Impact of Feedback on the ISM of Extreme Starburst Galaxies

Home , I Zwicky 18

Veronica Menacho This thesis presents an analysis of the ionised gas properties and the impact of strong stellar feedback on the interstellar medium of the Impact of feedback on the ISM of starburst galaxy Haro 11. extreme starburst galaxies

The case of Haro 11

Veronica Menacho Menacho Impact of feedback on ISM of extreme starburst the galaxies

Veronica Menacho Menacho

ISBN 978-91-7911-216-5

Department of Astronomy

Doctoral Thesis in Astronomy at Stockholm University, Sweden 2020

Impact of feedback on the ISM of extreme starburst galaxies The case of Haro 11 Veronica Menacho Menacho Academic dissertation for the Degree of Doctor of Philosophy in Astronomy at Stockholm University to be publicly defended on Friday 11 September 2020 at 10.00 in sal FA32, AlbaNova universitetscentrum, Roslagstullsbacken 21.

Abstract Blue compact galaxies (BCGs) are compact, metal-poor, starbursting galaxies with characteristics similar to what is expected for the young high-redshifted galaxies. BCGs are among the most active in producing a large number of massive star clusters, each containing thousands of massive stars. During their short life, massive stars are continuously injecting energy, heat and momentum into the ISM via their intense radiation, stellar winds, and later on supernova explosions. These feedback mechanisms impact directly the star's surroundings, but when this feedback originates from a concentration of massive star clusters, it can strongly affect the condition of the gas of the entire galaxy. This thesis presents a detailed analysis of the ionized gas condition and the effect of strong feedback in Haro 11, an extreme starbursting BCG and the closest Lyman continuum (LyC) leaking galaxy. We exploit the spectro-photometric capabilities of the MUSE instrument, by slicing the galaxy spectra in a sequence of maps in velocity bins, in order to obtain a 3D information of the galaxy. Haro 11 has a rich population of massive and predominantly young star clusters, concentrated in three compact knots within its 4 x 4 kpc$^2$ centre. We find that the localised stellar feedback is strongly impacting the global kinematics and the condition of the gas up to further distances in the halo. Many kpc-scale structures such as filaments, shells and bubbles were traced in our data. Moreover, the strong feedback seems to have developed kpc- scale bubbles, outflows and galactic ionized cones with drastic consequences for the likely escape of Ly$\alpha$ and LyC photons, gas and metals out of the galaxy. The extended halo around Haro 11 is governed by photoionization processes and/ or shocks from recurrent supernovae originated in the central starburst region. Due to the galaxy's extreme ISM condition, commonly used emission lines diagnostics produce, in part, large discrepancies in the ionized gas properties. The results presented in this work highlight: a) the strong impact of stellar feedback affecting the ISM at all scales in starburst systems; b) the fact that traditional relations drawn up from averaged measurements of emission lines or from simplified models, fail in probing the condition of the gas in extreme environments. This is an appeal to revisit the standard relations by including more realistic models where several physical processes are simultaneously at work; c) the method applied here can be used to explore in detail the high sensitive, high spatial-resolution data from future facilities such as JWST/ELT.

Keywords: Galaxies, stellar feedback, ISM, Integral field spectroscopy.

Stockholm 2020 http://urn.kb.se/resolve?urn=urn:nbn:se:su:diva-182190

ISBN 978-91-7911-216-5 ISBN 978-91-7911-217-2

Department of Astronomy

Stockholm University, 106 91 Stockholm

IMPACT OF FEEDBACK ON THE ISM OF EXTREME STARBURST GALAXIES

Veronica Menacho Menacho

Impact of feedback on the ISM of extreme starburst galaxies

The case of Haro 11

Veronica Menacho Menacho ©Veronica Menacho Menacho, Stockholm University 2020

ISBN print 978-91-7911-216-5 ISBN PDF 978-91-7911-217-2

Printed in Sweden by Universitetsservice US-AB, Stockholm 2020 Cover image:

3D view of the ionised gas architecture of Haro 11. It was constructed by slicing up the Halpha spectrum of Haro 11 in bins of 50 km/s in a velocity range - 400 to 350 km/s.

1. Abstract

Blue compact galaxies (BCGs) are compact, metal-poor, starbursting galaxies with characteristics similar to what is expected for the young high-redshifted galaxies. BCGs are among the most active in producing a large amount of massive star clusters, each containing thousands of massive stars. During their short life, massive stars are continuously injecting energy, heat and momentum into the ISM via their intense radiation, stellar winds, and later on supernova explosions. These feedback mechanisms impact directly the star’s surroundings, but when this feedback originates from a concentration of massive star clusters, it can strongly affect the condition of the gas of the entire galaxy. This thesis presents a detailed analysis of the ionised gas condition and the effect of strong feedback in Haro 11, an extreme starbursting BCG and the closest Lyman continuum (LyC) leaking galaxy. We exploit the spectro- photometric capabilities of the MUSE instrument, by slicing the galaxy spectra in a sequence of maps in velocity bins, in order to obtain a 3D information of the galaxy. Haro 11 has a rich population of massive and predominantly young star clusters, concentrated in three compact knots within its 4 x 4 kpc2 centre. We find that the localised stellar feedback is strongly impacting the global kinematics and the condition of the gas up to further distances in the halo. Many kpc-scale structures such as filaments, shells and bubbles were traced in our data. Moreover, the strong feedback seems to have developed kpc-scale bubbles, outflows and galactic ionised cones with drastic consequences for the likely escape of Lyα and LyC photons, gas and metals out of the galaxy. The extended halo around Haro 11 is governed by photoionisation processes and/or shocks from recurrent supernovae originated in the central starburst region. Due to the galaxy’s extreme ISM condition, commonly used emission lines diagnostics produce, in part, large discrepancies in the ionised gas properties. The results presented in this work highlight: a) the strong impact of stellar feedback affecting the ISM at all scales in starburst systems; b) the fact that traditional relations drawn up from averaged measurements of emission lines or from simplified models, fail in probing the condition of the gas in extreme environments. This is an appeal to revisit the standard relations by including more realistic models where several physical processes are simultaneously at work; c) the method applied here can be used to explore in detail the high sensitive, high spatial-resolution data from future facilities such as JWST/ELT.

1 2 CHAPTER 1. ABSTRACT 3

A papa, que construyo mis alas, A mama que me enseno a volar 4 CHAPTER 1. ABSTRACT 2. List of Papers

The following papers, referred to in the text by their Roman numerals, are included in this thesis.

PAPER I The impact of stellar feedback from velocity-dependent ionized gas maps - a MUSE view of Haro 11 V. Menacho, G. Östlin, A. Bik, L. della Bruna, J. Melinder, A. Adamo, M. Hayes, C. E. Herenz and N. Bergvall, MNRAS, 487, 3183-3198 (2019). DOI:10.1093/mnras/stz1414

PAPER II Ionised gas properties of the extreme starburst galaxy Haro 11. Tem- perature and metal abundance discrepancies V. Menacho, G. Östlin, A. Bik, To be submitted to MNRAS

PAPER III Mapping the dominant excitation mechanisms in the extreme starburst galaxy Haro 11 with MUSE Veronica Menacho, Arjan Bik and Göran Östlin MNRAS, draft version

PAPER IV The ionised halo of Haro 11 – a density bounded nebula G. Östlin, V. Menacho, A. Bik, A. Adamo, L. Della Bruna, M. Hayes, J. Melinder and N. Bergvall. To be submitted to MNRAS

PAPER V The source of leaking ionising photons from Haro 11 – Clues from HST/COS spectroscopy of knots A, B and C G. Östlin, E. T. Rivera-Thorsen, V. Menacho, M. Hayes, A. Runnholm, S. Oey, M. Mass-Hesse, G. Micheva, A. Adamo, A. Bik, J. Cannon, M. Gronke, D. Kunth, P. Laursen, J. Melinder, M. Messa and L. Smith. To be submitted to ApJ

Reprints for papers were made with permission from the publishers.

5 6 CHAPTER 2. LIST OF PAPERS 3. Author’s contribution

PAPER I - V I developed a graphical user interface (GUI) in phython which aims to guide a user towards the MUSE data reduction processes. This software has two main functionalities: First, it organises the large amount of raw files and the files generated from the reduction processes and second, it allows for each data reduction step, to configure the parameters/flags and execute the MUSE pipeline. I reduced the MUSE data of Haro 11 (P.I. G. Östlin) following the standard reduction procedures, except for the sky images that was created manually to avoid oversubtraction of the brightest lines in the halo. Papers 1–5 are based on this reduced data cube. PAPER I In this paper I led the entire data analysis and interpretation of the results under guidance of my supervisors. This work is based on spatially and spec- trally resampled emission lines that thereafter were split in 50 kms−1 bins. I developed the software needed to produce the results. I perform all steps to create the final velocity bins maps, except for the stellar absorption correction that was performed by a collaborator. I wrote the entire article and produced almost all figures and all tables. The HST image was provided by a collaborator. PAPER II In this paper, I led the entire work under guidance of my supervisors. This work presents integrated line maps and maps in three velocity ranges tracing the gas at blueshifted, central and redshifted velocities. In this work we study the physical conditions of the gas and the kinematics of the galaxy. I developed all the scripts and performed all the steps needed to create the final results. I wrote the entire article and produced all figures and tables. PAPER III In this paper, I led the entire work under guidance of my supervisors. This work also presents integrated line maps and velocity-dependent maps for the gas at blueshifted, central and redshifted velocities, however it focus on the excitation mechanisms of the gas: photoionisation or shocks. I developed the scripts to create the final results. The final maps were partially created with the help of GLUE, a python package developed to visualise iterative data. I wrote

7 8 CHAPTER 3. AUTHOR’S CONTRIBUTION the entire article and produced all figures and tables. PAPER IV This paper is based on my work on radial profiles of the physical properties of the gas in Haro 11. I created radial profiles of several emission lines and derived all physical properties of the gas. I developed the scripts to extract radial profiles of several integrated lines maps and I created most of the radial profile figures and the two maps presented in this paper. PAPER V This paper is based on HST/COS observations and MUSE data of the star forming knots A, B and C. I provided all information from the MUSE spectra. My work consisted in extracting integrated line flux in an aperture of 1 arcsec to thereafter apply the COS vigneting function. Finally I derived the kinematics and physical condition of the ionised gas of all three knots A, B and C.

Re-use of material from the licentiate thesis of the author: Part of the written material in this thesis, including paper I was included in my Licentiate thesis (2019). Chapter 1 was partially re-used and improved in chapter 1 and 2 in this thesis. The subsections 2.3.5 - 2.3.7 are new. Chapter 2 and 3 were partially re-used in chapters 3 and 4. The subsections 4.1.1, 4.1.2, 4.3.3, 4.3 and 4.4 are new Figures The source, and copyright if apply, of each ﬁgure is speciﬁed the caption. Contents

1 Abstract 1

2 List of Papers 5

3 Author’s contribution 7

Abbreviations 11

List of Figures 13

4 Introduction 15 4.1 Preamble ...... 15 4.2 Lyman continuum leaking galaxies as important sources for the reionisation ...... 16 4.3 Blue compact galaxies ...... 18 4.4 About the production of massive stars and massive star clusters ...... 21

5 The Interstellar Medium 25 5.1 Phases of the ISM and some basic mechanisms at work in the ISM ...... 25 5.1.1 Heating and cooling ...... 27 5.1.2 Equilibrium conﬁgurations in the ISM ...... 27 5.2 Basic processes at work in the warm ionised medium ... 28 5.2.1 HII regions ...... 29 5.3 Emission line diagnostics ...... 30 5.3.1 Ionised gas kinematics ...... 31 5.3.2 Ionisation parameter ...... 34 5.3.3 Shock tracer ...... 36 5.3.4 Dust attenuation ...... 36 5.3.5 Density and temperature ...... 38 5.3.6 Metallicities ...... 40

9 10 CONTENTS

5.3.7 Temperature and metallicity variations ...... 47

6 The impact of stellar feedback 51 6.1 Energy released by massive stars and star clusters ..... 51 6.1.1 Energy output of ionising radiation, stellar winds and supernova ...... 53 6.2 Structures of stellar feedback origin ...... 59 6.2.1 Observations of large scale structures ...... 60 6.3 Bubbles and superbubbles ...... 63 6.3.1 Physics of bubbles - superbubbles ...... 63 6.3.2 Bubbles in simulations ...... 67 6.4 Galactic winds, ﬁlaments and galactic fountain ...... 69

7 The violent nature of Haro 11 73 7.1 Haro 11 in the literature ...... 73 7.1.1 The cluster population in Haro 11 ...... 74 7.1.2 Complex kinematics ...... 78 7.2 MUSE data, observations, data reduction and methods .. 81 7.3 The warm ionised ISM of Haro 11 - Results and summary of papers I-V ...... 83 7.3.1 Paper I. The impact of stellar feedback from velocity- dependent ionised gas maps – a MUSE view of Haro 11...... 83 7.3.2 Paper II. Ionised gas properties of the extreme starburst galaxy Haro 11. Temperature and metal abundance discrepancies...... 88 7.3.3 Paper III. Deciphering the excitation mechanisms in Haro 11. A new perspective from the MUSE data. 89 7.3.4 Paper IV.The large ionised halo of Haro 11 – a density bound nebula ...... 90 7.3.5 Paper V.The source of leaking ionising photons from Haro 11 – Clues from HST/COS spectroscopy of knots A, B and C ...... 90

8 Summary and outlook 93 8.1 Future Work ...... 96

Sammanfattning xcvii

Acknowledgements xcix

References ci Abbreviations

AGN Active Galactic Nuclei BCG Blue Compact Galaxy CNM Cold Neutral Medium GMC Giant Molecular Cloud HIM Highly Ionized Medium HST Hubble Space Telescope ISM InterStellar Medium LaTeX Lamport TeX LMC Large Magellanic Cloud LyC LYman Continuum MUSE Multi Unit Spectroscopic Explorer NFM Narrow Field Modus RT Rayleight-Taylor (instability) SFR Star Formation Rate SMC Small Magellanic Cloud SN SuperNova SNR SuperNova Remnant SSC Super Star Cluster WFM Wide Field Modus WIM Warm Ionized Medium WNM Warm Neutral Medium WR Wolf-Rayet star

11 12 ABBREVIATIONS List of Figures

4.1 HST/ACS image of I Zwicky 18 ...... 20 4.2 The cyclical process of formation, evolution and death of stars. 22 4.3 The central region of 30 Doradus...... 23

5.1 Kinematics of Eso 338 ...... 32 5.2 OI-BPT diagram of Eso 338 tracing shocks ...... 35 5.3 Timescales in the production of elements ...... 43

6.1 Mechanical and ionising energy output from winds and supernovae ...... 56 6.2 Energy output from winds and supernovae ...... 57 6.3 Momentum injected by diﬀerent components ...... 58 6.4 Projection of a shell ...... 60 6.5 Bubbles and ﬁlaments in the Milky Way ...... 61 6.6 Structure of a bubble ...... 65 6.7 Evolution of a bubble in simulations ...... 69 6.8 Galactic wind in M82 ...... 70

7.1 Haro 11 seen with the MUSE instrument and HST telescope . 74 7.2 Cluster population of Haro 11 ...... 75 7.3 Massive cluster populations in the knots ...... 76 7.4 Supernova rates in the knots ...... 79 7.5 MUSE instrument ...... 81 7.6 WFM and NFM of MUSE ...... 82 7.7 3D view of the Hα emission ...... 84 7.8 3D view of the ionised gas structure of Haro 11 ...... 85 7.9 Sketch of the ionised gas components of Haro 11 ...... 87

13 14 LIST OF FIGURES 4. Introduction

4.1 Preamble

Galaxies are the basic building blocks of the universe at large scales. In terms of baryonic matter, they are composed of stars, gas and dust. The fraction of these components can differ considerably between galaxies, and gives rise to a large diversity in the galaxy population. Young galaxies are in general gas rich, metal poor and have younger stellar populations, while the most evolved systems have a predominantly older stellar population, and higher metallicities. However, galaxies are even more diverse if we also take into account their appearance and ongoing evolutionary processes. These properties, obtained from observations across a wide spectral range – from gamma to radio wavelengths – include: sizes and morphological appearance, intrinsic velocities, star formation rate and star formation history, spectral line properties and spectral energy distributions (SED), etc. Each of these properties reveals potentially important clues about the evolutionary history, and possible evolutionary path of a galaxy. Galaxies evolve mainly by converting their cold-gas content into stars. From young ages, galaxies grow either by accreting large amounts of cold gas from the cosmic web, or by merging with other galaxies. Hydrodynamical simulations have shown that gas accretion and merger events happen several times along the evolution of galaxies (Schaye et al. 2015; Springel et al. 2005). In each event, large amounts of stars are produced in a short time, which is called a starburst episode. At the same time, the strong stellar feedback produced by the most massive stars can potentially regulate the precipitated formation of new stars. tar formation will continue as long as molecular gas remains available.Galaxies that are forming stars are called active galaxies. However, the cold gas supplied cannot last indefinitely. At low redshifts, there exists a large population of passive galaxies which are not producing stars, or that have very low star formation rates (SFR). Although the majority of these systems 12 are massive galaxies (M>10 M ), there also exists a small fraction of less- massive (younger) systems. While the effects of gas consumption in merger events dominate the quenching of star formation at earlier times, the cessa- tion of star formation at later epochs is a direct consequence of the galaxy’s

15 10.6 excessive mass growth (M>10 M ) (Peng et al. 2010). In this thesis, I focus on Haro 11, a particular starburst galaxy with Lyα emission and in which Lyman continuum (LyC) emission has been detected. Haro 11 is a blue compact galaxy (BCG) that is undergoing an extraordinary episode of star formation. Its extreme condition is triggered by the ongoing merger event of two gas rich progenitors. Haro 11, as well as other BCGs, has similar properties to high-z galaxies; however being the closest detected LyC emitter, Haro 11 offers a unique opportunity to study in great detail not only the processes that drive galaxy evolution, but also the mechanisms that facilitate the escape of LyC radiation. This thesis presents the work of five papers that focus on: a) the impact that strong stellar feedback has in creating large scale structures and on the ionisation structure of the galaxy; b) the properties of the ionised gas under extreme conditions; c) the governing excitation mechanisms: photoionisation or shocks, at sub-kpc scales d) the ionisation budget of the galaxy; e) the analysis of LyC sources in Haro 11. In this literature review, I will first present in chapter one the importance of Lyman continuum leakage in galaxies in a cosmological context, and the properties of Blue Compact Galaxies. Then in the second chapter, I will describe the physical processes at work in the ISM and the emission line diagnostics used to examine the conditions of the ionised gas. In the third chapter, I will show the origin and effect of strong stellar feedback on the ISM of (relatively) low-mass galaxies. Finally, in chapter four, I will review the literature on Haro 11 and the main results of my work.

4.2 Lyman continuum leaking galaxies as important sources for the reionisation

In the evolution of the universe described by the well established ΛCDM model, there is a transition phase from the ‘dark ages’, when matter in the universe was neutral and opaque to radiation, to the ‘ages’ after ‘reionisation’, when the intergalactic medium (IGM) became ionised and transparent. Prior to this transition phase, dubbed as the epoch of reionisation, overdense regions collapsed into cold dark matter halos, initially forming the ‘first’ stars and shortly after, the first galaxies. Great amounts of ionising photons may have leaked out of these young galaxies, starting in this way the epoch of reionisation of the intergalactic medium. Although it is not clear when this period began (around z∼12), it is thought to have finished at redshift ∼6 (Zaroubi 2013), as evidenced in the spectra of quasars. In this transition phase, the spectra blue-wards from the Lyα line is characterised by multiple absorption lines (dubbed as the ‘Lyα forest’) caused by neutral hydrogen gas along the line of

16 sight (Fan et al. 2006). These absorption segments increase with redshift until they become completely absorbed, and mark the point where the intergalactic medium is dominated by neutral gas. Two main sources may have contributed to the reionisation of the universe: galaxies and active galactic nuclei (AGN). AGN are capable of efficiently producing immense amounts of energetic photons (Madau et al. 2004), many of which will escape to the IGM (Telfer et al. 2002). Star forming galaxies however, might be favoured due to their larger population at that epoch in comparison to AGN (Atek et al. 2015a,b; Haardt and Madau 2012). A fundamental goal in astrophysics is to understand how the first galaxies were formed, and to what extent they contributed to the reionisation of the universe. One of the main barriers that has yet to be overcome in reaching this goal is the lack of clear clues about how the first stars were formed in terms of mass and clustering. The physics of star formation predicts the formation of massive stars up to 1000 M at extremely low metallicities, called Pop III stars (Omukai 2008). Individual stars of such masses would surely have had a great impact on their surrounding gas. However, only a clustering of such stars would allow them to impact far enough out to reach the intergalactic medium. It is therefore widely discussed whether or not radiation from the first galaxies contributed substantially to the reionisation of the universe (Bouwens et al. 2015; Robertson et al. 2015). Galaxies with primordial gas content (i.e. primordial metallicities) have yet to be found, but a new generation of telescopes with highly sensitive instruments, such as JWST, GMT, TMT, ELT, and even SKA facilities will provide insight into the young universe and its young galaxy population. To date, only a few galaxies with redshifts greater than 6 have been observed. The farthest spectroscopically confirmed galaxy has a redshift of 11.1, which signifies that the observed light was emitted when the universe was about 400 million years old (Oesch et al. 2016). Even if, at such a high redshift, young galaxies in formation may be more common than at present epochs, they are rarely detected due to their low surface brightness. Moreover, they are too distant to be well spatially resolved, and thus the study of these systems in detail remains difficult (Bouwens et al. 2010a,b). Young galaxies are low-mass, gas-rich systems with low metallicities and a clumpy morphology (Elmegreen 2003; Genzel et al. 2017; Tacconi et al. 2013). These galaxies are undergoing an intense episode of star formation triggered by gas infall and/or merger events with other primordial systems. Their extreme condition favours the formation of large amounts of massive stars, whose intense ionisation field and mechanical/radiative feedback is be- lieved to facilitate the escape of ionising photons to the intergalactic medium (IGM). Cosmological studies have shown that in order to reionise the universe,

17 about 20% of the produced ionising radiation has to escape the galaxies (this quantity is called the escape fraction) (Atek et al. 2015a,b; Haardt and Madau 2012). Given the challenges to observe such galaxies, astronomers have instead targeted low redshift systems in order to measure the escape fraction of ionising radiation. Surprisingly, only a few galaxies in the redshift span from 0 to 4, were detected to leak LyC photons (Bergvall et al. 2006; Borthakur et al. 2014; de Barros et al. 2016; Izotov et al. 2016a,b, 2018; Leitherer et al. 2016; Puschnig et al. 2017; Vanzella et al. 2016, 2018, Izotov et al. 2018a). Nevertheless, this number is currently rising due to observational campaigns targeting galaxies with particular characteristics that seem to be typical of LyC emitters (Izotov et al. 2016b, 2018, and others.). These LyC leakers appear to be very compact in size and have extremely high SFR per area (Jaskot and Oey 2013; Marchi et al. 2018). Additionally, they have high ionisation parameter ratios (Jaskot et al. 2019; Nakajima and Ouchi 2014), weak UV ionisation absorption lines (Heckman et al. 2011), and strong mechanical feedback (Heck- man et al. 2011). Such galaxies in the local universe are known as ‘green pea’ galaxies, while others are blue compact galaxies (BCG) (Amorín et al. 2017, 2012; Cardamone et al. 2009; Izotov et al. 2011; Jaskot and Oey 2013). There are two mechanisms that favour the escape of LyC: The ﬁrst is the formation of a density bound nebula, a condition that is fulﬁlled whenever the number of ionising photons is larger than the number of recombinations. In this scenario, the excess of ionising photons will escape in any direction to the IGM. The second mechanism requires the presence of galactic holes in a medium that is not transparent to LyC photons. These holes can directly connect the areas producing ionising radiation (hot stars, young massive star clusters) to the IGM. The latter may be the dominant mechanism among the observed LyC leakers (Chisholm et al. 2015).

4.3 Blue compact galaxies

Blue compact galaxies are low redshift galaxies, compact in appearance and mostly metal poor. They are undergoing an intense episode of star formation, which is mainly triggered by an ongoing merger event, gas infall or tidal disruption from a close by passage. These characteristics are similar to the expected properties of young galaxies, and therefore BCGs are usually referred to as ’local analogues’. The great advantage of these ’local laboratories’ is pre- cisely their close distance, making them ideal targets to study at high spatial and spectral resolution the properties of young galaxies, and the physical processes that drive galaxy evolution (Kunth and Östlin 2000; Östlin et al. 2001; Papaderos et al. 1996; Thuan 2008). BCGs have oxygen abundances from 3 to 50 times lower (12+log(O/H)=7.1

18 – ∼8.3) than the metallicity of the Sun. Detailed analysis of their stellar populations indicate that these are more evolved systems with an active starburst episode. The age distribution of their stellar population implies that they might have had bursts of star formation episodes in the past, followed by longer qui- escent periods (Fanelli et al. 1988; Searle and Sargent 1972; Thuan 2008). Thus, past star formation has to some degree metal-enriched their ISM (Kong et al. 2003; Schulte-Ladbeck et al. 1999). Fig. 4.1 shows I Zwicky 18, one of the most metal poor galaxies ever observed. Although no galaxy with an extremely low metal content has yet been observed, there is indirect evidence of high redshift galaxies with extremely low metallicities. Bouwens et al. (2010a) presented a sample of z∼7 galaxies that show extremely blue UV-continua. The authors suggest that the observed UV slope (β-Slope) cor- −5 responds to galaxies with metallicities up to 0.5 ×10 Z , that are probably still populated by the first generation of stars (Pop III stars); however no direct evidence was found. It is interesting to note that up to now, there is still a lack of direct evidence for galaxies at any redshift or HII regions with metallicities lower than 12+log(O/H)∼7.0 (Hirschauer et al. 2016; Izotov et al. 2018, 2019). Thus, it seems that even low-metallicity galaxies were at least metal-enriched by the Pop III generation of stars (Kunth and Sargent 1986; Thuan et al. 2005). In BCGs, intense star formation is often localised in the central region, either as a unique compact region, or as a few sub-regions forming a complex structure. In this starburst environment, the most massive stars and star clusters are formed. The spectra of the emitting gas in these regions are characterised by strong emission lines and relatively high blue continua, which is an indi- cation of their young and massive stellar population (Kunth and Östlin 2000). Most of these galaxies are rich in neutral gas, mainly concentrated in dense clumps in the starburst region (Thuan et al. 2016). Surprisingly, some of the most metal poor systems show dust emission (Cormier et al. 2012; Hirashita and Hunt 2004) and a very weak H2 emission, but no signatures of polycyclic aromatic hydrocarbons (PAH) have been found (Cormier et al. 2012; Wu et al. 2006). Their low metallicities and extreme UV radiation fields are likely responsible for destroying the PAH component (Wu et al. 2007). BCGs appear to have two modes of star formation: an intense or active mode, and a less-intense mode. In the active mode, the star formation rate (SFR) is elevated, a condition that favours the formation of large amounts of 5 super star clusters (SSC, M≥ 10 M ) in compact regions (Thuan et al. 1997). These galaxies show neutral gas emission and a hot dust content. In the less- intense mode, the star formation happens at low rates in extended areas. In this mode, SSCs are almost absent in the starburst region and the dust is cooler due to a less intense ionisation field (Hirashita and Hunt 2004). Nevertheless, the ionising radiation is harder (more energetic) in low metallicity galaxies (Gu-

19 Figure 4.1: HST/ACS image of I Zwicky 18, the second most metal poor galaxy ever discovered. I Zwicky 18 has an oxygen abundance 40 times lower than that of the sun. Two separate star forming regions constitute this irregular galaxy. In them, thousands of stars are formed in star clusters that are easily recognised as bright spot in the image. (Credit NASA, ESA, Y. Izotov and T. Thuan).

seva et al. 2000) and the cooling is less eﬃcient. In these systems, emission lines requiring hard ionising radiation, such as He II (54.4 eV), [O IV]λ4227 (∼54.9 eV), [Fe V]λ4227 (∼54 eV), [Ne V]λ3346,3426 (∼96 eV) are very fre- quently identiﬁed (Thuan and Izotov 2005). Although the origin of this hard radiation is still under debate, the energetic photons are likely produced by hot Wolf-Rayet stars (WR) and/or luminous high-mass X-ray binaries. Alter- −1 natively, fast shocks (vshock > 400 km s ) propagating in a dense ISM could also provide the mechanism for collisionally excited lines (Thuan and Izotov 2005).

