Study of CO Emission in Nine Hot Dust-Obscured Galaxies at z ∼3

(1)

Study of CO Emission in Nine Hot Dust-Obscured Galaxies at z ∼ 3

Dissertation in partial fulfillment of the requirements for the degree of

Master of Science with a Major in Astronomy and Space Physics

Uppsala University

Department of Physics and Astronomy

[Timothy Faerber]

[May 2021]

(2)

Dissertation in partial fulfillment of the requirements for the degree of

Master of Science with a Major in Astronomy and Space Physics

Uppsala University

Department of Physics and Astronomy

Approved by Dr. Erik Zackrisson

Supervisors, Prof. Kirsten Kraiberg Knudsen Dr. Bitten Gullberg

Dr. Jan Scholtz Dr. Sabine König Examiner, Dr. Andreas Korn

[May 2021]

(3)

II

Abstract

Massive galaxies evolve through different phases including starburst-dominated and

active galactic nuclei (AGN)-dominated phases. These phases are predicted to be

prevalent at earlier times (z ∼ ₂ − 3). In this thesis I present high-sensitivity observations

from the Atacama Large Millimeter/submillimeter Array to investigate mid-J (J

_upper

= 4

and 5) CO emission in nine Wide-field Infrared Survey Explorer-selected hyperluminous,

hot dust-obscured galaxies (Hot DOGs). These sources are thought to represent a

transition phase between starburst- and AGN-dominated galaxies at z ≈ 2.5 − 5. All

nine sources are detected in continuum and line emission. I conclude that the sources

are gas-rich with M

_gas

≈ 10

¹⁰⁻¹¹

M

. Previous far-infrared spectral energy distribution

decomposition revealed that six of the sources have significant cold dust components

suggesting high star-formation rates (SFR ≈ 2000 − 9000 M

yr

⁻¹

). The molecular gas in

the sources is shown to follow roughly the same star-formation trend as a smaller sample

of Hot DOGs and other populations of star-forming and quasar-host galaxies at low- and

high-redshift. The resolved CO emission line data displays large velocity dispersions

(FWHM ≈ 400 − 900 km s

⁻¹

) consistent with other high-z star-forming and quasar-host

galaxies. For a subset of the sources, the line data shows disturbed morphologies and

velocity gradients possibly consistent with rotation or galaxy interaction. The results

from this analysis suggest that the studied sources are heavily dust-obscured quasars

undergoing extreme starburst episodes. The estimated gas and dynamical masses of the

sources are consistent with other populations of massive galaxies at low- and high-z,

indicating that they likely represent a stage in the evolution of massive galaxies.

(4)

Abstrakt / Sammanfattning

Massiva galaxer genomgår flera stadier i sin utveckling, inklusive ett tillstånd domin- erat av exceptionellt hög stjärnbildning (s.k. starburstgalaxer) och ett dominerat av en aktiv galaxkärna (AGN). Dessa tillstånd förväntas vara allmänt förekommande i det unga Universumet (z ∼ 2 − 3). I den här mastersuppsatsen presenterar jag CO observationer av hög känslighet (övergången från J

upper

= 4 och 5) från the Atacama Large Millimeter/submillimeter Array, för nio hyperluminösa, stoftskymda kvasarer (QSOs, eller kvasaren) utvalda från the Wide-field Infrared Survey Explorer. Dessa objekt förmodas representera en övergång vid z ≈ 2.5 − 5 mellan det tillstånd som domineras av extrem stjärnformation och det tillstånd som domineras av den aktiva galaxkärnan.

Alla nio objekt har detekterats i både kontinuum och linjemission. Jag avslöjar att dessa objekt innehåller stora mängder gas, M

gas

≈ 10

¹⁰⁻¹¹

M

. Tidigare dekompositioner av långvågig infraröd spektral energifördelning påvisade att sex stycken av dessa objekt innehåller en signifikant mängd kallt stoft, vilket tyder på hög stjärnformationstakt (SFR ≈ 2000 − 9000M

yr

⁻¹

). Den molekylära gasen i dessa galaxer följer i stora drag samma stjärnbildningstrend som ett mindre urval av Hot DOGs och andra populationer av stjärnbildande galaxer och kvasar-värdgalaxer vid låga och höga rödförskjutningar.

Den högupplösta CO spektraldatan uppvisar stora hastighetsdispersioner (FWHM

≈ 400 − 900 km s

⁻¹

), vilket är i överensstämmelse med andra stjärnbildande och kvasar-

värdgalaxer vid höga rödförskjutningar. För en delmängd av objekten påvisar linjedatan

störd morfologi och hastighetsgradienter potentiellt överensstämmande med rotation

eller galaxinteraktion. Analysen presenterad i detta arbete indikerar att de studerade

objekten är kraftigt stoftskymda kvasaren som genomgår perioder av extrem stjärn-

formation. Objektens uppskattade gas och dynamiska massor är också förenliga med

massorna för andra populationer av massiva galaxer vid låga och höga rödförskjut-

ningar, vilket tyder på att de troligtvis representerar ett stadium i evolutionen av massiva

galaxer.

(5)

IV

Acknowledgements

The completion of this thesis from my home in the United States during the COVID-19 pandemic would not have been possible if not for the amazing support and guidance that I received along the way.

I would first like to thank my incredible team of supervisors, Professor Kirsten Knud- sen, Dr. Bitten Gullberg, Dr. Jan Scholtz and Dr. Sabine König. Their expertise in the field and unwavering patience in assisting me has helped me to take my scientific knowledge and research skills to the next level. I have learned so much about galaxy evolution and the process of reducing raw data into interpretable results in these past six months thanks to my wonderful supervisors. I acknowledge support from the Nordic ALMA Regional Centre (ARC) node based at Onsala Space Observatory. The Nordic ARC node is funded through Swedish Research Council grant No 2017-00648. Additionally, I want to thank Dr. Madeleine Yttergren for translating the abstract of this thesis from English into Swedish. I also want to thank Dr. Erik Zackrisson for being my coordinating supervisor and Dr, Andreas Korn for being my examiner for this thesis.

I also would like to thank my previous supervisors, professors, teachers and classmates for instilling a sense of wonder and curiosity in me. Particularly, I want to thank Professor Sally Oey (the University of Michigan) for being the first scientist to recognize my potential and inspire me to turn my passion for astronomy into a career. Professor Oey was my supervisor for my first research project that I conducted while I was earning my B.S. degree, and has remained a wonderful contact ever since, always offering invaluable advice regarding navigating a career in academia.

I would also like to thank my family and friends who have supported me throughout this year; one that has been difficult for everyone given the global pandemic that we currently find ourselves in. I have truly learned how crucial social interaction and human connection are, especially when spending so much time isolated working on specific scientific research problems. I want to thank my Mother for lending me her unconditional support and regularly asking me for updates about my research. Throughout my thesis and during the entire COVID-19 pandemic, listening to and discovering new music has been a lifeline for me - sharing this with my Father in the evenings has made the long days and nights of research with limited access to the outside world more bearable.

Finally, I want to thank Jerry Garcia and Robert Hunter for the endless jams and thought-provoking lyrics that got me through many long nights of coding and writing.

“There is a road, no simple highway, between the dawn and the dark of night."

