A Dissertation entitled
High Mass X-ray Binaries in Nearby Star-forming Galaxies
by Blagoy Rangelov
Submitted to the Graduate Faculty as partial fulfillment of the requirements for the Doctor of Philosophy Degree in Physics
Dr. Rupali Chandar, Committee Chair
Dr. John-David Smith, Committee Member
Dr. S. Thomas Megeath, Committee Member
Dr. Scott Lee, Committee Member
Dr. Andrea H. Prestwich, Committee Member
Dr. Patricia R. Komuniecki, Dean College of Graduate Studies
The University of Toledo August 2012 Copyright 2012, Blagoy Rangelov
This document is copyrighted material. Under copyright law, no parts of this document may be reproduced without the expressed permission of the author. An Abstract of High Mass X-ray Binaries in Nearby Star-forming Galaxies by Blagoy Rangelov
Submitted to the Graduate Faculty as partial fulfillment of the requirements for the Doctor of Philosophy Degree in Physics The University of Toledo August 2012
High Mass X-ray Binaries (HMXBs), in which a compact object, either black hole or neutron star, is accreting material from a young, massive donor star, often dominate the high-energy emission from nearby star-forming galaxies. These high mass pairs are believed to form in star clusters, where most massive star formation takes place, but to become displaced from their parent clusters either because they are dynamically ejected or because their parent cluster has dissolved. We have con- ducted a systematic study of the formation and evolution of bright HMXBs in eight nearby galaxies, by detecting HMXBs from their X-ray emission in Chandra X-ray
Observatory observations, and identifying their parent clusters and donor stars in optical observations taken with the Hubble Space Telescope. We use the X-ray and optical properties of these systems to determine the ages of the binaries, whether the compact objects are black holes or neutron stars, and to constrain the masses of the donor stars.
iii This dissertation is dedicated to Iva, whose unwavering support and inspiration have carried me through this journey. Acknowledgments
This dissertation could not have been written without my advisor, Prof. Rupali
Chandar, who guided me through my academic program at the University of Toledo.
She never accepted less than my best efforts. Thank you.
I also wish to express my gratitude to Dr. Andrea Prestwich for her tremendously helpful and invaluable assistance and support. I would also like to thank my other Ph.D. committee members: Prof. JD Smith, Prof. Tom Megeath, and Prof. Scott
Lee.
v Contents
Abstract iii
Acknowledgments v
Contents vi
List of Tables ix
List of Figures x
1 Introduction 1 1.1 X-rayBinaries...... 3
1.2 High Mass X-ray Binaries in Nearby Galaxies ...... 6
1.3 HighMassX-rayBinariesandStarClusters ...... 8
1.4 TwoofNASA’sGreatestObservatories ...... 10
1.4.1 The Hubble Space Telescope ...... 10
1.4.2 The Chandra X-ray Observatory ...... 12 1.5 DissertationOutline ...... 12
2 The Connection Between X-ray Binaries and Star Clusters in NGC
4449 14
2.1 Introduction...... 15 2.2 X-ray Observations from Chandra ...... 16
2.2.1 DataandReduction ...... 18
vi 2.2.2 SourceDetectionandX-rayProperties ...... 18
2.2.3 X-rayColorsandModels...... 22
2.3 Optical Observations from HST ...... 25
2.3.1 DataandPhotometry ...... 25
2.3.2 ColorMagnitudeDiagramofDonorStars ...... 27 2.3.3 Cluster Selection ...... 29
2.3.4 ClusterAge,MassandSizeEstimates...... 29
2.4 Formation and Disruption of the Clusters ...... 33
2.5 Spatial Correlation Between X-ray and Optical Sources ...... 37
2.5.1 Correlation Between the Positions of XRBs and Star Clusters 37
2.5.2 Optical Sources That are Coincident With HMXBs ...... 40 2.6 Discussion...... 41
2.6.1 PropertiesofStarClustersClosesttoHMXBs ...... 41
2.6.2 A Population of Very Young, Massive Black Hole Binaries in
NGC4449...... 43
2.6.3 Processes Responsible for the Spatial Displacement Between
BHBsandStarClusters ...... 45 2.6.4 The Nature of Older X-ray Binaries in NGC 4449 ...... 47
2.7 SummaryandConclusions ...... 49
3 X-ray Binaries and Star Clusters in the Antennae: Optical Cluster Counterparts 52
3.1 Introduction...... 53
3.2 DataandSourceCatalogs ...... 55
3.2.1 X-rayObservationsandCatalogofXRBs...... 55
3.2.2 Optical Observations and Catalog of Star Clusters ...... 59
3.3 Astrometric Matching of the X-ray and Optical Catalogs ...... 60
vii 3.4 OpticalStarClusterCounterpartstoXRBs ...... 61
3.5 Discussion...... 69
3.5.1 Constraining the Nature of XRBs within Star Clusters . . . . 69
3.5.2 The Nature of HMXBs Associated with Star Clusters in the
Antennae ...... 70 3.6 SummaryandConclusions ...... 72
4 High Mass X-ray Binaries in Nearby Starburst Galaxies 74
4.1 Introduction...... 74 4.2 X-ray Observations from Chandra ...... 77
4.3 Identification ofOpticalCounterpartstoXRBs ...... 78
4.4 Results...... 84
4.5 Discussion...... 88
4.5.1 The Nature of the Ultra Luminous X-ray Source in Henize 2-10 88
4.5.2 X-rayColorsofXRBs ...... 88 4.5.3 Correlation Between the Positions of XRBs and Star Clusters 90
4.5.4 CandidateDonorStars...... 93
4.6 SummaryandConclusions ...... 93
5 Summary and Future Prospects 95
References 98
viii List of Tables
2.1 Chandra Observations ...... 20
2.2 X-raysourcecatalogue ...... 21
2.3 Model parameters for X-ray sources with > 50counts...... 23
2.4 HST Images ...... 26
2.5 Starclustercatalogue...... 30 2.6 Properties of Star Cluster Closest to XRBs in NGC 4449 ...... 41
3.1 OpticalcounterpartstoX-raysources...... 63
4.1 SampleGalaxies ...... 77
4.2 Chandra Observations ...... 78
4.3 CatalogueoftheXRBCandidates...... 79
4.4 CloseststarclusterstoXRBs ...... 81 4.5 DonorstarsofXRBs ...... 83
ix List of Figures
1-1 Artist’sconceptionofanX-raybinary ...... 2
1-2 Accretionontoacompactobject ...... 4
1-3 Canonical spectral states of BHX-ray binaries ...... 7
1-4 OpticalandX-rayimagesofM101 ...... 9
1-5 PredictednumberofHMXBs ...... 11
2-1 ColorimageofNGC4449 ...... 17
2-2 Locations of star cluster and X-ray sources on the V band HST image . 19
2-3 X-ray color-color diagram of X-ray binaries in NGC 4449 ...... 24
2-4 Color magnitude diagram of optical point sources ...... 28
2-5 Optical two-color diagrams of all star clusters ...... 31
2-6 Mass-agediagramofallclusters ...... 34 2-7 Age distribution of clusters in NGC 4449 ...... 35
2-8 Cumulative distribution of the displacement between X-ray binaries and
starclusters...... 38
2-9 Same as Figure 2-8, but excluding all BHBs and the known LMXB . . . 50
3-1 OpticalimageoftheAntennaegalaxies ...... 57
3-2 The locations of the 22 XRBs that are coincident with a star cluster... 58 3-3 A2.5′′ × 2.5′′ region around each of the 22 coincident sources ...... 66
4-1 HST imagesofthesixstarburstgalaxies ...... 76
4-2 X-ray color-color diagram of X-ray binaries in nearby starburst galaxies . 85
x 4-3 Cumulative distribution of the displacement between X-ray binaries and
star clusters in six starburst and Antennae galaxies ...... 86
4-4 Color magnitude diagram of optical point sources in six starburst galaxies 87
4-5 Retained and ejected from star clusters BH and NS ...... 91
xi Chapter 1
Introduction
A major goal of astronomy in the last century has been to understand the broad sweep of stellar evolution. How do stars form and evolve? Our understanding of stellar evolution has come a long way from Hertzsprung-Russell diagrams to detailed population synthesis models. We now know that molecular clouds form thousands of protostars which then spend the majority of their lives on the main sequence before evolving into giants or supergiants, and after spectacular deaths, end their lives as compact objects − white dwarfs (WDs), neutron stars (NSs), or black holes (BHs). White dwarfs are the final evolutionary state of all stars whose initial masses do not exceed ∼ 8 M⊙. The remnants of main sequence stars with higher mass are NSs or BHs, depending on how much mass is left in their cores at the end of their main sequence (nuclear burning) lifetime (Heger et al. 2003). White dwarfs are compact
< objects with masses ∼ 1.4 M⊙, the so-called Chandrasekhar limit (Chandrasekhar 1931). In stars that have core masses that exceed this limit at the end of their main sequence lives, the atoms are crushed, and the electrons and the protons fuse together to form an object composed almost entirely of neutrons − neutron stars. If the mass of the stellar remnant is more than ∼ 3 M⊙, neutron degeneracy pressure cannot halt the gravitational collapse, and the object will form a black hole. In this dissertation we will focus on NSs and BHs.
1 Figure 1-1 Artist’s conception of an X-ray binary. Compact object is accreting mass from companion star. The massive disk formed around the compact object is heated to such high temperatures that X-rays are emitted. (Credit: NASA/CXC).
Over the last several decades, observing the universe at different wavelengths led to the discovery of previously unknown classes of objects. X-ray binaries (XRBs), for example, were an important and unexpected discovery. Shortly after the launch of the Uhuru satellite, the discovery of two regularly pulsating sources fundamentally changed our understanding of bright X-ray stars (Giacconi et al. 1971). The observa- tions revealed that both sources showed Doppler shifts and X-ray eclipses (Schreier et al. 1972), which led to the conclusion that these X-ray emitters form a close binary. Their short orbital periods indicated that the two companions were close enough to interact and exchange mass. These systems are extremely X-ray luminous (for ex-
38 −1 ample, Sco X-1 has LX ≈ 10 erg s , Farinelli et al. (2008)), and such high X-ray emission cannot be explained by typical stellar atmospheric processes alone. It was recognized immediately that a different mechanism is responsible for the production of the X-rays, such as an accretion disk around a compact object, and that these
2 binaries contain a compact object and a main sequence or evolved star in a tight
orbit (Figure 1-1).
We now classify X-ray binaries into two broad categories: 1) low mass X-ray binaries (LMXBs), in which material is being stripped from a low-mass star, and
2) high mass X-ray binaries (HMXBs), in which a compact object accretes mass
> from a high-mass star ( ∼ 8M⊙). LMXBs are found in older stellar systems such as globular clusters, halos and bulges, whereas HMXBs are younger and typically found in galactic disks and other sites of recent star-formation. This dissertation focusses on HMXBs, and hereafter (unless otherwise stated) by XRB we mean HMXB.
The goal of this thesis is to constrain the types of luminous HMXBs found in eight nearby star-forming galaxies, including seven dwarf starbursts and the merging Antennae. Properties of HMXBs provide clues to the mass spectrum and number density of compact objects, and on the formation and evolution of massive pairs of stars. Our primary goals are to: 1) identify X-ray sources in observations taken with the Chandra X-ray Observatory, 2) determine the physical properties (e.g., ages and masses) of all the optical counterparts (donor stars and star clusters) to these X-ray sources using data from the Hubble Space Telescope (HST ). These results will allow us to determine the nature of the XRB populations in these galaxies, and to constrain
the age and type of individual donor stars and compact objects.
1.1 X-ray Binaries
HMXBs (Figure 1-1) are binaries where X-ray emission is produced as material is accreted from the young “donor” star onto the compact object (NS or BH). Such emission is usually wind-driven, that is, the strong stellar wind ejects enough material that it creates an accretion disk around the compact object (Tauris & van den Heuvel 2006). The accreted gas gets heated to high temperatures as it travels deep into the
3 Figure 1-2 Diagram illustrating wind-fed accretion. The stellar wind shed by a mas- sive O or B star creates an accretion disk surrounding a compact object (Credit: tt.astro.su.se).
gravitational potential well of the compact object. While colliding winds of a close
pair of massive stars can also produce X-rays, the emission tends to be less luminous
on average than the HMXBs we will focus on. HMXBs can be divided into two general categories based on the type of donor star: 1) a Be star (Be/X-ray binary), or 2) a supergiant star (SG/X-ray binary). Be stars rotate very rapidly, close to the breakup speed, and as a result matter escapes the star and forms an equatorial disk. An orbiting compact object “diving” through the disk of such a star could significantly increase the amount of X-ray emission.
Supergiant stars lose a substantial amount of mass through powerful winds. X-rays are produced as the compact object plunges through the stellar wind at some point during its orbit. Supergiants can lose mass erratically (e.g. outflows can contain large clumps of material) that can result in rapid changes of the X-ray emission.
4 When matter is transferred onto the compact object a large amount of gravita-
tional energy is released in the form of X-ray emission. If accretion takes place with
a mass transfer rate M˙ onto a compact object (NS or BH), which has a mass Mx and a radius R, the resulting steady state release of gravitational potential energy
(also called “accretion-luminosity”) will be Lacc = GMxM/R˙ , where G is Newton’s gravitational constant. Above the limiting (or “Eddington”) luminosity LEdd a star of mass M has such high radiation pressure that its outer layers are ejected. Thus,
LEdd is defined as the luminosity at which the radiation force balances the gravita-
38 tional force exerted by the star, which can be simplified to LEdd ≈ 1.3×10 (M/M⊙). The majority of the HMXBs in our Galaxy and the Magellanic Clouds have compan- ions of early spectral type (O or B) that lose mass in the form of a stellar wind
3 −1 −5 −1 (vwind ≈ 10 km s ), with high mass loss rates ∼ 10 M⊙ yr . A compact object moving through a medium will gravitationally capture some of this material (Fig- ure 1-2, Bondi & Hoyle (1944)).
Using the equation in the previous paragraph, we calculate that a NS with mass
38 −1 1.4 M⊙ cannot have LEdd exceeding ∼ 2 × 10 erg s , assuming steady state accre- tion. Under these assumptions we can calculate the observed X-ray luminosity of a XRB and crudely constrain the mass of its compact object. However, the most accu- rate way to estimate masses of the compact objects is through X-ray eclipses (which occur when the orbital plane of the binary is along the line of sight to the observer), using standard light curve analysis (e.g., Pietsch et al. (2004)). For pulsars in binary system, X-ray timing analysis, where spin period changes due to Doppler shifts can be detected, have also yielded orbital parameters (Townsend et al. 2011). The evolution of two high mass stars in a close binary can be a complex process.
In such systems, mass transfer from one star to the other is common. Other physical processes, such as tidal synchronization, mass loss, orbital circularization, a common envelop phase etc., play a major role in the evolution of these close binaries. Theoret-
5 ical simulations, which include these processes, have been used to follow the evolution
of isolated binaries (e.g., StarTrack; Belczynski et al. (2002)). These simulations pre-
dict an onset of X-ray emission, X-ray luminosities over time, and have been used
to model a number of properties, including populations of XRBs, albeit with a large
number of assumptions (Belczynski & Ziolkowski 2009). BH-XRBs go through different accretion states (Ferreira et al. 2006; Remillard &
McClintock 2006), which affect their X-ray luminosity and spectra. Figure 1-3 shows the canonical spectral states of BH X-ray binaries. The quiescent state is characterized
−9 by a very low accretion rate (M˙ ∼ 10 M⊙) with a hard X-ray component. HMXBs in this state have weak X-ray emission and will not be detected in most extragalactic studies. In the Hard state, the jet emitting disc (JED) may become optically thick, due to the higher M˙ . The JED’s intrinsic emission is weak when compared with the
outer disc. During the Soft state the JED disappears entirely. The luminous disc
has strong radial temperature gradient and the spectrum of the disc as a whole can
be modeled assuming multi-temperature disc blackbody spectra. This type of model
is commonly compared with X-ray observations of XRBs. The Intermediate state is
generally observed during the transition between the Hard and Soft states.
1.2 High Mass X-ray Binaries in Nearby Galaxies
The Einstein X-ray Observatory (Giacconi et al. 1979), the first imaging X-ray
telescope, opened up the systematic study of the X-ray emission from nearby galaxies.
Einstein images, which have a resolution of ∼ 45′′, showed extended and complex X-
ray emission, and gave the first clear detection of individual luminous X-ray sources
in galaxies just beyond the Milky Way (Fabbiano 2006).
More recently, Chandra images of star-forming galaxies beyond the local Group revealed a multitude of bright point sources (see Figure 1-4), most of which are now
6 Figure 1-3 The canonical spectral states of BH X-ray binaries. a) Quiescent state: the JED occupies a large zone in the accretion disc. b) Hard state: the pair beam creation threshold is still not reached. c) Soft state: there is no zone anymore within the disc where an equipartition field is present. No JED, hence neither magnetohy- drodynamics (MHD) jet nor pair beam. d) Luminous Intermediate state between the Hard and the Soft states: the high disc luminosity combined with the presence of a MHD jet allows pair creation and acceleration along the axis, giving birth to flares and superluminal ejection events. (Credit: Ferreira et al. (2006)).
7 generally accepted as HMXBs produced during recent star formation (Fabbiano 1989).
The young stellar ages within starbursts suggest that the associated X-ray sources are likely to be young objects. We know these sources are short-lived, because they
result from the evolution of a massive binary system where the more massive star has
undergone a supernova event. They are thus good tracers of recent star formation,
and the number of HMXBs is likely to be related to the host galaxies star formation
rate (SFR). Hence, nearby star-forming galaxies offer a unique environment to study
the population of young ( ∼< 100 Myr) HMXB. Most XRBs in the Galaxy and Magellanic Clouds are NS binaries. A study by
Belczynski & Ziolkowski (2009) shows that neutron star XRBs significantly outnum- ber black hole XRBs in our Galaxy by a factor of ∼ 30. These X-ray emitting binaries are found primarily in the field regions of galaxies, and not in more dense clusters and/or associations of stars. The fraction of BH vs. NS XRBs is poorly understood in galaxies beyond the Magellanic Clouds. Mineo et al. (2012) identified over one thousand X-ray sources in 29 galaxies located within 40 Mpc, using high quality X-ray data available in the Chandra archive.
Mineo et al. (2012) concluded that the majority of the point sources in these galaxies are HMXBs, powered by the accretion of matter from a massive donor star in a binary system. The authors also showed that the shape of the X-ray luminosity distribution of HMXBs is similar in different galaxies, and that their collective luminosity and number scale with the total star-formation rate of the host galaxy.
1.3 High Mass X-ray Binaries and Star Clusters
Fall et al. (2005) found that at least 20%, and possibly 100% of the massive stars in the Antennae galaxies form in compact star clusters (with most of the star-
−1 formation occuring in compact star clusters at a rate of ≈ 20 − 30 M⊙ yr ). The
8 Figure 1-4 Left: Color BVI image of the M101 spiral galaxy taken with HST. The bright blue clumps are regions where new stars have formed. The yellowish core consists mainly of old stars. The dark brown dust lanes are colder and denser re- gions where interstellar clouds may collapse to form new stars. All of these features are shaped into a beautiful spiral pattern by a combination of gravity and rotation. Right: Chandra’s image of M101, taken in X-ray light, shows the high-energy fea- tures of this spiral galaxy. Populations of point-like sources are easily detected above a generally cooler diffuse emission from the hot interstellar medium. Note that lu- minous X-ray sources are relatively sparse by comparison with the underlying stellar population, and can be detected individually with the Chandra sub-arcsecond reso- lution (Credit: NASA/ESA/CXC) results are similar for many other star-forming galaxies. Studies show that most O- stars in the Galaxy are also found in clusters and associations., and only ≈ 4% found in the field (de Wit et al. 2004, 2005). Given the strong constraints on the (high) fraction of massive stars formed in clusters, we consider it highly likely that HMXBs in star-forming galaxies form within star clusters. Kaaret et al. (2004) previously suggested that XRBs may have formed in young star clusters. Studying three starburst galaxies, they found that HMXBs are prefer- entially located near star clusters, albeit with a significant displacement (200 pc on average). Because star clusters are very good tracers of the star formation activity in galaxies, this suggests that the X-ray sources are young objects formed with the
9 recent star formation. Kaaret et al. (2004) also found that there is an absence of
bright X-ray sources with large displacements from clusters, suggesting that there might be a correlation between the maximum possible brightness of an X-ray source
and its motion.
To understand how far an HMXB can travel from its parent cluster, we have to know the XRB’s age. A close young binary has a chance to go into an HMXB phase only after the initially more massive star ends its life as a compact object (BH or NS).
