|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
ENVIRONMENTS OF X-RAY SOURCES IN EXTERNAL GALAXIES
DAVID M. CLARK(
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
S2007 David M. Clark
To my parents and grandparents.
I would like to thank my adviser, Dr. Stephen Eikenberry, for giving me an exciting
thesis project, issuing insightful answers when I had questions and guiding me towards
achieving skills I need to become a solid researcher. I thank the rest of my committee,
Dr. Fred Hamann, Dr. Ata Sarajedini, Dr. Vicki Sarajedini, and Dr. David Tanner, for
their support. I am appreciative to my collaborators Dr. Bernhard Brandl, Dr. Andy Ptak
and Dr. Ed Colbert for their extensive comments on several of my papers, which have
become integral parts of my thesis. I thank Alicol Claw..ph !- r~11~ for providing me with the
spectrum of X-37 and Dr. Hamann and Dr. V. Sarajedini who identified this source as a
background quasar. I would like to recognize my additional collaborators including Dr.
.John Wilson, Dr. .Joseph Carson, Dr. C'!s .Il. -I Henderson, Dr. Tom Hayward, Dr. Don
Barry, and Dr. .J.R. Houck. I thank the two, UF system administrators, K~en Sallot and
David Edmeades, for making sure the computer system in the department ran smoothly. I
am especially thankful to my many friends in graduate school. I am particularly grateful
to Michelle Edwards for her constant support, the many long conversions, and willingness
to proof read my papers. Valerie Mikles also provided me with plenty of good advice
for my research. Last, but far from least, I recognize the constant support of my family
throughout graduate school; my Mom, Sarah, my Dad, Tom, and my brother, Andrew. I
also acknowledge partial funding through an NSF CAREER award (NSF---I:~ :I) and the
NSF grant AST-0507547.
TABLE OF CONTENTS
ACK(NOWLEDGMENTS .......... . .. .. 4
LIST OF TABLES ......... ..... .. 7
LIST OF FIGURES ......... .... .. 8
ABSTRACT ............ .............. 10
1 INTRODUCTION ......... .. .. 12
2 DEEP NEAR-INFRARED IMAGING, AND PHOTOMETRY OF THE ANTENNAE
GALAXIES WITH WIRC ......... .. .. 17
2.1 The Data ......... ... .. 18
2.1.1 Observations ......... .. .. 18
2.1.2 Image Morphology ......... ... .. 19
2.1.3 Cluster Identification ........ ... .. 19
2.1.4 Photometry ......... . .. 20
2.2 Summary ......... ... .. 21
3 INFRARED COUNTERPARTS TO CHANDRA X-RAY SOURCES IN THE
ANTENNAE ............ ............ 23
3.1 Data Analysis ......... . .. 24
3.1.1 Astrometric Frame Ties . ..... .. 24
3.2 Results and Discussion . . .. .. .. 26
3.2.1 Identification of IR Counterparts to Chandra Sources .. .. .. 26
3.2.2 Photometric Properties of the IR Counterparts .. .. .. .. 28
188.8.131.52 Color magnitude diagrams .. ... .. 28
184.108.40.206 Ksluminosity . ..... .. 29
220.127.116.11 Cluster mass estimates ... ... .. .. 30
18.104.22.168 Non-detections of IR counterparts to X-ray sources .. 31
3.3 Conclusions ......... ... .. 32
4 A FIRST ESTIMATE OF THE X-RAY BINARY FRACTION AS A FUNCTION
OF STAR CLUSTER MASS IN A SINGLE GALACTIC SYSTEM .. .. .. 45
4.1 XRB-to-Cluster Mass Fraction . ..... .. 45
4.2 Effects of Age ........ . .. 48
4.3 Comparison with Models ......... .. .. 50
4.4 Summary and Conclusions ........ ... .. 53
5 MULTIWAVELENGTH STUDY OF CHANDRA X-RAY SOURCES IN THE
ANTENNAE ... ......... ............ 59
5.1 Observations and Data Analysis . .... .. 60
5.1.1 Infrared and Optical Imaging . . 60
5.1.2 Identification of Optical Counterparts to Chandra Sources .. .. 61
5.1.3 Photometry ......... . .. .. 63
22.214.171.124 Constant aperture photometry .. .. .. .. 63
126.96.36.199 PSF aperture photometry .... .. .. 64
5.2 Results and Discussion ......... .... .. 65
5.2.1 Spectral Evolutionary Models .... .. . 65
5.2.2 Estimating Av From Hca Equivalent Widths .. .. .. 67
5.3 Conclusions ........ .. .. 69
6 THE ULTRALUMINOUS X-RAY SOURCE X-37 IS A BACKGROUND QUASAR
IN THE ANTENNAE GALAXIES . ...... .. 79
6.1 Observations and Data Analysis . .... .. 80
6.1.1 Infrared and Optical Images ...... ... .. 80
6.1.2 Spectroscopy ......... .. .. 80
6.1.3 Photometry ......... . .. 81
6.2 Discussion ......... . .. .. 82
7 ENVIRONMENTS OF X-RAY POINT SOURCES IN THE DWARF STARBURST
GALAXY NGC 1569 ......... . .. 87
7.1 Observations and Data Analysis . .... .. 87
7.1.1 Infrared Imaging ......... ... .. 87
7.1.2 Astrometric Frame Ties . .... .. 88
7.1.3 Infrared Photometry ....... ... .. 89
7.2 Results and Discussion . ... .. .. .. 90
7.2.1 Identification of IR Counterparts to Chandra Sources .. .. .. 90
7.2.2 Photometric Properties of the IR Counterparts .. .. .. .. 91
188.8.131.52 Color magnitude diagrams .. ... .. 91
184.108.40.206 Ksluminosity . ..... .. 92
7.2.3 XRB-to-Cluster Mass Fraction .... .... . 93
7.2.4 Comparison with Models . ..... .. .. 95
7.3 Conclusions ........ .. .. 95
8 CONCLUSIONS . ..... .... .. .10
REFERENCES . ..._. ......_ .. 114
BIOGRAPHICAL SK(ETCH ......... .. .. 118
LIST OF TABLES
3-1 Common Sources Used for the 2MASS/WIRC Astrometric Frame Tie .. .. 34
3-2 Common Sources Used for the Chandra/WIRC Astrometric Frame Tie .. .. 34
3-3 Potential IR Counterparts to Chandra X-Ray Sources ... .. .. .. 35
3-4 Summary of Potential IR Counterpart Properties. ... ... .. 36
3-5 Summary of K(S-Test Results. ... ... .. 37
5-1 Potential Optical Counterparts to Chandra X-Ray Sources .. .. .. 71
5-2 Bruzual-CHl .) lot Model Fits . .. ... ... .. 72
5-3 X-17 and X-27 Cluster Counterpart Properties .... .. .. 72
6-1 Absolute Magnitude Comparison to Quasars at z m 0.26 .. .. .. 84
7-1 Common Sources Used for the Chandra/FLAMINGOS Astrometric Frame Tie. 97
7-2 Potential IR Counterparts to Chandra X-Ray Sources ... .. .. .. 98
7-3 Summary of Potential IR Counterpart Properties. ... ... .. 99
LIST OF FIGURES
2-1 Antennae Ks versus optical image comparison. .... .. .. 22
:3-1 Antennae IR counterpart to X-:36 X-ray source. .... .. .. :38
:3-2 Subintages of Antennae IR counterparts to X-ray sources. .. .. .. .. :39
:3-3 K,-band Antennae image. ......... .. .. 40
:3-4 Antennae IR color magnitude diagrams. ...... .. . 41
:3-5 MyK histogram of Antennae clusters . .... .. 42
:3-6 Cluster mass histogfram of Antennae clusters. ... .. .. 4:3
:3-7 Histogram of limiting cluster MsK for Antennae. .... .. .. 44
4-1 if as a function of cluster Ks fux. ....... ... .. 54
4-2 if assuming a range of instant age bursts for Antennae clusters. .. .. .. .. 55
4-3 if computed assuming a uniform and a PL cluster age distribution. .. .. .. 56
4-4 if computed assuming cluster ages as derived by 1\engel et al. (2005). .. .. 57
4-5 Suninary of if computed using four different cluster age distributions. .. .. 58
5-1 Subintages of strictly optical counterpart candidates to Antennae X-ray sources. 7:3
5-2 Subintages of IR and optical counterpart candidates to Antennae X-ray sources. 74
5-3 I-hand Antennae imaging with counterparts overlaid. ... .. .. 75
5-4 Example of a Bruzual-CHl I) lot model fit to cluster UBVIKI,. .. .. .. .. 76
5-5 Color-color and Ho~ equivalent width plots used to determine ages for the cluster
counterparts to X-17, X-27 and X-48. . ..... .. 77
5-6 Suninary of Bruzual-CHl .) lot model fits. .... .. .. 78
6-1 U and Ks counterparts to X-:37 ......... ... .. 85
6-2 X-:37 spectrum ......... .. .. 86
7-1 NGC 1569 color image. ......... . .. 100
7-2 NGC 1569, large image with IR counterparts. .... ... .. 101
7-:3 NGC 1569, small image with IR counterparts. .... ... .. 102
7-4 IR counterparts to X-ray sources in field of NGC 1569. ... .. .. 10:3
IR counterparts to X-ray sources associated with NGC 1569. ......
NGC 1569 C\!lI Is comparing magnitude cutoff and 10 SNR cutoff. ....
NGC 1569 Cill~s highlighting counterparts to X-ray sources........
M~s histogram of NGC 1569 clusters. .......
if as a function of cluster Ks fux for NGC 1569. ......
Uniform distribution of ty for NGC 1569. ......
.. .. 104
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
ENVIRONMENTS OF X-RAY SOURCES IN EXTERNAL GALAXIES
David M. Clark
M li- 2007
C'I I!1-: Stephen S. Eikenberry
Star clusters provide a unique opportunity to study both the environments and
progenitors associated with compact objects. Star-forming galaxies are abundant in both
star clusters and X-ray point sources. The latter are candidates for X-ray binaries (XRBs)
containing a compact object left behind after the death of a massive star. I study the
environments of compact objects by focusing on X-ray point sources and massive star
clusters in the two star-forming galaxies, the Antennae and NGC 1569.
I develop a successful technique for this study using the Antennae. I establish an
X-ray/IR astrometric frame tie with an rms positional uncertainty of ~0.5". I find 19
IR counterparts within 1.5" of an X-ray source. Performing an IR photometric study, I
find that the cluster counterparts are more luminous and massive than the general cluster
population in the Antennae. I define the quantity, rl, relating the fraction of observed
X-ray sources per unit mass as a function of cluster mass. I find a constant value of rl
=6 x10-s My 1, which I demonstrate indicates more massive clusters are more likely to
harbor XRBs only because they have more stars.
Using my IR-to-X-ray frame tie as an intermediary, I match Chandra X-ray
positions to HST optical positions. Applying spectral photometric models to IR/optical
counterparts I determine cluster mass, age and metallicity, which further characterize
the environments of Antennae XRBs. My analysis also includes a multiwavelength and
spectroscopic study of the unusual X-ray source, X-37. After finding an optical and
IR counterpart to this X-ray source, an optical/IR specturealdhisocesa
background quasar at a redshift of z = 0.26.
Extending my study to the dwarf, starburst galaxy, NGC 1569, I produce a frame tie
between ground-based, IR and Chandra, X-ray images with an rms positional uncertainty
of 0.2". I then identify seven cluster counterparts within 0.6"1 of an X-ray source. Unlike
the Antennae, I do not find a trend in luminosity or mass for these cluster counterparts.
Computing rl for this galaxy, I find a value of 3.3x10-6 1\~-
During the past few decades, X-ray observations have revealed a wide range in
compact objects front neutron stars and stellar mass black holes to super massive black
holes at the centers of galaxies. While there are abundant theories concerning compact
object formation, the specifics of their origins remain uncertain. For example, how
are compact object properties related to their progenitors mass, angular montentunt,
kinematics, etc.? Furthermore, what types of stars produce compact objects, under what
conditions and in what environments? In this dissertation I will explore some of these
issues by investigating the environments of stellar mass compact objects in nearby, <20
Mpc, galaxies. I will also demonstrate that the IR wavelength regime is a powerful tool for
this type of investigation.
In this study, I will use X-ray sources to identify compact objects. Alany of these
X-ray sources are X-ray binaries (XRBs) consisting of a compact object accreting matter
front a closely orbiting star. Compact objects are inextricably linked to massive stars.
Lada and Lada (2003) showed that these stars preferentially form in young stellar clusters.
Massive stars usually end their lives in supernovae, producing a compact remnant. This
remnant can he kicked out of the cluster due to dynamical interactions, stay behind
after the cluster evaporates, or remain embedded in its central regions. This last case is
particularly interesting as the compact object is still in situ, allowing investigations into
its origins via the ambient cluster population.
Star-fornxing galaxies are ideal targets for this work, as they contain many young
clusters and X-ray sources. Unlike galactic XR Bs, these X-ray sources have more
well-deternxined (or at least uniform) distances, nxinintizing all! ri ~ uncertainty in
luminosity. Comparing clusters containing XRBs to those without thent can contribute
statistics on the formation environments of compact objects. Finding lone XRBs can
provide insight into their kinematics.