- Robert Hunter

(6)

List of Tables

Table 2.1: Introduction of nine Hot DOGs used for analysis. . . . 21 Table 2.2: Calibration . . . . 22 Table 3.1: Reduced χ

²

(χ

²

per-degree-of-freedom) for single and double Gaussian

fits of observed CO emission line (left panel in Figs. 4.1 - 4.9). The form of Gaussian producing the best fit was used to characterize the emission line. The only source in the sample that was better fit with a double Gaussian than with a single Gaussian is W0831+0140. . . . 25 Table 3.2: CO emission line fitting parameters for nine Hot DOGs. . . . 26 Table 3.3: Resolution information for observations and sources in arcsec. . . . 30 Table 3.4: CO emission-derived size estimates in kpc for the six Hot-DOGs from

the sample that could be spatially resolved and fit with 2-dimensional Gaussian profiles (see Table 3.3). . . . 31 Table 3.5: Dust continuum brightness measurements and size estimates for nine

Hot DOGs in the sample. . . . 33 Table 4.1: Mass estimates of nine Hot DOGs. . . . 35 Table 4.2: L

_FIR

estimates for nine Hot DOGs from sample (Fan et al., 2016b).

L

_{FIR, SF}

represents the cold-dust emitting component where sctive star-

formation occurs. These components were decoupled from the total FIR

luminosities of their host galaxies through SED fitting performed by

Fan et al. (2016b). For those sources that have not had their SEDs fitted,

L

_{FIR, SF}

was not available. There are very high uncertainties associated

with these values, as several simplifying assumptions were made. . . . 41

(9)

List of Figures VIII

List of Figures

Figure 1.1: Massive Galaxy Evolution via Merger . . . . 3

Figure 1.2: Star-Formation History in the Universe . . . . 4

Figure 1.3: Merger Fraction of Massive Galaxies vs. Redshift . . . . 6

Figure 1.4: Quasar Luminosity Density vs. Redshift . . . . 7

Figure 1.5: WISE Color Selection Function . . . . 8

Figure 1.6: WISE Color-Color Plot . . . . 9

Figure 1.7: Hot DOG Spectral Energy Distributions (SEDs) . . . . 10

Figure 1.8: CO SLEDs for High-z Galaxies . . . . 12

Figure 2.1: Phase Offset for Baseline Pair . . . . 14

Figure 2.2: Dirty image of source given uv-coverage . . . . 15

Figure 2.3: Atacama Large Millimeter/submillimeter Array . . . . 16

Figure 2.4: Natural vs. Briggs Weighting . . . . 18

Figure 2.5: Moment Map Examples . . . . 20

Figure 3.1: Region used for spectra extraction . . . . 23

Figure 3.2: Emission line spectra of sources . . . . 24

Figure 3.3: FWHM . . . . 27

Figure 3.4: Moment Maps of Sources . . . . 28

Figure 3.5: Continuum Maps of Sources . . . . 32

Figure 4.1: Histogram of Molecular Gas Mass Estimates . . . . 36

Figure 4.2: Histogram of Dynamical Mass Estimates . . . . 37

Figure 4.3: FWHM Histogram . . . . 38

Figure 4.4: L’CO-FWHM . . . . 39

Figure 4.5: L

⁰_CO(1-0)

vs. L

_FIR

. . . . 42

Figure 4.6: W0831+0140 . . . . 44

Figure 4.7: W1322-0328 . . . . 45

Figure 4.8: W2246-0526 . . . . 46

Figure 4.9: W2246-7143 . . . . 47

Figure 4.10: W2305-0039 . . . . 48

(10)

List of Acronyms

AGN

Active Galactic Nuclei

ALMA

Atacama Large Millimeter/sub-millimeter Array

FIR

Far-infrared

Hot DOGs

Hot-Dust Obscured Galaxies

IR

Infrared

QSOs

Quasi-Stellar Objects

SEDs

Spectral Energy Distributions

SLED

Spectral Line Energy Distribution

SMBH

Supermassive Black Hole

WISE

Wide-field Infrared Survey Explorer

(11)

Introduction 1

1 Introduction

1.1 Galaxies in the Universe

We live in an “island universe" (Shapley and Curtis, 1921) known as the Milky Way galaxy. Observations of the gas and stars in the disk of the Milky Way have shown that it has a flat spiral morphology (Kapteyn, 1922) with ∼ 100 − 400 billion stars. In the early 20th century, spectroscopic observations of what were previously thought to be nebulae within the Milky Way revealed that these objects are in fact receding away from us at speeds far greater than the escape velocity of the Milky Way (Slipher, 1915).

This information and the relative brightness of supernovae in the Andromeda galaxy (then referred to as the Andromeda nebula; Curtis, 1917) versus the Milky Way galaxy suggest that other structures like the Milky Way galaxy exist far outside of its potential well.

Studies of extragalactic sources have shown that galaxies come in a wide variety of shapes, masses, colors, and compositions (gas phases, metallicity, etc.; Hubble, 1926).

This range in the observed physical properties of galaxies has lead astronomers to classify

them into three main categories: spiral, elliptical and irregular galaxies (De Vaucouleurs,

1959). Irregular galaxies show a range of messy, irregular (asymmetric) morphologies

and are much less common in the local universe than they are at high-z (Chapman et al.,

2003), as there was much more galaxy interaction in the early universe (Hung et al.,

2015). In general, elliptical galaxies are compact ellipsoids with a wide range in stellar

mass that is, on average, higher than that observed in spiral galaxies (M

_∗_,el

≈ 10

⁹⁻¹³

M

; M

∗,sp

≈ 10

⁹⁻¹²

M

; De Lucia et al., 2006; Rubin, 1983). Elliptical galaxies display

an older (and therefore redder) stellar population than spiral galaxies because of a lack

of ongoing star formation in their interstellar medium (ISM) due to their deficiency of

molecular gas (Springel et al., 2005). The reason that some galaxies in the local universe

appear as “young" spiral galaxies while others appear as “old" elliptical galaxies is

likely some event that occurs early on in their evolution, causing them to deplete their

reservoir of star-forming gas through various physical processes. Constraining the

physical processes that deplete galaxies of their star-forming gas and investigating the

physical events the trigger these processes are challenges faced in trying to understand

the evolution of massive galaxies.

(12)

1.2 Galaxy evolution

In the local universe (z 0.1) we expect to find relatively clear-cut examples of spiral and elliptical galaxies (with the occasional exception; Wen et al., 2009) well-situated in their own potential wells due to limited galaxy interaction and merger activity (Hopkins et al., 2008). Due to the finite speed of light (c ≈ 3 × 10

⁸

m s

⁻¹

), when observing objects at high-z (z ≥ 1) we are looking back in time to study the Universe as it was ∼ 8 − 13 billion years ago. Consequently, galaxies located at different redshifts represent different stages in the evolution of galaxies and display different morphologies. Galaxies in the early universe represent an early stage in galaxy evolution. For this reason they show irregular morphologies, having not settled into their final evolutionary states. Currently, a challenge in extragalactic astronomy is observing sources that may represent transitions between various classifications of galaxies throughout cosmic history.