20 4.4 About the production of massive stars and massive star clusters

Star formation is an important driver of galaxy evolution. Independent of the mechanisms that trigger it, star formation implies the consumption of cold gas, the metal-enrichment of the gas and the release of radiation and mechanical energy into the ISM. The formation, evolution and death of stars happen in a self-sustained cycle of connected events that starts and ﬁnishes in a cloud of gas (see Fig. 4.2). In a simpliﬁed scenario, star formation begins when dense molecular clouds start to collapse due to their self gravity. While this happens, the temperature and density of the central core rises considerably until hydrogen begins to fuse into helium. Radiative pressure holds further gravitational collapse and the star remains in hydrodynamical equilibrium. The process of burning hydrogen continues for the longest part of the star’s life. In massive stars, when the hydrogen is exhausted in the nucleus, the core slightly contracts, increasing its temperature considerably. Heavier elements, such as carbon (C), oxygen (O) and nitrogen (N) are produced. When the resources for these nuclear reac- tions are exhausted, the exterior envelope expands outwards. The subsequent fate of the star depends on its mass. In low mass stars, the outer envelope is ejected forming a ’planetary nebula’, while the already degenerate inner core collapses, forming a white dwarf. Massive stars undergo a supernova explosion. This is a violent explosion that propels the star’s outer layers away, leaving a neutron star or a black hole in the case of a massive core (M≥ 20 M ). The expelled metal-enriched material will then be the fuel for future star formation. Star forming galaxies normally have extended regions of star formation. Inside them, there exists complex arrangements of molecular clouds at all scales. At large scales, giant molecular clouds (GMC) can have masses up 9 to 10 M (Wilson et al. 2003), but only a small fraction of their mass is converted into stars (Lada et al. 2010). One giant molecular cloud can form up to thousands of stars. Massive clouds start to fragment systematically due to turbulence in the gas, forming smaller dense clumps where stars form. The hierarchy of these structures is also maintained in the stellar outcome. If the GMC belongs to an even larger structure, they will be part of an extended clustering system (Bastian et al. 2009). Therefore, GMCs form associations and cluster of stars. In a giant molecular cloud that is in process of forming a star cluster, the most massive stars radiate hard UV photons and begin to rapidly ionise the surrounding medium. At the same time, shocks driven by ionisation fronts and radiation pressure compress the intercluster medium and can trigger more

21 Figure 4.2: Cyclical process of formation, evolution and death of stars. The starting point is the central ’star-forming nebula’ where stars are born. The left cycle shows the evolutionary path of low-mass stars, while massive stars follow the right cycle. The size of the cycle represents the life time duration. Low mass stars live longer. Finally, both type of stars expel their outer envelope; this is the material that becomes the fuel for future star formation (Credit: NASA and the Night Sky Network).

22 Figure 4.3: The central region of 30 Doradus. At the centre sits the super star cluster R136, which is surrounded by a cavity of rarefied and hot gas. This region also contains other less massive clusters and OB-associations. The image shows several cavities, some of which are open. These cavities were most likely created by stellar winds and supernova explosions of the most massive stars within the massive star clusters. (Credit: Image: ESO/IDA/Danish 1.5 m/R. Gendler, C. C. Thöne, C. Féron, and J.-E. Ovaldsen). star formation. In general however, more matter is ionised than converted into stars. This mechanism regulates and stops the formation of stars in a cluster. Additionally, it facilitates the formation of a porous medium where part of the ionised gas will stream away, diminishing the molecular cloud mass and inducing turbulence both in the molecular cloud and in the ISM (Krumholz and McKee 2005). Star clusters are commonly defined as overdensities of stars that are gravi- tationally bound. Star clusters constitute a few percent of the GMC mass. They 7 have sizes ranging from 1 to 10 pc and masses up to ∼ 10 M . Star clusters 5 with masses M≥ 10 M are dubbed ’super star clusters’ (SSC) (Adamo et al. 2010, 2011; Messa et al. 2018). In massive systems, all cluster members are formed in a short time. In the first couple of Myr, new born stars are still embedded in the parental gas, but starting from 3 Myr onwards, the gas is mostly cleared by the effect of feedback from the most massive stars (Hollyhead et al. 2015). Star clusters in general, and specifically the most massive ones, are preferentially born in starburst systems. Star forming regions with a clustering system of stars and star clusters are

23 clearly seen in images of nearby galaxies (Bastian et al. 2009; Elmegreen et al. 2005; Messa et al. 2018). Usually, there are several bright spots of star formation not distributed uniformly, but that appear to be interconnected and form a larger network within a star forming region. In spiral galaxies, star formation happens mostly along the extended spiral arms. In small compact galaxies however, it is mostly concentrated in the central region and rarely concentrated at one side of the galaxy as is the case in I Zwicky 18 (Fig. 4.1) and 30 Do- radus in the Large Magellanic Cloud (LMC). In Fig. 4.3, we show a close view of 30 Doradus, the most active starburst region in the local group. The HST image shows several star clusters, association of stars and even massive single stars. The most massive star found in this region has a mass of about 315 M (Crowther et al. 2010), considerably more massive than the cutoﬀ limit of all initial mass function (IMF) distributions which are used to sample the stellar population of a star cluster (Doran et al. 2013). Among the star clusters at the centre of 30 Doradus, the super star cluster R136 has a stellar mass of about 5 M∼4.5×10 M and is the most massive in the local group. It has nearly one hundred massive stars (mostly OB stars, but also some Wolf-Rayet stars) that are releasing large amounts of ionising photons into its surrounding gas. It releases about 50% of the total ionising photons produced in 30 Doradus (Do- ran et al. 2013). The intensive photoionisation and radiative feedback from the most massive stars has shaped bubbles, ﬁlaments and cavities in the surrounding gas and dust, which are clearly visible in the image. The radiative and mechanical energy released by massive stars can drastically (alter) a star formation region and even an entire galaxy. This mechanism is described in the third chapter.

24 5. The Interstellar Medium

5.1 Phases of the ISM and some basic mechanisms at work in the ISM

The information presented here is mostly taken from Draine (2011); Oster- brock (2006) and lecture notes from Pogge (2011). The space between stars, e.g. the interstellar medium (ISM), is filled with gas and dust. The gas is mainly composed of hydrogen, followed by helium and elements heavier than helium which are labelled as ’metals’ in astrophysics. The most abundant metals are oxygen (O), nitrogen (N), carbon (C) and iron (Fe). The condition of the interstellar gas is linked to its physical properties: temperature, density, chemical composition and the degree of ionisation of the gas, all of whose values can vary significantly. For certain configurations of parameters, the gas is in thermal and energy equilibrium, which is established by the continuous balance between heating and cooling. The configurations that predominate are related to specific characteristic states or phases of the ISM. Draine (2011) defines seven phases of the ISM:

1. Coronal gas or Hot ionised Medium (HIM). This is a collisionally, highly ionised gas that has been heated by supernovae (SNe) to temperatures of T≥30000 K. This hot gas has low density and fills most of a galaxy’s halo. Moreover, it also fills the interior of bubbles and can be found in some outflows. In this gas, high ionisation emission lines such as [OIV] are produced. The coronal gas cools by soft X-ray emission and radio synchrotron emission. 2. HII gas or Warm ionised Medium (WIM). This is the gas surrounding hot massive stars (OB-type, WR, massive binaries), that has been photoionised by their UV photons. If this gas is expanding and therefore not in pressure equilibrium, it is mostly diffuse, like the gas in the galactic halo and the gas filling the space between clouds. Otherwise, in denser environments this gas is in pressure equilibrium, such as for ’HII regions’, where the photoionised gas is confined within a more neutral and colder gas (parental clouds). The WIM or HII-gas has a temperature of about 10000 K, densities ranging from 101 to 104 cm−3, and cools

25 mostly by optical line emission (recombination or collisional excitation lines) and free-free (continuum) emission. 3. Warm HI or Warm Neutral Medium (WNM). This gas is predominantly neutral, and its density and temperature are about 0.6 cm−3 and 5000 K respectively. The heating mechanisms acting in this gas are mostly cosmic ray, photoionisation and photoelectric heating in the dust. This gas is in pressure equilibrium and cools by line emission in the optical range, as well as ﬁne structure line emission in the infrared (IR). It is also observed at 21cm in radio. 4. Cool HI or Cold Neutral Medium (CNM). This is the neutral gas at temperatures of about 100 K, and densities of about 30 cm−3. The heating mechanisms are identical to those acting in the WIM: photoelectric heating in the dust, cosmic rays and photoionisation. It cools by ﬁne structure emission lines, but is additionally observed by absorption lines in the optical and UV, as well as by 21cm emission and absorption.

5. Diﬀuse H2 molecular gas. This is predominantly molecular gas at temperatures of about 50 K and densities of about 100 cm−3. The heating mechanisms here are similar to the mechanisms in the WIM. It also cools by ﬁne structure emission lines and is mostly observed in various CO- line emissions.

6. Dense H2 molecular gas. This is the gas and dust that builds a molecular cloud. It has high extinction, low temperatures (T≤ 50 K) and high densities (from 103 to 106 cm−3). The heating mechanisms are the same as in the WIM. It cools mainly by CO emission. 7. Cool stellar outflows. This is the gas expelled at low velocities (v≤30 km s−1) by evolved stars. The outflow densities vary from 1 to 106 cm−3 and the temperatures are in general lower than T≤ 1000 K. These outflows are traced by absorption in the optical and UV range, and by emission in the IR. The cold gas is additionally traced by CO lines and HI radio emission1.

The ISM is a dynamic system with diﬀerent sources of energy and pressure such as thermal, hydrodynamic, magnetic, cosmic rays, starlight and cosmic background radiation; all with approximately the same energy density. More- over, there is a constant interchange of energy and momentum in the ISM,

1It is important to note that outﬂows can have a multiphase nature, i.e. they can have a hot, warm and cold component. Even more so, most galactic outﬂows consist mainly of hot wind material, or warm clouds embedded within a hot wind (Heckman et al. 2015; Vijayan et al. 2019) without a cold component.

26 so that transitions between gas phases can occur in the regions where either heating or cooling dominates.

5.1.1 Heating and cooling Heating happens when the interstellar gas absorbs energy. It is associated with processes that release free electrons to the ISM or increase the kinetic and thermal energy of these electrons. Photoionisation heating is the principal heating mechanism in the warm ionised medium. The sources of this heating mechanism are mostly OB stars, but also WR stars, which release large amounts of UV photons into the ISM. The injected heating is then given by the energy difference between the photon energy and the ionisation potential of the electron, which is of a few eV. Another source of heating is ionisation from cosmic rays, which consist mainly of protons or atomic nuclei with energies of about 1-10 MeV. In the case of cosmic rays, after the ﬁrst ionisation, the newly released electrons have energies of ∼35 eV, high enough to produce at least one more ionisation through collision. In a cold neutral medium, photoelectric heating is the dominant mechanism. This mechanism consists of photons heating the surface of dust grains, which as a consequence release free electrons into the ISM. On the other side, cooling is achieved when energy is carried away from the various (ISM) regions or galaxy. It is linked to the recapture of free electrons at a high energy level, and its associated cascade of radiative transitions to the ground state. Thus, cooling is deﬁned by all processes that produce discrete emission (emission lines), and continuum radiation from particles in the gas characterising the emerging spectrum of the nebula. The most important cooling mechanisms in the warm ionised medium are the recombination of ions with free electrons (mostly ionised hydrogen), and the radiative de-excitation of collisionally excited metals. Among the metals, the most important coolants are oxygen, carbon, nitrogen, iron and silicon in the optical range. In the hot coronal gas, thermal Bremsstrahlung is the most important cooling mechanism, while in the cold neutral gas, thermal and non-thermal emissions due to collisions of dust grains with atoms, molecules or electrons dominate. The strength of the cooling lines which involve collisions depends not only on the metal abundance, but also on the density and temperature of the gas.

5.1.2 Equilibrium conﬁgurations in the ISM The equilibrium states that deﬁne a gas phase are the following:

• Kinetic equilibrium. In the ISM, the majority of collisions are elastic. The timescale for an electron-electron collision is shorter than for an

27 electron-hydrogen collision. If the former happens, the velocities of both electrons become thermalized, i.e. in kinetic equilibrium. In this case, their velocities follow a Maxwellian distribution that is characterized by a kinetic temperature. This electron temperature characterizes the temperature of the gas in a region.

• Ionisation equilibrium. This is the balance between ionisation and recombination, where the ionisation rate is set by the radiation ﬁeld, and the recombination rate by the density and electron velocities. The recombination rate increases with higher electron density and lower electron velocity.

• Pressure equilibrium. This equilibrium is set by the pressure balance in a system (HII region, molecular clouds or even a whole galaxy). There are several sources of pressure, such as thermal, magnetic, cosmic rays, hydrodynamic (from moving gas, e.g. stellar winds, outﬂows, SN blastwaves, etc.) and radiative pressure. The strength of these sources can be established by their energy densities, which is the energy per unit volume (eV cm−3) available to be released into the ISM .

5.2 Basic processes at work in the warm ionised medium

This thesis focuses on the condition of the warm ionised gas that emits mostly in the optical wavelength range. Therefore, I address here the physical processes that take place only in this particular gas phase. For most galaxies, more than 50% of the warm ionised gas is diffuse and resides outside classical star forming regions. In the Antennae galaxy for example, the diffuse gas mass is about 60% of the total ionised gas mass (Weil- bacher et al. 2018). The remaining dense gas is found in HII regions or in a wide number of structures such as planetary nebulae, novae, supernova remnants, filaments, shells from bubbles, or even close to the galactic nucleus. HII regions are dense regions that have been photoionised by photons (≥13.6 eV) of OB stars embedded in the nebula. They emit strong emission lines and a weak continuum. Planetary nebulae are expelled stellar envelopes of AGB progenitors. The resulting hot white dwarf remnant photoionises the expanding shell, which is characterised by strong emission lines and an overabun- dance of highly ionised elements. Supernova Remnants (SNRs) are rapidly expanding regions that have been ionised by the passage of blastwaves produced by supernova explosions.

28 5.2.1 HII regions

HII regions are regulated by three components: photoionisation equilibrium, thermal balance and hydrodynamics. Photoionisation equilibrium described before, is deﬁned as follows:

Z ∞ L a ξn ν ν dν = (1 − ξ)2n2 α(H,T) (5.1) H 2 H ν0 4πr hν Here the left hand term is (Number of hydrogen (H) atoms per volume) × (ionising photon flux) × (photoionisation cross section). The right hand term is (Number of electrons per volume) × (number of protons per volume) × (recombination coefficient). The recombination coefficient depends on the temperature and the number density of hydrogen atoms; ξ is the fraction of neutral hydrogen. Photons with energy equal to or higher than the ionisation potential of a 2 bound electron will eject it from the atom with a kinetic energy of 1/2mev = h(ν − ν0) , while the latter becomes ionised. When this electron moves into the ISM, it will be rapidly thermalised due to collisions with other free electrons, and its velocity will lie within the Maxwellian velocity distribution of the gas. The recombination of free electrons depends on the electrons’ thermal velocity distribution and the gas density. When an electron is captured by an ion, the atom usually ends up in an excited state. The lifespan of the electron in the excited state is shorter than the typical photoionisation time, and thus it will cascade down to a ground state, releasing a photon in each transition. There are two approximations for optical depth with respect to ionising photons (Lyman continuum photons) in the ambient medium: in an optically thin medium, the number of photons able to ionise the medium is larger than the number of recombinations, thus the excess of ionising photons will escape the nebula. In an optically thick medium however, all the ionising photons are absorbed by the nebula. Here, it is useful to use the "on the spot approxima- tion" to simplify the solution of the gas’ condition within the nebula: i.e. all photons emitted by recombination directly to the ground state will be locally absorbed. Therefore, only photons released by cascade transitions to the second lower level are taken into account for the energy deposition (cooling ) into the ISM. There are two important cases which must be considered for recombination processes. Case A recombination assumes a medium optically thin to the Lyman series. Thus, a substantial fraction of ionising photons will escape the nebula. Case B recombination instead assumes an optically thick medium where Lyman photons are absorbed and re-emitted immediately. Stars with temperatures ≥25000 K (hotter than B-type stars) emit a great

29 amount of photons capable of ionising hydrogen atoms. The energetic photons of stars with temperatures ≥40000 K will additionally singly-ionise helium (He+, ionisation potential ∼24.4 eV). Wolf-Rayet (WR) stars in particular emit the hardest photons capable of doubly-ionising helium (He++, ionisation potential ∼54.5 eV). Thus, a hot star will ionise its surrounding gas up to a transition point further from the star where the gas becomes mostly neutral. This transition is extremely sharp, and defines the Strömgren Radius. The species with the largest ionisation potential (He++ or He+) will be ionised closest to the star. It is important to note that radiation pressure can become important in a HII region, especially if there is a significant amount of dust present. Radiation pressure is the transfer of momentum from photons to the ambient matter. The exerted pressure is significant around the most massive stars since it strongly depends on the temperature of the star (P∝T4). The heating and cooling mechanisms of photoionisation and recombination were addressed before. Another important source of cooling is collisionally excited emission lines. Here the coolants are metals, mostly O, N, C, Fe and Si. In a nebula with typical temperatures of ∼10000 K (warm ionised phase), the thermal energy of free electrons is of a few eV, which is comparable with typical excitation potentials of the ground state fine structure levels of these metals. Thus, when free electrons impact with the metals, they will excite the bound electrons to a higher energy level, which in absence of further collisions, will radiatively de-excite while emitting photons of discrete energy (producing emission lines), which will likely escape the nebula. The efficiency of this cooling mechanism is directly proportional to the metal abundance of the gas. There are two density regimes that are relevant for collisions. The low density regime is fulfilled when each collision results in radiative de-excitation. On the other side, the high density regime is fulfilled when each collisionally excited electron is collisionally de-excited. The density that separates these two regimes is called the critical density. In the low density limit, the strength of the line rises with increasing density as ∝ n2 until it reaches the critical density, while in the high density regime, the line strength increases as ∝ n. In general, the critical density from each atom/ion is sufficiently high such that only a small fraction of the gas can meet this condition; therefore, this is not an important contributor of cooling in a galaxy.

5.3 Emission line diagnostics

In the optical wavelength range, various emission lines can be used to trace diﬀerent properties of the warm ionised gas. The recombination lines of the Balmer series (mainly Hα and Hβ) are commonly used as tracers of the kine-

30 matics of the warm ionised gas, the dust attenuation, the star formation rate and other parameters. Collisionally excited lines are used as diagnostics of the temperature, electron density, metallicity, ionisation of the gas, tracers of fast shocks and other parameters of the gas.

5.3.1 Ionised gas kinematics

The kinematics of the ionised gas is usually derived from the Hα line, but also from strong metal lines such as [OIII]λ5007, [NII]λ6583 or [SIII]λ9069 in the optical range. The velocities are derived from the spectral shift with respect to the kinematic centre of the galaxy and the velocity dispersion values from the width of the line. Galaxies can be classified based on their internal motions in two types of system. If the gas and stars follow ordered motions, they are rotationally supported systems and have a disk-like morphology. Otherwise, if the matter follows a mostly random motion, they are a pressure supported system. Spi- rals and ellipticals are the common representatives of rotationally and pressure supported systems respectively. Irregulars and distorted galaxies whose morphology has been affected by external interactions like mergers or close by passages correspond to a special group of pressure supported systems which do not have relaxed kinematics. Galaxies at high redshift (z≥5) are clumpy and irregular (Förster Schreiber et al. 2006). A great fraction of these are merger systems with chaotic kinematics, ruled by a combined effect of merger dynamics and intense feedback. In the KMOS project that comprises ∼600 galaxies, the main investigators Förster Schreiber et al. (2006, 2009); Wisnioski et al. (2015) found that the brightest galaxies from the peak of cosmic star formation (z∼2-3) onwards were found to be predominantly rotationally supported systems with large velocity −1 dispersions (vdis ∼50 - 100 km s ). However, Rodrigues et al. (2017) investigated the morphology of galaxies classified as disk systems and found that two thirds show signs of external gravitational interaction. At low redshift, massive galaxies are mainly rotationally supported systems, except the most massive ellipticals which are pressure supported. The kinematics of low-mass galaxies, which dominate in number density, are much more complex. Most of them are irregular and are dominated by feedback processes and/or galaxy interactions. Observational studies have shown that the velocity dispersion decreases with decreasing redshift. Up to now however, there is still no consensus about what causes the evolution of this velocity dispersion. Some authors have claimed that the increasing stability of the rotating disk is associated with the decrease of galaxy gas fraction with time (Genzel et al. 2011; Wisnioski et al.

31 Figure 5.1: Kinematics of Eso 338 at different scales from Bik et al. (2018). In the leftmost image, the halo shows irregular kinematics where a galactic outflow is superimposed. The middle and rightmost images trace the outflow to its origin, where the massive nuclear cluster ’cluster 23’ is located.

2015). Other works suggest that the higher velocity dispersions seen at higher redshifts are coupled to enhanced turbulent motions driven by the feedback resulting from elevated star formation rate surface densities (Lehnert et al. 2013). A third school of thought relies on elevated external gravitational instabilities at early times, such as intense gas accretion and galaxy mergers (Bournaud et al. 2011). In this thesis, I focus on the ionised gas kinematics imprinted in a special type of galaxy: a compact galaxy undergoing extreme star formation while in process of a merger event. Haro 11, as well as other BCGs, Green-Pea galaxies, and in general high-z galaxies, are undergoing an intense episode of star formation triggered either by gas infall or merger processes. In these systems, the overall gas kinematics do not only trace the dynamics of the merger, but also clearly show the imprints of stellar feedback in their interstellar medium (ISM). Their large population of young and massive stars releases large amounts of kinetic and radiative energy into the ISM, which is capable of shaping the structure of the ionised gas and of accelerating gas to high velocities. Several studies, particularly from dwarf galaxies, have shown that strong stellar feedback is able to create filaments, expanding shells and outflows, which leave footprints in the kinematics of these galaxies (see. e.g. Marlowe et al. 1995; Martin 1998). One clear example of a kpc-scale outflow is displayed in Eso 338 (see Fig. 5.1). Bik et al. (2018) found a complex biconical outflow in this galaxy that seems to be launched by stellar winds and supernovae originating from one of the most massive clusters in this system 7 (Cluster 23, Mdyn ∼ 10 M ). Thanks to the spectro-photometric capabilities of instruments like MUSE or ALMA, ionised/neutral gas structures can be now studied in velocity bins.

32 By extracting maps of the gas within a specific velocity range, it is possible to obtain a clear view of the size-velocity evolution of many structures of feedback origin. I explore this method in my thesis and prove it to be greatly powerful in tracing the different kinematic components of the gas such as outflows, expanding shells and tidal interactions (Menacho et al. 2019). The dispersion velocity aided by the line profile gives clues about the turbulence in the gas, and is directly linked to SN-induced shocks and powerful stellar winds in star forming regions. Thus, it is not surprising that large velocity dispersion values are commonly derived in starburst systems. Moreover, the velocity dispersion and the line profile also help to disentangle the kinematic component of the progenitor galaxies in the case of merger systems.

Escape velocities and mass ejection Although escape velocities are relatively slow in low-mass galaxies due to their small masses, powerful outflows with speeds ranging from hundreds or even thousands of kilometres per second are still required for gas to escape into the IGM. Outflows with velocities up to 1000 km s−1 are mostly found from absorption features in the UV spectral range, while they are hardly detected from emission lines. The escape velocity of an outflow in a non rotationally- supported galaxy can be calculated assuming a model for the mass distribution in a gravitational potential. In the most simplistic model, we assume a spher- ically symmetric isothermal potential with an outer cutoff limit at rmax. The escape velocity of an outflow at radius r is (Marlowe et al. 1995): √ p vesc = 2 · vcirc 1 + ln(rmax/r) (5.2) The circular velocity can be approximated by applying the virial theorem: p vcirc = G · Mtot/rmax. (5.3)

The total mass is estimated from the observable ’baryonic’ matter (MB) and the dark matter (MDM), yielding Mtot = MB + MDM. The quantity derived above is valid for an outflow along the line of sight. However, this is in general not the case, thus the escape velocity needs to be corrected for the effect of inclination. The intrinsic radius rint of a measured outflow robs flowing in a direction with an angle α relative to line of sight is:

rint = robs/sin(α) (5.4) ◦ ◦ Replacing r with rint in equation 5.2, outﬂows with α = 45 and 80 need apparent velocities larger than 95% and 70% their escape velocities respectively to overcome their gravitational potential. In Haro 11, we derive an escape velocity of 400 km s−1 assuming that 80% of its matter is dark matter,

33 and that the outflow is flowing almost along the line of sight. If we consider that most outflows are flowing at 45◦ of inclination with respect to the line of sight, we derive an outflow escape velocity of about 250 km s−1.

5.3.2 Ionisation parameter

Ionisation parameters are derived from emission lines of two ions with different ionisation potentials. Ionisation parameters are commonly used to trace the degree of ionisation in the ISM. They measure the flux density of energetic photons, which gives clues about the strength of the ionisation field in the ISM. Several ions in a variety of combinations have been used to evaluate the ionisation state in galaxies. e.g: [OIII]λ5007/Hα, [OIII]λ5007/[SII]λ6716,λ6731, [OIII]λ5007/[OII]λ3727, HeI λ5876/Hβ and [SII]λ6717/[NII]λ6583 (Bik et al. 2018; Keenan et al. 2017; Zastrow et al. 2011, 2013, among others.). The combined information of different ionisation parameters provides useful insight into the different zones of HII regions (Pellegrini et al. 2012). Highly ionised ions, which require energetic photons (EUV, X-ray photons) to be produced, are located close to ionisation sources, such as O-type stars, while intermediately ionised atoms are located in a transition zone between the hot central zone and the ambient medium. In this thesis we trace the ionisation structure of Haro 11 with the line ratio [OIII]λ5007/Hα given that both lines are the most intense in the optical range, and thus trace the faintest gas structures. Oxygen atoms need extreme UV photons (Ephot >35 eV; ≥350 Å) released by O-type stars to be doubly- ionised, while hydrogen atoms only need photons of lower energy (Ephot >13.6 eV). The [OIII]λ5007 emission line is produced when thermal electrons of energy ∼ 2.5 eV collide with doubly ionised oxygen ions and excite their bound electrons to a higher level, which afterwards de-excite, producing the afore- mentioned emission line. Consequently, an increase in [OIII]λ5007/Hα values indicates an increase of energetic photons capable of doubly ionising oxygen. Ionisation parameter mappings at resolutions of pc to kpc scales are highly useful for highlighting structures of feedback origin such as: ionisation channels, outflow paths, shells and cavities. These mappings applied to Haro 11 reveal various outflows as well as cavities and channels of highly ionised gas. Moreover, ionisation mapping can be used as a tracer of optical depth for LyC photons. It has been shown that the [OIII]λ5007/[OII]λ3727 ratio is a powerful indirect tracer of escaping LyC radiation (Izotov et al. 2018; Jaskot et al. 2019), since a highly ionised halo suggests a density bound system where ionising photons can easily escape the galaxy.

