Department of Physics
Master of Science
in Astronomy and Space Physics Thesis Dissertation
Ultraluminous sources in X-ray sky surveys
Author:
Miquel Colom i Bernadich
Supervisors:
Dr. Axel Schwope Prof. Erin O’Sullivan Topic Reader:
Prof. Erik Zackrisson
June 10, 2020
ABSTRACT
Ultraluminous X-ray sources (ULXs) are extragalactic, non-nuclear, point-
like X-ray sources whose luminosity supersedes that of the Eddington limit
of an accreting stellar mass black hole (L > 10
39erg/s). Most of them
are powered by black holes and neutron stars undergoing genuine super-
Eddington accretion, with a small handful of candidates being consistent
with sub-Eddington accretion on to an intermediate mass black hole. In this
thesis, we explore the populations of ULXs in the sky surveys of ESA’s X-
ray satellite, XMM-Newton, and the MPE’s newly launched X-ray telescope,
eROSITA. We do so by correlating them with the HECATE list of galaxies
to build two X-ray non-nuclear catalogues, and comparing the yields between
the very different surveys and previous works. To build a catalogue, we use
other reference lists of contaminant objects, such as the Gaia data releases,
the SIMBAD database or the SDSS survey to look for contaminating ob-
jects of diverse nature, such as foreground stars or background quasars, in
order to make sure that our resulting ULX samples are as clean as possible
with catalogue data only. Our results include attesting that the ninth data
release of XMM-Newton provides an improvement in quantity and quality
with respect to older data releases used in previous works, and that the
eROSITA survey is currently in a very preliminary stage. The two new cat-
alogues contain 12,952 and 3,720 non-nuclear X-ray sources, out of which
914 and 132 are ULX candidates with an expected ∼ 25% fraction of unde-
tected contaminants. This constitutes a very significant contribution to the
already known ∼ 300 ULX candidates. Since the sky coverage and depth
of the XMM-Newton and eROSITA surveys are vastly different, only 19 of
the ULX candidates are shared between the catalogues. ULX candidates are
found in preferentially in star-forming galaxies, but a subset of very bright
objects (L > 5 × 10
40erg/s) prove to be more common in elliptical galaxies,
in contradiction to what has been established in the literature.
Popul¨ arvetenskaplig sammanfattning
Ett dubbeltstj¨ arnsystem d¨ ar den ena komponenten ¨ ar ett svart h˚ al (eller
en neutronstj¨ arna) som stj¨ al massa fr˚ an den andra stj¨ arnan i systemet kallas
f¨ or “X-ray binaries”. Gasen som stj¨ als fr˚ an stj¨ arnan f˚ ar extremt h¨ og kinetisk
energi n¨ ar den faller mot det svarta h˚ alet och den f¨ orvandlas till v¨ armeenergi
n¨ ar gasen som stulits tr¨ affar gasen i det svarta h˚ alets ackretionsskiva. Till slut
blir v¨ armen s˚ a h¨ og att gasen b¨ orjar avge r¨ ontgenstr˚ alning som kan observeras
fr˚ an jorden. Det finns dock en ¨ ovre gr¨ ans f¨ or hur mycket r¨ ontgenstr˚ alning
som kan skickas ut. Detta ¨ ar f¨ or att om gasen i det svarta h˚ alets ackre-
tionsskiva tr¨ affas av r¨ ontgenstr˚ alning med tillr¨ ackligt h¨ og energi s˚ a kommer
gasen att sl˚ as ut ur omloppsbanan och detta leder till att ackretionsskivan
tillslut f¨ orsvinner. Energin som kr¨ avs f¨ or att detta ska ske kallas f¨ or Edding-
tongr¨ ansen och beror p˚ a hur massivt det svarta h˚ alet ¨ ar. Desto h¨ ogre massa
som det svarta h˚ alet har, desto h¨ ogre energiniv˚ a kan r¨ ontgenstr˚ alningen som
avges fr˚ an ackretionsskivan uppn˚ a utan att skivan “bl˚ ases bort”. Dock har
det uppt¨ ackts att det finns objekt i andra galaxer som uts¨ ondrar r¨ ontgen-
str˚ alning ¨ over deras Eddingtongr¨ ans. Dessa objekt kallas f¨ or ultralumin¨ osa
r¨ ontgenk¨ allor. Dessa objekt observeras oftast i spiral eller oj¨ amna galaxer
d¨ ar det finns tillr¨ ackligt mycket gas f¨ or att skapa de enorma stj¨ arnorna som
bildar “X-ray binaries”. Anledningen till varf¨ or dessa objekt kan uts¨ ondra
r¨ ontgenstr˚ alning med energi ¨ over vad Eddingtongr¨ ansen s¨ ager ¨ ar m¨ ojligt ¨ ar
f¨ or att det ¨ ar endast de yttersta delarna av ackertionsskivan som “bl˚ ases
bort” av str˚ alningen. En annan m¨ ojlighet ¨ ar att objektet best˚ ar av ett kon-
stigt svart h˚ al som ¨ ar mera massivt ¨ an de som skapas av supernovor men
mindre ¨ an de som bildas i galaxers centrum. I denna avhandling anv¨ ander
vi data fr˚ an r¨ ontgenstr˚ alningsteleskopen XMM-Newton och eROSITA, som
b˚ ada ligger i omloppsbana runt jorden, f¨ or att skapa tv˚ a nya stj¨ arnkataloger
av ultralumin¨ osa r¨ ontgenk¨ allor. Dessa kataloger konstrueras genom att f¨ orst
hitta r¨ ontgenstr˚ alningsk¨ allor som ligger i redan k¨ anda galaxer, sedan avg¨ or
vi ifall dessa r¨ ontgenk¨ allor redan tillh¨ or k¨ anda stj¨ arnor eller kvasarer. Vi
v¨ aljer sedan de ljusaste r¨ ontgenk¨ allorna som inte ¨ ar redan katalogiserade och
unders¨ oker ifall dom ¨ ar m¨ ojliga ultralumin¨ osa r¨ ontgenk¨ allor. ¨ Over 1000 nya
m¨ ojliga ultralumin¨ osa r¨ ontgenk¨ allor i det lokala universumet. F¨ orv˚ anansv¨ art
visar det sig att de ljusaste objekten finns oftare i elliptiska galaxer som har
mindre m¨ angd gas.
