reliable optical-NIR counterparts, We conclude that these 42 sources are reliable ALMA sources, referred to as the ASAGAO catalog. Note that the probability of the chance projection (P-value; [8]) in the first step is estimated to be ∼0.03 based on the number density of the ZFOURGE sources in our ASAGAO map. This indicates that ∼19 (= 631 × 0.03) out of 87 nearby optical-NIR objects are mistakenly identified due to the chance projection. The number of the sources rejected in the second step is 45 (= 87 − 42) which is much larger than 19. This supports that a high purity of the real sources is maintained in the ASAGAO catalog. The ASAGAO catalog is listed in electronic supplementary material Table 4. The details of the selection process are presented in [9].

4.1.2 Flux Density and Position Measurement For the flux density and noise evaluations, we conduct a 2D elliptical Gaussian fitting for the ALMA observed images with aegean [10]. Given the uncertainty of the fitting, we adopt the peak flux density if the integrated flux density is smaller than the peak flux density. Otherwise, we adopt the integrated flux density. We estimate the background and the noise by utilizing the BANE package that applies the 3σ clipping to the signal map, calculates the standard deviation in a sparse grid of pixels, and then interpolates to produce a mock noise image. The flux densities of our ASAGAO sources in the range of ∼0.3−2.3 mJy are listed in electronic supplementary material Table 4. To test the validation of our flux measurements, we compare our results with other studies for some of our ASAGAO sources that have also been identified with previous ALMA Band 6 surveys in the same region [5, 11, 12]. We find that our flux measurements agree with other ALMA study results within the errors. We present further details of the flux and noise measurements and the comparison with other ALMA Band 6 surveys in [9]. For the position measurement, first we test which LR or HR map has a less positional uncertainty by performing Monte-Carlo (MC) simulations. This is because the ALMA sources in the HR map has a better spatial resolution, but might be more significantly affected by noise fluctuations with the reduced surface brightness due to the high spatial resolution. In the MC simulations, we produce 1,000 artificial sources that have a symmetric 2D profile and uniform distributions of total flux densities of 0.3−1.0 mJy and Re of 0. 1−0. 3 taken from from previous ALMA size measurement results for high-redshift star-forming galaxies (e.g., [1, 13–15]). We then carry out the following five steps. (I) injecting the artificial source into the uv-visibilities of

4.1 Data Analysis

53

the ASAGAO map individually at a random position, and create a LR map around the artificial source position with a pixel scale of 0. 01. (II) checking whether a peak pixel is detected above the 3.5 σ level within a radius of 0. 5 from the artificial source center. (III) If the peak pixel is detected, we also produce a HR map with the same pixel scale as that of the LR map. (IV) measuring the source positions in the LR and the HR maps by obtaining the peak pixel positions within a radius of 0. 5. (V) obtaining spatial offsets of the peak pixel positions from the artificial source center in both the LR and HR maps. We repeat the steps of (I)−(V) 1,000 times, changing the injected artificial sources in the step (I). Note that we do not determine the source center by performing any profile fitting. This is because systematic uncertainties might be large for those objects detected as low S/N levels as our ASAGAO sample. The MC simulation results show that the average offset of ∼0. 06 and ∼0. 10 in the LR and HR map, respectively. This indicates that the LR map has less uncertainties in the measurements for the source centers with the typical source properties of total flux densities of 0.3−1.0 mJy and Re of 0. 1−0. 3. The ASAGAO source centers are thus measured based on the peak pixel positions in the LR map. The potential effect of the positional uncertainty of ∼0. 06 on the stacking analysis is discussed in Sect. 4.1.5.

4.1.3 Multi-wavelength Properties of Our Sample The details of the physical properties for the ASAGAO sources are presented in [16], while here we briefly summarize general physical properties of the ASAGAO sources, such as photometric redshifts (z phot ), spectroscopic (z spec ), stellar mass (Mstar ), star-formation rate (SFR), and IR luminosity (L IR ). We obtain z spec and z phot from the ZFOURGE catalog [3], which shows the redshift distribution falling in the range of z = 0.66−4.36 with a median value of z = 1.97. We evaluate the SFR, Mstar , and L IR values by performing magphys [7] based on the rich multi-wavelength data of 47 bands from the rest-frame UV to FIR wavelengths in GOODS-S, including Spit zer /MIPS (Rieke et al. 2004, 24 μm), H er schel/PACS (Poglitsch et al. 2010, 100 and 160 μm), and H er schel/SPIRE (Griffin et al. 2010, 250, 350, and 500 μm), in addition to the ZFOURGE photometry. We adopt BC03 templates [17], the [18] initial mass function, and the dust extinction of [19]. We fix the redshift in the SED fitting. In the SPIRE bands, we use the 24-μm source positions as the priors and de-blend the SPIRE images in almost the same manner as [20]. We list z spec , z phot , SFR, Mstar , and L IR in electronic supplementary material Table 4. We obtain the 16th−84th percentiles of SFR = 10.0−352 M /yr, log(Mstar ) = 10.2−11.6 M , and log(L IR ) = 11.0−12.7 L  , and the ASAGAO sources generally fall in the massiveend of the main sequence of the SFR−Mstar relation. We confirm that the redshift distribution and the SFR−Mstar relation is well aligned with previous ALMA results for the faint SMGs (e.g., [5, 21, 22]), which suggests that the ASAGAO sources are typical faint SMGs.

54

4 Insterstellar Medium Scale II: Galaxy Morphology

The rest-frame optical properties such as n opt , Re,opt , and the morphological classification are also analyzed with public HST catalogs in GOODS-S. For n opt and Re,opt , the ASAGAO catalog is cross-matched within a radius of 0. 5 with the catalog of [23] where n and Re are evaluated from the deep HST J - and H -band images for H -band selected sources down to H  24.5. Here we use the J -band (H -band) results for sources at z = 0.5−1.5 (1.5−3.5) for n opt and Re,opt of the ASAGAO sources. We then obtain 21 out of 42 ASAGAO sources whose n opt and Re,opt are reliably (flag = 0) measured in [23], where the median values are estimated to be n opt = 1.5 ± 0.01 and Re,opt = 3.6 ± 0.1 kpc. For the morphological classification, we carry out the same cross-matching procedure with the NIR morphology catalog of [24] that complete the visual classification for HST/H -band selected sources (H < 24.5) in GOODS-S. The classification is primarily performed with the H -band, but the J -band image along with V - and I -band ACS images are also used to help the classification. The sources are classified into five categolies of disk, spheroid, irregular, compact, and unclassifiable, and three additional interaction classes of merger, interaction, and non-interaction. The interaction classes are determined by checking the existence of close pairs and tidal features. We find that 29 out of the 42 ASAGAO sources are cross-matched with the catalog of [24] within a radius of 0. 5. Because a larger radius of 2. 5, corresponding to ∼20 kpc at z = 2.5, is adopted in [25] to search major merger systems, we also conduct the cross-matching with the larger search radius. This offers us an additional one ASAGAO source in the crossmatching, which include in the following analysis not to miss major merger pairs. We find that the majority of the ASAGAO sources are classified as the disk galaxies. This is a consistent result with the median Sérsic index result of n opt = 1.5 ± 0.01. We study the L IR dependence of n opt , Re,opt , and the morphological classification in Sect. 7.1.1.2. We note that these rest-frame optical results are limited by the H -band magnitude down to ∼ 24.5. However, 34 out of 42 ASAGAO sources are brighter than 24.5 mag in the H band, suggesting that we obtain typical rest-frame optical properties for the ASGAO sources in these analyses.

4.1.4 Visibility-Based Stacking To examine the average morphology of high-z dusty star-forming galaxies, we carry out the stacking analysis for the ASAGAO sources. For the stacking, we use stacker [26] which is a stacking tool on the uv-visibility plane for interferometric data. Here we focus on 33 ASAGAO sources that are located at z = 1 − 3 to remove a slight difference in the rest-frame FIR wavelengths and potential effects from the different redshifts such as the redshift evolution of Re,FIR (Sect. 3). The 33 ASAGAO sources fall in an L IR range of ∼ 1010 − 1013 L  and have a median value of 1011.98 L  . We adopt spatial centers based on the ASAGAO peak positions that are measured at 1.2-mm wavelength with ALMA Band 6 in Sect. 4.1.2 for the stacking. Figure 4.1 presents the HR map that is obtained from the visibility-based stacking for the 33 ASAGAO sources. The rms noise is decreased down to 8.1 μJy/beam, and

4.1 Data Analysis

55

Fig. 4.1 (Left) HR (natural-weighted) 1. 6×1. 6 image after the visibility-based stacking for the 33 ASAGAO sources at z = 1−3 (a), the best-fit S´ersic profile after beam convolution (b), and the residual image (c). The black contours denote the 4−28σ levels with a 4σ -level step, while the white contours indicate the −2 and −4σ levels. The synthesized beam (0. 21 × 0. 17) is presented at the bottom left in each panel. (Right) Radial profile of the surface brightness in the observed frame. The red circles and dashed line are the observed values and the best-fit S´ersic profile for the stacked image, where the error-bars are evaluated by the random aperture method. The black dashed and solid lines denote the best-fit S´ersic profiles with the fixed values of n = 1 and n = 4, respectively. The magenta curve presents the point source profile that corresponds to the synthesized beam of the ALMA observation. The gray shade indicates the standard deviation of the pixel values in the stacked image. The top axis depicts the radius in the kilo-parsec scale for the case that a source resides at z = 2. The figure is reproduced from Fig. 1 of [27] by permission of AAS

the stacked source is detected at the 29σ level at the peak. This high S/N level meets a requirement to securely measure the S´ersic profile (e.g.,[28, 29]).

4.1.5 nFIR and Re,FIR Measurements We measure n FIR and Re,FIR for the stacked source with galfit [30] which is a profile fitting tool on the image plane. We carry out the galfit task for the 1. 6 × 1. 6 HR map of the stacked source by utilizing a point-spread function (PSF) map from the synthesized beam image in the same size. We set initial parameters as total flux density of 1.0 mJy, Re = 0. 2, n = 1, axis ratio of 1.0, and position angle of 0 deg as initial parameters, where the sky value is fixed at 0. We fix the source center at the stacking center and evaluate the errors by the bootstrap method. We obtain the best-fit values and their errors of Re,FIR = 0. 11 ± 0. 02 and n FIR = 0.96 ± 0.10. To compare the size measurement on the uv-visibility plane, we also perform uvmultifit [31] that is a profile fitting tool on the uv-visibility plane. Because one cannot vary the

56

4 Insterstellar Medium Scale II: Galaxy Morphology

n FIR value in uvmultifit, here we measure the Re,FIR value with a fixed value of n FIR = 1. We obtain the best-fit value of Re,FIR = 0. 10 ± 0. 03, which agrees with the galfit result within the error. In the left panel of Fig. 4.1, we show the stacked, bestfit S´ersic profile, and the residual images. In the right panel of Fig. 4.1, we present the radial profile of the surface brightness of the stacked source, where three S´ersic profiles are presented: the best-fit, n FIR = 1 (fixed), and n FIR = 4 (fixed). These three profiles demonstrates that the best-fit result is close to n FIR ∼ 1 instead of n FIR ∼ 4. We note it is hard from the current results to conclude which n FIR ∼ 1 or n FIR ∼ 4 is more likely at r > 0. 4 (∼3.4 kpc at z = 2). However, because SMGs are reported to have the rest-frame optical effective radius Re,opt of ∼3−4 kpc in previous studies (e.g., [32–34], see also Chap. 3), it is reasonable to interpret that the major part of the stacked source is represented by the best-fit result at r ≤ 0. 4. Interestingly, we find a slight offset between the best-fit and the stacked profiles near the center (r  0. 1). In Sect. 7.1.4, we discuss possible physical mechanisms of this slight offset near the center, while we mainly focus on the major part of the stacked source profile represented by the best-fit n FIR ∼ 1 profile at r ∼ 0. 1−0. 4 in this paper. Given the positional uncertainty of ∼ 0. 06 for the ASAGAO sources evaluated in Sect. 4.1.2, we perform MC simulations to evaluate how significantly the n and Re measurements are affected by the positional uncertainty through the stacking process. In the MC simulations, first we create 33 model sources whose surface density profile is defined by the S´ersic profile with a fixed axis ratio of 0.75. These 33 model sources are then injected with random angles and offsets to a two-dimensional plane whose box and grid sizes are the same as the ALMA image used in our n FIR and Re,FIR measurements. We make the offsets of these model sources follow a Gaussian distribution whose average and standard deviation are 0. 06 and 0. 01, respectively. After the random injections, we estimate the average radial profile from these 33 model sources. Finally, we apply the minimum chi-square method to the average profile and obtain the best-fit parameters of Re and n in the S´ersic profile. We repeat this process 2,000 times, changing the S´ersic parameters in the ranges of Re = 0. 04 − 0. 4 and n = 0.5 − 4.0. In Fig. 4.2, we show the input and output values of n and Re in the MC simulations. In the input n range at  0.6, the output n values are systematically underestimated. This is because the positional uncertainty smoothes peaky radial profiles with large n values through the stacking process (cf. [35]). We also find that the output Re deviates from the input Re , This is also ascribed to the smoothing effect through the stacking, because the smoothed profile provides the output Re larger than the input Re , if the input Re is relatively smaller than the positional uncertainty. In the case the input Re is relatively large on the other hand, the smoothed profile with small n is challenging to fit the outskirts. This makes the output Re slightly smaller than the input Re . The best-fit results of n FIR and Re,FIR are corrected by these MC simulation results, and we obtain n = 1.2 ± 0.2 and Re,FIR = 0. 12 ± 0. 03. This Re,FIR value corresponds to 1.0 ± 0.2 kpc with the median redshift of z = 2.00 for our ASAGAO sources used for the stacking.

4.1 Data Analysis Fig. 4.2 MC simulation results of the relationship between input and output of S´ersic index n (top) and effective radius Re (bottom). The black dots denote 1,000 model sources. In the MC simulation for the n (Re ) measurement, the input Re (n) values are fixed at Re = 0. 15 (n = 1.0). The axis-ratio is also fixed at 0.75. The red circles and the error-bars present the median and the 16–84th percentiles of the bins. The figure is reproduced from Fig. 2 of [27] by permission of AAS

57

58

4 Insterstellar Medium Scale II: Galaxy Morphology

Table 4.1 Additional source catalog from archive ID R.A. Dec. S/N (J2000) (1) (2) (3) 105 254 255 256 260 580 581 592 648 649 651 653

34.568703 53.203598 53.389057 53.030369 52.937637 34.418846 34.421337 34.422413 53.118809 53.020367 53.137592 53.148853

−4.919106 −27.52034 −27.99159 −27.85580 −27.57688 −5.219619 −5.220879 −5.181027 −27.78287 −27.77991 −27.70021 −27.82118

15.3 19.8 26.0 43.3 26.7 28.4 16.2 21.4 49.0 21.0 16.4 39.9

(4)

log(L IR ) (L  ) (5)

(6)

2.30 2.72 2.67 2.12 2.33 2.87 2.90 2.72 2.31 2.01 2.45 2.58

12.90+0.06 −0.07 12.70+0.14 −0.57 12.64+0.23 −0.33 12.60+0.20 −0.26 12.31+0.35 −0.13 12.48+0.03 −0.03 12.33+0.05 −0.06 12.51+0.04 −0.05 12.84+0.10 −0.10 12.78+0.04 −0.05 12.45+0.10 −0.10 12.83+0.10 −0.10

1 2 2 2 2 1 1 1 3 3 3 3

z phot

Ref.

(1) ALMA source ID presented in [15] (2) ALMA source center (3) ALMA Peak S/N (4) Photometric redshift (5) IR luminosity (6) Reference of the L IR measurement [15, 17–19, 36, 41]

4.1.6 Additional Sample from Archive To comprehensively cover a wide range of L IR for high-z dusty star-forming galaxies, we compile an additional sample of bright SMGs in archive. We use the ALMA source catalog made from the ALMA-DUST dataset (Sect. 3), where 1034 ALMA sources (S/N≥5) are identified in ALMA Band 6 and 7 maps. To obtain as reliable results as the stacked ASAGAO sources, we select 12 individual ALMA sources that are observed with a synthesized beam size (natural-weighted map) of < 0. 3 and detected with S/N > 15 at z =1–3. In this additional sample, we adopt L IR derived in previous ALMA studies that perform the SED fitting with from optical to FIR bands (e.g., [36]). Otherwise, we estimate L IR in the same manner as Sect. 3, assuming the modified blackbody with the spectral index of β = 1.8 (e.g.,[37, 38]) and the dust temperature of Td = 35 K (e.g., [39, 40]). The additional sample takes the L IR range of 1012.3 − 1012.9 L  . The properties of the additional sample is listed in Table 4.1. Note that we do not perform any stacking for the additional sample, because we select the ALMA sources in the additional sample with high S/N levels (S/N > 15), We measure the n FIR and Re,FIR values individually with galfit in the same manner as the stacked source, where no corrections due to the stacking (see Sect. 4.1.5) are applied. We obtain the median values of n FIR and Re,FIR of n FIR = 1.1 ± 0.3 and Re,FIR = 1.3 ± 0.8 kpc, respectively.

4.2 Result

59

Fig. 4.3 Rest-frame FIR properties of n FIR (left) and Re,FIR (right) as a function of L IR (Fig. 3 of [27], reproduced by permission of AAS). The red filled circle is obtained from the visibility-based stacking for the 33 ASAGAO sources at z = 1−3. The L IR error-bar represents the 16th−84th percentiles of the L IR distribution for the ASAGAO sources, while the n FIR and Re,FIR error-bars are evaluated by the bootstrap method and the MC simulations. The red open squares present the additional sample of the 12 individual bright ALMA sources at z = 1−3. The red filled square indicates the median value of the 12 individual ALMA sources, where the error-bars denote the16th−84th percentiles of the distribution. The black open triangles and squares are the previous ALMA results of [1, 15], respectively. In the right panel, the black open squares are estimated by fixing n FIR = 1 that we do not present in the left panel. The red shaded regions are the best-estimates of the constant n FIR (left) and the Re,FIR −L IR relation (right). The constant n FIR is estimated from the stacked ASAGAO and the median value of the 12 individual ALMA sources, while the best-fit Re,FIR −L IR relation is obtained in Sect. 3 [15]

4.2 Result 4.2.1 Size and Morphology in FIR Figure 4.3 presents our stacking results of n FIR (left panel) and Re,FIR (right panel) for the ASAGAO as a function of L IR with the red circle. To test the L IR dependence of n FIR and Re,FIR , Fig. 4.3 also shows the individual and median values of the additional sample of the 12 ALMA sources at z = 1 − 3 with the red open and filled squares. In the left panel of Fig. 4.3, the black open triangles show bright SMGs identified in previous ALMA high-resolution 870-μm observations, called the An ALMA Survey of Submillimeter Galaxies in the Extended Chandra Deep Field South (ALESS; e.g., [1]) for comparison. Our stacking result explores the n FIR measurement down to L IR ∼ 1011 L  . The individual ALMA sources show the median n FIR value of n FIR = 1.1 ± 0.3, which is consistent with the stacking result from the ASAGAO sources of n FIR = 1.2 ± 0.2 (see Sect. 4.1.5). We find that our n FIR measurements of stacking and individual results both fall in the n FIR range consistent with that of the

60

4 Insterstellar Medium Scale II: Galaxy Morphology

ALESS results of [1] with a median n FIR = 0.9 ± 0.2. These results imply that the exponential-disk profiles with n FIR ∼ 1 is a representative surface brightness profile among dusty star-forming galaxies at z = 1 − 3 without a significant dependence on L IR . Assuming a constant value, we obtain the best-fit n FIR value of 1.2 ± 0.2. This confirms that it is validate the assumption of fixing n FIR = 1 in the Re,FIR measurements in Sect. 3. In the right panel of Fig. 4.3, the black open squares show the individual Re,FIR measurement results independently obtained in Sect. 3. We find that our Re,FIR measurements of stacking and individual results both agree with the independent results from Sect. 3. The individual results from the additional sample shows the median value of Re,FIR = 1.3 ± 0.8 kpc. This is consistent with the median value of the stacking result of Re,FIR = 1.0 ± 0.2 kpc within the 1σ error, while this may suggest a positive correlation between L IR and and Re,FIR . In fact, these two median values (Sect. 3). are consistent with the FIR size−luminosity relation of Re,FIR ∝ L 0.28±0.07 IR However, we note that we do not quantitatively investigate this potential correlation in FIR size−luminosity relation from these results due to the requirement on the intensive analysis for the incompleteness correction (e.g., [15, 42]), which is beyond the scope in this chapter.

4.2.2 Size and Morphology in Optical We also verify the L IR dependence of n opt and Re,opt . Figure 4.4 shows the n opt (left panel) and Re,opt (right panel) values for the ASAGAO sources at z = 1−3 as a function of L IR . The individual and median values are presented with the blue open and filled circles, respectively. For comparison, we also present previous individual and median results based on follow-up HST/H -band observations for the ALESS sources [34] with the black open and filled squares, respectively. In the right side of each panel, we present histograms of the ASAGAO and ALESS sources. In the histograms of Fig. 4.4, we find that the distributions of both n opt and Re,opt are similar between the ASAGAO sources and the ALESS sources. The KolmogrovSmirnov test (KS test) is used to examine whether these distributions of n opt and Re,opt are different statistically, and we find that neither of n opt and Re,opt can rule out the possibility that their histograms are originated from the same parent sample. We also carry out the Spearman’s rank test for the n opt and Re,opt distributions from our ASAGAO and the ALESS sources. The Spearman’s rank test shows that n opt and Re,opt have the correlation with L IR at the ∼ 2σ and ∼ 0σ levels, respectively, supporting that neither of n opt and Re,opt depend significantly on L IR . Although n opt may have a week anti-correlation with L IR , the statistical significance level is poor with the large scatter with the current dataset. We thus fit constant values to both Re,opt and n opt of the ALESS and ASAGAO sources, where the best-fit values are estimated to be Re,opt = 3.2 ± 0.6 kpc and n opt = 0.9 ± 0.3.

4.2 Result

61

Fig. 4.4 Rest-frame optical properties of n opt (left) and Re,opt (right) as a function of L IR (Fig. 4 of [27], reproduced by permission of AAS). The blue open circles are the ASAGAO sources at z = 1 − 3 whose n opt and Re,opt values are reliably (flag = 0) measured in [23]. The blue filled circles are the median values of the ASAGAO sources. The black open squares show the previous ALMA results of [34] for the bright SMGs identified in the ALESS survey. The black filled squares are the median values of the ALESS results. The error-bars of the blue filled circles and black squares denote the16th−84th percentiles of the distribution. The blue shaded regions are our best-estimates of constant n opt (left) and Re,opt (right) that are derived from the ASAGAO and ALESS sources. The red shaded regions are the same assignment as in Fig. 4.3. The blue and black histograms on the right-side of both panels denote the ASAGAO and ALESS sources, respectively. The blue and black arrows present the median values of the ASAGAO and ALESS sources

4.2.3 Comparison Between FIR and Optical We compare the n and Re measurements in the rest-frame FIR with those in optical wavelengths. In Fig. 4.4, we show our n and Re measurements in the rest-frame FIR and optical wavelengths with the red and blue shaded regions, respectively. In the left panel of Fig. 4.4, we find that the n value is mostly constant and unchanged between these wavelengths with n FIR ∼ n opt ∼ 1. In the right panel of Fig. 4.4 on the other hand, we find that Re,opt is generally larger than Re,FIR . This is consistent with previous ALMA results that report that dusty star forming activity is taking place in a compact regions compare to the stellar distribution (e.g., [1, 2, 14, 15, 41]). We note that the results of n FIR ∼ n opt ∼ 1 and Re,FIR < Re,opt does not necessary indicate the existence of a compact dusty disk with dynamical rotations. However, the existence of inner disk-like structures such as bars, oval disk, and spirals are reported in local disk galaxies, which is thought to play an important role in building up the central bulge (e.g.,[43]). This implies that there is the inner disk-like structures where compact dusty star-formation is taking place, which contributes to the central bulge formation. In fact, not only in the local galaxies, such compact rotating molecular gas disks, compared to the rest-frame optical sizes, are reported in distant massive starforming galaxies at z = 2.5 which contains the compact dusty star-forming regions

62

4 Insterstellar Medium Scale II: Galaxy Morphology

[44]. These results give arise a potential common picture for the high-z dusty starforming galaxies that dusty star formation occurs in a compact dusty disk embedded in a larger stellar disk.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44.