34 Figure 5.2: OI-BPT diagram of Eso 338 tracing shocks at diﬀerent velocities from Bik et al. (2018). Red dashed lines trace shocks at diﬀerent magnetic parameters. Blue lines show the strength of the line, which increases with increasing [OI]λ6300/Hα ratios and ionisation values.

35 5.3.3 Shock tracer Even if shocks are difficult to trace in galaxies, theoretical models provide powerful diagnostic line ratios to infer their strength (e.g. Allen et al. (2008)). The most common line ratio diagnostic in the optical range is [OI]λ6300/Hα. Neutral oxygen can be found mixed in a neutral and a lowly ionised medium. The [OI]λ6300 line is produced by collisions with thermal electrons able to transfer an energy of at least 1.9 eV to the oxygen atom. These collisions excite the bound electrons of neutral oxygen atoms to a higher level, which thereafter de-excite, emitting the [OI]λ6300 line. Fast shocks increase the kinetic energy of free electrons, and therefore a single electron can excite a large number of atoms, thereby increasing the intensity of the [OI]λ6300 line. This rise in energy is not sufficient to collisionally ionise the most common atoms (Hydrogen atoms need T∼ 40 000 ◦K), but it is sufficient to boost collisional excitations. The strength of the [OI]λ6300/Hα ratio depends on the metallicity and the magnetic parameter of the ISM. Allen et al. (2008) provided several shock diagnostics that are commonly used in literature. Additionally to a shock tracer (such as [OI]λ6300/Hα), their shock model takes as an input parameter a tracer of the ionisation degree of the gas. The reason for this dependence is that after the passage of a shock wave, the post-shocked area cools down and emits photons that can ionise the pre-shocked area, increasing the ionisation degree faster than the velocity of the shock. Thus, the photon density that is emitted is proportional to the shock velocity. Fig. 5.2 shows the OI-BPT diagram of ESO 338 (Bik et al. 2018). Over- plotted on it are line ratios for different shock velocities calculated from the shock models of Allen et al. (2008). The models were derived with the metallicity of Eso338-IG04 (comparable with the metallicity of Haro 11) and different magnetic parameters. Shocks at all velocities are traced in the gas that has relatively high [OI]λ6300/Hα values, while the strength of the shock correlates strongly with the ionisation parameter. Starburst galaxies have typical shock velocities of hundred of km s−1, while AGN can have shocks velocities up to 1000 km s−1 (Davies et al. 2000; Newman et al. 2012).

5.3.4 Dust attenuation The effect of dust on the emitted light of galaxies has been observed and recognised since the time when photographic plates started to be used at the beginning of the twentieth century. Obscured lanes and filaments were promptly recognised, especially in edge-on systems with a dusty disk like NGC 681, or even in the centre of our own Milky Way. It became evident that this effect depends on wavelength, since blue-band images showed stronger absorption

36 than images taken solely in red optical bands. Over the years, and aided by the development of better instruments with higher spectral and spatial resolution, the nature and effect of dust in galaxies has become better understood. Cosmic dust is composed of dust grains and aggregates of dust grains. The chemical composition consists of ice, various carbon compounds, metals, complex organics and silicates. There are two main types of dust grain: "graphite" or "amorphous" which is carbon-rich and "silicate" which is silicate-rich. Dust grains form from the condensation of elements on the cold surface of evolved stars or in the ISM. A large fraction of these dust grain are composed mainly of carbon, followed by nitrogen, sodium, magnesium, silicon and iron. Thus, if a galaxy has a considerable amount of dust, the elements (metals) depleted into dust grains need to be considered when estimating chemical abundances of the galaxy. Dust grains are in general irregularly shaped, porous and have sizes that span from 10−3.5 micron to about one micron. The largest dust particles are aggregates of many sub-micrometer grains that are compacted in a random order. The grain size distribution follows a power law of the form N(a)da∼a−3,5, such that most of the dust mass is concentrated in the largest dust particles (Draine 2011; Stein and Soifer 1983). The energy of the photons absorbed by dust is converted into vibrations that are afterwards mostly re-emitted in the IR spectral range. The absorption is universally proportional to wavelength, and gives rise to the common spectra of dusty systems with strong absorption in the short wavelength range, and weaker absorption in the long wavelength range compared to the intrinsic spectra of the galaxy. Dust attenuation as a function of wavelength has been observed and characterised for different types of galaxies that have slightly different dust populations and dust-size distributions. These are: starburst galaxies, the Milky Way, the SMC and the LMC. One of the main differences between these is the presence or absence of the dust bump at 2175 Å which is associated with graphite. For example, the SMC attenuation curve does not show this bump, suggesting that the SMC has a very low dust content of graphite type (Galliano et al. 2018). Dust attenuation in galaxies can be calculated from the UV, optical and IR ranges of the spectra, or even as a byproduct of spectral energy distribution (SED) fitting procedures. The attenuation for each wavelength segment (i.e. UV, blue-range in optical, ...) gives clues on the amount of dust affecting the stars that emit light in that specific wavelength section. For example, the attenuation derived in the UV provides information about the dust composition around UV-photon emitters, which are usually the most massive stars of a galaxy. Similarly, the dust present in the warm ionised phase will affect attenuation in the optical regime. In the optical range, the Balmer decrement is used to compute the attenuation as follows:

37 Rαβ_obs = 100.4[κ(Hβ)−κ(Hα)]∗E(B−V) (5.5) Rαβ_int

where Rαβ_obs is the observed Hα/Hβ flux ratio and the Rαβ_int is the intrinsic Hα/Hβ flux ratio. The intrinsic ratio is derived from atomic physics assuming a temperature and optically thick or thin gas. For HII regions, which are commonly ionisation bound (case B recombination) and have average temperatures of 104 K, the intrinsic Hα/Hβ ratio is 2.86 (Osterbrock 2006). κ(Hβ) and κ(Hα) are the extinction coefficients 1 obtained from the observed attenuation curves. E(B-V) is the colour excess of the gas due to the dust attenuation. The E(B-V) is generally used when showing attenuation due to dust in the ISM of galaxies. When correcting the spectra for dust attenuation, the following equation applies:

0.4∗E(B−V)∗κ(λ)att.curve F(λ)obs = F(λ)int ∗ 10 (5.6)

where κ(λ)att.curve is the extinction coeﬃcient for a given attenuation curve. For starburst galaxies, it is common to use the Calzetti attenuation curve (Calzetti et al. 2000); whereas for galaxies with characteristics similar to the LMC, SMC and MW, the respective attenuation curves are used. The most commonly used papers describing these curves are: Fitzpatrick (1999)) for the LMC; Prevot et al. (1984) for the SMC; and Seaton (1979) and Cardelli et al. (1989) for the Milky Way.

5.3.5 Density and temperature The average density of the gas is measured from the intensity ratio of collisionally excited lines of the same ion. The excitation energies of these lines are approximately the same, such that the relative excitation rates depend almost exclusively on the ratio of their collision strengths. The temperature is also measured from collisionally excited lines of the same ion, but the lines that are used arise from diﬀerent excitation energies, such that their excitation rates depend almost exclusively on the temperature. These lines originate from transitions populating the upper and lower level, which are called auroral and nebular lines respectively. Many ions are used to determine the density and temperature. The most common are [OIII] and [SIII] as they are the brightest lines that can trace both

1There are many online sources which provide the extinction coeﬃcients as functions of wavelength. One of the more commonly used is provided by the York University in Canada: http://www.cadc-ccda.hia-iha.nrc-cnrc.gc.ca/community/YorkExtinctionSolver/coeﬃcients.cgi

38 parameters to further distances. In the optical range, the following line ratios are used for determinations of density:

OII λ3726/[OII]λ3729 SII λ6720/[SII]λ6730

whereas for the temperature these are:

SIII λ6312/[SIII]λ9069 OIII λ4363/[OIII]λ5007 NII λ5755/[NII]6548 OI λ5577/[OI]λ6300

Each diagnostic traces the gas at the location where its ion is found. For example, the temperature obtained from the [OIII] lines is heavily weighted towards the gas that is highly ionised. In the same manner, the [SIII] line is used to trace the temperature of an intermediately ionised gas, and the [NII] line a slightly less ionised gas. The neutral oxygen line [OI], on the other side, represents a gas that is mostly neutral or lowly ionised. Thus, the density and temperature derived from different ion-tracers are not completely comparable. One of the strongest constraints in deriving temperatures is the weakness of the auroral lines. These are from 10 to 100 times weaker than the Balmer lines. In the optical range, the [OIII] and [SIII] auroral lines are brighter than the [NII] or [OI] lines in starburst galaxies. Thus, these lines are used to derive the temperature to the farthest part of the galaxy, while other lines can only be used for determining the temperature in the central regions. As it was pointed out before, the temperatures derived from different tracers are not comparable. In cases where the temperature of each ionisation zone is required, as is the case for metallicity estimations, these can be estimated based on empirical relations from the measured temperature of a different ionisation zone. Relations between the [SIII] and [OIII]-temperatures are found, e.g, in the following studies: Garnett (1992) developed temperature relations for a three-zone model, with the [OIII]-temperature tracing a highly ionised gas, the [SIII] ions the intermediately ionised gas, and the [OII] ions tracing a lowly ionised gas; Izotov et al. (2016b) took as basis the [OIII]-temperature and derived relations to obtain the [SIII] and [OII]-temperatures as a function of the gas metallicity; Binette et al. (2012) present a relation for the [OIII]- [SIII] temperature derived from extragalactic HII-regions while analysing the origin of the [SIII] and [OIII]-temperature discrepancies found in the data.

39 Pérez-Montero and Díaz (2003) present relations for the [OII]-[OIII]-[SIII]- temperatures based on observations of HII regions. In Paper 2, we derive gas density from the [SII] lines and temperature from the [SIII], [NII], and [OI] lines. Given that the [SIII] auroral line is brighter than the other two, we were able to derive with it the temperature to the largest distance yet in the halo. From the other two ions, we obtained the temperature only from the central starburst. Nevertheless, we use the [SIII] temperature values, characterising the temperature in the intermediately ionised gas, to derive the temperature in the highly and lowly ionised gas. We used simple relations presented by Garnett (1992), who addressed the temperature stratiﬁcation of a nebula in a three-zones model, separating the lowly, intermediately and highly ionised gas.

5.3.6 Metallicities Chemical abundances provide fundamental information on the evolution of galaxies. The metal content in a galaxy is linked to the past star formation activity, the inﬂow of pristine gas, the outﬂow of metal-enriched gas, and the mechanisms that are involved in the mixing of metals in the ISM. On spatially resolved scales, chemical abundances provide clues about the ongoing mechanisms that are metal-enriching or metal-diluting the gas, and whether a galaxy was assembled inside-out our outside-in (i.e. SF happened from inside to outside, or vice-versa) (Maiolino and Mannucci 2019). The term "metallicity" (Z) refers to the total mass of metals relative to the total mass of baryonic matter:

Z = Mmetals/Mbaryons (5.7) In astrophysics, all elements heavier than helium are considered to be metals. The abundance of a chemical element is usually expressed relative to the hydrogen abundance as follows:

12 + log(X/H) = 12 + log(NX/NH) (5.8)

where NX and NH are the number density of the element X and H (hydrogen). A value of 12 was added in this notation with the only purpose that the derived abundances are always positive. The abundance ratio of two elements X and Y is:

log[X/Y] = log(NX/NH) − log(NY /NH) = log(NX/NY ) (5.9) Given that oxygen is the most abundant element metal produced, the oxygen abundance is usually used as the tracer of metallicities of the ISM.

40 12 + log[O/H] = 12 + log(NO/NH) (5.10)

The metallicity of a region or galaxy is often compared to the metallicity of the Sun. If it is higher/lower than the solar value of 12+log(O/H) ∼ 8.86 (Asplund et al. 2009), it is referred as super-solar/sub-solar metallicity. It should be noted however, that the expression 12+log(O/H) is an approx- imation of the real metallicity of the ISM (Maiolino and Mannucci 2019).

Stars and the production of chemical elements

Elements like hydrogen (H) and most of helium (He) and lithium were formed by primordial nucleosynthesis after the Big Bang. From the remaining elements, the largest fraction of heavier elements originates from stellar nucleosynthesis processes within massive stars and in supernova explosions. Young stars, while on the main sequence, burn hydrogen in their cores through the pp-chain or CNO-chain. The CNO-chain take place in stars with mass M≥1.3 M . When the hydrogen inside the core is exhausted and the star evolves off the main sequence, helium starts to burn, producing carbon (C, isotope 12C) and oxygen (O, 16O). Outside the core, an upper layer of hydrogen continues to burn. The subsequent evolution depends on the mass of the star. Stars with masses lower than 2 M undergo a helium flash event (a huge amount of helium burns in a short time, and radiative pressure becomes dominant over gravitational force), which expels the outer envelope, forming a planetary nebula and leaving behind a white dwarf remnant. In the CNO-cycle of intermediate mass stars (M ∼ 2 – 8 M ), nitrogen is produced in a secondary reaction at the expense of oxygen and carbon, which reduces the abundances of C and O, and considerably increases the N abundance. Given that the CNO-cycle is favoured at high metallicities, the production of nitrogen can be considerably enhanced in metal-rich environments. In intermediate mass stars once the helium is exhausted in the core, helium and hydrogen continue to burn in outer shells. In this phase, called the Asymptotic Giant Branch AGB, many He-flash events occur. These events favour a con- vective process, bringing up part of the enriched elements from the core to the surface. In these He-flashes, a fraction of the star’s C/N-enriched mass is released in strong stellar winds. Finally, most of these metals are released in the last evolutionary stage, when these stars expel their outer envelopes, leaving behind a white dwarf. In massive stars (M>8 M ), once the helium in the core is exhausted, other heavier elements such as Neon (Ne), sodium (Na), magnesium (Mg), silicon (Si) and oxygen are produced. Some of these additionally start to burn when

41 the temperature rises considerably, while the lighter elements are still produced in the upper layers of the star. This process is halted when the core becomes composed of iron (Fe) and nickel (Ni). These stars then undergo a core-collapse supernova. The supernova explosion originates a shock wave that propagates outwards, heating the stellar ejecta which undergoes nucleosynthesis processes. This results in an enrichment of mostly alpha-elements (isotopes that are multiple of 4, e.g O, S, C) (Karakas and Lattanzio 2014). White dwarfs accreting mass from companion stars, which thereafter undergo Type Ia supernovae, are important sources of iron, but also of silicon, sulphur and other iron-peak elements. Binary neutron star mergers are also a rich source of iron elements.

Timescale of the production of the main elements

Fig. 5.3 shows the time needed to produce the most common elements (O, C, N, Fe) after a single star formation episode. These heavy elements are created mainly in the last stage of stellar evolution, therefore their production timescales can be seen as the time for when these elements are released into the ISM. The largest fraction of oxygen and other alpha-elements is produced by massive stars (M > 8 M ) that have short lifetimes. Consequently, they start to enrich the ISM shortly after the onset of star formation, from about ∼3 Myrs to about 40 Myrs. While alpha-elements, especially oxygen, compose the vast majority of core-collapse supernovae ejecta, a smaller fraction of nitrogen, carbon and iron produced in massive stars is also released to the ISM by supernovae. In the upper panel of ﬁg. 5.3, the "CC-SNe" label shows the timescale when massive stars explode as supernovae and the fraction of O, N, C and Fe produced and released during that period (<40 Myr) (Karakas and Lattanzio 2014). Carbon and nitrogen are mainly produced in intermediate-mass stars that live longer, therefore the nitrogen/carbon-enrichment of the ISM is delayed from 50 to a few hundred of Myrs with respect to the onset of star formation (Karakas and Lattanzio 2014). Iron and other iron peak elements are also subject to a delayed enrichment (from 50 Myr to a few Gyr), given that they are mainly produced in type Ia supernovae. In the upper panel of Fig. 5.3, the "Ia-SNa + AGB" label shows the timescale when evolved intermediate-mass stars (AGB-stars) eject their outer envelopes. This timescale overlaps with the time when SNe-I event happens (Maiolino and Mannucci 2019, and references therein). The lower panel of Fig. 5.3 shows the cumulative mass of metals produced after the onset of a single star formation episode. This plot clearly shows that

42 Figure 5.3: Production timescales of the main elements O, N, C and Fe from Maiolino and Mannucci (2019) and Vincenzo et al (in prep.). In a single star formation episode, oxygen, and in general alpha-elements, are released shortly after the onset of star formation by supernovae from the most massive stars. In the same event, other elements such as carbon, nitrogen and iron are also released, but in lower abundance. Most of the carbon and nitrogen are then released later by intermediate-mass stars. Even more delayed is the enrichment of iron, which is released mainly in supernovae type-Ia.

43 most of the oxygen is formed and released shortly after star formation happens. The enrichment of iron increases constantly after an initial ﬂattened increase. Carbon-enrichment is boosted in the beginning by core-collapse SNe, and later by AGB stars. The enrichment of nitrogen is dominated by the intermediate- mass stellar population.

Methods used to derive the metallicity

Three main techniques are used to derive chemical abundances. The first method, the "direct method", is based on direct measurements of temperature and intensity ratios of recombination or collisionally excited lines with respect to the Hβ line. The second method uses photoionisation models to interpret the strength of the brightest lines and by doing so, constrains and derives the gas- phase metallicity. The third method, called the "strong line method", uses the strength of some of the brightest lines relative to Hβ . The latest is calibrated from photoionisation models or empirical relations based on HII regions where the metallicity has been previously derived from the temperature-dependent method. Dust particles are made of elements that are not always taken into account when computing elemental abundances. Photoionisation models generally include dust depletion and yield metallicities in terms of the total metallicity of the system. The direct method instead computes the metallicities of the gas phase only. All three methods have strengths and weaknesses. The direct method is preferred since it directly measures the element abundances from relations derived from atomic physics. The underlying idea is that the strength of a metal-line is defined by the elemental abundance and emissivity, with the latter depending on temperature and density. The main weakness arises from an in- accurate or wrong derivation of the temperature. The second method relies on models that can be exploited in a large parameter range such as: different ionisation fields, dust content, effect of radiation pressure, multi-fitting features, etc. However these models are still simplistic, especially in the assumed geom- etry of the HII region, and the use of fixed relations between various elemental abundances, which are not necessarily valid in many star forming regions. The advantage of the third method is that since it is built from strong lines, it can be used where lines are in general weak, e.g. it can trace the metallicity to the farthest distances in the halo or be used in high-z systems. In the second paper, I present metallicities derived from the direct method and the strong line method.

44 The direct method

The direct method can be applied to collisionally excited lines (CEL) in the optical range, or metal recombination lines (REC) which are mostly found in the IR range. Both options differ in how they depend on temperature. The metallicities derived from CEL/REC have a strong/weak dependence on the temperature. A strong constraint in using the REC lines, which are the most direct way to derive elemental abundances, is the weakness of the metal recombination lines. These are from 100 to 1000 times weaker than the Balmer lines, such that they are limited to high signal-to-noise HII regions (Maiolino and Mannucci 2019). Here I will describe the metallicities derived from the CEL lines. In the multi-phase medium of galaxies, ions with a characteristic degree of ionisation populate an ambient medium that is defined by temperatures in a specific range of values. Singly ionised species such as O+ ,N+ and S+ populate the low ionisation zone, while others with higher ionisation potentials like O+2 are found in the highly ionised gas. Additionally, some species like S+2 seem to populate a transition zone with an intermediate ionisation field (Garnett 1992). Given that it is almost impossible to compute the temperature in each of these zones due to the auroral lines of some ions being up to 100 times weaker than the Balmer lines, many authors depart from the temperatures derived from the [OIII] or [SIII] lines, tracing the temperature of the high and intermediate ionisation zones respectively, and infer the temperature of the remaining zones on the basis of empirical relations. In paper 2, we used the [SIII]-temperature for the intermediate ionisation gas, and followed the pre- scription from Garnett (1992) to infer T[OII] and T[OIII] for the low and high ionisation gas. For the O+ ,N+ and S+ species, we used T[OII], for O+2,T [OIII], and for S+2, T[SIII]. We used the formulation of Izotov et al. (2006) to derive the chemical abundances of the most common elements. These authors derived empirical relations from the abundances of metal-poor galaxies sampled from the SDSS survey. The ionisation correction factors (ICF) were derived from photoionisation models and are used to correct the abundances by taking into account the ions absent in the data. The [OII]λ7320+7330/Hβ and [OIII]λ5007/Hβ ratios, with their corresponding temperatures, are used to estimate the O+ and O+2 abundances. In similar way, the [NII]λ6548+6584/Hβ , [SII]λ6717+6731/Hβ and [SIII]λ6312/Hβ ratios with their corresponding temperatures are used to derive the N+,S+ and S+2 abundances respectively. The oxygen abundance is then computed from the O+ + O+2 abundances after assuming that the abundance of O+3 ions is negligible. The nitrogen abundances are derived from the N+ abundances and an ICF that depends on the oxygen abundance. The

45 sulphur abundances are derived from the S+ and S+2 abundances and the corresponding ICF.

Strong line methods There is a long list of strong line calibrators that are used to derive metallicities. Here, I will limit the list to calibrators used in the optical range. Some of them use two or more strong collisionally excited lines, while others measure the line strengths relative to the Balmer lines. Nevertheless, this is an indirect method that can have strong degeneracy with some parameters such as ionisation. A complete list of strong line calibrators can be found in Maiolino and Mannucci (2019) and Pérez-Montero and Díaz (2005). The most popular diagnostics are:

• R2 = log([OII]λ3727/ Hβ ) This is strongly degenerate with the ionisation parameter.

• R3 = log([OIII]λ5007 /Hβ ) This also depends strongly on the ionisation parameter.

• R23 = log(([OII]λ3727 + [OIII]λ5007,4958)/Hβ ) This is a widely adopted calibration since it is less aﬀected by the ionisation ﬁeld of a HII region. The R23 method is double branched and its values switch between (12+log(O/H)∼ 8 – 8.5), which makes this calibrations dependent on metallicity.

• N2 = log([NII]λ6583 /Hα ) This method primarily traces the nitrogen abundances and cannot be used if a galaxy does not follow the common N/O-O/H relation on which this diagnostic is built. Additionally, it is sensitive to the ionisation structure of the region.

• S2 = log([SII]λ6716,λ6731 /Hβ ) This method is strongly degenerate with the ionisation parameter.

• O32 = log([OIII]λ5007 /[OII]λ3727 ) This line ratio is mostly used to trace the ionisation parameter. This method has a secondary dependence on the metallicity and is constructed from the metallicity-ionisation parameter relation.

• O3N2 = R3 - N2 This method is sensitive to the ionisation parameter and the nitrogen to oxygen ratio.

Almost all of these relations are strongly degenerate with the ionisation parameter. In cases where the lines involved are distant from each other in wavelength, they are sensitive to the dust attenuation correction.

46 Some authors have constructed other methods from combinations of the line ratios presented above, aiming to correct for the ionisation parameter dependence. In paper 2, we used one of these methods proposed by Pilyugin and Grebel (2016). These authors calibrated their methods based on HII regions whose metal abundances were previously derived using the temperature- dependent (direct) method. The "S-calibration" (S-calib) presented in Pilyugin and Grebel (2016) uses as a basis the following line ratios:

• N2 = [NII]λ6548+6584/Hβ

• S2 = [SII]λ6717+6731/Hβ

• R3 = [OIII]λ5007/Hβ

The computation of the oxygen abundances is separated ininto two branches based on the N2 values. The upper branch is computed when log(N2) ≥ -0.6 as follows:

12 + log(O/H) = 8.424 + 0.030log(R3/S 2) + 0.751logN2 (5.11) +(−0.349 + 0.182log(R3/S 2) + 0.508logN2) × logS 2

and the lower branch is computed when log(N2) < -0.6 as follows:

12 + log(O/H) = 8.072 + 0.789log(R3/S 2) + 0.726logN2 (5.12) +(1.069 − 0.170log(R3/S 2) + 0.022logN2) × logS 2

This calibration was found to have a very low scatter (<0.2) over a large range of metallicities.

5.3.7 Temperature and metallicity variations The estimation of the physical condition in a star forming region assumes that the photoionised region has a homogeneous temperature, density and chemical composition. At ﬁrst glance, these assumptions are valid since the gas in HII regions has the same origin, with the same chemical composition. How- ever, due to ionisation stratiﬁcation, the temperature can vary by few percent. Galaxies, and even HII regions, are complex systems with physical conditions that often cannot be reproduced by simple relations, as at least one of these assumptions is not valid. Moreover, the estimation of temperature can be far

47 from trivial given that some physical processes can alter the strength of temperature sensitive lines. Density inhomogeneities and ionisation stratification of the ISM can also produce variations in the temperature. Determining the correct temperatures is crucial for the estimation of metal abundances. Here I will summarise the origin of the temperature and abundance discrepancies (Peimbert et al. 2017, and references therein). It has long been known that some HII regions and Planetary Nebulae show temperature discrepancies when these are estimated from collisionally excited lines (CEL) and from metal recombination lines (RLs) of the same element. Some discrepancies were also found when comparing the CEL temperatures such as T([OIII]) with temperatures derived from HeII, T(HeII) or from the Balmer continuum T(Bac). Torres-Peimbert et al. (1980) were the first to address this problem and found that the resulting abundance discrepancies were in general systematically higher (factor 5-20) in some Planetary Nebulae. Es- teban (2002) summarises the origin of these fluctuations as follows:

• High density condensation.- In high density condensation, the critical density of a particular nebular line can be reached while remaining be- low the higher critical density of auroral lines. Thus, the nebular lines are collisionally de-excited but not the auroral lines, which leads to an over- estimation of temperature and an underestimation of metal abundances. This problem can be solved by using other temperature-sensitive lines with higher critical densities.

• Chemical inhomogenities.- If the gas is not homogeneously mixed, the gas can be found in two-phases. The RLs are emitted in the high-density, low temperature, metal-rich condensations that could have been formed by recent ejections of stellar material. These clumps are then embedded in low-density, high-temperature, metal-poor diﬀuse gas where the CELs are produced.

• Shocks.- Peimbert et al. (1991) was the first to investigate the effect of shocks on the emission lines of photoionised gas. These shocks can originate from strong stellar winds or supernovae. The latter is an important contributor of mechanical energy, given that more than half of all O-type stars explode within HII regions. The authors found that the strength of the auroral lines start to rise when the velocity of the shock is about 60 km s−1 (or 90 km s−1 in a composite model: stellar radiation + shocks), and increases up to a factor of 2.5 when the shock velocity is about 160 km s−1 . At larger shock velocities, this factor slightly decreases to about 2.4. This effect will considerably increase the temperature of the gas, while the metal abundances will be underestimated.

48 • Conduction fronts.- These fronts originate from a conductive layer situated between the hot material of shocked stellar winds and the visible plasma. Their eﬀect is similar to that of the shocks.