Acknowledgments
We would like to thank Dr. Iris Traulsen for her invaluable expertise and technical aid both during our work time in the office and during these shaky times of confinement. Also to Konstantinos Kovlakas for providing the HECATE list of galaxies, and to Jan Kurpas for providing the astrometrical tools used to draw many of the figures presented in this report.
This thesis has made use of data obtained from the 3XMM and 4XMM XMM-Newton Serendipitous Source Catalogs compiled by the 10 institutes of the XMM-Newton Survey Science Centre selected by ESA, and prelimi- nary data from the eROSITA All Sky Survey provided by the MPE through Dr. Georg Lamer on April 27, 2020. We have used images from the Pan- STARRS1 Surveys and the cross-match service provided by CDS, Strasbourg.
And finally, the author of this thesis gives his personal thanks to his
flatmates for being their constant company during the lockdown.
TABLE OF CONTENTS
Page
LIST OF TABLES . . . . iii
LIST OF FIGURES . . . . iv
CHAPTER 1 Ultraluminous X-ray Sources . . . . 1
1.1 Introduction . . . . 1
1.2 X-ray astronomy, a brief historical summary . . . . 2
1.3 X-ray binaries . . . . 5
1.3.1 Their nature and the Eddington limit . . . . 5
1.3.2 Galactic Black Hole Binaries . . . . 6
1.4 Ultraluminous X-ray sources . . . . 8
1.4.1 Their nature . . . . 8
1.4.2 Alternatives to super-Eddington accretion . . . . 10
1.4.3 ULXs as Intermediate Mass Black Holes . . . . 11
1.4.4 A model of super-Eddington accretion . . . . 12
2 ULX yields from different XMM-Newton data releases. . . . 13
2.1 Introduction . . . . 13
2.2 The XMM-Newton Observatory . . . . 14
2.3 Data sample. . . . 15
2.4 The filtering process. . . . 16
2.4.1 Matching detections with galaxies. . . . 17
2.4.2 Removing known contaminants, phase 1 . . . . 18
2.4.3 Removal of high flag sources . . . . 19
2.4.4 Removal of more known contaminants, phase 2 . . . . 19
2.5 Results summary . . . . 20
2.6 Comparisons between our and older works. . . . 22
2.7 Conclusions. . . . 27
3 ULX catalogues from the XMM-Newton and eROSITA surveys . . . 30
3.1 Introduction . . . . 30
3.2 The eROSITA telescope . . . . 31
3.3 Data samples . . . . 32
3.4 The filtering process . . . . 33
3.4.1 Matching sources with galaxies . . . . 34
3.4.2 Identification of known contaminants, automatic phase . . . . 35
3.4.3 Identification of known contaminants, SIMBAD phase . . . . 36
3.4.4 Manual identification of contaminants . . . . 37
3.5 Identification of ULXs . . . . 40
3.6 Building luminosity sub-samples . . . . 41
3.7 Results . . . . 42
3.8 Discussion . . . . 47
3.8.1 Properties of ULX candidates . . . . 47
3.8.2 Intermediate Mass Black Holes in the catalogues . . . . 49
3.8.3 The contaminant population . . . . 50
3.8.4 The performance of eRASS1 . . . . 53
3.8.5 Comparison with older catalogues . . . . 57
3.9 Conclusions . . . . 58
REFERENCES . . . . 64
APPENDIX
LIST OF TABLES
2.1 XMM-Newton specifications . . . . 15 2.2 The amount of detections of sources, sources, and parent galaxies for myDR4
and myDR9 after every step in our filtering process. . . . 29 2.3 Breakdown of the source types for E19’s, myDR4 and myDR9 at the end
of our filtering pipeline. . . . 29 3.1 eROSITA specifications . . . . 31 3.2 Breakdown of detections, sources, and parent galaxies for myNewDR9 and
myER1. . . . 43 3.3 Breakdown of detections, sources, and parent galaxies for objects of quality. . 43 3.4 Mean hardness ratios in the complete ULX and bright ULX sub-samples from
myNewDR9 and their standard deviation. . . . 46 3.5 The complete ULX sub-samples and their host halaxies, divided according to
morphology groups. . . . 48 3.6 Breakdown of contaminants. . . . 50 3.7 Mean hardness ratios of the contaminant populations from myNewDR9 and
their standard deviations. . . . . 50
LIST OF FIGURES
2.1 Example images of galaxy NGC 2631. . . . 20
2.2 Luminosity distribution of the X-ray sources in myDR4 and myDR9. . . . 21
2.3 Dispersion of the the galaxy size changes in HECATE versus RC3 & Dispersion of galaxy distance versus B magnitude. . . . 22
2.4 Astrometric TOPCAT maps of lost sources from E19 and new ones in myDR4 23 2.5 Optical images sources classified as extended in 3XMM-DR4 but present in E19 . . . . 24
2.6 Detections of sources classified as extended in 3XMM-DR4 but present in E19. 25 2.7 Detections from myDR4 disregarded in myDR9 . . . . 26
3.1 myNewDR9 example of OBJ_TYPE =“optical counterpart”. . . . 38
3.2 myER1 example of OBJ_TYPE =“optical counterpart” . . . . 38
3.3 myNewDR9 example of OBJ_TYPE =“unresolved double, optical counter- part” . . . . 39
3.4 myNewDR9 example of OBJ_TYPE =“spurious detection” . . . . 39
3.5 myNewDR9 and myER1 complete sub-samples . . . . 41
3.6 Distance distributions of sources and ULX candidates of quality . . . . 44
3.7 Luminosity distributions of the sources of quality in myNewDR9 and myER1 according to the morphological types of parent galaxies . . . . 45 3.8 Hubble type distributions of the complete ULX and bright ULX subsamples
in myNewDR9 and myER1 . . . . 45 3.9 Hardness ratio dispersion from objects in the complete ULX and bright ULX
subsamples from myNewDR9 . . . . 47 3.10 Hardness ratio dispersions of different populations from myNewDR9. . . . 51 3.11 Correlation of 4XMM-DR9 and myERASS1 sources. . . . . 53 3.12 Astrometric map and eROSITA photon image of galaxy NGC 1291, with some
sources recovered from 4XMM-DR9 . . . . 55 3.13 XMM-Newton images, eROSITA photon image, and astrometric map of galaxy
NGC 4649. . . . 56
Chapter One
Ultraluminous X-ray Sources
1.1 Introduction
Ultraluminous X-ray sources are interesting from many points of view. From the cosmological
point of view, they are a window to the stellar activity in the local universe and a very
valuable tool for the search of intermediate mass black holes. For astrophysics, they represent
some of the most extreme cases of stellar evolution. They are also a window to the most
extreme physical environments of the universe, involving super-Eddington accretion on to
stellar mass black holes and magnetized neutron stars. Studies in other wavelengths, or
even other astronomy messengers such as neutrinos (Mushtukov et al., 2018), should help
us to unveil the nature of these objects even further. In order for the astronomy community
to continue its research in this field, it is necessary to have a large sample of objects to
study. The objective of the projects contained in this thesis is to build two new catalogue
resources for the study of ULXs. The most recent cataloguing work on ULXs was published
by Earnshaw et al., 2019 contains 384 were ULX candidates, out of which 9 could potentially
be IMBH. Older works, such as Swartz et al., 2011 and Walton et al., 2011, list 107 and
470 objects. These works have used X-ray data from the XMM-Newton or Chandra X-ray
telescopes. We make use of the most recent data release of XMM-Newton, and survey data
from the recently launched eROSITA telescope, to expand upon these previous works.