In the past decade, Chandra and XMMI observations revealed extremely luminous
(1039-1042 eTgS S-1), off-nuclear X-ray point sources, classified as ultra luminous X-ray
sources (ULXs), in many nearby galaxies. Given that the Eddington luminosity of neutron
stars is ~ 1038 eTgS S-1, ULXs are 10 to 104 t1mes more luminous. Assuming luminosity
scales as the mass of the compact companion, this implies a mass range much larger than
stellar mass black holes (BHs), but smaller than the 106 to 10s A / ones at the centers of
galaxies. Several authors (e.g., Fabbiano, 1989; Zezas et al., 1999; R~oberats and Warwick,
2000; Makishima et al., 2000) -11---- -r these massive (100-10, 000 Me) compact sources
outside galactic nuclei are intermediate mass black holes (IMBHs), a new class of BHs.
While IMBHs could potentially explain the observed high luminosities, other theories
exist as well, including beamed radiation from a stellar mass BH (K~ing et al., 2001);
super-Eddington accretion onto lower-mass objects (e.g., Moon et al., 2003; Begelman,
2002); or supernovae exploding in dense environments (Plewa, 1995; Fabian and Terlevich,
1996). In addition, some background quasars can masquerade as ULXs on the face of a
foreground galaxy (Gutii~rrez and L6pez-Corredoira, 2005).
An ideal target to begin this investigation is the Antennae galaxies (NGC 4038/4039).
This merger is the closest pair of interacting galaxies and contains numerous X-ray sources
and star clusters. High resolution X-ray images using Chandra revealed 49 point sources in
the Antennae (Zezas et al., 2002a). I will assume a distance to the Antennae of 19.3 Mpc ,
which implies 10 sources are ultraluminous X-ray sources (ULXS) with X-ray luminosities,
Lx > 1039 eTgS S-1. COnSidering new observations of red giant stars in the Antennae
indicate a distance of 13.8 Mpc (Saviane et al., 2004), this ultraluminous X-ray source
population could decrease by roughly a half.
HST images of the Antennae display a plethora of star clusters in the spiral arms
and bridge region (Whitmore et al., 1999). Broad and narrow band photometric studies
indicate many of these clusters are massive, ~106 Me, and ~10 Myr old (e.g. Whitmore
and Zhang, 2002; Mengel et al., 2005). While most clusters are young, Whitmore et al.
(1999) indicate two additional populations: 1) a ~-100 1\yr population in the northeast
quadrant and 2) a ~-500 1\yr population that most likely formed during the initial
encounter. These authors also identified several globular clusters in the field around the
Previous X-ray work tried matching X-ray positions directly to the optical (Zezas
et al., 2002b). Due to the small field of view of the WFPC2 camera on HYT, the authors
were unable to locate sources in common and therefore made an absolute astrometric
match between C'htndrce and KY T coordinates. Considering the absolute astrometric
accuracy of WFPC2 is 0.9"-1.5"' (Biretta, 2000), Zezas et al. (2002b) defined matches as
optical sources within 2" of an X-ray source position. In many case, they found random
offsets between clusters and X-ray sources, including ITLXs. They -II---- -r many of these
X-ray sources are runaway XRBs escaping from their parent cluster. Unfortunately, the
chance alignment of unrelated objects is high: their models show 6+2 X-ray sources are
false matches with 8+4 optical sources.
In my approach to finding counterparts to X-ray sources in the Antennae, I chose
IR, J and Ks images, to make a frame tie to the C'htndrce X-ray images from Fabbiano
et al. (2001). Utilizing the similar dust-penetrating properties of these wavelengths,
I demonstrate the power of this approach to finding counterparts to X-ray sources. I
then perform a photometric study of these counterparts and compare this to the general
population of clusters.
One issue I am particularly interested in is the relationship between the observed
XRB frequency and cluster mass. By finding the observed number of XRBs as a function
of cluster mass, I can ;?i something about the XR B formation rate in star clusters.
Recent theoretical models of young, massive cluster evolution provide a framework for
comparison to my observational study. Two in particular, (Oskinova, 2005; Sepinsky et al.,
2005), incorporate XRB formation in their models. Oskinova (2005) uses a population
synthesis code to study the evolution of X-ray emission in young, massive clusters.
Sepinsky et al. (2005) investigate the role of supernova kicks in XRB expulsion front the
parent cluster using the population synthesis code, StarTrack. They also incorporate the
XRB formation rate for a range in cluster mass. In this work, I compare my observed
number of XRBs per cluster mass in the Antennae to that predicted by these models.
I then extend my analysis to optical wavelengths by fitting X-ray positions to
optical, HST positions using my IR frame tie as an intermediary. While a direct
optical/X-ray frame tie is difficult, an optical/IR frame tie is substantially easier. Most
bright KY T sources will show up in the IR, facilitating a coordinate match between
them. Incorporating the previously made IR/X-ray frame tie, I can make an astronietric
solution across all three wavelengths. Combining UBVI and J/K, photometry of cluster
counterparts, I fit spectral evolutionary models to these sources to derive masses, ages and
My niultiwavelength project resulted in an intriguing discovery with respect to the
unusual X-ray source, X-:37 as designated by Zezas et al. (2002a). At the distance to these
interacting galaxies it would have an X-ray luminosity, L,. = 4.5 x10"9ergs s-l, making it
a ITLX. A previous attempt to match X-ray positions directly to KY T positions indicated
that this object has a significant (> 1.0") offset front a nearby optical source (Zezas et al.,
2002b). This spawned discussions of whether this is a runaway X-ray binary escaping from
its parent cluster (Zezas et al., 2002h; Miller et al., 2004; Fabbiano, 2004). But, matching
X-ray positions directly to the optical is difficult due to the crowded KY T field and image
rotation. Here, I use X-:37 to demonstrate the power of my frame tie technique and the
necessity for followup, spectroscopic observations to ITLX counterparts.
I take my niultiwavelength program a step further and apply it to the dwarf,
starburst galaxy, NGC 1569. This system contains two super star clusters (SSCs)
(Arp and Sandage, 1985) and 16 X-ray sources on the face of the galaxy, which range
in luminosity from Lx = ~5x10"-:3X1036 ergs s-l (11 Istin et al., 2002). At a distance of
2.2 Mpc (Israel, 1988), it is much closer than the Antennae and allows me to probe much
lower star cluster masses, down to ~-1000 AI, Using J and Ks observations acquired with
FLAMINGOS on the K(PNO 4-ni telescope, I carry out a photonietric analysis of IR star
cluster counterparts to X-ray sources in this galaxy. I then compare the results to the
vastly different galactic system of the Antennae.
In my dissertation, C'!s Ilters 2 through 4 discuss the Antennae and the niultiwavelength
properties of counterparts to X-ray sources. In C'!s Ilter 5, I present a case study of
the multiply wavelength counterpart to the Antennae X-ray source, X-37. I extend
my research program to NGC1569 in Chapter 6. C'!s Ilter 7 presents a suninary and
conclusions found front my work.
DEEP NEAR-INFRARED IMAGING, AND PHOTOMETRY OF THE ANTENNAE
GALAXIES WITH WIRC
The Antennae galaxies, NGC 40:38/:39 (Arp 244), are probably the best-known
example of a pair of interacting galaxies. At a distance of only 19.3 Mpcl (Whitmore
et al., 1999) the Antennae system has been thoroughly studied over a large range of
Numerous observations that cannot he listed here individually have been made
at far-infrared, sub-nxillinteter and radio wavelengths. They generally agree that most
of the emission at longer wavelengths contes front the highly extincted overlap region.
The largest molecular complexes have masses of (:3 6) x 10slMy typically an order of
magnitude larger than the largest structures in the disks of more quiescent spiral galaxies
(Wilson et al., 2000). These authors also found an excellent correlation between the
CO emission and the 15pm entission seen by ISO (ili .Ilel et al., 1998). Recent nmid-IR
observations at slightly higher spatial resolutions with Spitzer (Wang et al., 2004) showed
that the rate of star formation per unit mass in the active areas is comparable to those in
starburst and some ultra-luntinous galaxies.
The first deep optical analysis of the Antennae with the Wide Field Camera on
HST (Whitmore and Schweizer, 1995) showed over 700 point-like objects. Subsequent
observations with 0 FPC2 (Whitmore et al., 1999) increased the sensitivity by :3
magnitudes in V hand and revealed between 800 and 8000 clusters in four age ranges:
(i) ages of < 5 Myr around the edges of the overlap region and 5 10 Myr in the western
loop, (ii) ages ~ 100 Myr in the northeastern star formation region, (iii) interniediate-age
clusters of ~ 500 Myr and (iv) old globular clusters front the progenitor galaxies. While
Whitmore et al. (1999) and Fritze-v. Alvenslehen (1999) studied the statistical properties
1 assuming Ho = 75kn1l-1 -Mp -l
of the cluster population, Gilbert et al. (2000) and Mengel et al. (2002) investigated the
properties of selected "super star
X-ray observations with Chandra (Zezas et al., 2002a) revealed 49 sources, including
several ultra-luntinous X-ray (ITLX) sources with X-ray luntinosities of Lx > 10"9 erg s-l
s---I -Hi.;~! these are binary accretion sources.
So far, most studies of the star clusters in the Antennae have been focused on a
single wavelength regime with few exceptions: Zhang et al. (2001) studied the relationship
between young star clusters and the interstellar niediunt based on observations ranging
front X-rays to the radio wavelengths, and Whitmore and Zhang (2002) correlated
optically detected star clusters with their radio counterparts front (Neff and ITlvestad,
2000). K~assin et al. (200:3) combined UBVRJHK images to derive extinction maps for
the Antennae and found several red clusters.
In this chapter I discuss the initial data analysis before coninencing with a detailed
investigation of the properties of star clusters associated with X-ray sources in the
Antennae in the following three chapters. I present the deep near-infrared observations of
NGC 40:38/40:39 obtained with 0 IRC and briefly describe the instrument. I then discuss
the infrared morphology of the Antennae galaxies, how I identified star clusters and the
method used to measure photometry of them.
2.1 The Data
My collaborators, Dr. Bernhard Brandl and Dr. Joseph Carson, obtained near-infrared
images of NGC 40:38/40:39 taken on 2002 March 22 using the Wide-field InfraRed
Camera (WIR C) on the Palomar 5-ni telescope. At the time of these observations,
WIRC was being coninissioned and was equipped with an under-sized HAHAII-1 array
(prior to installation of the full-sized HAHXII-2 array in September 2002), providing a
~ 4.7 x 4.7-arentinute field of view (FOV) with ~ 0.25"' pixels ("WIRC-1("-see Wilson
et al. (200:3) for details). Conditions were non-photonietric due to patches of cloud passing
through. Typical seeing-linlited images had stellar full-width at half-nlaxiniun of 1.0"! in
K., and 1.3" in J. We obtained images in both the .7- (1.25pn1) and Ks-band (2.15pnt)
independently. Our observing procedure consisted of a set of 20 randomly dithered
pointing controlled by a simple macro. This method produced a large effective FOV and
greater redundancy. At each position we took two 7.27 s exposures in the K~-band and
eight 29.07 s exposures in the J-hand. The shorter exposures in K., accounted for the high
sky background flux in this hand. The total, on-target exposure resulted in 19.38 s in both
.7 and K.,
2.1.2 Image Morphology
As seen in the K., hand image of the Antennae (Figure 2-1), the two nuclei are
the dominant features. In comparison to the optical, the northern nucleus (NGC 40:38)
appears quite different in the IR than the southern nucleus (NGC 40:39). The prominent
dust lane across the nucleus of NGC 40:38 is lacking in the IR. In the Third Reference
Catalogue of Bright Galaxies (de Vaucouleurs, 1991) NGC 40:38 is classified as a SB(s)ni
pec (BJ = 10.59), while NGC 40:39 is classified as SB(s)ni pec (BJ = 10.69). The K.
image clearly shows the spiral structure of NGC 40:38, a characteristic of Sh-type galaxies.
The overlap region contains a considerable amount of extinction in the optical and is
accentuated by the many dark, dust lanes. The effects of extinction all but disappear in
the IR, revealing a multitude of clusters. This is apparent in Figure 2-1, where I compare
my K., image of the Antennae with an optical image of this galaxy pair.
2.1.3 Cluster Identification
I identified IR point sources as compact 'smudges' in or near the Antennae.
Considering the variety of cluster shapes and complex background, my collaborators
and I felt this was the best method to ID cluster candidates for the photonietric analysis.
While most of these sources are likely associated with the Antennae, I used Table 4 in
Whitmore et al. (1999) to identify eight sources which are foreground stars (see Table
2-1). I also identified an additional source as a background quasar and discuss this object
in C'!s Ilter 6 as well as Clark et al. (2005). After removing these nine contaminates from
my sample, this left me with 224 clusters.