Studies of galaxies in the local and distant universe have revealed that every massive galaxy with a bulge component contains a Supermassive Black Hole (SMBH) with M

_BH

≈ 10

⁵⁻¹⁰

M

at its center (Kormendy and Richstone, 1995; Harrison, 2017). Gas rapidly accreting onto the central SMBH becomes extremely hot and dense as it is accelerated to relativistic speeds, causing thermal excitation of the gas resulting in a phenonenom known as an Active Galactic Nuclei (AGN). AGN are one of the most energetic astrophysical sources in the Universe (Padovani et al., 2017). When unobscured at optical wavelengths, these objects appear as extremely bright point sources known as Quasi-Stellar Objects (QSOs), or quasars. Understanding how AGN activity affects the star-forming potential of a galaxy is an important step in understanding the evolution of massive galaxies.

Throughout their evolution, massive galaxies take on a variety of shapes and sizes (and therefore classifications) depending on the physical processes dominating their energy output at any given time. A popular model for the evolution of massive galaxies predicts that mergers of gas-rich galaxies produce starburst galaxies, quasars through SMBH accretion, and eventually, red elliptical galaxies with little-to-no active star formation (Hopkins et al., 2006). The gravitational interaction between merging systems exerts tidal torques on their baryonic matter, leading to rapid inflows of gas into the center of the merger remnant (Hernquist, 1989). This results in high gas densities in the galactic nuclei of merger remnants, triggering starburst episodes and fueling rapid SMBH growth (Hopkins et al., 2008). These merger-driven starburst episodes are thought to be responsible for producing some of the properties observed in (ultra)-luminous infrared galaxies ((U)LIRGs) and sub-milliimeter galaxies (SMGs; Hopkins et al., 2006). It is theorized that over time, the large reservoir of cool molecular gas fueling this starburst activity is depleted via SMBH accretion, consumption through starburst episodes, and outflows caused by heating from AGN feedback and supernova-driven winds. This removes the molecular gas from the quasar-host galaxy, suppressing its star-formation and causing it to appear as an optically bright, unobscured quasar (Hopkins et al., 2008).

Due to a deficiency in the gas needed to fuel star formation and SMBH accretion, this

(13)

Introduction 3

quasar is believed to secularly evolve into a red elliptical galaxy with an old stellar population (Hopkins et al., 2008). This process is outlined in Figure 1.1, which was taken directly from Hopkins et al. (2006).

Figure 1.1: Evolution of a massive galaxy from a merger-induced starburst galaxy to an unobscured quasar. (a) Shows an isolated, gas-rich spiral galaxy; (b) shows the halo of the galaxy from (a) after having accreted other neighboring galaxies; (c) shows galaxy interaction and merging of two gas-rich galaxies; (d) shows merger-induced starburst episodes and buried AGN activity in the merger remnant of (C); (e) shows process in which the merger remnant from (c) consumes and expels its star-forming gas due to starbursts shown in (d), SMBH accretion and dust/gas expulsion - this is the stage in massive galaxy evolution that the sources studied in this thesis are thought to represent; (f) shows the optically unobscured quasar that is left over as a result of (e); (g) shows the secular decay of the quasar shown in (f); (h) shows the final stage the merger remnant - a gas-deficient, red elliptical galaxy. This figure is from Hopkins et al. (2006)

Throughout the history of the Universe, cosmic star-formation rate (SFR) density and black hole (BH) accretion density have been observed to reach their peak at z ∼ 2 − 3, suggesting that this is represents an important epoch in galaxy evolution. This noteworthy time in the the evolution of massive galaxies is popularly referred to as

“cosmic noon" (Förster Schreiber and Wuyts, 2020). Figure 1.2 shows the observed star-

(14)

formation rate (SFR) density (in units of M

yr

⁻¹

Mpc

⁻³

) as a function of the age of the Universe in Gyr. X-ray observations of AGN across a redshift range of z ∼ 0 − 8 by Aird et al. (2015) predict that the cosmic BH accretion density and cosmic SFR density should peak at similar redshifts (see Figure 1.2) .

Figure 1.2: Observations of SFR density from Madau and Dickinson (2014). The best fit of these observations is shown with the dashed red line. The solid black line represents the estimated BH accretion rate (Myr⁻¹Mpc⁻³) estimated by Aird et al. (2015) and the shaded grey region represents the 99 per cent confidence interval for the parameters in the model used to estimate accretion rates such as bolometric corrections, radiative efficiency, and systematic errors in observational data. BH accretion density values are scaled up by a factor of 1500. This figure is from Aird et al. (2015).

It has been suggested through observations (Hopkins et al., 2007) and modeling (Hopkins et al., 2006) that the physical mechanism driving this concurrent peak in star formation and BH accretion is the merging of massive, gas-rich galaxies. These mergers funnel large amounts of gas into the central ∼ few pc of merger remnants over timescales significantly shorter than the freefall time ( 10

⁹

years; Hopkins et al., 2008; Hernquist, 1989).

It has been theorized and observed that the merger fraction of galaxies increases with redshift, as the Universe was younger and more turbulent with higher incidences of galaxy interaction (Khochfar and Burkert, 2001). Figure 1.3, adopted from Hopkins et al.

(2008), shows the predicted merger fraction of massive galaxies (M

∗

> 10

¹⁰

M

) as a

function of redshift from z = 0 to z = 6. This redshift-dependency of the cosmic merger

(15)

Introduction 5

fraction of massive galaxies is likely responsible for both SFR and AGN activity reaching their peak in cosmic history at z ∼ 2 − 3 (Gruppioni et al., 2013).

As mergers are thought to play an important role in triggering extreme cases of rapid star-formation and SMBH accretion resulting in starburst galaxies and quasars, it is important to understand the different types of mergers and how they might impact the evolution of their remnants. For the case of this thesis, major mergers (mass ratios of

≤ 3 : 1 so that the interacting systems are both massive with similar masses) between gas-

rich systems are of particular interest, as they are thought to trigger significant starburst

episodes and AGN activity (Hopkins et al., 2008). It is still up for debate whether many

minor mergers could also reproduce the same observed phenomena (Bournaud et al.,

2007). If these mergers occur on timescales significantly shorter than the Hubble time,

gravitational pertubations are thought to tidally strip the gas content of the merging

galaxies (Hopkins et al., 2008). This causes the rapid infall of gas to the central ∼ pc

of the merger remnant resulting in intense central starburst episodes and rapid SMBH

accretion (Hopkins et al., 2008).

(16)

Figure 1.3: Merger fraction of massive galaxies vs. redshift. This figure illustrates that merger fraction is predicted to increase with redshift. The lines of various styles represent estimations using various models and assumptions outlined in Hopkins et al.

(2008).The solid black line represents the default model used in Hopkins et al. (2008) for the abundance of mergers and merging pairs. The dotted line shows a different halo occupation model, and the short dashed line represents a model that uses a different fit for the subhalo mass functions. Group capture/collisional model (solid blue line), angular momentum capture cross sections (long-dashed blue line), and simple dynamical friction considerations (dotted-dashed blue line) fits are also shown. The different points represent observations from various sources in the literature. This figure is from Hopkins et al. (2008).

Given the observational and theoretical evidence for a relationship between mergers and quasars, Hopkins et al. (2008) concluded that every major merger between star- forming, gas-rich

¹

galaxies will trigger a quasar. This build-up of black hole mass resulting in optical quasars has been observed to peak around z ≈ 2 (Hopkins et al., 2008). Figure 1.4, adapted from Hopkins et al. (2007), shows the predicted quasar luminosity density (L

Mpc

⁻³

) in the Universe as a function of z.