In this chapter, we make an introduction to the objects of our study. We star with a brief overview of the history of X-ray astronomy, in order to introduce the reader to some of its tools, usefulness, and limitations. We follow with a description of X-ray binaries, their properties, and an explanation of the Eddington limit in Section 1.3. And in Section 1.4, we delve into the nature of ultraluminous X-ray sources, overviewing their observational properties and the astrophysical processes shaping them. Chapter 2 serves as the report of a project in which we analyzed the identity of the resources at our hand by trying to reproduce some of our previous work. It also serves as a preliminary ground for Chapter 3, where we carefully describe the building of two new catalogues of non-nuclear X-ray sources and present our results.
1.2 X-ray astronomy, a brief historical summary
X-ray astronomy focuses on the high energy band (0.1-200 keV) of the electromagnetic spec- trum. As such, it is a window to some of the most energetic events and extreme environments of the universe. However, it comes with one very important limitation: X-rays are prone to atmospheric absorption, so much that not even the most powerful X-ray sources can be studied from the ground, making modern X-ray astronomy exclusively a domain of satellite astronomy. But the history of X-ray astronomy and its development began much before satellites had been invented.
It was the Sun what sparked the motivation for the study of the X-ray sky. In 1949,
a US military rocket equipped with photomultiplier tubes was launched to study the Sun’s
role in the dynamics of Earth’s ionosphere, confirming the hypothesis of it being a source of
ionizing radiation. As it is well explained in the review paper of Santangelo and Madonia,
2014, due to total X-ray absorption by Earth’s atmosphere, observations of this kind relied
on the short time frame offered by ballistic missile flights above the air column before falling
back to Earth, and for the following years researchers kept focusing on solar studies. It was
not until 1962 that the first great discovery was made. A rocket mission with two working Geiger-Müller counters, designed by researchers in American Science and Engineering Society (AS&E) and the MIT, was launched to measure X-rays from the Moon. The experiment failed to see any lunar radiation, but it detected a very bright X-ray source in the Scorpio constellation instead (Giacconi et al., 1962), an event that sparked the scientific interest in extrasolar X-ray astronomy. In 1963, another rocket with more sensitive instrumentation was launched, not only confirming the source in Scorpio, but also discovering a new one in the Crab Nebula and the presence of an X-ray radiation background (Bowyer et al., 1964). It was immediately proposed that neutron stars were responsible for such bright X- ray objects (Morton, 1964), and postulated that the X-ray background was of extragalactic origin (Bowyer et al., 1964). A remarkable boom of publications in X-ray astronomy occurred in the immediate years (Santangelo and Madonia, 2014), with the discovery of tenths of sources in the galactic plane (Friedman et al., 1967), and the observation of X-ray emission from supernova remnants and radio galaxies (Byram et al., 1966), and the first modellings of X-ray binary emission (Truemper, 1978), with the use of balloons carrying scintillators up to 40 km of height.
The next revolution came with the launch of the UHURU satellite, the first permanent
X-ray observatory in Earth’s orbit, in 1970 by the AS&E-MIT collaboration under NASA’s
supervision. Designed with sensitivity in the energy range of 2-20 keV and a spatial reso-
lution if 30’ (Giacconi et al., 1971), the satellite was active for a time-span of 3 years, and
the resulting survey catalogue contained up to 339 X-ray sources (Forman et al., 1978). An
equally remarkable satellite was the Einstein Observatory. Constructed by the same collab-
oration and launched in 1978, it was the first X-ray imaging telescope in the sense that it
allowed for the focus of incoming photons through a design of mirrors called Wolter Type-1
(Giacconi et al., 1969). Its high-resolution imager instrument reached a resolution of 2" on
axis, while it’s main imaging proportional counter reached 1’ (Giacconi et al., 1979). This
would set the trend of building all future X-ray satellites as telescopes with camera receptors
on their focal points, a property that carries on to the XMM-Newton and eROSITA tele- scopes. At the end of its life, the Einstein Observatory had unveiled that coronas, galaxies, and even hot gas in galaxy clusters also emitted X-rays
1.
At the end of the previous centuries and the beginning of the current one, many other
satellites have been designed to follow on the work of the Einstein Observatory in a true
explosion for X-ray astronomy. To describe all of these missions would mean to write an
entire review on the topic, a job already done in Santangelo and Madonia, 2014. However, we
will make some honorable mention here and remark their achievements, and refer the reader
to Santangelo and Madonia, 2014 for more details. The protagonists of X-ray astronomy at
the end of the previous century ROSAT, BeppoSAX and XMM-Newton, some of the greatest
symbols of the European contribution to the field. ROSAT was designed by the German
Aerospace Center and launched in 1990 with the far-UV range and soft X-rays in mind,
particularly useful for mapping the X-ray diffuse emission and for stellar studies. By contrast,
BeppoSAX, built by the Italian Space Agency, was launched in 1996 and was sensitive in
the 0.1-300 keV, playing a crucial role in unveiling the nature of Gamma-Ray Bursts. XMM-
Newton, launched in 1999 by the ESA, has been one of the most successful missions on
X-ray astronomy in history, and as of today it is still actively performing observations upon
request. Other telescopes of importance are NASA’s Chandra, launched in 1999 and still
active, its main attraction being an astonishing spatial resolution of 0.5", and NuSTAR from
2012, with a revolutionary resolution of 18" in the X-ray. And of course, this section has to
mention the Spektr-RG observatory containing the eROSITA telescope. Built by the Max
Planck Institute for Extraterrestrial Physics and launched by Roscosmos in 2019, its novelty
is that instead of staying on Earth’s orbit, it has been put in a halo orbit around L2, from
where it is currently performing a full survey of the X-ray sky (Merloni, 2019). Amongst
other objectives, the mission has been designed is to make cosmological studies from galaxy
clusters (Merloni et al., 2012).