A few sources were extremely faint and I was not able to get accurate photometry
on all 224 clusters. Therefore, I performed aperture photometry on 222 clusters in the
J-band and 221 clusters in the K,-band using a self-written IDL program. I found that
the full width at half maximum (FWHM) was 3.5 pixels (0.9") in the K, image and 4.6
pixels (1.2") in the J band. I used a photometric aperture of 5-pixel radius in K, band,
and 6-pixel radius in J band, corresponding to ~ 3a of the Gaussian PSF.
Background subtraction is both very important and very difficult in an environment
such as the Antennae due to the brightness and complex structure of the underlying
galaxies and the plethora of nearby clusters. In order to address the uncertainties in
background subtraction, I measured the background in two separate annuli around
each source: one from 9 to 12 pixels and another from 12 to 15 pixels. Due to the high
concentration of clusters, crowding became an issue. To circumvent this problem, I
emploi- II the use of sky background arcs instead of annuli for some sources. These
were defined by a position angle and opening angle with respect to the source center.
All radii were kept constant to ensure consistency. In addition, nearby bright sources
could shift the computed central peak position by as much as a pixel or two. If the
centroid position determined for a given source differed significantly (> 1 pixel) from the
apparent brightness peak due to such contamination, I forced the center of all photometric
apertures to be at the apparent brightness peak. For both annular regions, I calculated
the mean and median backgrounds per pixel.
Multiplying these by the area of the central aperture, these values were subtracted
from the flux measurement of the central aperture to yield 4 flux values for the source in
terms of data numbers (DN). Averaging these four values provided me with a flux value
for each cluster. I computed errors by considering both variations in sky background, asky,
and Poisson noise, wasd. I computed asky by taking the standard deviation of the four
measured flux values. I then calculated the expected Poisson noise by scaling DN to e-
using the known gain of WIRC (2 e- DN-1, Wilson et al., 2003) and taking the square
root of this value. I added both terms in quadrature to find the total estimated error in
I then calibrated my photometry using the bright 2MASS star '2MASS 12014790-1851156'
at 12"01m47.90", -18 51ml5.7" which is listed in the 2MASS database with J = 13.065,
aJ = 0.02 1!n I and Ks = 12.771, aKs = 0.02 mag. This star was in the WIRC FOV for
the Antennae observations.
In this chapter I presented near-IR, J and K, observations of the Antennae originally
acquired by my collaborators, Dr. Bernhard Brandl and Dr. Joseph Carson. I discussed
the data reduction, cluster identification and photometry of these sources in this galaxy
pair. In the next two chapters I will perform a detailed study of the photometric
properties of the IR counterparts to X-ray sources.
Figure 2-1. Antennae, WIRC Ks image on the left and UBVI optical image on the
right. Both image are ~-4'x4'. Notice the prominent dust lanes in the optical
disappear in the IR image.
INFRARED COUNTERPARTS TO CHANDRA X-RAY SOURCES IN THE
Recently, high resolution X-ray images using Chandra have revealed 49 point sources
in the Antennae (Zezas et al., 2002a). I will assume a distance to the Antennae of 19.3
Mpc (for Ho=75 km s-l Mpc-l), which implies 10 sources have X-ray luminosities greater
than 1039 eTgS S-1. COnSidering new observations of red giant stars in the Antennae
indicate a distance of 13.8 Mpc (Saviane et al., 2004), I point out this ultraluminous
X-ray source population could decrease by roughly a half. Typically, masses of black
holes produced from standard stellar evolution are less than ~ 20 Me (e.g., Fryer and
K~alogera, 2001). The Eddington luminosity limit implies that X-ray luminosities > 1039
ergs s-l correspond to higher-mass objects not formed from a typical star. Several
authors (e.g., Fabbiano, 1989; Zezas et al., 1999; Roberts and Warwick, 2000; Makishima
et al., 2000) -11---- -r these massive (10--1000 Meo) compact sources outside galactic
nuclei are intermediate mass black holes (IMBHs), a new class of BHs. While IMBHs
could potentially explain the observed high luminosities, other theories exist as well,
including beamed radiation from a stellar mass BH (K~ing et al., 2001); super-Eddington
accretion onto lower-mass objects (e.g., Moon et al., 2003; Begelman, 2002); or supernovae
exploding in dense environments (Plewa, 1995; Fabian and Terlevich, 1996).
Compact objects tend to be associated with massive star formation, which is strongly
suspected to be concentrated in young stellar clusters (Lada and Lada, 2003). Massive
stars usually end their lives in supernovae, producing a compact remnant. This remnant
can be kicked out of the cluster due to dynamical interactions, stay behind after the
cluster evaporates, or remain embedded in its central regions. This last case is of
particular interest to me as the compact object is still in situ, allowing me to investigate
its origins via the ambient cluster population. The potential for finding such associations
is large in the Antennae due to large numbers of both X-ray point sources and super star
clusters; a further incentive for studying these galaxies.
In C'!s Ilter 2, I presented J and K, photometry of ~220 clusters in the Antennae.
My analysis of (J K,) colors indicated that many clusters in the overlap region suffer
from 9-10 mag of extinction in the V-band (Brandl et al., 2005). This result contrasts
with previous work by Whitmore and Zhang (2002) who associated optical sources
with radio counterparts in the Antennae (Neff and Ulvestad, 2000) and argued that
extinction is not large in this system. Here, I continue the analysis of these Antennae IR
images by making a frame tie between the IR and Chandra X-ray images from Fabbiano
et al. (2001). Utilizing the similar dust-penetrating properties of these wavelengths, I
demonstrate the power of this approach to finding counterparts to X-ray sources. By
comparing the photometric properties of clusters with and without X-ray counterparts, I
seek to understand the cluster environments of these X-ray sources. In 93.1 I discuss the
IR astrometric frame tie to Chandra X-ray images. ~3.2 explains our matching technique
and the photometric properties of the IR counterparts. I conclude with a summary of my
results in ~3.3.
3.1 Data Analysis
3.1.1 Astrometric Frame Ties
The relative astrometry between the X-ray sources in NGC 4038/4039 and images at
other wavelengths is crucial for successful identification of multi-wavelength counterparts.
Previous attempts at this have suffered from the crowded nature of the field and confusion
between potential counterparts Zezas et al. (2002b). However, the infrared wave band
offers much better hopes for resolving this issue, due to the similar dust-penetrating
properties of photons in the Chandra and K, bands. (See also Brandl et al. (2005) for a
comparison of IR extinction to the previous optical/radio extinction work of Whitmore
and Zhang (2002).) I thus proceeded using the infrared images to establish an astrometric
frame tie, i.e. matching Chandra coordinates to IR pixel positions.
As demonstrated by Bauer et al. (2000), I must take care when searching for X-ray
source counterparts in crowded regions such as the Antennae. Therefore, my astrometric
frame tie used a simple approach hased on solving a two-dimensional linear mapping
function relating right ascension and declination coordinates in one image with :r and y
pixel positions in a second image. The solution is of the form:
rl = axri + but + c, (:31)
di = dxri + eux + f (:32)
Here rl and dr are the right ascension and declination, respectively, for a single
source in one frame corresponding to the :ri and yl pixel positions in another frame.
This function considers both the offset and rotation between each frame. Since I am
interested in solving for the coefficients a-f elementary linear algebra indicates we need
six equations or three separate matches. Therefore, I need at least three matches to fully
describe the rms positional uncertainty of the frame tie.
I first used the above method to derive an approximate astrometric solution for
the WIRC Ks image utilizing the presence of six relatively bright, compact IR sources
which are also present in images from the Two Micron All-Sky Survey (2MASS). We
calculated pixel centroids of these objects in both the 2MASS and WIRC images, and
used the 2MASS astrometric header information to convert the 2MASS pixel centroids
into RA and Dec. These sources are listed in Table :3-1. Applying these six matches to my
fitting function I found a small rms positional uncertainty of 0.2", which demonstrates an
accurate frame tie between the 2MASS and WIRC images.
Using the 2MASS astrometric solution as a baseline, I identified seven clear matches
between C'htndrce and WIR C sources, which had bright compact IR counterparts with no
potentially confusing sources nearby (listed in Table :3-2). I then applied the procedure
described above, using the C'htndrce coordinates listed in Table 1 of Zezas et al. (2002a)
(see that reference for details on the C'htndrce astrometry) and the WIR C pixel centroids,
and derived the astrometric solution for the IR images in the X-ray coordinate frame.
For the 7 matches, I find an rms residual positional uncertainty of ~ 0.5"' which I adopt
as my la position uncertainty. I note that the positional uncertainty is an entirely
emp~irical quantity. It shows the achieved uncertainty in mapping a target from one image
reference frame to the reference frame in another band, and automatically incorporates all
contributing sources of uncertainty in it. These include, but are not limited to, systematic
uncertainties (i.e. field distortion, PSF variation, etc. in both Chandra and WIRC) and
random uncertainties (i.e. centroid shifts induced by photon noise, flatfield noise, etc. in
both Chandra and WIRC). Thus, given the empirical nature of this uncertainty, I expect
it to provide a robust measure of the actual mapping error-an expectation which seems to
be borne out by the counterpart identification in the following section.
To further test the accuracy of my astrometric solution I explored the range in rms
positional uncertainties for several different frame ties. Specifically I picked ten IR/X-ray
matches separated by <1", which are listed in Table 3-3 (see $3.1). Of these ten I chose
24 different combinations of seven matches resulting in 24 unique frame ties. Computing
the rms positional uncertainty for each, I found a mean of 0.4" with a la uncertainty of
0.1". Considering the rms positional uncertainty for the frame tie used in my analysis falls
within la of this mean rms, this indicates I made an accurate astrometric match between
the IR and X-ray.
3.2 Results and Discussion
3.2.1 Identification of IR Counterparts to Chandra Sources
I used the astrometric frame tie described above to identify IR counterpart candidates
to Chandra X-ray sources in the WIRC Ks image. I restricted my analysis to sources
brighter than K, ~ 19.4 mag. This is the K, sensitivity limit which I define in my
photometric analysis below (see $220.127.116.11). Using the 0.5"' rms positional uncertainty of my
frame tie, I defined circles with 2a and 3a radii around each Chandra X-ray source where
I searched for IR counterparts (Figure 3-1). If an IR source lay within a 1.0"! radius (2a)
of an X-ray source, I labeled these counterparts as -1I nisg!". Those IR sources between
1.0"! and 1.5"' (2 3a) from an X-ray source I labeled as p.. --h!. ~" counterparts. I found
a total of 13 strong and 6 possible counterparts to X-ray sources in the Antennae. These
sources are listed in Table 3-3 and shown in Figure 3-2. Of the 19 X-ray sources with
counterparts, two are the Antennae nuclei (Zezas et al., 2002a), one is a background
quasar (Clark et al., 2005), and two share the same IR counterpart. Therefore, in my
analysis of cluster properties, I only consider the 15 IR counterparts that are clusters.
(While X-42 has two IR counterparts, I chose the closer, fainter cluster for my analysis.)
I then attempted to estimate the level of "contamination" of these samples due to
chance superposition of unrelated X-ray sources and IR clusters. This estimation can
be significantly complicated by the complex structure and non-uniform distribution of
both X-ray sources and IR clusters in the Antennae, so I developed a simple, practical
approach. Given the < 0.5"' rms residuals in our relative astrometry for sources in Table
3-2, I assume that any IR clusters lying in a background annulus with radial size of
2.0"-3.0"1 (4 6a) centered on all X-ray source positions are chance alignments, with
no real physical connection (see Figure 3-1). Dividing the total number of IR sources
within the background annuli of the 49 X-ray source positions by the total area of these
annuli, I find a background IR source surface density of 0.02 arcsecond-2 HarT CitTaTGr
X-ray sources. Multiplying this surface density by the total area of all -1I nisg!" regions
(1.0"! radius circles) and "possible" regions (1.0"-1.5"' annuli) around the 49 X-ray source
positions, I estimated the level of source contamination contributing to my -1I nlig!" and
"p<--t!.IR counterpart candidates. I expect two with a la uncertainty of +0.2/-0.1]
of the 13 -1 in isg" counterparts to be due to chance superpositions, and three with a la
uncertainty of +0.5/-0.31 of the six "p< .--h! I counterparts to be chance superpositions.
1 Found using confidence levels for small number statistics listed in Tables 1 and 2 of
This result has several important implications. First of all, it is clear that I have a
significant excess of IR counterparts within 1.0"! of the X-ray sources-13, where I expect
only two in the null hypothesis of no physical counterparts. Even including the p.-u
counterparts out to 1.5", I have a total of 19 counterparts, where I expect only five front
chance superposition. Secondly, this implies that for any given -1Inisg"~ IR counterpart,
I have a probability of ~ 85' (11/1:3 with a lo- uncertainty of 0.:31 ) that the association
with an X-ray source is real. Even for the "p< .--h!I I counterparts, the probability of true
association is ~-50' These levels of certainty are a tremendous intprovenient over the
X-ray/optical associations provided by Zezas et al. (2002b), and are strong motivators for
follow-up nmulti-wavelength studies of the IR counterparts. Finally, I can also conclude
from strong concentration of IR counterparts within ~ 1" of X-ray sources that the frame
tie uncertainty estimates described above are reasonable.