Hodge JA, Swinbank AM, Simpson JM et al (2016) ApJ 833:103 Tadaki K-I, Genzel R, Kodama T et al (2017) ApJ 834:135 Straatman CMS, Spitler LR, Quadri RF et al (2016) ApJ 830:51 Rujopakarn W, Dunlop JS, Rieke GH et al (2016) ApJ 833:12 Dunlop JS, McLure RJ, Biggs AD et al (2017) MNRAS 466:861 Gaia Collaboration, Brown AGA, Vallenari A et al (2016) A&A 595:A2 da Cunha E, Charlot S, Elbaz D (2008) MNRAS 388:1595 Downes AJB, Peacock JA, Savage A, Carrie DR (1986) MNRAS 218:31 Hatsukade B, Kohno K, Yamaguchi Y et al (2018) PASJ 70:105 Hancock PJ, Murphy T, Gaensler BM, Hopkins A, Curran JR (2012) MNRAS 422:1812 Walter F, Decarli R, Aravena M et al (2016) ApJ 833:67 Franco M, Elbaz D, Béthermin M et al (2018) A&A 620:A152 Franco M, Elbaz D, Béthermin M et al (2015) ApJ 810:133 Franco M, Elbaz D, Béthermin M et al (2015) ApJ 799:81 Fujimoto S, Ouchi M, Shibuya T, Nagai H (2017) ApJ 850:1 Yamaguchi Y, Kohno K, Hatsukade B et al (2020) PASJ 72:69 Bruzual G, Charlot S (2003) MNRAS 344:1000 Chabrier G (2003) PASP 115:763 Charlot S, Fall SM (2000) ApJ 539:718 Liu D, Daddi E, Dickinson M et al (2018) ApJ 853:172 Hatsukade B, Ohta K, Yabe K et al (2015) ApJ 810:91 Yamaguchi Y, Tamura Y, Kohno K et al (2016) PASJ 68:82 van der Wel A, Franx M, van Dokkum PG et al (2014) ApJ 788:28 Kartaltepe JS, Mozena M, Kocevski D et al (2015) ApJS 221:11 Le Fèvre O, Abraham R, Lilly SJ et al (2000) MNRAS 311:565 Lindroos L, Knudsen KK, Vlemmings W, Conway J, Martí-Vidal I (2015) MNRAS 446:3502 Fujimoto S, Ouchi M, Kohno K et al (2018) ApJ 861:7 van der Wel A, Bell EF, Häussler B et al (2012) ApJS 203:24 Ono Y, Ouchi M, Curtis-Lake E et al (2013) ApJ 777:155 Peng CY, Ho LC, Impey CD, Rix H-W (2010) AJ 139:2097 Martí-Vidal I, Vlemmings WHT, Muller S, Casey S (2014) A&A 563:A136 Targett TA, Dunlop JS, McLure RJ et al (2011) MNRAS 412:295 Targett TA, Dunlop JS, Cirasuolo M et al (2013) MNRAS 432:2012 Chen C-C, Smail I, Swinbank AM et al (2015) ApJ 799:194 Paulino-Afonso A, Sobral D, Ribeiro B et al (2018) MNRAS, arXiv:1709.04470 da Cunha E, Walter F, Smail IR et al (2015) ApJ 806:110 Chapin EL, Pope A, Scott D et al (2009) MNRAS 398:1793 Planck Collaboration, Abergel A, Ade PAR et al (2011) A&A 536:A21 Kovács A, Chapman SC, Dowell CD et al (2006) ApJ 650:592 Coppin K, Halpern M, Scott D et al (2008) MNRAS 384:1597 Barro G, Kriek M, Pérez-González PG et al (2016) ApJ 827:L32 Kawamata R, Ishigaki M, Shimasaku K et al (2018) ApJ 855:4 Kormendy J, Kennicutt RC Jr (2004) ARA&A 42:603 Tadaki K-I, Kodama T, Nelson EJ et al (2017) ApJ 841:L25

Chapter 5

Circumgalactic Medium Scale: Metal-Enriched Gas Halo

5.1 Data Analysis In this chapter, we use our original and archival deep ALMA Band 6/7 dataset of ALMA-z6CII (Table 2.1).

5.1.1 3D Position in ALMA Cube Before carrying out stacking for the rest-frame FIR continuum and the [C ii] line, we estimate source centroids in the ALMA 3-dimensional (3D) data cubes for the 18 [C ii] line sources through the following six steps: (1) In the velocity range of ∼100–900 km s−1 , producing fiducial [C ii] velocity-integrated maps where S/N of the [C ii] line emission is maxmized. (2) Spatially smoothing the fiducial [C ii] velocity-integrated maps with a uv-taper of 0. 6 and measuring positional centroids based on the peak pixel positions (pixel scale = 0. 01). (3) Creating [C ii] spectra at the fiducial source centroids with an aperture diameter of 1. 2. (4) Fitting a single Gaussian to the [C ii] spectra and evaluating FWHMs and peak frequencies of the [C ii] line emission. (5) Re-producing velocity-integrated maps with the velocity ranges of 2× the FWHMs. (6) In the same manner as step (2), measuring final spatial centroids in the new velocity-integrated map. Note that we use the uv-tapered map instead of the naturally-weighted map in steps of (2) and (6) because positional measurements in smoothed maps are reported to have smaller uncertainties than in the naturally-weighted maps through Monte-Carlo simulations in the uv-visibility

Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/978-981-16-4979-0_5.

© Springer Nature Singapore Pte Ltd. 2021 S. Fujimoto, Demographics of the Cold Universe with ALMA, Springer Theses, https://doi.org/10.1007/978-981-16-4979-0_5

63

64

5 Circumgalactic Medium Scale: Metal-Enriched Gas Halo

plane [1]. In Table 2.4, we list the final source centroids in the ALMA 3D data cubes i.e., source redshifts and spatial centers measured by the above procedure.

5.1.2 ALMA Visibility-Based Stacking The visibility-based stacking is performed via the following procedure. First, the visibility data is split into the rest-frame FIR continuum and the [C ii] line data sets. For the rest-frame FIR continuum data set, we extract the line-free channels with a velocity range of ≥ 2×FWHM from the original visibility. For the [C ii] line data set, we extract the [C ii] line channels across a velocity range of ± 50 km s−1 from the original visibility. This is because close companions might be contaminated in a wider velocity range [2, 3]. Second, we re-write the coordinate of the visibility data sets as “00:00:00.00 00:00:00.0” at the source center with stacker [4]. Third, we combine the visibility data sets for the stacking sample of ALMA-z6CII with the concat task. Fourth, based on the scatter of visibilities, including the effects of channel width, integration time, and system temperature, we re-calculate the data weights with the statwt task for the combined visibility data sets Finally, we obtain stacked data sets of the rest-frame FIR continuum and the [C ii] line. The central frequency in the stacked data set for the [C ii] line is 271.167 GHz, which corresponds to the [C ii] redshift at z = 6.01. Given that the redshift of z = 6.01 as the weighted average value for our stacking sample of ALMA-z6CII, we adopt the angular scale of 1 = 5.7 kpc in the following analyses in this Chapter. We note that we use the H -band peak positions as the common stacking center for the ALMA-HST sample, while we confirm that our stacking results are not affected from the difference in the choice of the common stacking center (see Sect. 5.1.3). Figure 5.1 shows the uv-visibility coverage of the ALMA-ALL sample after the visibility-based stacking. For comparison, Fig. 5.1 also presents another uv-visibility coverage for an individual data set (i.e., before stacking) is also presented. The stacked data of the ALMA-ALL sample shows the sufficient sampling in the uv-visibility coverage, especially at the short baselines (500 kλ), which enables us to recover the diffuse and extended emission. In Fig. 5.2, we present the natural-weighted maps after the visibility-based stacking for the rest-frame FIR continuum and the [Cii] line data sets of the ALMA-ALL and ALMA-HST samples. The r.m.s noise level in the stacked map is decreased down to 4.1 (8.3) µJy/beam with the synthesized beam size of 0. 43 × 0. 36 (0. 74 × 0.58), where the stacked source is detected at the 21σ (20σ ) and 10σ (8σ ) significance levels for the [Cii] line and the rest-frame FIR continuum, respectively, for the ALMA-ALL (ALMA-HST) sample. We find that the [C ii] line emission is spatially resolved and extended over a radius of ∼10 kpc scale. Based on the random-aperture method, the spatially resolved [C ii] line emission is detected at the 9.3 σ level in the aperture radius of 10 kpc even after masking the emission in a central area up to 2×FWHM of the ALMA synthesized beam. Because the extended structure is not always modeled

5.1 Data Analysis

65

Fig. 5.1 uv-visibility coverage for individual and stacked data (Fig. 2 of [5], reproduced by permission of AAS). For the individual data, we present Hz3 data as an example. For the stacked data, the uv-visibility coverage less than 500 kλ is well sampled in circular symmetrically, which enables us to investigate the diffuse, extended structures

by the clean algorithm properly, we adopt the dirty maps throughout this chapter for both the rest-frame FIR continuum and the [C ii] line. Before we study physical properties of the extended [C ii] line emission, we test whether any potential systematics cause the extended morphology through the stacking. In Fig. 5.3, we summarize various test results with the radial surface brightness profile of the stacked [C ii] line (red squares). First, we compare the radial surface brightness profiles between the individual (i.e., before stacking) and the stacked results. The top left panel of Fig. 5.3 shows the individual results for several [C ii] line sources (black solid lines). To investigate the reliable individual results including the diffuse and extended emission, here we present only those individual results whose lines are detected at S/N  10 with an ALMA beam size of  0. 8 to recover the diffuse and extended emission. We find that the stacked results are consistent within the scatter with these individual results. This suggests that we obtain a faithful representative of the 18 [C ii] line sources from our ALMA stacking result. Second, we address potential effects from the sample variance. We derive new 18 [C ii] radial profiles by making the 18 newly stacked data sets composed of 17 [C ii] line sources, i.e., in each newly stacked data we remove one source from the full sample. The top right panel of Fig. 5.3 shows the 16–84 percentiles of these 18 radial profiles with the red shaded area. The red shaded area also extends up to the radius of ∼ 10 kpc scale. This suggests that the existence of the extended [C ii] line emission is not affected by the sample variance Third, we further verify if any specific data properties cause the extended [C ii] line structure. Here we again derive newly stacked [C ii] line radial profiles by removing the sources that are (I) reported to have companions (Hz2, Hz6, Hz8, and WMH5; [2, 3]), and (II) taken with the lowest resolutions (BDF2203, NTTDF6345, and WMH13). The top right panel of Fig. 5.3 presents the [C ii] line radial profiles in the newly stacked data. We find that the extended structure found in

66

5 Circumgalactic Medium Scale: Metal-Enriched Gas Halo

Fig. 5.2 Natural-weighted 4 × 4 field image after the visibility-based stacking of the [Cii] line and the dust continuum for the ALMA-ALL (left) and ALMA-HST (right) samples (Fig. 3 of [5], √ reproduced by permission of AAS). The red and green contours denote the 2, 2 2, 4, ... ×σ levels of the √ [Cii] line and the dust continuum emission, while the white contours indicate the −2σ and −2 2σ levels. The synthesized beams are presented at the bottom left in each panel

the original stacked data is reproduced in the newly stacked [C ii] line radial profiles. This indicates that neither the contamination of the companions nor the bias to the low-resolution data cause the extended [C ii] line structure. Fourth, we investigate the surface brightness dimming effect among the redshift in our sample. We divide the 18 [C ii] line sources into two subsamples: high-redshift (z > 6) and low (z < 6) samples, and derive the newly stacked [C ii] line radial profiles. The bottom left panel of Fig. 5.3 shows the newly stacked [C ii] line radial profiles in these two subsamples. We find that the extended structure found in the original stacked data is again reproduced in the [C ii] line radial profiles in both subsamples. This shows that the

5.1 Data Analysis

67

Fig. 5.3 Radial surface brightness profile of the [C ii] line (red filled squares) and dust-continuum (green filled squares) emission for the ALMA-ALL sample (Fig. 4 of [5], reproduced by permission of AAS). Top Left: The black solid curves denote the individual results from five [C ii] line sources whose [C ii] lines are detected with high S/N levels and with the ALMA beam sizes of  0. 8 to recover the diffuse, extended structures. The gray shades indicate the error range of the individual results. The black dashed curve presents the synthesized ALMA beam in the ALMA-ALL sample. Top Right: The red shade shows the 16–84 percentile of the sample variance (see text). The open symbols indicate the re-stacked results without the [C ii] line sources that are (I) taken with the lowest resolutions (BDF2203, NTTDF6345, and WMH13; upward triangle), and (II) reported to have companions (WMH5, Hz2, Hz6, and Hz8; downward triangle). Bottom Left: The restacked results for the low- (z < 6; leftward triangle) and high- (z > 6; rightward triangle) redshift subsamples among the 18 [C ii] line sources. Bottom Right: whose peak S/N ratio is reduced down to the level comparable to the dust continuum map. All radial profiles are normalized to the peak value of the [C ii] line

68

5 Circumgalactic Medium Scale: Metal-Enriched Gas Halo

surface brightness dimming effect does not affect our stacking results, neither. Fifth, we compare the structures of the rest-frame FIR continuum and the [C ii] line in the same significance level by enhancing the noise level in the stacked [C ii] line map. To do so, we create a random noise map and smooth it with the ALMA beam. We then make the stacked [C ii] line map combined with the noise map. Changing the noise levels, we obtain the noise-enhanced [C ii] line map where the peak SNR is decreased down to the same level as the rest-frame FIR continuum. We create the 50 [C ii] line maps whose noise level is enhanced with above procedures. In the bottom right panel of Fig. 5.3, the error bars of the red squares denote the 16–84th percentile of the [C ii] radial profile in the noise-enhanced maps. We find that the [C ii] line profile still exceeds more than the rest-frame FIR continuum in the noise-enhanced maps. This suggests that the difference in the dynamic range does not affect the different structures between the rest-frame FIR continuum and the [C ii] line.

5.1.3 HST/H-Band Stacking For the ALMA-HST sample, we have also carried out image-based stacking for their deep HST H -band maps from the archive. Because we identify several objects in the H-band maps around some of the [C ii] line sources, we perform the following four steps before stacking. (1) Cutting out 8 × 8 stamps of the H -band maps around the [C ii] line peak. (2) Re-gridding the pixels to make their scales down to 0. 01. This is the same scale as our ALMA images. (3) Cross-matching photometric redshift catalogs [6, 7] and the [C ii] line source positions with a radius of 2. 0, and identifying low-z interlopers around the [C ii] line peaks. 4) Fitting S´ersic profiles [8] with galfit [9] to those low-z interlopers and removing them from the H -band maps. We then obtain an average stacked map. Here we adopt the stacking weight based on the noise levels of the ALMA [C ii] line maps, because the [C ii] line stacking based on the uv-visibility plane is weighted by the visibility scatter, which is comparable to the noise levels on the ALMA [C ii] line maps. In the H -band stacking, we smooth the H -band maps with the uv-tapered ALMA beams in a consistent manner with the [C ii] line stacking, where we adopt the peak positions of the smoothed H -band maps as the stacking centroids. The panel (f) of Fig. 5.4 shows the H -band map of the ALMA-HST sample after the stacking. To compare the size and morphology in the ALMA and the HST maps directly, we convolve the HST map where a PSF resembles the one of the ALMA image. We use galfit to obtain a kernel which achieves the convolution from the H band PSF to the ALMA beam. We adopt a sum of three independent S´ersic profiles that have a common center to model the kernel. Figure 5.4 summarizes a schematic overview: the best-fit kernel converts the H band PSF to the ALMA beam. First the HST PSF (panel a) is convolved with the best-fit kernel (panel b), and we obtain the mock ALMA beam (panel c). Then the actual ALMA beam (panel d) is subtracted from the mock ALMA beam, which produce the residual map (panel e). In the residual map, we find that the difference

5.1 Data Analysis

69

Fig. 5.4 Schematic overview to obtain the mock HST/H-band image whose spatial resolution is matched to the stacked ALMA image for the ALMA-HST sample: a HST/H−band PSF, b the best-fit kernel composed by three S´ersic profiles obtained with galfit, c the best-fit ALMA beam model obtained with galfit, d the ALMA synthesized beam in the stacked ALMA image for the ALMA-HST sample, e the residual between c and d, f the stacked HST/H−band image for the ALMA-HST sample, and g the stacked HST/H−band image obtained by convolving f with b. The red contours present 3, 5, 10, 20, 30, 40, and 50% of the PSF or beam response. The blue contours √ denote the 2, 2 2, 4, ... ×σ levels of the rest-frame UV continuum emission. The cutout sizes are 2 × 2 and 4 × 4 for the panels of a)−e and f−g, respectively. The figure is reproduced from Fig. 2 of [5] by permission of AAS

between the mock and actual ALMA beams are less than ∼ 1.8% within a radius of 1. 0 on the residual map. This indicating that the ALMA beams is well reproduced from the best-fit kernel and the H -band PSF. We finally apply the best-fit kernel to the stacked H -band map (panel f), where we obtain the mock H -band map whose PSF resembles that of the stacked ALMA map.

70

5 Circumgalactic Medium Scale: Metal-Enriched Gas Halo

5.2 Result 5.2.1 Discovery of [CII] Halo Figure 5.5 shows the surface brightness radial profiles of rest-frame FIR, UV continuum, and the [Cii] line derived from the visibility-based stacking for the ALMA-ALL (squares) and the ALMA-HST (circles) samples. For a fair comparison by exploiting the common stacking center, the ALMA-HST results are obtained by re-performing the ALMA visibility-based stacking with the H -band image peak positions. On the other hand, we cannot re-perform the ALMA visibility-based stacking by using the common stacking center with the ALMA-ALL sample, because of the lack of the HST/H -band images. In Fig. 5.5, the radial profiles of the ALMA-HST and ALMA-ALL show an excellent agreement in both [C ii] line and rest-FIR continuum. We find that the [C ii] line emission is extended over a radius of ∼10-kpc scale, in contrast to the rest-frame FIR

Fig. 5.5 Radial surface brightness profiles for the ALMA-HST (circles) and ALMA-ALL (squares) samples (Fig. 6 of [5], reproduced by permission of AAS). The radial values are estimated by the median of each annulus. The red, green, and blue symbols denote the [C ii] line, rest-frame FIR, and rest-frame UV continuum emission. The rest-frame UV continuum profile is directly derived from the mock HST/H -band image whose resolution is matched to that of the ALMA image. The black dashed and solid curves denote the ALMA synthesized beams in the stacked images of the ALMA-HST and ALMA-ALL samples, respectively. All radial profiles are normalized to the peak value of the [Cii] line. The green and red symbols are slightly shifted along the x-axis for clarity

5.2 Result

71

and UV continuum. Because previous HST studies report that normal star-forming galaxies at z ∼ 6 are compact with the typical effective radius less than 1 kpc (e.g., [10]), the ∼10-kpc scale structure corresponds to the surrounding circumgalactic medium (CGM) beyond the galaxy scale, and thus we dubbed it as the [C ii] halo. The presence of the [C ii] halo indicate that the ionized carbon exists and emit the [C ii] line in the CGM areas even without the stellar continuum. In Sect. 7.2, we discuss potential physical origins of this CGM-scale [Cii] halo. We also find that the radial profiles of the rest-frame UV and FIR continuum agree with each other within the 1 σ errors. We note that the rest-frame UV continuum is slightly extended more than the ALMA beam and thus spatially resolved, whereas the rest-frame FIR continuum is likely to follow the ALMA beam and thus not spatially resolved. This suggests that the rest-frame FIR continuum is intrinsically more compact than the rest-frame UV continuum, which is well aligned with the recent ALMA results that high-z star-forming galaxies generally have more compact sizes in the rest-frame FIR wavelength than in the rest-frame UV and optical wavelengths (e.g., [1, 11–14]).

5.2.2 Effect of [CII]–UV Offset In recent ALMA studies, a spatial offset has been reported between the [Cii]-line and the rest-frame UV emitting regions in a star-forming galaxy at z ∼ 7 (e.g., [15]). In order to examine the potential effect of this spatial offset in our stacking analysis, we compare the ALMA and HST stacking results with the ALMA-HST sample by adopting two different stacking centers: ALMA [C ii] line and HST/H band continuum peak positions. In Fig. 5.6, we show the stacking results derived with the stacking centers based of the ALMA [C ii] line and the HST/H -band continuum peak positions with the cross and circles symbols. We find that both the rest-frame FIR and UV continuum profiles are still more compact than the [C ii] line profile in both cases. This suggests that the [C ii] line is emitted from even out side of the stellar distribution traced by the rest-frame UV and FIR continuum, and shows that the [Cii]–UV offsets does not cause of the extended structure of the [C ii] line.

5.2.3 Radial Ratio of L [CII] to Total SFR We test whether faint satellite galaxies are attributed to the extended [C ii] line structure. For the test, we evaluate radial ratios of the [C ii] line luminosity L [CII] to SFR that is obtained from the rest-frame UV and FIR continuum. Because the ALMAHST and ALMA-ALL results are consistent with each other (Fig. 5.5), we adopt the stacked result of the rest-frame UV continuum from the ALMA-HST sample, while

72

5 Circumgalactic Medium Scale: Metal-Enriched Gas Halo

Fig. 5.6 Radial surface brightness profiles for the ALMA-HST sample derived with different stacking centers (Fig. 7 of [5], reproduced by permission of AAS). The red, green, and blue symbols denote the [C ii] line, rest-frame FIR, and rest-frame UV continuum emission. The color crosses and circles are the stacking results based on the stacking centers of the [C ii] line and the HST/H band peak positions, respectively. The rest-frame UV continuum profile is directly derived from the mock HST/H -band image whose resolution is matched to the ALMA image. The black solid curve denotes the ALMA synthesized beam. All radial profiles are normalized to the peak value of the [C ii] line. The green and red symbols are slightly shifted along the x-axis for clarity

we use the stacked results of the [C ii] line and rest-frame FIR continuum from the ALMA-ALL sample to minimize the errors in the following estimates. We first evaluate the radial profile of L [CII] . For the ALMA-ALL sample, the weighted-average FWHM of the [C ii] line width and the redshift are estimated to be 270 km s−1 and z = 6.01, respectively. Since the stacked [C ii] line map is produced in the velocity-integrated width of 100 km s−1 , we calculate the luminosity difference between the velocity-integrated values with the ranges of 100 km s−1 and 270 km s−1 in a single Gaussian line profile and apply the difference to the stacked values to recover L [CII] in the velocity-integrated width of FWHM. We second estimate the radial SFR value. We derive the obscured (SFRIR ), un-obscured (SFRUV ), and total SFR (SFRtotal ) with the equations in [18] of SFRIR [M yr −1 ] = 3.88 × 10−10 L IR [erg s−1 ], −1

−10

−1

SFRUV [M yr ] = 4.42 × 10 L UV [erg s ], SFRtotal = SFRIR + SFRUV ,

(5.1) (5.2) (5.3)

5.2 Result

73

Fig. 5.7 Left: Radial profiles of ΣSFR (left axis) and Σ L [CII] (right axis). The blue, green, and black circles indicate ΣSFRUV , ΣSFRIR , and ΣSFRtotal , respectively, based on Equations (1)–(3). The −1 red circles denote Σ L [CII] normalized to ΣSFR with the ratio of L [CII] /SFRtotal = 107 [L  M yr] that is the average value in the local star-forming galaxies [16]. Right: Ratio of L [CII] /SFRtotal as a function of SFRtotal . The filled (open) red circles indicate our stacking results at a radius of ≥ 4 kpc (< 4 kpc). The black dots denote local dwarf galaxy results in the global scale reported in [16]. The green crosses present the extended [C ii] line emission calculated from the local LIRG results in [17]. We assume the area of 1 kpc2 for the L [CII] and SFRtotal estimates in our stacking and the local LIRGs results. At the radius of > 7 kpc, the SFRtotal value in our stacking results becomes negative due to the noise fluctuations on the low surface brightness of the rest-frame UV and FIR continuum emission, where we evaluate the lower limit of the ratio by using the upper limit of SFRtotal . The figure is reproduced from Fig. 8 of [5] by permission of AAS

where L IR is the integrated IR flux density estimated by a typical modified blackbody with spectral index βd of 1.8 [19]and dust temperature Td of 35 K [20], respectively. L UV is the rest-frame UV luminosity that is estimated from the observed luminosity in the HST H -band. Finally, we obtain the radial ratio of L [CII] /SFRtotal by dividing the radial L [CII] values by the radial SFR values. In Fig. 5.7, the left panel shows the surface brightness radial profiles of L [CII] (Σ L [CII] ) and SFRtotal (ΣSFR ), and the right panels presents radial ratio of L [CII] /SFRtotal as a function of SFRtotal . In the right panel, the red filled and open circles indicate our stacking results in the outer (radius of ≥ 4 kpc) and central (< 4 kpc) regions, respectively. For comparison, we also show L [CII] /SFRtotal ratios of the local dwarf galaxies measure in the whole galaxy scale [16] in the right panel. We find that there is a negative trend between L [CII] /SFRtotal and SFRtotal , which might be caused by the [C ii] deficit in the high ionization state in the ISM around the intensely star-forming regions (e.g., [21–24]). We identify that the highest ratios (> 108 L  M−1 yr) fall at the outer regions and not compatible with typical values of the local dwarf galaxies (< 3 × 107 L  M−1 yr; black dots in the figure). These results suggest that the [C ii] halo is not likely driven by satellite galaxies. We note that there are other types of high-z galaxies with SFRtotal > 10 M such as quasarhost, submillimeter, and star-forming galaxies show ratios of 106 ∼ 107 L  M−1 yr

74

5 Circumgalactic Medium Scale: Metal-Enriched Gas Halo

(e.g., [25–28]). However these high-z galaxies are still difficult to explain the highest ratios in our stacking results (> 108 L  M−1 yr). In the local Universe, [17] report that the [C ii] line emission is spread over ∼ 1 − 10 kpc scale around luminous infrared galaxies (LIRGs), and we also compare our results with this spatially resolved emission around the local LIRGs. In the right panel of Fig. 5.7, the L [CII] /SFRtotal ratios of the extended emission around local LIRGs are presented with the green crosses. The ratios around the local LIRGs (< 3 × 107 L  M−1 yr) are yet lower than those of our stacking results (> 108 L  M−1 yr). Therefore, the [C ii] halo at z ∼ 6 is different from the extended [C ii] line emission observed in the local Universe, which imply a potential redshift evolution of the [C ii] halo. In Sect. 7.2.1, we discuss possible origins of the [C ii] halo.