1Mergers between gas-poor systems will not provide the material necessary to trigger starburst episodes and AGN activity through the build-up of black hole mass.

(17)

Introduction 7

Figure 1.4: Quasar luminosity density vs. redshift. This figure shows that the luminosity density of quasars is observed to peak at z≈2. The shaded region represents a 1σ

uncertainty. The points represent observations and the solid line shows a fit of these points. The dotted line shows a fit from Richards et al. (2006). This figure is from Hopkins et al. (2007).

Figure 1.4 illustrates that the quasar luminosity density of the Universe is observed (observations are represented by the black points in Figure 1.4) and modeled to peak at z ∼ 2 − 3.

1.2.1 Transition between starburst and quasar

To strengthen the model of galaxy evolution described above, we need to observe sources that represent a transition phase between starburst- and AGN-dominated galaxies.

Recently, a Wide-field Infrared Survey Explorer (WISE)-selected class of hyperluminous

(L

IR

> ₁₀

¹³

_L

) galaxies at z ∼ _2.5 − 4.6 (Tsai et al., 2015) known as Hot-Dust Obscured

Galaxies (Hot DOGs) has been discovered (Eisenhardt et al., 2012). Hot DOGs are gas-

rich (M

gas

≈ 10

¹⁰⁻¹¹

M

; Eisenhardt et al., 2012; Penney et al., 2020; Fan et al., 2018),

suggesting that they are part of the evolutionary sequence of massive galaxies. They

have signifcant hot dust components (T > 60 K) thought to be heated by an obscured

AGN (Assef et al., 2020). Observations have also revealed that Hot DOGs display a

wide range of high SFRs (SFR ∼ 100 − 1000 M

; Tsai et al., 2015). Based on mapping

of their molecular gas that revealed disturbed kinematics, it has been suggested that

Hot DOGs have a high merger fraction ( ∼ 62%; Fan et al., 2016a). Given these previous

observations, Hot DOGs have been suggested to represent this “blowout" (see Figure

(18)

1.1) transition phase where a heavily dust-obscured quasar undergoes intense starburst episodes that are eventually quenched, causing them to appear as highly dust-reddened, IR-luminous quasars (Hopkins et al., 2008).

Hot DOGs are selected using the “W1W2 dropout" method in which sources are clearly detected in the WISE 12 µm (W3) or 22 µm (W4) bands, and faint or undetected in the 3.4

µm

(W1) and 4.6 µm (W2) bands (Wu et al., 2012). Figure 1.5 shows the spectral response functions for each WISE band.

Figure 1.5: Right panel: Weighted mean WISE spectral response functions for W1, W2, W3 and W4 bands. Left Panel: Weighted mean WISE relative spectral response functions normalized to a single peak value on a logarithmic scale. This figure is from Wright et al. (2010).

Figure 1.6 shows the distribution of W1W2-dropouts in WISE color-color space relative

to other WISE-selected sources (Eisenhardt et al., 2012). This figure shows the rarity of

W1W2-dropouts in WISE color–color space.

(19)

Introduction 9

Figure 1.6: W ISE color-color plot. The black points and grey regions represent sources in the 313.4 deg²area of the all-sky release with Galactic latitude b>80 deg and<0.3 mag errors in W1 (3.4µm), W2 (4.6µm), representing about 8% of all WISE catalog sources in this region of the sky. The red points represents W1W2-dropouts selected over the whole sky excluding the Galactic plane and bulge. The blue point and error bars represent the source studied in Eisenhardt et al. (2012). This figure was adapted from Eisenhardt et al. (2012)

This method of selection was used as it is the most effective at finding luminous

dust-obscured galaxies with spectral energy distributions (SEDs) dominated by hot dust

emission at redshifts of z ∼ 1 − 4 (Assef et al., 2015). Infrared (IR) SED analysis of Hot

DOGs has shown that they have strong cold (10 − 100µm flux from star formation) and

hot (1 − 10µm flux due to AGN activity) dust emission components (Fan et al., 2016b,

2017). Figure 1.7 shows the median rest-frame IR SED of 22 Hot DOGs studied in Fan

et al. (2016b).

(20)

Figure 1.7: Median rest-frame IR SED of 22 Hot DOGs analyzed in Fan et al. (2016b). The red line represents the median SED of the sample, while the grey lines represent the individual SEDs of each Hot DOG in the sample. The blue and red dashed lines represent median SEDs of Type-1 and Type-2 QSOs, respectively. The dashed magenta line shows the SED for Arp 220 (a ULIRG from the literature). This figure shows that at λ∼10 µm, the luminosity of Hot DOGs roughly matches what has been observed in Type-1 and Type-2 QSOs. In this frequency range, Hot DOGs are also shown to be far more luminous than typical starburst galaxies (Arp 220 used as representative source) and slightly more luminous than a Type-1 Seyfert galaxy (Mrk 231 used as a representative source). It is theorized that this is due to heating from their strong AGN components, something that traditional starburst galaxies lack (Fan et al., 2016b). This figure was taken from Fan et al. (2016b).

1.3 Cool (molecular) gas in galaxies

In order to understand the evolution of galaxies, we need to investigate the processes involved in converting gas into stars. It is therefore important to constrain the physical properties of the gas that fuels star formation in galaxies (Carilli and Walter, 2013).

Directly preceding star-forming nuclear fusion due to perturbations and gravitational

instabilities in molecular clouds, the molecular gas in the ISM of galaxies cools (T

_kin

∼

10 − 20 K). Thermal excitation of molecular gas through collisions provides information

on the average kinetic temperature and density of the emitting gas (Carilli and Walter,

2013). By tracing the presence and physical parameters of molecular gas in high-z

(21)

Introduction 11

galaxies at different cosmic epochs we can better understand their potential to form stars (Carilli and Walter, 2013).

Molecular hydrogen (H

₂

) is the most abundant species in molecular clouds (Franco and Cox, 1986). However, given its high excitation temperatures , only a very small portion of the molecular gas content in galaxies can be constrained using emission from H

₂

. The high excitation temperatures of H

₂

are a result of its lack of a permanent dipole moment (and therefore large spacing between energy levels). This lack of a permanent dipole moment also results in ultraviolet (UV) dissociation of H

2

. For this reason other species of molecules with lower excitation temperatures and relatively high number densities are used to trace the presence of H

₂

. The most commonly used species for tracing the presence of molecular gas in galaxies is carbon monoxide (CO). CO is the second most abundant molecule after H

₂

in molecular clouds (Liu et al., 2013), has a permanent electric dipole that shields it from UV dissociation (Stahler and Palla, 2008) and is excited to emission at low temperatures ( ∼ 5 K for first transition) due to collisions with H

₂

molecules (Carilli and Walter, 2013). Due to its molecular composition, CO has different levels of excitation at different energies that trace the thermal properties of its environment (Duffendack and Fox, 1927).