Figure :3-3 is a 4.3' x 4.3' K. image of the Antennae with X-ray source positions
overlaid. I designate those X-ray sources with counterparts using red circles. Notice that
those sources with counterparts lie in the spiral arms and bridge region of the Antennae.
Since these regions are abundant in star formation, this seems to indicate many of the
X-ray sources in the Antennae are tied to star formation in these galaxies.
3.2.2 Photometric Properties of the IR Counterparts
18.104.22.168 Color magnitude diagrams
Using the 219 clusters that had both .7 and K., photometry, I made (.7 Ks) versus
K., color magnitude diagrams (Figure :3-4). I estimated a sensitivity limit by first finding
all clusters with signal-to-noise ~ So-. The mean .7 and K., magfnitudes for these clusters
were computed separately and defined as cutoff values for statistical analyzes. This yielded
19.0 nmag in .7 and 19.4 ning in K., I note that the X-ray clusters are generally bright in
the IR compared to the general population of clusters. While the IR counterpart for one
X-ray source (X-:32) falls below my J-hand sensitivity limit, its K., magnitude is still above
the K. cutoff. Therefore, I retained this source in our analysis.
I then broke down the X-ray sources into three luminosity classes (Figure 3-4).
I took the absorption-corrected X-ray luminosities, Lx, as listed in Table 1 of Zezas
et al. (2002a) for all sources of interest. These luminosities assumed a distance to the
Antennae of 29 Mpc. I used 19.3 Mpc (for Ho=75 km s-l Mpc-l) instead and so divided
these values by 2.25 as -II_t---- 4.1 in Zezas et al. (2002a). I defined the three X-ray
luminosities as follows: Low Luminosity X-ray sources (LLXs) had Lx < 3 x1038 eTgS
s-l, High Luminosity X-ray sources (HLXs) were between Lx of 3x1038eTgS S-1 and
1x1039 eTgS S-1, While Lx > 1x1039 eTgS S-1 Were Ultra-Luminous X-ray Sources (ULXs).
In Figure 3-4 I designate each IR counterpart according to the luminosity class of its
corresponding X-ray source. There does not appear to be a noticeable trend in the IR
cluster counterparts between these different groupings.
To further study the properties of these IR counterparts, I calculated M~Ks foT all IR
clusters. I calculated reddening using the observed colors, (J Ks)obs, (hence forth the
"color method"). Assuming all clusters are dominated by O and B stars, their intrinsic
(J K,) colors are ~0.2 mag. Approximating this value as 0 1!! I_ this allowed me to
estimate AKs aS ~ (J K,)obs/1.33 using the extinction law defined in Cardelli et al.
(1989, hereafter CC11L). Since these derived reddenings are biased towards young clusters,
older clusters will have abnormally high Ks luminosities.
For IR counterparts to X-ray sources, I also computed X-ray-estimated AK, using
the column densities, NVH, liSted in Table 5 of Zezas et al. (2002a). Here, NVH is derived
by fitting both a Power Law (PL) and Raymond-Smith (RS Raymond and Smith, 1977)
model to the X-ray spectra. Using the CC11L law, AKs iS defined as 0.12Av. Taking Av =
5 x 10-22 mag cm2 1H, I COuld then derive AKs
Then I compared AK, calculated using the "color method" to AK, found using the
above two NVH models. I found the "color method" matched closest to NVH(PL) for all
except one (the cluster associated with Chancine source :32 as designated in Zezas et al.
In Figure :3-5, I plot histograms of the distribution of Ks-band luminosity, ~Ks.
Figure :3-5 di-pl oni~ all clusters as well as over plotting only those with X-ray counterparts.
Notice that the clusters with associated X-ray sources look more luminous. To study
whether this apparent trend in luminosity is real, I compared these two distributions using
a two-sided K~olmogorov-Snxirnov (K(-S) test. In my analysis, for statistical purposes,
I only included clusters below A~K, < -13.2 nmag. Restricting my study to sources
with t;ood" photometry, I first defined a limit in Ks, 18.2 1!! .:- using the limiting J
magnitude, 19.0 ning as stated above, and, since the limit in K. is a function of cluster
color, the median (J Ks) of 0.8 nmag. Subtracting the distance modulus to the Antennae,
:31.4 1!! .-- front this K., limit, I computed our cutoff in A~K,. Since all clusters with
X-ray sources fall helow this cutoff, our subsaniple is sufficient to perform a statistical
The K(S test yielded a D-statistic of 0.37 with a probability of :3.2 x 10-2 that the
two distributions of clusters with and without associated X-ray sources are related.
Considering the separate cluster populations as two probability distributions, each can he
expressed as a cumulative distribution. The D-statistic is then the absolute value of the
nmaxiniun difference between each cumulative distribution. This test indicates that those
clusters with X-ray counterparts are more luminous than most clusters in the Antennae.
22.214.171.124 Cluster mass estimates
Whitmore et al. (1999) found '71' of the bright clusters observed with the Hubble
Space Telescope have ages <20 Myr. Therefore, in this study I will assume all clusters
are typically the same age, ~-20 Myr. This allows me to make the simplifying assumption
that cluster mass is proportional to luminosity and ask: Does the cluster mass affect the
propensity for a given progenitor star to produce an X-ray binary? I estimated cluster
mass using K., luminosity (AhK,). Since cluster mass increases linearly with flux (for
an assumed constant age of all clusters), I converted M~Ks to flux. Using these data as
binned in the M~K, histogram (Figure 3-5), I calculated an average flux per bin. By
computing the fraction of clusters per average flux, I am in essence asking what is the
probability of finding a cluster with a specific mass. Since those clusters with X-ray
sources are more luminous, I expect a higher probability of finding an X-ray source in a
more massive cluster. As seen in Figure 3-6, this trend does seem to be true. Applying a
K(S-test between the distributions for all clusters and those associated with X-ray sources
for clusters below the M~Ks completeness limit defined in the previous section, I find a
D-statistic of 0.66 and a probability of 7.2 x 10-3 that they are the same. Hence, the two
distributions are distinct, indicating it is statistically significant that more massive clusters
tend to contain X-ray sources.
While I assume all clusters are ~20 Myr above, I note that the actual range in ages
is -1-100 Myr (Whitmore et al., 1999). Bruzual-CHl I) ht spectral photometric models
(Bruzual and C'I I) h~,t, 2003) indicate that clusters in this age range could vary by a factor
of roughly 100 in mass for a given Ks luminosity. Thus, I emphasize that the analysis
above should be taken as -II__- -r11.- rather than conclusive evidence, and note that in
C'!. Ilter 4 (see also Clark et al., 2007b) I explore this line of investigation and the impacts
of age variations on the result in depth.
126.96.36.199 Non-detections of IR counterparts to X-ray sources
To assess whether my counterpart detections were dependent on reddening or their
intrinsic brightness, I found limiting values for M~Ks for those X-ray sources without
detected IR counterparts. I achieved this by setting all clusters Ks magnitudes equal to
my completeness limit defined for the C \!lls (19.4 in .-z. see $188.8.131.52) and then finding
M~K,(lim) for each using AK, calculated for that cluster. Since M~K,(lim) is theoretical and
only depends on re ad. 1.11.r I could now find this limit for all X-ray sources using an AKs
estimated from the observed NVH ValueS. Thus I considered all IR counterparts (detections)
and those X-ray sources without a counterpart (nondetections). If nondetections are
due to reddening there should not exist a difference in M~K,(lim) between detections and
nondetections. In contrast, if nondetections are intrinsically fainter, I expect a higher
M~K,(lim) for these sources. In the case of detections, I considered reddening from both
the "color method" and the NVH(PL) separately. I could only derive nondetections using
NVH(PL) reddening. Figure 3-7 shows M~K,(lim) appearS higher for all nondetections. To
test if this observation is significant, I applied a K(S-test to investigate whether detections
and nondetections are separate distributions. I found a D-statistic of 0.82 and probability
of 8.8 x 10-6 that these two distributions are the same using the "color method" for
detect ions. Considering the NVH(PL) reddening method for detect ions instead, the
D-statistic drops to 0.48 and the probability increases to 3.9 x 10-2. Since both tests
indicate these distributions are distinct, the observed high M~K,(lim) for nondetections
seems to be real. This leads to the conclusion that these sources were undetected because
they are intrinsically IR-faint, and that reddening does not pll li the dominant role in
I summarize these statistics in Table 3-4. Here I calculated the mean K,, (J Ks),
and M~K, for three different categories: 1) all clusters, 2) clusters only connected with
X-ray sources, and 3) these clusters broken down by luminosity class. I also include
uncertainties in each quantity. Notice that the IR counterparts appear brighter in K,
and intrinsically more luminous than most clusters in the Antennae, although there is no
significant trend in color. I also summarize the above K(-S test results in Table 3-5.
I have demonstrated a successful method for finding counterparts to X-ray sources in
the Antennae using IR wavelengths. I mapped Chandra X-ray coordinates to WIRC pixel
positions with a positional uncertainty of ~ 0.5'. Using this precise frame tie I found 13
-1Inisg!" matches (< 1.0"! separation) and 5 p.. --!Ible" matches (1.0"! 1.5"' separation)
between X-ray sources and IR counterparts. After performing a spatial and photometric
analysis of these counterparts, I reached the following conclusions:
1. I expect only 2 of the 13 -1 in isg" IR counterparts to be chance superpositions.
Including all 19 IR counterparts, I estimated 5 are unrelated associations. Clearly, a large
ill Ui~ R~y of the X-ray/IR associations are real.
2. The IR counterparts tend to reside in the spiral arms and bridge region between
these interacting galaxies. Since these regions contain the heaviest amounts of star
formation, it seems evident that many of the X-ray sources are closely tied to star
formation in this pair of galaxies.
3. A K. vs. (J Ks) C111)> reveal those clusters associated with X-ray sources are
brighter in K. but there does not seem to be a trend in color. Separating clusters by the
X-ray luminosity classes of their X-ray counterpart does not reveal any significant trends.
4. Using reddening derived (J Ks) colors as well as front X-ray-derived NVH, I found
K(,-band luminosity for all clusters. A comparison reveals those clusters associated with
X-ray sources are more luminous than most clusters in the Antennae. A K(S-test indicates
a significant difference between X-ray counterpart clusters and the general population of
5. By relating flux to cluster luminosity, sintplistically assuming a constant age for
all clusters, I estimated cluster mass. Computing the fraction of clusters per average flux,
I estimated the probability of finding a cluster with a specific mass. I find more massive
clusters are more likely to contain X-ray sources, even after I normalize by mass.
6. I computed a theoretical, limiting Mgs for all counterparts to X-ray sources in
the Antennae using X-ray-derived reddenings. Comparing detections to non-detections, I
found those clusters with X-ray source are intrinsically more luminous in the IR.
In the next chapter in which I explore the effects of cluster mass on XRB formation
rate (Clark et al., 2007a), I will investigate the effects of age on cluster luminosity and
hence our cluster mass estimates.
Table 3-1: Common Sources Used for the 2MASS/WIRC Astrometric Frame Tie
12 01 51.66
12 01 50.40
12 01 54.56
12 01 54.95
-18 51 34.7
-18 52 12.2
-18 53 04.0
-18 53 05.8
" Units are magnitudes, from WIRC photometry. Values in parentheses indicate
Table 3-2: Common Sources Used for the Chandra/WIRC Astrometric Fr-ame Tie
-18 52 04.80
-18 51 46.60
-18 52 03.20
-18 53 11.10
-18 53 03.20
-18 52 14.00
-18 53 15.10
Chandra Source ID 0
24 .. .. .. .. .. .. .
34 .. .. .. .. .. .. .
36 .. .. .. .. .. .. .
37 .. .. .. .. .. .. .. .
" ID numbers follow the naming convention of Zezas et al. (2002a).
b Units are magnitudes, from WIRC photometry. Values in parentheses indicate
Table 3-3: Potential IR Counterparts to Chandra X-Ray Sources
Chandra Src ID 0
11. .. .. .. .. .. .. .
24. .. .. .. .. .. .. .
,33. .. .. .. .. .. .. .
37. .. .. .. .. .. .. .
42. .. .. .. .. .. .. .
25. .. .. .. .. .. .. .
" ID numbers follow the naming convention of Zezas et al. (2002a)
b Chandra coordinates with an uncertainty of 0.5" (Zezas et al., 2002a)
c Positional offsets in units of seconds of arc from the Chandra coordinates to the
nates of the proposed counterpart
SUnits are magnitudes, from WIRC photometry. Values in parentheses indicate uncertainties in
the final listed digit.
Table :3-4: Suninary of Potential IR Counterpart Properties.
Category~ K eg a (J -K) oap_ Afx ag a
all clusters 16.72 0.08 0.82 0.03 -15.:33 0.09
X-ray sources 15.72 0.27 0.95 0.11 -16.30 0.35
LLX 15.85 0.36 0.84 0.11 -16.16 0.39
HLX 15.09 0.37 1.1:3 0.28 -17.14 0.54
ITLX 16.82 0.55 0.88 0.02 -14.67 0.51
" IUncertainties in each value.