The donor star in an HMXB is also a massive star, and therefore there is a limited amount of time when material can be accreted onto the compact object. This puts an upper limit on the predicted age of HMXBs to be ∼ 100 Myr (mass of ∼ 8 M⊙, depending on the exact metallicity).
Linden et al. (2010) present model predictions of the numbers and ages of HMXBs.
They trace the number of XRBs (above a given luminosity) as a function of time since formation for different metallicities (from 0.4 Z⊙ to 1Z⊙). This is presented in Figure 1-5. The models predict a strong production of XRBs at ages of 4 − 5 Myr.
At older ages the number of HMXBs depend strongly on the metallicity. Our study of XRB populations in nearby galaxies can test this result directly.
While HMXBs can be identified from Chandra, X-ray observations alone cannot tell us about the ages and nature of compact objects and donor stars. We need to combine X-ray observations with deep optical observations with high spatial resolu- tion.
1.4 Two of NASA’s Greatest Observatories
1.4.1 The Hubble Space Telescope
Since its launch in 1990, the excellent resolution and sensitivity of the Hubble Space Telescope have led to many new discoveries that have transformed our understand- 10 Figure 1-5 Adapted from Figure 1 in Linden et al. (2010). Predicted number of > 36 −1 HMXBs at Z=(Z⊙, 0.4Z⊙, 0.2Z⊙, 0.05Z⊙, 0.02Z⊙) with LX ∼ 10 erg s (bright, > 39 −1 top) and LX ∼ 10 erg s (ULX, bottom) over the first 20 Myr after a starburst of 6 10 M⊙ (see Linden et al. (2010) for details).
11 ing of the universe. For example, HST has drastically improved our understanding and ability to study individual stars and star clusters in nearby galaxies. HST was designed to be accessible to NASA’s space shuttles, and its instrumentation has been upgraded four times over its 22 yr lifetime. Among the imaging cameras that have been installed over time are the Wide Field Camera 2 (WFPC2; resolution of 0.1′′ for the Wide Fields and 0.05′′ for the Planetary Camera), the Advanced Camera for Sur- veys (ACS; resolution of 0.05′′ for the Wide Field Channel and 0.025′′ for the High
Resolution Camera), and the Wide Field Camera 3 (WFC3; 0.04′′). Observations from all of these cameras are used in this dissertation.
1.4.2 The Chandra X-ray Observatory
The Chandra X-ray Observatory, launch in 1999, has revolutionized the ability to detect and study XRBs in nearby galaxies. Its prime instrument is the Advanced
CCD Imaging Spectrometer (ACIS). ACIS (which includes two sub-arrays − ACIS-
I and ACIS-S) consists of 10 CCD chips and provides images as well as spectral information in the 0.2 − 10 keV range. Chandra’s unmatched resolution of ∼ 0.5′′ makes it possible to resolve individual X-ray sources in nearby galaxies, and allows us to study X-ray binary populations in great detail. Chandra ACIS observations of 8 nearby galaxies are used in this dissertation.
1.5 Dissertation Outline
The rest of this dissertation is organized as follows. In Chapter 2 we discuss the connection between X-ray binaries and star clusters in the nearby starburst galaxy
NGC 4449, which has excellent observations at both X-ray and optical wavelengths.
In Chapter 3 we investigate the properties of optical cluster counterparts to X-ray binaries in the merging Antennae galaxies. Chapter 4 presents the results of our
12 analysis of the X-ray binary populations in six additional nearby starburst galaxies
(He 2-10, NGC 1569, NGC 1705, NGC 3125, NGC 4214, and NGC 5253). A summary and planned future work are given in Chapter 5.
13 Chapter 2
The Connection Between X-ray
Binaries and Star Clusters in NGC 4449∗
We present 23 candidate X-ray binaries with luminosities down to 1.8 × 1036 erg s−1, in the nearby starburst galaxy NGC 4449, from observations totaling 105 ks
taken with the ACIS-S instrument on the Chandra X-ray Observatory. We determine
count rates, luminosities, and colors for each source, and perform spectral fits for
sources with sufficient counts. We also compile a new catalog of 129 compact star
clusters in NGC 4449 from high resolution, multi-band optical images taken with the Hubble Space Telescope, doubling the number of clusters known in this galaxy. The
UBVI,Hα luminosities of each cluster are compared with predictions from stellar
evolution models to estimate their ages and masses. We find strong evidence for a
population of very young massive, black-hole binaries, which comprise nearly 50% of
the detected X-ray binaries in NGC 4449. Approximately a third of these remain
within their parent star clusters, which formed τ ∼< 6 − 8 Myr ago, while others have
∗Blagoy Rangelov, Andrea H. Prestwich, & Rupali Chandar Published in the As- trophysical Journal, 2011, 741, 86
14 likely been ejected from their parent clusters. We also find evidence for a population
of somewhat older X-ray binaries, including both supergiant and Be-binaries, which
appear to be associated with somewhat older τ ≈ 100−400 Myr star clusters, and one
X-ray binary in an ancient (τ ≈ 10 Gyr) globular cluster. Our results suggest that
detailed information on star clusters can significantly improve constraints on X-ray binary populations in star-forming galaxies.
2.1 Introduction
Images of nearby starburst galaxies taken in X-rays with the Chandra X-ray Ob- servatory are spectacular, showing a multitude of bright point sources. It is now
generally accepted that most of these sources are high mass X-ray binaries produced
during recent star formation. HMXBs are binaries where one member of the system is
a compact object, either a black hole or neutron star, and the other is a young, mas- sive star. X-ray emission is produced as material is accreted from the young “donor”
star onto the compact object. HMXBs can be divided into two general categories
based on the type of donor star: 1) a Be star (Be/X-ray binary), or 2) a supergiant
star (SG/X-ray binary).
Previous studies have suggested an intriguing connection between HMXBs and young stellar clusters: many of the former are found close to, but not coincident with, young star clusters. These observations are broadly consistent with a scenario where
X-ray binaries (XRBs) form in star clusters, but have sufficiently large velocities that they are expelled from their parent cluster (Zezas et al. (2002), Kaaret et al. (2004)).
There are three mechanisms that might displace HMXBs from their birth sites. The binary could be given a “kick” during an asymmetric supernova explosion that forms the black hole or neutron star, or it could be ejected via dynamical interactions with other stars in a dense cluster core (McSwain et al. 2007). A third scenario that could
15 explain the observed spatial displacement between HMXBs and young star clusters
is that X-ray binaries formed in clusters that have since dispersed.
In this chapter, we investigate the relationship between HMXBs and star clusters using the best X-ray and optical observations currently available for the starburst galaxy NGC 4449. We analyze observations taken with the Chandra X-ray Observa- tory to select X-ray binaries in NGC 4449, and optical images taken with the Hubble
Space Telescope to detect star clusters and to estimate their ages and masses. An op-
tical color image of NGC 4449 based on the HST images is shown in Figure 2.1. Our
primary goals are to constrain the nature of X-ray binaries and in particular HMXBs,
and to explore the relationship between HMXBs and star clusters in NGC 4449.
The rest of this chapter is organized as follows. In Section 2.2 we present the X-ray observations from Chandra, including data reduction, source detection and
some basic models used for interpretation. In Section 2.3 we present the optical
observations from the HST and a new catalog of compact star clusters, including
their integrated colors and luminosities, and we derive their masses and ages. The
mass and age distributions of the clusters and what they tell us about the formation
and disruption of the clusters are the subject of Section 2.4, while Section 2.5 presents the spatial correlation between the clusters and the candidate HMXBs. Section 2.6
synthesizes these results to constrain the nature of the HMXBs in NGC 4449, and
Section 2.7 summarizes the main results of this work.
2.2 X-ray Observations from Chandra
NGC 4449 is an irregular, star-forming galaxy located at a distance of 3.82 ±
0.27 Mpc (Annibali et al. 2008). With an integrated magnitude of MB = −18.2 it is somewhat more luminous than the Large Magellanic Cloud (Hunter 1997). NGC
−1 4449 has a current star formation rate of ≈ 1.5 M⊙ yr (Thronson et al. 1987) and
16 Figure 2-1 Color image of NGC 4449 produced using the HST /ACS observations described in Section 3. The F435W (≈ B band) image is shown in blue, the F555W (≈ V band) image in green, and the red image combines the F814W (≈ I band) and ′ F658N (Hα) filters. North is up and east is to the left. This image is 5 along the long side. (Credit: NASA, ESA, A. Aloisi (STScI/ESA), and The Hubble Heritage (STScI/AURA)-ESA/Hubble Collaboration)
17 a near-solar present-day gas abundance (12 + log[O/H]=8.83; Grevesse & Sauval
(1998)).
2.2.1 Data and Reduction
We use three sets of archival Chandra observations of NGC 4449, with integration
times of 30 ksec (ObsID: 2031, PI: Heckman), 15 ksec (ObsID: 10125. PI: Long), and
60 ksec (ObsID: 10875, PI: Long), to detect X-ray point sources in NGC 4449. Basic
information for the three sets of observations is given in Table 1. The data were taken
with the Advanced CCD Imaging Spectrometer (ACIS) instrument on the Chandra telescope on February 4, 2001 in “faint” mode (ObsID: 2031), and March 4 and 7,
2010 in “vfaint” mode (ObsIDs: 10125 and 10875). The galaxy was positioned on the
back-illuminated S3 CCD chip. We processed the data using the Chandra Interactive
Analysis of Observations (CIAO) software (version 4.2) and Chandra Calibration Data
Base (CALDB) version 4.3.02, and restricted the data to the energy range between
0.3-8 keV. The observations were filtered in three energy bands – 0.3-1 keV (soft), 1-2 keV (medium) and 2-8 keV (hard).
2.2.2 Source Detection and X-ray Properties
We use CIAO’s Mexican-hat wavelet source detection routine wavdetect (Freeman
et al. 2002) to create source lists. Wavelet scales of 1.4, 2, 4, 8, and 16 pixels and a
detection threshold of 10−6 were used, which typically results in 1 spurious detection
per million pixels. The output sources were examined visually to verify each detection, and to correct the source catalog when multiple detections occurred. A catalog of
X-ray point sources detected in the Chandra observations is presented in Table 2.2, and the locations of these sources are shown in Figure 2-2.
2http://cxc.harvard.edu/ciao/
18 X31
X29
X28
X44 X26 X24 X23 X22
X43
X21 X41 X20
X19 X18
X15 X13 X14
X42 X46
X12
X11 X10
X6
Figure 2-2 V band (F555W filter) image of NGC 4449 taken with the ACS camera on the HST . North is up and east is to the left. The locations of star clusters selected from optical HST images are shown as circles. The locations of X-ray sources selected from Chandra X-ray observations are labeled and shown as crosses.
19 Table 2.1. Chandra Observations
ObsId Date PI Exposure (ks)
2031 2001-02-04 Heckman 30 10125 2009-03-04 Long 15 10875 2009-03-07 Long 60
A number of measurements and estimates are made for each source − total number of counts and the counts measured in the soft, medium, and hard energy bands. We calculate two X-ray colors, a “soft” color defined as H1=(M − S)/T , and a “hard” color defined as H2=(H − M)/T , where S, M and H are the total counts measured in the soft, medium and hard bands (Prestwich et al. 2003), and T is the total number of counts in all three bands. The luminosity of each source is estimated by fitting its spectrum with a power law model with a photon index (the slope of the power law)
Γ = 1.5. These values are compiled in Table 2.2.
There are eleven sources which have sufficient counts (> 50) for a crude spectral fit.
We were able to fit ten of these with simple X-ray spectral models, as summarized in Table 2.3. One source, X15, is in a region with high background from diffuse emission and we were unable to obtain a satisfactory fit. X-ray spectra and responses (including sensitivity of the instrument and CCD) were extracted using standard
CIAO software. Spectra were grouped for a minimum of 15 counts per bin. The fits were performed using XSPEC V 12.0 over an energy range 0.2−8.0 keV. For each source, we tried three models: a simple power law (PL), a MEKAL model3, and a
4 20 multi-color disk (MCD ) model. A fixed foreground Galactic column nH =1.5 × 10
3An emission spectrum from hot diffuse gas, e.g. Liedahl et al. (1995) 4A superposition of multi-temperature blackbody spectra expected from optically-thick accretion disk; the model only constrains the temperature of the inner disk Tin, e.g. Mitsuda et al. (1984)
20 cm−2 was assumed in each fit. The fits allow for an additional variable column due
to absorption intrinsic to the source. The best fit model and parameters for each
source are given in Table 2.3. Observed X-ray fluxes and estimated luminosities were calculated for the 0.3 − 8 keV energy range (both uncorrected for absorption) using the best fit parameters in the table.
Table 2.2: X-ray source catalogue
1 ID RA DEC Soft Hard LX Color Color (erg s−1) X8 187.00308 44.07569 0.20 0.00 4.55E+36 X14 187.00572 44.09145 0.15 -0.50 9.41E+36 X18 187.01634 44.09551 0.01 -0.09 7.68E+37 X13 187.02855 44.09120 0.31 -0.42 9.21E+36 X6 187.02988 44.07074 0.34 0.14 1.47E+37 X29 187.03043 44.12264 0.04 -0.07 3.34E+37 X10 187.03065 44.08164 -0.33 -0.16 5.67E+37 X12 187.03891 44.08566 -0.06 -0.32 3.01E+38 X20 187.04055 44.09808 0.23 -0.13 1.01E+38 X11 187.04576 44.08327 0.34 -0.26 3.05E+36 X23 187.04671 44.11066 -0.24 -0.26 2.25E+37 X15 187.04985 44.09188 -0.44 -0.04 7.66E+37 X24 187.04997 44.11149 0.02 -0.29 5.81E+37 X21 187.05013 44.09955 -0.28 -0.25 1.19E+37 X28 187.05533 44.11557 -0.01 -0.32 6.93E+37 X31 187.06829 44.12765 -0.08 0.41 9.53E+36 X22 187.07433 44.10939 0.41 -0.13 8.59E+38 X19 187.07917 44.09585 -0.90 -0.02 2.01E+37 X41 187.04336 44.09946 -1.00 0.00 1.43E+37 X42 187.06327 44.08829 -0.41 -0.57 1.78E+36 X43 187.08341 44.10589 0.41 0.09 3.36E+36 X44 187.05553 44.11308 0.13 0.09 1.13E+36 X46 187.04041 44.08867 -0.20 -0.23 7.59E+36 X262 187.04569 44.11343 0.11 -0.35 2.57E+38
1X-ray luminosities (0.3 − 8 keV range) are derived by fitting a power law (Γ = 1.5) to the
20 −2 data. An assumed Galactic nH = 1.5 × 10 cm is applied.
2Supernova remnant. 21 2.2.3 X-ray Colors and Models
After examining the resulting X-ray source catalog, we eliminate some sources from further consideration in this work for the following reasons. First, we remove all sources located beyond the optically luminous portion of NGC 4449 (R ≈ 3-4 kpc),
since these are almost certainly background galaxies and not associated with NGC
4449 itself. We also eliminate one source that is a confirmed supernova remnant in
NGC 4449 (Patnaude & Fesen 2003), and another which appears to be a spurious
detection of diffuse emission near the center of the galaxy. Two additional X-ray sources are coincident with foreground stars and were also eliminated. Our final
catalog contains 23 sources that we believe are X-ray binaries. These are listed in
Table 2.2.
The observed H1 vs. H2 colors of the X-ray binaries are shown in Figure 2-3. Two
models are shown for reference. The red line shows predictions from disk blackbody
models with temperatures ranging from 0.1 − 1.0 keV. Black hole binaries typically have disk blackbody temperatures of kT ≈ 1.0 keV. The orange line shows the effect of adding absorption to the disk blackbody models. The green line represents a power law with increasing photon index from 1.0 − 3.0. Accreting low mass neutron star binaries typically have absorbed power law spectra with photon index ∼ 2.0. Figure
2-3 suggests that NGC 4449 has a mix of X-ray binary populations typical of a galaxy that has had on-going star formation, with some sources better following disk black- body models and others in the absorbed portion of the diagram. The nature of the X-ray binaries will be discussed in more detail in Section 2.6.
22 Table 2.3. Model parameters for X-ray sources with > 50 counts
1 2 2 3 3 ID Net Best Fit nH Photon Tin T χ / d.o.f. Flux LX Counts Model (×1022 cm−2) index (keV) (keV) (erg s−1 cm−2) (erg s−1)
+8.0 −2 X18 141.3 ABS*PL (4.5−4.0)×10 1.5±0.3 − − 2.4 / 7 3.65E-14 6.30E+37 +12.5 −2 +0.6 X29 84.8 ABS*PL (2.2−2.2 )×10 1.26−0.4 − − 0.43 / 3 2.48E-14 4.30E+37 a +40 −3 +0.16 X10 172.7 ABS*(PL+MEKAL) (9.5−9.0)×10 2.1±0.4 − 0.46−0.4 7.0 / 7 2.72E-14 4.70E+37 +4.0 −3 X12 1197.7 ABS*MCD (4.0−8.5)×10 − 0.60±0.5 − 85.9 / 71 1.58E-13 2.76E+38 +0.4 X20 201.4 ABS*PL 0.19±0.1 1.7−0.3 − − 13.72 / 11 4.80E-14 8.35E+37 b +0.2 X23 106.0 ABS*MCD 0.037−0.04 − 0.37±0.13 − 5.28 / 4 1.13E-14 1.95E+37 +0.06 +0.4 X24 186.8 ABS*PL 0.19−0.08 2.4−0.3 − − 7.54 / 10 3.20E-14 5.66E+37 +0.38 X21 58.9 ABS*MCD 0.21−0.38 − 0.23±0.1 − 2.89 / 2 6.00E-15 1.68E+37 a +0.04 X28 309.8 ABS*MCD 0.04−0.03 − 0.64±0.1 − 24.7 / 17 3.98E-15 6.98E+37 +0.16 X22 1411.6 ABS*PL 0.6±0.07 1.9−0.14 − − 86.73 / 85 4.34E-13 7.62E+38 23
2Supernova remnant. 1The inner temperature of the multi-color disk model. 2MEKAL model temperature. 3Observed X-ray fluxes and estimated luminosities for the 0.3 − 8 keV energy range, uncorrected for absorption. aPossible line emission. bExcess flux just over 1 keV. 1.0 increasing hardness
DBB+ABS 0.5
LMXB
0.0 Soft Color (H1) -0.5 PL
DBB -1.0
increasing hardness
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 Hard Color (H2)
Figure 2-3 X-ray color-color diagram of X-ray binaries in NGC 4449. The H1 and H2 colors are defined in Section 2. Candidate high mass black hole binaries are shown as squares, and the rest of the sample is shown as triangles. The asterisks show the X-ray sources that are coincident with clusters. Theoretical tracks are shown as the open symbols connected with solid lines. The green triangles represent a power law (PL) with an increasing photon index from 1.0 to 3.0, and the red circles show predictions for diskblack-body models (DBB) with increasing temperature from 0.1 to 1.0 keV. The orange squares are a disk black-body model with T=0.9 keV and an 20 20 increasing hydrogen column density (DBB+ABS) nH in steps of 0, 1 × 10 ,5 × 10 , 1 × 1021, 5 × 1021, and 1 × 1022 cm−2. See text for details.
24 2.3 Optical Observations from HST
It has been suggested that the dense environments found in compact star clusters
may be very efficient in producing XRBs (e.g., McSwain et al. (2007)). Regardless
of whether or not XRBs form within compact clusters, the clusters are good tracers
of star formation in galaxies, and hence can provide important constraints on the XRB population. The goals of this section are two-fold: (1) select a new catalog of
compact star clusters in NGC 4449, to correlate with the locations of the XRBs, and
(2) measure luminosities and colors for any point source coincident with an XRB and
hence likely to be the donor star (discussed in Sections 2.5.2 and 2.6.4). While a
catalog of clusters in NGC 4449 already exists (Gelatt, Hunter & Gallagher 2001), it
is based on partial imaging taken with the WFPC2 instrument on board HST , and there are now deeper, higher resolution data with full coverage of NGC 4449 available
from the Wide Field Channel (WFC) camera of the Advanced Camera for Surveys
instrument.