5.2.4 Scale Length of [CII] Halo Based on two-component fitting with galfit, we further characterize detail surface brightness profiles of the [Cii] line. We assume that the two components are attributed to the central and the halo components. First for the central component, we model the surface brightness profile with the S´ersic profile. Here we fix the parameters in the S´ersic profile that are estimated from the stacked HST/H -band image (Fig. 5.4f. The stacked H -band image provides us with the best-fit S´ersic index n = 1.2 ± 0.01 and effective radius re of re = 1.1 ± 0.1 kpc that fall in the general range of normal starforming galaxies at z ∼ 6 [10]. Second for the halo component, we model the surface brightness profile with the exponential profile. The exponential profile is defined as Cn exp(−r/rn ) where rn is the scale length and Cn is a constant, and has been used for scale-length measurements of the Lyα halo around the high-z star-forming galaxies (e.g., [29–33]). To make the fitting stable, we fix the central peak positions at the same locations in both central and halo components. In top panel of Fig. 5.8, we show the best-fit results in the two component fitting based on the S´ersic+exponential profiles. We obtain the best-fit scale-length value for the halo component of rn = 3.3 ± 0.1 kpc, which corresponds to the effective radius of re = 5.6 ± 0.1 kpc. Because the central component, measured by the stacked HST/H -band image has re =1.1 ± 0.1 kpc, these results indicate that the [C ii] halo is extended more than the central galactic component composed of the stellar continuum by a factor of ∼5. Note that we confirm that we obtain a consistent result of re = 5.1± 1.7 kpc for the halo component from the visibility-based profile fitting with uvmultifit [34]. To examine any potential connections between the [C ii] halo and the Lyα halo that is homogeneously identified around normal star-forming galaxies at z ∼ 3–6 (e.g., [32, 33]), Fig. 5.8 compares the surface brightness profile of the [Cii] line with that of the Lyα line. For the [Cii] line emission, we show the stacked result of the ALMA-ALL sample, given the higher significance level detection than the ALMA-HST sample. For the Lyα line emission, we adopt recent results with the deep MUSE data for individual high-z LAEs of [33], where the authors systematically

5.2 Result

75

Fig. 5.8 Top: Two-component S´ersic+exponential profile fitting for the [C ii] line, averaged over 18 galaxies (Fig. 9 of [5], reproduced by permission of AAS). The red-dashed curves represent the best-fit results of the central stellar continuum and outer halo components, while the solid red curve denotes the sum of the best-fit two-component results. The solid blue curve and the shaded region indicate the median and the 16–84th percentile of the radial surface brightness profile of the Lyα lines in a recent control sample from MUSE [33]. For the Lyα line, we convolve the best-fit results of the S´ersic+exponential profiles with the ALMA beam. The red and blue arrows with error bars show the best-fit scale lengths of the [C ii] and the Lyα halo components, respectively. Bottom: Residuals in the best-fit results of one- (left) and two-component (right) profile fittings

76

5 Circumgalactic Medium Scale: Metal-Enriched Gas Halo

carry out the fitting with the two-component S´ersic+exponential profile. To perform a fair comparison, we select LAEs whose physical properties are consistent with our stacked sample (Table 2.4). As a result, we use the best-fit results of 6 LAEs at z > 5 with MUV  −21 mag and EWLyα < 100 Å. In 5.8, we find that the surface brightness profile of the Lyα line is comparable to that of the [Cii] line. The median scale length among the 6 LAEs is estimated to be 3.8 ± 1.7 kpc which agrees with the best-fit value of 3.3 ± 0.1 kpc for the halo component in the [C ii] surface brightness profile. These results may imply that the extended [Cii] line structure is related to the physical origins of the Lyα halo. We note that we confirm that the extended morphology of the [C ii] line emission is hard to be reproduced only with a single extended component. We also conduct the one-component (i.e., central alone) fitting to the [C ii] surface brightness profile, and the bottom panel of Fig. 5.8 shows the residuals of the [C ii] surface brightness profiles from the best-fit results of the one- and two-component fittings with uvmultifit. We find a bump at a radius of ∼ 1 over the errors appears in the one-component fitting, while the two-component fitting result shows the residuals broadly consistent with zero. These results indicates that the extended [C ii] line structure consists of a combination of the central plus halo components, instead of the single extended component alone.

5.2.5 [CII] Spectrum Stacking We also conduct the [C ii] spectrum stacking for the ALMA-ALL sample to verify whether a broad wing feature, a good probe for the on-going outflow activities, is observed. The [C ii] spectrum stacking is performed in the same manner as previous ALMA studies [27, 35]. To minimize the potential contamination of the close companions [2, 3], we adopt a relatively small aperture diameter of 0. 4 for the individual spectra In Fig. 5.9, we show the best-fit two Gaussian for the stacked [C ii]-line spectrum, where we model the line profile with the combination of the core and broad components. For stable results, we fix the velocity centers of the both components at 0 km/s in the fitting. We obtain the best-fit FWHMs of 296 ± 40 km/s and 799 ± 654 km/s for the core and broad components, respectively. To evaluate the significance level of the broad component, we calculate the velocity-integrated intensity in the velocity range of ±400–800 km/s and estimate it to be the 3.2σ level. It is reported that the stacked Keck spectra, whose stacking sample includes 6 out of our 18 [C ii] line sources, shows that the rest-frame UV metal absorption lines are blue-shifted with −1 from the [C ii]-systemic redshift [36]. the central outflow velocity of 440+110 −140 km s These results may suggest the tentative (3.2-σ ) broad wing feature might be indeed produced by the outflow. However, we note other possibilities that are attributed to the broad wing feature. One possibility is that the faint continuum is mistakenly regarded as the broad wing feature. Although the continuum subtraction has been performed for the [C ii] line

5.2 Result

77

Fig. 5.9 ALMA [C ii]-line spectrum averaged over the ALMA-ALL sample (Fig. 10 of [5], reproduced by permission of AAS). The spectrum is derived with the aperture diameter of 0. 4. The red curves denote the best-fit two (= core + broad) Gaussian component model. The shade regions indicate the velocity ranges in which the velocity-integrated intensity is tentatively detected at the 3.2σ level

data cubes of the 4 galaxies whose continuum is detected individually, the faint continuum from the rest of the 14 (= 18 − 4) galaxies could appear in the deep stacked spectrum after all. Another possibility is that the satellite galaxies. The [C ii] line emission from individual faint satellite galaxies are smoothed in the stacking procedure, which might be regarded as the broad wing feature. Given the low significance of the broad wing feature and these possibilities other than the outflow, here we do not conclude the existence of the outflow from the spectrum stacking results.

5.2.6 Comparison with Model We compare our observational results with two independent numerical simulations for star-forming galaxies having halo masses of Mhalo ∼ 1011 − 1012 M at z ∼ 6. Note that our sample is characterized by the average MUV value of  − 21 mag (Table 2.4) which corresponds to Mhalo ≈ 1011 − 1012 M based on the MUV –Mhalo relation [37]. First set is a star-forming galaxy of Althæa in a zoom-in simulation [38–40]. The dust radiative transfer (RT) and hydrodynamical simulations are combined together, which offers realistic predictions down to a spatial resolution of 30 pc for the spatial distribution of the rest-frame FIR and UV continuum emission as well as the [C ii] line. A full description of the dust RT and hydrodynamical simulations are summarized in previous studies [38–40]. We note that a post-processing step on snapshots of the hydrodynamical simulation calculates the dust RT. The [C ii] line emission is also calculated in post-processing in the same manner as [41] by utilizing the pho-

78

5 Circumgalactic Medium Scale: Metal-Enriched Gas Halo

Fig. 5.10 Left: 4 × 4 fake-color image for Althæa at z = 6.0 (top) and Halo-12 (bottom) in the zoom-in simulations (red: [Cii] line, green: rest-frame FIR continuum, blue: rest-frame UV continuum). Right: Radial surface brightness profiles of the [Cii] line (red curve), rest-frame FIR (green curve), and UV (blue curve) continuum emission estimated in the zoom-in simulations via stacking procedure. The solid and dashed color lines present the Althæa and Halo-12 results, respectively. The black dashed curve denotes the ALMA synthesized beam. The circles indicate the ALMA-HST stacking results whose colors are assigned in the same manner as the left panel. The figure is reproduced from Fig. 11 of [5] by permission of AAS

toionization code of cloudy [42]. We include calculations of the CMB suppression [43–46] in the process. Second set is another zoom-in cosmological hydrodynamic simulations based on the smoothed particle hydrodynamics (SPH) code Gadget- 3 [47] with the subgrid models developed in the First Billion Year (FiBY) project (e.g., [48]) and Overwhelmingly Large Simulations (OWLS) project [49] where the general properties of the high-redshift galaxy population are well reproduced (e.g., [50]). In this set, four different halos are used: Halo-12, Halo-A, Halo-B, and Halo-C that have Mhalo = (5 − 10) × 1011 M at z = 6.2 − 6.5. The details of Halo-12 is presented in [51–53], while we newly obtain the latter three halos from the simulation for this paper with similar initial conditions but with different merger histories. The minimum gravitational softening length corresponds to εg = 200 pc (comoving), and thus a ∼25 pc resolution is achieved at z = 7 for gravity. The SPH smoothing length is also allowed to adaptive down to 10% of εg , where the hydrodynamic resolution reaches a several parsecs at z = 6 − 7. As a post-process with “All-wavelength Radiative Transfer with Adaptive Refinement Tree” (ART2 code: [54, 55]), the RT calculation including the dust absorption/re-emission is conducted. This calculation solves for ionization structures of ISM/CGM and offers the SED over a wide wavelength range. The [C ii] line emissivity is computed based on the ionized carbon abundance.

5.2 Result

79

A full description of hydrodynamic simulations, RT, and the [C ii] line calculations are presented in [51, 53, 56]. The left panel of Fig. 5.10 shows color composites of the rest-frame UV, FIR continuum, and the [Cii] line of Althæa at z = 6.0 (top) and Halo-12 at z = 6.5 (bottom) in the zoom-in simulations. In both simulations, the [C ii] surface brightness morphology clearly shows extended structures over the radius scale of 10 kpc with surrounding satellite clumps and filamentary structures. These pictures of the extended [C ii] structure around the central galaxies is likely consistent with our observational results. To have a quantitative comparison, we conduct the stacking for the zoom-in simulations in the same manner as the observations. For the first set of the Althæa simulation, we produce 12 snapshots in the redshift range within 6.0 ≤ z ≤ 7.2. For each snapshot, the surface brightness is computed from face-on and three random angles, where we calculate the [C ii] line emissivity only within the velocity width of 100 km s−1 around the velocity center to match the visibility-based stacking procedure in the observations. We obtain a total of 48 (=12 × 4) snapshots of Althæa. Then, 9 out of 48 snapshots are randomly selected – the same sample size used for the stacking of the ALMA-HST sample. Finally, we perform the stacking for the 9 snapshots, and the stacked image is convolved with the ALMA beam. We refer to [57] for a full analysis of how the morphological results change from different viewing angles and evolutionary stages. For the second set of the zoom-in simulation, the surface brightness is calculated from three orthogonal angles for four different halos (Halo-12, Halo-A, Halo-B, and Halo-C), and we obtain a total of 12 (=4 × 3) snapshots. We then conduct the stacking and the convolution in the same manner as the first set. The right panel of Fig. 5.10 presents the radial surface brightness profiles derived from two independent zoom-in simulations after the stacking and the convolution with the ALMA beam. For comparison, the observational results from the ALMAHST sample are also presented, which is obtained in Sect. 5.2. We find that the overall trend of the rest-frame FIR and UV continuum in the observation are reproduced in both simulations within the errors. However, we also find that the extended [Cii] line structure in the observation is not reproduced in either of the simulations. In Althæa, although it reproduces the trend that the [C ii] line is extended more than the rest-frame UV and FIR continuum, the [C ii] line intensity at r > 5 kpc in the simulation is still fainter than the observed one. In Halo-12, the [C ii] line is the least extended. These results indicate that the existence of the [C ii] halo challenges current hydrodynamic simulations of galaxy formation.

References 1. Fujimoto S, Ouchi M, Kohno K et al (2018) ApJ 861:7 2. Jones GC, Willott CJ, Carilli CL et al (2017) ApJ 845:175 3. Carniani S, Maiolino R, Amorin R et al (2018) MNRAS 478:1170

80

5 Circumgalactic Medium Scale: Metal-Enriched Gas Halo

4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57.

Lindroos L, Knudsen KK, Vlemmings W, Conway J, Martí-Vidal I (2015) MNRAS 446:3502 Fujimoto S, Ouchi M, Ferrara A et al (2019) ApJ 887:107 Ilbert O, McCracken HJ, Le Fèvre O et al (2013) A&A 556:A55 Skelton RE, Whitaker KE, Momcheva IG et al (2014) ApJS 214:24 Sérsic JL (1963) Boletin de la Asociacion Argentina de Astronomia La Plata Argentina 6:41 Peng CY, Ho LC, Impey CD, Rix H-W (2010) AJ 139:2097 Shibuya T, Ouchi M, Harikane Y (2015) ApJS 219:15 Shibuya T, Ouchi M, Harikane Y (2015) ApJ 799:81 Hodge JA, Swinbank AM, Simpson JM et al (2016) ApJ 833:103 Hodge JA, Swinbank AM (2015) ApJ 810:133 Fujimoto S, Ouchi M, Shibuya T, Nagai H (2017) ApJ 850:1 Maiolino R, Carniani S, Fontana A et al (2015) MNRAS 452:54 De Looze I, Cormier D, Lebouteiller V et al (2014) A&A 568:A62 De Looze I, Cormier D, Lebouteiller V et al (2014) ApJ 788:L17 Murphy EJ, Condon JJ, Schinnerer E et al (2011) ApJ 737:67 Planck Collaboration, Abergel A, Ade PAR et al (2011) A&A 536:A21 Coppin K, Halpern M, Scott D et al (2008) MNRAS 384:1597 Díaz-Santos T, Armus L, Charmandaris V et al (2013) ApJ 774:68 Spilker JS, Marrone DP, Aravena M et al (2016) ApJ 826:112 Gullberg B, Swinbank AM, Smail I et al (2018) ApJ 859:12 Ferrara A, Vallini L, Pallottini A et al (2019) MNRAS 489:1 Capak PL, Carilli C, Jones G et al (2015) Nature 522:455 Rybak M, Calistro Rivera G, Hodge JA et al (2019) ApJ 876:112 Decarli R, Walter F, Venemans BP et al (2018) ApJ 854:97 Venemans BP, Neeleman M, Walter F et al (2019) ApJ 874:L30 Steidel CC, Bogosavljevi´c M, Shapley AE et al (2011) ApJ 736:160 Matsuda Y, Yamada T, Hayashino T et al (2012) MNRAS 425:878 Momose R, Ouchi M, Nakajima K et al (2014) MNRAS 442:110 Momose R, Ouchi M, Nakajima K et al (2016) MNRAS 457:2318 Leclercq F, Bacon R, Wisotzki L et al (2017) A&A 608:A8 Martí-Vidal I, Vlemmings WHT, Muller S, Casey S (2014) A&A 563:A136 Bischetti M, Maiolino R, Carniani S et al (2019) A&A 630:A59 Sugahara Y, Ouchi M, Harikane Y et al (2019) ApJ 886:29 Harikane Y, Ouchi M, Ono Y et al (2018) PASJ 70:S11 Pallottini A, Ferrara A, Bovino S et al (2017) MNRAS 471:4128 Pallottini A, Ferrara A, Gallerani S et al (2017) MNRAS 465:2540 Behrens C, Pallottini A, Ferrara A, Gallerani S, Vallini L (2018) MNRAS 477:552 Vallini L, Gallerani S, Ferrara A, Pallottini A, Yue B (2015) ApJ 813:36 Ferland GJ, Chatzikos M, Guzmán F et al (2017) Rev Mexicana Astron Astrofis 53:385 da Cunha E, Groves B, Walter F et al (2013) ApJ 766:13 Zhang Z-Y, Papadopoulos PP, Ivison RJ et al (2016) R Soc Open Sci 3:160025 Pallottini A, Gallerani S, Ferrara A et al (2015) MNRAS 453:1898 Lagache G, Cousin M, Chatzikos M (2018) A&A 609:A130 Springel V (2005) MNRAS 364:1105 Johnson JL, Dalla Vecchia C, Khochfar S (2013) MNRAS 428:1857 Schaye J, Dalla Vecchia C, Booth CM et al (2010) MNRAS 402:1536 Cullen F, McLure RJ, Khochfar S, Dunlop JS, Dalla Vecchia C (2017) MNRAS 470:3006 Yajima H, Nagamine K, Zhu Q, Khochfar S, Dalla Vecchia C (2017) ApJ 846:30 Arata S, Yajima H, Nagamine K, Li Y, Khochfar S (2019) arXiv e-prints, arXiv:1908.01438 Arata S, Yajima H, Nagamine K, Li Y, Khochfar S (2019) MNRAS 488:2629 Li Y, Hopkins PF, Hernquist L et al (2008) ApJ 678:41 Yajima H, Li Y, Zhu Q, Abel T (2012) MNRAS 424:884 Arata S, Yajima H, Nagamine K, Abe M, Khochfar S (2020) arXiv e-prints, arXiv:2001.01853 Kohandel M, Pallottini A, Ferrara A et al (2019) MNRAS 487:3007

Chapter 6

Cosmic Structure Scale: Number Density and Clustering

6.1 Data Analysis In this chapter, we analyze the large dataset of ALMA-FAINT (Table 2.1) composed of the multi-field deep ALMA data that became public by 2015 June. In the following subsections, we first show our data analyses. Our mm maps are observed in the wavelength range of 1.03−1.32 mm. The majority of our data are taken with Band 6, where our data set consists of 63 Band 6 maps and 4 Band 7 maps (electronic supplementary material Table 2). The average wavelength is estimated to be 1.23 mm among our major data of the ALMA Band 6 maps. We thus derive the number counts based on a flux density at 1.2 mm, S1.2 mm and scale the flux densities of the mm maps to S1.2 mm with the flux density ratios summarized in electronic supplementary material Table 2. We calculate all of the flux density ratios by assuming a modified blackbody whose dust temperature Td and spectral index βd are similar to those of typical SMGs; Td = 35 K (e.g., [1, 2]) and βd = 1.8 (e.g., Chain et al. [4]; Planck Collaboration [3]). The source redshift is assumed to be z = 2.5 that is a median redshift of SMGs (e.g., [5–7]). For example, the data of the previous studies defined with the wavelengths different from 1.2 mm, we estimate S1.2mm with the ratios of S1.2mm /S850μm = 0.37, S1.2mm /S870μm = 0.39, S1.2mm /S1.1mm = 0.77, and S1.2mm /S1.3mm = 1.28, for the data with the wavelengths different from 1.2 mm.

6.1.1 Source Detection We use SExtractor version 2.5.0 [8] to perform the source extraction for our ALMA maps before primary beam corrections. For the single-pointing, independent field maps, only the regions with the primary beam sensitivity greater than 50% is used

© Springer Nature Singapore Pte Ltd. 2021 S. Fujimoto, Demographics of the Cold Universe with ALMA, Springer Theses, https://doi.org/10.1007/978-981-16-4979-0_6

81

82

6 Cosmic Structure Scale: Number Density and Clustering

for the source extraction. For the mosaic map, we perform source extraction within the regions whose relative sensitivity to the deepest part of the mosaic map is greater than 50%. We assume that the sources are not spatially resolved out in our ALMA maps and identify sources whose positive peak count exceeds the 3.0σ level. We refer to the catalog of these sources as the 3.0σ -detection catalog. Given that the centers of the fields have the main science targets of the archived-data ALMA observations, we remove the objects located at the map centers from the 3.0σ -detection catalog.

6.1.2 Spurious Sources Since the detection is simply defined with a peak pixel count, many spurious sources should be included in the 3.0σ -detection catalog. We thus conduct negative peak analysis [10–12] for the deep (a), medium-deep (b), and cluster (c) datasets, which enables us to evaluate the number of the spurious sources. Each pixel in our ALMA maps is multiplied by −1, and for the negative peaks we perform the source extraction in the same manner as those in Sect. 6.1.1. Figure 6.1 shows the numbers of positive and negative peaks in our ALMA maps. We interpret the excess of the positive to negative peak numbers as the real source numbers. To quantitatively evaluate the real source numbers as a function of signal-to-noise ratio SNR, we model the distribution of the positive and negative peaks in Fig. 6.1 by c = an × 10−bn ×SNR , Nnp c Npp

= an × 10

−bn ×SNR

(6.1)

+ ap × 10

−bp ×SNR

,

(6.2)

c c and Nnp are the cumulative where an , bn , ap , and bp are free parameters, and Npp numbers of positive and negative peaks, respectively. We define spurious source rates f sp by the ratio of negative to positive peak numbers as a function of SNR,

f sp (SNR) =

c (SNR) Nnp c (SNR) Npp

.

(6.3)

We estimate the best-fit functions of Eqs. (6.1) and (6.2) and evaluate f sp . In Fig. 6.2, we show f sp as a function of SNR, which indicates that the spurious source rates are almost identical between the field data of deep (a) and medium-deep (b), while that of the cluster data (c) is higher than (a) and (b) at a given SNR. This result may suggest that the data depth does not affect the spurious source rates, but the mapping modes do. The complex distribution of the data depths in the mapping data of (c) might be another cause of the different relation between spurious source rate and SNR from relatively smooth distributions of the data depth in the single pointing data of (a) and (b). From the f sp results, we use sources down to an SNR whose spurious source rate is 70% which corresponds to 3.4, 3.4, and 3.9σ levels

6.1 Data Analysis Fig. 6.1 Total numbers of positive and negative peaks in the deep (a), medium-deep (b), and cluster (c) sub-datasets as a function of SNR (Fig. 6.2 of [9], reproduced by permission of AAS). The thick solid and dashed lines show the total numbers of the positive and negative peaks, respectively. The thin solid and dashed curves denote the best-fit model functions. The red lines are selection limits for our source catalog. See text for more details

83

84

6 Cosmic Structure Scale: Number Density and Clustering

Fig. 6.2 Cumulative spurious source rates as a function of SNR (Fig. 3 of [9], reproduced by permission of AAS). The solid, dashed, and dotted lines indicate the deep (a), medium-deep (b), and cluster (c) sub-datasets, respectively. The red lines at SNRs of 3.4 and 3.9 are the selection limits (SL) of our source catalog

in the data (a), (b), and (c), respectively. We apply these selection limits to the 3.0σ detection catalog and obtain 133 sources: 11 from the cluster data (c) and 122 from the field data (a and b). We present the source catalog of these 133 faint ALMA sources in electronic supplementary material Table 5. Note that we do not include an LBG at z = 7.5 behind A1689 (A1689-zD1; [13]) in our source catalog. This is because A1689-zD1 falls outside of the primary beam with a >50% sensitivity which is one of our selection criteria. Nevertheless, we have checked at the position of A1689-zD1 in our data and confirmed a source with an SNR of 4 whose flux is comparable to the measurements reported in Watson et al. [13] within the ∼1σ error.

6.1.3 Completeness and Flux Boosting We next evaluate the detection completeness by performing Monte-Carlo simulations. Based on the assumption that the sources are spatially unresolved in our ALMA maps, a flux-scaled synthesized beam is injected into a map as an artificial source on a random position. We scale these artificial sources in the SNR range of 3.0–7.0 with a step of 0.2. We then perform the source extraction in the same manner as the real data (Sect. 6.1.1) and regard the source is recovered, if an artificial source is extracted within a distance of the synthesized beam size from the input position. We iterate the above procedure 100 times for all of our ALMA maps at an SNR bin. In Fig. 6.3, we display the completeness estimates for three sub-datasets. A function of 1 − exp(aSNR − b) is fitted to these completeness estimates, where a and b are free parameters, and Fig. 6.3 also presents the best-fit functions. Note that we conduct these Monte-Carlo simulations for the completeness estimate in the data before

6.1 Data Analysis

85

Fig. 6.3 Completeness as functions of SNR for three sub-datasets (Fig. 4 of [9], reproduced by permission of AAS). The circles (red line), squares (blue line), and diamonds (green line) represent the completeness values (the best-fit functions) of the ID 6, 61 and 67 maps, respectively

primary beam corrections, because the source extraction in the real data (Sect. 6.1.1) is performed before primary beam corrections. We also examine potential flux boosting in our ALMA sources, which is generally caused by undetected faint sources [14, 15]. Here we use the artificial objects that are detected in the Monte-Carlo simulations during the completeness estimates, where we compare the output and input flux densities. We find that the ratios of the output to input flux densities are almost identical to one with a variance of ∼5% over the wide SNR range. This small flux boosting is probably because of the high angular resolution of ALMA. Therefore, we conclude that the flux boosting is negligible in our ALMA sources and do not apply any corrections for the potential flux boosting.

6.1.4 Flux Measurement We perform the 2D Gaussian-fitting routine of imfit in CASA to evaluate the integrated flux densities of our 133 objects as their source fluxes. Here we adopt the integrated flux density estimates of imfit only for sources that have an SNR of ≥5 and are classified as ‘spatially resolved’, because the large systematic uncertainties could be included in the integrated flux density estimates for low-SNR sources due to positive noise at around the source position. We use the peak flux values from the best-fit Gaussian with imfit for the rest of the sources. We confirm that the peak flux values of imfit, Simfit , and of SExtractor, SSEx , are consistent with each other within the 1σ error, Simfit /SSEx = 0.93 ± 0.15. Given that the gravitational lensing effects would make the sources distorted in the cluster data of A1689, we use a 1 circular gaussian uv-taper and produce a lowresolution map of A1689, which is in the same manner as the one produced by Watson et al. [13]. The beam size of this low-resolution map is 1. 31 × 1. 12. In this map,

86

6 Cosmic Structure Scale: Number Density and Clustering

we conduct the flux measurements with imfit at the source positions determined in the original map, in the same manner as the field sources described above. Here we define flux measurements in the non-tapered and tapered maps as Sno taper and Staper , respectively. Without two sources having signatures of source confusions by nearby objects, we obtain Staper /Sno taper = 1.01 ± 0.19 on average. However, a relatively high fraction of Staper /Sno taper = 1.23 is obtained in the strongly-lensed source of A1. We thus decide to adopt Staper for the sources in the cluster data of A1689, except for those two sources potentially confused by nearby objects. We note that we also perform the same test of tapering by making low-resolution maps for some of the field data. We find that Sno taper and Staper are almost identical in this case and thus decide to use Sno taper for the field data. To test the reliability of our flux measurements, we also perform Monte-Carlo simulations for the flux recovery. First, we select data of ID22 and ID55 as typical data sets whose beam sizes of ∼0. 8 and ∼1. 2 are equal to the two peaks of bimodal distribution of the beam size in our data. We then create 200 model sources with elliptical Gaussian profiles whose luminosities are normalized to the peak corresponding to SNRs in 3–7. Here we assume two source size cases for the model sources: 0. 2 and 0. 5 in the major-axis, that are taken from recent ALMA high-resolution studies for high-redshift submm sources (e.g., [16, 17]). These model sources are injected to the maps of ID22 and ID55 at random positions. Finally, in the same manner as our flux measurements for the real sources, we estimate flux densities of the model sources and compare the flux estimates with the input values. We find that the flux recovery ratio (defined by the ratio of the output to input fluxes) is 1.0 ± 0.1 (1.0 ± 0.2) for ID22 (ID55) data in the case that the input major-axis size is 0. 2. Similarly, we find 0.8 ± 0.1 (0.9 ± 0.2) for ID22 (ID55) data in the case that the input major-axis size is 0. 5. These results suggest that our flux measurements generally recover the true values within the 1 − 2σ uncertainties and that our flux measurements are reliable. We note that our data is not suffered from the missing fluxes of interferometric observations. With the shortest-baseline configuration that we use, ALMA proposer’s guide1 shows that the missing flux effects occur in sources with 2. 0−3. 0 scale spatial extensions at Band 6/7. The 2. 0−3. 0 scales significantly exceed our ALMA source sizes that are generally 0. 5. Therefore, the missing flux effect is negligible in our data.