Another mechanism that can affect the determination of temperature and abundance is the so called κ-distribution. This mechanism was first introduced by Nicholls et al. (2012) and later investigated by Dopita et al. (2013), who explored the possibility of a different distribution than the common Maxwell- Boltzmann distribution for free electrons in HII regions or Planetary Nebulae. These authors showed that in extreme conditions where several energy transport processes (shocks, plasma wave, ion heating, fast electrons produced by an X-ray/extreme-UV) take place, the distribution of free electrons follows a Lorenzian-like distribution. One outcome of such a distribution is that the CEL lines are weak, but not the Balmer lines. This mechanism affects all methods, since the metallicities are derived from the strong line ratios relative to the Balmer lines, or in the case of the photoionisation models, the spectral lines cannot be properly interpreted. Ferland et al. (2016) however shows that in HII regions and Planetary Nebulae, the time needed to thermalize these electrons is shorter than the heating/cooling timescales, thus this mechanism can be ruled out for most of the ionised nebulae. While temperature fluctuations provide a simple explanation for discrepancies found in the temperature and abundance estimations, there are still difficulties in explaining all properties observed in a nebulae. Moreover, the energy sources or physical mechanisms producing such fluctuations are in most cases unknown.

49 50 6. The impact of stellar feedback

6.1 Energy released by massive stars and star clusters

Massive stars are the main contributors of energy and momentum to the ISM of star forming galaxies. In starburst galaxies, they establish the dynamics of the system and are the principal engine behind the evolution of many dwarf galaxies. Their population is sampled from a mass distribution dubbed the Initial Mass Function (IMF), which describes the initial distribution of stars with masses between M and M+∆M (Salpeter 1955). In general terms, the number of stars sampled decreases with increasing mass; consequently, only a small fraction of the newborn stars are massive. This means that a low mass star cluster produces only a few massive stars, while a massive star cluster produces a large number of massive stars. Thus, the largest populations are 5 found in super star clusters (Msc ≥10 M ). The most important physical properties of massive stars are shown in Table 6.1. The lifespan of stars is inversely proportional to their mass, while their luminosity and temperature, which tell us about the strength of their radiation ﬁeld, increase exponentially. Nevertheless, although the lifespan of massive stars is very short, they considerably aﬀect their surroundings. When the most massive stars (O-type, M≥30 M ) are on the Main-Sequence (still burning hydrogen in their core), they release large amounts of UV photons that ionise the surrounding medium. The ionisation front around these stars will then expand as long as the massive stars remain on the Main-sequence.

◦ Mass (M ) Luminosity (L ) Teﬀ ( K) Main-Sequence lifespan (Myr) 5 600 17 000 70 10 10 000 22 000 20 15 17 000 28 000 10 25 80 000 35 000 7 60 790 000 44 500 3.4

Table 6.1: Physical properties of massive stars taken from the Astralian telescope national facility website www.atnf.csiro.au

51 Afterwards, the amount of UV photons released is substantially reduced as the star’s surface cools down. The eﬀective temperature decreases by a factor of approximately 2, which implies a drop in the production of ionising photons by a factor of at least ∼ 10, until the star explodes as a supernova (Tenorio- Tagle and Bodenheimer 1988). While the massive stars are still on the main sequence, radiative pressure exerts a pressure on the matter, which starts to expand.

Massive OB-type stars on the Main-Sequence also eject part of their mass into their surroundings through strong stellar winds. For mass-loss rates up −5 −1 to 10 M yr and for stars with M>20 M , this continuous energy output over their lifetimes is comparable with the energy released by a typical supernova explosion (Tenorio-Tagle and Bodenheimer 1988). After ∼ 4 Myrs, the most massive stars end their life as supernovae, releasing mechanical energy 51 of ∼10 ergs into the ISM. Wolf-Rayet stars (WR, M≥50 M ) are a special subgroup of massive stars whose hydrogen envelopes have been removed by strong stellar winds, leaving the extremely hot stellar core uncovered. These stars are an important source of highly energetic photons capable of doubly ionising He II (54.4 eV), and form the most powerful stellar winds with typical velocities of thousands of km s−1.

Less massive stars (M∼8–15 M ) do not produce a signiﬁcant amount of ionising photons and their stellar winds are considerably weaker. Instead, they inject mechanical energy to the ISM at the end of their lives when they explode as supernovae.

Thus, a star cluster with a mixed population of stars releases ionising and radiative energy during the ﬁrst ∼4 Myr of its evolution, when the most massive stars have not yet exploded as SNe. In this time, the ionisation and radiative feedback reach their maximum. At the same time, but in a longer time span, massive stars inject mechanical energy in the form of kinetic energy and momentum into the ISM through stellar winds. Starting from ∼4 Myr onwards, the mechanical energy is almost exclusively provided by supernova explosions, which last until 40 Myr, when the less massive stars explode as supernovae (Leitherer et al. 1999).

The impact of any of these feedback components depends strongly on the star cluster mass. In an association of stars and low-mass clusters, supernovae dominate the mechanical energy input. Ionising, radiative feedback and stellar winds become important in massive star clusters.

52 6.1.1 Energy output of ionising radiation, stellar winds and supernova Despite the complexity of the mechanisms involved in stellar feedback, the energy, momentum and pressure exerted onto the ISM can be roughly estimated from a few parameters obtained from observations. The radiation pressure is the momentum transferred from photons to matter and is proportional to the opacity of the matter to radiation. This pressure increases considerably if the ambient medium is dense and dusty. From Veilleux et al. (2005), the radiation pressure exerted onto the surrounding medium is:

34 p˙rad = 1.3 × 10 τLbol,11 [dyne] (6.1)

where τ is the optical depth of the medium to radiation and Lbol,11 is the 11 bolometric luminosity in units of 10 L . The rate of mechanical energy injected into the ISM is proportional to the SFR:

41 −1 −1 E˙∗ = 7 × 10 (SFR/M yr )[ergs ] (6.2) The momentum injection rate from SN and stellar winds to the ISM is then:

33 −1 p˙∗ = 5 × 10 (SFR/M yr )[dyne] (6.3) The injected momentum drives turbulence in the ISM, while the injected energy regulates the thermal phases of the ISM (Kim and Ostriker 2018). However, a more precise study of the impact that stellar feedback has on an ambient gas with specific characteristics requires a feedback model with different input parameters. This model requires: a) The gas density and its distribution in the ambient medium where the cluster or star forming region is located; b) the number of massive stars per mass bin, which is sampled from a given Initial Mass Function (IMF); c) the total power released by stellar winds and supernovae, each of which is usually set to be 1051 ergs. In the case of stellar winds, this energy is released over the star’s lifetime; and d) the supernova rate that depends on the cluster mass. Additionally, these models can be configured for two types of star formation: an instantaneous mode, which is the case for star clusters where all stars are born at the same time, or a continuous mode, that is representative for a star forming region where stars are born continuously during a period of time. For a cluster population with 250 massive stars, the supernova rate is given by an average of one supernova every 0.2 Myr over a period of 40 Myr. If a star cluster has an even larger population of these stars, the frequency of these events increases considerably i.e. the time between supernovae decreases. Moreover, in a clustering of star clusters (e.g. the star forming knots in Haro 11), the supernova rate is considerably elevated

53 given that all cluster members contribute to SNe. In the case of a star forming region with a continuous star formation rate, the supernova rate depends on the star formation rate parameter. One common software that computes evolutionary synthesis models of star clusters or star formation regions is Starburst 99 (Leitherer and Heckman 1995). In this software, the following input parameters can be configured: a) metallicities: from sub-solar to solar; b) IMF: Salpeter or Kroupa; c) star formation laws: instantaneous star formation (normalised to a total mass of 6 −1 10 M ) and continuous star formation with a rate of of 1 M yr . d) Top cutoff masses: depends on the models, from 30 and 100 M . Using Starburst 99, Leitherer et al. (1999) has presented results of stellar feedback outputs along the evolution of star clusters and star forming regions. In Fig. 6.1 and 6.2 we show the evolution from 1 Myr to 1 Gyr of the luminosity, energy, momentum and pressure outputs from a star cluster with an instantaneous burst in the first column and from star forming region with a continuous star formation in the second column. Both models have a metallicity of 0.02 Z , a Salpeter IMF and a high cutoff mass of 100 M . In Fig. 6.1 and 6.2 we report on different panels (a-j) the main results. These are:

• Panel a) shows that the mechanical luminosity from stellar winds and supernovae combined remains constant the ﬁrst ∼40 Myr for an instantaneous burst of star formation, while panel b) shows that it increases with time for continuous star formation. As we clearly see, supernovae are the main contributors (or long-time contributors) of mechanical feedback.

• Panel c) shows that the ratio of the luminosity of ionising photons over the bolometric luminosity (nonradiative) evolves similarly for both the continuous and instantaneous star formation cases. In the ﬁrst 3-4 Myr, the ionising luminosity is at its maximum while the most massive stars (all born at the same time) are still on the main sequence. After this time, it drops continuously. Panel d) shows that for a continuous SF law, the gradient of the ionising photon luminosity to bolometric luminosity ratio is shallower over time.

• Panel e) shows that in the ﬁrst 4 Myr after an instantaneous burst, almost all the matter released to the ISM comes from stellar winds. Thereafter and up to 40 Myr, the mass loss rate is dominated by SNe. Panel f) shows however, that for a continuous SF law, SWs dominate the mass loss rate for a longer time (10 Myr) and are then replaced by SNe, which continuously inject hot matter into their surroundings.

54 • Panel g) shows that the ratio of the mechanical to bolometric luminosity increases somewhat until ∼40 Myr and then drops substantially. The nonradiative energy becomes enhanced once the strong winds from Wolf-Rayet stars start to operate. Afterwards, SNe take over. For the continuous SF law seen in panel h) however, the ratio is lower but remains constant.

• Panels i) and j) show the total energy injected into the surrounding media for a Salpeter IMF and mass cutoﬀ at 30 and 100 M . For an instantaneous SF law i), the energy increases when SNe start to explode and remains constant for about 10 Myr; however for a continuous SF law j), the injected energy increases continuously with time. Given that the the IMF with a lower mass cutoﬀ samples more stars, the energy output of such stellar population dominates over the population sampled with a heavy IMF due to the large number of less massive stars that will explode as SNe.

These results show that during a timescale of 1 Gyr, the feedback released to the ISM by a star cluster or a region with continuous star formation is substantially different. This is mainly because all massive stars in a star cluster are acting on the ISM in the first 5 Myr and momentum output is over after 40 Myrs. Contrary to that, in a continuous star forming region, the energy and momentum output increase with time, reaching their maximum after 40 Myrs. This means that in a star cluster, we know with precision when the stellar radiation, strong stellar winds or supernova operate, and consequently we can estimate the timescale when feedback mechanisms such as ionisation radiation, radiative feedback and shocks are at their maximum. In a continuous star formation region it is considerably more difficult to disentangle the impact of these mechanisms on the ISM. Using a more realistic evolutionary synthesis model, Agertz et al. (2013) studied the energy and momentum output from stellar feedback of massive star clusters and their impact on the surrounding medium. The authors found that before the first SNe start to explode, stellar winds and radiative pressure have considerably cleared out dense regions, which reduces subsequent star formation. Stellar winds from massive stars launch in total about 10 M of hot gas at v∼3000 km s−1 into their surroundings, approximately the same hot gas mass that is ejected by SN at comparable velocities (Leitherer et al. 1999). At solar metallicities, the specific luminosity is dominated by radiation over stellar winds and SNe, however the momentum injected into the ISM in all three processes is approximately the same:p ˙rad ∼ (Lbol/c) ∼ p˙SN ∼ p˙winds ∼ (Lmech/v). Fig. 6.3 shows the momentum and energy injected by star clusters with

55 Figure 6.1: Evolution of different feedback components from a stellar cluster (instantaneous SF, first column) and a star forming region (continuous SF, rate of −1 1 M yr , second column) presented in Leitherer et al. (1999). The panels: a) and b) show the evolution of the mechanical luminosity output from stellar winds and SNe; c) and d) show the ratio of the ionising luminosity over the bolometric luminosity for stellar winds and SNe; and e) and f) show the mass loss rate from stellar winds and SNe. 56 Figure 6.2: Evolution of different feedback components from a stellar cluster (instantaneous SF, first column) and a star forming region (continiuous SF, rate −1 of 1 M yr , second column) presented in Leitherer et al. (1999). The panels: g) and h) show the ratio of the mechanical luminosity over the bolometric luminosity. i) and j) show the total energy input for a Salpeter IMF with upper mass limit of 30 and 200 M .

57 Figure 6.3: The left panel shows the momentum injected in the ISM by radiative pressure, stellar winds and supernovae for solar metallicity (solid lines), and 0.01 the solar metallicity (dashed lines) as a function of time. The radiation pressure is strongly dependent on the mass of the stellar clusters, and may dominate over the mechanical momentum input of supernovae or stellar winds in massive clusters. The right panel shows the energy released to the ISM by the shocked ejecta of stellar winds and SNe. This results are presented in Agertz et al. (2013).

5 6 masses of 10 and 10 M presented in Agertz et al. (2013). The panel on the left shows the evolution of the integrated momentum that is injected into the ISM with solar (solid lines) and sub-solar (0.01 Z , dashed lines) metallicities. Radiation pressure and stellar winds deposit momentum into the ISM only 5 during the ﬁrst 4 Myr. In a cluster with M∼10 M and solar-like metallicities, the momentum deposited by radiative pressure and supernovae is comparable, while the momentum injected by stellar winds is about 30%. In a sub-solar metallicity gas however, the momentum deposited by supernovae dominates. 6 In star clusters with masses M≥ 10 M and solar-like metallicities, radiative pressure injects about an order of magnitude higher momentum to the ISM that stellar winds and SN feedback. In all cases, stellar winds only deposit a small fraction of the total momentum, and their impact is even lower for sub-solar metallicity environments. The right panel of Fig. 6.3 shows that the energy injected into the ISM by the shocked SN ejecta dominates over the energy injected by stellar winds. This simulation shows two main results: the strength of radiative feedback increases with increasing metallicity, so that at solar-metallicities it can have similar eﬀects in the ISM to those of supernova explosions; and second, this radiative feedback increases considerably in clus- 6 ters with mass ≥10 M . Murray et al. (2011) already stressed the importance of radiation pressure in starburst galaxies with young and massive star clusters. This pressure in-

58 creases considerably in dense and dusty media (see equation 6.1). The au- 6 thors found that radiative feedback from star clusters with M≥ 10 M is the main driver of the expansion of HII regions where they are embedded. More- over, this pressure may be powerful enough to develop superwinds and thus launch galactic-scale outflows of cold gas (T∼ 104 K) into the intergalactic medium (Murray et al. 2011; Nath and Silk 2009). Cold gas outflows seen in absorption in some galaxies might only be explained by radiative feedback, as a domination of mechanical feedback would subject the cold clouds to Kelvin- Helmholtz instabilities, or even evaporate them in a short time (∼1Myr) (Mur- ray et al. 2011). Thus, simulations that include radiation pressure feedback have demonstrated that this mechanism plays a significant role in starburst galaxies that have massive star cluster populations. Specifically, it suppresses the star formation and affects the metal distribution in those galaxies (Murray et al. 2011; Nath and Silk 2009; Wise et al. 2012).

6.2 Structures of stellar feedback origin

In galaxies with intense star formation, structures at every scale shaped by stellar feedback are ubiquitous. In our galaxy, but also in several nearby galaxies, structures such as bubbles, shells, filaments, loops, arcs, holes and cavities were observed in a wide wavelength range (Deharveng et al. 2010; Heiles 1984; Hunter et al. 1993; Martin et al. 2002; Sivan.J 1974, among others.). Most of these structures were traced in optical, IR or radio observations. On the other side, X-ray observations are crucial to trace the hot rarefied gas embedded in bubbles or cavities. These structures are created by a localised injection of large amounts of energy into the ISM. The energy sources range from individual massive hot stars, which form small-scale structures of several pc, through OB-associations that create structures of hundreds of pc, to star clusters and star cluster associations that form kpc-scale structures. Although these arrangements are natural prod- ucts in such a violent ISM, only a small fraction of them have been observed in the Milky Way and in nearby galaxies, mainly due to lack of high spatial and spectral resolution instruments. With the new generation of highly sensitive instruments and telescopes such as MUSE and ALMA, the number of detected features of stellar feedback origin has rapidly increased. Bubbles, shells, arcs, loops, filaments, holes and cavities are transient objects that are visible for a few Myr until they lose their identity or dissolve. In broad terms, a bubble surrounded by a shell forms first. Once it breaks due to instabilities, part of the shell can be expelled to larger distances forming arcs or loops. The interior hot gas leaks out, clearing holes and cavities in the ISM, and dragging part of the shell on its way out, which forms the filaments. The

59 Figure 6.4: Projections of a shell seen at diﬀerent velocities. Here v=0 is the systemic velocity of the shell. (Sketch made by the author). physics describing bubble evolution is presented in a later subsection and is mostly extracted from Weaver et al. (1977) and Veilleux et al. (2005).

6.2.1 Observations of large scale structures

In the 70’ s and 80’ s, several HI surveys revealed pc- to hundred of pc-scale HI shells in the Milky Way (Heiles 1984; Hu 1981; Weaver et al. 1977, See Fig. 6.5). Some of these shells appeared as rings whose radii smoothly changed as function of velocity. This effect is due to a uniform and symmetric shell that expands isotropically. Fig. 6.4 shows the projection effect in velocity of such structures. If this shell were observed in velocity bins, each bin would show a different part of the shell. At the centre, the projected image is a large ring that will shrink with increasing or decreasing velocity. However, in reality most of the shells observed were either identified only in a narrow velocity range or show somewhat distorted morphology over a large velocity range. Moreover, the majority of the largest kpc-scale shells were traced either in their approaching or in their receding part. (Heiles 1984). Shells were also traced in the warm ionised gas phase and by dust emission. Some of those ionised gas shells are surrounded by a neutral gas shell. Well known examples are the Cygnus superbubble, the Gum Nebula and the Orion shell. All seem to be created by OB associations of stars (Brand and Zealey 1975; Sivan.J 1974). More recently, the Spitzer satellite has mapped the Milky Way in the micron wavelength range, which traces dust emission. These observations (24- 100 µm) revealed a large number of bubbles (see top image in Fig. 6.5), most of which are associated with HII regions (Churchwell et al. 2006, 2007). De- harveng et al. (2010) and Kendrew et al. (2016) studied the morphology and

60 Figure 6.5: Top: Bubbles and ﬁlaments in part of the Milky Way galactic plane as seen from the Herschel satellite. The hot rareﬁed gas is seen in blue, while dark colors show the regions where cold gas and dust is present. Credit ESA/Herschel/PACS, SPIRE/Hi-GAL Project. Bottom: Image of the star forming region NGC 3576 showing a large arc. Image Credit & Copyright: Desert Hollow Observatory.

61 the energy sources of about one hundred of them, and found that the majority were triggered by feedback of O-type stars. The authors found that these bubbles, of a few pc in size, contain not only hot ionised gas, but also hot dust in their interiors. Moreover, one quarter of these bubbles seem to have triggered star formation in their shells by compression of the existing dust in ionisation fronts. Maciejewski et al. (1996) reported on a large expanding HI shell with a size of ∼320 pc in diameter in the Aquila region of our own Milky Way. The authors estimated that energy injections of about 100 SN over 10 Myrs were needed to create the observed remnant. An even larger expanding HI shell was found by McClure-Griffiths et al. (2002). This structure of 1.4 kpc diameter is expanding away from the spiral arms into the low density halo gas, a condition that likely favoured its rapid expansion. When a bubble bursts open in the side of the halo, it creates large chimneys where hot gas streams out. One such HI galactic chimney was observed by McClure-Griffiths et al. (2003) in the Sagittarius-Carina spiral arm. It has a length of 1 kpc and a width of 0.6 kpc. Its interior wall shows plenty of small scale filaments, loops and drips. Other small-scale structures such as the arc in the region NGC 3582 (see Fig. 6.5) are also seen in other regions of the Milky Way. The first extragalactic shells were found from deep Hα observations in the LMC (Braunsfurth and Feitzinger 1983; Meaburn et al. 1981). These have sizes up to kpc in radius and are surrounding massive OB star associations, star clusters or star cluster complexes. Their interior gas is hot, rarefied and highly ionised. High resolution images show a filamentary structure in their interior borders, most likely due to the effect of radiative pressure and mechanical feedback. Inside them, star formation is still active in HII regions. Small-scale shells are also visible around SN remnants, and inside young and massive HII regions (Braunsfurth and Feitzinger 1983; Meaburn et al. 1981). The most important star forming region in the LMC is "30 Doradus", the central part of which can be seen in Fig. 4.3. This dynamic region has several shells (LMC2, LMC3 and LMC4) centred on massive star clusters and OB associations. Large-scale shells in complex structures become larger by merging with smaller shells around individual sources. The shell complex in 30 Doradus is embedded in an even larger shell of HI gas, a fact that hints to large-scale propagated star formation inside a large cloud (Dopita et al. 1985). Similar shells were found in other galaxies in our neighbourhood. Brinks and Bajaja (1986) have mapped HI holes with sizes from 0.1 to 1 kpc in M31. Although kpc-scale structures are not common in the MW, they appear to be frequent in galaxies with intense star formation (localised or general). Kamphuis et al. (1991) found in M101 a HI superbubble of 1.5 kpc, that is expanding at 50 km s−1 towards the halo. de Blok and Walter (2000) have

62 reported on a HI shell of 2 × 1.4 kpc in NGC 6822 that does not show evidence of expansion. Curiously, the authors did not find an obvious mechanism that could have formed this supershell. Rand and Stone (1996) found an even larger expanding HI supershell in NGC 4631. This structure has a diameter of 2.8 8 kpc and mass of ∼ 10 M . The authors suggest two possible mechanisms that might have originated it: stellar feedback of thousands of massive stars or a high velocity cloud. Hunter et al. (1993) found several ionised supershells with sizes ranging from hundreds of pc to 1.2 kpc, in a small sample of amorphous and irregular galaxies that are undergoing an intense episode of star formation. Beside the shells, the authors reported on kpc-scale ionised filaments. Other luminous Kpc-scale filaments were observed in M82, NGC 3079 and NGC 1569, all with a length of up to ∼3 kpc (Cecil et al. 2001; Lynds and Sandage 1963; Martin et al. 2002). These filaments are associated with superwinds originated from the blow-away of bubbles. Thus, these filaments may be pieces of the shell that has been dragged out by the outflowing hot gas from the bubble interior (Hunter et al. 1993). Kpc-scale bubbles need energy injections of thousands of OB stars in a relatively short interval of time. The ages of the observed superbubbles range from few to one hundred of Myr (de Blok and Walter 2000; Kamphuis et al. 1991; Rand and Stone 1996).

6.3 Bubbles and superbubbles

6.3.1 Physics of bubbles - superbubbles Here I describe the wind model of Weaver et al. (1977) presented in Kim et al. (2017), which provides basic results that are in agreement with most observed bubbles. It has been proven in simulations that a continuous wind solution can reproduce the results of models which include interval energy injections from supernovae. Given a star cluster mass Mcl, the number of SN explosions is: NSN = Mcl/m∗. Where m∗ is the star mass formed per SN, which is sampled by an 3 IMF. For a Mcl ≥ 10 M using a Kroupa IMF, m∗ ∼ 100 M for a time range of 3-40 Myr. The interval time between SN is then:

−1 ∆tSN,6 = tli f e/NSN = 0.4Mcl,4 [Myr] (6.4)

4 Where Mcl,4 is the star cluster mass in units of 10 M and the energy per SN 51 is assumed to be 10 ergs (E51). The total energy released in the bubble is E = E t SB SN ∆tSN Stellar winds and supernovae deliver energy to the surrounding medium. If radiative losses are insigniﬁcant, an expanding bubble is produced by the

63 overpressure exerted in the ambient medium. Fig. 6.6 shows the structure of a bubble, which is described as follows: The hot interior ejecta (stellar wind or SN matter) moves outwards with supersonic velocities. The density in this −2 area (A) drops as function of radius nw(r) ∝ r . The outflowing ejecta meets a shock in R1 (called "terminal shock"). When shock waves moving outwards meet the inner boundary R1, thermal conduction (at work in region B) almost fully converts the kinetic energy into thermal energy in zone (B). The shock in R1 is adiabatic due to the higher density in zone (B). Zone (B) is a hot (T≥ 106 K) region of shocked ejecta with a relatively homogeneous low density. This zone is separated from the shell (C) by a boundary region called the "contact discontinuity". The shell consists of swept up interstellar matter, which can be neutral (HI), ionised gas (HII) or both (a layer of ionised gas and an exterior layer of neutral gas). In the contact discontinuity, the temperature of the gas sinks to the temperature of the shell. Here, electron thermal conduction is at work. Around 40% of the heat is radiated away through collisions of metals producing UV lines such as [OIV]λ1035. The remaining 60% produces evaporation of matter from the shell, which mixes with the hot gas of the shocked ejecta in region (B). This process is the main supplier of matter in region (B). Lastly, R2 shows the boundary of the shell with the ambient interstellar medium. The interior, rarefied hot gas of a bubble is a soft X-ray emitter. The evolution of the bubble can be described in three phases: i) In the first phase, the bubble rapidly expands. ii) The second phase is characterised by formation of the shell, consisting of swept up matter. iii) In the last phase, radiative losses in the shell become important. 2/5 The first phase is very short and the shell expands as R1(t) ∝ t with a velocity v1(t) ∝ r1/t. Energy is supplied to the system at a constant rate through stellar winds or supernovae. The surrounding matter is constantly swept up while the bubble grows rapidly following an adiabatic expansion, which is faster than the sound speed of the ambient medium. If the energy supplied is high, then radiative losses are negligible. The first phase finishes when the time scale for radiative cooling of swept up matter is similar to the age of the system. In the second phase, the swept up matter collapses and forms the shell. The zones of the bubble described before and shown in Fig. 6.6 are established. Thermal conduction in R2 becomes important, as well as the evaporation of the shell mass. The density in the shocked ejecta zone varies as function of time and can increase considerably due to radiative cooling effects. The shell temperature ranges from 8000 to 10000 K. The terminal shock radius (R1) depends on the energy input and on the energy that escapes the shocked ejecta zone (B). In this phase, the shell expands as follows:

64 Figure 6.6: Structure of a bubble: In the centre sits an energy source, which can be a single massive star, a star cluster or a group of star clusters. This object is surrounded by a hot rareﬁed cavity, where the supersonic hot ejecta from stellar winds or supernova is released. (B) is a zone of shocked ejecta. (C) is a zone of swept up matter, which constitutes the shell. All these structure is surrounded by the ambient medium. Adapted sketch from the Fig. 1 of Weaver et al. (1977).