Table 3-5: Summary of K(S-Test Results.
3.2 x 10-2
9.6 x 10-2
8.8 x 10-6
3.9 x 10-2
M\Ks(lim) (NH) C
" Probability distributions are the
b K~olmogorov-Smirnov D-statistic.
c C \!: "color method" for detec-
tions, NH: NVH(PL) method for detec-
tions. See text for details.
Figure 3-1. IR counterpart to X-36 overlaid with areas of positional uncertainty centered
on the X-ray source position. Black: 1.0"! radius circle for strong sources, red:
1.0"-1.5" annulus for possible sources, and blue: 2.0"-3.0"1 annulus used to
estimate background source contamination.
X-24 (N nucleus) X-2P4 X-29 S nuiles. 51
Figure 3-2. Subimages highlighting X-ray positions for all X-rays sources with IR
counterparts. Positional error circles are 1.0"! in radius. Image field-of-view is
10.0"x10.0"1 and north is up. I label -lIning!" matches in black and p.. --!Ihle"
matches in red.
Figure 3-3. K,-band image of the Antennae showing positions of X-ray sources. The
field-of-view is 4.3 x 4.3'. Red circles designate those sources with IR
counterparts. Notice these tend to reside in the spiral arms and bridge region
between the galaxies.
-1 0 1 2 3 4
X ULX d
../ q .)
gw/ X-ray counterparts
-1 0 1
(upper left) (J K,) vs. Ks C111)> for all clusters with photometry. In each of
the three plots, the dashed line shows my cutoff for statistical analyzes. (upper
right) Here clusters with X-ray counterparts are designated. Note that one
IR counterpart falls below my sensitivity limit. Since its K, magnitude is still
above the K, cutoff, I retained this source in my analysis. (lower right) These
clusters are broken down by their luminosity class. LLX: Lx < 3x10 ergs
s-l, HLX: 3x10 ergs s-l < Lx < 1x10 'ergs s-l, and ULX: Lx > 1x10 '
Figure 3-5. M~Ks histogram for all clusters. Pinned by 0.2 magnitudes. Dashed line shows
cutoff for statistical on~ lli. -
z -6 i! -
Figure 3-6. This figure is the same as Figure 3-5, except that I have plotted MsK versus
the number of clusters per hin divided by the mean flux in that hin. Arguing
that mass is proportional to flux, this graph shows the probability of finding a
cluster with a given mass.
-15 -14 -13 -12 -11
Theoretical, 1KlM)y histograms for all X-ray sources with counterparts,
detections, and those without, nondetections, using hins of 0.2 magfnitudes.
For detect ions, reddening computed using NVH(PL) (top) and "color method"
A FIRST ESTIMATE OF THE X-RAY BINARY FRACTION AS A FUNCTION OF
STAR CLUSTER MASS IN A SINGLE GALACTIC SYSTEM
In C'!s Ilter 3 I performed an extensive study of the XRB environments in the
Antennae galaxies using Chandra X-ray images and J and Ks IR images (Brandl et al.,
2005, henceforth Paper I). In this chapter I will expand on this study by exploring the
relationship between XRBs and cluster mass in the Antennae (see also Clark et al.,
Recent theoretical models of young, massive cluster evolution provide a framework for
comparison to my observational study. Two in particular, (Oskinova, 2005; Sepinsky et al.,
2005), incorporate XRB formation in their models. Oskinova (2005) use a population
synthesis code to study the evolution of X-ray emission in young, massive clusters.
Sepinsky et al. (2005) investigate the role of supernova kicks in XRB expulsion from the
parent cluster using the population synthesis code, StarTrack. They also incorporate the
XRB formation rate for a range in cluster mass. I will compare my observations of the
XRB formation rate in the Antennae to those predicted by these models.
I begin this chapter by defining a quantity, rl, relating the XRB fraction to cluster
mass in the Antennae (@4.1). In my analysis, I estimate cluster mass using K, luminosity,
which depends non-trivially on the assumed cluster age. I explore cluster age/1uminosity
relations and their impact on my mass estimates in ~4.2. I then compare rl to the
measured value predicted by theoretical cluster evolutionary models and present
conclusions in 94.3.
4.1 XRB-to-Cluster Mass Fraction
In (I Ilpter 3 I used K, luminosity to estimate cluster mass, which assumes that all
clusters are the same age. I then demonstrated that XRBs are more common in more
massive clusters in the Antennae (see $184.108.40.206 and Figure 3-6).
This result is not surprising-as star cluster mass increases, so does the number of
massive stars in it. Through stellar evolution, a certain fraction of these stars will die
in supernova explosions, leaving behind neutron star or black hole remnants. In turn, a
fraction of these stellar remnants will retain/acquire a mass-donating companion star,
becoming a detectable XRB. Thus, through sheer numbers of stars in more massive
clusters, I expect a greater likelihood of finding XRBs in them. This leads me to two
interesting questions: 1) quantitatively, what range in cluster masses tend to harbor
XRBs and 2) is there some intrinsic property of massive cluster physics that favors the
production of XRBs, beyond simple scaling with mass?
For the first time, I am able to answer these questions for the Antennae galaxies.
Never before has there existed a large enough sample of IR-to-X-ray associations to allow
meaningful statistics. In my approach to answer these questions I explore the relationship
between the number of X-ray detections per unit mass as a function of cluster mass in the
Antennae. I can formalize this expression in the following equation:
Nx (TT = ot (17 (-17) --TT(4-1)
Here, NVx(lA) is the number of detected X-ray sources with an IR cluster counterpart,
Nol(lA) is the number of detected clusters, and l(lA) is the fraction of X-ray sources per
unit mass, all as a function of cluster mass, -TT .
If l(lA) increases or decreases over a range in 14, this means there could be
something peculiar about massive cluster physics to favor or suppress XRB formation.
In contrast, a constant r(lT) across all if would indicate that more massive clusters are
more likely to have an XRB simply because they have more stars.
While r(if ) is a powerful tool in studying XRB formation rates, it requires that I
know the mass of each star cluster. However, extrapolating the masses of the Antennae
clusters from my photometric data required models that called for estimates of ages and
metallicities. While I successfully constrained these inputs and determined cluster masses
(see below and ~4.2), I first sought to compute l(lA) in terms of a purely observable
quantity-flux. Calculating rl(M ) as a function of K,-band flux, rl(FKs), allOWed me to
investigate non-model dependent trends in rl(Mc).
I calculated rl(FKs) foT CluSterS With a K,-band luminosity brighter than the -13.2
mag cutoff using bin sizes of FK, = 4 x 106 (Figure 4-1). This bin size was small enough
to show a trend in rl(FKs), but large enough to contain at least two clusters with X-ray
sources, allowing me to assign error bars to each value of rl(FKs). The errors plotted on
the graph are uncertainties in the mean value of rl(FK, ) added in quadrature with the
Poisson uncertainty in each bin. Due to the small sample size per bin, I computed these
uncertainties using the small number statistics formulae described in K~eeping (1962).
Figure 4-1 shows that 17(Fh,) is r.oughly consistent with a constant value of 5.4 x 10-8 F-c
with an uncertainty of op = 1.8 x 10-" FI .
In essence, rl(FKs) is comparing two different mass distributions, NotI(FK,), the mass
distribution for all clusters in the Antennae and N~x(FKB) FKs, the mass distribution
for clusters with X-ray sources, normalized by flux. If there is a constant number of
X-ray sources per unit cluster mass as -II__- -1. .1 by Figure 4-1, then these two mass
distributions should be the same. We can further corroborate this result by comparing
NVx(FK,) FK, and NotI(FK,) using a two-sided K~olmogorov-Smirnov (K(S) test. The
K(S-test yielded a D-statistic of 0.75 and a probability of 0.107 that they are related.
Considering the separate cluster mass populations as two probability distributions, each
can be expressed as a cumulative distribution. The D-statistic is then the absolute value
of the maximum difference between each cumulative distribution. This test quantitatively
demonstrates that there is nothing peculiar about massive clusters in the Antennae with
associated XRBs. I also computed the Pearson r linear correlation coefficient between
NVx(FK,) FK, and NotI(FK,), finding a value of 0.99. Since a value of 1 means a perfect
linear fit, this value of r further substantiates the observed relationship in rl(FKs).
I then converted rl(FKs) into the more conventional units of solar mass. Selecting
the mass-normalized M~Ks liSted in the Bruzual-CHl .) lot (BC) cluster evolutionary
models (Bruzual and C'I Ia lot, 2003) for a 20 Myr cluster as a typical value in the
Antennae (Whitmore et al., 1999), I converted the model M~K, oO FK, using the standard
relationship between luminosity and flux. Multiplying rl(FKs) by FK, I COnVerted rl(FKs)
to solar masses: assuming a cluster metallicity of z = 0.02, rl = 5.8 x 10-slMy with
an uncertainty of ay = 7.9 x 10-9My 1, while assuming a metallicity of z = 0.05,
rl= 7.9 x 10-'.\/-1 with an uncertainty of aeq-= 1.2 x 10-81Al
While I assumed all clusters are ~20 Myr old, I note that the actual range in ages
should be ~1-100 Myr (Whitmore et al., 1999). The BC models indicate that clusters in
this age range could vary by a factor of as much as 100 in mass for a given K, luminosity.
Since rl is a function of mass, incorrectly assigning cluster ages has the potential to
significantly impact rl. In the next section, I explore how differences in cluster age can
affect the value of rl.
4.2 Effects of Age
I investigated the effect differences in cluster age has on rl by assuming three
individual age distributions for the Antennae clusters: an instant burst in which
all clusters are the same age, a uniform distribution, and a distribution of the form,
dN/ldv oc 7-r- (Fall et al., 2005). In each case, I assumed all clusters have solar metallicity
(z = 0.02) (Whitmore et al., 1999).
In the instant burst case, I assigned the same age to describe all clusters and picked
several such values in the range 1-100 Myr. Applying the BC models, I produced several
different cluster mass distributions. For each distribution I computed a mean rl (Figure
4-2). The mean for these values is rlinstant = 4.7 x10-s Myo~ with a standard deviation of
ainstant = 1.8 x10-s My 1. Comparing rlinstant to rl computed assuming constant cluster
ages of 20 Myr (rl20) and 100 Myr (rlloo) shows that both rl20 and rlloo fall within ~1e of
rlinstant and that all three values of rl differ by less than a factor of two (Figure 4-5).
Next, I assumed a uniform age distribution for my Antennae cluster sample. Picking
ages at random from a uniform distribution between 1-100 Myr, I assigned an age to
each cluster in my sample. Applying the BC models, I computed each cluster's mass
based on the assigned age, produced a mass distribution, and then calculated a mean rl.
Performing a Monte Carlo (jlC) simulation, I recreated cluster mass distributions 10,000
times, producing a large sample of rl's with a mean of rlniform = 3.2x10-s My, and
Juniform = 7.4x10-9 Mgo~ (Figure 4-3). While there is at least a la difference between
rluniform and rl20, a ulllfOrm age distribution is not the most applicable to the Antennae.
As demonstrated by Fall et al. (2005), most of the clusters in this galaxy pair are young,
with a median cluster age of ~ 20 Myr.
A more realistic approach is assuming the cluster ages are defined by a power law
(PL): dN/ldv oc 7-r- (Fall et al., 2005). These authors derived their relationship using
HST UBV1IH, observations of ~11,000 clusters. Fitting BC models to photometry of
each cluster, they generated an age distribution. In my analysis I picked ages at random
according to this distribution. Mirroring the procedure used for the uniform distribution
case above, I created a sample of 10,000 rl values. I found a mean for this sample of rppe
= 5.2 x10-8 My] and a apt = 9.5 x10-9 My Interestingly, rlp falls well within la
of rl20 (Figure 4-3). This is not surprising considering most clusters are ~-20 Myr old.
Furthermore, it illustrates that by assuming an instant burst of 20 Myr, as I did initially,
will not significantly affect my results, since rl20 and rlp vary by a factor of less than two.
In a final scenario, I used the cluster ages listed in the electronic table available
through Mengel et al. (2005) to fit ages to 144 clusters in my sample, including all
15 clusters associated with X-ray sources. Mengel et al. (2005) derived ages by using
three age indicators-UBVI and K, broadband photometry to break the age/reddening
degeneracy, He~ and Bry emission to identify clusters less than 7 Myr, and CO band-head
absorption from narrow-band images for clusters ~10 Myr. They then fit these data to
theoretical spectra for ages < 500 Myr using a X2 millimization technique (for details, see
Mengel et al., 2005).