6.1.5 Mass Model To analyze the cluster data (C) of A1689 at z = 0.183, We construct its mass model with the parametric gravitational lensing package glafic [18]. We make the mass model in the same manner as Ishigaki et al. [19]. Three types of mass distributions are included in our mass model: cluster member galaxy halos, cluster-scale halos, and 1

Table A-1 of the ALMA proposers’ guide: https://almascience.eso.org/proposing/proposersguide.

6.1 Data Analysis

87

external perturbation. To evaluate the mass distributions, we first produce a galaxy catalog of A1689 based on source extractions with optical images of g475 , r625 , and i 775 bands taken with HST. We estimate the cluster-scale halos with three brightest member galaxies placed in the core of the cluster. We then select the cluster member galaxies with the color criteria of r625 g475 − r625 g475 − r625

< 24, 1 > − (r625 − 24) + 0.7, 18 1 < − (r625 − 24) + 1.3, 18

(6.4)

where we carry out the source extraction and photometry with SExtractor. We calculate the external perturbation with the theoretical model, assuming that the perturbation is weak (e.g., [20]). To determine the best-fit mass model, we perform the standard χ 2 minimization and optimize free parameters of the mass profiles, using the spatial positions of multiple images of high-redshift galaxies behind the cluster presented in the literature [21, 22]. In the best-fit mass model, all of offsets between the model and observation positions of the multiple images are achieved within 1. 0 in the image plane. The best-fit mass model is presented in Fig. 2.4. We then estimate magnification factors (μ) at our source positions, assuming our source redshift at z = 2.5. We note that our statistical results is unchanged by this assumption. This is because most of mm sources in the target fields at the highgalactic latitude should reside at z > 1 where the μ values do not largely depend on a redshift. For example, if we calculate the magnification factors for sources at z = 7.5 that correspond to the redshift of A1689-zD1 and evaluate the differences between the magnification factors for sources at z = 2.5 (μ2.5 ) and z = 7.5 (μ7.5 ), we find that the average difference, Δμ ≡ (μ7.5 − μ2.5 )/μ2.5 , is 0.36 ± 0.25 for our 11 sources in A1689. For the conservative estimate, we add this lensing magnification difference of 0.36 into the errors of the intrinsic flux density estimates for our sources in A1689. Accordingly, these uncertainties are propagated to our major results including number counts in Sect. 6.2.1.

6.1.6 Survey Area We evaluate survey areas of our survey from the data of a, b, and c. The survey areas are defined by the high sensitivity regions that are detailed in Sect. 6.1.1. Because the sensitivities of our ALMA maps are not spatially uniform, the survey areas depend on the flux densities. In Fig. 6.4, we present our estimates of the survey areas of a, b, and c. In the field data of a and b, the decrease of the primary beam sensitivity corresponds to the increasing radius from the phasecenter of the observation. In other words, the

88

6 Cosmic Structure Scale: Number Density and Clustering

Fig. 6.4 Survey areas as a function of intrinsic flux densities (Fig. 5 of [9], reproduced by permission of AAS). The dashed, dotted, and solid lines represent the survey areas of the field (a and b), cluster (c), and total (a, b, and c) datasets, respectively

required intrinsic flux densities to be detected increase from the center to the edge of the data. First, we evaluate the radius where sources with a given intrinsic flux density can be detected at our selection limit SNR. We then calculate the circular areas with the radius obtained in each map, and sum up these areas to estimate the total survey areas as a function of intrinsic flux density. In the cluster data of c, we cannot regard the spatial distribution of the sensitivity as a simple function of radius, because the cluster data are mapped out by mosaic observations. Moreover, the gravitational lensing effect in the cluster data allows us to identify intrinsically faint sources. We thus make an effective magnification map by multiplying the sensitivity and the magnification factor estimated with the mass model of Sect. 6.1.5 at each pixel. We then estimate the survey area of the cluster data where a source with a given intrinsic flux density can be observed above our selection limit SNR.

6.2 Result 6.2.1 Number Counts at 1.2 mm The outputs obtained in Sect. 6.1 allow us to derive differential number counts at 1.2 mm. In the derivation of the number counts, we use the 122 sources identified in the ALMA Band 6 maps alone, because unknown systematics could be included in the 11 sources in the Band 7 in the flux scaling to 1.2 mm fluxes (Sect. 6.1). To derive the number counts, We generally follow methods used in the previous ALMA studies to derive the number counts [10–12]. A source with an intrinsic flux density of S has a contribution to the number counts, ξ , which is calculated by

6.2 Result

89

ξ(S) =

1 − f sp (S) , C(S)Aeff (S)

(6.5)

where Aeff is the survey area, and C is the completeness. For f sp (Eq. 6.3), Aeff , and C, we use the outputs obtained in Sect. 6.1. Then, a sum of the contributions for each flux bin is estimated by n(S) =

Σξ(S) , Δ log S

(6.6)

where Δ log S is the scaling factor which depends on the flux bin size. Because we adopt a flux bin size of 0.25 for the 1-dex width logarithmic differential number counts, the scaling factor in our estimate is 0.25. We evaluate the errors that include both flux uncertainties of the identified sources and Poisson statistical errors of the source numbers. For the Poisson uncertainty, we use the values presented in Gehrels [23] that are applicable for the small number statistics, because the numbers of our sources are small in each flux bin. For the flux uncertainty, We include the ALMA system’s flux measurement uncertainty whose typical value is ∼10% (Sect. 2.2.1) and the random noise. In addition, we add the lensing magnification uncertainties for our 11 sources in the A1689 data (see Sect. 6.1.5). To evaluate how flux uncertainty affect the number counts, we conduct Monte-Carlo simulations. We first create a mock catalog of the faint ALMA sources whose flux densities follow the Gaussian probability distributions with standard deviations corresponding to the flux uncertainty from system (+ the lensing magnification) uncertainties and the random noise. We then derive the number counts for each mock catalog in the same manner as what we apply to our real sources. Repeating this processes1000 times, we calculate the standard deviation of the number counts per flux bin due to the flux uncertainty. In conjunction with the standard deviation of the Poisson errors and the flux uncertainty, the 1σ uncertainties of the number counts is finally obtained. Figure 6.5 and Table 6.1 show the differential number counts with the associated 1σ uncertainties. We also derive the cumulative number counts at 1.2 mm that are summarized in Table 6.1, following the same manner as those deriving the differential ones. In Fig. 6.5, we caution that the faintest data point at ∼0.002 mJy (red open circle) consists of a single source with a very high magnification factor of μ > 100. Such a high magnification factor generally includes large systematic uncertainties related to the mass model. Moreover, the large error bar of the faintest data point does not offer important constraints on the number count estimates. Therefore, the faintest data point is not used in the following discussion. Except for this very highly magnified object, the intrinsic fluxes of our sample take the range from 0.018 to 1.2 mJy. We thus regard the flux density coverage of our study as ∼0.02−1 mJy.

90

6 Cosmic Structure Scale: Number Density and Clustering

Fig. 6.5 Differential number counts at 1.2 mm (Fig. 6 of [9], reproduced by permission of AAS). The red filled circles are our number counts derived from our faint ALMA sources. The red curve and the gray region denote the best-fit Schechter function and the associated 1σ error. In the Schechter function fitting, we use our number counts and the previous measurements shown with the black filled stars (Sc12; [24]) and the magenta filled diamonds (Ca15 1.1 mm; [12]). Note that the 1.1 mm results of Ca15 are obtained from ALMA sources, none of which are used in our number count measurements. Because the rest of the existing ALMA studies include some ALMA data covered by our study, for our Schechter function fitting we do not include the previous ALMA measurements shown with the magenta open squares (Ka13; [25]), open inverse triangles (H13), open stars (O14), open diamonds (Ca15), open triangles (Si15; [26]), and open pentagons (Ot15; [27]). Note that the faintest data point of our study (red open circle) is also removed from the sample of our Schether function fitting, due to the possible large systematic uncertainties. The pale-green region presents a flux range where no reliable number-count data bin exists (see text). We do not use the faintest data point of Sc12 (black open star) for our fitting, because this data point has unrealistically small errors. The dashed and dotted curves show the model predictions for number counts based on cosmological hydrodynamic simulations with GADGET- 3 (Sh12; [28]) and semi-analytic models (M15; [29]), respectively. These model predictions are roughly consistent with our results. For the previous measurements of the number counts at 1.1 mm and 1.3 mm different from our 1.2 mm band, we scale the flux densities with the methods shown in Sect. 6.1. Since the flux scaling factors for 1. 0. We thus perform Monte-Carlo simulations to evaluate the probability that the offsets between optical-NIR and mm sources is magnified to >1. 0 in the image plane. First, 100 artificial sources are placed in the source plane around an ALMA source in A1689. We make the positions of these artificial sources follow the Gaussian distributions having the standard deviation of 0. 4. Then, the offsets between the ALMA source and the artificial sources are calculated in the image plane based on our mass model. We repeat this process all of 11 ALMA sources identified in A1689. We find that ∼50% of the artificial sources are magnified to have >1. 0 offsets in the image plane. Given the number of optical counterparts identified in A1689 is 2, a total of 4 (=2/[50%]) optical counterparts are suggested to exist without lensing distortion effects (+the intrinsic offsets) from our simulation results. Therefore, the true counterpart-identification fraction of A1689 is estimated to be 67%(=4/6) that shows an excellent agreement with the counterpart-identification fraction of SXDS (65 ± 22%). If the lensing distortion effects (+the intrinsic offsets) is really taking place, 2 (=4 − 2) sources should have optical counterparts with an offset by >1. 0 from the ALMA sources in the image plane. To test this hypothesis, we refer the possible distribution of the artificial optical counterparts from our Monte-Carlo simulations and search for the possible optical-NIR counterparts whose offsets are >1. 0 from the ALMA sources. We find such possible optical counterparts in 4 sources of A1689. Some of these 4 sources could be real optical counterparts, and we thus confirm the hypothesis that the lensing distortion effects (+the intrinsic offsets) could be the reason for the difference of the counterpart-identification fractions between the blank field of SXDS and the cluster of A1689. The multi-band photometry, including Spitzer (3.6−24 µm) and Herschel (250−500 µm) data, are listed in Table 6.4 for the optical-NIR counterparts. Because the spatial resolutions of the Spitzer and Herschel data are worse than the search radius of 1. 0, we determine the counterparts in the Spitzer and Herschel data based on the visual inspection. This allows us to identify sources residing at a position slightly beyond the search radius of 1 . About a half of our faint ALMA sources are detected in the Spitzer data, while no sources are identified in the Herschel bands. These results indicate that our faint ALMA sources are not the dusty starburst galaxies whose SFRs exceed > a few 100M yr −1 and/or that they are very high redshift sources (z > 4). We also test whether red spectral energy distribution (SED) objects are missed in our search. These objects could be not detected in optical-NIR (0.4−2 µm) bands but only in Spitzer and Herschel bands. We adopt the visual inspection technique described above and search for Spitzer and Herschel counterparts of our faint ALMA sources with no optical-NIR counterparts. We find neither Spitzer nor Herschel coun-

23.76

27.42

24.84

···

27.24

25.76

26.31

25.69

>28.23 27.14

S2

S3

S4

S5

S6

S7

24.70

25.30

26.35

26.18

>28.23 26.68

26.00

23.45

>28.23 >28.93 28.50

>28.23 25.96

23.87

S10

S11

S12

S13

S14

S15

27.30

22.79

24.76

25.00

25.63

26.59

27.32

21.96

24.58

24.47

25.29

26.71

23.52

i

z

26.62

21.55

24.75

24.05

25.25

25.01

22.88

J

H

20.42

24.05

22.92

24.43

···

···

K

20.08

24.22

22.66

24.41

···

··· 19.94

4.5 µm 20.11

5.8 µm 20.12

8.0 µm

250 µm 350 µm 500 µm

>18.56 >14.35 >14.55 >14.15

24 µm

21.84

>22.85 >22.81 >18.56 >14.35 >14.55 >14.15

19.61

20.06

19.95

20.24

>18.56 >14.35 >14.55 >14.15

>25.27 >25.16 >22.85 >22.81 >18.56 >14.35 >14.55 >14.15

21.68

>25.27 >25.16 >22.85 >22.81 >18.56 >14.35 >14.55 >14.15

>25.27 >25.16 >22.85 >22.81 >18.56 >14.35 >14.55 >14.15

20.02

3.6 µm

>26.18 >25.31 >25.94 >25.27 >25.16 >22.85 >22.81 >18.56 >14.35 >14.55 >14.15

20.81

24.53

23.21

24.80

···

···

···

···

···

···

A1

A2

27.35

26.61

g475

23.40

26.09

22.96

21.63

22.65

24.02

25.65

20.93

22.45

23.62

25.07

20.24

21.55

22.36

25.90

20.22

21.18

21.76

24.60 20.88

21.26

25.67

20.45

26.28

25.78

r625

23.08

25.68

20.78

20.46

>22.81 >18.56 >14.35 >14.55 >14.15

22.55

19.89

19.91 22.73

19.66

19.69 22.33

19.35

19.31

16.66

15.95

>14.35 >14.55 >14.15

>14.35 >14.55 >14.15 >22.81 >18.56 >14.35 >14.55 >14.15

19.62

19.51

>25.27 >25.16 >22.85 >22.81 >18.56 >14.35 >14.55 >14.15

20.40

25.84

24.79

i 775

22.84

25.59

25.36

z 850

22.42

23.55

H160

21.25

>26.75 24.36

24.29

J110

21.72

···

···

···

21.16

18.45

4.5 µm

20.12

18.72

5.8 µm

20.32

19.10

8.0 µm

19.89

16.94

24 µm

>14.55 >14.75 >14.35

250 µm 350 µm 500 µm

>18.56 >14.35 >14.55 >14.15

>25.35 >24.85 >23.95 >23.45 >20.05 >14.55 >14.75 >14.35

18.25

3.6 µm

20.27

>28.24 >27.22 >26.18 >25.31 >25.94 >25.27 >25.16 >22.85 >22.81 >18.56 >14.35 >14.55 >14.15

>28.33 >28.24 >27.22 >26.18 >25.31 25.22

22.94

24.53

26.25

22.96

The u ∗ magnitudes are MAG_AUTO values given by our SExtractor photometry. Similarly, BV Rc i  z  magnitudes are MAG_AUTO values listed in the Furusawa et al. [64]’s catalogs. The Spitzer photometry values are total magnitudes estimated from the SEIP aperture flux densities with the aperture correction, while the HST photometry values (for A1 and A2) are the total magnitudes calculated from MAG_APER values of our SExtractor measurements. The aperture corrections for the Spitzer and HST photometry are provided by the Spitzer and HST websites, respectively (see text). The lower limits correspond to the 3σ levels. The S1 and S2 objects are not observed in u ∗ band and J H K bands †The secure photometry cannot be carried out due to the blending with a bright neighboring objects for the B-500 µm bands. See Fig. 6.8. ††The positions of the optical-NIR counterpart are far from the ALMA source center by ∼1. 2

···

···

ID

23.49

23.22

24.66

26.25

>28.23 25.30

S9††

25.14

23.86

24.57

S8†

Abell 1689

Rc

23.85

>28.93 >28.56 >28.33 >28.24 >27.22 >26.18 >25.31 >25.94 >25.27 >25.16 >22.85 >22.81 >18.56 >14.35 >14.55 >14.15

25.11

25.61

26.69

25.91

···

S1

24.24

B

24.46

u∗

ID

V

Table 6.4 Photometry of the optical-NIR counterparts

SXDS

104 6 Cosmic Structure Scale: Number Density and Clustering

6.2 Result

105

terparts whose optical-NIR fluxes are undetectable. These results decline the possibility that our faint ALMA sources are dusty starburst galaxies. To further examine AGN signatures of our faint ALMA sources, we cross-match our faint ALMA sources with the public catalogs of the radio 1.4 GHz and X-ray 0.5-10 keV sources. No counterparts are identified within a 1. 0 − 3. 0 radius around our faint ALMA sources. Thus, our faint ALMA sources are not the population with the AGN signatures. To summarize the multi-wavelength properties of the optical-NIR counterparts of our faint ALMA sources with Spitzer, Herschel, radio, and X-ray data, our faint ALMA sources have signs of neither dusty starbursts nor AGNs. we caution that we complete these cross-match analysis only in SXDS because of the lack of the available radio and X-ray data in A1689.

6.2.5.2

Lensed ALMA Sources in A1689

Two optical counterparts of A1 and A2 are gravitationally lensed objects identified in the A1689 cluster. Figure 6.9 presents the optical image at the position of A1. We identify that the optical and mm emission has a small positional offset of ∼0. 8, which can be explained by the positional uncertainty of the ALMA image (Sect. 6.2.5.1).4 The optical counterpart of A1 corresponds to one of the known multiple images, dubbed 5.2, while the rest of two multiple images are referred to as 5.1 and 5.3 [61]. In the ALMA image, the mm emission is marginally detected at the position of 5.1 at the 3σ level, while no signals above the 3σ level are identified at the position of 5.3. In Table 6.5, we summarize the physical properties of 5.1–5.3. We discuss the different mm emissivities among these multiple images in Sect. 6.2.5.1. The redshift of 5.1 and 5.3 is spectroscopically determined at z = 2.60 (Richard et al. in preparation, see [61]). Although a spectroscopic redshift of 5.2 (i.e. A1) has not yet been determined, we thus regard the redshift of A1 as z = 2.60. We convert the intrinsic 1.3-mm flux of A1 to the 1.2-mm one with the flux scaling factor (Sect. 6.1), and obtain 0.020 ± 0.008 mJy after correcting the primary beam sensitivity and the lensing magnification of μ = 22.8. Previous SMA 870 µm observations did not detect 5.1 and 5.3 above the 3σ level, but only 5.2 [71]. The intrinsic flux of 5.2 in Chen et al. [71] is measured to be 0.085 ±0.035 mJy at 870 µm. With the flux scaling factor from 870 µm to 1.2 mm (Sect. 6.1), this corresponds to 0.033 ± 0.014 mJy at 1.2 mm, which is consistent with our measurement of 0.020 ± 0.008 mJy within the 1σ errors. The optical-NIR counterpart of A2 is identified with the lensing magnification factor of μ = 4.9, which indicates the intrinsic 1.2 mm flux of A2 to be 0.077 ± 0.035 mJy after correcting the primary beam attenuation and the lensing magnification. Neither z photo nor z spec of A2 are available in the literature. 4

We caution that the offset could be explained by a real positional difference between the opticaland mm-emitting regions. However, it is required to perform ALMA data analysis for a positional accuracy better than this study.

106

6 Cosmic Structure Scale: Number Density and Clustering

Fig. 6.9 False-color image of A1 (Fig. 10 of [9], reproduced by permission of AAS). The used images and color assignment are the same as Fig. 2.4. The center of the cyan circles pinpoint the positions of the 5.1 and 5.2 sources, which are two of the multiple images of the galaxy at z spec = 2.60 [21, 61]. The radius of the cyan circle is 1. 0 that corresponds to the search radius for the counterpart identifications. The red contours show the ALMA 1.3 mm emission at the 2, 3, 4, and 5 σ levels. The white curves represent the lines for the lensing magnification of μ = 50

We compare the physical properties between these lensed ALMA sources with optical-NIR counterparts identified in A1689 (Table 6.5) and the previously-known lensed sources, SMMJ16358 (e.g., [73]) and Cosmic Eylash (e.g., [74]), identified by single-dish observations. SMMJ16358 and Cosmic Eyelash have intrinsic SFRs of ∼500 and 210 M yr −1 , respectively. The SFR values are inferred to be 1.3 and 0.5 mJy at 1.2 mm based on the S F R − L IR relation [36]. On the other hand, the opticalNIR counterparts of our lensed ALMA sources have intrinsic fluxes of ∼0.02−0.08 mJy at 1.2 mm. Thus, our ALMA study has identified a population with the intrinsic flux  10 times fainter, i.e., less star-forming, than previously known lensed mm sources. Despite the lack of optical-NIR counterparts, we also find a very strongly lensed source candidate, A3, that shows the double-peak morphology, dubbed A3a and A3b, in the ALMA image. We present the ALMA image of A3, including A3a and A3b, in the upper panel of Fig. 6.10. A3 falls very close to the critical curve for a z = 3 source (Fig. 6.10), and the critical curve crosses between the double peaks of A3a and A3b. More quantitatively, we examine whether multiple images of a single lensed source behind A1689 can explain these positions of double peaks. Based on our mass model, we calculate predicted positions of multiple images made by a lensed source at z = 2 − 6, which is shown in cyan lines in the upper panel of

(3)

13:11:29.064

13:11:29.224

13:11:34.120

(2)

···

A1

···

(1)

5.1

5.2

5.3

28.33

r625

>28.24

i 775

>27.22

z 850

159.5

(7)

μ

(1)−(7) correspond to (2)−(8) in Table 6.5. These values are for the A3a component (Fig. 6.10). Because A3a and A3b are confused in the uv-taper image, these fluxes are obtained on the image without uv-taper

A3

(1)

ID

Table 6.6 Very strongly lensed source candidate

6.2 Result 109

110

6.2.5.3

6 Cosmic Structure Scale: Number Density and Clustering

Color and Luminosity Properties

To understand the population of the faint ALMA sources with the detectable opticalNIR continuum emission, we investigate color (i.e. SED) and luminosity properties of the optical-NIR counterparts, S1-15 and A1-2, First of all, we refer the public photo-z catalog of Williams et al. [62] estimated from the optical-NIR SEDs, and find that +0.08 +0.03 , 1.57+0.07 photo-z values of S11, S12, S15 are 1.54−0.02 −0.02 , and 1.45−0.02 , respectively. +0.23 We also find that photo-z values of S4 and S6 are 0.77−0.03 and 0.67+0.01 −0.02 , respectively, and thus that S4 and S6 reside at low redshift (z < 1). This indicates that at least ∼10% (= 2/17) optical-NIR counterparts are low-z objects. Nevertheless, because the majority of mm sources are thought to be high-z galaxies and AGNs [75], we interpret that the rest of 15 (=17 − 2) optical-NIR counterparts are candidates of high-z galaxies. We thus use high-z populations (z > 1) from the literature for the comparison of colors and luminosities. The major high-z populations are LAEs, BzKs, DRGs, SMGs, LBGs (including BX/BM), and AGNs. Note that LAEs, SMGs, and DRGs are generally very faint in optical continua. To perform a reliable comparison with our optical-NIR counterparts, we focus on the other populations of LBGs (BX/BM) and BzKs. Comparison with the LBG BX/BM Populations The BX and BM galaxies generally reside in the redshift ranges of 2.0  z  2.5 and 1.5  z  2.0, respectively [78], and overlapped with the LBG population, while the LBG population also includes sources at z  3. The BX and BM galaxies are defined by the color criteria of G − R and Un − G. We adopt these color criteria to identify the BM and BX galaxies in our optical-NIR counterparts, but the Un , G and R bands are not included in the photometry bands of our data. Therefore, we alternatively adopt a photometry set of BV − Rc i  and U − BV , following the procedure in Ly et al. [76]. The Rc i  and BV photometry are defined by  x1 f R + (1 − x1 f i  ) , 3630 μJ y   x2 f B + (1 − x2 f V ) , BV = −2.5log 3630 μJ y

Rc i  = −2.5log



(6.11)

where f X is the flux density per unit frequency (erg s−1 cm−2 Hz−1 ) in a band ‘X’ corresponding to B, V , Rc , and i  . The coefficients of x1 and x2 are 0.207 and 0.314, respectively. Based on these photometric relations and the original color criteria in [78], we derive the color criteria of the BX objects as BV −Rc i  ≤ 0.2(u − BV ) + 0.4, U −BV ≥ BV − Rc i  + 0.23, BV −Rc i  ≥ −0.2, U −BV ≤ BV − Rc i  + 1.0.

(6.12)

6.2 Result

111

Fig. 6.11 Two-color diagram for our optical-NIR counterparts (Fig. 12 of [9], reproduced by permission of AAS). The cyan and magenta-shaded regions are the BX and BM selection windows that are defined in the BV − Rc i  and u ∗ − BV color plane [76]. The BX and BM galaxies reside at z  2 − 2.5 and z  1.5 − 2.0, respectively. The yellow shades indicate the approximate region of the selection window for z ∼ 3 LBGs [77], where we assume that the original photometry of Un , G and R bands are roughly the same as those of our u ∗ , B, and Rc bands, respectively

Similarly, we derive another color criteria of the BM objects as U −BV ≥ −0.1, BV −Rc i  ≤ 0.70(U − BV ) + 0.280, BV −Rc i  ≤ −0.382(U − BV ) + 0.853, BV −Rc i  ≥ −0.2, U −BV ≥ BV − Rc i  + 0.23.