CO(1-0) (T

_ex

≈ 5.5 K; n

_crit

≈ 2.1 × 10

³

cm

⁻³

) is the best tracer of CO in high-z galaxies

(Emonts et al., 2014), but since it has such a low excitation temperature, it is not normally

the dominant CO transition line in AGN-dominated galaxies (Narayanan and Krumholz,

2014). For this reason, it is common to observe high-z galaxies in other CO rotational

transitions that are then scaled to estimate the CO(1-0) flux using models based on

observed CO spectral line energy distributions (SLEDs) and radiative transfer modeling

(Carilli and Walter, 2013). Modeling and observations have shown that mid-J CO

transitions such as CO(4-3) trace regions where the molecular gas is slightly hotter

(T

_ex

≈ 55.3 K) and denser (n

_crit

≈ 8.7 × 10

⁴

cm

⁻³

) than when in its ground state (Carilli

and Walter, 2013). High-J CO transitions such as CO(7-6) trace the hotter (T

_ex

≈ 154.9

K), denser (n

crit

≈ 4.5 × 10

⁵

cm

⁻³

) gas in regions of ongoing star-formation, while even

higher-J CO transitions such as CO(9-8) trace the even hotter (T

_ex

≈ 248.9 K), denser

(n

_crit

≈ 8.7 × 10

⁵

cm

⁻³

) gas likely heated by AGN activity (Solomon and Vanden Bout,

2005; Carilli and Walter, 2013). Mid-J CO emission is used to study molecular gas in

high-z galaxies because it is generally brighter than the other CO transitions for rapidly

star-forming galaxies. Figure 1.8 shows the observed CO SLEDs for a sample of SMGs

studied in Narayanan and Krumholz (2014).

(22)

Figure 1.8: CO SLEDs for a sample of SMGs. The blue line represents the Rayleigh-Jeans limit of thermal emission. This figure shows that, on average, starburst galaxies peak in mid-J CO emission. This figure is from Narayanan and Krumholz (2014).

1.4 The goal of this thesis

Hot DOGs have been theorized to represent a transition phase at z ∼ 3 between dusty starburst galaxies and unobscured quasars (Fan et al., 2018; Hopkins et al., 2008), how- ever fairly little is known about their average physical properties due to limited observa- tions.

This thesis aims to characterise the physical properties of Hot DOGs. I focus on the

molecular gas, which is likely the fuel for star formation in the ISM of galaxies. Specifi-

cally, in this thesis I use high-resolution, high-sensitivity interferometric observations of

mid-J CO lines in nine Hot DOGs at z ≈ 3 − 4.6 to characterize the physical properties,

spatial distribution and kinematics of their molecular gas, allowing for further analysis

to establish their hierarchical role in galaxy evolution.

(23)

Methodology 13

2 Methodology

2.1 Radio interferometry

In order to probe the angular scales necessary to resolve the gas kinematics and deter- mine the gas distribution and morphology on ∼ kpc scales in objects located within the redshift range of the sample, very high angular resolutions are needed (. 0.5 arcsec); Fan et al., 2018). The angular resolution (θ) of a telescope is a measure of the finest angular size that can be resolved with that telescope. The angular resolution of a single-dish telescope is directly proportional to the ratio between the observed wavelength (λ

_obs

) to the telescope’s diameter (D):

θ

∝

λ_obs

D . (2.1)

It is obvious from this relationship that to achieve a finer angular resolution for a given wavelength, the telescope diameter must be increased. Due to constraints of modern technology, the largest single-dish telescope in the world is a 500m telescope called the Five-hundred-meter Aperture Spherical Radio Telescope (FAST; Li and Pan, 2016). Upon its completion in 2016, FAST surpassed the Arecibo Observatory as the largest-ever single-dish telescope on Earth. Even with a diameter of 500m, the resolution (2.9 arcsec in the L-band; ν

_L

≈ 1 − 2 GHz) and frequency coverage (70 MHz - 3 GHz;

Gibney, 2019) is not sufficient to probe the angular scales and sky frequencies required to resolve CO emission in sources on ∼ kpc scales at z ∼ 3, which is ∼ 0.5 arcsec (Fan et al., 2018).

The most efficient method used for resolving the emission from gas and dust in the interstellar medium in high-z galaxies is interferometry. By correlating the signals from numerous telescopes, the angular resolution can be increased as such:

θ

∝

λ_obs

B

_max

, (2.2)

where the maximum baseline, B

_max

, is the largest separation between any two tele- scopes in the array.

In order to correlate all of the antennas together to generate one aperture in a practice

referred to as aperture synthesis, the array is split into pairs of antennas, or baselines. Each

baseline measures a single point (known as a visibility) in what is known as the uv-plane,

sometimes referred to as the aperture plane. The distance between the antennas in each

baseline of an array is plotted in the uv-plane, adding to the total uv-coverage of the

(24)

array. To account for the phase difference between the antennas in each baseline pair, a reliable antenna near the center of the array is chosen as the reference antenna, which is set to have zero phase relative to all other antennas. Baseline pairs including this antenna are used to calibrate the phase offset between the antennas due to wave fronts hitting each antenna at a slightly different time (see Figure 2.4). This effect is greater when observing at higher frequencies, as the phase offset will account for more wavefronts per second.

Figure 2.1: Phase offset (φ) when observing a source of at a given elevation (θ) with a baseline pair (a, b) separated by a given distance (d).

Increasing the number of baselines in the array will result in greater uv-coverage and more accurate characterization of the sky emission. During observations the rotation of the Earth is also used to increase the uv-coverage of the array. The more extensive the uv-coverage is, the more accurately it will characterize the sky emission.

Using this uv-coverage, the Fourier transform of the sky brightness is measured in the

uv-plane. If the source observed is a point source, this Fourier transform produces the

dirty beam of the array. For this reason, a theoretical point source is used to determine the

dirty beam of an array. Convolving the measured visibility function with the dirty beam

produces a dirty image of the source. Figure 2.2 shows the uv-coverage (top left panel)

and dirty beam (top right panel) for an example array configuration. The bottom left

panel shows the dirty image bottom right panel shows the sky emission of an example

source.

(25)

Methodology 15

Figure 2.2: Top left panel: uv-coverage for an example interferometer array configuration. Top right panel: Dirty beam of example array configuration. Bottom left panel: Dirty image of example source observed with array. Bottom right panel: sky emission of example source. This figure was adapted from Thiébaut and Giovannelli (2009).

The Atacama Large Millimeter/submillimeter Array (ALMA)

At the present time one of the most advanced interferometers observing at millime-

ter/submillimeter wavelengths is the Atacama Large Millimeter/sub-millimeter Ar-

ray (ALMA), located in the Atacama Desert in Chile. ALMA consists of 66 high-

sensitivity antennas that can be placed in compact or extended configurations depending

on the science goals of the observation. For this thesis I will use observations from ALMA

as it is the only observing facility in the southern hemisphere that is currently capable of

resolving the angular sizes required to characterize molecular gas kinematics on a ∼ kpc

scale in galaxies at z & 3. An accurate artist’s rendition of ALMA is shown in Figure 2.3.

(26)

Figure 2.3: An artist rendering of the Atacama Large Millimeter/submillimeter Array located in the Atacama Desert, Chile. Image taken from

https://www.britannica.com/topic/Atacama-Large-Millimeter-Array

2.2 Calibration and imaging

2.2.1 Calibration

In order to ensure that the observations accurately characterize the emission from the sources in my sample, imperfections in the receivers (or antennas) and atmospheric affects that can cause absorption and phase delays must be accounted for. This step in the data handling process is referred to as calibration. There are generally two steps of calibration, online calibration and offline calibration. Online calibration steps are performed at the telescope while the observations are being carried out, while offline calibration steps are performed remotely after the observations have been carried out.