1.3 X-ray binaries
1.3.1 Their nature and the Eddington limit
A key step for understanding the exceptionally of Ultraluminous X-ray Sources (ULXs), is to know the physics of one of the most studied objects in X-ray astronomy: X-ray binaries.
These kind of objects are born in the latter stages of an evolved binary pair. As a short review, stars evolve into red giants and supergiants when leaving the main sequence. During this phase, their radii become hundreds of times larger than the original one. Stars with initial mass M > 8 M end their lives at the end of this stage in a spectacular supernova, leaving behind either a neutron star (NS, M
NS≈ 1.4 M ) or a stellar mass Black Hole (StBH, M
StBH. 15 M
, Casares et al., 2017). The most massive stars have very short time-spans due to the extremely high temperatures and pressure at their cores. For instance, a 20 M
star will go through all these phases in less than 10 million years. The formation of an X-ray binary requires an original binary pair with at least one of the components satisfying M > 8 M . In this system, the most massive star will complete their life cycle first and turn into a NS or a StBH. If the pair remains closely bounded after the supernova event, when the companion star abandons the main sequence its expanded outer layers overfill the Roche lobe and mass transfer onto the main star occurs. In this scenario, the gravitational energy of the in-falling gas turns into thermal energy when it accumulates at the accretion disk around the compact remnant. Therefore, the mass transfer between the stars turns into very energetic radiation emission that may reach temperatures of k
BT ∼ 10 keV (Tauris and van den Heuvel, 2006). A very simple modelling of this phenomena, parameterized by the mass of the accreting object M , its radius R, and the transfer rate ˙ M , predicts a typical luminosity of
L = GM R
M ≈ 0.1 ˙ ˙ M c
2, (1.1)
for and object with the size and mass of a NS or a StBH. However, this luminosity has a limit, as when it becomes too high, radiation pressure will blow off the accretion disc, quenching the luminosity of the objects. This is known as the Eddington limit, and it is expressed as
σ
pL
4πcR
2. GM m
pR
2⇒ L
Edd≈ 4πGm
pc σ
pM = 1.26 × 10
38M M
, erg/s (1.2) where σ
pis the proton cross-section and m
pits mass. These objects adhere to this criteria, as most X-ray binaries showcase luminosities of 10
34−10
38erg/s (Tauris and van den Heuvel, 2006), and the extra galactic X-ray populations showcase break around the 10
40erg/s mark in star-forming galaxies (Swartz et al., 2011; Wegner et al., 2000), and around the 5×10
38erg/s mark in LMXB-dominated elliptical galaxies (Kim and Fabbiano, 2004; Kong et al., 2007).
Accounting for the exposed information, the Eddington limit for X-ray binaries is usually set at the L
Edd= 10
39erg/s mark (Kaaret et al., 2017).
There are a few ways to classify X-ray binaries. For instance, if the companion star is now more massive than the accreting remnant, the system is called a High Mass X-ray Binary (HMXB). The contrary case is named a Low Mass X-ray Binary (LMXB). In general, HMXB are stellar wind-driven and more frequently found in the galactic plane, while in LMXB it is the pure Roche Lobe overfill what causes the mass transfer, and they are more commonly associated with the galactic bulge and globular clusters (Tauris and van den Heuvel, 2006;
Casares et al., 2017). NS-powered X-ray binaries are also distinguishable from StBH-powered ones by the presence of coherent X-ray pulsations from the hotspots in the magnetic poles of the rotating star (van den Heuvel, 1993), and cyclotron emission lines in highly magnetized ones (Truemper, 1978).
1.3.2 Galactic Black Hole Binaries
Galactic black hole binaries (GBHB) are a well known kind of sub-Eddington (L
X<
10
39erg/s) X-ray binaries populating the disc and bulge the Milky Way, composed by a
can be modelled with two main components (Remillard and McClintock, 2006):
• A thermal one, parameterized by the temperature of the accretion disc at the Innermost Stable Circular Orbit (ISCO) around the black hole, usually at a few keV.
• And a power law component N (E)dE ∝ E
−ΓdE originating from an optically thick Comptonizing corona of electrons around the disc, with 1.4 < Γ < 2.1 and a cut-off in between 30 keV and 100 keV.
Usually, GBHBs are either found in the thermal state, when the thermal component dom- inates over the power law; or in the hard state, when the contrary is true. Some transient sources spend most of their time in a low luminosity hard state, but transition to the ther- mal state in outbursts that can last a few months, while others switch from one to the other without such harsh contrasts (Remillard and McClintock, 2006), even though in general the thermal state is brighter that the hard state. Additionally, the hard state has a compact radio counterpart show high-frequency quasi periodic oscillations (QPOs) with frequencies of 100-450 Hz in 2:3 resonance, interpreted as ongoing instabilities at the inner edge of the accretion disc (Kaaret et al., 2017; Remillard and McClintock, 2006). The radio emission is thought to originate from streams of relativistic particles, or jets, in a phenomenon that is called a microquasar and that is completely analogous to the quasars (QSOs) formed in the nucleus of active galaxies (AGNs), albeit with much shorter evolution time-scales (Romero et al., 2017).
The importance of paying attention to these objects is that, as it will be shown in the
following section, GBHBs are thought to be a directly neighbouring population to that of
ULXs, sharing many properties with them, but at the same time having very distinctive
traits that set each apart from the other.
1.4 Ultraluminous X-ray sources
1.4.1 Their nature
Ultraluminous X-ray Sources (ULXs) are extragalactic, non-nuclear, point-like, X-ray sources with luminosities close to, or exceeding, that of the Eddington limit of an accreting stellar- mass black hole. Extragalactic means that, while they are always associated with a host galaxy, none of them have been discovered in the Milky Way. Non-nuclear implies that they are not located at the centre of their host galaxy, i.e., they are not AGNs. Point-like means that their X-ray emission is not extended, and it is thought to come from one astronomical object or a very narrow region around it.