(6.13)

We applying these alternative color criteria to our optical-NIR counterparts. There are 10 out of 17 optical-NIR counterparts that are observed with the photometric system of u ∗ BV Rc i  required in the color criteria. Except for the obviously low-z objects of S4 and S6, we use 8 (=10 − 2) optical-NIR counterparts. In Fig. 6.11, we display the two color diagrams of these 8 sources together with the color selection criteria. We find that 6 sources fall in or near the selection windows of z ∼ 3 LBGs or z ∼ 2 BX/BM, where we include S11 at the border of the selection window. S9 locates at the upper right corner of Fig. 6.11. These colors around the S9 position are comparable to z  4 LBGs (see, e.g., [77]). It is thus likely S9 is explained by a LBG at z  4. On the other hand, S3 deviates from any selection windows. We conclude that 7 out of 8 optical-NIR counterparts have colors consistent with either z  3 LBGs or z ∼ 2 BX/BM. Given the two low-z objects of S4 and S6,

112

6 Cosmic Structure Scale: Number Density and Clustering

70% (= 7/[8 + 2]) of the optical-NIR counterparts meet the color criteria for the LBG (BX/BM) population. We also compare magnitudes between the 10 opticalNIR counterparts and the known z  3 LBGs and z ∼ 2 BX/BM. We find that the Rc magnitudes take the range of 23 − 27 mag in all of the 10 optical-NIR sources (Table 6.4), which is the typical brightness of z  3 LBGs and z ∼ 2 BX/BM [77, 78]. We thus conclude that a majority (∼70%) of the optical-NIR counterparts of the faint ALMA sources belong to the LBG (BX/BM) population. Comparison with the BzK Populations In contrast with the LBG (BX/BM) population, the BzK population also includes moderately reddened dust-poor star-forming galaxies (a.k.a sBzK), in addition to blue star-forming galaxies, at 1.4  z  2.5 [79]. We apply the color criteria of the BzK population to our optical-NIR counterparts. There are 10 out of the 17 opticalNIR counterparts that have the sufficient photometric measurements of B, z  , and K bands (Table 6.4).5 Removing obviously low-z objects of S4 and S6 again, we analyze 8 optical-NIR counterparts in this color selection. Because the photometry system in SXDS used in our study is different from that of previous BzK studies (e.g., [79]), we again correct the color selection criteria for BzK galaxies. For this correction, we follow the same manner as [80] given by, (B − z)Daddi+04 = (B − z)SXDS + 0.3, (z − K )Daddi+04 = (z − K )SXDS + 0.1.

(6.14)

In Fig. 6.12, we present the B − z and z − K colors of the 8 optical-NIR counterparts. We find that 5 out of the 8 optical-NIR counterparts are classified as sBzK, while the rest of the three sources (S3, S9 and S10), have colors different from sBzK or pBzK galaxies whose redshift takes the range of z  1.4 − 2.5. Because S3 also shows the different colors from those of the LBGs and BX/BM populations (Sect. 6.2.5.3), S3 would fall at z  1.4 that is out of the redshift window of BzK. Although the colors of S9 and S10 are similar to those of stars, in Sect. 6.2.5.3, the LBG and BX/BM galaxy selection results show that the colors of these two sources agree with z  3 galaxies. Thus, S9 and S10 would be galaxies at z  3 that fall out of the redshift window of BzK. In this way, we confirm that the results of the BzK selection are consistent with those of LBG and BX/BM galaxy selection. In summary, given the two low-z objects of S4 and S6, we find that 50%(=5/[8 + 2]) of the optical-NIR counterparts are classified as sBzK galaxies at z  1.4 − 2.5. We then again compare magnitudes of these optical-NIR counterparts (Table 6.4), and find that the optical and NIR magnitudes are consistent with those of the typical sBzK galaxies (e.g., [79]). Comparison with the AGN Population Although no X-ray nor radio continua are detected from our optical-NIR counterparts, we also study the mid-IR color property to further test for the AGN. This is because the bright mid-IR bright color is generally caused by the hot-dust emission, which 5

The K band photometry is missed in the A1689 region misses.

6.2 Result

113

Fig. 6.12 Two-color diagram of the BzK galaxy selection in the SXDS photometry system (Fig. 13 of [9], reproduced by permission of AAS). The red circles denote the 7 optical-NIR counterparts. Star-forming galaxies at z  1.4 − 2.5, called sBzK galaxies, are selected in the region beyond the solid line defined by (z − K ) − (B − z) ≥ −0.2. The pBzK galaxies, passively evolving galaxies at z  1.4 − 2.5, are selected in the triangular region defined with the sold and dashed (z − K > 2.5) lines at the upper right corner. Galactic stars fall in the region below the dotted line (See also [79, 80])

can be a good probe for the AGN. Six counterparts of S1, S6, S11, S12, S15, and A1 are detected in the mid IR of 8 and 24 µm bands, and we compare the mid-IR colors between the six mid-IR bright counterparts and the local templates of a starburst (Arp 220) a typical AGN (Mrk 231). In Fig. 6.13, we show the diagnostic color diagram of S8.0µm /S4.5µm vs. S24µm /S8.0µm that separates starburst (SB) and AGN populations [81]. The dashed and dotted lines are taken from Fig. 4 of [81], which indicates the color tracks of Mrk 231 and Arp 220 in the redshift range of 0.5−3.0. All of our six mid-IR bright counterparts fall near the color track of Arp 220 (SB), and far from that of Mrk 231 (AGN). This suggests that the six counterparts are not AGN, but SB. IRX-β Relation To evaluate the SED properties quantitatively, we calculate the rest-frame UV continuum slope β and the IR-to-UV luminosity ratio IRX(≡L IR /L 1600 ) for 6 faint ALMA sources of S4, S6, S11, S12, S15, and A1 whose spectroscopic or photometric redshifts have been measured. Here we define L IR as the integrated infrared flux density in a range of rest-frame 8–1000 µm. We estimate β with two broadband magnitudes, m 1 and m 2 : β=−

m1 − m2 − 2, 2.5 log(λ1c /λ2c )

(6.15)

114

6 Cosmic Structure Scale: Number Density and Clustering

Fig. 6.13 Color-color diagram for the AGN diagnostics (Fig. 14 of [9], reproduced by permission of AAS). The red circles and upper limits are our five mid-IR bright counterparts. The dotted and dashed lines present the color tracks of Mrk 231 (AGN) and Arp 220 (SB), respectively, that are redshifted from z  0.5 to 3.0. The numbers labeled to the lines represent the redshifts

where λ1c and λ2c are the central wavelengths of the two broadband filters. We use (m 1 , m 2 ) = (u ∗ , B), (u ∗ , B), (V , i  ), (V , i  ), (V , i  ), (r625 , z 850 ) for S4, S6, S11, S12, S15, and A1, respectively. These IRX and β values are shown in Fig. 6.14, together with those of optically selected star-forming galaxies at z ∼ 2 and nearby (U)LIRGs. In Fig. 6.14, we also present 5.1, the multiple image of A1 (Table 6.5), whose ALMA continuum is also marginally detected (∼3σ ). For the lensed sources of A1 and 5.1, we use the lensing magnification values estimated at the optical (submm) source positions, assuming that the positional offsets between the optical and submm sources are real, to derive the intrinsic optical (submm) luminosities. In Fig. 6.14, all of the six faint ALMA sources show the IRX-β relations comparable to the Calzetti and SMC law curves. At a given β, all these six faint ALMA sources locates far below the (U)LIRGs. The clear difference between (U)LIRGs and the faint ALMA sources indicates that the faint ALMA sources are not the objects whose infrared luminosities are just fainter than (U)LIRGs by 1−2 order(s) of magnitude, i.e., miniature (U)LIRGs. We caution that these results hold on the 59% of the faint ALMA sources that have the optical-NIR counterparts. The rest of the 41% of the faint ALMA sources without optical-NIR counterparts could have low L 1600 values, suggested by the fact of no optical-NIR counterparts, and IRX values as high as those of (U)LIRGs. These 6 faint ALMA sources consist of 2 low-z sources (S4 and S6) and 4 high-z sources (S11, S12, S15, and A1), where we define low-z and high-z by z < 1 and z ≥ 1, respectively. In Fig. 6.14, one low-z source of S4 is placed at the region of z = 2 optically selected star-forming galaxies [82], while the other low-z source of S6 has large IRX and β values. In contrast, all of the four high-z sources have the IRX-β relations well aligned with those of the z = 2 optically selected galaxies. These results imply that low-z sources would be widely distributed in the IRX-β plot, but that majority of high-z sources are similar to z = 2 optically selected galaxies.

6.2 Result

115

Fig. 6.14 Infrared-to-UV luminosity ratio IRX as a function of the UV slope β (Fig. 15 of [9], reproduced by permission of AAS). The red filled circle is the spectroscopically-confirmed A1 (5.2) source at z spec = 2.60. The red open circles denote the faint ALMA sources with the photometric redshifts. The open black circle is the 5.1 source of the A1 multiple image system that is marginally detected in the ALMA data (Table 6.5). The blue diamonds (arrow) represent(s) the measurements (the upper limit) of the z ∼ 2 optically-selected galaxies obtained by the stacking analysis [82]. The black squares indicate nearby (U)LIRGs [83, 84]. The solid, dashed, and dotted curves denote the IRX-β relation of the extinction curves for the Calzetti law [85], the updated Calzetti law [86], and the SMC law [87]. For the Calzetti law curve, we shift the original relation of the Calzetti law [85] by +0.24 dex, because [85] define their infrared luminosities at 40−120 µm that is different from our definition (8−1000 µm)

Note that we confirm that 5.2 (i.e., A1) and 5.1 are the multiple images of a single high-z source based on the similar IRX-β values. What are the Optical-NIR Counterparts of the Faint ALMA Sources? In Table 6.7, we summarize the results in this subsection. We find that 7 out of the 10 optical-NIR counterparts (including the low-z galaxies) meet the color criteria of LBG, BX/BM, and sBzK as well as the optical magnitudes consistent with these populations. Among the rest of the three counterparts, two are low-z objects and the remaining one is probably the foreground/background object due to the chance projection (Sect. 6.2.5.1). These results suggest that not all but the majority of the optical-NIR counterparts of our faint ALMA sources belong to the LBGs, BX/BM, and sBzKs populations. Moreover, neither AGN nor (U)LIRG signatures are identified from any optical-NIR counterparts. We thus conclude that the majority of our faint ALMA sources with a detectable optical continuum (∼25 mag) are UV bright star-forming galaxies of LBG, BX/BM, and sBzKs with no AGN. The clustering analysis results in Sect. 6.2.4 show a weak clustering signal of bg < 3.5 comparable to those of sBzKs and LBGs, which supports this conclusion.

34.612686

34.613739

34.437805

34.446648

34.446007

34.387825

34.411190

34.305553

34.194702

34.197121

34.442795

34.441357

34.747295

34.274231

34.587593

197.871590

197.869086

S1

S2

S3

S4

S5

S6

S7

S8

S9

S10

S11

S12

S13

S14

S15

A1

A2

1.22

1.03

1.26

1.26

1.25

1.23

1.30

1.25

1.25

1.23

1.23

1.25

1.30

1.25

1.32

1.32

−4.588505

−5.494339

−5.007715

−5.008159

−5.220457

−4.745051

−5.067293

−5.056259

−5.059778

−4.911066

−4.912045

−4.858103

−4.860201

−5.318766

−1.345783

−1.337143

0.31 ± 0.08

0.30 ± 0.06

0.32 ± 0.09

0.32 ± 0.09

0.37 ± 0.10

0.32 ± 0.09

0.53 ± 0.10

0.31 ± 0.08

0.43 ± 0.12

0.29 ± 0.07

0.30 ± 0.08

0.41 ± 0.11

0.049 ± 0.014

0.047 ± 0.014

0.056 ± 0.015

0.31 ± 0.09

0.36 ± 0.07

(2)

(1)

1.22

Sλobs obs (mJy)

λobs (mm)

−4.583063

Decl. (J2000)

0.077 ± 0.033††

0.020 ± 0.008††

0.21 ± 0.10

0.43 ± 0.12

0.26 ± 0.11

0.29 ± 0.09

0.56 ± 0.11

0.37 ± 0.10

0.39 ± 0.15

0.22 ± 0.09

0.14 ± 0.09†

0.42 ± 0.13

0.050 ± 0.017

0.044 ± 0.016

0.032 ± 0.010

0.31 ± 0.10

0.43 ± 0.09

(3)

corr S1.2mm (mJy)

···

···

2.60

···

1.45+0.03 −0.02 ···

···

···

···

···

··· ···

···

+0.08 1.54−0.02 +0.07 1.57−0.02

···

···

···

···

···

···

···

··· ···

···

···

0.77+0.23 −0.03 0.67+0.01 −0.02

···

···

···

(5)

z spec

···

···

···

(4)

z photo

···

···

Y

···

···

Y

Y

N

N

···

···

N

···

···

N

N

N

N

···

N

Y ···

···

N

···

N

···

···

(7)

pBzK

···

Y

···

N

···

···

(6)

sBzK

···

···

Y

···

···

Y

···

N

N

···

···

N

N

···

Y Y

··· ···

Y

···

N

···

···

···

···

N

(9)

AGN

···

Y

···

Y

···

N

···

···

(8)

BX/BM +LBG

Notes (1): Wavelength in the observed frame. (2): Peak flux density of the SExtractor measurement (Sect. 6.1.1) with the primary beam correction at the observed wavelength. (3): The best-estimate source flux density at 1.2 mm. The source flux is estimated with the 2D Gaussian-fitting routine of imfit in CASA (Sect. 6.1.4) with the primary beam correction and the flux scaling to 1.2 mm (Sect. 6.1). For A1 and A2 sources, the lensing magnification corrections are also applied. The lensing magnification factors are estimated at the ALMA flux peak positions of A1 and A2. (4): Photometric redshift estimated by Williams et al. [62]. (5): Spectroscopic redshift obtained by Limousin et al. [61]. (6), (7), (8), and (9): “Y” indicates the objects that meet the color selection criteria of sBzK, pBzK, BX/BM, and AGN populations, respectively. “N” represents the sources escaping from the color space of the selection criteria. “Y ” is presented for the sources having colors consistent with the populations †The photometry of imfit may be biased, due to the systematic noise. The SExtractor peak flux measurement suggests 0.32 ± 0.08 mJy. †† The magnification uncertainty (Sect. 6.1.5) is also included in the error

R.A. (J2000)

ID

Table 6.7 Properties of our optical-NIR counterparts

116 6 Cosmic Structure Scale: Number Density and Clustering

6.2 Result

117

We caution that the optical-NIR counterparts are identified in 59% of our faint ALMA sources. In other words, the above conclusion can be applied roughly about a half of the faint ALMA sources, and the rest half of the faint ALMA sources are still unclear about their origins. One possibility is that the rest half of the faint ALMA sources are the dust obscured star-forming galaxies that have been missed in the optical–NIR observations. In fact, recent ALMA studies also report the existence of the dark objects in the optical–NIR wavelengths that are not detected in any HST bands even in CANDELS fields, but reliably detected with the ALMA (i.e. HST– dark objects; [88]). We discuss the existence of the objects missed in the optical–NIR observations in Sect. 7.4.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36.

Kovács A, Chapman SC, Dowell CD et al (2006) ApJ 650:592 Coppin K, Halpern M, Scott D et al (2008) MNRAS 384:1597 Chapin EL, Pope A, Scott D et al (2009) MNRAS 398:1793 Planck Collaboration, Abergel A, Ade PAR et al (2011) A&A 536:A21 Chapman SC, Blain AW, Smail I, Ivison RJ (2005) ApJ 622:772 Yun MS, Scott KS, Guo Y et al (2012) MNRAS 420:957 Simpson JM, Swinbank AM, Smail I et al (2014) ApJ 788:125 Bertin E, Arnouts S (1996) A&A 117:393 Fujimoto S, Ouchi M, Ono Y et al (2016) ApJS 222:1 Hatsukade B, Ohta K, Seko A, Yabe K, Akiyama M (2013) ApJ 769:L27 Ono Y, Ouchi M, Kurono Y, Momose R (2014) ApJ 795:5 Carniani S, Maiolino R, De Zotti G et al (2015) A&A 584:A78 Watson D, Christensen L, Knudsen KK et al (2015) Nature 519:327 Austermann JE, Aretxaga I, Hughes DH et al (2009) MNRAS 393:1573 Austermann JE, Dunlop JS, Perera TA et al (2010) MNRAS 401:160 Ikarashi S, Ivison RJ, Caputi KI et al (2014) arXiv:1411.5038 Ikarashi S, Ivison RJ, Caputi KI et al (2015) ApJ 799:81 Oguri M (2010) PASJ 62:1017 Ishigaki M, Kawamata R, Ouchi M et al (2015) ApJ 799:12 Kochanek CS (1991) ApJ 373:354 Coe D, Benítez N, Broadhurst T, Moustakas LA (2010) ApJ 723:1678 Diego JM, Broadhurst T, Benitez N et al (2015) MNRAS 446:683 Gehrels N (1986) ApJ 303:336 Scott KS, Wilson GW, Aretxaga I et al (2012) MNRAS 423:575 Karim A, Swinbank AM, Hodge JA et al (2013) MNRAS 432:2 Simpson JM, Smail I, Swinbank AM et al (2015) ApJ 807:128 Oteo I, Zwaan MA, Ivison RJ, Smail I, Biggs AD (2015) arXiv:1508.05099 Shimizu I, Yoshida N, Okamoto T (2012) MNRAS 427:2866 Makiya R, Enoki M, Ishiyama T et al (2015) arXiv:1508.07215 Knudsen KK, van der Werf PP, Kneib J-P (2008) MNRAS 384:1611 Chen C-C, Cowie LL, Barger AJ et al (2013) ApJ 776:131 Scoville N, Aussel H, Brusa M et al (2007) ApJS 172:1 Hodge JA, Karim A, Smail I et al (2013) ApJ 768:91 Schechter P (1976) ApJ 203:297 Fixsen DJ, Dwek E, Mather JC, Bennett CL, Shafer RA (1998) ApJ 508:123 Kennicutt RC Jr (1998) ARA&A 36:189

118 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88.

6 Cosmic Structure Scale: Number Density and Clustering Adelberger KL, Steidel CC, Giavalisco M et al (1998) ApJ 505:18 Robertson BE (2010) ApJ 716:L229 Mo HJ, White SDM (2002) MNRAS 336:112 Sheth RK, Tormen G (1999) MNRAS 308:119 Webb TM, Eales S, Foucaud S et al (2003) ApJ 582:6 Blain AW, Chapman SC, Smail I, Ivison R (2004) ApJ 611:725 Weiß A, Kovács A, Coppin K et al (2009) ApJ 707:1201 Williams CC, Giavalisco M, Porciani C et al (2011) ApJ 733:92 Hickox RC, Wardlow JL, Smail I et al (2012) MNRAS 421:284 Grazian A, Fontana A, Moscardini L et al (2006) A&A 453:507 Hayashi M, Shimasaku K, Motohara K et al (2007) ApJ 660:72 Quadri R, van Dokkum P, Gawiser E et al (2007) ApJ 654:138 Blanc GA, Lira P, Barrientos LF et al (2008) ApJ 681:1099 Furusawa J, Sekiguchi K, Takata T et al (2011) ApJ 727:111 Lin L, Dickinson M, Jian H-Y et al (2012) ApJ 756:71 Ouchi M, Shimasaku K, Okamura S et al (2004) ApJ 611:685 Adelberger KL, Steidel CC, Pettini M et al (2005) ApJ 619:697 Overzier RA, Bouwens RJ, Illingworth GD, Franx M (2006) ApJ 648:L5 Lee K-S, Giavalisco M, Gnedin OY et al (2006) ApJ 642:63 Gawiser E, Francke H, Lai K et al (2007) ApJ 671:278 Ouchi M, Shimasaku K, Furusawa H et al (2010) ApJ 723:869 Fry JN (1996) ApJ 461:L65 Peebles PJE (1993) Principles of physical cosmology. Princeton University Press, Princeton Scoville N, Sheth K, Aussel H et al (2015) arXiv:1505.02159 Limousin M, Richard J, Jullo E et al (2007) ApJ 668:643 Williams RJ, Quadri RF, Franx M, van Dokkum P, Labbé I (2009) ApJ 691:1879 Ueda Y, Watson MG, Stewart IM et al (2008) ApJS 179:124 Furusawa H, Kosugi G, Akiyama M et al (2008) ApJS 176:1 Oliver SJ, Bock J, Altieri B et al (2012) MNRAS 424:1614 Simpson C, Martínez-Sansigre A, Rawlings S et al (2006) MNRAS 372:741 Ashby MLN, Stanford SA, Brodwin M et al (2013) ApJS 209:22 Bouwens RJ, Illingworth GD, Franx M et al (2009) ApJ 705:936 Bouwens RJ, Illingworth GD, Franx M et al (2015) ApJ 803:34 Downes AJB, Peacock JA, Savage A, Carrie DR (1986) MNRAS 218:31 Chen C-C, Cowie LL, Barger AJ, Wang W-H, Williams JP (2014) ApJ 789:12 Chen C-C, Smail I, Swinbank AM et al (2015) ApJ 799:194 Kneib J-P, van der Werf PP, Kraiberg Knudsen K et al (2004) MNRAS 349:1211 Swinbank AM, Smail I, Longmore S et al (2010) Nature 464:733 Casey CM, Narayanan D, Cooray A (2014) Phys. Rep. 541:45 Ly C, Malkan MA, Hayashi M et al (2011) ApJ 735:91 Steidel CC, Adelberger KL, Shapley AE et al (2003) ApJ 592:728 Steidel CC, Shapley AE, Pettini M et al (2004) ApJ 604:534 Daddi E, Cimatti A, Renzini A et al (2004) ApJ 617:746 Yuma S, Ohta K, Yabe K (2012) ApJ 761:19 Ivison RJ, Greve TR, Serjeant S et al (2004) ApJS 154:124 Reddy N, Dickinson M, Elbaz D et al (2012) ApJ 744:154 Trentham N, Kormendy J, Sanders DB (1999) AJ 117:2152 Goldader JD, Meurer G, Heckman TM et al (2002) ApJ 568:651 Meurer GR, Heckman TM, Calzetti D (1999) ApJ 521:64 Takeuchi TT, Yuan F-T, Ikeyama A, Murata KL, Inoue AK (2012) ApJ 755:144 Bouchet P, Lequeux J, Maurice E, Prevot L, Prevot-Burnichon ML (1985) A&A 149:330 Franco M, Elbaz D, Béthermin M et al (2018) A&A 620:A152

Chapter 7

Discussion

7.1 Insterstellar Medium Scale 7.1.1 Do Mergers Trigger Dusty Starbursts? 7.1.1.1

Close Companion and Star-Formation Mode

In Sect. 3.2.4, we find that about a half of our ALMA sources have the dusty restframe FIR-emitting regions largely separated from those of the rest-frame UV and/or optical emission. This may suggest a possibility of the dusty starbursts triggered by the major mergers. For testing the possibility, we investigate the rest-frame UV and optical morphology and the star-formation mode of our ALMA sources. Given that the pre-selection bias in the rest-frame UV and optical bands exists in the archive data, we use the OC5S-mmT sources whose central targets in the initial ALMA observations are selected in the submm/mm bands (Sect. 3.2.1). Firstly, the OC5S-mmT sources are cross-matched with the 3D-HST sources [1]. The high spatial resolution of the HST images is required to obtain homogeneous and reliable morphology results. We identify that 56 OC5S-mmT sources in the 3DHST source catalog. In Fig. 7.1, we show the false-color HST images of these 56 OC5S-mmT sources. Secondly, we classify the morphology of the 56 OC5S-mmT sources and test whether they are the major mergers or not. If the HST map shows major merger pairs of the OC5S-mmT source, we regard it as the major mergers. For the major merger pair identification, we perform the following three steps: (i) selecting the optical-NIR objects the spatial offset within a radius of 2. 5 from the OC5S-mmT, (ii) choosing the optical-NIR objects with the Δz phot < 1 from the object selected in (i), and (iii) identifying the optical-NIR objects with Mstar greater than 10% of that of the object selected in (ii). Note that the radius of 2. 5 in (i) is equal to ∼20 kpc at z = 2.5, which is used in Le Fèvre et al. [3] to classify the major merger system. Even with a larger radius of 6 (∼50 kpc at z = 2.5) in (i), we confirm that no additional OS5S-mmT sources are identified, which ensures that we have not missed © Springer Nature Singapore Pte Ltd. 2021 S. Fujimoto, Demographics of the Cold Universe with ALMA, Springer Theses, https://doi.org/10.1007/978-981-16-4979-0_7

119

120

7 Discussion

Fig. 7.1 5 × 5 fake-color HST images of the 56 OC5S-mmT sources that are identified in the 3D-HST regions (red: H160 , green: J125 , blue: I814 ). The red contours indicate ALMA mm band intensity from the 5 to 30σ levels with a 2σ -level step. The red cross represents the ALMA source center. The red ellipse shows the synthesized beam size. The white circle denotes the 2. 5 search radius for the optical-NIR objects. The large blue cross presents the optical-NIR counterparts of the ALMA sources, while the small blue cross indicates the major merger pair. Note that ID661 is placed at the edge of the coverage of the HST observations. The figure is reproduced from Fig. 15 of Fujimoto et al. [2] by permission of AAS

7.1 Insterstellar Medium Scale

121

any potential major mergers with our classification. We find that the major merger pairs exist in 27% (=15/56) of the OC5S-mmT sources. Because our classification of (i)−(iii) is limited by the spatial resolution of HST/H -band (0. 18) with which cannot identify close major merger pairs, we regard these 27% of the OC5S-mmT sources as “early/mid-stage major merger” We refer to the rest of 73% (=41/56) of the OC5S-mmT sources as “isolated galaxy”. Thirdly, we investigate the star-formation modes of the isolated galaxies and the early/mid-stage major mergers. In the left panel of Fig. 7.2, we show the Mstar − SFR relation for the 56 OC5S-mmT sources. In the right panel, we present the histograms for the isolated galaxies and the early/mid-stage major mergers as a function of ΔMS, where we define ΔMS as the offset from the main sequence of the SFR–Mstar relation with a unit of sSFR. We find that the early/mid-stage major mergers has a histogram similar to that of the isolated galaxies. The KS-test result for these two histograms shows that we cannot rule out the possibility that the same parent sample produces the samples of the early/mid-stage major mergers and the isolated galaxies. A possible explanation for the the similar histograms is that the isolated galaxies are also dominated by the major mergers such as late-stage one. To further examine whether the morphology of the isolated galaxies show the hint of the mergers, we cross-match the isolated galaxies with the detail morphology catalog of HuertasCompany et al. [4] that evaluate the probabilities of the five classes of disk, spheroid, irregular, compact, and unclassifiable based on the Convolutional Neural Network technique. Here, we define the objects with the irregular probability over 50% as

Fig. 7.2 Left: Our ALMA sources in the Mstar −SFR plane for investigating the star-formation mode (Fig. 16 of [2], reproduced by permission of AAS). The circles represent the 56 OC5SmmT sources whose optical-NIR counterparts are detected in the 3D-HST regions. The number of the circles indicates the number of the multiple components in the major merger system that is indicated by the rest-frame UV-optical morphologies. The solid lines and the shade regions are the main sequences of the Mstar −SFR relation and the associated 1σ errors, respectively, at z = 1−2, 2−4, and 4−6 that are estimated by Speagle et al. [5]. The colors correspond to the redshift ranges in the color-bar scale. Right: Histograms of the isolated galaxies (bottom) and the major mergers (top) as a function of ΔMS, where ΔMS is the difference in sSFR from the main sequence

122

7 Discussion

the merger-like morphology. The 25 isolate galaxies are studied in the morphology catalog, and we find that only 4 out of the 25 isolated galaxies are classified as the merger-like morphology. Even if we change the threshold of the irregular probability, we confirm that the fraction of the the merger-like morphology among the isolated galaxies is not changed significantly. This indicates that the isolated galaxies are not dominated by the late-stage mergers. If we combine our selections of the isolated galaxies with the merger-like morphology and the early/mid-stage major mergers, only ∼30−40% of our ALMA sources are classified as the major mergers. We thus conclude that not only major merger, but also other mechanism(s) trigger the dusty starburst. We caution again that the spatial resolution of the data regulate the results of the visual-based morphology classification. Huertas-Company et al. [4] conduct the morphology classification based on the HST/H -band images with the spatial resolution of 0. 18. On the other hand, recent ALMA observations achieve ultra-high resolutions of 0. 015 − 0. 05 for high-z dust starbursts [6, 7]. Although the wavelengths are different between our analysis and these recent ALMA observations, we may obtain some hints for understanding the mechanisms of the dusty starburst. The ALMA ultra-high resolution maps unveil that four out of five high-z SMGs are resolved into several small clumps down to ∼200−300 pc scales with the high SFR density of ∼300−3000 M yr−1 kpc−2 within the central kiloparcsec. Since such high SFR density values are comparable to those of the local major merger of Arp220, the main trigger can be explained by the major mergers [7]. Although the origins of the compact clumps identified in the ultra-high resolution maps are still under debated (cf. [8]), there recent ALMA results suggest the possibility that some of the isolated galaxies are also resolved into the merger-like morphologies after all by future ultra-high resolution observations.