2.3.1 Data and Photometry
ACS/WFC imaging of two positions within NGC 4449 was taken in the F435W
(≈ B), F555W (≈ V band), F814W (≈ I band), and the F658N (Hα) filters, on November 10-11, 2005 (Proposal ID GO: 10585, PI: Aloisi). Four individual exposures
were taken in each filter at each pointing. The ACS/WFC has a pixel scale of 0.049′′,
or a projected scale of 18.4 pc per arcsecond at the assumed distance of 3.82 ±
0.27 Mpc (Annibali et al. 2008) to NGC 4449. We downloaded the ACS data from
the Hubble Legacy Archive5 (HLA). The HLA combines the individual flatfielded
exposures for a specific filter together using the PYRAF task Multidrizzle, and outputs geometrically corrected images. For the NGC 4449 observations, the HLA
5http://hla.stsci.edu/
25 Table 2.4. HST Images
HST Instrument Proposal ID Filter Exposure time (s)
WFPC2 6716 F336W 2×520 ACS/WFC 10585 F435W 4×3660 (PosA) 4×3478.91 (PosB) F555W 4×2460 F658N 4×360 F814W 4×2060
used stars from the U.S. Naval Observatory catalogue to astrometrically correct the images.
We also use available F336W (≈ U) band imaging of two pointings within NGC 4449, taken with the WFPC2 camera (Proposal ID GO: 6716, PI: Stecher). Each position has two exposures. The WFPC2 has four CCDs − the Planetary Camera (PC) has a scale of 0.0456′′ pix−1, and the three Wide Field (WF) CCDs have a scale of
0.0996′′ pix−1. Note that there are two sets of images for each WFPC2 observation available in the HLA: a combined WFPC2 image including all four CCDs with the PC resampled to the same resolution as the three WF CCDs, and an image of only the
PC, at its original pixel scale. We have used the combined WFPC2 image when our cluster candidates were in one of the three WFs and WFPC2-PC for objects located in the PC. Details on pointings and exposure times for the HST data used here are given in Table 2.4. We identified ∼200,000 sources in each ACS pointing (in the V band), using the
IRAF6 DAOFIND task. These include star clusters and bright, individual stars in
NGC 4449, as well as some foreground stars and background galaxies. We perform
6IRAF is distributed by the National Optical Astronomy Observatories, which are operated by the Association of Universities for Research in Astronomy, Inc., under cooperative agreement with the National Science Foundation.
26 circular aperture photometry of all detected sources using radii of 1 and 3 pixels
and background annuli of 8 and 13 pixels, using the PHOT task in IRAF. We use
zeropoints on the VEGAMAG system taken from Table 9 in Holtzman et al. (1995)
for the WFPC2/F336W filter, and from Table 10 in Sirianni et al. (2005) for the
ACS observations. Corrections for inefficiency in the charge transfer were determined for the WFPC2/F336W photometry based on the formulae given by Dolphin (2000;
we have used the most recent characterization available from A. Dolphin’s website
http://purcell.as.arizona.edu/wfpc2calib/). We also applied aperture corrections to
extrapolate from our fixed aperture radius to the total magnitude in each filter, based
on the concentration index (CI, the magnitude difference measured between a radius of 1 and 3 pixels) measured for each source. We obtained a linear fit between the mea- sured CI and aperture corrections for ≈20 hand-selected, relatively isolated sources, and used this fit to estimate the aperture corrections for the rest of the sources.
2.3.2 Color Magnitude Diagram of Donor Stars
The colors and luminosities of individual stars give an estimate of their masses and hence a constraint on their ages. In several cases, the optical images reveal a single point source within the positional uncertainty of an XRB, which is likely the donor star. Although a full analysis of the spectral energy distributions for these stars is beyond the scope of this chapter and will be presented in a future work, we can use our photometry to broadly assess their ages. In Figure 2-4, we compare the measured luminosities and colors for these matched optical sources with solar metallicity stellar isochrones with ages of 10 Myr, 100 Myr, and 1 Gyr from the Padova group (Girardi et al. (2008); Marigo et al. (2008)). These provide a rough guide to the ages/types of the candidate donor stars, and suggest that the bright stellar sources associated with
X10 and X24 are supergiant stars, while those associated with X6, X31, and X14 are significantly fainter and hence may be somewhat older.
27 18
20 X 10 X 24 10 Myr
22 X 14
X 6
F555W 100 Myr m 24 X 31
26 1 Gyr
28
0 1 2 3 (mF555W - mF814W)
Figure 2-4 Color magnitude diagram of optical point sources which are unique coun- terparts to X-ray sources. The solid lines are Padova isochrones (for Z =0.02) of 10 Myr, 100 Myr and 1 Gyr (from left to right).
28 2.3.3 Cluster Selection
We measure the sizes of all detected objects using the Ishape software (Larsen
1999). This gives a better measure than CI of the sizes of relatively bright, isolated
star clusters. Ishape convolves analytic profiles with the PSF, and determines the best fit to each source. We assume a KING30 profile, a single mass King (1966) model with a fixed ratio of tidal to core radius of 30 and fit the data within a radius of 5 pixels. For more details on Ishape and estimating the sizes of clusters, see Larsen
(1999).
Most star clusters are slightly broader than the PSF at the distance of NGC 4449.
The biggest challenge in selecting compact star clusters from the entire source list is to separate them from chance blends and superpositions in the crowded star forming regions. We use the following critieria to select star cluster candidates: (1) mV ≤ 22
(MV ≈ −6 at the assumed distance of 3.82 Mpc for NGC 4449); (2) CI > 1.3; (3) FWHM > 0.2 pixels, i.e. at least 0.2 pixels broader than the PSF, as measured by
Ishape; and (4) no “neighbors”, i.e. other object detected within a 5 pixel radius. A
final by-eye inspection was made to throw out some remaining blends (at the ≈ 20 −
30% level). The final catalog contains 129 candidate star clusters, double the number published previously by Gelatt, Hunter & Gallagher (2001). Basic information for the selected clusters, including their locations and measured photometry, is given in
Table 2.5. The cluster locations are shown in Figure 2-2.
2.3.4 Cluster Age, Mass and Size Estimates
Figure 2-5 compares predictions from solar metallicity stellar population models
of Charlot & Bruzual (2007; hereafter CB07, private communication; see also Bruzual
& Charlot (2003)), with colors measured for our final cluster sample. Crosses mark the predicted colors for 106, 107, 108, 109, and 1010 yr clusters, starting from the
29 Table 2.5. Star cluster catalogue
Number RA DEC U-B V-I Age1 log(τ/yr)
1 187.05396 44.083114 -2.106 -0.295 6.40 2 187.05370 44.083248 -2.195 -0.395 6.02 3 187.06067 44.083354 -0.303 1.080 9.30 ···
Note. — This table is published in its entirety in the electronic edition of the Astrophysical Journal. A portion is shown here for guidance regarding its form and content. 1Typical uncertainties are approximately ±0.3 in log τ. upper left. The arrow shows the direction of reddening for a Galactic-type extinction law (Fitzpatrick 1999).
The model predictions match our cluster photometry relatively well. The majority of the clusters have blue integrated colors, indicating that they are fairly young, with ages τ ∼< few × 108 yr. At the top of the diagram there are a few young clusters in crowded regions that fall above the models, likely due to some contamination from neighbors in the lower resolution U band images (we see a similar effect for the colors of clusters in M51; Chandar et al. (2010a)). The few sources with U-B colors significantly redder than the model predictions are somewhat faint in the U band, leading to large uncertainties in this color. The bottom panel of Figure 2-5 shows the B − V vs. V − I two-color diagram, and includes some objects for which no U band photometry was measured, either because they are outside the field of view or because they are too faint in this band. The sources which fall significantly below (redward of) the model predictions in B −V have strong nebular line emission (which is observed in the narrow-band F658N (Hα) filter), moving them off the model track.
30 -2 Av = 0.4
-1 F435W - m
F336W 0 m
1
-0.5 Av = 0.4
0.0
F555W 0.5 - m
F435W 1.0 m
1.5
2.0 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0
mF555W - mF814W
Figure 2-5 Optical two-color diagrams of all star clusters in our sample. The solid line shows predictions from the single stellar population models of CB07 and the crosses show the following predicted ages: log(τ/yr) = 6, 7, 8, 9 and 10, starting from the upper-left. The arrow shows the direction of reddening.
31 While two-color diagrams like those shown in Figure 2-5 are useful for visualizing the evolution of clusters, we actually use a χ2 minimization technique to estimate the ages of the clusters. This means that points outside the models (e.g., the “high” points in Figure 2-5) can be reasonably well fit.
We estimate the age τ and extinction AV for each cluster as we have done in previous works (see e.g., Fall et al. (2005) for details), by performing a least χ2 fit comparing observed magnitudes with the predictions from the CB07 stellar population models with solar metallicity (Z =0.02), which appear to match the measured colors of clusters in NGC 4449 reasonably well. The best fit combination of τ and AV for each
2 2 obs mod 2 cluster returns the minimum χ : χ (τ, AV )= Pλ Wλ(mλ −mλ ) , where the sum runs over all available broad-band (UBVI) filters and the F658N narrow-band filter, but requires a minimum of three measurements (including the V band) to estimate age and extinction. The weights W are related to the photometric uncertainty σλ as
2 2 −1 Wλ = [σλ +(0.05) ] . The F658N filter includes both stellar continuum and nebular line emission, and is dominated by line emission from ionized gas for the youngest
(τ ∼< several × 106 Myr) objects, and by continuum emission from stars for clusters with ages of τ ∼> 107 yr. This enables us to use the narrow-band filter as a fifth data point in many cases, regardless of the age of the cluster.
We tested our method by estimating ages and masses from our χ2 analysis for different assumptions regarding extinction and by comparing our photometry with the predictions from a model with lower metallicity (Z =0.008). In general, we find that the age estimates are quite robust for very young clusters (τ ∼< 6 × 106 yr) that have nebular emission, and for clusters with ages between 3 × 107 yr ∼< τ ∼< 109 yr. Clusters older than ≈ 109 yr suffer from the well-known age-metallicity degeneracy, and we cannot tell if they are approximately a Gyr or older than ≈ 10 Gyr from the data presented here. Eleven clusters in our catalog have integrated colors similar to those of metal-poor globular clusters in our Galaxy (and in many other galaxies) with
32 0.8 ∼< V − I ∼< 1.3 and 0.6 ∼< B − V ∼< 0.9. The dating of clusters with no nebular emission and with ages between ≈ 6 × 106 yr and ≈ 3 × 107 yr is degenerate in some cases, with two nearly equally good combinations of age and extinction. We estimate that typical uncertainties are ≈ 0.3 in log τ for clusters with τ ∼< 109 yr, typical for this method (e.g., Fall et al. (2009)).
The mass of each cluster is estimated from the total V band luminosity, corrected for extinction, and the age-dependent mass-to-light ratios (M/LV ) predicted by the CB07 models.
2.4 Formation and Disruption of the Clusters
The mass and age distributions of a population of star clusters provides important clues to the formation and disruption of the clusters. In Figure 2-6 we show the mass and age estimates for star clusters in NGC 4449. The solid line represents MV = −7, the approximate completeness limit of our sample, and shows that our sample does not contain clusters over the same mass range at all ages, because clusters fade over time.
The mass-age diagram shows some small scale features, such as a sparsely populated region in the range 7.0 ∼< log(τ/yr) ∼< 7.6 yr. This particular feature occurs where the predicted colors loop back on themselves, covering a small region in color space over a relatively long time, and effectively resulting in a gap. This artificial, empty stripe and similar features do not affect our conclusions. Qualitative trends in the distribution of cluster ages and masses are apparent from the mass-age diagram. We see that NGC 4449 has formed relatively massive clusters
> 4 9 (M ∼ 10 M⊙) more or less continuously over the last 10 yr. A number of clusters (approximately half of our sample), formed very recently, in the last τ ∼< 107 yr, and a number of clusters clearly formed ≈ 1−3×108 yr ago. As mentioned in Section 3.3, our sample includes ≈ 11 candidate ancient globular clusters, implying that NGC 4449
33 107
Nuclear Cluster 106 ) . O 105
104 Mass (M/M
103
102 6 7 8 9 10 log (τ/yr)
Figure 2-6 Mass-age diagram of all clusters in our sample. The solid line shows the approximate magnitude limit of MV = −7.0. began forming stars approximately a Hubble time ago. This is broadly consistent with the results of Annibali et al. (2008), who found that stars in NGC 4449 have formed over at least the last ≈ 109 yr, with tentative evidence for earlier star formation as well, based on the colors and luminosities of individual stars. Qualitatively, the overall distribution of cluster ages and masses in NGC 4449 appears to be similar to that in other galaxies with very different environments and star formation histories, such as the merging Antennae galaxies (e.g., Fall et al. (2005), and the more quiescent
Magellanic Clouds (Chandar et al. 2010a), although the gap between 107 − 108 yr is somewhat more prominent. We present a more quantitative analysis below. The left panel on Figure 2-7 shows the age distribution for clusters in NGC 4449 in two different intervals of mass. Note that the data points for each distribution are restricted to a mass-age range that is not affected by incompleteness (i.e., above the solid line), so the observed shapes are caused by the formation and disruption of
34 4 103 log (M/MO .) = 4.3-6.0 log τ < 7
log (M/MO .) = 4.0-4.3 log τ = 8.0-8.4 3 102 ) + const
τ 2
1
dN/dM + const 10 log (dN/d 1
0 6 7 8 9 10 103 104 105 106 log (τ/yr) Mass (M/MO .)
Figure 2-7 The age distribution of clusters in NGC 4449 in the indicated intervals of mass is shown on the left, and the mass function in the indicated intervals of age is shown on the right. The lines show the best fits to the distributions. See text for more details. the clusters and not by their fading out of our sample. Although there are few data points, these distributions appear to decline steeply, and can be approximated by a simple power law, dN/dτ ∝ τ γ , with best fits of γ = −0.91 ± 0.13 for clusters more
4 massive than 2 × 10 M⊙, and γ = −1.21 ± 0.21 for clusters with masses between
4 1−2×10 M⊙. These values of γ are the same within the uncertainties. We find that
γ can change by up to ≈ 0.3 if different bin sizes and centers are used, but the overall decline starting at very young ages does not change. Based on these experiments, we take γ = −1.0 ± 0.3.
The right panel on Figure 2-7 shows the mass function for clusters in NGC 4449 in two different intervals of age. These mass functions can be approximated by a power law, dN/dM ∝ M β, with best fit values of β = −2.25 ± 0.15 for τ < 107 yr, and
β = −2.07 ± 0.19 for τ =1 − 2 × 108 yr, which are the same within the uncertainties.
We take β to be −2.16 ± 0.3, the mean and range of the values above. Because the mass and age distributions appear to be at least approximately inde-
35 pendent of one another, the joint distribution of cluster ages and masses g(M, τ) can be approximated as g(M, τ) ∝ M βτ γ, with β ≈ −2 and γ ≈ −1, for τ ∼< few × 108 yr
> 4 and M ∼ 10 M⊙, similar to the form observed for clusters in the Magellanic Clouds (Chandar et al. 2010a), M83 (Chandar et al. 2010b), and in the Antennae (Whitmore et al. 2007; Fall et al. 2010).
Chandar et al. (2010c) compiled results for the age distribution of star clusters in over a dozen different galaxies, including dwarf irregular, spiral, and merging galaxies.
All of these different galaxies appear to have similar shapes to their (mass-limited) age distributions, where they decline from the present to the past (over the past ≈ few×108 −109 yr). This shape for the age distribution has previously been interpreted as due primarily to the gradual, early disruption rather than to the formation of the clusters, since it is far more likely that the clusters in all of these galaxies have similar disruption histories than it is that they have similar formation histories.
After their formation in the dense cores of giant molecular clouds, different phys- ical processes cause clusters to lose mass and to eventually disrupt. Fall et al. (2009) and Chandar et al. (2010a) have suggested the following approximate sequence and timescale for these processes: (1) removal of ISM by stellar feedback, τ ∼< 107 yr; (2) continued stellar mass loss, 107 yr ∼< τ ∼< 108 yr; and (3) tidal disturbances by passing molecular clouds, τ ∼> 108 yr. It is likely that the cluster mass and age, which can be approximated by power-law distributions, result from a complex situation that involves several of these disruption processes. We refer the interested reader to Fall et al. (2009) and Chandar et al. (2010a) for more details. On longer timescales, mass-loss is driven by the escape of stars due to internal two-body relaxation, or evaporation.
36 2.5 Spatial Correlation Between X-ray and Opti-
cal Sources
2.5.1 Correlation Between the Positions of XRBs and Star
Clusters
Kaaret et al. (2004) previously suggested that X-ray sources in three different starburst galaxies (M82, NGC 1569 and NGC 5253) may have formed in young star clusters, because they are preferentially located near star clusters, albeit with a signif- icant displacement (≈ 200 pc on average). This conclusion was based on a comparison between the cumulative distribution of displacements measured between actual X-ray sources and their closest clusters, and simulated displacements from a population of X- ray sources distributed randomly relative to the clusters. In order to explore whether or not X-ray binaries form in compact star clusters, here we compare the positions (of the entire sample) of 23 candidate X-ray binaries (presented in Section 2.2) with those of the star clusters (Section 2.3), to look for a statistically significant spatial correlation between the two.
We repeat the spatial correlation analysis performed by Kaaret et al. (2004) for our sources in NGC 4449. We identify the nearest star cluster to each X-ray binary, and measure the (projected) distance between the two. For XRBs that are not coincident with clusters, this gives a lower bound on the displacement of the source from its parent cluster, since the closest cluster is not necessarily the parent. The results for
NGC 4449 are shown in Figure 2-8 as a cumulative distribution of the separations between the binaries and the closest cluster. The figure shows that there are 7 XRBs that are within 100 pc of a star cluster, and nearly all are within ≈ 400 pc of a star cluster. Interestingly, three HMXBs in NGC 4449 are spatially coincident with very young star clusters (within the 1σ positional uncertainty of ≈ 0.5′′ for Chandra),
37 25
20
15 Number
10
5
10 100 1000 Distance [ pc ]
Figure 2-8 Cumulative distribution of the displacement (in parsecs) between X-ray binaries and star clusters. The black (solid) line shows the result for sources in NGC 4449. The other lines show results from our Monte Carlo simulations. The red solid line is the median of 1000 random distributions where the the locations of the X-ray sources are drawn from a distribution that follows the light of the galaxy. The dashed lines show the 1 σ deviation from this median curve. The blue solid line represents the median of 1000 random distributions where the X-ray source locations are drawn randomly from across the face of the galaxy. suggesting that these XRBs formed within the clusters. To determine whether the clustering of XRBs near star clusters is statistically significant, we compare the observations with simulated X-ray source populations.
We consider two different randomly distributed cases, i.e. where the X-ray sources are not associated with the clusters. First, we generate random sets of (23) uniformly
distributed sources, as done in Kaaret et al. (2004), and find spatial displacements
from the star clusters using the procedure described above. The simulation was run 1000 times, with the cumulative distribution determined in exactly the same way as
for the actual observations, and the median result shown as the blue line in Figure 2-8.
38 The uniformly distributed random sources show a clear difference from the observed
spatial displacements, similar to the results found by Kaaret et al. (2004) for three
different starburst galaxies (not including NGC 4449).
Next, we select random sets of (23) sources distributed in a more realistic fashion, one where the XRBs fall off radially like the stellar light in NGC 4449 itself. This was accomplished by fitting the radial luminosity profile of NGC 4449 in a 2MASS
K band image, by a broken power-law, φ(L) ∝ Lα, with α = −0.79 over the radial range 150 − 800 pc, and α = −2.28 over the radial range 800 − 3000 pc. The K
band should give a reasonably good measure of the distribution of older stars that
form the stellar backbone of NGC 4449, rather than the distribution of young star
clusters. The median result from 1000 runs of the random simulations described above are shown as the solid red line in Figure 2-8. In this case, the results of the
randomly drawn spatial displacements are more similar to the observations, except at
the small displacement end, with the observations now overlapping partly with the 1σ uncertainty line from the simulation at separations ∼> 100 pc. We find that a radial profile of NGC 4449 determined from far-ultraviolet images taken with GALEX gives very similar results to those based on the K-band profile. Our results indicate that the specific recipe used to populate synthetic X-ray sources in the image can affect the interpretation of the results, i.e. whether or not the X-ray sources are associated with star clusters. In general, a uniform distribution is much more likely to place synthetic
X-ray sources in the outer portions of the galaxy when compared with the second method; because there are few X-ray sources located in the outskirts of NGC 4449, this procedure causes the simulation and observations to look substantially different. We conclude that, in the case where the sources are distributed randomly but in a fashion similar to the underlying galaxy light, the random distribution does not differ significantly from the observed one, except at the smallest separations, where we
find several XRBs closer to star clusters than predicted by the random distributions.