!1/5 E51 3/5 Rshell(t) = 60 t6 [pc] (6.5) ∆tSN,6 namb,0

where namb,0 = ρ/(1.4mH) is the hydrogen number density of the ambient medium at time t6 in units of one Myr. Radiative cooling gains importance with time. The third phase starts when parts of the shell where the density is high, start to cool radiatively. For a continuous energy injection model, applied for massive clusters with many SN explosions prior the shell formation and/or a low ambient gas ( ∆tSN,6namb,0 <

65 022 0.044E51 ), this occurs in a time:

4 0.28 −0.71 −0.28 tshell_cool(t) = 1.8 × 10 E51 namb,0 ∆tSN,6 [yr] (6.6) otherwise for less massive clusters with a low supernova rate it occurs at:

4 0.22 −0.55 tshell_cool(t) = 4.4 × 10 E51 namb,0 [yr] (6.7) The ﬁnal evolution of the bubble depends on the eﬃciency of radiative losses. The expansion of the shell is calculated as follows:

0.37 −0.37 −0.62 Rshell_cool(t) = 5.5 E51 ∆tSN,6 namb,0 [pc] (6.8) The inner and outer shocks (R1 and R2) vary as function of time as follows: 0.58 0.44 Rt(t) ∝ t and Rt(t) ∝ t . If the supplied energy becomes unimportant, the remnants of the shell still expand due to momentum conservation. The hot rarefied interior gas remains in overpressure relative to the ambient medium. After a while, if the shell does not experience Rayleigh-Taylor instabilities, radiative loses and the ambient gas pressure become important and thus shell expansion stops. In zone B, almost 90% of the thermal energy will be radiated away. In the shell, the cooling gas can form small-scale high density condensations of gas, which can form a new generation of stars. The stability of the shell depends strongly on the ambient medium’s density and its distribution. If a bubble is expanding into a gas with a density gradient, it may begin to accelerate and experience Rayleigh-Taylor (RT) instabilities. The expanding shell becomes Rayleigh-Taylor unstable under two conditions: The first is when the shell moves into a density gradient and begins to accelerate. After the RT condition is established, the shell fragments, while the acceleration stops and the interior hot gas starts to leak out. The second condition is met by a discontinuous supply of energy through supernova explosions. The hot interior gas causes the shock waves of subsequent supernova explosions to decelerate to the speed of sound. If this condition is met before the shock wave reaches the shell, the thermal and kinetic pressure of the shock wave will be larger than the shell pressure. The shell will accelerate while it becomes RT unstable (Tenorio-Tagle and Bodenheimer 1988). In the theory of shocks, at increasing temperature, the shock velocity decreases. For bubbles centred on massive clusters, the larger number of SNe will increase the temperature of the bubble rapidly if radiative losses are not sufficient to dissipate the energy input. Thus, shock waves can become subsonic after many supernovae. Once the shell is fragmented, it usually follows the "blowout" phenomenon. Part of the shell are dragged out, forming filaments. If the energy is deposited predominantly on one side of the shell or if there is a (stronger) density gradient on that side, the blowout will occur only on that side.

66 Rayleigh-Taylor instabilities most likely occur when the bubble is still expanding. In a HII region, a bubble may begin to form when the stars are still embedded in the parental cloud. Once the bubble reaches the borders of the molecular cloud, it will burst open and form larger ﬁlaments. The circumfer- entially oriented arcs seen in several galaxies may be shell remnants that were accelerated by overpressurization of the interior hot gas on that side (McCray and Kafatos 1987) (See the Fig. 6.5).

6.3.2 Bubbles in simulations

Bubbles can originated from radiation pressure, stellar winds and supernova explosions. Radiation pressure and stellar winds, as they are released at the same time, can conjointly form bubbles up to several hundreds of pc (e.g Cyg OB2, Abbott et al. 1981). However, most of the observed bubbles, and especially superbubbles (kpc in radius), have increased their size mainly due to the energy input of supernova explosions (Hunter et al. 1993; Kamphuis et al. 1991; Marlowe et al. 1995). The gas density and its distribution are fundamental parameters in the evolution of bubbles. Simulations using a homogeneous, stratified, exponential or a Gaussian gas distribution do not produce the same results. Bubbles expanding in density gradients are accelerated and after some time may become Rayleigh-Taylor unstable. On the other hand, higher gas density produces smaller bubbles (Tenorio-Tagle and Muñoz-Tuñón 1998). Tomisaka and Ikeuchi (1986) ran a simulation of a bubble created by 360 SN, with a SN explosion every 0.2 Myr. The bubble developed in an ambient medium with a density of a) 0.1 cm−3 and b) 1 cm−3. When the expansion velocity became 8 km s−1, the radius of the bubble with a density of 0.1 cm−3 was about 1.3 kpc, while this was smaller for the higher density model. The same model was run with a stratified gas distribution inside a disk galaxy. The dense ambient gas model developed a thick, egg-shaped shell with a hot interior gas, while in a low density ambient medium, the model developed a somewhat larger, thick cylindrical-shaped shell, with the side perpendicular to the disk opened and leaking hot interior gas into the halo. The shell was completely ionised due to the low density ambient medium. In that model, the cooling time was larger than the expansion time. In the case of a higher density ambient medium, this condition can be reversed; radiative cooling can be very efficient and can lead to the formation of a neutral gas shell. The rate of SN events is another fundamental parameter. In low-mass star clusters, the time between supernova explosions is large enough such that once the last SN has exploded, the shell velocity drops substantially. If the velocity dispersion of the surrounding gas (turbulent medium) is larger than the expan-

67 sion of the shell, which likely happens for low-mass clusters, the shell and all the interior hot gas will merge with the ISM (Kim and Ostriker 2015, 2018). Early models encountered difficulties developing larger shells. For instance, when the number of massive stars is on the order of thousands, only a homogeneous ISM can develop a few kpc-scale shell. However this condition is utopic given that there is stratification of the ambient gas at large scales, especially in disk galaxies. As an example, this condition is not met when a shell is accelerated into the halo, becomes RT unstable and thereafter fragments. In a two-phase model (cold and hot gas phases), most of the momentum is released into a low-density, intercloud medium before formation of the shell. At the time of shell formation (when the hot gas mass is at its maximum), the momentum has decreased to about half of the initial momentum injected by the SN. Magnetic forces do not play a role in the first stages but do in the last stages of the shell evolution. When a large number of SNe explode in the same location and in a short interval of time, the evolution of the shell is abrupt. The momentum increases if the ambient density is low (Kim and Ostriker 2015). The left panel of fig 6.7 shows three snapshots in the evolution of a bubble presented in Kim et al. (2017). The bubble grows by continuous injection of SN energy in intervals of 0.01 Myr in an ambient medium with a hydrogen −3 density of navg = 1cm . The density, pressure and temperature of the bubble are shown in the first, second and third rows of the first panel respectively. The first column shows the bubble when the shell starts to cool radiatively. At this stage, the pressure inside the hot interior is elevated, keeping the bubble nearly spherical. For this specific configuration of input parameters, the bubble reaches the scale height (H, defined by the drop in the density for a factor of ∼2.7. See the second column) shortly after the shell starts to cool radiatively. Thus, the bubble evolves to this stage without considerable energy loss while keeping its morphology. Here 44% of the total injected energy is kept in the bubble while 15% is converted into thermal energy in the hot medium. The right column shows the bubble when it reaches 2H. Its morphology has changed considerably due to RT instabilities in the shell. The second panel shows some physical properties (density, temperature, ram pressure, thermal pressure and velocity, from top to bottom respectively) of the ‘zoom in’ region marked with a white box on the top-right of the first panel. The right panel shows how the interior hot gas interacts with the warm shell matter, and this in turn, with the ambient medium (warm and cold gas). The red contours show the boundary of the interior hot gas (T∼ 106 K) within the shell. This boundary is very clear in all maps. The interior rarefied hot gas is highly overpressurized and has a high velocity. It will remain hot if the SN rate is high, otherwise it will cool down. Within this gas there are some islets of cooler and warmer dense gas. The hot interior gas is surrounded by a thin and somewhat dense

68 gas with temperature of about 104 K (visible in the density, pressure and velocity maps). Lastly, a strong forward shock progresses from the boundary of the shell into the ambient medium (Kim et al. 2017).

Figure 6.7: Snapshots of a bubble model presented in Kim et al. (2017). The ﬁrst leftmost column shows the bubble when the shell starts to radiatively cool. The bubble reaches one scale height in the middle column and 2 scale heights for the third column. The top, middle and bottom rows show the density, pressure and temperature respectively. The rightmost panel shows the density, temperature, ram pressure, thermal pressure and velocity (top to bottom images) of the region displayed in the white box at the top-right of the ﬁrst panel.

In general, the mass and volume of the hot gas increase continuously while SN explosions still happen, and decrease after that, while the radius of the remnant still increases (Kim and Ostriker 2015).

6.4 Galactic winds, ﬁlaments and galactic fountain

Large-scale winds can be powered by stellar feedback processes or AGN. Here I will describe galactic winds driven by stellar feedback mechanisms. Kim et al. (2017) suggested that only shells that break out before radiative cooling starts to become important can drive hot winds with high mass loading factors. Thus, powerful sources with a minimum cluster mass on the order of 6 Mcl ≥ 10 M (massive super star clusters) are required. The energy needed to

69 Figure 6.8: Galaxies showing galactic winds: a) M82 is a prototype of galactic winds (Image credit: NASA, ESA and the Hubble Heritage Team (STScI/AURA)). b) NGC 1482. Top: composite image of Hα in red, [NII] in green and X-ray in blue. Bottom: in black are the shock-excited regions ([N II]λ6583/Hα≥1) showing the regions affected by the wind, while red shows the disk galaxy (Images taken from Veilleux et al. 2005). c) NGC 3070. The x-ray emission (Chandra) is shown in blue, and the Hα ionised gas emission is shown in gold. In the same panel, the top-right image is a zoom in of the central part and clearly displays in gold the superbbuble with its filamentary pillars. In violet is the X-ray emission which is co-spatial with the filaments. (Images taken from Veilleux and Bland-Hawthorn 1998). inflate a superbubble however, can come from multiple clusters. In fact, it is common to find a clustering of star clusters in starburst galaxies. In Haro 11 for example, each of the three star forming regions (knots A, B and C) contain

70 6 a few clusters with masses Mcl ≥ 10 M and many more less massive clusters (Adamo et al. 2010; Hayes et al. 2007), that are capable of driving kpc-scale superbubbles (Menacho et al. 2019). Galactic winds were first observed in the nearby starburst galaxy M82 by Lynds and Sandage (1963). These bipolar galactic winds developed after a kpc-scale superbubble break out that was inflated by the nuclear starburst of the galaxy. In disk galaxies, superbubbles develop faster in the direction perpendicular to the disk plane due to the lower density. These winds are then easily detected in high-energy observations. Most of the galactic winds were observed and analysed from UV and optical absorption lines, while some of them were also traced in emission lines and even in continuum emission over a wide wavelength range from X-ray to radio (Heckman et al. 2015; Veilleux and Bland-Hawthorn 1998; Veilleux et al. 2005). These emission lines are produced by radiative losses and deceleration of wind material. The detection of winds is somewhat difficult due to their low gas density. Line ratios involving shock excitation lines such as [SII]/Ha or [OI]/Ha are important tracers of winds. Galactic winds were observed in the centre of the Milky Way, powered most likely by central young star clusters (Morris and Serabyn 1996; Yusef- Zadeh et al. 2000). Galactic winds were also observed in other galaxies such as: M82, NGC 253, NGC 3079, NGC 4631, NGC 1482 and NGC 891, Arp 220. The morphology of the outflows appears to be diverse: they can be egg- shaped (NGC 3079, Cecil et al. (2001)), bipolar with double loop (Arp 220), biconical (M82, Bland and Tully (1988); Lynds and Sandage (1963)) or they might have a filamentary morphology (NGC 1567, Martin et al. (2002)). Fil- aments were detected in Hα from few pc to 10 Kpc in size. X-ray emission seems to co-exist with some of these filamentary structures. Some of them are limb-brightened, a fact that suggests that these are sitting in a hot hollow structure. Some filaments do not show these characteristics and are interpreted as cold disk gas that was entrained by the wind. Filaments are transient structures and will vanish with time. Galactic winds were detected with velocities ranging from several km s−1 to more than thousand of km s−1 (Chisholm et al. 2015; Heckman et al. 2015). Winds in NGC 3079 reach up to 1500 km s−1 (Cecil et al. 2001) . There seems to be a correlation between the gas phase seen in the outflows and the outflow velocity. Hot gas outflows were detected with the highest velocities (Veilleux et al. 2005). Interestingly, the mass-load fraction carried by the winds decreases with increasing star formation rate (Heckman et al. 2015). The outflow gas mass detected from the warm ionised gas and neutral gas ranges from 105 7 to 10 M , although this mass is a lower limit, since not all the gas flowing in the winds may be detected (Veilleux et al. 2005).

71 Few galaxies have powerful winds that can escape the gravitational potential of the galaxy. Assuming that halo drag is unimportant, about 5 to 10% of the neutral material may escape in starburst driven winds (Rupke et al. 2005; Veilleux et al. 2005). Dust was also observed in galactic winds, which may be entrained in the winds, given that it was usually found in the filaments (Cecil et al. 2001; Radovich et al. 2001). Galactic fountains normally form from dense warm gas, which is accelerated by kinetic energy from SN feedback to velocities that are not high enough to escape the gravitational potential. Thus, this gas will rise to a certain altitude and then will turn and fall back, forming a galactic fountain (Kim and Ostriker 2018). Consequently, SN explosions in discrete episodes will lead to periods where inflows dominate and others where outflows dominate in the fountain flow.

72 7. The violent nature of Haro 11

7.1 Haro 11 in the literature

Haro 11 is a well known starburst blue compact galaxy. It stands out among others, not only for being one of the most luminous BCGs in the local universe, but mainly because it is the ﬁrst of few galaxies where Lyman continuum emission has been detected (Bergvall et al. 2006; Leitet et al. 2011). Morphologi- cally, it is a system in an advanced stage of a major merger between a gas rich and a more evolved progenitor, whose appearance, kinematics and projected orientation resemble the Antennae galaxy (Östlin et al. 2015).

10 With a stellar mass of 1.6 ×10 M (Östlin et al. 2001, 2015), Haro 11 is located at the high mass limit in the (not well defined) classification of BCGs. In contrast to most BCGs, it has a low gas content. The neutral gas reservoir of 8 5 × 10 M detected in 21cm emission and absorption (MacHattie et al. 2014; Pardy et al. 2016) is about three times lower in mass that the ionised gas mass 9 (∼1.37 × 10 M from Menacho et al. 2019). The molecular gas however, is not well constrained. Its mass ranges between half to three times the neutral gas as inferred by CO and dust measurements (Cormier et al. 2012). Haro 11 is also a Lyα emitter (Hayes et al. 2007), whose radiation constitutes of a diffuse halo component attributed to resonance scattering, and of concentrated emission from knot C, possibly related to an outflow. The left panel in Fig. 7.1 shows the integrated Hα map (Voronoi binned to a S/N≥5 in Hα) obtained with MUSE. The huge ionised halo extends over 30 kpc in diameter down to a sensitivity level of 1.5 ×10−20 erg s−1 cm−2 arcsec−2. In addition to the overall spherical distribution of the ionised gas, there are some striking features. For instance, the arc in the north-east and the network of filaments and clumps in the southern hemisphere. The image in the right panel was taken with the the Hubble space telescope (HST) and shows the arrangement of stars, warm gas, and in less proportion, dust in the central 4 × 3 kpc of the galaxy. The centre encloses three bright knots: A, B and C and an obscuring dusty arm west of knot A and B (dubbed the ’ear’ in Östlin et al. 2015)

73 Figure 7.1: Haro 11 seen with the MUSE instrument and HST telescope. The left image shows the MUSE Hα map from Menacho et al. (2019) displaying the full extent of the ionised halo. The right HST composite images show central 5x5 kpc, which focus on the starburst region. Credit: ESO/ESA/Hubble and NASA.

7.1.1 The cluster population in Haro 11

Adamo et al. (2010) studied the young star cluster population of Haro 11, and found that approximately four stars out of ten were born within star clusters, making Haro 11 among the most efficient known star cluster factories. Around 200 clusters were identified with masses ranging from 104 to ∼107.28 M (Östlin et al. 2015). Figures 7.2 and 7.3 show the distribution of star clusters identified in Haro 11. The most massive clusters populate the centre of the star forming regions knots A, B and C. Knots B and C are the most massive 8 star forming regions, with total stellar masses close to and higher than 10 M (Östlin et al. 2015). In the region of knot A marked in Fig. 7.3 there are about 35 star clusters 3.8 6.7 with masses between 10 to 10 M . Of these, 16 clusters have masses 5 M≥10 M and are aged between 3.5 to 20 Myrs, but most of them are very young (<5 Myrs). Two star clusters were identified to have stellar mass M≥106 M , both are about 20 Myrs old. In the region of knot B marked in Fig. 7.3, about 34 star clusters were 5 identified by Adamo et al. (2010). About 19 of them have masses M≥10 M and ages of about 3.5 Myrs with the exception of one that is about 20 Myrs 6 old. Five clusters have stellar masses M≥10 M , one of which is the older cluster. In the region of knot C marked in Fig. 7.3, about 22 star clusters were identified. Östlin et al. (2015) noted however that Adamo et al. (2010) grossly underestimate the total flux of the central compact knot, whose virial mass is

74 Figure 7.2: Cluster population of Haro 11 from Adamo et al. (2010) superimposed on the image from the HST. The image shows the position of the clusters identiﬁed in Haro 11. Additionally, the colour illustrates the age (blue is young, red is old) and the mass of the clusters (size of marks: small/big circles for the low-mass/mass-rich clusters). The youngest clusters (M<3.5 Myr) are located mainly in knot B and the dusty arm. Knot A has a spread cluster population, with ages from 1 to 50 Myr. Knot C has a predominantly older cluster population 5−6 ∼ 10 Myr. Some old (>1 Gyr) and massive clusters (10 M ) were identiﬁed towards the northwest and southeast part of knot B. Superimposed images made by the authors using the HST image (credits: ESO/ESA/Hubble and NASA) and the cluster population by Adamo et al. (2010).

75 Figure 7.3: Cluster populations of the three knots A, B and C identiﬁed by Adamo et al. (2010). The image shows in blue, green and red colours, the cluster populations of knots A, B and C respectively. The darker tones show the very young cluster populations while lighter tones indicated older cluster popu- 5 lations. Clusters with masses M≥10 M are marked with a yellow ring and the 6 most massive clusters with masses M≥10 M have an additional black ring. We clearly see that most of the star clusters in the knots are super star clusters. Knot B is a factory of very massive star clusters, but knot C host the most massive 8.4 nuclear star cluster of the galaxy (Mvir =10 M ). The cluster population in knot A is predominantly younger, but it contains also some clusters of about 20 Myrs. The population of knot B is homogeneously young while in knot C it has a mixed population of young and old clusters. (Image made by the author).

76 8.4 about 10 M within 40 pc. The excess in the SED ﬁtting of knot C suggests that this nuclear region host more than one stellar population and is more consistent with a nuclear star cluster. More than half of the clusters identiﬁed in Adamo et al. (2010) in knot C are well separated from this central region, therefore their mass and ages might be well constrained by these authors. The average age from these clusters is about 10 Myrs. The majority (∼90%) of the star clusters in Haro 11 were formed during the last 40 Myr, the age of the current starburst. The cluster population is very young in this galaxy, with a cluster age distribution that peaks at 3.5 Myr (Adamo et al. 2010; Östlin et al. 2001). Knot B especially, has a very massive and young star cluster population that is releasing large amounts of ionising photons. The cluster population in knot C is on average 10 Myr old, somewhat older than in knot B. Knot A and the dusty arm have a mixed cluster population, but the young cluster population dominates. Thus, Haro 11 is a young and rich factory of super stellar clusters. Each of them comprises a large number of massive stars at ages that are actively contributing to stellar feedback. Thus, we are capturing Haro 11 at a time when the ionisation and mechanical feedback produced by its stellar components is at its maximum.

Supernova rate in the star forming knots A, B and C

Given the importance of massive stars for metal-enrichment of the ISM, I analysed the supernova rate in the star forming regions in paper II. This analysis helps to identify the clusters that have contributed the most to enriching the ISM with metals, and also provides information about the places where these metal are released. The star cluster population in Haro 11 is very massive and young. I used the derived star cluster masses and ages by Adamo et al. (2010), except for knot C whose mass is underestimated (Östlin et al. 2015). Knot C has properties of 8.4 a nuclear star cluster (Mvir ∼10 , re f f = 40 pc from Östlin et al. 2015) with ongoing star formation for perhaps hundreds of Myrs, though the last ∼40 Myrs in a star bursty episode. Thus, using the stellar mass and ages of the clusters in the galaxy I computed the number of supernova events in time steps of 0.2 Myrs for each cluster mass using Starburst 99 (Leitherer et al. 1999). We set up the following parameters in the Starburst 99 simulation: i) Fixed star formation law for all cluster, but knot C. For knot C we use a continuous −1 8.2 star formation rate of 4 M yr (equivalent to a mass assembly of 10 in 40 Myrs); ii) a Kroupa IMF with a high cut oﬀ mass of 120 M ; iii) Supernova explosions happens for stars M> 8 M ; iv) Stellar synthesis models of Geneva v00.

77 This quantity is then used to derive the number of supernova events that have happened so far in each star cluster, as well as the number of supernova events that have happened in the last 5 and 10 Myrs in each star forming knot. Fig. 7.4 shows the star cluster distribution in knot A (blue colour), knot B (green colour) and knot C (red colour). The star clusters with more than 200 SNe in the last 10 Myrs are shown with a yellow ring, while the clusters with more than 1000 SNe are additionally shown with a black ring. The cluster population in knot A is predominantly young, but knot A also hosts several clusters older than 5 Myrs. I found 9 clusters where more than 200 SNe events happened. Of these, four star clusters had more than 1000 SN events. Almost all of these clusters are located in the core of the knot, while only two with >200 SNe are in the outskirts of the star forming region. The star cluster with the largest amount of SN events is about 20 Myrs old and had about 11000 SN events in the last 10 Myrs. It is located slightly east of the knot A core. Although the population of knot B is very massive, most of the star clusters are very young; therefore I found only 6 star clusters where more than 200 SN events happened, two of which had more than 1000 SN events. The cluster with most SNe (∼7000 SNe, Age∼20 Myrs) is situated slightly to the north compared to the core of knot B. While knot C is the most compact star forming condensation, with less star clusters than the other knots, it dominates the number of SN events in the starburst region. About 12 star clusters had more than 200 SNe, 6 of which had more than 1000 SNe. From them, the nuclear cluster in knot C dominate the number of supernova in all the galaxy. Thus, more than 80% of the supernova events in the last 10 Myrs happened in this super massive star cluster. This analysis points out that most of the metals released into the ISM in the last 10 Myrs originate mainly in knot C, followed by knot A, and in less proportion, in knot B. Although we might expect a tight correlation of the metal abundance and the supernova events in these knots, this is not evidently clear in Haro 11. Knot B has the most metal rich gas, while knot C has a low metal abundance. We study the physical characteristics of all three knots and try to explain this discrepancy in Paper II.

7.1.2 Complex kinematics

The ISM of Haro 11 is rich in gas structures that have been traced at wavelengths ranging from X-ray to radio. These structures give insight into the complex kinematics of this merger system. Prestwich et al. (2015) studied the X-ray hot gas in this galaxy and found hard X-ray emission coincident with knot B that hints at a high energy source, most likely an intermediate mass

78 Figure 7.4: Supernova rates in the star forming regions. The blue, green and red dots show the positions of the star clusters identiﬁed by Adamo et al. (2010). The star clusters with more than 200 SN events in the last 10 Myrs are shown with a yellow ring. The star clusters with more than 1000 supernovae in the last 10 Myrs are shown additionally with a black ring. In knots A, B and C there are four, two and six clusters respectively that had more than 1000 SN events in the last 10 Myrs. In the centre of knot C sits the most massive nuclear star cluster of the system. (Image made by the author)

79 black hole in a low accretion mode. Knot C is a compact soft X-ray emitter that was linked to strong stellar winds released by X-ray binaries populating this knot. Additionally, soft X-rays from thermal emission are detected in a central structure surrounding the main knots, and the dusty arm. Several kinematic components have been reported at all wavelengths in Haro 11. In the warm UV gas phase, Grimes et al. (2007) found a low-velocity outflow at − 80 km s−1 and FWHM ∼300 km s−1, and a weaker component at − 280 km s−1 associated with a high-velocity outflowing wind. Rivera- Thorsen et al. (2017) further studied the low ionisation UV metal absorption lines (mainly from SiII) that trace the neutral ISM, towards knot C, and found evidence for clumpy ISM moving with velocities ranging from − 400 to 200 km s−1. The Lyα emission analyzed by Hayes et al. (2007) shows an intense peak at the base of knot C and hints of a bipolar outflow, which might favour the shift of the Lyα emission from the resonance wavelength, thus allowing these photons to escape the galaxy. Despite the high SFR in Haro 11, it has been notoriously difficult to detect the 21 cm HI emission, since this system has very little neutral gas content. Pardy et al. (2016) found that the bulk motion of the neutral HI gas is moving at +56 km s−1 and a FWHM of 77 km s−1. Kinematics of the warm ionised gas in the optical range were mainly analysed by Östlin et al. (2015) using high resolution (R∼ 10000) integral field spectroscopy of the [SIII] line. The authors emphasise the multi-component nature of the galaxy within velocities ranging from − 130 to 130 km s−1. Three components were identified: one related to the dynamics of the merger, a second linked to the dusty arm and a third at − 100 km s−1 around knot A. In the centre of Haro 11, there is a velocity gradient that was reported by Östlin et al. (2015) and James et al. (2013), who suggest rotational motions. Moreover, James et al. (2013) interpreted it as a rotation disk. In Menacho et al 2020 (Paper II) we show however, that this gradient is due to the motion of the progenitor cores in process of coalescence. Our MUSE observations cover a large field of view and we do not find any sign of rotation in the halo. Moreover, the kinematics of the galaxy is dominated by virial motions and by gas that has been accelerated by the effect of feedback. In Menacho et al. (2019) (Paper I) we show the complexity of the ionised gas components of Haro 11, that were traced in velocity bins of 50 km s−1 from − 400 to 350 km s−1. The ionised gas and the stars at the centre seem to have the same kinematic properties. Namely, they are tracing the motion of the merger dynamics in the system. These studies give insight into the complex kinematics of this merger system. Moreover, there is a puzzling offset between the gas traced in all phases. The neutral gas seems to be concentrated at systemic velocities, while the ionised gas is moving at velocities where no 21 cm emitting gas is seen.

80 Figure 7.5: MUSE instrument. Credit Eric Le Roux / Service Communication / UCBL / MUSE.

7.2 MUSE data, observations, data reduction and methods

MUSE is the most powerful integral-field spectrograph (see Fig. 7.5) in the optical range at the moment. Its unique spectro-photometric capabilities: high sensitivity, high spatial resolution, high resolving power and simultaneous spectral coverage in a large wavelength range; allows MUSE to provide a 3D view (2D spatial, 1 spectral) of the observed objects in the sky. MUSE was principally built at CRA-Lyon by a consortium of seven institutions. MUSE is mounted at the Very Large Telescope (UT4) and began to operate at the beginning of 2014. The field of view is split into 24 channels which are directed to a spectrograph. At the same time, each of them is split into 48 slices which work as mini-slits. There are two modes of observation (see Fig. 7.6): a Wide Field Mode (WFM) with a field of view of 1 × 1 arcmin, and a Narrow Field Mode (NFM) with a field of view of 7.5 × 7.5 arcsec. In my work, I have used spectral data taken exclusively in the WFM. This wide field mode has a spatial sampling of 0.2 × 0.2 arcsec and the spectral element is 0.25 Å / pxl. Thus, in the whole field of view, 90000 spectra are sampled, one for each pixel of 0.2 × 0.2 arcsec.

81 Figure 7.6: Two operational modes of MUSE: the Wide Field Mode and the Narrow Field Mode. Table taken from the ESO-MUSE web portal (https://www.eso.org/sci/facilities/develop/instruments/muse.html).