Following the method discussed in (4.1, I used the BC models to convert AtK to
mass for those clusters with Mengel et al. (2005) age estimates. Using cluster hins of
3x106AL, in mass, I computed three values for ty(Af,) (see Figure 4-4). The errors plotted
on the graph are uncertainties in the mean value of ty(Al.) added in quadrature with the
Poisson uncertainty in each hin. Again, I used the small number statistic formulae in
Keeping (1962) to compute these errors. I found a mean in ty(M-,) of 5.8x10-slAC and
o = 1.9 x 10-slAf~
I suninarize my age analysis in Figure 4-5. Comparing the distributions in ty for the
instant burst, uniform and power law age distributions, all values for ty are within a factor
of two. Even if I assume an instant burst of 10 Myr, ty differs by no more than a factor of
four (see Figures 4-2 and 4-3). These relatively small variations indicate cluster age has
little effect on ty and hence my results.
My observed lack of dependence between ty and an assumed cluster age distribution is
not surprising considering the recent work in Fall (2006). These authors discussed how the
cluster mass distribution in the Antennae is independent of the cluster age distribution.
Since if depends on this mass distribution, my small range in ty corroborates the results in
4.3 Comparison with Models
Through the quantity ty, my investigation showed that the XRB formation rate per
unit mass is independent of cluster mass in the Antennae. I estimated cluster mass by
fitting BC spectrophotonietric models to cluster Mrs,, assuming all clusters are ~20
Myr. Recognizing that this method depends on cluster age, I explored how different
assumptions for the Antennae cluster age distribution will affect ty and showed that ty
varies by less than a factor of four. This small variation in ty demonstrated that this
quantity does not strongly depend on the assumed age distribution of clusters in the
Inow proceed by comparing my observed value for rl = 6 x10-s Mg- (z = 0.02) to
that predicted by models of young, massive clusters. Specifically, I chose the theoretical
work of Oskinova (2005) and Sepinsky et al. (2005).
In a recent study presented in Oskinova (2005), the author modeled X-ray emission
from young, massive star clusters, assuming a closed system with constant mass, no
dynamics and all stars are coeval, with cluster metallicities of either z = 0.02 or z = 0.008.
These models predict ~2-5' of all OB stars in a cluster should produce high mass X-ray
binaries (HMXBs). In the models in Oskinova (2005), all clusters are assumed to have
masses of Met = 106 Meo with stellar masses ranging from 1-100 Me. Considering the
Salpeter initial mass function (IMF) of the form ~((M) = i1,sif-2.35 and defining stars
with masses > 8 Meo as "OB -I Il for our purposes here, I estimated I0' of all stars
in the model clusters are OB stars. Therefore, 1-3x10-3 Of all StarS in a CluSter With an
initial mass of 3 x106 Me (set by the Salpeter IMF) should produce an XRB. Since the
Salpeter IMF implies there are 7x10s stars in a cluster, then these stars should produce
7-22x103 XRBs-orders of magnitude greater than the ~49 observed in the Antennae. In
fact, only 15 of these sources are associated with clusters. Expressing rl as a fraction of
XRBs-to-cluster mass, the models in Oskinova (2005) -11--- -in ranges from 3-7x10-4
My 1. These values are greater by at least a factor of 1000 from my estimates for rl.
Clearly, this predicts a much larger formation rate of compact object binaries than what I
observed in the Antennae. Oskinova (2005) note that they were unable to detect HMXBs
in three massive (~ 104 Me) clusters which they predict should contain between 1-3
HMXBs. If my observed value for rl accurately describes the formation rate of HMXBs,
then it is not surprising that Oskinova (2005) fail to find any. As pointed out by these
authors, future modeling of HMXB formation is needed to understand the discrepancy
between the predictions and observations of XRB populations in starburst galaxies.
In another study, Sepinsky et al. (2005) use the binary evolution and population
synthesis code, StarTrack (B.1.I. i-in-1:; et al., 2002), to investigate the rate of XRB
formation and ejection from young, massive clusters. This program tracks stellar
parameters such as radius, luminosity, mass and core mass. The simulations are stopped
at the formation of a compact object. The models include mass transfer in binaries and
include transient XRBs. Sepinsky et al. (2005) consider cluster masses ranging from
5 x 104 Me to 5 x 106 Me and cluster ages from 1 to ~-20 Myr and compute the average
number of XRBs within 1-1000 pc of the cluster center.
Considering the typical cluster age in the Antennae is 20 Myr, Sepinsky et al. (2005)
predict a 5 x 104 Me cluster should contain 0.13 XRBs, while 15 XRBs should reside in a
5 x 106 Men cluster. Here I assume that an XRB is associated with a cluster if it is within
100 pc. This separation is equivalent to 1.0"! at the distance of the Antennae (for Ho=75
km s-l Mpc-l), which is my criteria for an XRB-cluster association (Chapter 3). These
model predictions for XRB detections assume a limiting X-ray luminosity of Lx = 5x1035
ergs s-l, but the observed limiting luminosity in the Antennae is 2 x1037 eTgS S-1. Using
the X-ray luminosity function (XLF) for the Antennae defined in Zezas and Fabbiano
(2002), I scaled the XRB results of Sepinsky et al. (2005) to estimate what these models
would predict for the observed number of XRBs in the Antennae clusters. Using a XLF
power law slope of a~ = -0.45 (Zezas and Fabbiano, 2002), the models predict 0.02 XRBs
are observed in a 5 x 104 Me cluster, while 2.7 XRBs should be seen in a 5 x 106 Me
cluster, at the luminosity limits of the X-ray observations. Expressing these model results
as a fraction of XRBs-to-cluster mass, I can directly compare them to my observed value
for rl in the Antennae. Doing so, Sepinsky et al. (2005) predict rl ranges from 4 5 x 10-7
My 1, at least a factor of 10 higher than my predictions for rl. As mentioned by Sepinsky
et al. (2005), several caveats exist for their models including: 1) assumed binary fraction
of unity which could lead to over estimates of the mean number of XRBs per cluster, 2)
the stellar, power-law IMF can affect the XRB fraction per cluster, and 3) changes in
the half-mass radius can strongly influence the median XRB distance from the cluster.
These factors could potentially explain the discrepancies between their models and my
4.4 Summary and Conclusions
In this chapter I introduced the quantity, ty, relating the fraction of X-ray sources
per unit mass as a function of cluster mass. Applying this function to the Antennae,
I revealed several important environmental implications for the X-ray sources in the
Antennae. Specifically, if predicts a far different relationship between XR B formation
and cluster mass than that predicted by Oskinova (2005) and is broadly consistent with
that predicted by Sepinsky et al. (2005). Clearly, future cluster modeling with particular
emphasis on the relationship between the number of XRBs in a galaxy and the galactic
cluster environment is essential to explain my current observations. In addition, I plan
to extend my observational study to additional starburst galaxies. I can then address
whether ty depends on an individual galaxy or is constant for all galactic environments.
S 10.0- -
-18 -17 -16 -15 -14
Figure 4-1. This figure dli- pha (FK,) plotted versus M~K,. The bins are FKs = 4 x 106 i
size and designated by the histogram. Error bars are the uncertainties in the
mean value of rl added in quadrature with the Poisson uncertainty in each bin.
The dotted line is the mean of the four rl(FKs) ValueS.
5,5 6,0 6,5 7,0 7.5 8.0 8.5
Figure 4-2. Assuming an instant burst of star formation in the Antennae, I plot the mean
value of if for a range in ages between 1 100 1\yr. Notice the factor of ~4
range in ty as well as the degeneracy in if in this age range. Error hars are
uncertainties in ty.
2x10-8 4x10-8 6x10-8 8x10-8 1 x10-7
Comparison between uniform and PL Monte Carlo simulations of rl. The
peaks of each distribution vary by a factor of ~2 in rl, indicating rl does
not significantly change when I assume different age distributions for the
Antennae. I also plot the values of rl for instantaneous bursts at four different
II I I I I I I I I I I I I I I I I I I I I I I I I I I I I I
0 5.0 x106 1.0x107 1.5x107 2.0x107 2.5x107
Here l(Meo) is plotted versus cluster mass in units of Me. In this case, I
computed cluster mass using ages provided by Mengel et al. (2005) (see 94.2).
The bins are Meo 3 x106 Meo in size and designated by the histogram. Error
bars are the uncertainties in the mean value of rl added in quadrature with the
Poisson uncertainty in each bin. The dotted line is the mean value of the three
a b c d e f
Here I summarize how age affects rl, assuming four different age distributions
for the Antennae clusters: instant burst (a), uniform (b), power law (c)
and derived ages by Mengel et al. (2005) (d). Each value is the mean rl and
includes 1-o- error bars. See text for details. Also included is rl for an instant
starburst of 20 Myr (e) and 100 Myr (f).
MULTIWAVELENGTH STUDY OF CHANDRA X-RAY SOURCES IN THE
The numerous X-ray point sources and young, massive star clusters in the Antennae
makes this galaxy pair an ideal laboratory for studying the environments of X-ray binaries
(XRBs). Chandma observations revealed 49 X-ray point sources ranging in luminosity from
1038 1040 eTgS S-1 (Zezas et al., 2002a). While most are XRBs with either a black hole or
neutron star, those sources with Lx > 1039 eTgS S-1 are more unusual objects classified as
ultraluminous X-ray sources (ULXs). Some theories -II__- -r these ULXs are intermediate
mass black holes with masses from 10 -1000 Me (e.g., Fabbiano, 1989; Zezas et al., 1999;
Roberts and Warwick, 2000; Makishima et al., 2000), but there remains a considerable
amount of controversy concerning their origins (e.g., K~ing et al., 2001; Moon et al., 2003;
Compact objects tend to be associated with massive star formation, which some
theories -II---- -r is predominant in young stellar clusters (Lada and Lada, 2003). In my
previous chapters on the Antennae I found a close association between compact objects
and clusters, identifying 15 possible IR counterparts to X-ray sources. Many of these
counterparts reside in the spiral arms and l Il !I ls-; region of the Antennae -locations
predominant in massive star formation. Those X-ray sources without counterparts could
be compact objects that escaped their parent cluster or remained behind after their cluster
evaporated. In OsI Ilpter 3, I -II__- -a third possibility, that some X-ray sources do have
counterparts, but these are too faint to see in the IR images.
In this chapter, I extend my study to optical wavelengths using HST images of
the Antennae. The higher sensitivity of HST will allow me to search for additional
counterparts to X-ray sources. In addition, combining multi-band photometry in the
optical and IR, I can use spectral evolutionary models to measure cluster properties. In
95.1 I discuss my IR observations and analysis of optical HST archival images. I describe
my photometric analysis of counterpart cluster properties in 95.2. I summarize my results
5.1 Observations and Data Analysis
5.1.1 Infrared and Optical Imaging
I base this study in part on IR J (1.25 pm) and K, (2.15 pm) band images of the
Antennae. My collaborators and I initially analyzed these images in Brandl et al. (2005)
and report the details of their reduction in that work. In summary, I acquired 20-minute
total exposures in each filter using the Wide-field InfraRed Camera (WIRC) on the
Palomar 5-m telescope during the night of 22 March, 2002.
I supplemented the IR wavelengths with optical HST images (Whitmore et al., 1999)
obtained from the HST archive. This data set consists of images in the following filters:
F336W(U), F439W(B), F555W(V), and F814W(I). Those authors also included narrow
band, F658N images centered on the redshifted He~ line at the distance to the Antennae.
I used the F658N filter to derive reddening and ages for some optical counterparts (see
In my efforts to understand the environments of the Antennae X-ray sources, I
extended my frame tie between the IR and X-ray discussed in (I Ilpter 3 to optical, HST
images. Tying Chandra X-ray coordinates directly to HST positions is nontrivial due to
field crowding in the HST images. Instead, I used my excellent frame tie between the
IR and X-ray as an intermediary. Initially, I identified 12 sources in common with the
IR and optical. Again, using the mapping method described in (I Ilpter 3, I matched IR
pixel positions to HST right ascension and declination. Due to image rotation between
the WF and PC cameras I made separate frame ties between the IR and each camera
field. Image rotation also exists between filters, which forced me to make additional frame
ties between the I-band image and each of the other four bands. In all cases, the rms
positional uncertainty was less than 0.6". Applying my IR-to-optical astrometric fits to
previously derived IR x/y pixel positions of X-ray sources, I found the HST positions for
all Chandra X-ray sources.
Using the IR as an intermediary between the optical and X-ray frame tie has
many advantages over direct optical/X-ray matches. First, the IR penetrates dust in
similar v- .--s to X-rays, facilitating the identification of counterparts to X-ray sources.
Furthermore, while bright IR sources may be obscured in the optical, the converse
is generally untrue-any bright HST source shows up prominently in the IR images,
enabling an excellent optical/IR frame tie, and thus closing the astrometric loop at all
wavelengths. Previous attempts to match X-ray and optical positions produced many
possible counterparts, but with poor accuracy-as nar Ilny as T.~' of the matches are chance
coincidences (Zezas et al., 2002a).