7.1.1.2

Morphological Classification

We also check the L IR dependence of the detail rest-frame optical morphological classification for our ALMA sources. To carry out the fair comparison based on the homogeneous dataset, we use the ASAGAO sources at z = 1 − 3 (Chap. 4). For the sake of statistically reliable results, we also verify ALMA sources identified in previous studies in GOODS-S [9, 10], and apply the same morphological classification to these ALMA sources as described in Sect. 4.1.3. The entire ALMA sample combining the ASAGAO and the previous ALMA sources span the L IR range of ∼1011 − 1013 L  , where we divide the whole sample into two sub-samples: the high-z ULIRG (L IR > 1012 L  ) and LIRG (L IR ≤ 1012 L  ) samples. In Fig. 7.3, we show the histograms of the rest-frame optical morphology for the (U)LIRG samples. We also present the same histogram for the optically-selected galaxies classified in Kartaltepe et al. [12]. We find that the LIRG sample have higher and low fractions of the disk and the irregular/merging galaxy, respectively, than the ULIRG sample. The lower fraction of the irregular/merging galaxy indicates that the dusty star formation in the LIRG sample is not mainly triggered by the on-going

7.1 Insterstellar Medium Scale

123

Fig. 7.3 Histograms of the rest-frame optical morphology with the deep HST/H -band images for the ALMA sources at z = 1 − 3 (Fig. 5 of [11], reproduced by permission of AAS). The morphological classification contains six classes: spheroid (sph.), disk, irregular (irr.), point source (ps.), merger (mer.), and interacting (int.) as in Kartaltepe et al. [12]. The magenta and red histograms indicate the high-z ULIRG (L IR > 1012 L  ) and LIRG (L IR ≤ 1012 L  ) samples, respectively. The error-bars denote the Poisson uncertainty presented in Gehrels [13]. The number of galaxies in the ULIRG and LIRG samples are 23 and 19, respectively. The gray histogram is obtained from the entire sample of the >50,000 optically-selected galaxies (H < 24.5) in the morphology catalog of Kartaltepe et al. [12]. Note that the sum of the percentages exceeds 100% because the morphological classifications are not mutually exclusive. The magenta histogram is slightly shifted along the x-axis for clarity

merger processes. In fact, the fraction of disk galaxy in the LIRG sample is consistent with that of the optically-selected galaxies within the errors. These results imply that dusty star formation in the LIRG sample at high-z generally occurs in a secular star-formation mode in the regular disk galaxies. In contrast, the ULIRG sample has shows higher fractions of the irregular/merging galaxy than the other samples. This indicates that the on-going merger process plays a key role for triggering the dusty star formation in the ULIRG sample. Nonetheless, the disk galaxy is the most abundant population even in the ULIRG sample, implying that other mechanisms than the on-going merger process also cause the dusty star formation in the high-z ULIRGs. Note again that the rest-frame optical morphology classification is limited by the spatial resolution of HST/H -band (∼0. 18). In both LIRG and ULIRG samples, the objects classified as the disk galaxy might be resolved into other morphology in future higher resolution observations.

124

7 Discussion

Fig. 7.4 Size−stellar mass relation for our ALMA sources and other galaxy populations (Fig. 13 of [2], reproduced by permission of AAS). The red open circles present the Re values in the rest-frame FIR for the OC10S sources. The blue and black circles denote the Re values in the rest-frame optical wavelength for the star-forming and quiescent galaxies at z = 1−3, respectively, in the 3D-HST regions obtained by Shibuya et al. [14]

7.1.2 Size and Stellar Mass Relation In Sects. 3.2.5 and 4.2.3, we find that our ALMA sources have the rest-frame FIR size smaller than the rest-frame UV and optical sizes. To understand the role of the dusty star-forming galaxies in the context of the galaxy evolution, we examine the Re as a function of Mstar among the different wavelengths and galaxy populations. Figure 7.4 presents the relation between Re and Mstar . For the dusty star-forming galaxies, we plot the OC10S sources with the Re values in the rest-frame FIR wavelength in red circles. For comparison, we also plot the quiescent and the star-forming galaxies at z = 1−3 with the Re values in the rest-frame optical wavelength in black and blue circles, respectively, that are obtained from the recent HST results [14]. The black and blue circles show the sequence of the quiescent and star-forming galaxies, where the latter is placed below the former on the Re −Mstar plane. We find that there is a good agreement between the majority of our ALMA sources and the sequence of the quiescent galaxies, albeit the wavelengths are different. Given that the abundant stars produced in the intense star-formation in the dusty starburst activities likely form the major part of the future stellar mass distribution in the host galaxies, it might be reasonable that our ALMA sources and the quiescent galaxies have the similar Re values in the rest-frame FIR and optical wavelength that represent the current star-formation and the stellar mass distribution, respectively. This connection is also well aligned with the evolutionary scenario from the high-z dusty starbursts, the compact quiescent galaxies at z ∼ 2, and to the local elliptical galaxies. (e.g., [15–22]). Interestingly, we also find some of our ALMA sources that fall between the sequences of the quiescent and the star-forming galaxies. We may be witnessing

7.1 Insterstellar Medium Scale

125

the transition phase from the star-forming to the quiescent galaxies by the rapid mass assembly via the dusty starburst.

7.1.3 Evolutions of Size and Morphology In Sect. 7.1.2, we discuss the possibility of the direct connection from the dusty starforming to compact quiescent galaxies. However, it is required to test whether the radial profiles of the total star formation can explain these size and morphological evolution among the high-z star-forming galaxies. In this Section, we derive the evolution track of the galaxy size and morphology in the rest-frame optical wavelength and discuss the evolutionary path from the high-z dusty star-forming galaxies to the compact quiescent galaxies. For long years, it has been suggested an evolutionary path from high-z dusty starforming galaxies, the compact quiescent galaxy (cQG) phase at redshift z ∼ 1−2, and to local elliptical galaxies (e.g., [2, 8, 15–23]). In Sect. 4.2.2 we find that the dusty star-forming galaxies generally have the exponential-disk profile in the rest-frame optical wavelength of Re,opt ∼ 3 kpc and n opt ∼ 1, while the cQGs are generally have a different shape of the spheroidal profile of Re,opt  1 kpc and n opt ∼ 4 (e.g., [14, 24, 25]). To confirm the evolutionary path between the high-z dusty star-forming galaxies and the cQGs, we need to explain the increasing and decreasing trends of n opt and Re,opt , respectively, at the same time. Several evolution mechanisms have been discussed in the literature, such as “compaction” of the gas in star-forming galaxies due to disk instabilities [26], mergers of gas-rich galaxies (e.g., [17]), and inside-out growth of compact progenitors (e.g., [27, 28]). Here we alternatively test another evolutionary mechanism: the star-forming activity in the dusty star-forming galaxies alone regulates the evolutions of n opt and Re,opt . We caution that high SFRs even up to an order of ∼1000 M /yr could arise in the dusty starburst systems, but the depletion time scale is generally as short as ∼100−200 Myr (e.g., [29]) in those intense star-formation activities. On the other hand, normal star-forming galaxies could keep a secure star-formation with a long depletion time of ∼1 Gyr (e.g., [30]). This indicates that the the majority of the stars may be produced by the secular star-formation, instead of the intense star-formation (e.g., [31–34]). We thus investigate the evolutions of n opt and Re,opt also for the high-z LIRGs (L IR = 1011 −1012 L  ), in addition to the high-z ULIRGs (L IR > 1012 L  ), which we regard as the galaxies under the secular star-formation and the intense modes, respectively. To obtain the average properties of the high-z (U)LIRGs, we classify the 33 ASAGAO sources at z = 1 − 3 into two L IR samples of > 1012 L  and L IR ≤ 1012 that we regard as the representative samples of the high-z ULIRGs and LIRGs, respectively. For the average properties in the rest-frame optical and FIR wavelengths, we calculate the mean values of Mstar and L IR for the high-z ULIRG and LIRG samples. Following the average L IR values, we evaluate the general relation between L IR and S´ersic profiles in the rest-frame optical and FIR wavelengths from our

126

7 Discussion

best-fit results in Sect. 3.2. To evaluate the average property in the rest-frame UV wavelength, we use the deep HST/F606W image in GOODS-S and conduct the image-based stacking for the 33 ASAGAO sources. Because the massive galaxy centers are generally represented by the rest-frame FIR emission rather than the restframe UV emission (e.g., [22, 35]). we adjust the source centers for the stacking based on the ASAGAO source centers measured in ALMA Band 6. We then obtain the average values of the un-obscured rest-frame UV luminosity (L UV ), effective radius (Re,UV ), and S´ersic index (n UV ) from the best-fit S´ersic profile of the stacked HST/F606W image with galfit. We also calculate the un-obscured (SFRUV ) and the obscured SFR (SFRIR ) from the average L UV and L IR with the equation in Straatman et al. [36] of SFRUV = 2.4 × 10−10 L UV (L  ), SFRIR = 1.09 × 10

−10

L IR (L  ).

(7.1) (7.2)

The left panel of Fig. 7.5 shows radial surface density profiles of Mstar (Σ Mstar ), SFRIR (ΣSFRIR ), and SFRUV (ΣSFRUV ) for the high-z LIRGs. We calculate the summation of ΣSFRUV and ΣSFRIR and also show the radial surface density profile of the total SFR (ΣSFRUV+FIR ). The radial profile of Σ Mstar is derived from the rest-frame optical S´ersic profile with HST/F160W-band (Sect. 4.2.2), based on the assumption that the source centers are the same in the rest-frame FIR and optical wavelengths. In these radial surface density profiles, we find in both high-z ULRIGs and LIRGs that the central star-forming region is dominated by ΣSFRIR instead of ΣSFRUV . Based on these radial profiles with assumptions of no radial motions of the stars and a constant SFR during a time scale Δt, we calculate the sum of ΣSFRUV+FIR ×Δt and Σ Mstar and evaluate the evolutions of n opt and Re,opt for the high-z ULRGs and LIRGs. We assume two cases for Δt. The first case is the depletion time. From the gas fraction ( f gas ) as a function of Mstar (e.g., [37, 38]), we estimate the gas mass (Mgas ) and calculate the depletion times to be ∼210 and 640 Myr for the high-z ULIRGs and LIRGs, respectively. The second case is the cosmic time of ∼1.0 Gyr from z = 2 to 1.5. This is because the star-formation could continue in a longer time than the depletion time owing to the gas supply from inflows. In fact, f gas is suggested to be unchanged down to z ∼ 1 by the recent bathtub model results, due to the the balance of the gas consumption by the star-formation and the gas inflow/outflow (e.g., [26]). In the middle and right panels of Fig. 7.5, we show the results of the evolutions of Re,opt and n opt for the high-z LIRGs and ULIRGs in our model. For comparison, the distributions of z ∼ 1−2 quiescent and z ∼ 2 − 3 star-forming galaxies are also shown with the black and blue shaded regions, respectively. In the middle panel, we find that both high-z LIRGs and ULIRGs reproduce the decreasing trend of Re,opt well. On the other hand, we find in the right panel that the increasing trend of n opt is not reproduced. Since our model for the evolutions of Re,opt and n opt assumes no radial motions of the stars, these results suggest that forming the spheroidal profile

7.1 Insterstellar Medium Scale

127

Fig. 7.5 (Left) Radial surface density profile of SFR and Mstar for the high-z LIRGs. The red and blue dashed lines represent the radial profile of ΣSFRIR from our best estimates in the rest-frame FIR wavelength (Sect. 4.2.1) and ΣSFRUV from the stacking result with the deep HST/F606W image (see text), respectively. The black line indicates ΣSFRUV+FIR that is obtained by combining the red and blue dashed lines. The black shade denotes Σ Mstar estimated from our best estimates in the rest-frame optical wavelength (Sect. 4.2.2). (Middle) Re,opt evolution of the high-z (U)LIRGs. The red and magenta filled squares show the Re,opt evolutions of high-z ULIRG and LIRGs from z = 2 to z = 1.5 under the constant star-formation. The red and magenta open squares present the same Re,opt evolutions, but under the constant star-formation until the depletion time scales. The red and magenta lines indicate the evolutional tracks of ULIRG and LIRG, respectively. The blue and black shaded regions are 16th–84th percentiles of z = 1.5−3.0 star-forming (SFGs) and z = 0.5−2 quiescent galaxies (QGs) obtained from Shibuya et al. [14]. (Right) n e(opt.) evolution of high-z (U)LIRGs. The symbols and color assignments are the same as in the middle panel. The figure is reproduced from Fig. 6 of Fujimoto et al. [11] by permission of AAS

of n opt ∼ 4 requires other mechanisms such as dynamical dissipation from the starforming to quiescent galaxies. In fact, a cQG recently discovered at z = 2.1478 turns out to be a rotationally supported, fast-spinning galaxy with n opt = 1.01+0.12 −0.06 [39]. Unless the angular momentum in a post star-forming galaxy that had an exponentialdisk profile is lost by any dynamical dissipating events, it could be reasonable that the stellar distribution maintains the exponential-disk profile of n opt ∼ 1. Our model and the recent observational results consistently imply the existence of another step from the star-forming to quiescent galaxies to form the spheroidal stellar distribution. We caution that the initial Mstar values in our model are massive relative to the mainsequence of the star-forming galaxies at z ∼ 2. There remains a possibility that less massive, normal and/or dusty star-forming galaxies at higher redshifts are more likely the direct progenitor of the cQGs rather than z ∼ 2 dusty star-forming galaxies used in our model. The future deep observations will examine the evolutions of size and morphology from the less massive, normal and/or dusty star-forming galaxies at higher redshifts to fully understand the origin of the cQGs.

7.1.4 Compact Dusty Bulge Versus AGN In this Section, we discuss the origins of the compact component that we identify at the stacked source center for our ASAGAO sources (Sect. 4.1.5). In the right

128

7 Discussion

panel of Fig. 4.1, we find a slight offset between the best-fit and the stacked profiles near the center (radius  0. 1). The offset likely follows the ALMA beam profile, indicating that this offset is explained by an unresolved compact component at the center. Interestingly, Hodge et al. [8] also find in the high-resolution ALMA maps that the spatial profiles in some of bright SMGs also agree with a combined profile of Gaussian + point source (PS). The central compact component could be explained by the following two possible origins. One possibility is an AGN. In recent ALMA studies for local galaxies, a radius of ∼150-pc scale dust-continuum emission has been observed from AGN and surrounding regions (e.g., [40]). In our ALMA HR map, the half-width-halfmaximum (HWHM) of the ALMA beam is equal to ∼800 pc at z = 2. If the spatial distribution of the dust continuum from the AGN and surrounding regions is not significantly changed between z = 0 and z = 2, the central AGN cannot be resolved in our ALMA HR map, and thus observed as a PS profile. The other possibility is a compact dusty bulge with a spheroidal (n FIR ∼ 4) profile, probably related to the bulge formation at the center of the galaxy. To understand which origin is more likely to explain the compact component at the center of the galaxy, we perform a two-component fitting for the 1. 6 × 1. 6 HR map of the stacked source by using galfit. We model two different profiles of PS+disk and bulge+disk. In both cases, the initial parameters for the disk component are taken from the best-fit values obtained from the single S´ersic profile fitting (Sect. 4.1.5). To obtain stable results, we fix Re of the disk component and the central position of each component at the image center. In the best-fit results, we find that the two different models produce almost the same minimum chi-square values. However, we find that Re of the bulge component in the bulge+disk model reaches the minimum value in the fitting of 0. 0001, indicating that the best-fit bulge+disk model can be regarded as the PS+disk profile. Based on the best-fit result with the PS+disk model, we estimate the contribution of the PS component to the total flux density to be 1.5 ± 0.5%. Interestingly, recent studies 16 ALMA sources report a contribution of the AGN to the total flux density at the ∼1-mm band to be ∼0.5% from the SED decomposition of the star-formation and AGN [10], which is consistent with our above estimate within several factors. Moreover, Ueda et al. [41] report that a majority (∼70%) of ALMA mm-selected galaxies in the L IR range of 1011.5 −1012.8 L  are classified as X-ray AGNs. These results suggest that the compact component is likely explained by the central AGN. Although we cannot rule out a possibility that a PS-like very compact dusty bulge is formed at the center, it will be concluded in the future deep and high-angular resolution observations with ALMA.

7.2 Circumgalactic Medium Scale In Chap. 5, we discover the [C ii] halo around the normal star-forming galaxies at z ∼ 6. In the following subsections, we discuss the physical origin of the [C ii] halo.

7.2 Circumgalactic Medium Scale

129

7.2.1 Possible Origins of [Cii] Halo In Sect. 5.2.1, we find that the [Cii] line emission extends to the ∼10-kpc scale surrounding around the normal star-forming galaxies at z = 5 − 7. We also find that the radial profile of the extended [C ii] line is potentially related to the Lyα halo. In contrast to the previous reports of the 10-kpc scale carbon abundance around the dusty starburst galaxies at z ∼ 2 (e.g., [42, 43]), our results indicates the discovery of the universal cold carbon gas halos around early normal galaxies. To understand the physical origin of the [C ii] halo potentially associated with the Lyα halo, it is required to investigate the [C ii] halo from the following two aspects: what supports the [C ii] (and Lyα) line emissivity in the CGM? How the carbon abundance become enriched in the early phase of the CGM? First, we focus on the first question of the [C ii] line emissivity. The observational and theoretical studies suggest following five possible scenarios: (A) Satellite galaxies; (B) Circumgalactic (CG) PDR; (C) Photoionization; (D) Cold streams; and (E) Outflow. We illustrate these five possible scenarios in Fig. 7.6. The first scenario explains that the sum of satellite galaxies causes the extended morphology (Fig. 7.6A). If there are satellite galaxies around the central galaxies, the Lyα and [Cii] emission lines from these satellite galaxies could be observed as an extended structure around the central galaxy with a limited spatial resolution. In this scenario, the spatial distribution of the satellite galaxies defines the extended halo size and explains both extended components of the Lyα and [C ii] lines. It may be also possible that these satellite components are the tidal tails produced by the past merging events. The second scenario is a PDR extended over CG scale (Fig. 7.6B). Massive stars produce the ionizing photons, hν > 13.6 eV, and form the HII region on the central galactic scale. On the other hand, far-ultraviolet (FUV) photons, 6 eV < hν < 13.6 eV, penetrate the surrounding ISM deeper than the ionizing photons and make the PDR distribute more widely than the HII region. Then, the [Cii] line emission could be detected in the CG, if the PDR extends over the CG scale. Besides, the resonance scattering by the neutral hydrogen in the surrounding ISM (e.g., [45]) makes the Lyα line emission similarly spatially extended. The third scenario is photoionization (Fig. 7.6C). This scenario is similar to scenario (B), but the HII region and the surrounding PDR is much more extended due to an existence of strong ionizing sources and/or the ISM properties differ from scenario (B). In this case, the Lyα line emission is extended due to the fluorescence (e.g., [46]), instead of the resonance scattering in scenario (B). Although the carbon may be doubly ionized with less [Cii] line emission near the ionizing source centers in the highly ionized ISM, such deficit of the [Cii] line emission is consistent with the recent ALMA results that many star-forming galaxies at z > 5 do not always show clear [Cii] line detections at the stellar-continuum positions (e.g., [47]). The fourth scenario invokes cold streams (Fig. 7.6D). The intense star-formation in high-z galaxies is suggested to be fed by a cold and dense gas (∼104 K), dubbed cold streams, in cosmological hydrodynamical simulations (e.g., [48]). The cold

130

7 Discussion

Fig. 7.6 Illustrations of five possible origins of the [C ii] line emissivity in the CGM, with the potential association of the Lyα halo (Fig. 12 of [44], reproduced by permission of AAS). A satellite galaxies; B circumgalactic (CG) photodissociation region (PDR); C photoionization; D cold stream; and E outflow. The blue and red shades show the neutral and ionized hydrogen in ISM and CGM. The yellow stars represent the star-forming regions. The inner and outer dashed circle denote the effective radii (Re ) of the central and halo components of the [Cii] line emission, respectively

streams radiate Lyα as well as [Cii] emission lines powered by gravitational energy, and produce the extended Lyα and [Cii] line structures around a galaxy. Moreover, the shock heating could be also caused by the cold stream, which also produces the Lyα and [Cii] emission lines. The last scenario is outflow (Fig. 7.6E). The ionized hydrogen and carbon powered by the star-formation and/or AGN feedback form the extended Lyα and [Cii] emission lines, where a part of these line emission could be contributed by the associated process of the shock heating. We note that we cannot rule out the possibility that our ALMA sources have the past AGN activity and/or on-going faint AGNs, even though we choose ALMA sources not reported as AGNs.