The wavelength coverage ranges from 4800 to 9300 Å, though the extended mode ranges from 4600 to 9350 Å. The spatial resolution combined with the spectral information provide a 3D view of the spectral characteristics of the observed target. The average image quality of the data is about 0.6 arcsec of FWHM. The spectral resolution also depends on the wavelength, and is about 100 km s−1at the wavelength of the Hα line (at z = 0). Since 2017, adaptive optics with 4 laser guide stars has been implemented in order to considerably improve the spatial resolution of the WFM and NFM. The narrow field mode is at work since 2017. Each spaxel is 0.025 × 0.025 arcmin. The spectroscopic data used for this work was taken between 2014 and 2016 with the MUSE instrument (Bacon et al. 2010) in the wide field mode. The extended ionised halo of Haro 11 was covered by a 2x2 mosaic of 4 point- ings, whose adjacent borders overlap by 30". Owing to the mosaic design, the integrated exposure time varies from 6h 13min 20s in the central 0.5x0.5 arcsec, to 1h 33min and 20s in the corners. Standard data reduction procedures (Weilbacher et al. 2012) were applied with a small improvement of the sky subtraction, which was performed manually by masking the emitting line regions. The final data was then corrected by the underlying stellar absorption using the Python package pPXF (Cappellari and Copin 2003) with the E-Miles

82 stellar spectra library (Vazdekis et al. 2010). Stellar absorption primarily in the atmospheres of A-Type stars predominantly aﬀects the Balmer line.

7.3 The warm ionised ISM of Haro 11 - Results and summary of papers I-V

7.3.1 Paper I. The impact of stellar feedback from velocity-dependent ionised gas maps – a MUSE view of Haro 11. For the first paper, we produced a series of velocity-dependent maps in order to study both the internal architecture of Haro 11 and the degree of ionisation of the ISM. Additionally, we acquired insight into the areas that are affected by fast shocks. In order to do so, we extracted spectra of the lines Hα,Hβ, [OIII]λ5007 and [OI]λ6300. We then re-sampled each line cube in bins of 50 km s−1 which were then extracted in line maps. Each diagnostic was extracted in a velocity range from −400 km s−1 to 350 km s−1. For diagnostics involving emission lines separated in wavelength (the case of [OIII]λ5007/Hα), we uni- fied the spatial resolutions by convolving the maps with better resolution with a Gaussian kernel. This kernel was sampled from the resolution difference between the better resolution maps and the line with the worst resolution. In this way, all line maps of a diagnostic have the same spatial resolution. Finally, emission line maps of the Hα and [OIII]λ5007 lines, as well as the diagnostic line ratios [OIII]λ5007/Hα and [OI]λ6300/Hα were used for the analysis of the ionised gas of Haro 11. In this work, the architecture of the ionised gas was analysed by means of the Hα emission line, which traces the ionised gas structure. We used the [OIII]λ5007/Hα ratio to trace the level of ionisation to further distances in the halo. Finally, we used the [OI]λ6300/Hα ratio with the [OIII]λ5007/Hα ratio to trace the locations where fast shocks operate. In addition, estimates of the ionised gas mass that will escape the galaxy’s gravitational potential were calculated from the Hα and [OIII]λ5007 line strengths. Fig. 7.7 shows a 3D reconstruction of the extracted Hα maps. In the Hα maps, we discovered an arrangement of structures that change with velocity. Most of these structures became visible only in the velocity sliced maps. The halo is nearly circular in all velocity bins. At the sensitivity of our data, the size of the halo in the radial direction is about 7 times larger than the central starburst. Fig. 7.8 shows two perspectives of the 3D ionzation structure of the galaxy. These maps show only the very low and very high ionzation regions in blue ([OIII]λ5007/Hα≤0.33) and yellow ([OIII]λ5007/Hα≤1.7) respectively. The panel on the left shows the integrated ionisation map over the whole velocity

83 Figure 7.7: 3D view of the velocity sliced Hα emission maps. The units of the E-W (east-west) and S-N (south-north) side are in pixels (1 pxl = 0.2 arcsec). (Image produced by the author) range, while the panel on the right shows the ionisation values integrated along the east-west direction. Both maps show that the warm ionised gas in Haro 11 is predominantly highly ionised in the southwestern hemisphere, especially in a velocity range of − 300 to 350 km s−1 . On the other side, the integrated ionisation structure in the northeastern hemisphere appears to be assembled in rings of gas with slightly different ionisation. For instance, the first 3.5 kpc in radius from the centre do not have extreme ionisation values. This region seems to be covered by a thin shell of lowly ionised matter which coincides with the arc structure. This first shell is then covered by another thick, highly ionised shell with a thickness of ∼5 kpc. The gas outside of the (apparent) second shell is mainly lowly ionised towards the northeast, and is otherwise mixed. It is not clear whether these differently ionised shells exist, or if they are due to projection effects. Both 3D views images and the discussion of the ionised shells are not included in Paper 1, but they are included in my Licentiate thesis.

84 The left panel of Fig. 7.8 shows the locations of the knots. Knot C is in general lowly ionised, while knot A is highly ionised. Knot B does not have an extreme ionisation value and is shown as a red circle. On the top of knot B, there is a lowly ionised zone coincident with a redshifted outﬂow, which is traced up to 1000 km s−1 in the [OIII]λ5007 maps. The ionisation maps also trace two star forming regions, which have low ionisation, in the west- northwest (at v∼ − 400 to − 350 km s−1) and south (at v= − 300 km s−1).

Figure 7.8: 3D view of the ionised gas structure of Haro 11. In blue and yellow are shown the regions with [OIII]λ5007/Hα ratios ≤0.33 and ≥1.7 respectively. Low ionised regions (blue) are found in the northeast part of the galaxy while the highly ionised gas is found in the southeast part of the galaxy. (Image produced by the author)

The ionised gas structure appears to be shaped mainly by feedback processes, while there are only few structures linked to the merger dynamics. The tidal tails traced in the halo at central velocities are most likely shaped by tidal interaction of the merger. Additionally, a mini dusty arm that folds from knot A to the north, seems to be a tidal arm of the gas rich progenitor. In some parts of these tidal arms, gas starts to condense in compact clumps, likely forming tidal dwarf galaxies or star formation clumps in the halo. One of these shows emission in many lines, suggesting that star formation is taking place there. Others are only visible in the Hα emission and it is not clear if they are locally ionised, or if the ionising photons come from the central knots of the galaxy. The structures of stellar feedback origin that are seen in the Hα and particularly in the [OIII]λ5007/Hα maps are: some kpc-scale highly-ionised ﬁlaments (up to 9 kpc in radius) towards the southern halo, an arc at a distance of 3.5 kpc from the central starburst covering knot B and C, a circular lowly ionised

85 shell covering a highly ionised cavity around knot C, two highly ionised out- ﬂows emerging from knot C, small arcs oriented in a circular fashion around the central starburst towards the west and ﬁnally three highly ionised cones.

Insight into a kpc-scale superbubble The semi arc structure covering the starburst region and the highly ionised filamentary structure in the south strongly suggest the presence of a kpc-scale superbubble that might be powered by intense stellar feedback from the most massive stars in the central star forming knots. The shell seems to be open in the south, which at the moment of blow-out may have developed superwinds that dragged the shell and surrounding dense gas outwards, thus forming the filaments. Between the filaments, the diffuse and highly ionised gas suggests the presence of rarefied gas that is completely ionised. This low-column density gas may favour the escape of photons and matter to the IGM. Addition- ally to this main superbubble, we found evidence of a second smaller shell surrounding knot C. This lowly ionised shell surrounding the highly ionised cavity around knot C suggests that the massive stellar population of knot C has developed its own superbubble. Fig. 7.9 shows the architecture of Haro 11 as interpreted from the Hα and ionisation maps. There is a supershell (r∼ 3.8 kpc. Radius estimated in Pa- per II) centred on the galaxy centre and open in the south. The hot interior gas is flowing out in the south, and part of the dragged shell has formed filaments. There is a smaller shell (r∼ 1.7 kpc) surrounding knot C that seems to be fragmented, but with narrow gaps between fragmented parts. The sketch also shows the main star forming knots A, B and C as well as the dusty arm.

Insight into the mechanism that favours the escape of LyC radiation in Haro 11 Our analysis suggests that both a density bound scenario (given that the southwestern halo is highly ionised), and low-density channels could be the mechanisms that allow LyC to escape.

Evidence of ionising gas escaping the galaxy To calculate the escape velocity, we assumed a dark matter fraction and an inclination with respect to the line of sight for the gas ﬂowing out. We derived the luminosities from the Hα line for gas moving at velocities between 250 and 400 km s−1. At higher velocities, the underlying absorption is quite strong and was not corrected for properly, therefore we used the [OIII]λ5007 intensity (v> 400 km s−1) to derive the luminosity of the gas moving at higher velocities.

86 Figure 7.9: Sketch of the Haro 11 ionised gas components from Menacho et al. (2019). The central knots A, B and C, and the dusty arm are the engine of the galaxy. They are surrounded by a superbubble (r∼3.8 kpc) that is opened in the south. Knot C has developed a local r∼1.7 kpc superbubble. The shell of the largest bubble is most likely fully ionised, while the shell around knot C contains gas with lower ionisation values.

87 The mass that will escape the galaxy ranges from 6% for a dark matter fraction of 90% , to ∼35% for a dark matter fraction of 60%. It is most likely that the true value is between these two extremes. Given the intense feedback in Haro 11, the ionised gas is dominated by expansion.

7.3.2 Paper II. Ionised gas properties of the extreme starburst galaxy Haro 11. Temperature and metal abundance discrepancies.

In this paper, we present an analysis of the kinematics and the ionised gas properties of the ISM in Haro 11. For the latter, we show the integrated properties of the gas as well as the properties of the gas at blueshifted, central and redshifted velocities. We follow the same as procedure described in Paper I to extract maps in velocity bins. The kinematics were extracted from a single Gaussian fit and aimed to trace the position of both galaxy progenitors, as well as to map the regions with broad components in the galaxy. The velocity map shows complex motions dominated by both the merger dynamics (northeastern hemisphere), and the effect of strong feedback (southwestern hemisphere). Knots A and B are likely the core of the gas rich progenitor, while knot C seems to be the core of the more evolved progenitor. The velocity dispersion clearly shows the impact of frequent supernovae in knot C on inducing turbulence not only in its surroundings, but also in parts of the halo. Two structures traced in Paper I are distinguishable in the kinematics. The western highly ionised cone has slightly blueshifted velocities and a slightly higher velocity dispersion in comparison with the surrounding gas. Thus, this cone seems to be a galactic-scale outflow that is shocking the halo gas on its way out. We additionally found that the redshifted outflow close to knot B has one of the highest velocity dispersions of the galaxy. We also derived the kinematics of the [NII]λ6583 and [OIII]λ5007 lines tracing the nitrogen enriched gas and the highly ionised gas respectively. The most noticeable feature is found when comparing the [NII]λ6583 and Hα velocities. We found that the most blueshifted nitrogen enriched gas draws a semi-arc structure that fits well with the bright arc traced in the Hα maps. Thus, it is not clear if this structure is the prolongation of the supershell covering the knots whose blowing away create the filamentary structure towards the south. The E(B-V) map, which traces the dust distribution in the galaxy, shows that the dust is mostly concentrated in knots B and C, while the halo is almost dust-free. Around knot B, we uncovered what seems to be a dusty shell structure of few hundred pc in radius. Knot A has average properties for dust,

88 density and temperature. Knot B seems to have dense, colder gas, while knot C has hot rarefied gas. The temperature in the halo is somewhat inhomogeneous. We measured high temperatures (∼15000 K) in the lowly ionised zone towards the east-northeast of knot C. In this zone, we traced fast shocks with the data from Paper I. In the rest of the halo, the temperature varies between 8500 and 12000 K. We calculated the abundances following two methods: a direct method and a strong line method. In the direct method, the oxygen abundances have a strong dependence on the temperature. In the high temperature zones, we derived low metallicities, and the opposite for the low-temperature zones. In general, we found a low metallicity in knot C, high metallicity in knot B. In the halo we derive very low metallicities in the shocked zone, where we derived extremely high-temperatures. The high metallicities were derived in the low-temperature zones, which are the highly ionised filamentary structure and ionised cones. There we derived temperatures slightly lower than 10000 K. The abundances derived from the strong method are somewhat different. The metal-rich zones are concentrated in the centre, especially around knots B and C. The halo shows in general low metallicities. In both methods, we found that the blueshifted gas is in general more metal-enriched than the gas at central or redshifted velocities. Thus, we most likely traced the high-velocity, metal- enriched gas expelled by supernova explosions. In the nitrogen-abundance maps, we traced two nitrogen enriched regions that are coincident with young massive clusters with Wolf-Rayet signatures. We present a map showing the metallicity discrepancy between the estimations of both methods. Knot C and the lowly-ionised zone affected by shocks have the largest discrepancies. In the latter, shocks are likely affecting the auroral lines, which leads to overestimate the temperature and consequently, underestimate of abundances as well. In knot C, it is unclear which mechanism is affecting the gas. Large discrepancies are also found in the highly ionised zone. Here, the strong line method underestimate the metallicities.

7.3.3 Paper III. Deciphering the excitation mechanisms in Haro 11. A new perspective from the MUSE data.

In this paper we present a spatially resolved analysis of the excitation mechanisms of the galaxy with integrated line diagnostics and diagnostics of the gas at blueshifted, central and redshifted velocities. We follow the same procedure described in Paper I and II to extract maps in velocity bins. Here study the so called "BPT-diagrams" in velocity bins. We aim to study the excitation mechanisms: shocks or photoionisation in the many structures that were previously identiﬁed in speciﬁc velocity bins. Given that the ionis-

89 ing photons are released almost exclusively from the central knots and that the SN-induced shock waves propagate from the centre outwards, we can study these mechanisms as function of radius. Thus, we plot each diagnostic (e.g. [OIII]λ5007/Hβ and [OI]λ6300/Hα) relatively to the distance from the centre. We found some pattern in the diagnostic maps that seems to speciﬁc structures such as the highly ionised zones or the Hα bright arc (which is the knot’s C shell remnant or the large shell remnant covering the whole starburst region). We can trace in our diagrams the gas of these structures evolving from photoionised-dominated to shock-dominated with radius.

7.3.4 Paper IV.The large ionised halo of Haro 11 – a density bound nebula

In this paper, we present an analysis of the ionised gas conditions of the galaxy in radial maps. The radial maps were divided in two parts, one part tracing the gas rich progenitor, and the other used to trace the more evolved progenitor. Using the radial information of the [OIII]/Hα and [SII]/[OIII] maps as a tracer of the ionisation structure of the galaxy, we found that the halo is highly ionised up to distances where we can trace the gas with suﬃcient signal-to-noise (sn le 5, r∼16 kpc). This suggest that the halo is, at least to some extent, density bound, which allows the escape of ionising radiation. The total ionised gas 9 mass was calculated from the Hα luminosity and is about 10 M . Using a simple model that is characterised by the fraction of neutral gas observed, we estimated whether or not the ionising photons can escape the ISM of Haro 11. We found that the halo is to a great extent transparent to ionising photons, which allows these highly absorbed photons to escape. Using a simple calculation we estimated that about half of the LyC photons measured could have escaped by means of the density-bound halo.

7.3.5 Paper V. The source of leaking ionising photons from Haro 11 – Clues from HST/COS spectroscopy of knots A, B and C

This paper presents a UV-optical analysis of the neutral gas in the knots A, B and C, aiming to uncover the star forming knot where LyC is leaking. UV spectra were obtained from the HST-COS instrument while optical spectra were taken from the MUSE instrument. We also analysed silicon absorption lines that trace the lowly (Si II) and highly (Si IV) ionised gas. We found that knot C has a covering fraction of neutral gas of about 50%, while in knots A and B the covering fraction is almost 100%. This indicates that knot C is not completely opaque to LyC radiation, while knots A and B are. Another indirect tracer of the escape of LyC is the peak separation of the

90 Lyα emission. A narrow peak indicates a high probability of LyC escape. Knot C does not have a blueshifted peak, although there is a feature in absorption that hints to the presence of a blueshifted peak. Knots A and B have a peak separation of approximately 400 km s−1 , which is not suﬃciently narrow to imply an escape of LyC radiation. A third indirect method to test the regions where LyC escapes involves the ionisation level of the gas. A highly ionised gas favours the escape of LyC radiation. Although the covering fraction of neutral gas favours knot C as the LyC leaker, this knot is lowly ionised compared to knot A. Nevertheless, the ﬁrst method is the strongest indirect method to determine if LyC can escape from a region or not. Therefore, we conclude that knot C is likely the knot where LyC is escaping.

91 92 8. Summary and outlook

The work presented in this thesis takes advantage of the 3D capabilities of the MUSE instrument to study in detail the 3D ionised gas structure of Haro 11. The success of my work is owed to the high sensitivity and great spatial resolution of our data, allowing us to detect for the ﬁrst time the faint halo gas and other structures in the central starburst region. Previous works used long slit spectra or IFU with worse spatial sampling than our MUSE data, and focused either on the integrated properties of the galaxy or only on the three star forming knots A, B and C in the central starburst. Their most important result is: Haro 11 is a Lyman alpha and Lyman continuum emitter with a vast population of young and massive star clusters (Adamo et al. 2010; Bergvall et al. 2006; Hayes et al. 2007). Indeed, the current starburst started about 40 Myrs ago and peaked about 3.5 Myrs ago (Adamo et al. 2010). Haro 11 is currently an extreme emitter of ionising photons as inferred from the amount of ionised gas that is about three times that of the neutral gas mass (Pardy et al. 2016, and Paper I). The starburst region was found to have complex kinematics from spectral data with high spectral resolution (Östlin et al. 2015). The ionised gas conditions in the knots are diverse, and measured or derived values vary from author to author, but in general Haro 11 was found to be a metal-poor galaxy (Guseva et al. 2012; Hayes et al. 2007; James et al. 2013). In this thesis we present results from both the average properties of the galaxy and the knots, and from diverse discrete gas structures found in the galaxy that were analysed from a 3D perspective. To exploit and present the 3D information, we sliced up the galaxy into a sequence of maps in velocity bins. We used the velocity unit as a tracer of galaxy depth. These are the main contributions of my work to the general understanding of Haro 11:

• We uncovered several gas structures assembling the halo. Some of these were created by merger dynamics such as tidal tails, (star forming) gas clumps and a region where most of the lowly ionised gas was com- pressed by virial motions. Others structures originated from the impact of stellar feedback, such as the ﬁlamentary structure in the southern part of the galaxy (Papers I to III).

93 • We were able to trace the multi-scale impact of strong feedback at high spatial resolution. Most of the feedback signatures are imprinted in the ionisation mapping such as highly ionised cones at galactic scales, shell remnants, galactic-scale gas fountains and paths of highly ionised out- ﬂowing gas. Others, such as dust-shells, were traced in the dust attenuation map (Papers I and II).

• We ﬁnd the ISM of Haro 11 to be strongly aﬀected by radiative and momentum feedback from its massive stellar population. The analysis of the ionisation structure in the halo of Haro 11 suggests a partially density-bound halo towards the south. This makes this part of the halo a possible source of ionisation photon leakage from Haro 11 (Papers I and IV).

• We located several outﬂows originated either from winds of Wolf-Rayet stars or from massive star clusters. In addition, we have traced paths of highly ionised outﬂowing gas; one of which seems to transport and distribute metal-enriched gas to larger distances in the halo (Paper II).

• We found evidence of two fragmented kpc-scale superbubbles whose breakout may have strong consequences for the galaxy. The hot vented gas seems to have dragged matter along, forming the ﬁlamentary complex towards the south while clearing low-density paths where gas and metals likely escape the galaxy (Paper I and II).

• Although three star forming knots (A, B and C) are the main engines of the galaxy, only one (knot C) dominates the energy released in supernovae. This nuclear region is clearly inducing important turbulence, and shocks not only in its closest surroundings, but also farther out in the halo (Papers II and III).

• We ﬁnd knots A and B to be covered by neutral and ionised gas, while knot C has a considerably lower fraction of neutral gas along the line of sight. This makes knot C the most likely (main) source of the detected LyC radiation. We also detected dusty gas inﬂows on the front and back side of knot B. (Papers II and V)

• We ﬁnd knot C to have extreme conditions: It is a soft X-ray emitter, and the brightest FUV and Lyα source of the galaxy (Grimes et al. 2007; Hayes et al. 2007). It has frequent supernova events that propagate shocks unhindered in its extremely low density medium. Notwithstand- ing, it has molecular gas (Hayes in prep.) and dust shielded in dense clouds. Thus, a large range of energy transport processes might take

94 place in this knot; this seems to aﬀect the strength of the lines and therefore we were unable to get conclusive results of its physical properties. (Papers II - V)

• We derive extremely high temperatures in knot C when using the [NII]- temperature sensitive line, a factor 30 higher than the temperatures derived using [OI] or [SIII]-temperature sensitive lines. We suggest that at extreme conditions, there is at least one mechanism that can strongly enhance the strength of the [NII]-auroral lines only.

• We find slightly higher metal abundances for Haro 11 than previously reported. We also find large discrepancies in the metallicities derived from two different methods in places with very high ionisation, the region affected by shocks and knot C. While in the shocked area these discrepancies can be attributed to the effect of shocks in overestimating the temperatures, we do not find a straightforward explanation for such differing results in the other regions. (Papers II and III). (Paper II)

• We ﬁnd that a considerable amount of ionized gas might escape the galaxy. (Paper I)

In a general context, the work presented in this thesis highlights:

• The strong impact of stellar feedback (radiative, ionising and momentum) on the ISM of extreme starbursting galaxies. We illustrate this by means of ionised gas structures such as superbubbles, shells, arcs, as well as the ionisation structure of the galaxy such as outﬂows, galactic ionised cones and a partially density-bound halo.

• Extreme starburst galaxies may tend to develop kpc-scale superbubbles in their starburst regions. The breakout of these bubbles can develop powerful galactic winds able to drill holes in the ISM of galaxies. If these winds overcome the galaxy’s gravitational potential, LyC photons may escape the galaxy through the low-density holes. This could be the main mechanism behind the escape of LyC photons in most LyC leaking galaxies.

• We present evidence that stellar feedback is able to launch powerful out- ﬂows allowing the escape of a signiﬁcant fraction of gas and metals to the IGM.

• We find that traditional relations developed from average measurements or simplified models fail to probe the condition of the gas in extreme environments (such as knot C). Such a condition seems to affect the

95 strength of the emission lines. This result draws attention to the need of revisiting standard relations by including more realistic models where several physical processes are simultaneously at work.

• The method used in my work has proven to be eﬀective in exploiting 3D data, and can be applied to explore the high sensitivity, high spatial- resolution data from future facilities such as JWST or ELT.

8.1 Future Work

One possibility is to explore the ionised halo of Haro 11 from the recombination lines as well as from the metal lines. In this way, we can study the amount of ionising gas in the diﬀuse, yet bright halo of Haro 11, as well as the amount present in the central star forming HII regions. Another possibility is to exploit high spatial resolution AO IFU observations to look at the gas in the centre of the galaxy. There are still open questions about the physical condition of the gas in the knots, especially in knot C. It would be interesting to apply the technique of velocity-dependent maps to explore the ISM of other galaxies, especially the ones that are most aﬀected by stellar or AGN-feedback. This technique can also be applied to high spatial resolution observations of high-z galaxies with JWST and ELT.