Ever since their discovery, they have been a flourishing field of study in X-ray astronomy, as it is well showcased in recent reviews such as Bachetti, 2016 and Kaaret et al., 2017, from where this chapter draws most of its references. ULXs were first observed with the Einstein Observatory in the arms of several spiral galaxies (Fabbiano, 1989). They are, in fact, most commonly associated with star-forming galaxies, whose X-ray luminosity distribution has been reported several times to have a break around the 10
40erg/s mark (Swartz et al., 2011; Wegner et al., 2000), in compassion to elliptical ones where a break is usually found at the 5 × 10
38erg/s mark (Kim and Fabbiano, 2004; Kong et al., 2007). Interestingly enough, elliptical galaxies born from a recent merger event are an exception to this (Kim and Fabbiano, 2010). These facts would indicate that they share some characteristics with the main X-ray binary population.
Their spectra can be fitted with similar components to those of GBHBs, albeit with different values for their parameters, implying some physical differences. In particular, ULXs can be classified in five spectral classes (Sutton et al., 2013; Kaaret et al., 2017):
• Broad Disc ULXs. This spectral class, which dominates in the ULX threshold
(L
X≈ 10
39erg/s), consists in one single broad feature in the spectra. This class is
well described by a StBH at the high end of the mass distribution accreting at the Eddington limit or slightly above it.
• Soft ULXs. These are part of the super-Eddington regime (L
X> 10
39erg/s), and show a dominant thermal component at temperatures of k
BT < 1 keV. This spectral class can showcase a fractional short-term variability of up to 30%, mainly on the hard component of the spectrum.
• Hard ULXs. These also corresponds to the super-Eddington regime, albeit they are usually fainter that the ones from soft class. It has a harder dominating thermal component at typical temperatures of k
BT ≈ 2 keV, and it showcases a much milder short-term variability of . 10%. Several ULXs with long term luminosity variability have been seen to switch between the soft and hard states, suggesting that the same physical objects lie behind them.
• Ultrasoft ULXs. These are soft ULXs with little to no trace of the hard component in their spectrum.
• Upscaled GBHB. These objects are very much like GBHB, but their luminosities far exceed those of the even brightest ULXs, reaching L
X> 10
40erg/s or even L
X>
10
41erg/s and sometimes being called Hyperluminous X-ray Sources (HLXs) in the literature. They are, however, in a sub-Eddington regime. Like GBHBs, they can be described a thermal component and a power law and showcase transitions between the thermal and hard states, but with higher temperatures and QPOs of larger duration.
Sources with these characteristics are the most promising Intermediate Mass Black Hole (IMBH) candidates, but only two of them have been found so far.
The broad disc class is interpreted as the link between GBHB and the super-Eddington
ULX population. It is well described by models of a StBH at the high end of the mass
distribution (M ≈ 15 M ) accreting at the Eddington limit or very slightly over it. This
link is even more apparent when in variable broad disk ULXs, the hard and soft components
of the super-Eddington states make short appearances during periods of high luminosity (Sutton et al., 2013). It is therefore understood that the more luminous hard and soft ULXs states also correspond are also powered by StBHs.
1.4.2 Alternatives to super-Eddington accretion
Many feel uncomfortable with the idea of super-Eddington accretion on to compact objects.
In fact, the X-ray community has explored alternative scenarios. In this section we review two of the most popular ones: radiation beaming and the presence of high mass black holes.
A StBH at the high end of the mass distribution (M ≈ 15 M ) accreting at a sub- Eddington rate, but with a relatively high radiation beaming, could explain an apparent luminosity of L
X> 10
39erg/s without the necessity of super-Eddington accretion. Em- pirical evidence, however, directly contradicts this notion. Some ULXs showcase nebular counterparts both in other regions of the electromagnetic spectrum that can only be ex- plained with more or less isotropic emission or outflows. For instance, ULX Holmberg II X-1, a source 3 Mpc away with a registered emission rate of L
X≈ 10
40erg/s (Zezas et al., 1999), has been observed to be contained within HeII λ4686 ionized region (Kaaret et al., 2004). This region must have been carved by the X-ray emission itself, indicating rather weak beaming for this very luminous source. Another particularly interesting case is that of M82 X-2, a ULX that reaches a luminosity of L
X≈ 1.8 × 10
40erg/s (Kong et al., 2007).
Timing analyses unveiled the presence of coherent pulsation with a period of 1.37 s, revealing that it consists in a NS-powered ULX (Bachetti et al., 2014). The interesting fact about this object is that its pulse variability has a maximum value of 23%, indicating also that not even a NS-powered ULX can produce a strong beaming.
Another alternative explanation that enjoyed more popularity comes from realizing that
according to equation (1.2), a higher mass rises the Eddington limit. Therefore, sub-
Eddington accretion onto IMBHs with masses of M > 100 M ULXs could afford lumi-
nosities of L
X> 10
39erg/s. There are two main problem with this approach. The first one is that theoretical obstacles arise whee trying to explain such a high concentration of IMBHs (Madhusudhan et al., 2006; King, 2004). The second one is that in that case, ULX would be expected to behave like upscaled GBHBs, but as we have shown in the previous section, this is not usually the case. Perhaps the most definitive evidence against IMBHs is that some of the brightest ULXs are powered by a NS (Bachetti et al., 2014; Fürst, 2016;
Israel et al., 2017; Israel et al., 2016; Carpano et al., 2018), being them irrefutable proof for super-Eddington accretion, and indicating that perhaps even more undetected NSs dwell inside of bright ULXs. Nevertheless, there is strong evidence in favour of two particular ULXs being powered by an IMBH.
1.4.3 ULXs as Intermediate Mass Black Holes
The most famous IMBH candidate is ESO 243-49 HLX-1. Located 90 Mpc away, it has been
observed to reach the luminosity of L
X≈ 10
42erg/s. Most importantly, spectral transitions
typical from GBHB have been confirmed from this source, and the black hole mass has been
estimated to be of the order of M ≈ 1, 000 M
from the scaling of the temperature at the
ISCO with black hole mass in the thermal state, while a lower limit of M > 9, 000 M
is
deduced from the scaling of the luminosity (Servillat et al., 2011). However, the origin of this
object is rather peculiar, as it is thought to be the central black hole of a tidally disrupted
dwarf galaxy (Mapelli et al., 2013). Another famous candidate is the 400 M
black hole
M82 X-1. Reaching a luminosity of L
X≈ 10
41erg/s, due to its spectral behaviours being
similar to that of an upscaled GBHB it had been long suspected to host an IMBH, but the
recent measurement of resonant QPO at 3.32 and 5.07 Hz has confirmed for this to be the
case (Pasham et al., 2014).