5.1.2 Identification of Optical Counterparts to Chandra Sources
As I did in (I Ilpter 3, I defined circular areas of positional uncertainty centered
on each X-ray source position. Specifically, an inner aperture with a radius of 1.0"! and
two outer annuli, one from 1.0"-1.5" and another from 2.0"-3.0". Due to the highly
crowded HST fields I often found multiple sources within the central aperture, forcing
me to deviate from my criteria used in OsI Ilpter 3 to define counterpart candidates. I
narrowed my selection criteria by defining -1Ining!" matches as those X-ray sources with
only one match to an optical source in the central aperture and only ~ 1 source in the
outer annulus. "Possible" matches were those X-ray sources with matches to two optical
sources within the central aperture and only ~ 2 sources in the outer annulus. Using these
criteria, I identified 13 I-band counterparts to 10 X-ray sources, seven -1Ining!" matches
and three p.. --h!I I:" matches to six I-band counterparts. Five of these 10 X-ray sources
did not have previously identified counterparts in the IR. Using the V-band HST image,
I found an additional three counterparts to two X-ray sources, a -1Ining!" match and a
"p<--t!.match to two V-band. Both of these X-ray sources did not have counterparts in
the IR. I list all optical counterparts in Table 5-1. Comparing optical source positions, I
identified all of these counterparts across both the V and I hands. Of all 12 X-ray sources
with optical counterparts, one is the southern nucleus and another is a background quasar
(see ChI Ilpter 6 and Clark et al., 2005). Therefore, in my analysis of cluster properties,
I only considered the optical cluster counterparts to the remaining 10 X-ray sources. In
Figure 5-1 I display subimages of all X-ray sources with optical counterparts.
Using the technique developed in Chapter :3, I attempted to estimate the level of
"contamination" of these samples due to chance superpositions of unrelated X-ray sources
and optical clusters. First, I calculated an optical source density by counting the number
of sources in the outer, 2.0"-3.0"1 annulus around each cluster counterpart candidate
and divided this by the total area of these annuli. 1\ultiplying this density by the total
area of positional uncertainty around the -lI nisg!" and "p< .--h! I optical counterparts,
I estimated the level of source contamination contributing to my sample of optical
counterpart candidates. I expect five with a la uncertainty of +0.5/-0.31 of the seven
-1Inisg!" counterparts to be due to chance superpositions, and five with a la uncertainty
of +0.6/-0.41 of the six p.. --h!I. ~" counterparts to be chance superpositions. Clearly these
statistics indicate the n 1..! R~~ly of my optical counterparts are chance superpositions and
further demonstrates the difficulty of making such matches in the crowded, HST images of
I also increased my sample by considering the optical equivalents to the previously
identified 15 IR cluster counterparts (Chapter :3). These additional optical matches
were identified using the IR-to-optical astrometric frame tie to match IR counterpart
positions to HST positions. While in many cases a single IR counterpart split into
multiple optical counterparts, I labeled these conglomerations as a positive match. I found
optical counterparts to all IR cluster counterparts, increasing my total number of X-rays
1 Found using confidence levels for small number statistics listed in Tables 1 and 2 of
sources with counterparts to 22. Figure 5-2 dli ph subimagfes of those counterparts to
X-ray sources seen across all six IR and optical bands. Figure 5-:3 is an I-hand image of
the Antennae showing the positions of all counterpart candidates to X-ray sources.
I performed aperture photometry on my X-ray cluster counterpart sample using
two techniques. One method emploi- 0 the same sized aperture across all wavelengths to
estimate cluster colors, while a second method used apertures defined by the image point
spread function (PSF) to find photometric magnitudes.
220.127.116.11 Constant aperture photometry
In this work, I fit spectral evolutionary models to cluster colors. Following the
photometric procedure defined in Whitmore et al. (1999), I performed photometry on
all optical counterpart clusters using constant apertures across all bands. This method
gave me accurate measurements of cluster colors, since color indices are defined using
magnitudes measured within same-sized apertures. (But, as Whitmore et al. (1999) points
out, there may be an error of a few tenths of a magnitude in total magnitude.) For those
counterparts only seen in the optical, I picked a central aperture with a radius of 4 pixels
and a sky background annulus from 10-15 pixels. These are with respect to the WF chips.
In terms of the PC chip, these radii translate to :3 and 7-10 pixels for the central aperture
and background annulus, respectively.
In the case of those cluster counterparts seen across all six IR and optical bands,
I defined a constant photometric aperture as ~:$o- of the J-hand Gaussian PSF, where
the full width at half maximum (FWH1\) is 1.2". The background annulus had a radius
of ~6-10o- of the PSF. Considering HST resolved many of the IR sources into multiple
components, this large photometric aperture encompassed these conglomerations. Since
I expect that these sources are part of the same, larger structure with similar ages and
metallicities, it is appropriate to include photometry of them in spectral evolutionary
18.104.22.168 PSF aperture photometry
As I discuss in 95.2.1, I estimated cluster mass assuming mass is proportional
to cluster flux and only used the V and Ks bands. Therefore, I chose a more precise
technique to derive cluster magnitudes, specifically defining photometric apertures as
~ 3a of the Gaussian PSF. I also used this method to measure I magfnitudes for optical
counterparts listed in Table 5-1. The FWHM in these bands are 0.18" in V, 0.21" in I,
and 0.9" in Ks. Thus, the aperture radii are 2.5 pixels in V and 2.6 pixels in I for WFC,
and 5 pixels in Ks for WIRC. I did not find cluster counterparts on the PC chip field and
did not consider photometric parameters for this field.
I compensated for the crowded field in the Antennae by measuring sky background
in two separate annuli, from ~6-10a of the Gaussian PSF. In addition, I divided the
annuli into arcs for exceptionally crowded regions. Multiplying the mean and median sky
background in each annulus by the area of the central aperture and then subtracting this
from the central aperture flux yielded four separate flux measurements for the source.
I then averaged these four values to give me the source flux. When computing errors, I
considered variations in the sky background, asky, and Poisson noise, ames. To calculate
asky I found the standard deviation of the four flux values for each source. Dividing
the mean source flux by the instruments gain and then taking the square root of these,
we found ages. The known gain for WIRC is 2 e- DN (Wilson et al., 2003)2 While
Whitmore et al. (1999) used a gain of 7 e- DN for HST. Both asky and wasd were added in
quadrature to yield a total error in flux, afl,,
I used a bright, 2MASS star to establish the Ks magnitude zeropoint. I derived
HST F555W and F814W magfnitudes using zero points listed in Table 28.1 of the HST
Data Handbook (Voit, 1997). Applying color transformations defined in Holtzman et al.
2 At the time of the Antennae observations, WIRC was equipped with a Hawaii-1
1K x 1K detector and this is the gain for it.
(1995), I converted F555W and F814W magnitudes to Johnson V and I magnitudes,
respectively. I expressed errors in magnitude, am, as afl, divided by the mean flux. In
the case of the optical filters, am consists of the additional errors in the zeropoint and
color transformations, all of which were added in quadrature.
5.2 Results and Discussion
5.2.1 Spectral Evolutionary Models
After performing photometry on all X-ray source counterparts, I threw out nine
that had bad photometry. I defined "bad pho- nin.! 1 1y as those sources with negative
fluxes in any of the six bands (UBVIJ and Ks,). In many cases this occurred due to
poor sky subtraction in exceptionally crowded regions of the Antennae. Furthermore,
some faint optical counterparts were only seen the the V and I bands and thus was not
able to get photometry across all four optical bands for them. Thus I continued further
analysis on 16 counterparts to X-ray sources. These 16 clusters consist of nine sources
seen in all six filters, four with only UBVI photometry and three containing solely V
and I magnitudes (see Table 5-2). I fit Bruzual-C'l .) h >t (BC; Bruzual and C'l l .)h t, 2003)
spectral evolution models to the cluster magfnitudes available for each source to determine
mass, age and metallicity. These parameters gave me a more complete understanding
of XRB environments in the Antennae. While it is uncertain which strictly optical
counterparts are associated with an X-ray source, I proceeded separately with analysis of
them as well and interpreted their derived properties as -II_ -lR.; of the nearby X-ray
As discussed in 95.1.2, those clusters identified as strictly optical counterparts are not
necessarily believable as accurate matches to X-ray sources. Therefore, I urge caution in
interpreting the model results for these counterpart candidates.
Before performing the model fits, I de-reddened all magnitudes using Av derived
from X-ray measured column densities, NVH (Zezas et al., 2002a, Table 5). Zezas et al.
(2002a) fit two separate models to their X-ray spectra to compute NVH; a Raymond-Smith
(RS: Raymond and Smith, 1977) model and a power law (PL) model. I chose to compute
reddening using NVH(PL) because this provided a closer match to reddening calculated
using (J K,) cluster colors (see C'!. Ilter :3). Applying the following equation, A4-
5 x 10-22 mag cm2 1H, I found A44. Inserting A44 in the reddening law defined in Cardelli
et al. (1989, hereafter CC11L), I found the reddening in all six hands.
I estimated the error in reddening, a 4, by also computing reddening in each filter
using the extinction law defined in Rieke and Lehofsky (1985). Comparing these values to
the reddening computed using the CC11L law, I defined the difference in reddening as a4.
Adding a 4 and a, in quadrature, I computed a total error in de-reddened magnitude, an.r
Plotting the cluster de-reddened magnitudes, I produced a pseudo spectral energy
distribution (SED). Iteratively shifting the BC model SEDs, I found a best fit to the
cluster SEDs by minimizing X2 (see Figure 5-4). I also introduced reddening in the cluster
SEDs. By iteratively picking values for A44 in the range 0.0-3.0 mag and in steps of 0.1
1!! .-- I selected the A44 that contributed to my best model fit. I defined goods fits as those
models within one of the best fit X2 sum (Eg2 + 1). The resulting fits gave me a range in
metallicity, age, and reddening for each cluster. Each fit is listed in Table 5-2.
The shift in magnitude between model and data (A) gave me a mass estimate for
each IR /optical counterpart. a contains information on the distance modulus and mass
of the cluster. Subtracting off the known distance modulus to the Antennae, ma = :31.4
1!! I- I was left with a difference in luminosity between my cluster and a 1 Alo cluster
as listed by the BC models. Converting this difference into a change in flux gave me the
cluster mass in solar masses. Before computing this change in flux, I renormalized the
luminosity difference, expressing it in terms of magnitudes instead of colors. For fits only
to the optical bands, I renormalized the luminosity difference to Ms-, while I used Mxs for
fits across all bands.
5.2.2 Estimating Av From Hca Equivalent Widths
Three X-ray sources, X-17, X-27 and X-48, only had optical cluster counterparts and
did not have X-ray-measured NVH. Therefore, I could neither use (J K,) nor NVH oO
find cluster Av. Since my fitting technique to the BC models requires knowledge of Av, I
needed a separate method to find the properties of these three cluster counterparts.
I chose to adapt a method outlined in Whitmore and Zhang (2002) to analyze
this trio of clusters. As these authors point out, there exists a degeneracy between
reddening and age for clusters with Av >1 1!n I. the reddening vector can intersect the BC
evolutionary track in multiple locations. I was able to break this degeneracy by using He~
equivalent widths (EWs) to estimate cluster age.
Before measuring the Hca EWs from the narrow band F658N HST image, I estimated
the continuum level in our three sources using scaled I fluxes. I converted the resulting
continuum-subtracted He~ fux to units of energy using the following relation: 4.2 x1015
ergs s-l = 1 DN s-1.3 Applying the PHOTFLAM keywords listed in Table 28.1 of the
HST Data Handbook (Voit, 1997), I computed Hca EWs. I list my measured EWs in Table
5-3. Fitting these Hca EWs to Starburst99 spectral evolution models (Leitherer et al.,
1999), I found cluster ages. These models consist of evolution tracks for five different
metallicities ranging from z = 0.001-0.04. Thus, I calculated five separate age estimates
for each of my three clusters.
Next, I produced three separate color-color plots overlaid with BC evolutionary tracks
(Figure 5-5). Using equations 1 and 2 in Whitmore and Zhang (2002) I corrected for
excess He~ emission seen in the V and I bands and define these corrected magfnitudes as
Veor and Icor, respectively. Based on the Hca EW-predicted ages alone, I determined all
three clusters are <20 Myr. Since there is little discrimination between cluster color and
age in the (V Icor) vs. (B V) plot, I only used the (B Veor) vs. (U B) and (V Icor)
3 See http://www stsci.edu/instruments/wfpc2/Wfpc2 faq/wfpcnrphtfahml
vs. (U B) plots to estimate cluster age. Tracing the reddening vector backwards from
each cluster's colors, I searched for the intersection with the BC models that matched
closest to the age predicted by the cluster's Hca EW. Repeating this an~ lli--is for a range in
metallicities, I searched for the best match in age predicted by the two evolutionary track
models. I found a z = 0.008 provides the best fit for the cluster counterparts to X-17 and
X-27. I did not find a reasonable fit to X-48's counterpart. This cluster lies below the BC
models in color and its reddening vector never intersects the model. Therefore, I excluded
this cluster counterpart from further analysis.
After deriving age and metallicity for the cluster counterparts to X-17 and X-27, I
estimated their mass. Comparing the clusters' observed (V I) to the intrinsic (V I),
I calculated Av using the CC11L extinction law. I then used Av and the distance modulus
to the Antennae, md = 31.4 il to find each cluster's V luminosity, Myv. Comparing Myv
to Myv listed in the BC models for a 1.0 Me cluster with the same age and metallicity, I
found each clusters mass in solar masses. I summarize the properties for these two clusters
in Table 5-3.