96 Sammanfattning

Blå kompakta galaxer (Blue Compact Galaxies, eller BCGs) är kompakta, me- tallfattiga, aktivt stjärnbildande galaxer som i många avseenden liknar unga, starkt rödförskjutna galaxer. BCGs är särskilt aktiva producenter av massiva stjärnhopar, som var och en innehåller tusentals massiva stjärnor. Under sina korta liv injicerar massiva stjärnor kontinuerligt energi, värme och rörelse- moment till det interstellära mediet via sin intensiva strålning, sina stjärnvin- dar, och i slutändan via supernovaexplosioner. Sådana återkopplingsmekanis- mer påverkar stjärnans direkta omgivning, men när återkopplingen härstammar från en koncentration av massiva stjärnhopar, kan det även starkt påverka hela galaxens gastillstånd. Den här avhandlingen presenterar en detaljerad analys av den joniserade gasens tillstånd och effekten av stark återkoppling i Haro 11, som är en ex- tremt kraftigt stjärnbildande BCG och den närmaste Lyman-kontinuum (LyC)- läckande galaxen. Vi drar nytta av MUSE-instrumentets kapacitet att spektro- fotometriskt dela upp inkommande ljus i en serie bilder sorterade utefter has- tighet, och därmed ge information om galaxen i tre dimensioner. Haro 11 har ett rikt utbud av massiva och företrädelsevis unga stjärnhopar, som är kon- centrerade i tre kompakta knytpunkter inom dess 4 x 4 kpc2 stora centrum. Vi finner att den lokala återkopplingen från stjärnor har en stark inverkan på globala rörelsemönster och gasens tillstånd över långa avsånd inom galaxens halo. Många kpc-skaliga strukturer såsom filament, skal och bubblor kan spå- ras i våra data. Därutöver tycks den starka återkopplingen ha bildat kpc-skaliga bubblor, utflöden och galaktiska joniserade koner med drastiska konsekvenser för sannolikt flykt av Lyα- och LyC-fotoner, samt gas och metaller, ut ur galaxen. Den utsträckta halon runt Haro 11 regleras av fotojonisering och/eller chockverkan från återkommande supernovor som härstammar från de centra- la stjärnbildande regionerna. På grund av de extrema förhållandena i galaxens interstellära medium, resulterar en standarddiagnostik av emissionslinjerna ibland i stora avvikelser för den joniserade gasens egenskaper. Resultaten som presenteras i detta arbete belyser: a) den starka inverkan som återkoppling från stjärnor har på det interstellära mediet över alla skalor i stjärnbildande system; b) det faktum att traditionella relationer som härletts från genomsnittliga mätningar av emissionslinjer eller från förenklade modeller inte går att använda för att studera gastillståndet i extrema miljöer. Detta understryker behovet av att revidera standardrelationerna genom att inkludera mer realistiska modeller där flera olika fysikaliska mekanismer samverkar; c) att metoden som applicerats här kan användas för att i detalj studera mycket känsliga och högupplösta data från framtida teleskop som JWST och ELT. Acknowledgements

It was a long journey to here, where I can finally say: ’keep calm, I a doctor!’. On this long way I met wonderful people, some of them helped me to achieve my projects, some of them were collaborators, many of them were my friends and the rest my family, which is a mix of everyone. I want to start with my supervisors Göran and Arjan. Göran introduce me to Haro 11 and gave me the amazing MUSE data. Although I thought that there was already enough written about Haro 11 and I should jump to other projects, I became immediately attracted by the details I saw in the data, and by the many structures that Haro 11 displays when playing with the data cube. Göran was very patient and supportive all the times when I had trouble in my work. Thank you for teaching me to look at the big picture of my work, for your time taken to explain to me many things related to galaxies or Haro 11, for supporting my travels to conferences and observations, and for supporting my work even after my regular time. Arjan helped me alot, especially when I knew nothing. Thank you Arjan for always having time to clarify my doubts, to answer my questions and to solve my problems. Thank you for introducing me to many python scripts/packages that made my work easier, and a big thank for correcting my presentations and proposals very carefully. To the members of the galaxy group that were like my family in these years: Göran, Arjan, Matt, Angela, Jens, Armin, Alexandra, Lorenza, Mattia, Axel, Timmy, Töger, Rocio (of course) and the ones that are not anymore here: Sinead, Matteo, Christian, Michael and Katie. Thank you for the good time with nice conversation, for the glasses of wines, whiskies, schnaps, liquors and beers (non-galaxies might think we were drinking all the time. Do you know what? YES! and ’salud!’), for the cakes and cookies (that was the second favourite activity after drinking, of course!), for the food, for the races ran together, for the travels, for the pub days, for the bowling days, for the beer days, for the hiking days. I am so happy to have had the opportunity to share an amazing time with you. I will always remember our adventures together. Apart from that, thank you for sharing your knowledge in our group meetings, especially thank you Angela and Matt for sharing your interesting ideas and way of thinking with us. To my friends, the Italian mafia: Christina, David and Giovanni (I still find mozzarella boring, because "burrata" is the only mozzarella I believe in!). To Matteo, Sinead, Armin, Mattia, Alexandra, Illa, Emanuel, Anders and others that I probably forgot, thank you for the great moments in and outside Al- banova. Thank you for inviting me many times to your home, for cooking, for playing and for all the great times together. You are awesome! Except Armin when playing, then you are quite normal! To Rocio, thank you very much for your support, your help with adminis- trative/medical stuff. You were an angel that always had open arms to me. Un millon de gracias Rocio!! A Markus, my mentor, that has been on my side all this time. A big thank you to Quentin for helping me with the English grammar part. And Giovanni, thank you for supporting me all this time. I know, it was hard! You stood at my side and helped me rise like a phoenix out of the ashes. I thank my family in Chochilandia. I dedicate this thesis to my parents that were always supporting me in this journey. Gracias papa por trabajar duro, por ensenarnos a ser responsables y persistentes, por tu esfuerzo y dedicacion en el campo, por ensenarme a vivir al son del sol y la lluvia. Gracias mama por trabajar duro, por trabajar mucho y en silencio y con poca recompensa, por ensenarnos valores y principios, pero por sobre todo por ensenarnos a volar, a sentir que una frontera no es un limite, que hay que lanzarse a la aventura, que hay que sonar y lanzarse, que la carga se acomoda en el camino. A Moni y Pastor que creyeron en mi y tambien me apoyaron. Letztendlich, Ich danke meiner Familie in Österreich fur die schöne Zeit zusammen, für die Jobst Fam- ilienwandertäge und für eure Gastfreunschaft. Ein ganz besonderer Dank geht an Franz, meinen Fahrlehrer. References

D. C. Abbott, J. H. Bieging, and E. Churchwell. Mass loss from very luminous ob stars and the cygnus superbubble. ApJ, 250:645–659, November 1981. 67

A. Adamo, G. Östlin, E. Zackrisson, M. Hayes, R. J. Cumming, and G. Micheva. Super star clusters in Haro 11: properties of a very young starburst and evidence for a near-infrared ﬂux excess. MNRAS, 407:870–890, September 2010. doi: 10.1111/j.1365-2966.2010.16983.x. 23, 71, 74, 75, 76, 77, 79, 93

A. Adamo, G. Östlin, and E. Zackrisson. Probing cluster formation under extreme conditions: massive star clusters in blue compact galaxies. MNRAS, 417:1904–1912, November 2011. doi: 10.1111/j.1365-2966. 2011.19377.x. 23

O. Agertz, A. V. Kravtsov, S. N. Leitner, and N. Y. Gnedin. Toward a complete accounting of energy and momentum from stellar feedback in galaxy formation simulations. ApJ, 770:25, June 2013. 55, 58

M. G. Allen, B. A. Groves, M. A. Dopita, R. S. Sutherland, and L. J. Kewley. The mappings iii library of fast radiative shock models. ApJS, 178:20–55, September 2008. 36

R. Amorín, A. Fontana, E. Pérez-Montero, M. Castellano, L. Guaita, A. Grazian, O. Le Fèvre, B. Ribeiro, D. Schaerer, L. A. M. Tasca, R. Thomas, S. Bardelli, L. Cassarà, P. Cassata, A. Cimatti, T. Con- tini, S. de Barros, B. Garilli, M. Giavalisco, N. Hathi, A. Koekemoer, V. Le Brun, B. C. Lemaux, D. Maccagni, L. Pentericci, J. Pforr, M. Talia, L. Tresse, E. Vanzella, D. Vergani, G. Zamorani, E. Zucca, and E. Merlin. Analogues of primeval galaxies two billion years after the big bang. Nature Astronomy, 1:0052, March 2017. 18

R. O. Amorín, J. M. Vílchez, and E. Pérez-Montero. Unveiling the nature of the “green pea” galaxies. Astrophysics and Space Science Proceedings, 28:243, 2012. 18

Martin Asplund, Nicolas Grevesse, A. Jacques Sauval, and Pat Scott. The Chemical Composition of the Sun. ARA&A, 47(1):481–522, September 2009. doi: 10.1146/annurev.astro.46.060407.145222. 41

H. Atek, J. Richard, M. Jauzac, J.-P. Kneib, P. Natarajan, M. Limousin, D. Schaerer, E. Jullo, H. Ebeling, E. Egami, and B. Clement. Are ultra-faint galaxies at z = 6-8 responsible for cosmic reionization? combined constraints from the hubble frontier ﬁelds clusters and parallels. ApJ, 814:69, November 2015a. 17, 18

H. Atek, J. Richard, J.-P. Kneib, M. Jauzac, D. Schaerer, B. Clement, M. Limousin, E. Jullo, P. Natarajan, E. Egami, and H. Ebeling. New constraints on the faint end of the uv luminosity function at z ˜ 7-8 using the gravitational lensing of the hubble frontier ﬁelds cluster a2744. ApJ, 800:18, February 2015b. 17, 18

R. Bacon, M. Accardo, L. Adjali, H. Anwand, S. Bauer, I. Biswas, J. Blaizot, D. Boudon, S. Brau-Nogue, J. Brinchmann, P. Caillier, L. Capoani, C. M. Carollo, T. Contini, P. Couderc, E. Daguisé, S. Deiries, B. Delabre, S. Dreizler, J. Dubois, M. Dupieux, C. Dupuy, E. Emsellem, T. Fechner, A. Fleischmann, M. François, G. Gallou, T. Gharsa, A. Glindemann, D. Gojak, B. Guiderdoni, G. Hansali, T. Hahn, A. Jarno, A. Kelz, C. Koehler, J. Kosmalski, F. Laurent, M. Le Floch, S. J. Lilly, J.-L. Lizon, M. Loupias, A. Manescau, C. Monstein, H. Nicklas, J.-C. Olaya, L. Pares, L. Pasquini, A. Pécontal-Rousset, R. Pelló, C. Petit, E. Popow, R. Reiss, A. Remillieux, E. Renault, M. Roth, G. Rupprecht, D. Serre, J. Schaye, G. Soucail, M. Steinmetz, O. Streicher, R. Stuik, H. Valentin, J. Vernet, P. Weilbacher, L. Wisotzki, and N. Yerle. The muse second-generation vlt instrument. In Ground-based and Airborne Instrumentation for Astronomy III, volume 7735 of Proceedings of the SPIE, page 773508, July 2010. 82

N. Bastian, B. Ercolano, and M. Gieles. Hierarchical star formation in m33: properties of the star-forming regions. Ap&SS, 324:293–297, December 2009. 21, 24

N. Bergvall, E. Zackrisson, B. G. Andersson, D. Arnberg, J. Masegosa, and G. Östlin. First detection of Lyman continuum escape from a local starburst galaxy. I. Observations of the luminous blue compact galaxy Haro 11 with the Far Ultraviolet Spectroscopic Explorer (FUSE). A&A, 448(2):513–524, March 2006. doi: 10.1051/0004-6361:20053788. 18, 73, 93

A. Bik, G. Östlin, V. Menacho, A. Adamo, M. Hayes, E. C. Herenz, and J. Melinder. Super star cluster feedback driving ionization, shocks and outﬂows in the halo of the nearby starburst ESO 338-IG04. A&A, 619:A131, November 2018. doi: 10.1051/0004-6361/201833916. 32, 34, 35, 36

L. Binette, R. Matadamas, G. F. Hägele, D. C. Nicholls, G. Magris C., M. Á. Peña-Guerrero, C. Morisset, and A. Rodríguez-González. Discrepancies between the [O iii] and [S iii] temperatures in H ii regions. A&A, 547:A29, November 2012. doi: 10.1051/0004-6361/201219515. 39

J. Bland and B. Tully. Large-scale bipolar wind in m82. Nature, 334:43–45, July 1988. 71

S. Borthakur, T. M. Heckman, C. Leitherer, and R. A. Overzier. A local clue to the reionization of the universe. Science, 346:216–219, October 2014. 18

Frédéric Bournaud, Damien Chapon, Romain Teyssier, Leila C. Powell, Bruce G. Elmegreen, Debra Meloy Elmegreen, Pierre-Alain Duc, Thierry Contini, Benoit Epinat, and Kristen L. Shapiro. Hydrodynamics of High-redshift Galaxy Collisions: From Gas-rich Disks to Dispersion-dominated Mergers and Com- pact Spheroids. ApJ, 730(1):4, March 2011. doi: 10.1088/0004-637X/730/1/4. 32

R. J. Bouwens, G. D. Illingworth, P. A. Oesch, M. Stiavelli, P. van Dokkum, M. Trenti, D. Magee, I. Labbé, M. Franx, C. M. Carollo, and V. Gonzalez. Discovery of z ˜ 8 galaxies in the hubble ultra deep ﬁeld from ultra-deep wfc3/ir observations. ApJ, 709:L133–L137, February 2010a. 17, 19

R. J. Bouwens, G. D. Illingworth, P. A. Oesch, M. Trenti, M. Stiavelli, C. M. Carollo, M. Franx, P. G. van Dokkum, I. Labbé, and D. Magee. Very blue uv-continuum slope β of low luminosity z ˜ 7 galaxies from wfc3/ir: Evidence for extremely low metallicities? ApJ, 708:L69–L73, January 2010b. 17

R. J. Bouwens, G. D. Illingworth, P. A. Oesch, M. Trenti, I. Labbé, L. Bradley, M. Carollo, P. G. van Dokkum, V. Gonzalez, B. Holwerda, M. Franx, L. Spitler, R. Smit, and D. Magee. Uv luminosity functions at redshifts z∼4 to z∼10: 10,000 galaxies from hst legacy ﬁelds. ApJ, 803:34, April 2015. 17

P. W. J. L. Brand and W. J. Zealey. Cloud structure in the galactic plane - A cosmic bubble bath. A&A, 38: 363–371, February 1975. 60

E. Braunsfurth and J. V. Feitzinger. Young stars and bubbles in the Large Megellanic Cloud. A&A, 127: 113–131, October 1983. 62

E. Brinks and E. Bajaja. A high resolution hydrogen-line survey of Messier 31. III - H I holes in the interstellar medium. A&A, 169:14–42, November 1986. 62

Daniela Calzetti, Lee Armus, Ralph C. Bohlin, Anne L. Kinney, Jan Koornneef, and Thaisa Storchi- Bergmann. The Dust Content and Opacity of Actively Star-forming Galaxies. ApJ, 533(2):682–695, April 2000. doi: 10.1086/308692. 38

Michele Cappellari and Yannick Copin. Adaptive spatial binning of integral-ﬁeld spectroscopic data using Voronoi tessellations. Mon.Not.Roy.Astron.Soc., 342:345, February 2003. 82 C. Cardamone, K. Schawinski, M. Sarzi, S. P. Bamford, N. Bennert, C. M. Urry, C. Lintott, W. C. Keel, J. Parejko, R. C. Nichol, D. Thomas, D. Andreescu, P. Murray, M. J. Raddick, A. Slosar, A. Szalay, and J. Vandenberg. Galaxy zoo green peas: discovery of a class of compact extremely star-forming galaxies. MNRAS, 399:1191–1205, November 2009. 18

Jason A. Cardelli, Geoﬀrey C. Clayton, and John S. Mathis. The Relationship between Infrared, Optical, and Ultraviolet Extinction. ApJ, 345:245, October 1989. doi: 10.1086/167900. 38

G. Cecil, J. Bland-Hawthorn, S. Veilleux, and A. V. Filippenko. Jet- and wind-driven ionized outﬂows in the superbubble and star-forming disk of ngc 3079. ApJ, 555:338–355, July 2001. 63, 71, 72

J. Chisholm, C. A. Tremonti, C. Leitherer, Y. Chen, A. Woﬀord, and B. Lundgren. Scaling relations between warm galactic outﬂows and their host galaxies. ApJ, 811:149, October 2015. 18, 71

E. Churchwell, M. S. Povich, D. Allen, M. G. Taylor, M. R. Meade, B. L. Babler, R. Indebetouw, C. Watson, B. A. Whitney, M. G. Wolﬁre, T. M. Bania, R. A. Benjamin, D. P. Clemens, M. Cohen, C. J. Cyganowski, J. M. Jackson, H. A. Kobulnicky, J. S. Mathis, E. P. Mercer, S. R. Stolovy, B. Uzpen, D. F. Watson, and M. J. Wolﬀ. The bubbling galactic disk. ApJ, 649:759–778, October 2006. 60

E. Churchwell, D. F. Watson, M. S. Povich, M. G. Taylor, B. L. Babler, M. R. Meade, R. A. Benjamin, R. Indebetouw, and B. A. Whitney. The bubbling galactic disk. ii. the inner 20 ◦. ApJ, 670:428–441, November 2007. 60

D. Cormier, V. Lebouteiller, S. C. Madden, N. Abel, S. Hony, F. Galliano, M. Baes, M. J. Barlow, A. Cooray, I. De Looze, M. Galametz, O. Ł. Karczewski, T. J. Parkin, A. Rémy, M. Sauvage, L. Spinoglio, C. D. Wilson, and R. Wu. The nature of the interstellar medium of the starburst low-metallicity galaxy haro 11: a multi-phase model of the infrared emission. A&A, 548:A20, December 2012. 19, 73

Paul A. Crowther, Olivier Schnurr, Raphael Hirschi, Norhasliza Yusof, Richard J. Parker, Simon P. Good- win, and Hasan Abu Kassim. The R136 star cluster hosts several stars whose individual masses greatly exceed the accepted 150Msolar stellar mass limit. MNRAS, 408(2):731–751, October 2010. doi: 10.1111/j.1365-2966.2010.17167.x. 24

R. Davies, M. Ward, and H. Sugai. Shock excitation in interacting galaxies: Markarian 266. ApJ, 535: 735–747, June 2000. 36

S. de Barros, E. Vanzella, R. Amorín, M. Castellano, B. Siana, A. Grazian, H. Suh, I. Balestra, C. Vignali, A. Verhamme, G. Zamorani, M. Mignoli, G. Hasinger, A. Comastri, L. Pentericci, E. Pérez-Montero, A. Fontana, M. Giavalisco, and R. Gilli. An extreme [o iii] emitter at z = 3.2: a low metallicity lyman continuum source. A&A, 585:A51, January 2016. 18

W. J. G. de Blok and F. Walter. Evidence for tidal interaction and a supergiant h i shell in the local group dwarf galaxy ngc 6822. ApJ, 537:L95–L98, July 2000. 62, 63

L. Deharveng, F. Schuller, L. D. Anderson, A. Zavagno, F. Wyrowski, K. M. Menten, L. Bronfman, L. Testi, C. M. Walmsley, and M. Wienen. A gallery of bubbles. the nature of the bubbles observed by spitzer and what atlasgal tells us about the surrounding neutral material. A&A, 523:A6, November 2010. 59, 60

M. A. Dopita, D. S. Mathewson, and V. L. Ford. Shapley constellation iii - a region of self-propagating star formation. ApJ, 297:599–606, October 1985. 62

Michael A. Dopita, Ralph S. Sutherland, David C. Nicholls, Lisa J. Kewley, and Frédéric P. A. Vogt. New Strong-line Abundance Diagnostics for H II Regions: Eﬀects of κ-distributed Electron Energies and New Atomic Data. ApJS, 208(1):10, Sep 2013. doi: 10.1088/0067-0049/208/1/10. 49

E. I. Doran, P. A. Crowther, A. de Koter, C. J. Evans, C. McEvoy, N. R. Walborn, N. Bastian, J. M. Bestenlehner, G. Gräfener, A. Herrero, K. Köhler, J. Maíz Apellániz, F. Najarro, J. Puls, H. Sana, F. R. N. Schneider, W. D. Taylor, J. Th. van Loon, and J. S. Vink. The VLT-FLAMES Tarantula Survey. XI. A census of the hot luminous stars and their feedback in 30 Doradus. A&A, 558:A134, Oct 2013. doi: 10.1051/0004-6361/201321824. 24 Bruce Draine. Physics of the interstellar and intergalactic medium. Princeton university press, 2011. 25, 37

B. G. Elmegreen. Gas phase processes aﬀecting galactic evolution. Ap&SS, 284:819–832, April 2003. doi: 10.1023/A:1024054129337. 17

B. G. Elmegreen, D. M. Elmegreen, D. R. Vollbach, E. R. Foster, and T. E. Ferguson. On the Origin of Exponential Disks at High Redshift. ApJ, 634:101–108, November 2005. doi: 10.1086/496952. 24

C. Esteban. Are Temperature Fluctuations Out There? In William J. Henney, Jose Franco, and Marco Martos, editors, Revista Mexicana de Astronomia y Astroﬁsica Conference Series, volume 12 of Revista Mexicana de Astronomia y Astroﬁsica Conference Series, pages 56–61, Feb 2002. 48

X. Fan, M. A. Strauss, R. H. Becker, R. L. White, J. E. Gunn, G. R. Knapp, G. T. Richards, D. P. Schneider, J. Brinkmann, and M. Fukugita. Constraining the evolution of the ionizing background and the epoch of reionization with z˜6 quasars. ii. a sample of 19 quasars. AJ, 132:117–136, July 2006. 17

M. N. Fanelli, R. W. O’Connell, and T. X. Thuan. Spectral synthesis in the ultraviolet. ii - stellar populations and star formation in blue compact galaxies. ApJ, 334:665–687, November 1988. 19

G. J. Ferland, W. J. Henney, C. R. O’Dell, and M. Peimbert. The abundance discrepancy factor and t2ˆ in nebulae: are non-thermal electrons the culprits? RMxAA, 52:261, October 2016. 49

Edward L. Fitzpatrick. Correcting for the Eﬀects of Interstellar Extinction. PASP, 111(755):63–75, January 1999. doi: 10.1086/316293. 38

N. M. Förster Schreiber, R. Genzel, M. D. Lehnert, N. Bouché, A. Verma, D. K. Erb, A. E. Shapley, C. C. Steidel, R. Davies, D. Lutz, N. Nesvadba, L. J. Tacconi, F. Eisenhauer, R. Abuter, A. Gilbert, S. Gillessen, and A. Sternberg. SINFONI Integral Field Spectroscopy of z ˜ 2 UV-selected Galaxies: Rotation Curves and Dynamical Evolution. ApJ, 645:1062–1075, July 2006. doi: 10.1086/504403. 31

N. M. Förster Schreiber, R. Genzel, N. Bouché, G. Cresci, R. Davies, P. Buschkamp, K. Shapiro, L. J. Tacconi, E. K. S. Hicks, S. Genel, A. E. Shapley, D. K. Erb, C. C. Steidel, D. Lutz, F. Eisenhauer, S. Gillessen, A. Sternberg, A. Renzini, A. Cimatti, E. Daddi, J. Kurk, S. Lilly, X. Kong, M. D. Lehnert, N. Nesvadba, A. Verma, H. McCracken, N. Arimoto, M. Mignoli, and M. Onodera. The SINS Survey: SINFONI Integral Field Spectroscopy of z ˜ 2 Star-forming Galaxies. ApJ, 706:1364–1428, December 2009. doi: 10.1088/0004-637X/706/2/1364. 31

Frédéric Galliano, Maud Galametz, and Anthony P. Jones. The Interstellar Dust Properties of Nearby Galaxies. ARA&A, 56:673–713, September 2018. doi: 10.1146/annurev-astro-081817-051900. 37

Donald R. Garnett. Electron Temperature Variations and the Measurement of Nebular Abundances. AJ, 103:1330, April 1992. doi: 10.1086/116146. 39, 40, 45

R. Genzel, S. Newman, T. Jones, N. M. Förster Schreiber, K. Shapiro, S. Genel, S. J. Lilly, A. Renzini, L. J. Tacconi, N. Bouché, A. Burkert, G. Cresci, P. Buschkamp, C. M. Carollo, D. Ceverino, R. Davies, A. Dekel, F. Eisenhauer, E. Hicks, J. Kurk, D. Lutz, C. Mancini, T. Naab, Y. Peng, A. Sternberg, D. Vergani, and G. Zamorani. The Sins Survey of z ~2 Galaxy Kinematics: Properties of the Giant Star-forming Clumps. ApJ, 733(2):101, June 2011. doi: 10.1088/0004-637X/733/2/101. 31

R. Genzel, N. M. F. Schreiber, H. Übler, P. Lang, T. Naab, R. Bender, L. J. Tacconi, E. Wisnioski, S. Wuyts, T. Alexander, A. Beiﬁori, S. Belli, G. Brammer, A. Burkert, C. M. Carollo, J. Chan, R. Davies, M. Fos- sati, A. Galametz, S. Genel, O. Gerhard, D. Lutz, J. T. Mendel, I. Momcheva, E. J. Nelson, A. Renzini, R. Saglia, A. Sternberg, S. Tacchella, K. Tadaki, and D. Wilman. Strongly baryon-dominated disk galaxies at the peak of galaxy formation ten billion years ago. Nature, 543:397–401, March 2017. 17

J. P. Grimes, T. Heckman, D. Strickland, W. V. Dixon, K. Sembach, R. Overzier, C. Hoopes, A. Aloisi, and A. Ptak. Feedback in the local lyman-break galaxy analog haro 11 as probed by far-ultraviolet and x-ray observations. ApJ, 668:891–905, October 2007. 80, 94 N. G. Guseva, Y. I. Izotov, and T. X. Thuan. A spectroscopic study of a large sample of wolf-rayet galaxies. ApJ, 531:776–803, March 2000. 19

N. G. Guseva, Y. I. Izotov, K. J. Fricke, and C. Henkel. Vlt/x-shooter observations of blue compact galaxies haro 11 and eso 338-ig 004. A&A, 541:A115, May 2012. 93

F. Haardt and P. Madau. Radiative transfer in a clumpy universe. iv. new synthesis models of the cosmic uv/x-ray background. ApJ, 746:125, February 2012. 17, 18

M. Hayes, G. Östlin, H. Atek, D. Kunth, J. M. Mas-Hesse, C. Leitherer, E. Jiménez-Bailón, and A. Adamo. The escape of Lyman photons from a young starburst: the case of Haro11. MNRAS, 382:1465–1480, December 2007. 71, 73, 80, 93, 94

T. M. Heckman, S. Borthakur, R. Overzier, G. Kauﬀmann, A. Basu-Zych, C. Leitherer, K. Sembach, D. C. Martin, R. M. Rich, D. Schiminovich, and M. Seibert. Extreme feedback and the epoch of reionization: Clues in the local universe. ApJ, 730:5, March 2011. 18

T. M. Heckman, R. M. Alexandroﬀ, S. Borthakur, R. Overzier, and C. Leitherer. The systematic properties of the warm phase of starburst-driven galactic winds. ApJ, 809:147, August 2015. 26, 71

C. Heiles. H i shells, supershells, shell-like objects, and ’worms’. ApJS, 55:585–595, August 1984. 59, 60

H. Hirashita and L. K. Hunt. The role of dust in “active” and “passive” low-metallicity star formation. A&A, 421:555–570, July 2004. 19

Alec S. Hirschauer, John J. Salzer, Evan D. Skillman, Danielle Berg, Kristen B. W. McQuinn, John M. Cannon, Alex J. R. Gordon, Martha P. Haynes, Riccardo Giovanelli, Elizabeth A. K. Adams, Steven Janowiecki, Katherine L. Rhode, Richard W. Pogge, Kevin V. Croxall, and Erik Aver. ALFALFA Discovery of the Most Metal-poor Gas-rich Galaxy Known: AGC 198691. ApJ, 822(2):108, May 2016. doi: 10.3847/0004-637X/822/2/108. 19

K. Hollyhead, N. Bastian, A. Adamo, E. Silva-Villa, J. Dale, J. E. Ryon, and Z. Gazak. Studying the YMC population of M83: how long clusters remain embedded, their interaction with the ISM and implications for GC formation theories. MNRAS, 449:1106–1117, May 2015. doi: 10.1093/mnras/stv331. 23

E. M. Hu. High latitude h i shells in the galaxy. i. ApJ, 248:119–127, August 1981. 60

D. A. Hunter, W. N. Hawley, and J. S. Gallagher, III. A survey for extra-h ii region ionized gas structures in irregular galaxies. AJ, 106:1797–1811, November 1993. 59, 63, 67

Y. I. Izotov, G. Stasinska,´ G. Meynet, N. G. Guseva, and T. X. Thuan. The chemical composition of metal-poor emission-line galaxies in the Data Release 3 of the Sloan Digital Sky Survey. A&A, 448(3): 955–970, Mar 2006. doi: 10.1051/0004-6361:20053763. 45

Y. I. Izotov, N. G. Guseva, and T. X. Thuan. Green pea galaxies and cohorts: Luminous compact emission- line galaxies in the sloan digital sky survey. ApJ, 728:161, February 2011. 18

Y. I. Izotov, I. Orlitová, D. Schaerer, T. X. Thuan, A. Verhamme, N. G. Guseva, and G. Worseck. Eight per cent leakage of lyman continuum photons from a compact, star-forming dwarf galaxy. Nature, 529: 178–180, January 2016a. 18

Y. I. Izotov, D. Schaerer, T. X. Thuan, G. Worseck, N. G. Guseva, I. Orlitová, and A. Verhamme. Detection of high lyman continuum leakage from four low-redshift compact star-forming galaxies. MNRAS, 461: 3683–3701, October 2016b. 18, 39

Y. I. Izotov, T. X. Thuan, N. G. Guseva, and S. E. Liss. J0811+4730: the most metal-poor star-forming dwarf galaxy known. MNRAS, 473(2):1956–1966, January 2018. doi: 10.1093/mnras/stx2478. 19

Y.I. Izotov, G. Worseck, D. Schaerer, N. G. Guseva, T. X. Thuan, V.Fricke, A., and I. Orlitová. Low-redshift lyman continuum leaking galaxies with high [o iii]/[o ii] ratios. MNRAS, 478:4851–4865, August 2018. 18, 34 Y. I. Izotov, T. X. Thuan, and N. G. Guseva. J1234+3901: an extremely metal-deﬁcient compact star- forming dwarf galaxy at redshift 0.133. MNRAS, 483(4):5491–5498, March 2019. doi: 10.1093/mnras/ sty3472. 19

B. L. James, Y. G. Tsamis, J. R. Walsh, M. J. Barlow, and M. S. Westmoquette. The Lyman break analogue Haro 11: spatially resolved chemodynamics with VLT FLAMES. MNRAS, 430(3):2097–2112, April 2013. doi: 10.1093/mnras/stt034. 80, 93

A. E. Jaskot and M. S. Oey. The origin and optical depth of ionizing radiation in the “green pea” galaxies. ApJ, 766:91, April 2013. 18