1.4.4 A model of super-Eddington accretion
It is currently the consensus that ULXs are powered by StBH and NS accreting close to or at super-Eddington rates, with a handful of exceptions where an IMBH is required. In Kaaret et al., 2017 they show a consensus summary of the modelling of the super-Eddington ULX regime, from which we base this section. The current picture is that super-Eddington ULXs consist in HMXBs with very high mass transfer rates. In this situation, the accretion disc becomes thick at the inner regions, taking a shape of a torus that ends at the ISCO.
The hard component originates from the extremely hot region inner face of the torus-shaped disc. The soft component originates from cold, radiation-driven outflows. These winds take of from the inner face of the torus, where the radiation pressure is the highest. Taking the shape of a funnel, they spreads out beyond the torus-shaped disc in turbulent manner.
Unlike the Comptonized corona in GBHBs, these winds are optically thin, driving the power
law component of the spectrum to irrelevance. This geometry allows for super-Eddington
accretion, as photons escape perpendicularly the accretion disc. The perspective of the
observer decides which spectrum is observed. If the line of sight goes along the axis of the
accretion disc, then a hard ULX is seen, but if the inclination is higher then the outflow
winds dominate and the ULX is observed as soft. If the view is parallel to the accretion
disc, then the hard component is completely obscured and the ULX becomes supersoft. In
addition to this picture, we can add from Sutton et al., 2013 that the high variability in the
soft states is explained by the turbulent nature of the outflows, and that an increase of the
accretion rate leads to a narrowing of both the disc and the outflow funnel, explaining why
ULXs become softer when luminosity increases.
Chapter Two
ULX yields from different XMM-Newton data releases.
2.1 Introduction
The project contained in this chapter was originally conceived as the training ground for the methods that we planned to apply to build a catalogue of non-nuclear X-ray sources from the eRASS1 survey. The plan consisted in emulating the work of Earnshaw et al., 2019 (E19), by attempting to rebuild their catalogue of non-nuclear X-ray sources from the same initial sample, the fourth data release of the XMM-Newton survey, 3XMM-DR4
1(Rosen et al., 2016). However, in the end, the project was expanded to include the construction of a preliminary catalogue from the latest data release of the same survey, 4XMM-DR9
2(Webb et al., 2019), which allowed us to perform interesting comparisons, not only between our catalogue and previous works, but also between the two different versions of our catalogues and, by extension, between 3XMM-DR4 and 4XMM-DR9. In this chapter, we expose the methods and results for this particular project. For more definitive results from the construction of new catalogues of non-nuclear sources from 4XMM-DR9 and eRASS1, the reader is recommended to skip ahead to the following chapter.
1
https://xmmssc-www.star.le.ac.uk/Catalogue/3XMM-DR4/UserGuide_xmmcat.html
2
https://heasarc.gsfc.nasa.gov/W3Browse/xmm-newton/xmmssc.html
This introduction is followed by a brief introduction to the XMM-Newton Observatory.
In Section 2.3 we explain the data samples used for the construction of the catalogues. In Section 2.4 we talk about the data pipeline that filters out potential contaminants. In Section 2.5 we present our results and in Section 2.6 we compare them to those of E19. And finally, we present our conclusions in Section 2.7.
2.2 The XMM-Newton Observatory
The XMM-Newton Observatory is a cornerstone mission of the European Space Agency Horizon 2000 programme (Jansen et al., 2001). Launched in 1999, it has been performing query survey observations and it has been cited in more than 6,000 scientific papers
3. It caries three Wolter Type 1 X-ray telescopes and a conventional telescope for the optical/UV spectral range (Mason et al., 2001). The three X-ray telescopes are aligned to each other and contain further CCD detectors at their focal points, called the European Photon Imaging Cameras (EPIC), which are sensible in the energy band of 0.15-12 keV. Two of the cameras are Metal Oxide Semi-conductors CCD arrays (Turner et al., 2001) located behind Reflection Grating Spectrometers (Den Herder et al., 2001) that disperse about half of the incoming photons. The third one, called the pn camera (Strüder et al., 2001), receives an unobstructed beam of photons. On top of that, each camera has four available filters that block potential optical and excessive soft X-ray contamination from the targeted objects. All these elements cooperate to take images with a field of view of 30’, a Full Width at Half Maximum (FWHM, an estimate of visual resolution) of 6", a time-averaged sensitivity of ∼ 10
−14erg/s/cm
2, and a spectral resolution of ∼ 80 eV (ESA: XMM-Newton SOC, 2019).
These stats have allowed for a remarkable yield, and of particular interest for this work are the XMM-Newton serendipitous surveys. As their name indicates, these consist in searches for "accidental" sources appearing in XMM-Newton pixel maps. These are the catalogues
3
https://www.cosmos.esa.int/web/xmm-newton/home
that we have used to build lists of non-nuclear X-ray sources, with the objective of finding ULXs amongst them.
Camera Band FWHM/HEW Field of view Sensitivity
EPIC pn 0.15-12 keV 5"/14" 30’ 10
−14erg/s/cm
2EPIC MOS 0.15-12 keV 6"/15" 30’ 10
−14erg/s/cm
2Table 2.1 XMM-Newton specifications (ESA: XMM-Newton SOC, 2019) of use for this work.
2.3 Data sample.
The most recent data release of the serendipitous surveys corresponds to 4XMM-DR9, a catalogue
4of 810,795 detections of 550,124 independent X-ray sources, both point-like and extended, extracted from 11,204 XMM-Newton observations. For the purposes of this work, the data release of 3XMM-DR4 version of the catalogue has also been used, which contains 531,261 detections of 372.728 sources from 7427 observations. XMM-Newton catalogues.
Both 3XMM-DR4 and 4XMM-DR9 consist on lists of individual detections. In other words, the word source refers to the physical object in the sky, which may have been detected several times. When several detections are associated to a physical source, some of their parameters are averaged to compute the source parameters, which always have the prefix SC_ in front of them. In this work, we look both at detection parameters and source parameters to proceed with our filtering choices.