39 These sources are discussed in more detail in Section 2.6.
2.5.2 Optical Sources That are Coincident With HMXBs
Three of the candidate HMXBs (X12, X21 and X46) have young (τ ∼< 107 yr) star clusters located within their ≈ 0.5′′ positional uncertainties. There is also an XRB coincident with an old globular cluster (X29); this is almost certainly a low-mass X-ray binary (LMXB). Here, we check the probability that this spatial coincidence is due to chance superposition rather than to a physical association between the HMXBs and the clusters. We distribute 23 XRBs (the number in our sample) randomly throughout
NGC 4449 in 100,000 runs, and find one chance superposition approximately every
fifth run when all 129 clusters were included. This is ≈25 times less than the four real observed coincidences between star clusters and XRBs. The probability of a chance superposition is almost identical if we only consider the three coincident HMXBs and star clusters younger than 10 Myr, i.e. the most probable host clusters. Here, the random sources follow the luminosity profile of NGC 4449 (see Section 2.4.1).
If the random sources are uniformly distributed instead, we find coincidences only
∼3% of the time, approximately 1 out of every 30 runs. Our results indicate that the positional coincidence between the XRBs and clusters in these three cases has a high statistical significance, and it is highly unlikely the result of chance superposition.
In addition to the XRBs that are coincident with star clusters, several others are coincident with an optical point source: X6, X10, X14, X20, and X24, as presented in Section 2.3.2. These point sources are well within the body of NGC 4449. They have the colors and magnitudes expected of supergiant stars, and are almost certainly the donor star in the X-ray emitting binary. Other candidate HMXBs are located in crowded regions, making it more difficult to identify a unique optical counterpart to the XRB, because there are several sources in the HST image within the positional uncertainties of the X-ray source. Some of the X-ray sources (e.g., X18, X13, X20,
40 X11, and X19) do not have any obvious optical counterpart.
2.6 Discussion
2.6.1 Properties of Star Clusters Closest to HMXBs
The ages of star clusters in the vicinity of the X-ray binaries can help to constrain the ages and the nature of the HMXBs. Even when there is no unambiguous stellar or cluster counterpart to an X-ray source, it is plausible that the XRB is the same age as the stars and clusters around it. Here, we present the physical properties of the star cluster that is closest to each HMXB. In Table 2.6, for each HMXB we compile the distance to the closest star cluster in our catalog, and the age and mass of this cluster. We also include the X-ray luminosity of each source determined in
Section 2.2.1, and a column that lists whether the X-ray colors are better described by a disk black body or a power law model, or are in the absorbed portion of the diagram.
Table 2.6: Properties of Star Cluster Closest to XRBs in
NGC 4449
ID X-ray ID of Closest Cluster Age1 Distance Notes Model Cluster log(τ/yr) (pc) X8 Absorption 96 9.34 266 X14 Absorption 117 8.06 339 X18 ... 117 8.06 285 X13 Absorption 112 8.06 232 X6 Absorption 93 6.64 122 X29 ... 89 8.91 7 LMXB X10 Power-law 99 8.31 286 X12 Disk blackbody 104 6.96 10 Coincident X20 Absorption 33 8.86 222 X11 Absorption 102 8.36 89 X23 Disk blackbody 66 6.58 181
41 Table 2.6 – Continued ID X-ray ID of Closest Cluster Age1 Distance Notes Model Cluster log(τ/yr) (pc) X15 Power law 13 8.31 9 Coincident X24 Disk blackbody 69 6.84 68 X21 Disk blackbody 53 6.72 13 Coincident X28 Disk blackbody 81 6.72 136 X31 ... 87 6.70 630 X22 Absorption 70 6.38 133 X19 Supersoftsource 51 8.06 383 X41 Supersoftsource 50 6.90 225 X42 Disk blackbody 6 6.40 113 X43 Absorption 62 6.40 57 X44 Absorbed 75 6.16 134 X46 Disk blackbody−Power law 111 6.70 4 Coincident
1Uncertainties in the cluster age estimates are typically ±0.3 in log τ.
The colors of HMXBs can vary, because these binaries go through different emis- sion states, or because of their geometric orientation, since inclination effects through a disk can result in colors in the absorbed portion of the X-ray two-color diagram.
This means that it is not possible to definitively identify the type of HMXB (e.g., black hole vs. neutron star) or to determine its age, just from the X-ray properties.
However, different types of HMXBs do show some general trends in the color-color diagram. For example, black hole binaries often approximately follow the disk black- body models (Remillard & McClintock 2006). At high accretion rates the X-ray emission from black hole binaries is dominated by thermal emission from a disk, lead- ing to both high luminosities and soft X-ray colors. On the other hand, accreting low-mass neutron star binaries often have absorbed power-law spectra.
42 2.6.2 A Population of Very Young, Massive Black Hole Bi-
naries in NGC 4449
We can select very young binaries, which are likely to be high-mass black hole binaries (BHB), in two different ways: (1) from their location in the X-ray color-color diagram, and (2) from their proximity to very young star clusters. We first select candidate BHBs from Table 2.6 based on their X-ray colors. There are seven X-ray sources that are best described by disk black-body models: X12, X21, X23, X24, X28, X42 and X46. All of these X-ray sources are either coincident with (X12, X21, and X46), or fairly close to (within 200 pc) a very young τ ∼< 8 Myr star cluster.7 The coincidence with clusters is particularly important, because it establishes a direct connection between the BHBs and very young clusters (recall that in Section 2.5.1 we found that chance superpositions are highly unlikely). These seven sources have a median X-ray luminosity of 2.25 × 1037 erg s−1. The fact that they have the X- ray colors expected for BHBs and are also close to very young star clusters strongly supports that these sources are in fact, very young BHBs (discussed in more detail below).
Next, we select candidate black hole binaries based solely on their proximity
(within 200 pc) of a very young (τ ∼< 8 Myr) star cluster. These criteria return all seven XRBs described by the disk black body models, plus four additional sources: X6, X22, X43, and X44. All four of these are found in the “absorbed” portion of the
X-ray two-color diagram. However, they are located near star forming regions which contain very young τ ∼< 8 Myr star clusters, and that are further out in NGC 4449
> (with galactocentric distances Rg ∼ 1.2 kpc) on average than the seven sources dis- cussed previously. The stellar density of the galaxy has dropped significantly at these galactocentric distances, which strongly suggests that X6, X22, X43, and X44 are
7Source X21 also happens to have an older, τ ≈ 200 Myr cluster nearby.
43 associated with the nearby, recent star formation. In fact, random simulations return
> only a single XRB with Rg ∼ 1.2 kpc and within 200 pc of a young star cluster in 10,000 runs.
The physical association between BHBs and very young star clusters is also appar- ent as the small separations in the cumulative distribution of spatial displacements between XRBs and star clusters (Figure 2-8). This regime deviates strongly from random distributions, as we would expect if BHBs are physically associated with very young star clusters. We conclude that (at least some) BHBs form in, and not just near, compact star clusters, based on the fact that three candidate BHBs in our sample are spatially coincident with a very young star clusters and that a fourth is very close to a young cluster (within a projected distance of 13 pc).
The ages of τ ∼< 6 − 8 Myr estimated for the coincident and proximate star clusters to the 11 XRBs discussed above strongly suggests that these are black hole
(rather than neutron star) binaries. At these very young ages, only stars initially more
> massive than M ∼ 25−30 M⊙ will have had time to become supernovae (based on the Padova models for solar metallicity). While there is still some uncertainty about the exact range of stellar masses that end their lives as black holes rather than as neutron stars, most models predict that the transition between these two types of compact remnants occurs for star with initial masses somewhere in the range of ≈ 18 − 25 M⊙
(e.g., Fryer (1999)), below the initial stellar masses that have completed their main sequence lifetime in ≈ 6 − 8 Myr old star clusters. However, metallicity effects may
complicate the relationship between initial stellar mass and remnant type (e.g., Heger
et al. (2003)). Given these uncertainties we will refer to these as ’candidate’ BHBs. One of the XRBs that is best described by a disk blackbody model, X24, is not
coincident with a star cluster, but is coincident with a bright point source. The lack
of any other bright stars within the 1σ astrometric uncertainty suggests that this is
the high mass donor star in the XRB. The luminosity and color of this source are
44 consistent with isochrones that are ≈ 8 − 10 Myr, as shown in Figure 2-4, similar to the age of the nearest star cluster.
To summarize, the main results of this section are that we find strong evidence for a population of BHBs in NGC 4449, and that these massive binaries likely formed recently in compact star clusters. Three of the candidate BHBs appear to reside within their parent star clusters, while the others do not. These very young (τ ∼< 6 − 8 Myr) sources comprise a significant fraction of all X-ray emitting binaries brighter than ≈ few × 1036 erg s−1 in NGC 4449. BHBs therefore make up approximately 48% or 11 out of 23 XRBs.
2.6.3 Processes Responsible for the Spatial Displacement Be-
tween BHBs and Star Clusters
Black holes are the compact remnants of massive O stars. O stars typically form in (massive) clusters rather than in the field, although they can be dynamically ejected
from clusters into the field. For example, studies of field O stars in the Milky Way (de
Wit et al. 2005) show that the majority of field O stars appear to be runaways from
nearby star clusters. Moreover, many and possibly most O stars are born in binaries
(Garc´ıa& Mermilliod 2001; Larson 2001). Section 2.6.2 provided strong evidence
that BHBs probably form in star clusters. In the previous section, we found that while three of the candidate BHBs in
NGC 4449 are coincident with a star cluster, the majority of them are not. What
physical process(es) is responsible for this displacement between BHBs and their par-
ent clusters? We consider three different mechanisms that could lead to the apparent
displacement between BHBs and their parent star clusters:
• The parent cluster has dissolved and is therefore no longer visible, leading to an apparent displacement between the BHB and neighboring clusters.
45 • The BHB was ejected from its parent cluster during dynamical interactions with
stars in the dense cluster core (Poveda et al. 1967; Gies & Bolton 1986).
• The BHB was ejected from its parent cluster due to an asymmetric supernova
kick (Zwicky 1957; Blaauw 1961).
In Section 2.4, we suggested that the shape of the cluster mass and age distribu-
tions in NGC 4449 are primarily signatures of the disruption rather than the formation
of the clusters. Regardless of whether individual clusters dissolve partially or fully,
our results in NGC 4449 (and those in other galaxies), suggest that star clusters lose a
significant amount of mass very quickly, on timescales of only ≈ 10 Myr8. UV spectra
of starburst galaxies reveal that the dispersed field populations are dominated by B stars, whereas UV-bright star clusters are often dominated by O stars, consistent with
a scenario of rapid cluster dispersal and/or mass-loss (e.g., Tremonti et al. (2001);
Chandar et al. (2003)). Despite growing evidence that even massive star clusters, i.e.,
those most likely to host O stars, may disperse rapidly, we believe that this is unlikely
to be the mechanism responsible for the observed spatial displacement between very
young BHBs and star clusters. N-body simulations show that an unbound cluster retains the appearance of a bound cluster for 10 − 20 crossing times (Baumgardt &
Kroupa 2007), on the order of 10-20 Myr for typical clusters in our NGC 4449 catalog.
Since no cluster is observed at the locations of the majority of the BHBs, the early
dispersal of clusters can be ruled out as the origin of the spatial displacement between
very young, high mass BHBs and star clusters.
We conclude that many BHBs in NGC 4449 have been ejected from their parent clusters, either via dynamical kicks due to interactions with other stars in cluster
cores, or due to an asymmetry in the supernova explosion of the compact object,
8This early mass loss is not driven by relaxation of the cluster due to two-body interactions (e.g., Fall & Zhang (2001)). Relaxation-driven evaporation operates on significantly longer timescales than the ages of the BHBs, and ejects mostly low mass stars.
46 which nonetheless does not unbind the binary. We cannot differentiate between these
two mechanisms using the current observations. We can however, estimate lower and
upper limits to the ejection velocities. We estimate a lower kick velocity by assuming
that each BHB was ejected from its closest cluster very soon after it formed, and
divide the distance to this cluster by its age, neglecting for the moment uncertainties in the cluster age estimates. These lead to lower limits of 45, 10, 27, and 23 km/s for
X23, X24, X28, and X42, respectively. The other BHB binaries give limits of 28, 55,
23, and 84 km/s for X6, X22, X43, and X44. The cluster ages and hence the velocities
are uncertain by ≈ a factor of two, resulting in lower limits to the velocities between
≈ 5 − 160 km/s. Of course, dynamical ejection need not have occurred right after
the cluster formed, but could have occurred more recently. We estimate upper limits to the kick velocities by assuming that each BHB was ejected within the last 1 Myr,
i.e. we divide the distance by 1 Myr. This procedure gives a range of velocities (for
non-coincident sources) between ≈ 30 − 180 km/s.
2.6.4 The Nature of Older X-ray Binaries in NGC 4449
In the previous subsections we found that ≈ 11 of the 23 candidate XRBs in
NGC 4449 are likely very young, massive BHBs. The goal of this section is to better understand the ages and types of the remaining X-ray binaries.
We first constrain the number of low mass X-ray binaries (LMXBs) in NGC 4449,
systems consisting of a black hole or neutron star accreting from a low-mass compan- ion. Source X29 is almost certainly a LMXB, because it is spatially coincident with a cluster that has integrated colors similar to those of ancient Galactic globular clusters.
Its X-ray colors and luminosity however, are unremarkable when compared with the rest of the XRB sample, and therefore cannot be used to select ancient LMXBs in general. In their study of five early-type galaxies (ellipticals and lenticulars), which have formed LMXBs but not HMXBs, Kundu et al. (2007) find ≈ 25 − 60% of all
47 LMXBs are coincident with globular clusters and the rest are found in the field. As-
suming a similar fraction of cluster-to-field LMXBs in NGC 4449 implies that there
are ≈ 2 − 4 field and hence ≈ 3 − 5 total LMXBs in NGC 4449, ≈ 13 − 22% of our
total sample.
Our analysis implies that there are 7 − 9 remaining X-ray binaries in NGC 4449 that are neither very young, high mass BHBs nor very old LMXBs. These are likely
older than ≈ 10 Myr and younger than several billion years (age of LMXBs). Inter- estingly, from Table 2.6 we find that the closest clusters to X14, X18, X13, X10, X20,
X11, and X19 are all intermediate age, with τ ≈ 100 − 400 Myr. The separations between these X-ray binaries and their closest clusters is larger than those found be- tween BHBs and their closest clusters, with a median (mean) separation of ≈ 285 pc (250 pc).
These “intermediate” age HMXBs are likely neutron star X-ray binaries with either a Be or supergiant star for a companion. Accretion from the former occurs via a disk, while the latter are primarily wind fed. There are also differences in the optical luminosities of Be versus supergiant donor stars: Be-HMXBs in the Small
Magellanic Cloud have MV ranging from −2 to −5 (McBride et al. 2008), which at the distance of NGC 4449 corresponds to mV ≈ 26 − 23, while supergiant HMXBs in the Galaxy are significantly brighter (Chevalier & Ilovaisky 1998), with MV brighter
< than ≈ −6.5(mV ∼ 21.5). X10 and X14 are therefore candidate supergiant-HMXBs, because they have bright coincident point sources (see Section 3.2). X18, X13, X20,
X11, and X19 however, do not show any bright coincident point sources in the optical images down to mV ≈ 25, and are therefore likely Be-HMXBs. Be-X-ray binaries dominate the population of HMXBs in the Magellanic Clouds
(McBride et al. 2008). They turn-on approximately 20-50 Myr after a star formation event, as the first neutron stars are formed. The low mass of the neutron star and lower accretion rates mean that Be-X-ray binaries typically have lower X-ray lumi-
48 nosities than massive BHBs. However, most Be star X-ray binaries go into outburst
when the neutron star passes through the circumstellar disk of the Be star and the
accretion rate increases. We note that the candidate Be-HMXBs in NGC 4449 have
X-ray luminosities considerably higher than quiescent Be X-ray binaries. If they are
indeed Be X-ray binaries, then they are in outburst, and we are detecting only the highest luminosity sources. This would suggest there is a much larger population of
Be X-ray binaries in NGC 4449 below our detection threshold.
In Section 2.6.2 we used the spatial displacement diagram to support our conclu- sion that very young, massive BHBs form in compact star clusters. Here, we note that there is an inherent limitation in our ability to interpret this diagram for older binaries. While the association between very young clusters and XRBs shows up clearly, this is not true for intermediate age sources. Figure 2.9 shows the cumulative distribution of displacements between τ ∼> 10 Myr X-ray binaries and star clusters, where we have removed the BHBs and the known LMXB. After ≈ 50 − 100 Myr,
XRBs that have been ejected at even low velocities of a few km/s are no longer
obviously spatially associated with their parent clusters, since at separations larger
than ≈ 200 pc the distribution is consistent with the 1σ uncertainties from a random distribution.
2.7 Summary and Conclusions
In this chapter, we presented the discovery of 23 candidate X-ray binaries in
the nearby starburst galaxy NGC 4449, from Chandra/ACIS-S observations. We
measured count rates, luminosities, and colors for these sources.
We also presented a new catalog of 129 compact star clusters brighter than MV ≈ −7 in NGC 4449 from multi-band, high resolution, optical imaging taken with the ACS/WFC and WFPC2 cameras on-board the HST . This doubled the number of
49 12
10
8
6 Number
4
2
10 100 1000 Distance [ pc ]
Figure 2-9 Same as Figure 2-8, but excluding the eleven massive BHBs and the known LMXB.
compact star clusters known in this galaxy. Mass and age estimates for these clusters
> 4 show that NGC 4449 has formed relatively massive (M ∼ 10 M⊙) clusters more or less continuously over the last ≈ 109 yr. The joint distribution of cluster ages and masses appears to be similar to those found in other nearby galaxies such as the
Magellanic Clouds, M83, and the merging Antennae galaxies, albeit with somewhat larger uncertainties due to the smaller number of clusters, and can be approximated as g(M, τ) ∝ M βτ γ , with β = −2.16 ± 0.30 and γ = −1.0 ± 0.3.
Our main conclusion is that we have found clear evidence for a population of very young, high mass, black hole X-ray binaries in NGC 4449. We find 11 candidate high mass BHBs, nearly half of the sample of X-ray emitting binaries. Three of the BHB candidates are coincident, within the astrometric uncertainties of Chandra, with a very young, τ ∼< 8 Myr star cluster, and a fourth is nearly coincident with a very young cluster. The others are all within 200 pc of very young clusters. Based on these results,
50 we suggest that these massive BHBs form in star clusters, where most massive O stars
(the progenitors of BHBs) are born. Many are subsequently ejected from their parent
clusters either due to dynamical interactions within dense clusters or as the result of
an asymmetric supernova explosion. The observed displacement between BHBs and
very young star clusters is not caused by the dissolution of the parent clusters. The small separation between massive BHB candidates and their closest cluster clearly
deviates from randomly populated distributions, further supporting our conclusion
that these X-ray binaries have a direct relationship with young star clusters.
We found one X-ray binary in NGC 4449 that is coincident with an old star cluster,
and hence is almost certainly a LMXB. Based on the fraction of LMXBs found in the
field vs. in globular clusters in early-type galaxies, we estimate that there are ≈ 2 − 4 field LMXBs and hence 3 − 5 total LMXBs in NGC 4449, although they cannot be identified from their X-ray properties alone.
We suggest that the remaining XRBs are intermediate age supergiant and Be-
HMXBs. Although none of these remaining binaries are coincident with a star cluster, the closest cluster to these sources has an estimated age between 100 − 400 Myr.
In two cases, we identify coincident point sources in the HST images that have luminosities consistent with those expected of supergiant donor stars. The other candidate intermediate-age X-ray binaries do not have an obvious optical counterpart, consistent with the expected luminosities for Be star donors.
Based on these results, we conclude that high mass X-ray binaries, particularly massive, black-hole binaries, dominate the X-ray luminosity from NGC 4449. We suggest that the ages and locations of star clusters provide important insight and constraints on the different types of X-ray binaries in nearby star-forming galaxies.
51 Chapter 3
X-ray Binaries and Star Clusters in
the Antennae: Optical Cluster Counterparts∗
We compare the locations of 82 X-ray binaries (XRBs) detected in the merging
Antennae galaxies by Zezas et al. (2006), based on observations taken with the Chan- dra X-ray Observatory, with a catalog of optically selected star clusters presented by
Whitmore et al. (2010), based on observations taken with the Hubble Space Telescope.