Anne E. Jaskot, Tara Dowd, M. S. Oey, Claudia Scarlata, and Jed McKinney. New Insights on Lyα and Lyman Continuum Radiative Transfer in the Greenest Peas. ApJ, 885(1):96, Nov 2019. doi: 10.3847/ 1538-4357/ab3d3b. 18, 34

J. Kamphuis, R. Sancisi, and T. van der Hulst. An H I superbubble in the spiral galaxy M 101. A&A, 244: L29–L32, April 1991. 62, 63, 67

Amanda I. Karakas and John C. Lattanzio. The Dawes Review 2: Nucleosynthesis and Stellar Yields of Low- and Intermediate-Mass Single Stars. PASA, 31:e030, July 2014. doi: 10.1017/pasa.2014.21. 42

R. P. Keenan, M. S. Oey, A. E. Jaskot, and B. L. James. Haro 11: Where is the Lyman Continuum Source? ApJ, 848:12, October 2017. 34

S. Kendrew, H. Beuther, R. Simpson, T. Csengeri, M. Wienen, C. J. Lintott, M. S. Povich, C. Beaumont, and F. Schuller. The milky way project and atlasgal: The distribution and physical properties of cold clumps near infrared bubbles. ApJ, 825:142, July 2016. 60

C.-G. Kim and E. C. Ostriker. Momentum injection by supernovae in the interstellar medium. ApJ, 802:99, April 2015. 68, 69

C.-G. Kim and E. C. Ostriker. Numerical simulations of multiphase winds and fountains from star-forming galactic disks. i. solar neighborhood tigress model. ApJ, 853:173, February 2018. 53, 68, 72

C.-G. Kim, E. C. Ostriker, and R. Raileanu. Superbubbles in the multiphase ism and the loading of galactic winds. ApJ, 834:25, January 2017. 63, 68, 69

X. Kong, S. Charlot, A. Weiss, and F. Z. Cheng. Spectroscopic study of blue compact galaxies. iii. empirical population synthesis. A&A, 403:877–887, June 2003. 19

M. R. Krumholz and C. F. McKee. A General Theory of Turbulence-regulated Star Formation, from Spirals to Ultraluminous Infrared Galaxies. ApJ, 630:250–268, September 2005. doi: 10.1086/431734. 23

D. Kunth and G. Östlin. The most metal-poor galaxies. A&A Rev., 10:1–79, 2000. 18, 19

D. Kunth and W. L. W. Sargent. I zw 18 and the existence of very metal poor blue compact dwarf galaxies. ApJ, 300:496–499, January 1986. 19

C. J. Lada, M. Lombardi, and J. F. Alves. On the star formation rates in molecular clouds. ApJ, 724: 687–693, November 2010. 21

M. D. Lehnert, L. Le Tiran, N. P. H. Nesvadba, W. van Driel, F. Boulanger, and P. Di Matteo. On the self-regulation of intense star-formation in galaxies at z = 1-3. A&A, 555:A72, July 2013. doi: 10.1051/ 0004-6361/201220555. 32

E. Leitet, N. Bergvall, N. Piskunov, and B.-G. Andersson. Analyzing low signal-to-noise fuse spectra. conﬁrmation of lyman continuum escape from haro 11. A&A, 532:A107, August 2011. 73

C. Leitherer and T. M. Heckman. Synthetic properties of starburst galaxies. ApJS, 96:9–38, January 1995. 54 C. Leitherer, D. Schaerer, J. D. Goldader, R. M. G. Delgado, C. Robert, D. F. Kune, D. F. de Mello, D. Devost, and T. M. Heckman. Starburst99: Synthesis models for galaxies with active star formation. ApJS, 123:3–40, July 1999. 52, 54, 55, 56, 57, 77

C. Leitherer, S. Hernandez, J. C. Lee, and M. S. Oey. Direct detection of lyman continuum escape from local starburst galaxies with the cosmic origins spectrograph. ApJ, 823:64, May 2016. 18

C. R. Lynds and A. R. Sandage. Evidence for an explosion in the center of the galaxy m82. ApJ, 137:1005, May 1963. 63, 71

J. A. MacHattie, J. A. Irwin, S. C. Madden, D. Cormier, and A. Rémy-Ruyer. Detection of h i absorption in the dwarf galaxy haro 11. MNRAS, 438:L66–L70, February 2014. 73

W. Maciejewski, E. M. Murphy, F. J. Lockman, and B. D. Savage. The aquila supershell: A remnant of multiple supernovae. ApJ, 469:238, September 1996. 62

P. Madau, M. J. Rees, M. Volonteri, F. Haardt, and S. P. Oh. Early reionization by miniquasars. ApJ, 604: 484–494, April 2004. 17

R. Maiolino and F. Mannucci. De re metallica: the cosmic chemical evolution of galaxies. A&A Rev., 27 (1):3, February 2019. doi: 10.1007/s00159-018-0112-2. 40, 41, 42, 43, 45, 46

F. Marchi, L. Pentericci, L. Guaita, D. Schaerer, A. Verhamme, M. Castellano, B. Ribeiro, B. Garilli, O. L. Fèvre, R. Amorin, S. Bardelli, P. Cassata, A. Durkalec, A. Grazian, N. P. Hathi, B. C. Lemaux, D. Maccagni, E. Vanzella, and E. Zucca. Lyα-lyman continuum connection in 3.5 ≤ z ≤ 4.3 star-forming galaxies from the vuds survey. A&A, 614:A11, June 2018. 18

A. T. Marlowe, T. M. Heckman, R. F. G. Wyse, and R. Schommer. Observations of the impact of starbursts on the interstellar medium in dwarf galaxies. ApJ, 438:563–589, January 1995. 32, 33, 67

C. L. Martin. The impact of star formation on the interstellar medium in dwarf galaxies. ii. the formation of galactic winds. ApJ, 506:222–252, October 1998. 32

C. L. Martin, H. A. Kobulnicky, and T. M. Heckman. The metal content of dwarf starburst winds: Results from chandra observations of ngc 1569. ApJ, 574:663–692, August 2002. 59, 63, 71

N. M. McClure-Griﬃths, J. M. Dickey, B. M. Gaensler, and A. J. Green. The galactic distribution of large h i shells. ApJ, 578:176–193, October 2002. 62

N. M. McClure-Griﬃths, J. M. Dickey, B. M. Gaensler, and A. J. Green. Loops, drips, and walls in the galactic chimney gsh 277+00+36. ApJ, 594:833–843, September 2003. 62

R. McCray and M. Kafatos. Supershells and propagating star formation. ApJ, 317:190–196, June 1987. 67

J. Meaburn, D. L. Terrett, and J. C. Blades. The dynamics of giant ﬁlamentary shells in the large magellanic cloud. iii - n59a /dem 241/. MNRAS, 197:19–30, October 1981. 62

V. Menacho, G. Östlin, A. Bik, L. Della Bruna, J. Melinder, A. Adamo, M. Hayes, E. C. Herenz, and N. Bergvall. The impact of stellar feedback from velocity-dependent ionized gas maps - a MUSE view of Haro 11. MNRAS, 487(3):3183–3198, August 2019. doi: 10.1093/mnras/stz1414. 33, 71, 73, 74, 80, 87

M. Messa, A. Adamo, D. Calzetti, M. Reina-Campos, D. Colombo, E. Schinnerer, R. Chandar, D. A. Dale, D. A. Gouliermis, K. Grasha, E. K. Grebel, B. G. Elmegreen, M. Fumagalli, K. E. Johnson, J. M. D. Kruijssen, G. Östlin, F. Shabani, L. J. Smith, and B. C. Whitmore. The young star cluster population of m51 with legus - ii. testing environmental dependences. MNRAS, 477:1683–1707, June 2018. 23, 24

M. Morris and E. Serabyn. The galactic center environment. ARA&A, 34:645–702, 1996. 71 N. Murray, B. Ménard, and T. A. Thompson. Radiation pressure from massive star clusters as a launching mechanism for super-galactic winds. ApJ, 735:66, July 2011. 58, 59

K. Nakajima and M. Ouchi. Ionization state of inter-stellar medium in galaxies: evolution, sfr-m∗-z dependence, and ionizing photon escape. MNRAS, 442:900–916, July 2014. 18

B. B. Nath and J. Silk. Starburst-driven galactic outﬂows. MNRAS, 396:L90–L94, June 2009. 59

S. F. Newman, K. Shapiro Griﬃn, R. Genzel, R. Davies, N. M. Förster-Schreiber, L. J. Tacconi, J. Kurk, S. Wuyts, S. Genel, S. J. Lilly, A. Renzini, N. Bouché, A. Burkert, G. Cresci, P. Buschkamp, C. M. Carollo, F. Eisenhauer, E. Hicks, D. Lutz, C. Mancini, T. Naab, Y. Peng, and D. Vergani. Shocked superwinds from the z ˜ 2 clumpy star-forming galaxy, zc406690. ApJ, 752:111, June 2012. 36

David C. Nicholls, Michael A. Dopita, and Ralph S. Sutherland. Resolving the Electron Temperature Discrepancies in H II Regions and Planetary Nebulae: κ-distributed Electrons. ApJ, 752(2):148, June 2012. doi: 10.1088/0004-637X/752/2/148. 49

P. A. Oesch, G. Brammer, P. G. van Dokkum, G. D. Illingworth, R. J. Bouwens, I. Labbé, M. Franx, I. Momcheva, M. L. N. Ashby, G. G. Fazio, V. Gonzalez, B. Holden, D. Magee, R. E. Skelton, R. Smit, L. R. Spitler, M. Trenti, and S. P. Willner. A Remarkably Luminous Galaxy at z=11.1 Measured with Hubble Space Telescope Grism Spectroscopy. ApJ, 819(2):129, March 2016. doi: 10.3847/0004-637X/ 819/2/129. 17

K. Omukai. Low-metallicity star formation: the characteristic mass and upper mass limit. In L. K. Hunt, S. C. Madden, and R. Schneider, editors, Low-Metallicity Star Formation: From the First Stars to Dwarf Galaxies, volume 255 of IAU Symposium, pages 49–55, December 2008. 17

G. J. Osterbrock, D. E. & Ferland. Astrophysics of gaseous nebulae and active galactic nuclei. University Science Books, 2006, 2006. 25, 38

G. Östlin, P. Amram, N. Bergvall, J. Masegosa, J. Boulesteix, and I. Márquez. Dynamics of blue compact galaxies, as revealed by their hα velocity ﬁelds. ii. mass models and the starburst triggering mechanism. A&A, 374:800–823, August 2001. 18, 73, 77

G. Östlin, T. Marquart, R. J. Cumming, K. Fathi, N. Bergvall, A. Adamo, P. Amram, and M. Hayes. Kinematics of haro 11: The miniature antennae. A&A, 583:A55, November 2015. 73, 74, 77, 80, 93

P. Papaderos, H.-H. Loose, T. X. Thuan, and K. J. Fricke. Optical structure and star formation in blue compact dwarf galaxies. I. Observations and proﬁle decomposition. A&AS, 120:207–228, December 1996. 18

S. A. Pardy, J. M. Cannon, G. Östlin, M. Hayes, and N. Bergvall. Detection of H I in Emission in the LYα Emitting Galaxy Haro 11. AJ, 152:178, December 2016. doi: 10.3847/0004-6256/152/6/178. 73, 80, 93

M. Peimbert, A. Sarmiento, and J. Fierro. The Eﬀect of Shock Waves on the Spectra of H II Regions and Planetary Nebulae. PASP, 103:815, August 1991. doi: 10.1086/132886. 48

Manuel Peimbert, Antonio Peimbert, and Gloria Delgado-Inglada. Nebular Spectroscopy: A Guide on Hii Regions and Planetary Nebulae. PASP, 129(978):082001, August 2017. doi: 10.1088/1538-3873/ aa72c3. 48

E. W. Pellegrini, M. S. Oey, P. F. Winkler, S. D. Points, R. C. Smith, A. E. Jaskot, and J. Zastrow. The optical depth of h ii regions in the magellanic clouds. ApJ, 755:40, August 2012. 34 Ying-jie Peng, Simon J. Lilly, Katarina Kovac,ˇ Micol Bolzonella, Lucia Pozzetti, Alvio Renzini, Gianni Zamorani, Olivier Ilbert, Christian Knobel, Angela Iovino, Christian Maier, Olga Cucciati, Lidia Tasca, C. Marcella Carollo, John Silverman, Pawel Kampczyk, Loic de Ravel, David Sanders, Nicholas Scov- ille, Thierry Contini, Vincenzo Mainieri, Marco Scodeggio, Jean-Paul Kneib, Olivier Le Fèvre, San- dro Bardelli, Angela Bongiorno, Karina Caputi, Graziano Coppa, Sylvain de la Torre, Paolo Franzetti, Bianca Garilli, Fabrice Lamareille, Jean-Francois Le Borgne, Vincent Le Brun, Marco Mignoli, En- rique Perez Montero, Roser Pello, Elena Ricciardelli, Masayuki Tanaka, Laurence Tresse, Daniela Vergani, Niraj Welikala, Elena Zucca, Pascal Oesch, Ummi Abbas, Luke Barnes, Rongmon Bordoloi, Dario Bottini, Alberto Cappi, Paolo Cassata, Andrea Cimatti, Marco Fumana, Gunther Hasinger, Anton Koekemoer, Alexei Leauthaud, Dario Maccagni, Christian Marinoni, Henry McCracken, Pierdomenico Memeo, Baptiste Meneux, Preethi Nair, Cristiano Porciani, Valentina Presotto, and Roberto Scaramella. Mass and Environment as Drivers of Galaxy Evolution in SDSS and zCOSMOS and the Origin of the Schechter Function. ApJ, 721(1):193–221, Sep 2010. doi: 10.1088/0004-637X/721/1/193. 16

Enrique Pérez-Montero and Angeles I. Díaz. Line temperatures and elemental abundances in H II galaxies. MNRAS, 346(1):105–118, November 2003. doi: 10.1046/j.1365-2966.2003.07064.x. 40

Enrique Pérez-Montero and Angeles I. Díaz. A comparative analysis of empirical calibrators for nebular metallicity. MNRAS, 361(3):1063–1076, August 2005. doi: 10.1111/j.1365-2966.2005.09263.x. 46

L. S. Pilyugin and E. K. Grebel. New calibrations for abundance determinations in H II regions. MNRAS, 457(4):3678–3692, Apr 2016. doi: 10.1093/mnras/stw238. 47

R Pogge. Physic of the ism, 2011. Lecture ISM course. Notes from courses of Osterbrock,D and Dinerstain, H. 25

A. H. Prestwich, F. Jackson, P. Kaaret, M. Brorby, T. P. Roberts, S. H. Saar, and M. Yukita. Ultra-luminous x-ray sources in haro ii and the role of x-ray binaries in feedback in lyα emitting galaxies. ApJ, 812: 166, October 2015. 78

M. L. Prevot, J. Lequeux, L. Prevot, E. Maurice, and B. Rocca-Volmerange. The typical interstellar extinction in the Small Magellanic Cloud. A&A, 132:389–392, March 1984. 38

J. Puschnig, M. Hayes, G. Östlin, T. E. Rivera-Thorsen, J. Melinder, J. M. Cannon, V. Menacho, E. Zack- risson, N. Bergvall, and E. Leitet. The lyman continuum escape and ism properties in tololo 1247-232 - new insights from hst and vla. MNRAS, 469:3252–3269, August 2017. 18

M. Radovich, J. Kahanpää, and D. Lemke. Far-infrared mapping of the starburst galaxy ngc 253 with isophot. A&A, 377:73–83, October 2001. 72

R. J. Rand and J. M. Stone. Modeling the hi supershell in the edge-on-galaxy ngc 4631 as an energetic hvc impact. AJ, 111:190, January 1996. 63

T. E. Rivera-Thorsen, G. Östlin, M. Hayes, and J. Puschnig. Neutral ISM, Lyα, and Lyman-continuum in the Nearby Starburst Haro11. ApJ, 837:29, March 2017. doi: 10.3847/1538-4357/aa5d0a. 80

B. E. Robertson, R. S. Ellis, S. R. Furlanetto, and J. S. Dunlop. Cosmic reionization and early star-forming galaxies: A joint analysis of new constraints from planck and the hubble space telescope. ApJ, 802:L19, April 2015. 17

M. Rodrigues, F. Hammer, H. Flores, M. Puech, and E. Athanassoula. Morpho-kinematics of z ∼ 1 galaxies probe the hierarchical scenario. MNRAS, 465(1):1157–1180, February 2017. doi: 10.1093/mnras/ stw2711. 31

D. S. Rupke, S. Veilleux, and D. B. Sanders. Outﬂows in infrared-luminous starbursts at z ≤ 0.5. ii. analysis and discussion. ApJS, 160:115–148, September 2005. 72

Edwin E. Salpeter. The Luminosity Function and Stellar Evolution. ApJ, 121:161, January 1955. doi: 10.1086/145971. 51 Joop Schaye, Robert A. Crain, Richard G. Bower, Michelle Furlong, Matthieu Schaller, Tom Theuns, Claudio Dalla Vecchia, Carlos S. Frenk, I. G. McCarthy, John C. Helly, Adrian Jenkins, Y. M. Rosas- Guevara, Simon D. M. White, Maarten Baes, C. M. Booth, Peter Camps, Julio F. Navarro, Yan Qu, Alireza Rahmati, Till Sawala, Peter A. Thomas, and James Trayford. The EAGLE project: simulating the evolution and assembly of galaxies and their environments. MNRAS, 446(1):521–554, January 2015. doi: 10.1093/mnras/stu2058. 15

R. E. Schulte-Ladbeck, U. Hopp, M. M. Crone, and L. Greggio. A stellar population gradient in vii zw 403: Implications for the formation of blue compact dwarf galaxies. ApJ, 525:709–719, November 1999. 19

L. Searle and W. L. W. Sargent. Inferences from the composition of two dwarf blue galaxies. ApJ, 173:25, April 1972. 19

M. J. Seaton. Interstellar extinction in the UV. MNRAS, 187:73, January 1979. doi: 10.1093/mnras/187.1. 73P. 38

Sivan.J. A new look at the interstellar hydrogen through a very-wide-ﬁeld photographic hα survey of the whole milky way. A&AS, 16:163–172, 1974. 59, 60

Volker Springel, Simon D. M. White, Adrian Jenkins, Carlos S. Frenk, Naoki Yoshida, Liang Gao, Julio Navarro, Robert Thacker, Darren Croton, John Helly, John A. Peacock, Shaun Cole, Peter Thomas, Hugh Couchman, August Evrard, Jörg Colberg, and Frazer Pearce. Simulations of the formation, evolution and clustering of galaxies and quasars. Nature, 435(7042):629–636, June 2005. doi: 10.1038/nature03597. 15

W. A. Stein and B. T. Soifer. Dust in galaxies. ARA&A, 21:177–207, January 1983. doi: 10.1146/annurev. aa.21.090183.001141. 37

L. J. Tacconi, R. Neri, R. Genzel, F. Combes, A. Bolatto, M. C. Cooper, S. Wuyts, F. Bournaud, A. Burkert, J. Comerford, P. Cox, M. Davis, N. M. Förster Schreiber, S. García-Burillo, J. Gracia-Carpio, D. Lutz, T. Naab, S. Newman, A. Omont, A. Saintonge, K. Shapiro Griﬃn, A. Shapley, A. Sternberg, and B. Weiner. Phibss: Molecular Gas Content and Scaling Relations in z ˜ 1-3 Massive, Main-sequence Star-forming Galaxies. ApJ, 768:74, May 2013. doi: 10.1088/0004-637X/768/1/74. 17

R. C. Telfer, W. Zheng, G. A. Kriss, and A. F. Davidsen. The rest-frame extreme-ultraviolet spectral properties of quasi-stellar objects. ApJ, 565:773–785, February 2002. 17

G. Tenorio-Tagle and P. Bodenheimer. Large-scale expanding superstructures in galaxies. ARA&A, 26: 145–197, 1988. 52, 66

G. Tenorio-Tagle and C. Muñoz-Tuñón. The biconical kiloparsec structure generated by nuclear starbursts. MNRAS, 293:299, January 1998. 67

T. X. Thuan. Blue compact dwarf galaxies: Laboratories for probing the primordial universe. In L. K. Hunt, S. C. Madden, and R. Schneider, editors, Low-Metallicity Star Formation: From the First Stars to Dwarf Galaxies, volume 255 of IAU Symposium, pages 348–360, December 2008. 18, 19

T. X. Thuan and Y. I. Izotov. High-ionization emission in metal-deﬁcient blue compact dwarf galaxies. ApJS, 161:240–270, December 2005. 20

T. X. Thuan, Y. I. Izotov, and V. A. Lipovetsky. Hubble space telescope observations of the blue compact dwarf sbs 0335-052: A probable young galaxy. ApJ, 477:661–672, March 1997. 19

T. X. Thuan, A. Lecavelier des Etangs, and Y. I. Izotov. Abundances in the h i envelope of the extremely low metallicity blue compact dwarf galaxy sbs 0335-052 from far ultraviolet spectroscopic explorer observations. ApJ, 621:269–277, March 2005. 19

T. X. Thuan, K. M. Goehring, J. E. Hibbard, Y. I. Izotov, and L. K. Hunt. The h i content of extremely metal-deﬁcient blue compact dwarf galaxies. MNRAS, 463:4268–4286, December 2016. 19 Kohji Tomisaka and Satoru Ikeuchi. Evolution of a superbubble driven by sequential supernova explosions in a plane-stratiﬁed gas distribution. Publications of the Astronomical Society of Japan, 38:697–715, January 1986. 67

S. Torres-Peimbert, M. Peimbert, and E. Daltabuit. IUE and visual observations of the Orion nebula and IC 418 : the carbon abundance. ApJ, 238:133–139, May 1980. doi: 10.1086/157966. 48

E. Vanzella, S. de Barros, K. Vasei, A. Alavi, M. Giavalisco, B. Siana, A. Grazian, G. Hasinger, H. Suh, N. Cappelluti, F. Vito, R. Amorin, I. Balestra, M. Brusa, F. Calura, M. Castellano, A. Comastri, A. Fontana, R. Gilli, M. Mignoli, L. Pentericci, C. Vignali, and G. Zamorani. Hubble imaging of the ionizing radiation from a star-forming galaxy at z=3.2 with fesc≥ 50 %. ApJ, 825:41, July 2016. 18

E. Vanzella, M. Nonino, G. Cupani, M. Castellano, E. Sani, M. Mignoli, F. Calura, M. Meneghetti, R. Gilli, A. Comastri, A. Mercurio, G. B. Caminha, K. Caputi, P. Rosati, C. Grillo, S. Cristiani, I. Balestra, A. Fontana, and M. Giavalisco. Direct lyman continuum and ly α escape observed at redshift 4. MNRAS, 476:L15–L19, May 2018. 18

A. Vazdekis, P. Sánchez-Blázquez, J. Falcón-Barroso, A. J. Cenarro, M. A. Beasley, N. Cardiel, J. Gorgas, and R. F. Peletier. Evolutionary stellar population synthesis with miles - i. the base models and a new line index system. MNRAS, 404:1639–1671, June 2010. 83

C. Veilleux, S. Cecil and j Bland-Hawthorn. Colossal galactic explosions. Scientiﬁc American, (74-79), May 1998. 70, 71

S. Veilleux, G. Cecil, and J. Bland-Hawthorn. Galactic winds. ARA&A, 43:769–826, September 2005. 53, 60, 70, 71, 72

Aditi Vijayan, Chang-Goo Kim, Lucia Armillotta, Eve C. Ostriker, and Miao Li. Kinematics and Dy- namics of Multiphase Outﬂows in Simulations of the Star-Forming Galactic ISM. arXiv e-prints, art. arXiv:1911.07872, Nov 2019. 26

R. Weaver, R. McCray, J. Castor, P. Shapiro, and R. Moore. Interstellar bubbles. II - Structure and evolution. ApJ, 218:377–395, December 1977. doi: 10.1086/155692. 60, 63, 65

P. M. Weilbacher, O. Streicher, T. Urrutia, A. Jarno, A. Pécontal-Rousset, R. Bacon, and P. Böhm. Design and capabilities of the muse data reduction software and pipeline. In Software and Cyberinfrastructure for Astronomy II, volume 8451 of Proceedings of the SPIE, page 84510B, September 2012. 82

Peter M. Weilbacher, Ana Monreal-Ibero, Anne Verhamme, Christer Sandin, Matthias Steinmetz, Wol- fram Kollatschny, Davor Krajnovic,´ Sebastian Kamann, Martin M. Roth, Santiago Erroz-Ferrer, Raf- faella Anna Marino, Michael V. Maseda, Martin Wendt, Roland Bacon, Stefan Dreizler, Johan Richard, and Lutz Wisotzki. Lyman-continuum leakage as dominant source of diﬀuse ionized gas in the Antennae galaxy. A&A, 611:A95, Apr 2018. doi: 10.1051/0004-6361/201731669. 28

C. D. Wilson, N. Scoville, S. C. Madden, and V. Charmandaris. The mass function of supergiant molecular complexes and implications for forming young massive star clusters in the antennae (ngc 4038/4039). ApJ, 599:1049–1066, December 2003. 21

J. H. Wise, T. Abel, M. J. Turk, M. L. Norman, and B. D. Smith. The birth of a galaxy - ii. the role of radiation pressure. MNRAS, 427:311–326, November 2012. 59

E. Wisnioski, N. M. Förster Schreiber, S. Wuyts, E. Wuyts, K. Bandara, D. Wilman, R. Genzel, R. Ben- der, R. Davies, M. Fossati, P. Lang, J. T. Mendel, A. Beiﬁori, G. Brammer, J. Chan, M. Fabricius, Y. Fudamoto, S. Kulkarni, J. Kurk, D. Lutz, E. J. Nelson, I. Momcheva, D. Rosario, R. Saglia, S. Seitz, L. J. Tacconi, and P. G. van Dokkum. The KMOS3D Survey: Design, First Results, and the Evolution of Galaxy Kinematics from 0.7 <= z le 2.7. ApJ, 799(2):209, February 2015. doi: 10.1088/0004-637X/799/2/209. 31 Y. Wu, V. Charmandaris, L. Hao, B. R. Brandl, J. Bernard-Salas, H. W. W. Spoon, and J. R. Houck. Mid- infrared properties of low-metallicity blue compact dwarf galaxies from the spitzer infrared spectrograph. ApJ, 639:157–172, March 2006. 19

Y. Wu, V. Charmandaris, L. K. Hunt, J. Bernard-Salas, B. R. Brandl, J. A. Marshall, V. Lebouteiller, L. Hao, and J. R. Houck. Dust in the extremely metal-poor blue compact dwarf galaxy i zw 18: The spitzer mid- infrared view. ApJ, 662:952–958, June 2007. 19

F. Yusef-Zadeh, F. Melia, and M. Wardle. The galactic center: An interacting system of unusual sources. Science, 287, January 2000. 71

S. Zaroubi. The epoch of reionization. In T. Wiklind, B. Mobasher, and V. Bromm, editors, The First Galaxies, volume 396 of Astrophysics and Space Science Library, page 45, 2013. 16

J. Zastrow, M. S. Oey, S. Veilleux, M. McDonald, and C. L. Martin. An ionization cone in the dwarf starburst galaxy ngc 5253. ApJ, 741:L17, November 2011. 34

J. Zastrow, M. S. Oey, S. Veilleux, and M. McDonald. New constraints on the escape of ionizing photons from starburst galaxies using ionization-parameter mapping. ApJ, 779:76, December 2013. 34 —————————————————–