Just like in E19, we are only interested in point-like sources. Therefore, we excluded from the X-ray catalogues any source with an extension larger than the FWHM resolution of the telescope
5(we filter through SC_EXTENT < 6 arcsec in XMM catalogues). Furthermore, we also choose sources with a detection likelihood larger than 8 (SC_DET_ML > 8).
4
http://xmmssc.irap.omp.eu/
5
https://xmm-tools.cosmos.esa.int/external/xmm_user_support/documentation/uhb/xmmcomp.html
As for the list of galaxies, in E19 the Third Reference Catalogue of Bright Galaxies (RC3, De Vaucouleurs et al., 1991; Corwin et al., 1994) and the Catalogue of Neighbouring Galaxies (CNG, Karachentsev et al., 2004), which contain information on the angular dimensions of galaxies as well as their central position, are used and then correlated with the NASA/IPAC Extragalactic Database (NED
6) and the NED Distance Database (NED-D
7) to update their positions and distances. However, we found this method to be inefficient and unreliable, as the NED cone-search matching engine has proven to be slow and it did not always give the correct match from of coordinates listed in RC3 and CNG, due to the great number of listed objects, especially of high redshift galaxies, and the difficulty for filtering them. Many times this would lead to wrong distances and central positions for the galaxies, which implied wrong calculated luminosity and nuclear sources that were not identified as such, leading to ridiculous source parameters for stellar mass objects (L
X> 10
44erg/s in some cases). These mistakes were only possible to find upon manual inspection after running the pipeline. For this reason, we choose to use the Heraklion Extragalactic CATaloguE (HECATE) instead, which already comes with an accurate list of central position, dimensions and distances for 204,733 galaxies (Kovlakas et al., 2020).
From now on, we refer to the 3XMM-DR4-HECATE correlated data set as myDR4.
Likewise for 4XMM-DR9-HECATE, as myDR9.
2.4 The filtering process.
We have used the TOPCAT
8(Taylor, 2005) software and the STILTS
9commands as tools to manipulate the .fits tables containing the catalogues. The bash scripts written for this problem will be made available along with this report. Throughout this filtering process
6
http://nedwww.ipac.caltech.edu/
7
http://ned.ipac.caltech.edu/Library/Distances/
8
http://www.star.bris.ac.uk/mbt/topcat/
9
http://www.star.bris.ac.uk/~mbt/stilts/
we emulate the work of E19, so many of our steps are also present there. This filtering process has been applied both with 3XMM-DR4 data and 4XMM-DR9 data, creating myDR4 and myDR9 with the same pipeline. As we follow with this description, the reader is referred to Table 2.2, which shows the number of remaining objects after every step.
2.4.1 Matching detections with galaxies.
Galaxies-observations correlation. We start by correlating HECATE with the list of XMM-Newton observations within a radius of 15’, which corresponds to the telescope’s radius of view. From all the successfully correlated galaxies, the ones with isophotal semi- major axis larger than 15’ are removed, as they will correspond to nearby objects with an already well studied X-ray population. This results in galaxies from the local group, such as the Magellanic Clouds, M31 or M33, amongst others, being excluded from the study.
Detections-galaxies correlation. The next step is to add all detections of sources that fall inside of the isophotal ellipse of a galaxy within a 1-sigma certainty, i.e, whose position overlaps with the isophotal ellipse within positional uncertainty. If the position angle of the isophotal ellipse is known, the detection is branded with the flag MATCH_FLAG = 0.
Otherwise, if the detection is correlated with the minor isophotal circle it is branded as
MATCH_FLAG = 1, while if it is is just inside of the major isophotal circle but not the
minor one, it is branded as MATCH_FLAG = 2. This is a bit different from E19, where only
the inner circle is considered for galaxies with unknown position angle. This way, we build
a more comprehensive including sources that otherwise would have been excluded despite
belonging to the parent galaxy, while allowing the user to be conservative with the catalogue
usage according to their criteria.
2.4.2 Removing known contaminants, phase 1
Many of our found X-ray sources correspond to either Active AGNs or foreground stars in the Milky way. It is essential to remove these objects if we want to make a catalogue of non-nuclear sources with a particular focus on ULXs.
Central sources removal. The first objects we worry about are the central sources of parent galaxies, which constitute AGNs and are therefore not of our interest. As XMM- Newton has a FWHM resolution of 6", we decide to remove any source whose position overlaps with a circle of radius 3" around the central position of the parent galaxy within 3-sigma confidence, i.e., three times its positional uncertainty.
QSOs removal. The following objects we take care about are background Quasi-Stellar Objects (QSOs), which also constitute AGN sources with strong X-ray luminosity. With this purpose in mind, we correlate the Véron-Cetty & Véron catalogue of 168,940 known QSOs Véron-Cetty and Véron, 2010, and eliminate all sources which an angular distance of 10" from a known QSO. This general radius is chosen for two reasons: firstly, no positional uncertainties are listed in the QSOs catalogue, and therefore we are not able to perform a personalized match for every object; and secondly, it is of the order of magnitude of the XMM-Newton EPIC camera FWHM plus the typical source positional uncertainty.
Stars removal. To eliminate contaminating foreground stars, we repeat the exercise
with the remaining sources and the Tycho2 catalogue, which contains 2,539,913 known bright
stars (Høg et al., 2000). This catalogue is used both in E19 and in this work as the listed
stars are bright enough to have a significant X-ray emission. However, for the treatment of
data in Chapter 3, we use instead the most recent Gaia data release (Gaia Collaboration
et al., 2018) and compute upper limits for the X-ray luminosity of matched stars to decide
whether it corresponds or not to the source.
2.4.3 Removal of high flag sources
Both 3XMM-DR4 and 4XMM-DR9 assign a SUM_FLAG to every detection, which is a summary of the EPIC camera flags
10. In short, SUM_FLAG can take the following values:
0 if no flags were activated for the detection, 1 if the detection parameters may be affected, 2 if the detection is possibly spurious, 3 if the detection is in an area of the field where spurious detections may occur, and 4 if both 2 and 3 are raised.
Additionally, the SC_SUM_FLAG flag indicates the highest SUM_FLAG amongst the detections assigned to a source. We decide to exclude all detections with SUM_FLAG > 1, but the user should be wary of the source parameters for sources with SC_SUM_FLAG > 1.