7.2 Circumgalactic Medium Scale

131

7.2.2 From Observational and Theoretical Results Hints from Observational Results: In the observational results, we find that the [C ii] line spatially extends more than both the rest-frame UV and FIR dust continuum up to a radius of at least ∼7 kpc beyond the errors (Fig. 5.5). If we assume a constant line emissivity of [C ii] at a given stellar continuum [49], a large gap appears between radial profiles of the stellar continuum and the [C ii] line. This indicates that the large part of the [C ii] line emissivity of the [C ii] halo is insufficiently explained by the stellar continuum. Although the central to halo areas have different [C ii] line emissivities at a given stellar continuum, such outer areas is expected to have a metallicity as low as ∼1% of the galaxy center [50]. Even if low-mass, faint satellite galaxies produce the stellar continuum at the halo areas, the mass−metallicity relation [51] likewise suggests the lower metallicity in the satellite galaxies than in the central galaxy. Since the metallicity at the halo areas is expected to be significantly lower than the central areas [50, 51], Because less metallicity makes the [C ii] line emissivity decreased at a given stellar continuum [52], the different [C ii] line emissivities are furthermore difficult to explain the [C ii] halo by the stellar continuum. Moreover, our stacking results show that the L [CII] /SFRtotal ratio becomes higher towards halo areas where the high ratios cannot be explained either by the local star-forming or dwarf galaxies (Fig. 5.7). In contrast, the dwarf galaxies show the decreasing L [CII] /SFRtotal ratio as a function of SFR. This would be because the low-mass (∼dwarf) galaxies maintain the harder UV radiation field than the central, massive galaxies due to the fact that low mass galaxies are generally compact [14, 25]. Not only the local studies, recent high-z observation results also suggest that the low mass galaxies have the high ionizing radiation field [53]. In the high ionizing radiation, the carbon is doubly or triply ionized, where the [C ii] line emissivity is decreased and, again, hard to explain the extended [C ii] line component by the satellite galaxies. Our observational results thus rule out the possibility of the scenario (A), supporting the rest of four scenarios of (B)−(E). Hints from Theoretical Results: In the zoom-in simulation results, the large part of the [C ii] line luminosity in the [C ii] halo is not reproduced (Fig. 5.10). If the current assumptions in the simulations for the [C ii] line emissivity calculations are correct, the existence of the extended [C ii] line requires additional physical mechanism(s) to explain it. There are two possible mechanisms that are not included in the calculation of the [Cii] line emissivity in the simulations, but may contribute to the existence of the extended [C ii] line. The first mechanism is the shock heating (e.g., [54]). Although the shock process is followed by the hydro-dynamical calculation in the zoom-in simulation, the computation of the ionized carbon abundance in cloudy does not include the shock heating. Since the shock heating is caused in the galaxy merging or inflow/outflow processes, the scenarios of (A), (D), and (E) are supported. The second mechanism is the past/on-going AGN activities. The strong ionizing source

132

7 Discussion

of the past/on-going AGN activities forms a large area of the HII region and the surrounding PDR. Moreover, the AGN feedback may cause the shock heating, which also contributes to the [Cii] line emissivity. In this case, the scenarios of (C) and (E) are supported. Although we cannot rule out the possibility that modifying any parts of the assumptions in the zoom-in simulation recovers the extended [Cii] line luminosity, these results may suggest a hint that the shock heating and/or AGN activities contributes to the [Cii] line luminosity in the extended halo component. Note that 8 out of 12 sources in the ALMA-HST sample are placed at 5 < z < 6, where the CMB effect is less than the zoom-in simulation results at z = 6.0 − 7.2. Because the CMB effect significantly reduces the diffuse [Cii] line emission (e.g., [55, 56]), the redshift range slightly higher than the ALMA-HST sample may cause the insufficient [Cii] line luminosity in the zoom-in simulation results. We summarize the possible scenarios for the physical origin of the [Cii] line emissivity in the [C ii] halo. In the observational results, we rule out the scenario (A). In the zoom-in simulation results, if the assumptions in the calculation for the [Cii] line emissivity are correct, all of the scenarios except for (B) are possible. The possible scenarios are thus (C), (D), and (E) with the current best estimates of both observational and theoretical results.

7.2.3 What Made the Primordial CGM Metal-Enriched? Although there remain several possible scenarios that can give rise to [C ii] emission in the CGM, it is important that that outflows are required in all cases to enrich the carbon abundance of the primordial CGM gas around the early galaxies. There are two modes of outflows, hot-mode and cold-mode outflows. The hot-mode outflow is defined as the outflow of ionized (hot) hydrogen gas that is heated by massive star/AGN radiation and supernova (SN) explosions. The cold-mode outflow is the outflow of neutral (cold) hydrogen gas that is pushed by the radiative and kinetic pressures given by massive-stars, AGNs, and SNe. Here, the majority of [Cii] line is emitted in neutral (cold) hydrogen gas clouds. Moreover, the cosmic time at z ∼ 5−7 (∼1 Gyr) is too short to produce the [CII] emitting cold halos from hot gas expelled by the hot-mode outflow via recombination. Our finding of the [Cii] halo in the 18 normal star-forming galaxies implies that outflows are dominated by cold-mode outflows in the early star-forming galaxies. Since we find the consistent radial surface brightness profiles between the Lyα and [Cii] emission lines (Sect. 5.2.4), the Lyα and [Cii] halos may have the same physical origin. Future deep observations of both Lyα and [Cii] emission lines for individual high-z galaxies will enable us to comprehensively understand the CGM metal enrichment also with the theoretical simulations that includes the radiative transfers of these emission lines.

7.3 Cosmic Structure Scale

133

7.3 Cosmic Structure Scale In this Section, we assess the dust obscured star-forming galaxies via comparisons with the results of our ALMA and the optical-NIR observations.

7.3.1 Comparison with the IR Luminosity Function 0≤z≤4 We first examine the reliability of our number counts, comparing with the IR luminosity function (LF). Gruppioni et al. [57] derive 11 IR LFs at 0 ≤ z ≤ 4 based on the sources selected in Herschel PACS observations at 70, 100, and 160 µm, and obtain L IR measurements by IR SED template fittings with the PACS and SPIRE (250–500 µm) photometric data. We estimate S1.2 mm of these sources at 0 ≤ z ≤ 4 from the L IR measurements, assuming the modified black body SED with βd = 1.8+0.2 −0.3 . If we adopt a constant Td , it causes a large uncertainty in the SED estimate due to the wide range of the L IR measurements. We use the L IR − Td relation [58] to estimate Td for the modified blackbody with, Td [K] = 1.06 × (log L IR [L  ] − 9.5)2 + 27.0.

(7.3)

This L IR − Td relation is obtained by our modeling with the data points of Fig. 3 in Hwang et al. [58]. Because the L IR range of the data points is L IR ≥ 109.5 L  , we adopt a constant value of Td = 27.0 at L IR < 109.5 L  . From the S1.2 mm measurements, we derive eleven number counts from the eleven IR LFs. We refer this process as IR LF analysis. Figure 7.7 presents the 1.2-mm number counts estimated from the IR LF analysis with the color filled regions. We extrapolate the 1.2-mm number counts below the limiting luminosity of each IR LF, and these extrapolations are color translucent regions in Fig. 7.7. Because most of the 1.2-mm number counts are extrapolations below 0.1 mJy, we perform following analyses at S1.2 mm ≥ 0.1 mJy. There are eight 1.2-mm number counts that include the extrapolations at S1.2 mm ≥ 0.1 mJy. We investigate the uncertainties of the extrapolations in these eight 1.2-mm number counts. In Gruppioni et al. [57], the shapes of these eight IR LFs (0.6 ≤ z ≤ 4.2) are determined by fixing the faint-end power law slope, αp , at the value of −1.2. This αp value of −1.2 is obtained in the IR LF measurement for the local IR LF at 0 ≤ z < 0.3. However, the value of αp should evolve with redshift. In fact, recent optical studies have shown that αp of the LFs for star-forming galaxies are significantly steeper at high-z than low-z. Bouwens et al. [59] report that the αp value of the LF for star-forming galaxies is αp ∼ −1.8 at z = 4. To evaluate the uncertainties of extrapolations in the maximum case, we fix αp = −1.8, and derive the best-fit function with the data points of the eight IR LFs in Gruppioni et al. [57]. From IR LFs to the 1.2-mm number counts for all of the IR LFs, we also estimate the uncertainty

134

7 Discussion

Fig. 7.7 Number counts at 1.2 mm estimated from the 11 IR LFs at 0 ≤ z ≤ 4 (Gr13; [57]). The color regions show the eleven 1.2-mm number counts. The translucent regions with dashed lines denote the extrapolations for the eight 1.2-mm number counts that have the limiting luminosities at S1.2 mm > 0.1 mJy. The black shade represents a total of these eleven 1.2-mm number counts, and the width of the black shade indicates the uncertainties. The red line with the gray shade presents our number counts at 1.2 mm with ALMA and the 1σ error obtained in Sect. 6.2.1

of the βd of the modified black body. The color filled and translucent regions consist of these uncertainties of αp and the βd . We do not include the dispersion of the L IR − Td relation for the uncertainty estimate because we aim to obtain average values statistically. The black shade in Fig. 7.7 indicates a total of 1.2-mm number counts that sum over these eleven 1.2-mm number counts. The black shade is composed of the uncertainties of the eleven 1.2-mm number counts. Within the errors, the black shade is generally consistent with the red line and the gray region that is our ALMA 1.2-mm number counts. We thus conclude that our ALMA 1.2-mm number counts are reliable. Using our reliable ALMA and the eleven 1.2-mm number counts, we also estimate the redshift distribution, n(z), for the faint ALMA sources. Here we first calculate the integrated fluxes of the eleven 1.2-mm number counts, and then obtain eleven contribution fractions to the integrated flux of our ALMA 1.2-mm number counts. These eleven contribution fractions are divided in four redshift bins at 0 ≤ z ≤ 4 with Δz = 1. These contribution fractions of the four redshift bins are our n(z) estimate. We refer this process as n(z) analysis. We first perform the n(z) analysis for the mm sources down to 1.0 mJy to compare with the n(z) estimate of SMGs in the literatures. The blue filled histograms in the top panel of Fig. 7.8 present the result of the n(z) analysis with S1.2 mm > 1.0 mJy. The total contribution fraction up to z = 4 is ∼90%. We assume that the remaining ∼10% is contributed from the mm sources at z > 4, and show this remaining fractions with blue open histograms divided in the bins at 4 < z ≤ 6. We find that the general shape

7.3 Cosmic Structure Scale

135

Fig. 7.8 Redshift distribution of SMGs (S1.2 mm ≥ 1.0 mJy; Top panel) and the faint ALMA sources (0.1 ≤ S1.2 mm < 1.0 mJy; Bottom panel) estimated from the n(z) analysis. The blue filled (open) histograms show the contribution fractions of of these mm sources at 0 ≤ z ≤ 4 (4 < z ≤ 6). The solid and dashed lines present the n(z) results of SMGs derived in Yun et al. [60] and Simpson et al. [61], respectively. The red open histogram indicates the top hat assumption of n(z) for the faint ALMA sources that we adopt through this paper

of our n(z) estimate is consistent with the previous n(z) studies for SMGs [60, 61]. We thus conclude that the n(z) analysis is reliable. We next explore the n(z) analysis for the faint ALMA sources in the flux range of 0.1 − 1.0 mJy, and show the result in the bottom panel of Fig. 7.8. We also illustrate the n(z) shape of the top hat assumption with the red histogram. This top hat assumption has the redshift range of 1 ≤ z ≤ 4 with the median redshift value at z med = 2.5, which we adopt in the SFR estimate (Sect. 6.2.3) and the clustering analysis (Sect. 6.2.4). In the bottom panel of Fig. 7.8, the n(z) dispersion is larger than that of SMGs in the top panel, while the n(z) shape is generally consistent with the top hat assumption. We conclude that n(z) of the faint ALMA sources might have wider distribution than that of SMGs, and that our results of the analyses with the top hat assumption are reliable.

7.3.2 Comparison with the Stellar Mass Function 0 ≤ z ≤ 8 Next, we investigate whether there are dust obscured star-forming galaxies missing in the optical-NIR observations. The dust obscured star-forming activity should be included in the total star formation rate, SFRtotal , given by SFRtotal = SFRIR + SFRUV [M /yr], where SFRIR (SFRUV ) is the dust-obscured (UV) star formation rate.

(7.4)

136

7 Discussion

We first estimate the SFRtotal values of galaxies at 0 ≤ z ≤ 8 from the Mstar measurements, assuming the stellar mass functions (SMFs) and stellar mass (Mstar ) −SFRtotal relation. For the SMF measurements, we use 7 SMFs of Muzzin et al. [62] that consist of star-forming and quiescent galaxies at 0 ≤ z ≤ 4. We also use five SMFs at 4 ≤ z ≤ 8 of Song et al. [63] who derive the SMFs from UV LFs and a median linear correlation between the rest-frame UV absolute magnitude at 1500 Å(MUV ) and Mstar in the logarithmic scale. For the Mstar −SFR relation, we adopt a best-fit function form of star-formation main sequences (MSs) that are estimated by 25 independent studies [5], SFRtotal = (0.84 − 0.026 × t) log Mstar [M ] − (6.51 − 0.11 × t),

(7.5)

where t is the cosmic age in Gyr. We then calculate the SFRIR values from the SFRtotal estimates using the dust attenuation, AUV , defined as AUV ≡ 2.5 × log(SFRIR /SFRUV + 1) = 2.5 × log(SFRIR /(SFRtotal − SFRIR ) + 1).

(7.6)

The AUV values are estimated with the Mstar − AUV relation of Pannella et al. [64], AUV [mag] = 1.6 × log Mstar − 13.5.

(7.7)

We convert the SFRIR to L IR values using the SFR−L IR relation [65], and then obtain the S1.2 mm values in the same manner as Sect. 7.7. From the S1.2 mm measurements, we derive twelve 1.2-mm number counts. We refer these process starting from the twelve SMFs as SMF analysis. The color filled regions in Fig. 7.9 illustrates the seven 1.2-mm number counts estimated from the seven SMFs at 0 ≤ z ≤ 4. The color translucent regions indicate extrapolations below the limiting Mstar of each SMF. Since the Mstar values are not directly measured in the five SMFs at 4 ≤ z ≤ 8, the five 1.2-mm number counts estimated from these five SMFs are also presented with the color translucent regions. The width of the color filled and translucent regions are composed of the uncertainty of the extrapolations and βd in the same manner as the previous section. There are three 1.2-mm number counts that include the extrapolations at S1.2 mm ≥ 0.1 mJy. These three 1.2-mm number counts are estimated from three SMFs at 2 ≤ z ≤ 4. Because the αp values of these three SMFs are poorly determined in Muzzin et al. [62], we fix the αp value at −1.5 from the high-z studies [63]. We do not include the dispersion of the Mstar - SFRtotal , Mstar - AUV , and SFR - L IR relations, because here again our propose is to obtain average values statistically. The black shade in Fig. 7.9 shows a total 1.2-mm number counts that sum over the twelve 1.2-mm number counts. The width of the black shade consists of the uncertainties from the twelve 1.2-mm number counts. In Fig. 7.9, we find that the total 1.2-mm number counts are consistent with our ALMA 1.2-mm number counts

7.3 Cosmic Structure Scale

137

Fig. 7.9 Number counts at 1.2 mm estimated from the SMF analysis using the 12 SMFs at 0 ≤ z ≤ 8 (Mu13; [62] and So15; [63]). The color filled regions show the seven 1.2-mm number counts from the SMFs of Mu13. The color translucent regions with the dashed lines denote the extrapolations for three out of the seven 1.2-mm number counts that have the limiting Mstar above 0.1 mJy. The color translucent regions and the dashed lines also indicate the five 1.2-mm number counts from So15. The black shade represent the total of these twelve 1.2-mm number counts, and the width of the black shade indicates the uncertainties. The red line with the gray shade presents our ALMA 1.2-mm number counts and the 1σ errors obtained in Sect. 6.2.1. The two origins of the gap between the total and our ALMA 1.2-mm number counts are illustrated as the characters of “A” and “B” with arrows (see text for the details)

within the uncertainties at S1.2 mm ≤ 0.4 mJy. This result suggests that most of the faint ALMA sources at S1.2 mm ≤ 0.4 mJy can be explained by the optical star-forming galaxies if the extrapolations for the faint end of the present SMFs are correct in the reasonable range. On the other hand, Fig. 7.9 also shows that there is a large gap between the total and our ALMA 1.2-mm number counts at S1.2 mm > 0.4 mJy beyond the uncertainties. There are two origins that cause this gap. First origin is star-burst galaxies that are outliers on the Mstar −SFR relation with high SFRtotal values. The higher SFRtotal values at a fixed Mstar produces higher S1.2 mm estimates than that of the normal star-forming galaxies on MS, which causes the 1.2-mm number counts shift to bright side in the SMF analysis. To evaluate this outlier effects in the maximum case, 50% of galaxies are applied with another Mstar −SFR relation that is scaled for SMGs (MS × 3; [66]). We then derive another total 1.2-mm number counts in the same manner as the SMF analysis. The black dashed line in Fig. 7.9 represents another total 1.2-mm number counts. We find that the gap still remains between another total and our ALMA 1.2-mm number counts even in the maximum case of the outlier effects. This result indicates that there exists another contributor of this gap which is the second origin. The second origin is dust obscured star-forming galaxies that have been missed in the optical-NIR observations. In fact, recent studies report that ∼20% of the SMGs

138

7 Discussion

have no optical-NIR counterparts (e.g., [9, 67]). Moreover, we also confirm that ∼40% of the faint ALMA sources have no optical-NIR counterparts in Sect. 6.2.5.1. Because the galaxies with high Mstar values are highly dust obscured based on the Mstar − AUV relation, these mm sources without optical-NIR counterparts might have high Mstar values. In this case, the shapes of the SMFs are not well determined at the high Mstar regime. We conclude that the gap at S1.2 mm > 0.4 mJy consists of the two origins: (A) star-burst galaxies and (B) the uncertainty of the SMF shapes at the high Mstar regime due to the dust obscured star-forming galaxies missing in the optical-NIR observations.

7.4 Cosmic Star-Formation Rate Density The evolution of the cosmic star-formation rate density (SFRD) has been measured at high redshifts (e.g., [59, 68, 69]). These measurements are mainly based on the optical-NIR observations. However, there is a possibility that optical observations miss some of dust obscured star-forming galaxies. In this case, the SFRD measurements are underestimated. The direct observations for dust obscured star-formation are needed to measure the SFRD values correctly. In Sect. 4, we find that the dusty star-formation occurs in a compact dusty disk embedded in a larger stellar disk. In Sect. 5, we discover the existence of the [C ii] halo around z ∼ 6 normal star-forming galaxies extended over the radius of ∼10 kpc. We discuss the physical origin of these [C ii] and dust continuum halos and conclude that these halos are not originated by the satellite galaxies. In Sect. 6.2.1, we derive the 1.2-mm number counts down to 0.02 mJy and almost fully resolve the CIB under the resolutions comparable to the ISM scale ( 0. 5–1. 0). These results suggest that the dust obscured star-formation is taking place in the ISM scale, instead of widely distributed outside of the galaxies in the CGM and/or IGM scales, and we succeed in identifying the entire population of the ISM-scale dust obscured star-formation contributing to the CIB with ALMA. In this Section, we investigate the total cosmic star-formation rate density including the dust obscured part that may have been missed in the previous optical-NIR observations. Note that we reveal the possibility that there exist the dust obscured star-forming galaxies missing in the optical-NIR observations in the differential sense in Sect. 7.3.2, while we evaluate these dust obscured star-forming galaxies in the cumulative sense from the SFRD measurements in this Section. Figure 7.10 shows the cosmic SFRD measurements estimated from previous optical-radio observations at 0 ≤ z ≤ 6 [59, 70–73]. The black solid and dashed lines with gray shade present the best SFRDtotal estimate of Hopkins and Beacom [68] who perform a functional model fitting to the previous SFRD measurements. To estimate the SFRDtotal values including the dust obscuration, Hopkins and Beacom [68] calculate a sum of the SFRDUV and SFRDIR value that is obtained with the direct IR observations at z ≤ 1. However, the SFRDtotal measurements at 1 < z ≤ 3 are performed by adding the SFRDIR value of z = 1 to the SFRDUV values at 1 < z ≤ 3.

7.4 Cosmic Star-Formation Rate Density

139

Fig. 7.10 Cosmic SFRD measurements from our ALMA 1.2-mm number counts and previous optical-radio observations. The red circle represents our SFRDtotal measurement at 1 ≤ z ≤ 4 estimated from our SFRDIR and the average of the previous SFRDUV values [59, 70, 72]. The black solid and dashed lines denote the previous best estimate of the functional model fitting to the SFRDtotal and SFRDcorr measurements at 0 ≤ z ≤ 6, respectively [68]. The gray shade is the uncertainty of the previous best-estimate SFRDtotal value, and the black circle presents the average of the previous best-estimate SFRDtotal value at 1 ≤ z ≤ 4. For presentation purposes, we slightly shift the black circle along the abscissa. The blue symbols are the SFRDUV measurements from the optical observations, while the orange symbols are dust corrected values estimated from the blue symbols. The blue and orange regions are the uncertainties of the SFRDUV and SFRDcorr values that are presented in Fig. 18 of Bouwens et al. [59]. The light-red region shows the SFRIR measurement estimated from the direct IR observations at z ≤ 1. The SFRD measurement from radio observations are also plotted in the green pentagons at z ≤ 1.3

Moreover, the SFRDtotal measurements at z > 3 is estimated just by applying the dust corrections to the SFRDUV values (SFRDcorr ). With these dust obscuration estimates at z > 1, the SFRDtotal values might miss the dust obscured star-formation. We thus measure a SFRDtotal value from our ALMA 1.2-mm number counts, and compare the result with the value of Hopkins and Beacom [68]. In the SFRDtotal measurement, we first estimate the SFRDIR value from the integrated flux of our −2 ALMA 1.2-mm number counts down to 0.02 mJy (22.9+6.7 −5.6 Jy deg ; see Sect. 6.2.3). We convert the integrated flux to an integrated L IR value with the assumption of z = 2.5, β = 1.8, and Td = 35K. We then estimate the SFRIR value from the integrated L IR value. From SFRIR to SFRDIR , we adopt the top hat assumption at 1 ≤ z ≤ 4 for the survey volume that is confirmed as reliable in the Sect. 7.3.1. We combine this SFRDIR value and a volume weighted average SFRDUV value at 1 ≤ z ≤ 4 from the literatures in Bouwens et al. [59]. We then obtain the SFRDtotal value at

140

7 Discussion

1 ≤ z ≤ 4, and plot the result in Fig. 7.10 with the red circle. A volume-weighted average SFRDtotal value of Hopkins and Beacom [68] at 1 ≤ z ≤ 4 is also plotted in Fig. 7.10 with the black circle to fairly compare our and the previous SFRDtotal values. In Fig. 7.10, we find the excess of our SFRDtotal value to the average of the previous best-estimate SFRDtotal value by 40+20 −40 %. Although there is no excess within 1σ error, this result provides us the same indication as the result in the SMF analysis that there might be the dust obscured star-forming galaxies missing in the optical-NIR observations. In fact, the light-red region in Fig. 7.10 from the direct IR observations shows the similar excess to the orange regain that presents the SFRDcorr values at z ≤ 1. Moreover, the simple steps in this SFRDtotal estimates include much less uncertainties than the uncertainties of the SMF analysis except for the uncertainty of the survey volume. To evaluate the uncertainty of the survey volume, we also calculate these SFRDtotal values with another top hat assumption at 0 ≤ z ≤ 6. We confirm that the excess still remains even with another top hat assumption because both of our and the average of the previous best-estimate SFRD values are systematically decreased according to the increase of the survey volume. We thus conclude that there exist the dust obscured star-forming galaxies missing in the optical-NIR observations with 1σ significance level. The error bars of our SFRDtotal value consist of the uncertainties from the average of the previous best-estimate SFRDUV and our SFRDIR measurements. The upper error of our SFRDIR measurement is defined by the upper limit of the CIB value from the COBE observation. If the number count studies with ALMA reveal the flattening and/or truncation at S1.2 mm  0.02 mJy (Sect. 6.2.3), we can set stronger upper limit than the present one. On the other hand, the lower error of our SFRDIR measurement is estimated from the integrated flux error of our ALMA 1.2-mm number counts down to 0.02 mJy. Note that we confirm the reliability of our ALMA 1.2-mm number counts only at S1.2 mm ≥ 0.1 mJy with the IR LF analysis in Sect. 7.3.1. We thus also include the difference of the integrated fluxes down to 0.02 mJy between our ALMA and the total 1.2-mm number counts as part of the lower error. This lower error and the center value of the SFRDIR measurements can be increased as the ALMA observations provide us deeper data than Slimit = 0.02 mJy. These upper and lower limit improvements with further ALMA observations will enable us to address the excess of the SFRDtotal values beyond 1σ level.

References 1. 2. 3. 4. 5. 6. 7.

Momcheva IG, Brammer GB, van Dokkum PG et al (2016) ApJS 225:27 Fujimoto S, Ouchi M, Shibuya T, Nagai H (2017) ApJ 850:1 Le Fèvre O, Abraham R, Lilly SJ et al (2000) MNRAS 311:565 Huertas-Company M, Gravet R, Cabrera-Vives G et al (2015) ApJS 221:8 Speagle JS, Steinhardt CL, Capak PL, Silverman JD (2014) ApJS 214:15 Speagle JS, Steinhardt CL, Capak PL, Silverman JD (2016) ApJ 822:36 Iono D, Yun MS, Aretxaga I et al (2016) ApJ 829:L10

References 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61.