Within the 2σ positional uncertainty of ≈ 0.8′′, we find 22 X-ray binaries are coinci- dent with star clusters, where only 2-3 chance coincidences are expected. The ages of the clusters were estimated by comparing their UBVI, Hα colors with predictions
from stellar evolutionary models. We find that 14 of the 22 coincident XRBs (64%)
are hosted by star clusters with ages of ≈ 6 Myr or less. Based on these ages and
dynamical considerations, we suggest that these 14 sources are likely to be massive
binaries that have a BH as the compact object. All of the very young host clusters
> 4 5 are fairly massive and have M ∼ 3×10 M⊙, with many having masses M ≈ 10 M⊙.
∗Blagoy Rangelov, Andrea H. Prestwich, Rupali Chandar, & Bradley C. Whitmore Submitted for publication in the Astrophysical Journal
52 Five of the XRBs are hosted by young clusters with ages τ ≈ 10 − 100 Myr, while three are hosted by intermediate age clusters with τ ≈ 100 − 300 Myr. We conclude that precision age-dating of star clusters which are spatially coincident with XRBs in nearby star forming galaxies is a powerful method of constraining the nature of the
XRBs.
3.1 Introduction
Observations of the “Antennae” (NGC 4038/39) reveal that this merging system of two gas-rich spiral galaxies is full of point-like X-ray sources, many of which are probably high mass X-ray emitting binary star systems (e.g., Zezas et al. (2002)).
Most X-ray binaries in star forming galaxies are believed to have either a back hole or a neutron star as the compact source, formed after a massive star ends its life as a supernova. If this remnant is in a binary system with another star, mass transfer onto the compact object can result in X-ray emission. In this chapter, we will refer to X-ray binaries that have a massive donor star as high mass X-ray binaries.
−1 The Antennae have a fairly high rate of star formation, ≈ 20 − 30 M⊙ yr , with much of this star formation occuring in compact star clusters (e.g. Whitmore &
Schweizer (1995); Whitmore et al. (1999); Fall et al. (2005); Whitmore et al. (2010)).
Fall et al. (2005) found that at least 20%, and possibly 100% of the massive stars in the Antennae form in compact star clusters. Most O stars in the Galaxy are also found in clusters and associations. Studies of Galactic O stars currently found in the
field suggest that all but 4% of these likely formed in nearby clusters and associations, but were subsequently ejected (de Wit et al. 2004, 2005). Given the strong constraints on the (high) fraction of massive stars formed in clusters, we consider it highly likely that HMXBs in the Antennae formed within star clusters. Once formed, binary systems can: (1) remain within their host cluster, (2) be
53 dynamically kicked out of their parent/host cluster (either due to the supernova
explosion or due to dynamical interactions with other stars in the crowded centers of star clusters), or (3) be left behind after the parent cluster dissolves. The first scenario
leads to the strongest constraint on the ages of XRBs, via the age of the parent/host
star cluster. Therefore, precise age estimates for host clusters can provide important
information on XRBs. This chapter focuses on XRBs that remain within their host
cluster in the Antennae. The last two scenarios, which lead to a situation where
XRBs are close to but not coincident with star clusters, will be studied in a future paper.
Clark et al. (2011) recently compared X-ray sources in the Antennae from the
Zezas et al. (2006) catalog with near-infrared images from the WIRC camera on the Palomar 5-m telescope, and optical images from the WFPC2 camera on the
HST . The FWHM of the IR images is approximately 1′′, and that of the optical images is ≈ 0.22′′ (the three WF CCDs on the WFPC2 camera have a plate scale of 0.1′′ pix−1). Clark et al. (2011) found 32 likely IR counterparts to the X-ray sources. They estimated ages for a subset of 10 sources by comparing integrated
UBVIJK photometry with predictions from the STARBURST99 stellar population
models (Leitherer et al. 1999), and derived ages for all ten clusters of log(τ/yr) ≈
> 5 6.9−7.5. Their derived masses are M ∼ 10 M⊙ with several clusters having estimated
6 masses of ≈ 10 M⊙. Whitmore et al. (2010) recently studied star clusters in the Antennae from deeper,
higher resolution observations taken with the ACS camera on HST (the data and
observations are summarized in Section 3.2.2). These observations provide better
discrimination between star clusters, individual bight stars and background galaxies,
and better separate crowded young clusters, than the older WFPC2 observations. The
availability of this new cluster catalog allows us to revisit the properties of star clusters in the Antennae that host XRBs, using the highest quality observations currently
54 available. In this chapter, we investigate the ages and masses of star clusters that
are coincident with X-ray sources in the merging Antennae galaxies. We assume a
distance to the Antennae of 21 Mpc with a distance modulus of 31.71 mag (Schweizer et al. 2008; Riess et al. 2011). At this distance, 1′′ subtends a physical scale of 99.8 pc. Just over a quarter of the likely HMXBs in the Antennae are coincident or nearly coincident with a star cluster, and are the focus of this work. A follow-up paper will investigate the relationship between star clusters and X-ray binaries for the remaining three-quarters of the population that are not coincident with a star cluster.
This chapter is organized as follows. Section 3.2 summarizes the X-ray observa- tions and analysis of the X-ray binaries performed by Zezas et al. (2006). It also summarizes the optical observations and analysis of the star clusters performed by Whitmore et al. (2010). Section 3.3 describes the astrometric matching between the
X-ray and optical source catalogs. Section 3.4 presents the properties of star clusters that are coincident with X-ray binaries in the Antennae, and Section 3.5 discusses the implications of these results for the nature of the XRBs. We summarize our main conclusions in Section 3.6.
3.2 Data and Source Catalogs
3.2.1 X-ray Observations and Catalog of XRBs
The Antennae were observed with the ACIS-S instrument on the Chandra X-ray
Observatory as part of two programs (PI: Murray, Proposal Number: 01600062, and
PI: Fabbiano, Proposal Number: 03700413). There are seven individual observations with integration times between 37 and 75 ks, and a total exposure time of 411 ksec.
The limiting luminosity of the observations range between ≈ 2 × 1037 and 5 × 1037
erg s−1, depending on the exposure time and local background (see Zezas et al. (2006)
for more details). The publicly available online catalog (Zezas et al. 2006) contains
55 seven more sources that were detected in a merged exposure, thus increasing the total
number to 127. Of these, ninety-six are found within the main body and nuclear
region of the Antennae and hence are likely to be associated with the merger, while
the rest are almost certainly background galaxies.
The Chandra observations of the Antennae galaxies reveal both point sources and diffuse X-ray emission. Zezas et al. (2006) suggested that fifteen of their sources may be detections of diffuse emission rather than point sources. We visually inspected these sources in the observations, and agree that these detections are unlikely to be
XRBs except in two cases (X84 and X94), which have significant counts ( ∼> 20) in the hard band and are therefore unlikely to be associated with the diffuse plasma.
The remaining thirteen sources are eliminated from further consideration. We also eliminate X28, because we could not distinguish it spatially from X27 in the Chandra observations. This leaves a sample of 82 candidate XRBs in the Antennae. We show the locations of these sources on an optical image of the Antennae in Figure 3.1.
Zezas et al. (2006) extracted a number of parameters for the X-ray sources in the
Antennae, including locations, counts in the “soft” (S, 0.3 − 1.0 keV), “medium” (M,
1.0 − 2.5 keV), and “hard” (H, 2.5 − 7.0 keV) bands in each individual exposure, and from the combined image. Following Prestwich et al. (2003), we define two X-ray colors, H1 and H2, as follows. The soft color is defined as H1=(M − S)/T , and the hard color as H2=(H − M)/T , where S, M and H are the total counts in the bands defined above, and T is the total number of counts in all three bands. The
H1 and H2 values are determined from the published values in Zezas et al. (2006), after subtracting the corresponding background counts for each band. Note that the energy ranges in the S, M, and H bands are defined somewhat differently from those in Prestwich et al. (2003).
56 Figure 3-1 Optical image of the Antennae galaxies taken with the ACS camera on the HST . North is up and east is to the left. The locations of the star clusters used in this work are shown as circles. The locations of X-ray sources selected from Chandra X-ray observations are shown as crosses. See text for details.
57 Figure 3-2 Color image of the Antennae produced using HST /ACS images in B,V,I, and the narrow-band Hα observations (shown in purple) to highlight the sites of recent cluster formation. The locations of the 22 XRBs that are coincident with a star cluster, and which are the focus of this chapter, are identified and labeled.
58 3.2.2 Optical Observations and Catalog of Star Clusters
The XRBs detected in the Antennae will be compared with optically2 selected star clusters presented in Whitmore et al. (2010). This cluster catalog was based on HST observations of the Antennae taken with the ACS/WFC camera in the broad-band BVI and narrow-band Hα filters. The WFC on ACS has a pixel scale of
0.05′′ pixel−1. Point-like sources were detected using the DAOFIND task in IRAF, and photometry was performed using the IRAF task PHOT, with a 2-pixel aperture radius and inner and outer background annuli of 4 and 7 pixels, respectively. Corrections were applied to convert the aperture magnitudes to total magnitudes. Here, the
Hα filter contains both stellar continuum and nebular line emission (i.e., we did not perform any continuum subtraction). The instrumental magnitudes were converted to the VEGAMAG system using zeropoints from Sirianni et al. (2005) for ACS. In addition, U band photometry was measured for most sources from HST /WFPC2 images (the three WFC CCDs have a pixel scale of 0.1” and the PC CCD has 0.05”), using a 2 pixel aperture for the WFC CCDs (1.5 pixels for the PC) with zeropoints from Holtzman et al. (1995). The full catalog contains both individual stars and star clusters in the Antennae. Here, we select star clusters by restricting the sample to sources brighter than MV of −9, a limit that is brighter than nearly all individual stars.
We estimated the age and extinction of each cluster by performing a least χ2 fit comparing their UBVI,Hα magnitudes with predictions from the Bruzual & Charlot
(2003) models for simple stellar populations, assuming solar metallicity, a Salpeter initial stellar mass function, and a Galactic-type extinction curve. The mass of each cluster was estimated from the extinction-corrected V band luminosity and the (age-
2Optically selected cluster catalogs directly detect approximately 85% of clusters younger than a few Myr, i.e. the Antennae do not contain a significant population of very young, highly obscured star clusters (Whitmore & Zhang 2002).
59 dependent) mass-to-light ratio predicted by the Bruzual & Charlot models, assuming
a distance modulus of 31.71 mag to the Antennae (Schweizer et al. 2008; Riess et al. 2011).
In Whitmore et al. (2010) we assessed the accuracy of our photometric age esti- mates by comparing with ages determined for 16 clusters from ground-based spec- tra (Bastian et al. 2009). We found that our age estimates are generally within log (τ/yr) ≈ 0.3 or a factor of two of the spectroscopically derived ones for the age range of relevance here, τ ∼< 3 × 108 yr. For this work, we also created a continuum subtracted Hα image, which we used
to identified the sights of recent star formation, but not in the age-dating. An ex-
amination of this image shows that clusters with estimated ages of τ ∼< 6 Myr are associated with nebular emission in nearly all cases, as expected, again supporting the
accuracy of our cluster age determinations. This image will be used in Section 4 to
investigate the presence or absence of ionized gas associated with star clusters that are coincident with XRBs. For the mass estimates, the largest uncertainty comes from
uncertainty in the age determinations, which dominate the predicted mass-to-light
ratio. We previously estimated that the cluster masses are accurate at approximately
the factor of two level.
3.3 Astrometric Matching of the X-ray and Opti-
cal Catalogs
In order to find optical counterparts to XRBs in the Antennae, we first need to
match the X-ray and optical coordinate systems. We use the RA and Dec of X-ray
sources presented in Zezas et al. (2006). For the optical data, we use the combined BVI image available from the Hubble Legacy Archive (HLA; http://www.hla.stsci.edu);
sources in the HLA images have excellent relative positions, but there may be global
60 shifts, at the ≈ 1 − 2′′ level, relative to their absolute positions.
We first identified the optical counterparts to ten X-ray sources (X1, X2, X3, 6,
X8, X37, X60, X90, X107, and X117) in the outer regions of the HLA image. None of these sources are located within the main body of the Antennae, and hence are almost certainly background or foreground sources. Source X90 is a known QSO (studied by Clark et al. (2005)) and X107 is clearly a background galaxy in the optical image, while the remaining X-ray sources have point-like optical counterparts, suggesting that they may be QSOs. A comparison between the X-ray and optical coordinates for these ten sources gives a mean shift of 2.175′′, and a standard deviation (i.e. 1σ
positional uncertainties) of ≈ 0.4′′. This is similar to, although somewhat smaller than, the ≈ 0.6′′ positional uncertainty between the X-ray positions and the WIRC
IR images used in the Clark et al. (2011) study.
3.4 Optical Star Cluster Counterparts to XRBs
A comparison between the X-ray source positions and the optically selected star
cluster catalog in the main part of the galaxy reveals that 22 XRBs (see Figure 3.2) are
coincident with star clusters (15 XRBs have at least one cluster within 1σ or 0.4′′ and
7 others within 2σ or 0.8′′). Therefore, in ≈ 27% of the 82 candidate XRBs found in the main body of the Antennae, we can uniquely identify or strongly constrain the star cluster in which the XRB formed. A 2.5′′ × 2.5′′ portion of the optical
HST image showing the location of these 22 X-ray sources and coincident clusters is shown in Figure 3.3. Their locations are also shown in Figure 3.2. We compile the properties of all coincident objects in Table 3.1. This includes the coordinates,
S/N, and luminosity for the X-ray binaries from Zezas et al. (2006), and the cluster ages and masses estimated by Whitmore et al. (2010). In a few cases, particularly in the crowded sites of recent star formation, we find more than a single potential
61 cluster counterpart, and include information for all of these clusters, although in most
cases they have similar ages. In cases where an XRB has more than one potential
counterpart, we assume that the most massive cluster is the most likely host. In these
cases we only list clusters located within 1σ of the XRB.
Here, we check how many of the 22 candidate HMXBs found within 0.8′′ of a star cluster may be due to chance superposition. We populate the merger with 82 randomly distributed, synthetic XRBs in a fashion that follows the mass of the merger, and assume they are all bright enough to be detected. This is accomplished by smoothing the F814W (≈ I band) image, which best traces the overall mass of the
Antennae and is less sensitive to recent star formation than shorter wavelength images, and using this as a weight map for the random simulations. These simulations were run 1000 times, and show ≈ 2−3 chance coincidences (within 2σ) between the location
of XRBs and star clusters on average. This result suggests that the number of chance
coincidences is significantly lower than the number that we observe, and hence almost
all of the HMXBs that are coincident with star clusters are real associations and not
due to chance superposition.
The remaining 60 out of 82 (≈ 73%) of candidate XRBs in the Antennae are not coincident with a star cluster. We have visually confirmed that none of these are coincident with clusters fainter than our adopted magnitude limit. In 5 of the remaining 60 cases, we do find a relatively bright point source within 1σ with MV between ≈ −7 and −9, in the HST images. These are likely the donor stars in the
XRB. In the rest of the cases, no obvious optical counterpart to the candidate XRBs
is observed.
62 Table 3.1: Optical counterparts to X-ray sources
1 1 ID Our Ages Clark11+ Ages Our Masses Clark11+ Masses LX S/N Comments −1 log(τ/yr) log(τ/yr) log(M/M⊙) log(M/M⊙) log(erg s )
High probability XRB candidates X27 6.7 7.00−7.48 5.2 5.05−5.58 39.77 644.1 weak Hα emission X38 7.6 7.04−7.44 5.6 5.91−6.33 38.23 13.6 2σ X41 6.0 − 5.5 − 37.80 6.3 Strong Hα emission X45 7.7 − 5.0 − 37.40 3.8 2σ X47 6.7 − 4.5 − 38.35 37.0 2σ; weak Hα emission X49 7.7 − 4.9 − 38.15 20.4 weak Hα emission nearby X50 6.7 − 4.9 − 38.67 48.0 2σ; near NGC 4038 nucleus − − −
63 X52 7.5 5.8 1.6 X53 6.7 − 5.0 − 38.23 32.7 weak Hα emission ··· 6.5 ··· 4.9 ········· ··· 6.2 ··· 4.6 ········· X58 6.8 − 5.1 − 38.41 25.7 weak Hα emission X83 8.1 6.74−8.31 5.0 5.88−7.22 38.60 34.1 2σ X84 6.5 − 5.3 − 38.77 52.0 Hα emission ··· 6.6 ··· 5.0 ········· Hα emission ··· 6.4 ··· 4.8 ········· Hα emission X89 6.4 7.00−7.48 4.6 5.13−5.66 37.43 4.8 strong Hα emission ··· 6.5 ··· 5.4 ········· 2σ X91 8.3 − 6.2 − 37.37 4.3 X98 7.6 − 5.6 − 37.60 4.5 2σ ··· 7.7 ··· 5.1 ········· 2σ ··· 8.1 ··· 5.1 ········· 2σ X99 8.2 − 5.3 − 39.54 254.4 Table 3.1 – Continued
1 1 ID Our Ages Clark11+ Ages Our Masses Clark11+ Masses LX S/N Comments −1 log(τ/yr) log(τ/yr) log(M/M⊙) log(M/M⊙) log(erg s )
··· 6.7 − 4.7 − ··· ··· 2σ X101 6.6 7.12−7.19 5.7 5.97−5.99 38.02 10.3 Hα emission ··· 6.0 ··· 4.9 ········· Hα emission X121 6.7 − 5.5 − 37.53 5.5 Hα emission ··· 6.5 ··· 5.2 ········· Hα emission
Less certain XRB candidates X22 6.6 − 4.7 − 37.49 4.0 2σ; diffuse Hα emission X24 6.5 − 4.8 − − 2.8 strong Hα emission X36 6.6 − 4.5 − 37.97 9.0
64 X87 6.7 6.94−7.42 4.6 5.63−6.17 37.83 8.5 weak Hα emission
Notes. − In the comments we distinguish clusters that are within 2σ (0.8′′) rather than 1σ (0.4′′) of the candidate XRB.
1 From Zezas et al. 2006. Clark et al. (2011) found IR counterparts (within 2′′) to 32 of the X-ray sources
listed in the Zezas et al. (2006) catalog that are within the main body of the Antennae (see their Figure 1). Excluding sources that are beyond the field of view of the ACS
images used here, close to the northern nucleus, and those that may be detections of
diffuse plasma, we have 17 sources (coincident with clusters) in common with Clark
et al. (2011).
Sources X45 and X49 are in our list but not in Clark et al. (2011). They are
coincident with clusters that have ages of ≈ 50 Myr and are somewhat fainter than most of the very young clusters in our sample, and hence may not have been detected
in the WIRC observations used by Clark et al. (2011). Meanwhile, Clark et al. (2011)
list X7, X11, X86, and X94 as coincident sources, while we do not. These are missing
from our coincident list for various reasons. For X11, we find an optical counterpart,
but this appears to be a star, likely the donor star, rather than a cluster. X86 appears
to fall in a dust lane in the ACS observations, with no obvious clusters nearby, while X94 is close to, but not coincident with (≈ 2′′ away), a region of recent star formation
which shows some Hα emission, but no obvious clusters. Clark et al. (2011) find an extremely faint IR counterpart to source X7 which we do not detect. Based on these comparisons, our deep, high resolution optical observations appear to be at least as effective as ground-based infrared imaging for identifying the counterparts to X-ray sources, despite the relatively large dust content of the Antennae. These observations have the added advantage of higher resolution, and hence the ability to age-date the clusters more accurately.
X-ray point sources which are spatially coincident with star clusters are almost certainly X-ray binaries. We note here that the X-ray images of the Antennae do reveal a significant amount of diffuse X-ray emission coincident with some of the X- ray sources listed in the Zezas et al. (2006) catalog, and that in a few cases the listed sources have poor contrast with this diffuse emission. Following Zezas et al. (2006),
65 Figure 3-3 A 2.5′′ ×2.5′′ region around each of the 22 coincident sources is shown from the HST V band image. The two (green) circles represent 0.4′′ and 0.8′′ (one and two σ) positional uncertainties around each XRB. The (red) crosses identify clusters within the vicinity of each XRB.
66 we assume that these are likely faint XRBs, although it is possible that a few of
the most marginal cases may be spurious detections within this diffuse emission. In
Table 1 we list separately, at the end of the table, four X-ray sources from the Zezas
et al. (2006) work which have a somewhat lower probability of being XRBs because
of their low S/N in the X-ray images. The S/N from Zezas et al. (2006) is provided for all sources in Table 1.