Online calibration steps include pointing and focus calibration (to ensure that the target source is centered in the field of view and in focus), ensuring that each antenna is in the exact correct position, and calibrating any imperfections in the shape or reflecting surfaces of each antenna.

Offline calibration steps that were carried out as a part of this thesis include ampli-

tude calibration, atmospheric water vapor measurements, bandpass calibration, flux

calibration, and gain/phase calibration. Amplitude calibration measures the tempera-

ture contribution to each antenna from its own electrical components and the ambient

environment. Water vapor measurements account for the atmospheric water vapor

between the source and receivers at the time of observation (this can also change over

(27)

Methodology 17

the time of the observation - this is measured and accounted for as well) using a water vapor radiometer. Bandpass calibration uses a very bright point source (typically a quasar observed at high-frequency so that no flux is missed within a single beam) to calibrate the amplitude and phase for the target source in each channel over all spectral windows. Flux calibration uses a source of known brightness to define the conversion factor between temperature in Kelvin (K) on the receiver and brightness of the source on the sky in Jy. Finally, gain/phase calibration is used to monitor how the receiver is responding to the source elevation (given the weight load of the antenna) and to monitor the phase between receivers as the source moves across the sky and the atmosphere changes. To analytically account for all of the imperfections and inconsistencies that affect the observations, a set of calibration terms are applied to the true visibility function on the sky for antennas i and j (V

_ij^true

). The visibilites produced by antennas i and j (V

_ij^obs

) can be described by the function:

V

_ij^obs

= M

_ij

B

_ij

G

_ij

D

_ij

E

_ij

P

_ij

T

_ij

V

_ij^true

(2.3) where M

_ij

accounts for baseline-based errors, B

_ij

is the baseline response, G

_ij

is the gain amplitude and phase, D

_ij

is the instrumental polarization, E

_ij

represents the errors encountered due to the target source’s elevation, P

_ij

is the change in parallactic angle, T

_ij

accounts for variations in the length and opacity of the path of observation. The parameters are determined using least-squared fitting.

2.2.2 Imaging

After the data has been collected and calibrated, I use an imaging script in the Common Astronomy Software Applications (CASA; a Python package used for processing ALMA data; Jaeger, 2008) to generate maps of the sources in both continuum and line emission.

I generate data cubes to get a visual representation of the spatial and spectral energy distribution of the emission in the sources. To produce these data cubes I use a script that takes the spectrum at each pixel in the field of view to generate a raw three-dimensional (u, v, and frequency) data cube.

To deconvolve the dirty beam from the dirty images, leaving a clean image of the

source representative of the sky emission, I perform a task known as “cleaning" (in the

case of this analysis the Högbom clean algorithm (Högbom, 1974) was used). During

cleaning, the script iteratively runs through each channel determining and logging

the brightest points until a set threshold is reached. This threshold, determining to

what level the cleaning is done, could be a certain number of iterations or a limit to

the sensitivity (for example 3σ), defined by the user. This generates a file detailing the

coordinates and fluxes of all points above the set threshold which is then deconvolved

with the clean beam to produce a clean image of the source in each channel of the new,

clean data cube. The clean beam is an approximation of the main lobe of the dirty beam.

(28)

During this imaging process different weighting can be given to different baselines, allowing for a strategic trade-off between sensitivity and resolution depending on the nature of the observed source. Natural weighting, which is the default setting for cleaning, applies more weight to the shortest baselines in the configuration, producing ideal imaging for spatially extended, faint objects with a low signal-to-noise ratio (SNR).

Uniform weighting gives more weight to the longer baselines, producing higher-resolution imaging used for bright, compact sources where high angular resolution and lower sensitivity is needed. Briggs weighting is an alternative weighting scheme that allows the user to alter the robust parameter to explore different weightings. Using robust = 2 produces similar results as using natural weighting, whereas using lower robust values such as robust = − 2 gives the same results as uniform weighting. For the purposes of this analysis I use Briggs weighting with robust = 0.5 when cleaning the sources with SNR ≥ 10 and natural weighting when cleaning the sources with SNR < 10. Figure 2.4 shows the moment-0 map examples for one of the sources studied in this thesis using natural weighting and Briggs weighting with robust = 0, 0.5 and 1 to illustrate the difference that using natural versus Briggs weighting with varying robust parameters makes in the imaging process.

Figure 2.4: Examples of imaging using natural weighting (top left panel), Briggs weighting with robust=1 (top right panel), Briggs weighting with robust = 0.5 (bottom left panel) and Briggs weighting with robust = 0 (bottom right panel).

The synthesized beam of the observation (shown in the bottom left of each panel of

Figure 2.4) becomes smaller as the robust parameter is decreased (beam

_nat

= 0.39 arcsec

(29)

Methodology 19

× 0.30 arcsec; beam

_robust=1

= 0.37 arcsec × 0.29 arcsec; beam

_robust=0.5

= 0.32 arcsec × 0.26 arcsec; beam

_robust=0

= 0.28 arcsec × 0.23 arcsec).

The first step in the imaging process was to clean every channel in every spectral window, giving a clean data cube over the observed frequency range. Each observation is carried out in four separate spectral windows that cover different regions of the electromagnetic spectrum - channels are frequency ranges used to bin the observed spectral data. From this data cube I determine which channels contain line emission and which channels contain continuum emission (which are called line-free channels). From here, I bin all of the uv-data from the channels determined to be line-free into one channel which I clean to generate an image of the continuum. To generate a data cube of the line emission I take the original uv-data and perform a linear fit to the line-free channels, which is then subtracted from the data cube. This leaves a new uv-data set with only line emission

¹

that is cleaned to produce a data cube showing the line emission. Finally, I produce moment maps of the sources illustrating different moments (quantities describing the distribution of a value, in this case CO emission, throughout physical space) with respect to velocity in their molecular gas reservoirs. Taking the measured intensity (I

ν

) at each discrete frequency (dν) over the entire observed spectrum (ν) at each pixel gives the three moment maps that are used in this analysis:

• _{Moment-0 (} R

I

ν

dν): This shows the amount of gas present at each pixel in the image. I take the channels with line emission and collapse (sum) them into one single channel that describes the spatial distribution of the emission. A moment-0 map of CO emission in one of the sources studied in this thesis is shown in the left panel of Figure 2.5.

• Moment-1 ( R (

ν

) I

_ν

dν): The first derivative of moment-0. This map shows how fast the gas at each pixel is moving towards or away from the observer. The velocity offset of each pixel with respect to rest-frame of source is taken and compiled to generate a map showing the velocity of the emitting gas. An example of a moment-1 map is shown in the center panel of Figure 2.5.

• _{Moment-2 (} R (

ν²

) I

_ν

dν): The second derivative of moment-0. This map shows the velocity dispersion of the gas at each pixel. This is sometimes also referred to as the FWHM map. An example of a moment-2 map is shown in the right panel of Figure 2.5.

Figure 2.5 shows moment-0, moment-1 and moment-2 maps for one of the sources studied in this thesis.

1usually some continuum channels are left on either side of the channels containing line emission to show the relative brightness of the emission line to the noise of the spectrum, and that the continuum subtraction was executed properly

(30)

Figure 2.5: Examples of moment-0 (left panel), moment-1 (center panel), and moment-2 (right panel) maps.