This is a bit different from E19, where they eliminate all detections of the source instead.
2.4.4 Removal of more known contaminants, phase 2
As a last sanity check, we matched the results from the previous filtering with the NED
and SIMBAD (Wegner et al., 2000). These last steps are not included inside of the STILTS
pipeline but have been instead performed at a later stage. In particular, the SIMBAD
match was implemented through TOPCAT using the cross-match service provided by CDS,
Strasbourg. Just like in previous steps, we eliminated from our catalogue all sources that
within a search radius of 10" were matched with a QSO, but also with supernovae, as they
are too strong X-ray sources. It is left out to filter many other objects such as fainter stars,
cataclysmic variables and Cepheid stars. With this, we are left with the final catalogues
whose contents are described in the next section. These steps are called NED contaminants
removal and SIMBAD contaminants removal in Table 2.2.
Figure 2.1 XMM-Newton observation 0762610401, including galaxy NGC2631 with all detections highlighted with red crosses (left) and zoomed-in Pan-STARSS1 (Chambers et al., 2016) optical image of galaxy of the same galaxy (right). The de- tected sources are highlighted with red crosses, while a green cross point towards the galactic center. This example makes it clear why is it necessary to remove central sources, and shows how faint can non-nuclear X-ray sources be in visible light.
2.5 Results summary
We calculate the luminosity of all accepted detections from the 8-EPIC camera flux and the distance of the parent galaxy, taking into account the uncertainty of both. These values are listed in our catalogue as Luminosity and LuminosityErr. XMM-Newton catalogues also contain the average flux parameters of all detections associated with a source, which we analogously use to compute the source luminosity and its uncertainty, SC_Luminosity and SC_LuminosityErr. Thereafter, we define as ULX any source with at least one detection that holds Luminosity + LuminosityErr > 10
39erg/s, in order to include any source who has entered the ULX luminosity regime at some point. Analogously, we define as a bright ULX any source with at least one detection that holds Luminosity + LuminosityErr >
5 × 10
40erg/s. These latter ones are of great importance since they are candidates for
10
https://xmmssc-www.star.le.ac.uk/Catalogue/3XMM-DR4/col_flags.html#sumflag
Figure 2.2 Luminosity distribution (without uncertainties) of the entire X-ray source sample found in myDR4 (left) and myDR9 (right), plus a break down of the population according to morphological galaxy types. The yellow regions signal the ULX (L
X> 10
39erg/s) and bright ULX (L
X> 5 ∗ 10
40erg/s) regimes.
hosting exotic physics or even hosting an IMBH (Kaaret et al., 2017). In Table 2.3 we break down the contents of the catalogues according to this classification.
In E19, 2139 detections of 1314 sources found in 305 galaxies are listed. 384 of these are ULX candidates. In myDR4 there are contained 3,355 detections of 1,936 sources in 478 galaxies. 545 of these satisfy the ULX luminosity condition. In myDR9 we have found 6,743 detections of 3,103 sources in 714 galaxies. 952 of these satisfy the ULX luminosity condition.
This represents an increase of 47% and 136% from E19 if we account for the total sample of sources, and an increase of 36% and 137% if we account for the ULX sample. In Fig. 2.2, the luminosity distributions for the entire source sample from myDR4 and myDR9 are shown, plus a breakdown of the distribution according to the morphological type of galaxies.
A clear tendency of X-ray sources around and above the Eddington limit being discovered more frequently in Spiral and Irregular galaxies is seen, giving foot to associating ULX with star-forming regions. Furthermore, the X-ray population in Elliptical galaxies tends to spread towards lower luminosities. Also, the luminosity peak is just below 10
39erg/sec.
These results agree very well with the ones presented in E19. However, this should be taken
Figure 2.3 Dispersion of the fractions of HECATE and RC3 semi-major and semi- minor axes (left) and dispersion of distance against integrated B magnitude for HECATE and RC3 galaxies appearing in appearing in myDR4 and E19, respec- tively.
with a grain of salt, as our sample has a bias in favour of brighter sources since the farther away are the galaxies the less likely is that XMM-Newton detects low-luminosity objects.
This is corrected in Chapter 3, as complete sub-samples accounting for the camera sensitivity are built for the study of non-nuclear X-ray populations.
2.6 Comparisons between our and older works.
We have seen an increase in the number of detected sources from E19. One could naively expect to find the same objects in myDR4 and E19. Actually, we find some pronounced differences. If we do a simple SRCID cross-match between E19 and myDR4, we recover 1052 sources of the original 1314 from E19, which equates to an 80% of their sample. Furthermore, if we perform a galaxy correlation, myDR4 only recovers 270 out of the 306 galaxies from E19. As we will show below, most of these differences are due to the use of HECATE.
While E19 publishes a list of 1314 ULX candidates in 306 galaxies, we find 1936 sources in
478 galaxies. The pronounced increase in the number of involved galaxies (and consequently
in the number of ULX candidates) is due to the extension of HECATE towards fainter
Figure 2.4 Astrometric maps of the detections found in galaxies NGC 4649 (left, large), NGC 4647 (left, small) and NGC 4552 (right). The isophotal ellipses (inner circle in NGC 4552) from RC3 (green) and HECATE (purple) are also shown. Red dots represent detections present in myDR4, while the blue ones only appear in E19. From the maps it can be appreciated how slight changes in isophotal sizes lead to several sources from E19 being missed in myDR4 and vice versa.
galaxies. Many of these new galaxies, as shown in Fig. 2.3, contain measured magnitudes that reach fainter values than those found in E19, which in some cases also implies more distant galaxies. After a simple cross-check, we have attested that HECATE finds 300 out of the 306 galaxies appearing in E19. With this information, it is certain to say that the HECATE list of galaxies is more complete than the one built in E19 from RC3 and CNG.
To understand how, despite these facts, we only find 80% of sources and 76% of galaxies from E19 in myDR4, we need to state that HECATE uses updated HyperLeda
11galaxy isophotal dimensions which, due to an increase of photometric data and the use of an updated computation method Paturel et al., 1991, have different values from the ones listed in RC3 and CNG, especially for the faintest galaxies. Fig. 2.3 shows a great dispersion of changes of galaxy sizes in HECATE with respect to RC3. This point is illustrated by Fig. 2.4, where it is seen how sources appearing in E19 are lost in myDR4 for this reason. As a rough assessment of the validity of this argument, we have computed the amount of lost sky area
11