The star clusters that are coincident with XRBs in the Antennae can be divided into three broad ranges in age. Of the 22 coincident sources, 14 of the counterparts are very young with estimated ages of 6 Myr or younger, five of the counterparts are young with ages of ≈ 20 − 50 Myr, and three have intermediate ages with τ ≈
100 − 300 Myr. We discuss these three broad age categories in more detail below. XRBs coincident with very young clusters are found in the crowded star forming
regions of the Antennae, and as noted in Table 3.1, nearly all of these clusters appear
to have at least some associated Hα emission, which is a good indication of their
youth. All of the clusters that likely host an XRB and which are younger than 10
4 5 Myr have masses higher than 3 × 10 M⊙ and up to several ×10 M⊙, with a median
5 mass of ≈ 10 M⊙. The median X-ray luminosity of the XRBs with very young optical counterparts is ≈ 1038 erg s−1.
Five of the X-ray sources are hosted by clusters that are somewhat older, with estimated ages of ≈ 20 − 50 Myr. As expected, clusters with these ages tend to be more spread out within the Antennae than the very young clusters. The average
(and median) mass of coincident clusters in this age range is somewhat higher than
5 for the youngest clusters, with ≈ 3 × 10 M⊙. The X-ray luminosities of XRBs with 20 − 50 Myr old counterparts ranges from 3 × 1037 to 2 × 1038 erg s−1, with a median
luminosity of 1.4 × 1038 erg s−1, i.e., roughly the same as the < 10 Myr clusters.
Three X-ray sources (X83, X91, and X99) are hosted by even older clusters, which are ≈ 100−300 Myr old. Clusters with these ages are fairly well dispersed throughout
67 the Antennae, and no longer trace regions of the most recent star formation. The clusters in this age range that are coincident with XRBs have estimated masses of 105,
6 5 1.6×10 , and2×10 M⊙, mostly higher than those of younger clusters that host XRBs.
In fact, as we will discuss more fully in a future paper, one quarter of the (16) star
5 clusters with ages of 100 − 300 Myr and masses higher than ≈ 6 × 10 M⊙ are either coincident with or very close to an XRB, suggesting that clusters with these ages and masses are quite efficient at producing bright XRBs. The X-ray luminosities of the
XRBs coincident with intermediate age clusters tend to be higher than those hosted by younger clusters, with luminosities of 4 × 1038, 2 × 1037, and 3.5 × 1039 erg sec−1, although the statistics are obviously poor.
Clark et al. (2011) presented ages for clusters coincident with ten X-ray sources in the Antennae, one (X102) of which is outside of our field of view. Their age estimates are also listed in Table 3.1, and have fairly similar estimated ranges, log τ ≈ 6.9 − 7.5 for nearly all of these clusters. We note that they do not derive ages as young as
3 − 6 Myr for any cluster coincident with an X-ray binary, as we have found here for the majority of coincident clusters. Below, we compare the age results of coincident star clusters from Clark et al. (2011) with those from our work.
• We find ages between 3 − 6 Myr for clusters associated with X87, X89, and
X101. Hα emission is observed in all three cases, as expected for clusters with
these ages. Clark et al. (2011) find typical ages of τ ≈ 20 ± 10 Myr for clusters
coincident with these sources.
• We find an age between 20 − 50 Myr for the cluster associated with X38. Clark et al. (2011) find similar but slightly younger age, with τ ≈ 20 ± 10 Myr.
• We find an age of 100 − 300 Myr for the cluster associated with X83. Clark et
al. (2011) find a (poorly constrained) age for this cluster somewhere between
5 Myr and 200 Myr. 68 In addition to differences in the ages of coincident clusters, we derive systematically
lower cluster masses than Clark et al. (2011). The main reason for this discrepancy
is likely the poorer resolution of the IR images compared with that of HST , 0.25′′
vs. ≈ 0.05′′, and the relatively large aperture radius used by Clark et al. (2011) for their photometry. It is clear from their Figure 3.2 and our Figure 3.1 that where they identify a single cluster, we sometimes find several, as they also note when comparing their IR images with the higher resolution optical images taken with WFPC2.
3.5 Discussion
3.5.1 Constraining the Nature of XRBs within Star Clusters
At an age of 6 Myr, stellar evolution models predict that only stars initially more massive than ≈ 30 M⊙ will have evolved off the main sequence and become compact
remnants. While there is still some uncertainty about the exact range of stellar masses
that end their lives as NSs and those that become BHs, most models predict that, for
lower metallicities, the transition between NSs and BHs occurs for stars with initial
masses somewhere in the range of ≈ 18 − 25 M⊙ (e.g., Fryer (1999), Heger et al.
(2003)), well below 30 M⊙. For approximately solar metallicities, as in the Antennae,
stellar evolution models predict that massive stars develop substantial winds, which
cause enough mass loss that the end product is a NS rather than a BH. Evolution of
massive stars in tight binary systems however, is likely to be even more complicated
with mass transfer back and forth, making it difficult for stellar evolution models
alone to predict which stars become BHs and which ones become NSs. Independent clues to the nature of the compact objects in HMXBs comes from
their dynamics. N-body simulations show that NSs are far more susceptible to ejection
from their parent cluster than BHs. Portegies Zwart et al. (2007) found that the vast
majority of NSs (≥ 90%) are ejected from their parent clusters, while only ≈ 45% of
69 the BHs were expelled. This is because natal kicks imparted during the supernovae
explosions that formed the compact objects typically fling lower mass NSs from their
birth sites.
We have recently performed N-body simulations of star clusters, with no primor- dial binaries, over their first ≈ 300 Myr of evolution, tracking all NSs and BHs (Sen et al. 2012). Our results confirm those of Portegies Zwart et al. (2007), and extend them to candidate HMXBs. Our main result is that an HMXB found within its parent star cluster is very likely to have a BH and not a NS as the compact object (this is true for at least ages up to ≈ 300 Myr, the limit of our simulations). The reason for this is two-fold: i) BHs have a much higher probability of being retained within their parent clusters (as also found by Portegies Zwart et al. (2007)), and ii) the higher masses of the retained BHs, when compared with NSs, makes them much more likely to form a binary system (e.g., Garofali et al. (2012), Sen et al. (2012), Converse & Stahler
(2011)). Conversely, HMXBs found in the field and not with their parent clusters or associations are much more likely to have a NS as the compact object. It is possible that a population of primordial binaries may somewhat alter these conclusions, but this will depend on the initial mass ratios and orbital parameters of the binaries.
3.5.2 The Nature of HMXBs Associated with Star Clusters
in the Antennae
In Section 3.5.1 we presented dynamical arguments that suggest that HMXBs hosted by star clusters have a BH as the compact object, regardless of the metallicity- dependent evolutionary path that led to the formation of the compact object. We therefore conclude that most of the candidate HMXBs studied here have a BH as the compact object, because they still reside within their parent clusters. This suggests that at least a quarter (22 out of 82) of the luminous XRBs observed in the Antennae
70 have a BH as the compact object. Although quite uncertain, most XRBs in the
Galaxy and the Magellanic Clouds appear to have a neutron star as the compact object. According to Belczynski & Ziolkowski (2009) neutron star XRBs outnumber black hole XRBs by a factor of ∼ 30 in our Galaxy. These X-ray emitting binaries are found primarily in the field, and not in clusters/associations, and hence the large fraction of NS-binaries is consistent with the dynamical arguments presented above.
The ages of ≈ 3 − 6 Myr determined for very young star clusters in the Antennae that are coincident with HMXBs is consistent with predictions of HMXBs formation.
Linden et al. (2010) followed the evolution of primordial pairs of stars drawn from a
Salpeter (α = −2.35) initial mass function, with a minimum mass of 4 M⊙ for the primary, and the mass for the secondary drawn randomly from a flat distribution for the mass ratios. The initial binary separation was drawn from a flat distribution (in
5 log space) with an upper limit of 10 R⊙. They use the population synthesis code
StarTrack (Belczynski et al. 2008) to model processes such as stable mass transfer through Roche-lobe overflow and unstable common envelope phases. Linden et al.
(2010) show in their Figure 1 that in solar metallicity systems like the Antennae,
36 −1 bright (LX > 10 erg s ) HMXBs turn on suddenly at ages of approximately 4 Myr, with a production rate that is sharply peaked between 4 and 6 Myr and dropping off thereafter. These ages are quite similar to those listed in Table 3.1 for many coincident clusters, and are probably not due to a peak in the star formation rate 4−6 Myr ago, since: (i) dynamical simulations which reproduce the observed morphology of the merging Antennae do not find such a peak in the star formation rate (see e.g., Karl et al. (2011)); and (ii) it would be impossible to have such a well timed burst over the entire field of the Antennae since the communication time is ∼ 109 yr for a signal traveling at the typical random velocity of the interstellar medium of ≈ 10 km s−1.
(e.g., Fall et al. (2009); Whitmore et al. (2007)). Hence, our age analysis is consistent with predictions for the production time scale of HMXBs in high metallicity galaxies.
71 In the dwarf starburst galaxy NGC 4449 (Rangelov et al. 2011) we found three
HMXBs coincident with very young star clusters and one coincident with an intermediate- age one. Although NGC 4449 has a smaller number of HMXBs than the Antennae
(since it has fewer stars in general) leading to poorer statistics when analyzing dif- ferent HMXB sub-populations, there do appear to be similarities between the two systems. In both cases, ≈ 15% of all HMXBs are coincident with very young clusters
(τ ∼< 8 Myr). There are however, no HMXBs in NGC 4449 coincident with 30 − 50 Myr old clusters. We will further investigate this issue by adding more dwarf starburst galaxies to our sample in Chapter 4.
The Antennae galaxies contain 6 ultra-luminous X-ray sources (ULXs; LX > 1039 erg/s). Of these we find 2 that are coincident with star clusters: X27 with a very young (τ ≈ 3−6 Myr) cluster and X99 with an intermediate age (τ ≈ 100−200 Myr) cluster.
3.6 Summary and Conclusions
In this work, we compared the locations of 82 candidate XRBs (from Zezas et al. (2006)) with the locations of massive star clusters (from Whitmore et al. (2010)) in the merging Antennae galaxies. Our main conclusions are:
• 22 out of the 82 candidate XRBs are located within 0.8′′ of a star cluster,
i.e. within the 2σ positional uncertainty of the X-ray and optical observations.
Only 2-3 coincidences between XRBs and star clusters would be expected due to chance superposition, indicating that the XRBs likely formed within these
star clusters.
• We found that the ages of host clusters fall within three ranges: (1) very young,
with τ ≈ 3−6 Myr; (2) young, with τ ≈ 30−50 Myr; and (3) intermediate age,
72 with τ ≈ 100 − 300 Myr. Fourteen of the 22 XRBs are hosted by very young
clusters, five by young clusters, and three by intermediate age clusters.
• We found two ULX candidates that are coincident with a star cluster (one
cluster has an age of τ ≈ 5 − 6 Myr and the other τ ≈ 100 − 200 Myr).
• Direct N-body simulations of star clusters have shown that XRBs found within their parent clusters are highly likely to have a black hole rather than a neutron
star as the compact object (Garofali et al. 2012; Sen et al. 2012). We therefore
conclude that the 22 XRBs that are coincident with a cluster in the Antennae
are most likely black hole binaries of different ages.
Our results indicate that the relative locations of XRBs and star clusters combined with precise age dating of star clusters is a powerful approach for finding BH binaries and determining their ages.
73 Chapter 4
High Mass X-ray Binaries in
Nearby Starburst Galaxies
4.1 Introduction
As seen in the previous two chapters, unlike major galaxy mergers like the An- tennae, dwarf starbursts tend to have relatively few HMXBs. While the individual analyses from such galaxies lacks statistical significance to produce unambiguous re- sults on their separations from parent clusters or on the types of compact objects and donor stars, combining the data from multiple starburst galaxies allows us to draw more solid conclusions for galaxies of this type.
In this chapter we present the results from a study of HMXBs in six starburst galaxies − He 2-10, NGC 1569, NGC 1705, NGC 3125, NGC 4214, and NGC 5253
(Table 4.1). These galaxies are located within ≈ 10 Mpc, have high spatial resolution X-ray observations from Chandra and optical observations from HST. These galaxies are also known to have recently formed massive star clusters. HST images of the six starbursts are shown in Figure 4-1. We detect X-ray source populations and select candidate HMXBs in each galaxy, and select the closest cluster to each HMXB. We also identify a number of potential donor stars to the HMXBs from the HST images.
74 One important result from HST has been the discovery that much of the star formation in nearby galaxies occurs in young, compact star clusters (Whitmore et al.
2010). As shown in the previous two Chapters, HMXBs, whose progenitors are very massive stars, most likely form in young clusters. Thus, knowing the properties of the potential parent clusters of XRBs can provide important information about the XRBs that is not possible to obtain from X-ray observations alone, such as their ages and whether they have a black hole or neutron star as the compact object.
Most O- and early B-type stars are found in binaries (Sana & Evans 2011). Even single field high-mass stars are often believed to have been part of a multiple system in the past, then ejected by a supernova kick or by dynamical interaction. This suggests that star-formation works differently for high-mass than low-mass stars. There is both theoretical and observational evidence that suggests that high-mass stars often form as near-equal mass pairs. This likely explains the number of XRBs we see in low- mass system, such as low-mass star clusters, where it will be statistically unlikely to have a high-mass pair of stars if a random pairing is assumed. Indeed, recent results from numerical simulations of star clusters (Garofali et al. 2012) show that dynamical interactions cannot produce the XRB populations observed in nearby galaxies, rather high-mass primordial pairs are needed in order to explain the observations.
Supergiant donors are bright, with MV brighter than ≈ −6.5 (Chevalier & Ilo- vaisky 1998), while Be donors tend to be fainter, with typical MV ranging from −2 to −5 (McBride et al. 2008). Deep photometry with HST can therefore be used to distinguish between the two main classes of HMXBs.
This chapter is organized as follows. Section 4.2 presents the X-ray data, selection and basic measurements of the XRBs. Section 4.3 summarizes the optical observations and photometry of the closest star clusters and of potential donor stars to each XRB.
We present the results from our analysis of X-ray properties of XRBs and spatial correlation between HMXBs and star clusters in Section 4.4. In Section 4.5 we discuss
75 Figure 4-1 HST BVIHα images of the six starburst galaxies. Left to right, top to bottome: He 2-10, NGC 1569, NGC 1705, NGC 3125, NGC 4214, and NGC 5253. (Credit: NASA/ESA)
76 Table 4.1. Sample Galaxies
a a a Galaxy nH Distance MB log(O/H)+12 SFR Name (×1020 cm−2) (Mpc)
He 2-10 9.70 9.0 -17.7 8.7b 1.4 NGC 1569 22.61 3.3 -17.5 8.2c 1.5 NGC 1705 3.85 5.1 -15.8 8.2d 0.1 NGC 3125 5.71 10.8 -17.0 8.3e 0.2 NGC 4214 1.49 2.9 -14.0 8.2f 0.2 NGC 5253 3.88 3.2 -16.9 8.2g 0.1
Note. — aLee et al. (2009); bAllende Prieto et al. (2001); c Kobulnicky et al. (1999); dLee & Skillmannti (2004); e Engelbracht et al. (2008); f Ma´ız- Apell´aniz et al. (2002); gL´opez-S´anchez et al. (2007). the nature of the XRBs. We summarize our main conclusions in Section 4.6.
4.2 X-ray Observations from Chandra
We use twelve sets of archival Chandra observations of the six starburst galaxies in
our sample to detect X-ray point sources. Basic information is given in Table 4.2. The
data were taken with the Advanced CCD Imaging Spectrometer (ACIS) instrument
on the Chandra X-ray Observatory, positioning the galaxies on the back-illuminated S3 CCD chip. We processed the data using the Chandra Interactive Analysis of Obser-
vations (CIAO) software (version 4.3) and Chandra Calibration Data Base (CALDB)
version 4.3.01.
We use CIAO’s Mexican-hat wavelet source detection routine wavdetect (Freeman
et al. 2002) to create the source lists. The techniques used to detect the X-ray sources
and determine their properties are similar to those described in Section 2.2.2. X-ray
1http://cxc.harvard.edu/ciao/ 77 Table 4.2. Chandra Observations
Galaxy ObsId Date PI Exposure Name ks
He2-10 2075 2001-03-23 Martin 20 NGC1569 782 2000-04-11 Martin 100 NGC1569 4745 2004-01-10 Zezas 10 NGC1705 3930 2003-09-12 Heckman 60 NGC3125 4012 2003-02-16 Satyapal 5 NGC3125 8181 2007-03-16 Strickland 60 NGC4214 2030 2001-10-16 Heckman 30 NGC4214 4743 2004-04-03 Zezas 30 NGC4214 5197 2004-07-30 Zezas 30 NGC5253 2032 2001-01-13 Heckman 60 NGC5253 7153 2005-12-12 Zezas 70 NGC5253 7154 2006-05-20 Zezas 70
luminosities (0.3 − 8 keV range) are derived by fitting a power law (Γ = 1.5) to the
data. The optical images were used to help eliminate non-HMXB X-ray sources, such
as foreground galaxies and background stars, as described in Section 4.3. Table 4.3
gives basic information about all 46 X-ray sources associated with the six starburst
galaxies in our sample.
While individual X-ray sources in He 2-10, NGC 1569, NGC 4214, and NGC 5253 have been studied before, no systematic study of the X-ray properties of these sources have been attempted. Kaaret et al. (2004) previously detected XRBs in NGC 1569 and NGC 5253, and their results are similar to ours.
4.3 Identification of Optical Counterparts to XRBs
In order to find optical counterparts to XRBs in the six starbursts, we first need
to match the X-ray and optical coordinate systems for each galaxy. The RA and Dec
78 of X-ray sources detected in the Chandra observations have positional uncertainties of approximately 0.5′′. For matching purposes, we use V -band images available from the Hubble Legacy Archive (HLA; http://www.hla.stsci.edu).
Table 4.3: Catalogue of the XRB Candidates
1 ID RA DEC Net Soft Hard LX Counts Color Color log(erg s−1) He 2-10 X203 129.065922 -26.409508 45.22 0.39 -0.22 38.22 X204 129.063028 -26.409394 315.58 0.02 0.16 39.08 X205 129.066784 -26.408520 27.26 0.23 -0.30 37.99
NGC 1569 X306 67.701016 64.846293 215.66 0.42 -0.01 37.15 X307 67.739307 64.846855 287.66 0.45 -0.17 37.32 X308 67.700647 64.847482 213.65 0.18 -0.44 38.36 X309 67.702725 64.849683 249.01 0.31 -0.22 38.33 X323 67.719753 64.844135 16.66 0.10 0.80 − X324 67.725305 64.845346 30.70 0.01 -0.43 36.44 X325 67.722377 64.845579 30.96 0.22 -0.39 36.37 X328 67.707720 64.848736 212.25 0.20 -0.04 37.17 X329 67.698817 64.851022 40.81 0.11 0.03 36.23 X330 67.695644 64.851960 27.42 0.92 -0.84 36.17 X336 67.694012 64.848168 17.63 -0.07 -0.43 36.17
NGC 1705 X405 73.561804 -53.362187 10.67 0.37 -0.03 36.45
NGC 3125 X501 151.64187 -29.938130 212.27 0.41 0.06 38.83 X502 151.63862 -29.934296 246.34 0.52 -0.14 38.97 X503 151.63907 -29.934837 33.55 0.64 -0.35 38.13 X504 151.63858 -29.935240 14.20 0.03 -0.42 37.91 X505 151.64146 -29.938102 956.99 0.41 -0.12 39.67
NGC 4214 X601 183.936372 36.312793 41.01 0.30 0.18 37.04 X602 183.909333 36.322510 897.31 0.13 -0.16 38.50 X603 183.920271 36.327706 30.67 0.32 -0.09 36.88 X604 183.912418 36.328347 7.00 -1.00 0.00 36.62 X605 183.909013 36.329149 171.00 0.26 -0.22 37.77 X607 183.931489 36.336072 10.66 -0.16 -0.28 36.46 79 Table 4.3 – Continued
1 ID RA DEC Net Soft Hard LX Counts Color Color log(erg s−1) X608 183.908974 36.347449 29.67 0.42 -0.40 36.90 X609 183.914059 36.348695 13.00 -0.74 -0.13 36.65 X617 183.916667 36.311349 18.00 -0.85 0.15 36.80 X618 183.889400 36.316614 18.34 -0.93 -0.04 36.85 X628 183.913220 36.325318 30.34 0.47 -0.01 36.91 X646 183.921580 36.316132 10.67 0.12 0.28 36.29 X647 183.916795 36.326743 28.33 -0.39 -0.24 37.04
NGC 5253 X704 204.961077 -31.655714 82.99 -0.06 -0.14 37.16 X706 204.963214 -31.644693 20.67 0.05 0.76 35.89 X707 204.980746 -31.642713 33.06 -0.50 -0.19 36.75 X708 204.974156 -31.642396 43.68 0.20 -0.45 36.93 X709 204.985208 -31.640527 91.03 -0.65 -0.16 37.20 X710 204.983446 -31.640059 77.93 -0.02 -0.41 37.12 X711 204.982671 -31.639887 56.69 -0.24 -0.30 37.05 X712 204.984760 -31.639021 441.65 0.11 -0.30 37.92 X713 204.996885 -31.630692 20.00 0.02 -0.33 36.56 X714 205.017701 -31.625079 23.67 0.06 -0.17 36.60 X715 205.009085 -31.621954 52.33 -0.15 -0.20 36.99 X716 205.014086 -31.620633 61.66 0.47 -0.14 36.90 X749 204.989868 -31.636764 2423.28 0.24 0.21 38.52
1X-ray luminosities (0.3 − 8 keV range) are derived by fitting a power law (Γ = 1.5) to the
data. Galactic nH is applied (see Table 4.1).