These images along with continuum maps and spectral data allow analysis of the kinematics and physical properties of the molecular gas in the sources from the sample.

2.3 Sample used for analysis

To study the evolution of massive galaxies and the production of unobscured quasars,

objects with redshifts of z ∼ 2 − 3 are studied, as this is when SFR and quasar luminosity

have been observed to peak (see Figures 1.2 and 1.3). For this, I study nine WISE-selected

Hot DOGs The nine sources studied in this thesis were selected as they have W ISE

photometry in the literature (Tsai et al., 2015) and previously measured optical redshifts

placing them at z ≈ 3 − 4.6. Table 2.1 shows the sky position in right ascension (RA)

and declination (DEC), the optical redshift (z

_opt

), and references to the literature for

the discovery and any photometry that has been performed on the sources chosen for

analysis in this thesis. These nine sources are from ALMA project 2017.1.00358.S.

(31)

Methodology 21

Table 2.1: Introduction of nine Hot DOGs used for analysis.

Source RA (J2000) DEC (J2000) z

opt

Discovery/Photometry [arcsec] [arcsec]

W0116-0505 01:16:01.42 -05:05:04.20 3.173

¹ ⁶

W0134-2922 01:34:35.71 -29:22:45.40 3.047

³ ⁶

W0831+0140 08:31:53.26 01:40:10.80 3.910

⁵ ^{5 6}

W1248-2154 12:48:15.21 -21:54:20.40 3.318

² ⁶

W1322-0328 13::22:32.57 -03:28:42.20 3.043

⁴ ⁶

W2042-2345 20:42:49.28 -32:45:17.90 3.968

⁶ ⁶

W2246-0526 22:46:07.57 -05:26:35.00 4.593

⁶ ^{5 6}

W2246-7143 22:46:12.07 -71:44:01.30 3.458

⁶ ^{5 6}

W2305-0039 23:05:25.88 -00.39.25.70 3.106

² ⁶

1

Assef et al. (2020)

2

Fan et al. (2016b)

3

Fan et al. (2017)

4

Finnerty et al. (2020)

5

Jones et al. (2014)

6

Tsai et al. (2015)

Observations of the sample analyzed for this thesis were carried out with ALMA during Cycle 5 in Band 3. A summary of the observations is given in Table 2.2. For each source the receiver was tuned to the redshifted CO(4-3) line

²

using the optical redshifts from Table 2.1. This table gives the observing dates, number of antennas used for each observations, calibrator sources used, and representative frequencies for each spectral window. The sources listed twice in this table represent two separate execution blocks used for the observation.

2W2246-0526 was detected in CO(5-4) emission due to observational constraints, and therefore the receiver was tuned to the redshifted CO(5-4) line.

(32)

Table 2.2: Summary of the ALMA observations.

SourceDateNantFluxCalib.PhaseCalib.νspw,centralνcont,central [dd-mm-yyyy][GHz][GHz] W0116-050505-12-201747J0006-0623J0116-1136110.470108.485,96.485,98.402 W0134-2922[X5ef8]12-12-201743J2357-5311J0120-2701113.909111.987,99.989,101.789 W0134-2922[X2c9]13-12-201747J2357-5311J0120-2701113.909111.987,99.989,101.789 W0615-57161 07-12-201747J0519-4546J0550-5732104.805106.695,92.695,94.695 W0831+014007-12-201747J0750+1231J0839+010494.32496.203,106.504,108.308 W1248-215420-12-201744J1337-1257J1245-1616106.792105.000,94.809,93.109 W1322-0328[X304d]20-12-201745J1256-0547J1312-0424114.047112.112,100.111,102.003 W1322-0328[X3390]20-12-201742J1256-0547J1312-0424114.047112.112,100.111,102.003 W2042-324510-12-201744J2056-4714J0839+010492.89194.773,104.815,106.695 W2246-052607-12-201746J2134-0153J2229-0832102.891104.841,90.962,92.849 W2246-714303-12-201747J2357-5311J2229-6910103.412105.293,91.400,93.294 W2305-0039[X2fe3]10-12-201744J0006-0623J2301-0158112.275114.195,100.278,102.195 W2305-0039[X3151]10-12-201744J0006-0623J2301-0158112.275114.195,100.278,102.195

1The emission from W0615-5716 is very faint, and there were issues with the imaging process. It was excluded from this analysis, as further imaging was beyond the scope of this thesis.

(33)

Results 23

3 Results

3.1 Emission line spectra fitting results

To extract the emission line spectra of the sources in the sample (see Figure 3.2) the simplifying assumption was made that the sources are described by rectangular boxes.

The spectra at each pixel over the rectangular region determined to contain the line emission is then summed into a single spectrum used to characterize the line emission of the source. Figure 3.1 shows an example of the regions selected for spectral extraction for three sources in this analysis (W0831+0140, W2246-0526 and W2305-0039 - these three sources were chosen as examples as they both have confidently detected emission and display different morphologies).

Figure 3.1: Moment-0 maps for W0831+0140 (left panel), W2246-0526 (middle panel), and W2305-0039 (right panel) with regions used for spectral extraction shown with the black boxes

To fit Gaussian functions to these spectra, a fitting function (scipy.optimize.curve_fit)

was used in Python. Figure 3.2 shows the observed emission line spectra for each source

in the sample. The dashed red line represents the Gaussian fit that best describes the

observed spectra.

(34)

Figure 3.2: Observed emission line spectra for nine sources with Gaussian fits displayed with the dashed red line. For W0831+0140, the blue Gaussian represents the blueshifted component while the red Gaussian represents the redshifted component.

Table 3.1 shows the reduced χ

²

(χ

²

per degree of freedom) values for both single and

double Gaussian fits of the observed spectra. Only the Gaussian fit that best describes the

observed data is plotted in Figure 3.2. W0831+0140 is the only source with line emission

that was better described with a double Gaussian than a single Gaussian, indicating the

presence of two separate emission components separated by ∼ 420 km s

⁻¹

, possibly due

to rotation or a past/ongoing merger.

(35)

Results 25

Table 3.1: Reduced χ²(χ²per-degree-of-freedom) for single and double Gaussian fits of observed CO emission line (left panel in Figs. 4.1 - 4.9). The form of Gaussian producing the best fit was used to characterize the emission line. The only source in the sample that was better fit with a double Gaussian than with a single Gaussian is W0831+0140.

Source χ²/N χ²/N

Single Gaussian Double Gaussian

W0116-0505 0.30 0.33

W0134-2922 0.44 4.77

W0831+0140 0.25 0.07

W1248-2154 2.35 2.50

W1322-0328 0.58 0.62

W2042-3245 1.24 53.46

W2246-0526 0.07 0.08

W2246-7143 0.25 0.27

W2305-0039 8.07 8.61

Table 3.2 shows the parameters associated with the Gaussian fits and the SNR of the

moment-0 maps for the sources in the sample. z

CO

is the redshift determined for the

source using CO line emission. SNR

_spec

and SNR

_map

are the signal-to-noise ratios of the

source’s spectra and moment-0 maps, respectfully. S

_peak

is the peak of the Gaussian fit

(in mJy) chosen to describe the line emission. FWHM is the velocity dispersion of the

gas (described below Table 3.2), and I

_CO

is the velocity-integrated flux of the CO line (in

units of Jy km s

⁻¹

).