141

Hodge JA, Swinbank AM, Simpson JM et al (2016) ApJ 833:103 Hodge JA, Karim A, Smail I et al (2013) ApJ 768:91 Dunlop JS, McLure RJ, Biggs AD et al (2017) MNRAS 466:861 Fujimoto S, Ouchi M, Kohno K et al (2018) ApJ 861:7 Kartaltepe JS, Mozena M, Kocevski D et al (2015) ApJS 221:11 Gehrels N (1986) ApJ 303:336 Shibuya T, Ouchi M, Harikane Y (2015) ApJS 219:15 Lilly SJ, Eales SA, Gear WKP et al (1999) ApJ 518:641 Genzel R, Baker AJ, Tacconi LJ et al (2003) ApJ 584:633 Tacconi LJ, Genzel R, Smail I et al (2008) ApJ 680:246 Hickox RC, Wardlow JL, Smail I et al (2012) MNRAS 421:284 Toft S, Smolˇci´c V, Magnelli B et al (2014) ApJ 782:68 Chen C-C, Smail I, Swinbank AM et al (2015) ApJ 799:194 Chen C-C, Smail I, Swinbank AM et al (2015) ApJ 799:81 Barro G, Kriek M, Pérez-González PG et al (2016) ApJ 827:L32 Barro G, Kriek M, Pérez-González PG et al (2015) ApJ 810:133 van Dokkum PG, Franx M, Kriek M et al (2008) ApJ 677:L5 van der Wel A, Franx M, van Dokkum PG et al (2014) ApJ 788:28 Dekel A, Mandelker N (2014) MNRAS 444:2071 Nelson E, van Dokkum P, Franx M et al (2014) Nature 513:394 Wellons S, Torrey P, Ma C-P et al (2015) MNRAS 449:361 Carilli CL, Walter F (2013) ARA&A 51:105 Tacconi LJ, Genzel R, Neri R et al (2010) Nature 463:781 Elbaz D, Dickinson M, Hwang HS et al (2011) A&A 533:A119 Rujopakarn W, Rieke GH, Eisenstein DJ, Juneau S (2011) ApJ 726:93 Wuyts S, Förster Schreiber NM, van der Wel A et al (2011) ApJ 742:96 Schreiber C, Elbaz D, Pannella M et al (2016) A&A 589:A35 Tadaki K-I, Genzel R, Kodama T et al (2017) ApJ 834:135 Straatman CMS, Spitler LR, Quadri RF et al (2016) ApJ 830:51 Tacconi LJ, Neri R, Genzel R et al (2013) ApJ 768:74 Dessauges-Zavadsky M, Zamojski M, Schaerer D et al (2015) A&A 577:A50 Toft S, Zabl J, Richard J et al (2017) Nature 546:510 Izumi T, Kohno K, Aalto S et al (2015) ApJ 811:39 Ueda Y, Hatsukade B, Kohno K et al (2018) ApJ 853:24 Ivison RJ, Papadopoulos PP, Smail I et al (2011) MNRAS 412:1913 Falgarone E, Zwaan MA, Godard B et al (2017) Nature 548:430 Fujimoto S, Ouchi M, Ferrara A et al (2019) ApJ 887:107 Xue R, Lee K-S, Dey A et al (2017) ApJ 837:172 Cantalupo S, Porciani C, Lilly SJ, Miniati F (2005) ApJ 628:61 Carniani S, Maiolino R, Amorin R et al (2018) MNRAS 478:1170 Dekel A, Birnboim Y, Engel G et al (2009) Nature 457:451 De Looze I, Cormier D, Lebouteiller V et al (2014) A&A 568:A62 Pallottini A, Ferrara A, Bovino S et al (2017) MNRAS 471:4128 Mannucci F, Cresci G, Maiolino R, Marconi A, Gnerucci A (2010) MNRAS 408:2115 Vallini L, Gallerani S, Ferrara A, Pallottini A, Yue B (2015) ApJ 813:36 Stark DP, Richard J, Siana B et al (2014) MNRAS 445:3200 Appleton PN, Guillard P, Boulanger F et al (2013) ApJ 777:66 da Cunha E, Groves B, Walter F et al (2013) ApJ 766:13 Zhang Z-Y, Papadopoulos PP, Ivison RJ et al (2016) Royal Society Open Science 3:160025 Gruppioni C, Pozzi F, Rodighiero G et al (2013) MNRAS 432:23 Hwang HS, Elbaz D, Magdis G et al (2010) MNRAS 409:75 Hwang HS, Elbaz D, Magdis G et al (2015) ApJ 803:34 Yun MS, Scott KS, Guo Y et al (2012) MNRAS 420:957 Simpson JM, Swinbank AM, Smail I et al (2014) ApJ 788:125

142 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73.

7 Discussion Muzzin A, Marchesini D, Stefanon M et al (2013) ApJ 777:18 Song M, Finkelstein SL, Ashby MLN et al (2015) arXiv:1507.05636 Pannella M, Elbaz D, Daddi E et al (2015) ApJ 807:141 Kennicutt RC Jr (1998) ARA&A 36:189 da Cunha E, Walter F, Smail IR et al (2015) ApJ 806:110 Chen C-C, Smail I, Ivison RJ et al (2016) arXiv:1601.02630 Hopkins AM, Beacom JF (2006) ApJ 651:142 Madau P, Dickinson M (2014) ARA&A 52:415 Schiminovich D, Ilbert O, Arnouts S et al (2005) ApJ 619:L47 Le Floc’h E, Papovich C, Dole H et al (2005) ApJ 632:169 Reddy NA, Steidel CC (2009) ApJ 692:778 Smolˇci´c V, Schinnerer E, Zamorani G et al (2009) ApJ 690:610

Chapter 8

Conclusion

In this chapter, we summarize our main results and discussions in the scales of the ISM, the CGM, and the cosmic structure.

Insterstellar Medium Scale In Chap. 3, we analyse 1627 deep ALMA maps in Band 6 and 7 that became public by 2017 July and study the statistics of the rest-frame FIR effective radius (Re(FIR) ). The sample is composed of 1034 continuum sources detected at the submm and mm wavelengths, which is the largest sample ever made with the ALMA data. We identify 577 optical-NIR counterparts with the photometric redshifts at z = 0 − 6 and confirm from the redshift distribution that there exists no pre-selection bias in the initial ALMA observations. We measure the Re(FIR) values for the large sample based on the homogeneous uv-visibility size analyses with the exponential disk model (n = 1) Evaluating the We carefully evaluate the completeness in the source selection and the size measurement via Monte-Carlo simulations and address the L FIR − Re(FIR) relation and the Re(FIR) evolution. We discuss the physical origins of the dusty starbursts by comparing our results with the rest-frame UV morphology and size. The main findings of this section are summarized below. 1. Our ALMA sources show the Mstar values of ∼1010 −1011.5 M and the SFR values of ∼100−1000 M yr −1 . We find that the half are the high-mass end of the main sequence, while the rest of the half of our ALMA sources are the starbursts. The fraction of the starburst is decreased at high redshifts, which is consistent with the previous studies. 2. The redshift distribution is unchanged as a function of L FIR nor between ALMA Band 6 and 7. We obtain the median redshift of z med = 2.36, which is consistent with the previous studies. © Springer Nature Singapore Pte Ltd. 2021 S. Fujimoto, Demographics of the Cold Universe with ALMA, Springer Theses, https://doi.org/10.1007/978-981-16-4979-0_8

143

144

8 Conclusion

3. Owing to the largest ALMA sample, we successfully derive the Re(FIR) − L FIR relation in a wide redshift range of z = 0 − 6, and find a positive correlation at the >99% significance level. 4. We obtain the best-fit slope of α FIR = 0.28 ± 0.07 via the power-law fitting of IR Re(FIR) ∝ L αFIR . We confirm that the best-fit slope is not affected by the contamination either of the lensed sources, the AGNs, or the potentially low-quality of early ALMA cycles. We also find the redshift evolution of Re(FIR) : Re(FIR) decreases toward high redshifts at a fixed L FIR . The best-fit α FIR and the redshift evolution trend agree with the previous studies for the galaxy effective radius in the rest-frame UV (Re(UV) ) and optical wavelengths (Re(Opt.) ). 5. On the statistical basis, the average Re(FIR) values at z = 1−6 are smaller than those of Re(Opt.) and Re(UV) at a given Mstar range. On the individual basis, the comparison between the Re(FIR) and Re(Opt.) (Re(UV) ) in individual galaxies also supports the trend of Re(FIR)  Re(Opt.) (Re(UV) ). The dusty starbursts are suggested to be taking place in the compact regions by these statistical and individual results. 6. The median value of the spatial offset between the rest-frame UV-optical and FIR continuum is estimated to be 0. 24, which might be be explained by the astrometric uncertainty (∼0. 25) between ALMA and HST reported in the previous studies. There are some sources with the spatial offsets between the rest-frame UV-optical and FIR continuum larger than the astrometric uncertainty, indicative of the intrinsic spatial offsets between the rest-frame UV and FIR star-forming regions, instead of the misidentification for the optical-NIR counterparts due to the chance projection. 7. We test whether our ALMA sources are major mergers based on the rest-frame UV and optical morphology with the deep HST images. We identify 27% of our ALMA sources as the early/mid-stage major mergers, while the rest of the 73% as the isolated galaxies. The total fraction of the major mergers including the isolated galaxies with the merger-like morphology is estimated to be ∼30−40% in our ALMA sources. This indicates that the dusty starbursts are triggered not only by the major mergers but also the other mechanism(s). In Chap. 4, we conduct the visibility-based stacking to the ALMA sources identified in the ASAGAO survey and further study the S´ersic index n and the effective radius Re in the rest-frame FIR wavelength, n FIR and Re,FIR . The ALMA highresolution (∼0. 19) 1.2-mm imaging of the ASAGAO survey covers the 26 arcmin2 area in GOODS-S imaged with WFC3/IR on HST, which also enables us to obtain Re and n in the rest-frame optical wavelength, Re,opt and n opt . By using an additional sample of individual ALMA sources from the archive, we examine the morphology of the dusty star-forming galaxies in the rest-frame optical and FIR wavelengths as a function of L IR over a wide range of ∼1011 − 1013 L  . We then discuss the evolutionary sequence from the high-z dusty star-forming galaxies to the compact quiescent galaxies. We summarize the major findings as follows. 7. The33 ASAGAO sources shows a 29σ level detection via the visibility-based stacking, which enables us to securely measure the average S´ersic profile. Carefully evaluating the smoothing effect due to the positional uncertainty via Monte-

8 Conclusion

8.

9.

10.

11.

12.

145

Carlo simulations, we obtain the best-estimates of Re,FIR = 1.0 ± 0.2 kpc and n FIR = 1.2 ± 0.2. By adding individual measurements for 12 bright ALMA sources in the archive, we evaluate n FIR and Re,FIR as a function of L IR in the range of ∼1011 − 1013 L  . We find that the n FIR measurements are unchanged from n FIR = 1.2 ± 0.2 by L IR . On the other hand, the Re,FIR measurements fall on the positive power-law correlation between L IR − Re,FIR reported in the previous ALMA studies. The distributions of n opt and Re,opt show no significant difference between the bright SMGs identified in the ALESS survey and the less bright ASAGAO sources. The statistical test results indicate that neither n opt and Re,opt depend strongly on L IR . We obtain the best-estimates of Re,opt = 3.2 ± 0.6 kpc and n opt = 0.9 ± 0.3 in the L IR range of ∼1011 − 1013 L  . Comparing the rest-frame optical and FIR results, we find that n always takes the common value of ∼1 regardless of the wavelengths. We also find Re,FIR < Re,opt , which agrees with the previous ALMA studies. These results suggest a general picture of the high-redshift star-forming galaxy: the dusty star formation takes place in a compact disk-like structure embedded in a larger stellar disk. The best-fit S´ersic profiles in the rest-frame FIR+UV and optical provides us the radial surface density profile of SFRUV+FIR and Mstar , respectively. Assuming the constant star-formation, our simple model reproduces the decreasing trend of Re,opt from z ∼ 2 dusty star-forming galaxies to z ∼ 1 − 2 compact quiescent galaxies. However, the increasing trend of n opt is not reproduced, where other mechanism(s) are required such as kinematic dissipation. In the HR map of the stacked ASAGAO source, we find that there is a compact component at the center. The two-component fitting results show that the stacked profile is explained by a point source (PS)+disk profile rather than a bulge+disk profile. The PS component is likely to be originated by the central AGN, while one cannot rule out a possibility that there is a PS-like very compact dusty bulge at the center.

In Sect. 7.1, we discuss the physical origins of the ALMA submm/mm sources and the role in the galaxy evolution based on the above results. 13. We analyse the rest-frame UV and optical morphology with the deep HST images and test whether major mergers dominate in our ALMA sources. We find 73% of our ALMA sources classified as the isolated galaxies, while the rest of 27% as the early/mid-stage major mergers. When we regard the isolated galaxies with the merger-like morphology also as the major mergers, the fraction is increased to ∼30−40% in our ALMA sources. This indicates that the majority of the dusty starbursts are not the major merger system and that other mechanisms are required to trigger the dusty starbursts. 14. We study the detail morphology in the rest-frame optical wavelength for the z ∼ 1 − 3 ALMA sources and find that on-going merger processes are likely important in high-z ULIRGs, but not in LIRGs. However, we caution the galaxies classified as the isolated disk are the most abundant even in the high-z ULIRGs,

146

8 Conclusion

which implies that the dusty star formation even in the ULIRGs at z = 1 − 3 is triggered not only by the on-going merger process. 15. On the size−stellar mass plane, Re(opt.) of the quiescent galaxies at z = 1 − 3 and Re(FIR) of our ALMA sources shows the similar distribution. This agrees with an evolutionary sequence from the high-z dust starburst to the local elliptical galaxies: the intense star-formation of dusty starbursts in compact regions forms the major part of the stellar distribution of the host that are evolved to the compact quiescent and the local elliptical galaxies. 16. From the best-fit S´ersic profiles in the rest-frame optical and UV+FIR, we derive the radial surface density profile of Mstar and SFRUV+FIR , respectively. The observed trend of Re,opt decreasing from dusty star-forming galaxies at z ∼ 2 to compact quiescent galaxies at z ∼ 1 − 2 is well reproduced from our simple model which assumes a constant star formation. However, the same model does not reproduce the increasing trend of n opt . This indicates that the increasing trend of n opt requires other mechanism(s) such as kinematic dissipation. 17. We find that there is a compact component at the center in the HR map of the stacked ASAGAO source. The two-component fitting results suggest that a point source (PS)+disk profile more likely explains the stacked profile rather than a bulge+disk profile. We interpret that the central PS component is attributed to the AGN, although we cannot rule out the possibility that a PS-like very compact dusty bulge is formed at the center.

Cicrumgalactic Medium Scale In Chap. 5 and Sect. 7.2, we study the existence of the carbon-enriched gas halos around normal star-forming galaxies whose [Cii] line have been individually detected at z = 5.153 − 7.142 via the visibility-based stacking with the deep ALMA data. Our stacking approach achieves the well-sampled deep visibility data in the uv-plane, which allows us to explore the diffuse emission spread over the circumgalactic scale. Also utilizing the deep HST/H -band data, we compare the radial profiles of the restframe UV and FIR continuum as well as the [Cii] line emission. In Sect. 7.2, We discuss what physical origins form the extended [Cii] line emission. We summarize the major findings as follows. 18. The visibility-based stacking of our and archival deep ALMA data produces 10σ and 21σ level detections at the peak for the dust continuum and the [Cii] line emission, respectively, as average properties of 18 galaxies with SFR  10−70M at z = 5.153 − 7.142. The stacked [Cii] line clearly shows its morphology extended more than the distribution of the dust continuum. In fact, the stacking results of the radial surface brightness profiles of the [Cii] line are extended up to radius ∼10-kpc scale (9.2σ ), which is consistent within the scatters of the individual results.

8 Conclusion

147

19. We also perform the rest-frame UV continuum stacking for 9 out of the 18 [Cii] line sources whose deep HST/H -band data have been taken in the archive. In addition to the rest-frame FIR dust continuum, the radial surface brightness profiles of the [Cii] line extends more than the rest-frame UV continuum emission. We examine the radial ratio of L [CII] /SFRUV+FIR and find that the ratio becomes higher towards the halo areas, where the high ratios cannot be explained by the star-forming or dwarf galaxies. This indicates that the [C ii] halo is not originated by the faint, low-mass satellite galaxies. 20. The two-component (S´ersic+exponential) profile fitting results indicate that the extended [Cii] component has a scale length of 3.3 ± 0.1 kpc (=5.6 ± 0.1 kpc in effective radius) that is comparable to the Lyα halo universally found around the high-z star-forming galaxies. Based on the comparison in the effective radius, the extended [Cii] component has a larger radius by a factor of ∼5 than that of the central galactic component. 21. The state-of-art zoom-in simulation reproduces the radial surface brightness profile trends of the extended [Cii] line emission and of the rest-frame UV comparable to the rest-frame FIR continuum emission. However, the zoomin simulation does not reproduce a large part of the [Cii] line luminosity in the extended component, where the zoom-in simulation may miss other mechanisms associated to the feedback process to form the extended [Cii] halo component in early galaxies. 22. The existence of [C ii] haloes around these early galaxies raises two questions: what powers the [C ii] line emission in the CGM and how is the carbon abundance enriched at such an early cosmic epoch. Although there remain several possible scenarios that are attributed to the extended [C ii] emission in the CGM scale, outflows are essentially required in all cases to enrich the carbon abundance of the primordial CGM gas around the early galaxies. Our results are thus the first evidence of outflow activities in these early galaxies. 23. The same stacking procedure is applied to another ALMA dataset for 34 QSOs at z = 5.78–7.09, which produces not only the [C ii] halo but also the rest-frame FIR continuum halo extended over the radius of 10–15 kpc scale. The radial ratios of L [CII] /SFRIR cannot explain the extended [C ii] and rest-frame FIR continuum, suggesting that the strong outflow (radiation) originated from the QSOs expels (heats) the dust into (in) the CGM and the thermal dust emission in the CGM is identified as the rest-frame FIR continuum halo.

Cosmic Structure Scale In Chap. 6, we analyse one cluster, Abell 1689, and fifty blank-field maps taken by ∼120 deep ALMA Band 6 and 7 pointings and identify 133 faint ALMA continuum sources at 1.2 mm with a flux density of 0.02 − 1 mJy. This sample explores the faintest regime in the 1.2 mm survey known to date reaching down to 0.02 mJy, which is drawn from the complete archive search for the deep ALMA data open for

148

8 Conclusion

public by 2015 June and our deep ALMA projects. Using this faint 1.2-mm source sample, we derive the 1.2-mm number counts and estimate the contribution of the faint-source to the CIB. In Sect. 7.3, we also examine the physical origin of these faint 1.2-mm sources based on the statistical approach of clustering analysis as well as the individual approach of studying optical-NIR counterparts. We summarize the major findings as follows. 24. We derive the 1.2-mm number counts down to a flux density of 0.02 mJy by combining the faint and bright number counts from our and the previous studies, respectively. The number counts are consistent with the previous results using the single-dish lensing and ALMA data, and significantly push the flux density limit to the faintest regime. We find that the Schechter function well represents the number counts at 0.02 mJy 25. Based on the derived number counts, we estimate the total integrated 1.2-mm −2 down to the flux limit of Slimit of 0.02 mJy. flux density to be 22.9+6.7 −5.6 Jy deg +31 This corresponds to 104−25 % of the CIB measured by COBE observations. Since the CIB is fully resolved down to ∼0.02 mJy within the errors, the major contributors to the CIB at 1.2 mm are the individual sources with 0.02 mJy. In other words, the contribution to the CIB of the 1.2-mm sources fainter than ∼0.02 mJy should be negligibly small, suggesting a potential flattening and/or truncation in the number counts at the further faint flux density regime below ∼0.02 mJy. 26. We analyse the clustering properties of our faint ALMA sources via the countsin-cells method, assuming the sources redshift at z ∼ 1 − 4. We constrain the galaxy bias with an upper limit of bg < 3.5. This is smaller than those of massive galaxy populations of SMGs and DRGs, but comparable with those of UV-bright LBGs and sBzK populations. 27. We search for optical-NIR counterparts of our faint ALMA sources, focusing in two regions of Abell 1689 and SXDS that have rich multi-wavelength opticalNIR images. We identify 17 optical-NIR counterparts and find that 59% of our faint ALMA sources have the counterparts that are detectable in the optical-NIR continuum (m  28 mag). The mid-IR diagnostics indicates that there are no AGN activities in these sources, which is confirmed from the absence of the X-ray or radio counterparts in any of our faint ALMA sources. We apply the color selection criteria of LBGs, BX/BM, pBzK, and sBzKs to our optical-NIR counterparts that have photometric data sets required for these color selections. We find that a majority of the optical-NIR counterparts meet color criteria of LBG, BX/BM, and sBzK, where the magnitudes are also comparable to these populations. Because we find that 59% of our faint ALMA sources are detectable in the optical-NIR band and classified as LBG, BX/BM, and sBZK populations, we conclude that about a half of faint ALMA sources are are high-z star-forming galaxies with the dust extinction and SFR much smaller than SMGs. Besides, the optical-NIR counterparts of the faint ALMA sources have the IRX-β relation clearly different from (U)LIRGs, indicating that the faint ALMA sources are not

8 Conclusion

149

explained by miniature (U)LIRGs that are simply scaled down with the infrared brightness. 28. We investigate our ALMA 1.2-mm number counts, comparing with the eleven 1.2-mm number counts estimated from the eleven LFs at 0 ≤ z ≤ 4. We find that our ALMA and the total of the eleven 1.2-mm number counts are consistent within the uncertainties at S1.2 mm ≥ 0.1 mJy, and that the estimates for our ALMA 1.2-mm number counts are reliable. From the contribution fractions of these eleven to our ALMA 1.2-mm number counts, we confirm that the top hat assumption at 1 ≤ z ≤ 4 agrees with our n(z) estimate for the faint ALMA sources. We also derive twelve 1.2-mm number counts from twelve SMFs at 0 ≤ z ≤ 8. The total of the twelve 1.2-mm number counts, on the other hand, shows a deficiency against our ALMA 1.2-mm number counts at S1.2 mm ≥ 0.4 mJy beyond the uncertainties. Although the SMF analysis with another Mstar − SFR relation partly fills the deficiency, the remaining deficiency implies the possibility that these are the dust obscured star-forming galaxies missing in the optical-NIR observations.

From Inter-stellar and Circum-Galactic to Cosmic Structure Scale From the ISM and CGM to the cosmic structure scales, our results suggest that the dust obscured star-formation is taking place in the ISM scale, instead of widely distributed outside of the galaxies in the CGM and/or IGM scales, where we succeed in identifying the entire population of the dust obscured star-formation contributing to the CIB with ALMA. 29. We estimate the SFRDtotal values at 1 ≤ z ≤ 4, combining the results from the previous SFRDUV studies and our 1.2-mm number counts, with the top hat assumption of n(z). We find that our SFRDtotal estimate shows the excess to the average of the previous best-estimate SFRDtotal value by 40+20 −40 %. Although this excess is only at 1σ significance level, this result supports the possibility that there exist the dust obscured star-forming galaxies missing in the optical-NIR observations.

Curriculum Vitae

Seiji Fujimoto Cosmic DAWN Center, Niels Bohr Institute, University of Copenhagen Jagtvej 128, DK-2200 Copenhagen N, Denmark E-mail: [email protected]

Research Interests – Evolution of massive galaxies – Structure formation in the early universe – Interplay among galaxy, circumgalactic and intergalactic media

Work Experience – – – – –

2021–present EU co-fund INTERACTIONS Fellow, Cosmic DAWN Center 2019–present DAWN Fellow, Cosmic DAWN Center 2019–2019 ALMA Project Researcher, NAOJ/University of Waseda 2019–2019 ICRR Project Researcher, University of Tokyo 2016–2019 JSPS Research Fellow (DC1), University of Tokyo

© Springer Nature Singapore Pte Ltd. 2021 S. Fujimoto, Demographics of the Cold Universe with ALMA, Springer Theses, https://doi.org/10.1007/978-981-16-4979-0

151

152

Curriculum Vitae

Education – 2016–2019 Ph.D Graduate school of Science, Department of Astronomy, University of Tokyo (Advisor: Prof. M. Ouchi) – 2014–2016 M.Sc. Graduate school of Science, Department of Astronomy, University of Tokyo (Advisor Prof. M. Ouchi) – 2010–2014 B.Sc. Department of Astronomy, University of Tokyo (Advisor: Prof. K. Kohno)

Awards/Prizes – – – – –

2021 Inoue Research Award for Young Scientists 2019 University of Tokyo School of Science Research Award for Doctoral Students 2016 University of Tokyo School of Science Research Award for Master Students 2016 Institute for Cosmic Ray Research President’s Award 2015 University of Tokyo President’s Award for Students

Awarded Telescope Times (S. Fujimoto as a PI) • JWST: Cycle 1 GO 1567 12h • VLT: 108.22MK 30h • Subaru: S16A0033N 15h, S20A0045N 10h S20B0150S 4h S21B0108N 18h, S22A0094N 27h • ALMA: 2017.1.00531.S 18h, 2019. 1.00672.S 12h, 2019.1.00236.S 10h, 2019.2.00050.S 42h, 2021. 1.00055.S 16h, 2021.1.00247.S, 19h • NOEMA: DDT E19AD 9h, E20EO 5 h, E20EN 2h, S21DM 34h, W21EH 26h, W21EF 2h • JVLA: DDT 20A-520 13h, 21A-145 22h, 21A-162 23h • JCMT: M17BP073 24h, M18AP001 30h • SMA: 2020B-S051 24h

Outreach – 2021 Press Release, “ALMA Discovers Rotating Infant Galaxy with Help of Natural Cosmic Telescope” – 2019 Press Release, “Carbon Cocoon Surrounded Growing Galaxies—ALMA Spots Earlies Environment Pollution in the Universe—”

Curriculum Vitae

153

– 2016 Press Release, “ALMA Resolves the Cosmic Infrared Background Light” – 2016–2018 Outreach public talk “Science Lab” in Hikawa High School, Japan – 2017 Web Article “Beyond Connecting Dots”, School of Science News in U.Tokyo

International Conferences in the Past 5 years Summary: Oral talks in peer-reviewed and invited conference (18), other oral talks (26) Highlights – – – – – – – – – – – – – –

11. 2022, Resolving Extragalactic Universe with ALMA, Tokyo, Japan (Invited) 03. 2022, The Growth of Galaxies in the Early Universe—VII, Sesto, Italy (Invited) 03. 2022, Linking Galaxy Physics from ISM to IGM Scales, Sesto, Italy (Invited) 06. 2021, Summer All Zoom Epoch of Reionization Astronomy Conference 2, online 02. 2021, CIDER—The Cold ISM During the Epoch of Reionization, online 10. 2020, The rise of metals and dust in the Universe over a wide range of cosmic times, online 07. 2020, Summer All Zoom Epoch of Reionization Astronomy Conference, online 11. 2019, Ringberg Workshop, Ringberg, Germany (Invited) 11. 2019, Revolutionary Spectroscopy of Today as Springboard to Webb, Leiden, Netherlands (Invited) 10. 2019, ALMA 2019: Science Results and Cross-Facility Synergies, Cagliari, Italy 09. 2019, Views on the ISM in galaxies in the ALMA era, Bologna, Italy 09. 2019, Extremely Big Eyes on the Early Universe, Roma, Italy 09. 2018, Chili-Japan Academic Forum, Nikko, Japan (Invited) 08. 2017, Twenty years of Submillimeter Galaxies, Durham, England