The HLA images are corrected for geometric distortion leading to excellent relative astrometry within each image, but can have global shifts up to ≈ 2′′ relative to the Chandra
positions. In order to determine these shifts, we identified the optical counterparts to a
number of X-ray sources in the outer regions of the HLA images, beyond the optically
luminous portion of each galaxy. These are mostly background or foreground objects, since
in some cases we see spiral structure as expected for late-type background galaxies, and in
others diffraction spikes as expected for saturated foreground stars. When comparing the
locations of the X-ray and optical images, we assume an uncertainty of 0.5′′, the intrinsic
80 astrometric uncertainty of Chandra.
At distances of ≈ 10 Mpc and less, most star clusters are broader than the point-spread function. The closest cluster to each XRB was identified, using similar criteria to those used in NGC 4449 and described in Chapter 2.3.3. The galaxies were observed with different cameras on HST, either the WFPC2, ACS (WFC or HRC) or WFC3 cameras, in a variety of filters. The WFPC2 has a pixel scale of 0.1′′ pixel−1 for the WF CCDs, the HRC and
WFC on ACS have 0.025 and 0.05′′ pixel−1 respectively, and WFC3 has 0.046′′ pixel−1.
Photometry of the clusters was performed using the IRAF task PHOT, with a 3-pixel aperture radius. The instrumental magnitudes were converted to the VEGAMAG system using the calibrations given in Holtzman et al. (1995) (WFPC2), Sirianni et al. (2005)
(ACS), or from http://www.stsci.edu/hst/wfc3 (WFC3), as appropriate. We estimated the age and extinction of each cluster by comparing the measured optical magnitudes with predictions from the Bruzual & Charlot (2003) models for simple stellar populations. When no U band observations are available, we simply correct for foreground extinction assuming the Fitzpatrick (1999) Milky Way extinction model, and then select the model closest to the observed colors. While this can lead to larger uncertainties on age, we have made independent visual checks of the clusters, using techniques similar to those described in
Whitmore et al. (2010), to confirm the age determinations. Information on the closest clusters is presented in Table 4.4. In a few cases, more precise age determinations are available in the literature for the clusters from ultraviolet spectroscopy. These are noted in the Table.
Table 4.4: Closest star clusters to XRBs
ID DIstance U-B B-V V-I Age1 (pc) log(τ/yr) X203 61 · · · -1.20 -0.11 7.0 X204 92 · · · -1.06 0.27 6.6a X204 62 · · · -0.96 0.33 6.7a X205 237 · · · -1.20 -0.11 7.0 X306 80 ··· ··· 1.09 8.5 X306 106 ··· ··· 1.36 7-7.3b X307 230 ··· ··· -0.03 6.7 81 Table 4.4 – Continued
ID DIstance U-B B-V V-I Age1 (pc) log(τ/yr) X308 79 ··· ··· 1.09 8.5 X308 82 ··· ··· 1.36 7-7.3b X309 37 ··· ··· 0.68 6.8b X323 55 ··· ··· 0.06 6.7 X324 51 ··· ··· -0.02 6.7 X325 29 ··· ··· 1.00 8.0 X328 84 ··· ··· 1.36 7-7.3b X328 69 ··· ··· ··· 8.5 X329 40 ··· ··· 1.29 6.8b X329 86 ··· ··· 0.68 6.8b X330 14 ··· ··· 1.45 6.7 X330 75 ··· ··· 1.29 6.8b X336 19 ··· ··· 0.99 8.5 X405 339 ··· ··· 3.39 7.1c X501 466 -1.21 -0.06 -0.07 6.7 X502 338 -1.44 0.04 -0.19 6.7 X503 137 -1.44 0.04 -0.19 6.7 X504 76 -0.15 0.47 0.69 6.7 X504 117 -1.44 0.04 -0.19 6.7 X505 458 -1.21 -0.06 -0.07 6.7 X601 272 -0.54 0.09 0.53 8.5 X602 334 -1.36 -0.52 0.29 6.6d X603 119 -0.89 0.40 0.90 8.0 X603 141 -1.03 0.97 1.23 6.9d X604 146 -1.53 -0.52 0.29 6.6d X604 170 -0.45 0.19 0.96 7.2d X605 72 -0.45 0.19 0.96 7.2d X607 183 0.30 0.20 1.16 8.5 X608 112 -0.24 -0.22 0.48 7.7 X609 346 -0.24 -0.22 0.48 7.7 X618 331 -1.71 1.73 0.57 8.5 X628 84 -1.53 -0.52 0.29 6.6d X646 90 -0.70 0.66 1.00 8.5 X646 99 -1.16 -0.34 0.61 6.7 X647 41 -1.03 0.97 1.23 6.9d X647 122 -1.53 -0.52 0.29 6.6d X704 431 · · · 0.68 1.01 9.0 X706 378 · · · 0.63 1.00 9.0 X706 420 · · · 0.68 1.01 9.0 X707 112 · · · 0.06 0.64 8.0 X708 332 · · · 0.63 1.00 9.0 X708 380 · · · 0.78 1.15 8.5 X709 113 · · · 1.89 -0.67 6.6
82 Table 4.4 – Continued
ID DIstance U-B B-V V-I Age1 (pc) log(τ/yr) X710 92 · · · 1.89 -0.67 6.6 X711 40 · · · 1.89 -0.67 6.6 X712 112 · · · 1.89 -0.67 6.6 X713 92 · · · 0.38 0.70 8.5 X714 580 · · · 0.58 0.90 8.5 X715 381 · · · 0.58 0.90 8.5 X716 577 · · · 0.58 0.90 8.5 X749 32 · · · 0.76 0.92 9.0
Notes. − aChandar et al. (2003); b Westmoquette et al. (2007); cAnnibali et al. (2009); d
Ubeda´ et al. (2007).
1Typical uncertainties are approximately ±0.3 in log τ.
2XRB coincident with star cluster.
The HST images also reveal point sources that are coincident with candidate HMXBs in
a number of cases. For these, we determine colors and luminosities by performing aperture
photometry in a manner similar to that performed for the clusters. We estimate the age of
each star by performing a χ2 fit comparing observed magnitudes with the stellar population
models with solar metallicity from the Padova group (Girardi et al. 2008; Marigo et al.
2008). The results are presented in Table 4.5.
Table 4.5: Donor stars of XRBs
ID MV U-B B-V V-I Age log(τ/yr) X203 -7.30 · · · -0.79 1.02 6.97 X205 -4.10 · · · 0.32 -2.45 6.55 X306 -4.10 ··· ··· 0.02 7.79 X307 -5.20 ··· ··· 1.21 7.50 X308 -4.20 ··· ··· 2.03 7.61 X309 -7.90 ··· ··· 0.89 6.89 X324 -4.80 ··· ··· 0.54 7.76 X325 -4.60 ··· ··· 1.85 7.46 X328 -7.30 ··· ··· 0.32 7.18 83 Table 4.5 – Continued
ID MV U-B B-V V-I Age log(τ/yr) X329 -5.20 ··· ··· 0.37 7.67 X330 -4.20 ··· ··· 0.08 7.78 X336 -3.50 ··· ··· 1.35 7.75 X405 -4.40 ··· ··· ··· 7.27 X501 -7.40 -0.32 0.15 0.27 7.16 X502 -8.50 -0.50 0.30 0.83 6.80 X505 -4.70 0.50 -1.31 · · · 6.53 X601 -5.10 -0.21 0.12 1.51 7.40 X602 -3.10 -0.76 -0.30 1.63 7.85 X604 -5.50 -1.34 -0.25 0.25 7.42 X607 -6.30 0.89 1.26 2.77 6.97 X608 -3.70 -1.17 -0.45 0.38 7.81 X617 -3.40 2.78 0.04 1.74 7.88 X628 -5.20 -1.99 0.18 0.34 7.67 X646 -3.80 -1.64 0.22 0.36 7.86 X647 -4.60 -1.85 0.10 0.24 7.70 X647 -6.40 -1.20 -0.44 0.08 6.96 X703 ··· ··· ··· ··· 7.79 X704 ··· ··· ··· ··· 7.04 X708 -2.70 · · · 1.27 1.05 8.00 X712 -7.90 · · · -0.19 -0.09 6.78 X715 -4.40 · · · 0.52 0.84 7.79
4.4 Results
We detect a total of 46 XRBs in our sample of six starburst galaxies down to a luminosity
35 −1 of LX = 7.8 × 10 erg s (source X706). The XRBs have luminosities in the range
36 39 −1 38 −1 of 10 − 10 erg s , with the mean and median being LX = 2.7 × 10 erg s and
37 −1 LX = 1.1 × 10 erg s respectively. Sources X204 and X505 are candidate ultra luminous
> 39 −1 X-ray sources (ULXs; LX ∼ 10 erg s ). Figure 4-2 shows the two X-ray colors H1 vs. H2, which are also presented in Table 4.3.
The two model lines represent disk blackbody (red) and power law (green) emission, which were described in detail in Chapter 2. The orange line shows the effect of adding absorption
84 1.0
increasing hardness
DBB+ABS 0.5
0.0 Soft Color (H1)
-0.5 PL
DBB
increasing hardness -1.0
-1.0 -0.5 0.0 0.5 1.0 Hard Color (H2)
Figure 4-2 X-ray color-color diagram of X-ray binaries in all six starburst galaxies. Theoretical tracks are defined as in Figure 2.3 (see Section 2.2.3 for details). to the disk blackbody models. These two models and the associated physical processes were discussed in greater details in Section 2.2.3. It is interesting to note that Figure 4-2 reveals that the majority of XRBs occupy the so-called “absorbed” section of the diagram H1 ∼> 0 (See Section 4.5.2 for more details).
We compute the separation between the 46 XRBs and their closest clusters. We used
similar techniques described in Section 2.5.1. The results are presented in Figure 4-3 as
a cumulative distribution of the separations between the binaries and the closest cluster.
The figure shows that there are ≈ 20 XRBs that are within 100 pc of a star cluster. As
we found in NGC 4449, almost all XRBs are located within 400 pc of a star cluster. Six
X-ray sources are spatially coincident with star clusters (within the adopted 2σ positional uncertainty of 1.0′′). In comparison, we show the cumulative distribution of the separations
85 1.0
0.8
0.6 Fraction
0.4
0.2
0.0 10 100 1000 Distance [ pc ]
Figure 4-3 Cumulative distribution of the displacement (in parsecs) between X-ray binaries and star clusters in the six starburst (green line) and the Antennae (red line) galaxies. between the XRBs and their closest clusters in the Antennae. The two distributions are clearly different.
We assume that as we found in NGC 4449, the probability of a chance superposition be- tween an XRB and star cluster is very low. We also compare the distribution of XRB-cluster separations in the six starbursts studied here with that for NGC 4449. The Kolmogorov-
Smirnov (KS) statistic returns a probability of 0.69 that these distributions are drawn from the same parent population. Next, we compare the full sample of seven starburst galaxies, i.e., including NGC 4449, with the distribution of XRB-cluster separations for the 82 sources in the Antennae galaxies from Chapter 3. A KS test probability of only 0.007 suggests that these two distributions are likely different. Figure 4.3 shows that the two distributions are similar at small separations, but begin to deviate at separations larger than ≈ 100 pc, with larger separations found between XRBs and clusters in the Antennae than in the starburst galaxies. The implication of this result is discussed in Section 4.5.3.
86 -10
-8
Supergiants -6 V
M Be stars
-4
-2
0 1 2 3 (V-I)
Figure 4-4 Color magnitude diagram of candidate donor stars in six nearby dwarf starburst galaxies. The solid lines are Padova isochrones (for Z = 0.02) of 6 Myr, 10 Myr and 100 Myr (from top to bottom).
We detect optical point sources (likely the donor stars) within the positional uncertainty of ≈ 30 XRBs. The ages of these individual stars are derived by comparing the colors and luminosities (listed in Table 4.5) with solar metallicity stellar isochrones. The results are presented on a color-magnitude diagram shown on Figure 4-4. The approximate luminosity which separates supergiant donors from lower luminosity Be star donors is also marked in the Figure.
87 4.5 Discussion
4.5.1 The Nature of the Ultra Luminous X-ray Source in
Henize 2-10
We detect three X-ray sources associated with Henize 2-10. Reines & Deller (2012) have
6 recently claimed that there is an intermediate mass black hole (with ≈ 10 M⊙) at the center of this dwarf starburst galaxy. They report that this low luminosity active galactic nucleus is detected as a central unresolved radio source with the Very Large Array (VLA) that is spatially coincident with a hard X-ray point source observed with Chandra (our X204).
They also claim that no star cluster is seen at the location of the nuclear radio source
(within their positional uncertainty) even in deep near-infrared HST observations. We
aligned HST images from both WFPC2 and ACS cameras with the Chandra observations using several background galaxies outside the main body of He 2-10. Interestingly, we find that two star clusters are located within the 1σ positional uncertainty of X204. While the
VLA observations have excellent astrometric accuracy with uncertainties at the level of
0.1′′ × 0.03′′ (Reines & Deller 2012), Chandra’s positional uncertainty is much higher, close to 0.5′′, which places one cluster within 1σ of X204 even for the somewhat different HST -
Chandra alignment determined by Reines & Deller (2012). The ages of these two clusters were determined from UV long-slit spectra taken with the STIS instrument on-board HST
(Chandar et al. 2003) to be ≈ 5 − 6 Myr. Based on these results, we cannot rule out the possibility that X204 actually resides within a very young star cluster. N-body simulations show that this is insufficient time for an intermediate mass black hole to form (Vesperini et al. 2010). If this is the case, the X-ray emission likely originates from an HMXB.
4.5.2 X-ray Colors of XRBs
X-ray colors can be compared with model predictions in order to constrain the types of compact objects in HMXBs (Prestwich et al. 2003), although it can be difficult to make a unique classification because HMXBs tend to go through different “states” related to their
88 complex accretion history. The X-ray colors have the advantage of providing a spectral classification tool when only a limited number of photons are detected from a given source, which is certainly the case for most X-ray population studies in galaxies beyond the Local
Group. In addition, color-color diagrams are relatively assumption-free and, while difficult to study individual XRBs, offer a statistical way to discriminate among different types of possible X-ray sources.
As mentioned in the previous Section, Figure 4-2 shows that the majority of XRB colors in these starburst galaxies are found in the absorbed portion of the X-ray color-color diagram. There are two general reasons an X-ray source might appear here: 1) there is a significant amount of extinction, either foreground or within the binary itself, or 2) the
X-ray spectrum of the binary is intrinsically harder than predicted by the disk blackbody and power-law models. We can rule out foreground extinction, both in our Galaxy and in the host, as the cause. The foreground extinction is not significant for any of our sample galaxies, and a visual inspection of the locations of the HMXBs in the optical HST images shows that they are not located in regions of high extinction such as dust lanes or regions of high star formation. Furthermore, the spectral energy distributions of XRB-host clusters
< are best fit by fairly low values of extinction, with AV ∼ 1 mag. This suggests that the source of the absorption would have to reside within (or very close to) the binary itself.
One simple way to assess whether or not there is absorption within the binary itself is through geometrical arguments. For example, softer X-ray photons could be blocked along sight lines through the accretion disk. Accretion disks should have random orientations relative to us, therefore we would also expect some sight lines with low absorption, although the exact number would depend on the size, thickness and flaring of the accretion disk.
Another possibility is that clumps of material residing in a “cocoon” enshrouding the binary could absorb soft X-rays. If this material is clumpy, we would still expect some sight lines to be unabsorbed. Another argument that suggests that absorption is not responsible for the absorbed X-ray colors is that the supergiant donor stars do not show appreciable extinction.
This could occur if the donor star is not subject to the same absorbing column as the source of the X-rays. Assuming that the XRB orbits are randomly orientated relative to the line of
89 sight, we would expect to observe significant extinction in the donor star colors in at least a few cases. However, this is not what we find. We tentatively suggest that the observed colors of the XRBs implies that their intrinsic spectra are harder than predictions from simple disk blackbody and power-law models would suggest, although more sophisticated simulations should be performed before geometrical effects can be firmly rules out.
4.5.3 Correlation Between the Positions of XRBs and Star
Clusters
Another way to constrain the ages and types of compact objects in the HMXBs is through their proximity to their parent star cluster. We use the ages of star clusters in the surrounding area of the XRBs to constrain the ages of the HMXBs. While the best way to achieve this is by identifying cluster counterparts to HMXBs, it is very likely that an XRB has the same age as the stars and clusters around it. We also employ N-body simulations to help us better understand the connection between HMXBs and star clusters.
We ran an N-body simulation of a star cluster using the STARLAB software, with
4 ≈ 32, 000 stars (≈ 1.2 × 10 M⊙), with a virial radius of 2 pc, a concentration of c = 1, assuming a King model profile. The cluster did not have any primordial binaries. This simulation is similar to those run in Garofali et al. (2012). Figure 4-5 shows the results from our simulation. Here, most NSs are ejected from their parent cluster, while BHs are far more likely to be retained, at least up to ages of 300 Myr (Sen et al. 2012). This suggests that HMXBs that reside within their parent clusters are more likely to have a BH as their compact object than a NS, whereas those that are displaced from their parent clusters are much more likely to have a neutron stars as the compact object. STARLAB (similar to other N-body codes) assumes that the progenitor supernovae of BHs and NSs have similar momentum distributions, and therefore the larger masses of black holes result in lower kick velocities, and hence a higher retention fraction.
Garofali et al. (2012) show that XRBs do not form dynamically within clusters within the first 10 Myr, except in the most extreme, unrealistic conditions. They concluded that
90 Figure 4-5 The number of NSs and BHs, found in and out of clusters, as a function of time. The results are from a single N-body simulation run with 32,000 stars and a standard Kroupa IMF. The top panel shows that neutron stars are much more likely to be ejected from their parent clusters. The bottom panel shows that within clusters, black holes dominate over neutron stars, particularly at ages younger than 20 − 30 Myr.
HMXBs result from primordial binaries present from the cluster’s birth. Preliminary results which include primordial binaries are presented in Garofali et al. (2012); Sen et al. (2012).
Their findings have implications for the observational work presented here, and can be approximately broken into three categories: 1) XRBs coincident with star clusters younger than ≈ 30 Myr, 2) XRBs within 100 pc of a very young ( ∼< 10 Myr) cluster, and 3) XRBs displaced more than 100 pc from young (10 − 100 Myr) clusters.
N-body simulations clearly show (Figure 4-5) that BHs dominate the population of compact objects in very young clusters, at least up to ages of ≈ 30 Myr. Preliminary results from N-body simulations indicate that this result is not sensitive to different initial cluster conditions, such as varying binary fraction and binary period distributions (Sen et al. 2012). Therefore, HMXBs which reside within clusters that are younger than ≈ 30 Myr
have a high probability of having a black hole as the compact object.
The third situation, where HMXBs are located at distances further than 100 pc from
91 10 − 100 Myr old clusters, also provides fairly clear results. Here, N-body simulations show
that a very high fraction, ≈ 80 − 90% (depending somewhat on initial conditions) of NSs
are ejected from their parent clusters with fairly high velocities. This suggests that HMXBs
located further than approximately 100 pc have a high probability of having a neutron star
as the compact object. The second situation, an HMXB located within 100 pc of a very
young cluster, is the most difficult to assess. Here, we expect that the HMXB population
contains binaries with a mix of neutron stars and black holes, although the exact probability
depends on the initial conditions of the clusters, details of stellar evolution, mass loss, etc.
Another caveat regarding the simulations relates to predictions from stellar evolution
models. Modern studies indicate that massive stars with solar metallicity lose mass effi-
ciently through stellar winds, and are more likely to end their lives as NSs rather than BHs
(e.g., Heger et al. (2003)). In this situation, NSs would be produced earlier than shown in
our simulation, although this would not alter the dynamics. However, an earlier formation
and ejection would lead to larger separations than later formation scenarios.