(36)

Table 3.2: CO emission line fitting parameters for nine Hot DOGs.

Source z_CO(4-3) SNRmap1 SNRspec2 S_peak FWHM I_CO(4-3)

[mJy] [km s⁻¹] [Jy km s⁻¹] W0116-0505 3.1900±0.0005 9.01 6.16 1.12±0.12 613±75 0.73±0.12 W0134-2922 3.0590±0.0008 6.81 3.22 0.25±0.05 606±142 0.16±0.05 W0831+0140 3.9154±0.0006 7.09 3.52 1.87±0.13 897±63 1.76±0.27 W1248-2154 3.3250±0.0009 5.31 3.83 0.51±0.09 686±135 0.37±0.10 W1322-0328 3.0446±0.0001 21.07 16.11 3.24±0.13 464±21 1.60±0.09 W2042-3245 3.9690±0.0005 6.25 3.88 0.88±0.15 403±8 0.38±0.10 W2246-0526³ 4.6001±0.0004 10.82 7.84 1.30±0.11 599±58 0.83±0.11 W2246-7143 3.4630±0.0003 14.68 10.86 2.53±0.15 755±53 2.03±0.19 W2305-0039 3.1108±0.0001 20.92 16.07 4.06±0.17 536±25 2.32±0.14

1SNRmapwas calculated by taking the flux density of the emitting region averaged over the channels containing line emission divided by the noise outside of the emitting region averaged over the same channels. In the case that the emission is resolved, a multiplication factor of ^√¹

Nis added, where N is the number of primary beams that fit within the emitting region. This value is systematically higher in all of the sources in the sample.

2SNRspecwas calculated by taking ICO/(error of ICO), giving the ratio of the spectral detection to the spectral noise across all channels within the emitting region.

3CO(5-4) emission was detected for W2246-0526.

To determine the velocity dispersion of the emission lines in the sources, the full width

of the Gaussian fit at half of its maximum flux (FWHM) is taken (see Figure 3.3).

(37)

Results 27

Figure 3.3: Example of FWHM for a single Gaussian. Figure adapted from Leo (1994)

3.2 Imaging results

3.2.1 Moment maps

Figure 3.4 shows the moment maps for the sources included in this analysis. The left

panel of each row shows the moment-0 map of the CO emission, which illustrates

the spatial distribution of the molecular gas in the sources. The center panel of each

row shows the moment-1 map of the CO emission, which illustrates the velocity of

the molecular gas in the sources relative to the source rest-frame. The right panel of

each row shows the moment-2 map of the CO emission, which illustrates the velocity

dispersion (FWHM) of the gas in the sources.

(38)

Figure 3.4: Moment-0 (left), moment-1 (center), and moment-2 (right) maps for the sources in the sample. Moment-2 map velocity dispersion values were calculated as the FWHM at each pixel. For the moment-1 and moment-2 maps, only those pixels with detections that are≥three times the RMS of the field are shown. Moment-0 contours are overlaid in black representing 4σ, 5σ, 7σ, 9σ, 12σ, 15σ, and 18σ detections (when such high SNR detections were observed). The final three sources in the sample

(W2246-0526, W2246-7143, and W2305-0039) are shown on the next page.

(39)

Results 29

The sources show a range of morphologies and velocity profiles that I discuss in

greater detail in Chapter 4. Table 3.3 gives the resolution information in arcseconds for

the sources in the sample. The weighting scheme (natural or Briggs with robust = _{0.5) is}

shown along with the beam size of the observation. The position angle, major axis and

minor axis of the source are also given. Included is information about whether or not

the source could be resolved in CO emission.

(40)

Table 3.3: Resolution information for observations and sources in arcsec.

SourceBaselineWeighting1BeamSizePositionAngleMajorAxisMinorAxisResolved? [arcsec×arcsec][degrees][arcsec][arcsec]y/n W0116-0505Natural0.32×0.23-64.810.46±0.090.34±0.08y W0134-2922Natural0.38×0.28-61.41--n2 W0831+0140Briggs(robust=0.5)0.32×0.2673.210.44±0.060.38±0.05y W1248-2154Natural0.55×0.4724.38--n W1322-0328Briggs(robust=0.5)0.36×0.3556.820.38±0.060.21±0.06y W2042-3245Natural0.54×0.39-75.52--n W2246-0526Natural0.37×0.24-61.890.30±0.060.27±0.07y W2246-7143Briggs(robust=0.5)0.36×0.2513.160.39±0.070.29±0.09y W2305-0039Briggs(robust=0.5)0.34×0.2767.710.48±0.080.38±0.08y

1Natural weighting was used as the default weighting for imaging, giving the best possible sensitivity with the worst possible angular resolution with the interferometer setup of the observation. When the source has a high SNR and is resolved, it is worth considering using Briggs weighting, which allows for a robust parameter. This parameter ranges from (−2, 2), with robust=-2 returning the highest possible angular resolution with the worst possible sensitivity and robust=2 returning the same results as natural weighting.

2W0134-2922 is marginally resolved, but not resolved well enough to estimate the size of the emitting region.

(41)

Results 31

Table 3.4 shows the sizes of the CO emitting regions in kpc for the six sources that could be resolved. The major axis and minor axis are given along with the effective radius and emitting area of the source.

Table 3.4: CO emission-derived size estimates in kpc for the six Hot-DOGs from the sample that could be spatially resolved and fit with 2-dimensional Gaussian profiles (see Table 3.3).

Source Major Axis Minor Axis Re1 Emitting Area

[kpc] [kpc] [kpc] [kpc²]

W0116-0505 3.6±0.7 2.7±0.6 3.3±1.3 30.0±9.3 W0831+0140 3.2±0.4 2.7±0.4 3.0±0.8 26.9±5.2 W1322-0328 3.0±0.5 1.6±0.5 2.5±0.9 15.5±5.1 W2246-0526 2.0±0.4 1.8±0.5 2.0±0.9 11.5±3.9 W2246-7143 2.9±0.5 2.2±0.7 2.7±1.2 20.1±7.2 W2305-0039 3.8±0.6 3.0±0.6 3.5±1.3 34.9±9.4

1The effective radii (Re) of the emitting regions of the sources were estimated using the simplifying assumption that the region displays a uniform elliptical geometry so that Re= ^2A+B₃ , where A is the major axis and B is the minor axis of the ellipse.

3.2.2 Continuum maps

Figure 3.5 shows the continuum maps of the nine sources analysed in this thesis with

mid-J CO emission contours overlaid in black. For most of the sources the continuum

emission is more compact than the mid-J CO emission, as the fainter more extended

continuum emission is undetected at the sensitivity used for the observations. The mid-J

CO emission has a higher SNR then the continuum emission, therefore the fainter edges

of the emission are more easily distinguished from the noise in the field. Table 3.5 shows

the flux density, SNR and sizes of the continuum emission of the sources in the sample.

(42)

Figure 3.5: Continuum maps of nine Hot DOGs analyzed in this thesis. CO(4-3) (CO(5-4) for W2246-0526) moment-0 contours are overlaid in black representing 4σ, 5σ, 7σ, 9σ, 12σ, 15σ, and 18σ detections (when such high SNR detections were observed).

Study of CO Emission in Nine Hot Dust-Obscured Galaxies at z ∼3