In our sample of six starbursts, we find that six out of 46 (∼ 15%) XRBs are spatially
coincident with star clusters. Of these XRBs, 3 are coincident with very young ( ∼< 10 Myr), 1 with young (10 − 100 Myr), and 2 with intermediate age (≈ 100 − 300 Myr) star clusters.
As discussed in the previous paragraphs, most of these coincident HMXBs are likely to have
BHs as the compact objects. If we look at the closest clusters to XRBs for these six dwarf
starbursts, 24 out of a total of 46 XRBs (52%) are close to very young τ ∼< 10 Myr clusters, 6 (13%) are close to young ≈ 10 − 100 Myr clusters, and 16 (35%) are close to intermediate
age ≈ 100 − 300 Myr clusters. Much like in NGC 4449 we see that ≈ 50% of our X-ray
binary population is likely to have very young parent clusters, suggesting that they formed
quite recently. With only a few exceptions, all of these very young HMXBs are located
within 100 pc of very young star clusters, consistent with having similar, young ages.
The cumulative distribution of XRB-cluster separations (Figure 4-3) provides informa-
tion on the physical ejection mechanisms of HMXBs from their host clusters. In Section 4.4,
we found that the separations for XRBs in the merging Antennae galaxies appears to be
different from that in seven dwarf starbursts. This is an intriguing result, which implies
92 that different (or additional) physical processes may play a role in displacing XRBs from their parent star clusters. As seen from Figure 4-3, the separations for both Antennae and starburst galaxies are similar up to the first ∼ 100 pc. At larger separations the two distri- butions deviate from one another, with the Antennae having more XRBs that are further away from clusters. One possible explanation for this difference is that some NS HMXBs in the Antennae received a higher kick velocity than those in the dwarf starbursts.
4.5.4 Candidate Donor Stars
While donor stars cannot be directly identified in HMXBs that still reside within their parent clusters, important constraints on donor star masses and ages are possible from deep optical HST imaging of HMXBs that have been ejected from their birth sites.
For 33 of the XRBs in our sample we detect candidate donor stars. We constrain their ages by fitting isochrones to the wide band photometric data obtained from HST. As described in Section 2.6.4, known Be-HMXBs have MV ranging from −2 to −5, while SG-
HMXBs are usually brighter than −6.5. In Figure 4-4 we show the approximate dividing line between these two classes of HMXBs at MV ≈−6. Of the 33 candidate donor stars, 9 fall into the supergiant category. The rest are less luminous Be-HMXB candidates. From
Table 4.5 (and Figure 4-4) we see that the donor star candidates fall in the evolved portion of the CMD rather than along the main sequence, allowing for an estimate of masses and ages. We find that 25 have estimated ages between 10 and 100 Myr, while only 8 are younger than 10 Myr.
4.6 Summary and Conclusions
In this Chapter, we presented the discovery of 46 candidate X-ray binaries in the nearby starburst galaxies He 2-10, NGC 1569, NGC 1705, NGC 3125, NGC 4214, and NGC 5253, from Chandra/ACIS-S observations. We measured count rates, luminosities, and X-ray colors for these sources. We also presented the ages and separations of the closest cluster to each of these XRBs, from multi-band, high resolution, optical imaging taken with WFPC2, 93 ACS, and/or WFC3 cameras on-board the HST. Our main conclusions are:
• In He 2-10, we found that two optically luminous clusters are coincident with X-
ray source X204, within the positional uncertainties. This conclusion differs from
previous claims that this X-ray source does not have an optical counterpart. If the
X-ray emission originates from one of these clusters, than the source is unlikely to be
an intermediate mass black hole as claimed by Reines & Deller (2012).
• The X-ray colors of most XRBs in six starburst galaxies fall in the “absorbed” portion
of the diagram. This is either the result of significant extinction somewhere along the
line-of-sight, or because these sources have intrinsically harder X-ray spectra than
predicted by simple disk blackbody and power-law models. The optical observations
rule out foreground extinction and some simple geometrical arguments appear to rule
out absorption through the accretion disk within the binary itself. We tentatively
conclude that the observed colors of these sources indicate harder intrinsic spectra
than suggested by models that are commonly used to simulate X-ray emission from
these tight binaries.
• Six out of 46 XRBs are coincident with young star clusters, and hence likely host
BHBs. Nearly half of the sources are located further than 100 pc from a cluster.
Dynamical arguments and the results of N-body simulations suggest that these are
likely NS-XRBs. The remaining sources are located less than 100 pc away from a
cluster, and it is harder to assess if these are BH- or NS-binaries.
• A statistical analysis of the distances between XRBs and their closest (presumed
parent) clusters indicates that the HMXBs in the merging Antennae galaxies have a
different distribution of separations from their parent clusters than those in starbursts.
• Approximately 30 donor stars are identified in these starbursts. The majority of them
(25 out of 33) have estimated ages between 10 and 100 Myr. The optical luminosities
suggest that 9 of the donors are supergiants, while the rest are probably Be-XRBs.
94 Chapter 5
Summary and Future Prospects
Previous studies have established that HMXBs, in which a compact object, either a black hole or a neutron star, is accreting material from a young massive donor star, often dominate the high energy output from nearby star-forming galaxies. These sources are known to be spatially correlated with young star-forming regions in the Galaxy (e.g., Bodaghee et al.
(2012)), and have been found close to, but not coincident with, young star clusters in three nearby starburst galaxies (Kaaret et al. 2004). Observations taken with the Chandra X-ray
Observatory are extremely efficient at finding HMXBs, but it has proven difficult thus far to constrain basic properties, such as ages, dynamics, nature of the compact object (BH vs. NS), and donor star masses, for individual sources beyond the Galaxy and Magellanic
Clouds.
This dissertation has made significant progress in determining the basic properties of individual HMXBs in nearby starbursts and in the merging Antennae galaxies by combining high spatial resolution observations in the X-ray and optical. We produced catalogs of
HMXBs for seven nearby starburst galaxies (He 2-10, NGC 1569, NGC 1705, NGC 3125,
NGC 4214, NGC 4449, and NGC 5253), many of them newly discovered, including positions, luminosities, and X-ray colors using data from Chandra. We presented a new catalog of 129 compact star clusters in NGC 4449 from multi-band, high resolution, optical imaging taken with HST , more than doubling the number of previously known clusters in this galaxy. We identified a number of parent star clusters and donor stars in all eight galaxies studied here.
One key discovery is that approximately one-quarter of the HMXBs found in the merging 95 Antennae and in the dwarf starburst NGC 4449 actually reside within, and not just close to, compact star clusters. This result provides the strongest empirical evidence to date that
HMXBs form in star clusters. These results are broadly consistent with a scenario where
HMXBs form in the clusters, but become displaced from their parent clusters, either because
they are dynamically ejected or because the parent cluster has dissolved. The optical
observations allowed us to rule out cluster dissolution as the main mechanism responsible
for the observed displacement between HMBXs and clusters. This suggests that HMXBs
are often dynamically ejected from their birthsites, and implies that they have larger space
motions than if their parent clusters had simply dissolved.
We were able to determine the ages of HMXBs via ages of their parent clusters, using
multi-band observations taken with HST. A second key result from this dissertation is that
the majority of HMXBs in our target galaxies formed quite recently, only ≈ 3 − 8 Myr ago.
We were also able to estimate the masses and sizes of parent clusters to HMXBs in the
Antennae and NGC 4449, and find a large range for host cluster masses, more than a factor
of 100.
Age estimates for HMXBs, combined with their location relative to their parent star cluster, allowed us to determine whether the compact objects in these systems were black holes or neutron stars. Arguments based on predictions from stellar evolution models were presented in a published paper, Rangelov et al. (2011), and in Chapter 2 of this disserta- tion. During the course of this study however, we realized that these predictions may not apply particularly well to stars in binary systems, since they were developed to explain the evolution of massive single stars. Instead, in Chapter 3 (a paper submitted to the Astro- physical Journal for publication) and Chapter 4 (which will be submitted for peer-review), we presented arguments that suggest that dynamics are a better way of determining the nature of these compact objects, and which are independent of any metallicity-dependent evolutionary paths that lead to their formation. These arguments are based on the results of N-body simulations presented in Garofali et al. (2012) and Sen et al. (2012), which show that HMXBs that contain a BH are far more likely to be retained within their parent cluster than those containing NSs, because natal kicks imparted during the supernova explosion
96 that forms the compact object typically fling lower mass neutron stars from their birth sites.
A third key result from this work is that our sample starburst galaxies and the Antennae have formed a population of very young, black-hole binaries. Black-hole binaries are quite rare in the Milky Way, with fewer than a dozen such systems currently known.
This dissertation has laid out methodology that will allow us to determine the de- mographics of HMXB populations in many other nearby, star-forming galaxies, since the galaxies we have studied so far do not cover the full range of morphologies, metallicities, and environments. One preliminary result which demonstrates some of the intriguing dis- coveries is a comparison of the X-ray source populations in the spiral galaxies M51 and
M101. These two galaxies have the same star formation rates, within the errors (Lee et al.
2009), yet very different X-ray binary populations. While M101 has a large number of ob- servable donor stars in HST observations, M51 does not. At the distances of these galaxies
and the detection limit of the HST observations, massive donor stars, either supergiants
or massive main sequence stars, should be detectable. Therefore, X-ray sources that do
not reside within star clusters and which do not have detectable donor stars are likely to
have low-mass donors, and hence to be low-mass X-ray binaries. This indicates that despite
their similar star formation rates, the number of HMXBs in M51 and M101 are significantly
different.
The Chandra and HST public archives host X-ray and optical observations for 17 spirals
(including M51 and M101), dwarf starbursts and merging galaxies, located within ≈ 40 Mpc,
which can be analyzed in the same way as presented in Chapters 2, 3, and 4. With the new
knowledge and techniques that we have developed, these archival observations will be used
in the future to study the formation and evolution of HMXB populations in a large number
of nearby, star-forming galaxies.
97 References
Allende Prieto, C., Lambert, D. L., & Asplund, M. 2001, ApJ, 556, L63
Annibali, F., Aloisi, A., Mack, J., Tosi, M., van der Marel, R. P., Angeretti, L., Leitherer,
C., & Sirianni, M. 2008, AJ, 135, 1900
Annibali, F., Tosi, M., Monelli, M., et al. 2009, AJ, 138, 169
Antoniou, V., Zezas, A., Hatzidimitriou, D., & Kalogera, V. 2010, ApJ, 716, L140
Bastian, N., Trancho, G., Konstantopoulos, I. S., & Miller, B. W. 2009, ApJ, 701, 607
Baumgardt, H., & Kroupa, P. 2007, MNRAS, 380, 1589
Belczynski, K., Kalogera, V., Rasio, F. A., Taam, R. E., Zezas, A., Bulik, T., Maccarone,
T. J., & Ivanova, N. 2008, ApJS, 174, 223
Belczynski, K., Bulik, T., & Klu´zniak, W. 2002, ApJ, 567, L63
Belczynski, K., & Ziolkowski, J. 2009, ApJ, 707, 870
Blaauw, A. 1961, Bull. Astron. Inst. Netherlands, 15, 265
Bodaghee, A., Tomsick, J. A., Rodriguez, J., & James, J. B. 2012, ApJ, 744, 108
Bondi, H., & Hoyle, F. 1944, MNRAS, 104, 273
Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000
Chandar, R., Leitherer, C., Tremonti, C., & Calzetti, D. 2003, ApJ, 586, 939
Chandar, R., Leitherer, C., Tremonti, C. A., et al. 2005, ApJ, 628, 210
Chandar, R., Fall, S. M., & Whitmore, B. C. 2010, ApJ, 711, 1263 98 Chandar, R., et al. 2010, ApJ, 719, 966
Chandar, R., Whitmore, B. C., & Fall, S. M. 2010, ApJ, 713, 1343
Chandrasekhar, S. 1931, MNRAS, 91, 456
Chevalier, C., & Ilovaisky, S. A. 1998, A&A, 330, 201
Clark, D. M., et al. 2005, ApJ, 631, L109
Clark, D. M., et al. 2011, MNRAS, 410, 890
Converse, J. M., & Stahler, S. W. 2011, MNRAS, 410, 2787 de Wit, W. J., Testi, L., Palla, F., Vanzi, L., & Zinnecker, H. 2004, A&A, 425, 937 de Wit, W. J., Testi, L., Palla, F., & Zinnecker, H. 2005, A&A, 437, 247
Dolphin, A. 2000, PASP, 112, 1397
Engelbracht C. W., Rieke G. H., Gordon K. D. et al. 2008, ApJ678, 804
Fabbiano, G. 1989, ARA&A, 27, 87
Fabbiano, G. 2006, ARA&A, 44, 323
Fall, S. M., Chandar, R., Whitmore, B. C. 2005, ApJ, 631, 133
Fall, S. M., Chandar, R., Whitmore, B. C. 2009, ApJ, 705, 453
Fall, S. M., Krumholz, M. R., & Matzner, C. D. 2010, ApJ, 710, L142
Fall, S. M., & Zhang, Q. 2001, ApJ, 561, 751
Farinelli, R., Titarchuk, L., Paizis, A., & Frontera, F. 2008, ApJ, 680, 602
Ferreira, J., Petrucci, P.-O., Henri, G., Saug´e, L., & Pelletier, G. 2006, A&A, 447, 813
Fitzpatrick, E. L. 1999, PASP, 111, 63
99 Freeman, P. E., Kashyap, V., Rosner, R., & Lamb, D. Q. 2002, ApJS, 138, 185
Fryer, C. L. 1999, ApJ, 522, 413
Galloway, D. K., Psaltis, D., Chakrabarty, D., & Muno, M. P. 2003, ApJ, 590, 999
Garc´ıa, B., & Mermilliod, J. C. 2001, A&A, 368, 122
Garofali, K., Converse, J. M., Chandar, R., & Rangelov, B. 2012, submitted for publication
in ApJ
Gelatt, A. E., & Hunter, D. A., & Gallagher, J. S. 2001 PASP, 113, 142
Giacconi, R., Gursky, H., Kellogg, E., Schreier, E., & Tananbaum, H. 1971, ApJ, 167, L67
Giacconi, R., Branduardi, G., Briel, U., et al. 1979, ApJ, 230, 540
Gies, D. R., & Bolton, C. T. 1986, ApJS, 61, 419
Girardi, L., Dalcanton, J., Williams, B., et al. 2008, PASP120, 583
Grevesse, N., & Sauval, A. J. 1998, Space Sci. Rev., 85, 161
Heger, A., Fryer, C. L., Woosley, S. E., Langer, N., & Hartmann, D. H. 2003, ApJ, 591, 288
Holtzman, J., Burrows, C. J., Casertano, S., Hester, J., Trauger, J. T., Watson, A. M., &
Worthey, G. 1995a, PASP, 107, 1065
Hunter, D. 1997, PASP, 109, 937
Kaaret, P., Alonso-Herrero, A., Gallagher, J. S., Fabbiano, G., Zezas, A., Rieke, M. J. 2004,
MNRAS, 348, 28
Karl, S. J., Fall, S. M., & Naab, T. 2011, ApJ, 734, 11
King, I. R. 1966, AJ, 71, 64
Kobulnicky, H. A., Kennicutt, R. C., Jr., & Pizagno, J. L. 1999, ApJ, 514, 544
100 Kundu, A., Maccarone, T. J., & Zepf, S. E. 2007, ApJ, 662, 525
Larsen, S. 1999, A&AS, 139, 393
Larson, R. B. 2001, The Formation of Binary Stars, 200, 93
Lee, J. C., Gil de Paz, A., Tremonti, C., et al. 2009, ApJ, 706, 599
Lee, H., & Skillman, E. 2004, ApJ, 614, 698
Leitherer, C., et al. 1999, ApJS, 123, 3
Liedahl, D. A., Osterheld, A. L., & Goldstein, W. H. 1995, ApJ, 438, L115
Linden, T., Kalogera, V., Sepinsky, J. F., Prestwich, A., Zezas, A., & Gallagher, J. S. 2010,
ApJ, 725, 1984
L´opez-S´anchez, A.´ R., Esteban, C., Garc´ıa-Rojas, J., Peimbert, M., & Rodr´ıguez, M. 2007,
ApJ, 656, 168
Marigo, P., Girardi, L., Bressan, A., et al. 2008, A&A, 482, 883
Ma´ız-Apell´aniz, J., Cieza, L., & MacKenty, J. W. 2002, AJ, 123, 1307
McBride, V. A., Coe, M. J., Negueruela, I., Schurch, M. P. E., & McGowan, K. E. 2008,
MNRAS, 338, 1198
McSwain, M. V., Ransom, S. M., Boyajian, T. S., Grundstrom, E. D., Roberts, Mallory S.
E. 2007, ApJ, 660, 740
Mineo, S., Gilfanov, M., & Sunyaev, R. 2012, MNRAS, 419, 2095
Mitsuda, K., et al. 1984, PASJ, 36, 741
Patnaude, D. J., Fesen, R. A. 2003, ApJ, 587, 221
Pietsch, W., Mochejska, B. J., Misanovic, Z., et al. 2004, A&A, 413, 879
Portegies Zwart, S. F., McMillan, S. L. W., & Makino, J. 2007, MNRAS, 374, 95 101 Poveda, A., Ruiz, J., & Allen, C. 1967, Boletin de los Observatorios Tonantzintla y
Tacubaya, 4, 86
Prestwich, A. H., Irwin, J. A., Kilgard, R. E., Krauss, M. I., Zezas, A., Primani, F., &
Kaaret, P., & Boroson, B. 2003, ApJ, 595, 719
Rangelov, B., Prestwich, A. H., & Chandar, R. 2011, ApJ, 741, 86
Reines, A. E., & Deller, A. T. 2012, ApJ, 750, L24
Remillard, R. A. & McClintock, J. E. 2006, ARA&A, vol. 44, Issue 1, pp.49-92
Riess, A. G., et al. 2011, ApJ, 730, 119
Sana, H., & Evans, C. J. 2011, IAU Symposium, 272, 474
Sen, N., Converse, J. M., Chandar, R., & Rangelov, B. 2012, in prep.
Schreier, E., Levinson, R., Gursky, H., et al. 1972, ApJ, 172, L79
Schweizer, F., et al. 2008, AJ, 136, 1482
Shtykovskiy, P. E., & Gilfanov, M. R. 2007, Astronomy Letters, 33, 437
Sirianni, M., M.J. Jee, N. Bentez, J.P. Blakeslee, A.R. Martel, G. Meurer, M. Clampin, G.
De Marchi, H.C. Ford, R. Gilliland, G.F. Hartig, G.D. Illingworth, J. Mack, and W.J.
McCann 2005, PASP, 117, 1049
Tauris, T. M., & van den Heuvel, E. P. J. 2006, Compact stellar X-ray sources, 623
Thronson, H. A., Jr., Hunter, D. A., Telesco, C. M., Decher, R., & Harper, D. A. 1987,
ApJ, 317, 180
Townsend, L. J., Coe, M. J., Corbet, R. H. D., & Hill, A. B. 2011, MNRAS, 416, 1556
Tremonti, C. A., Calzetti, D., Leitherer, C., & Heckman, T. M. 2001, ApJ, 555, 322
Ubeda,´ L., Ma´ız-Apell´aniz, J., & MacKenty, J. W. 2007, AJ, 133, 932
102 Vesperini, E., McMillan, S. L. W., D’Ercole, A., & D’Antona, F. 2010, ApJ, 713, L41
Westmoquette, M. S., Exter, K. M., Smith, L. J., & Gallagher, J. S. 2007, MNRAS, 381,
894
Whitmore, B. C., Zhang, Q., Leitherer, C., Fall, S. M., Schweizer, F., & Miller, B. W. 1999,
AJ, 118, 1551
Whitmore, B. C., Chandar, R., & Fall, S. M. 2007, AJ, 133, 1067
Whitmore, B. C., et al. 2010, AJ, 140, 75
Whitmore, B. C., & Schweizer, F. 1995, AJ, 109, 960
Whitmore, B. C., & Zhang, Q. 2002, AJ, 124, 1418
Zezas, A., Fabbiano, G., Rots, A. H., Murray, S. S. 2002, ApJ, 577, 710
Zezas, A., Fabbiano, G., Baldi, A., Schweizer, F., King, A. R., Ponman, T. J., Rots, A. H.
2006, ApJS, 166, 211
Zwicky, F. 1957, Berlin: Springer, 1957
103