In the plots di11l-p Id in Figure 5-6, I include the cluster counterparts discussed
in ~5.1.4 where I fit BC models to cluster colors to find their properties. Notice the
reasonably good agreement for most of the clusters. Only two of these clusters, with the
X-ray sources X-34 and X-46, are outliers. Since I could only fit UB VI photometry to
these clusters, it is understandable that they have a wide range in ages.
I summarize the results from all model fits in Figure 5-6. These model fits indicated
a wide range in cluster properties. Only considering those cluster counterparts seen across
all six IR and optical bands, I found ages between ~106-10" Myr. Masses ranged from
10s-107 Me, with most ~106 Mo. The mean metallicity was z = 0.047 with a range from
z = 0.004-0.050. This is consistent or slightly higher than solar metallicity, which is the
value Whitmore et al. (1999) assumed in their 2.1, lli--- Finally, I did not find a trend
between these cluster properties and the luminosities of their associated X-ray sources.
If I include the (less reliable) cluster counterparts only seen in the optical, I found
ages between ~106-1010 Myr, a range in masses of ~103-10s Me, and metallicities from z
Using my initial IR-to-X-ray frame tie as an intermediary, I produced an accurate
frame tie between Chandra X-ray coordinates and optical HST positions, achieving a
positional uncertainty of less than 0.6". After producing this astrometric match, I searched
for optical counterparts to X-ray sources and performed a multiwavelength photometric
study of these counterparts. In my analysis, I reached the following conclusions:
1. I identified 8 -1ining"" matches (X-rays sources with ~-1 optical counterpart within
a 1.0"! radius) and 4 p.. --!Iub matches (X-ray sources with ~-2 optical counterparts
within a 1.0"! radius). Of these 12 matches, seven are not seen in my IR study of the
Antennae (ChI Ilpter 3). Considering the Inisg"~ matches to eight X-ray sources, I
estimated five of their optical counterparts are chance alignments. In the case of the
"p<--t!.matches to four X-ray sources, I estimated five of their optical counterparts
are chance alignments. These large numbers indicated most of my identified optical
counterparts are not necessarily believable and I interpreted the photometric properties
of these counterparts as only a so--- -- -lR.; description of the environments of the nearby
2. I found all 15 IR cluster counterparts present in the HST images. In many cases,
the high resolution and increased sensitivity of HST split the IR sources into individual
components. I expect that conglomerations are all part of the same, larger structure with
approximately the same age and metallicity. Therefore, I included these groupings in my
photometric study of cluster counterparts to X-ray sources.
3. Using a X2-minimization technique, I fit BC models to nine cluster counterparts
across all six, UBVIJK, bands and four cluster counterparts across the four UBVI
optical bands. An additional three strictly optical counterparts lacked X-ray-derived
NVH: a quantity necessary to estimate reddening. Therefore, I combined BC model fits to
cluster colors and Starburst99 model fits to cluster Hco equivalent widths to characterized
these clusters' properties.
Since I was uncertain if the strictly optical counterparts are real, I considered model
results across all optical and IR hands separately from models fits to only optical bands.
The BC model fits across all bands indicated the X-r w- onurce-associated clusters are
106-107 Al, in mass, ~ 106-10s Myr in age, with metallicities of z m 0.05.
While I can use multiwavelength photometry to describe cluster properties, there
remains some uncertainty in these characteristics due to errors in magnitude and model
limitations. In future work, I plan to acquire spectra of Antennae cluster counterparts and
refine estimations of their properties.
Table 5-1: Potential Optical Counterparts to Chandra X-Ray Sources
Src ID 0
"(Stlungt CP e
29. .. .. .. .. .
37. .. .. .. .. .
45. .. .. .. .. .
46. .. .. .. .. .
"Possible" CP e
25. .. .. .. .. .. .
Id CP e
12 01 52.15
12 01 53.00
12 01 54.77
12 01 55.71
-18 51 52.02
-18 52 09.59
-18 52 52.43
-18 52 32.16
" ID numbers follow the naming convention of Zezas et al. (2002a)
b Chandra coordinates with an uncertainty of 0.5" (Zezas et al., 2002a)
c Positional offsets in units of seconds of arc from the Chandra coordinates
to the WIRC coordinates of the proposed counterpart.
SUnits are magnitudes, from HST PSF photometry. Values in parentheses indicate uncertainties
in the final listed digit.
e CP stands for counterpart.
Table 5-2: Bruzual-CHl .) lot Model Fits
Chandra Src ID 0
42 b .. .. .. .. .. .
34. .. .. .. .. .. .. .
A, o(!M )
" ID numbers follow the naming convention of Zezas et al. (2002a)
b Due a one-sided age range, BC fits increased to EX2 + 1.921
Table 5-3: X-17 and X-27 Cluster Counterpart Properties
a Log(age/yr) a
Log (mass/A )
" Difference in agfe between that determined
usingf reddeningf vector intersection to BC model
and Hca EW.
1 "; C
Subimages highlighting optical counterparts to X-ray sources. Positional error
circles with 1.0"! radii centered on X-ray source positions are in yellow. Blue
circles are centered on cluster counterparts and have 0.3" radii equivalent to
the I-band photometric aperture. I label those X-ray sources with possible
counterparts using a '*'. All images are 6.0" x6.0".
ii II r ;;; ;;;;l;;;~i~liil
Figure 5-2. Subintages highlighting optical counterparts to X-ray sources with IR cluster
counterparts. Positional error circles with 1.0"! radii centered on X-ray source
positions are in yellow. Red circles are centered on cluster counterparts and
have 1.3" radii equivalent to the K,-band photonietric aperture. I label those
X-ray sources with possible counterparts using a '*'. All images are 6.0"x6.0".
I-hand Antennae HST image (Whitmore et al., 1999) showing positions of
X-ray sources. North is up and east is to the left. Blue circles are optical
counterparts identified in this paper. Red squares are optical counterparts to
X-ray sources with IR cluster counterparts identified in ChI Ilpter 3.
I~- I I- I
b: 1 5, 15
X-10 Counterpart (dereddened)
Example of a good y 2 fit betWeen a cluSter counterpart and a BC model. I
include the derived cluster properties and the quality of the fit parameters. In
the lower graph, notice the small residuals between the model and data.
* I* -
- 0.5 0.0 0.5
0.0 0.5 1.0
6 7 8 9 10
- 0.5 0.0 0.5
Color-color and He~ equivalent width plots used to determine ages for the
cluster counterparts to X-17, X-27 and X-48 (crosses). Dashed lines are
reddening vectors extending from cluster colors to BC models. Notice
reddening vector from X-48 counterpart does not intersect BC model,
indicating a poor fit. I indicate this cluster with a circle in the lower left plot.
Also included in these plots are clusters whose properties I determined using
X2-fitting to BC models. I fit UBV/IJK, magnitudes to clusters indicated by
diamonds and UB VI magnitudes to clusters indicated by triangles. Solid lines
in lower left plot indicate ranges in cluster age.
Summary of results from X2 f1ttillg to BC models. I fit the BC models to
UBV/IJK, photometry (o's), and UBVI/photometry (A). Error bars are ranges
in cluster parameters for models within one of the best fit X2 Sum. I also
include the two clusters I analyzed using Hca EWs (+'s). I divide these plots
by X-ray source luminosity into three separate regions: LLX: Lx < 3 x1038
ergs s-l, HLX: 3 x1038 eTgS S-1 < Lx < 1 x1039erTgS S-1, and ULX: Lx >
1x1039 eTgS S-1. Notice there is no obvious trend between cluster properties
and the luminosity of the associated X-ray source. Note: Those strictly optical
counterparts to X-ray sources are not necessarily accurate matches and these
cluster properties do not conclusively describe their nearby X-ray source's
THE ULTRALUMINOUS X-RAY SOURCE X-37 IS A BACKGROUND QUASAR IN
THE ANTENNAE GALAXIES
U~ltraluminous X-ray sources (ULX) are typically defined as point sources with
X-ray luminosities > 1039 eTgS S-1. Einstein (Long and van Speybroeck, 1983; Helfand,
1984; Fabbiano, 1989) and ROSAT (Roberts and Warwick, 2000; Colbert and Ptak,
2002) observations revealed many of these objects in nearby galaxies. Recent Chandra
observations indicate the Antennae (NGC 4038/4039) contain 9 ULXs (Zezas et al., 2002a,
assuming here and hence forth Ho = 75 km s-l Mpc-l), the largest number discovered so
far in a single galaxy (Fabbiano et al., 2003).
One X-ray source in the Antennae that has received considerable attention in
the literature is X-37, as designated by Zezas et al. (2002a). At the distance to these
interacting galaxies (19.3 Mpc) it would have an X-ray luminosity, L, = 4.5 x1039eTgS S-1
making it a ULX. A previous attempt to match X-ray positions directly to HST positions
indicated that this object has a significant (> 1.0") offset from a nearby optical source
(Zezas et al., 2002b). This spawned discussions of whether this is a runaway X-ray binary
escaping from its parent cluster (Zezas et al., 2002b; Miller et al., 2004; Fabbiano, 2004).
But, matching X-ray positions directly to the optical is difficult due to the crowded HST
field and image rotation.
In C'!s Ilter 3 (see also Clark et al., 2007b) I demonstrated that the IR is a powerful
method for finding counterparts to X-ray sources. This wavelength regime has similar
dust penetrating properties to X-rays, facilitating counterpart identification. In OsI Ilpter
5 I matched X-ray and optical HST positions with high precision by using the IR frame
tie as a go-between. Here I use X-37 to demonstrate the success of my method and show
the importance of obtaining follow-up spectra of counterparts to ULX candidates. 96.1
describes spectroscopic observations of the counterpart to X-37 and briefly discusses
my analysis of the IR and optical images. I summarize my results and present the
implications of this discovery in 96.2.
6.1 Observations and Data Analysis
6.1.1 Infrared and Optical Images
I continue with analysis of the IR and optical images of the Antennae by focusing
on the particularly interesting counterpart to the X-ray source, X-37. Using the
astrometric frame tie discussed in ('! .pter 3, I identified a bright, Ks = 16.2 mag source
0.3 + 0.5"' from the X-ray position of X-37 (see Figure 6-1). Note that this contradicts
previous reports of a significant offset between X-37 and the star cluster candidate, and
independently eliminates any need for the hypothesis that this X-ray source is a Tiru I..e .i
X-ray binary (Zezas & Fabbiano 2002; see also Miller et al. 2004a and Fabbiano et al.
2003). Using this astrometric frame tie between the X-ray and optical, HST images
discussed in ('! .pter 5, I found an optical source 0.6 + 0.6"1 from the X-ray position of
X-37 (Figure 6-1)-again, an insignificant offset.
My collaborator, Micol C! !in1ph r11~~ acquired an optical spectrum of X-37 on March
7, 2003, using the Low Resolution Imaging Spectrograph (Oke et al., 1995, LRIS) on
K~eck I. He observed this source in two separate slitmasks with a 400 lines mm- grating
and blaze wavelength of 8500 A~, yielding a total wavelength coverage on the red side of
5600-9400 A~. Seeing throughout the observations was ~ 0.8"'. He observed X-37 for a total
of 6000 seconds between the two masks with typical single exposure times of 600 seconds.
He reduced the spectrum with a combination of standard IRAF procedures and IDL
code using the NeAr arc-lamp spectrum for wavelength calibration. He used the standard
star Feigfe 67 to flux calibrate the spectrum.
He extracted spectra of the X-37 counterpart from each individual exposure in a 1.8"'
(approximately 2 FWHM) region centered on the emission peak. A background spectrum
extracted from a region 2" removed from the counterpart was sufficient for removing the
night sky lines. I point out the blended lines, Ha/I[NII] at ~ 6600 A~ and the [SII] pair at
~ 6750 A~, observed at the redshift of the Antennae (Figure 6-2). These spectral features
are at approximately the same levels in the on-source and background spectra, -II---- -1;!
that they are likely diffuse emission from the Antennae and not associated with X-37.
After presenting this spectrum to my colleagues, Dr. Fred Hamann and Dr. Vicki
Sarajedini, they noted the broad emission lines characteristic of a quasar. We then
looked for other spectral features to confirm this classification. The realization that
the broad line at 8296 A~ was red-shifted Hce prompted a search for additional lines
generally associated with quasars. We noted a plethora of such identifiers including the
nonrestframe wavelengths of HP at 6148 A~, [OI] at 7965 A~, [NII] at 8320 A~, and [SII] at
8510 A~. This corroborated our conclusion that X-37 is a quasar.
Using the observed, narrow [OIII] lines at 6270.45 a and 6330.94 a, Micol ClIn-IuplI~ .
measured a redshift of z=0.26. At the redshift to this quasar, X-37 would now have an
X-ray luminosity of L, = 1.4 x 1043 eTgS S-1. As I show in Table 6-1, this luminosity is
typical of the X-ray luminosity for quasars at a similar distance (Schartel et al., 1996).
To investigate how much of the continuum flux could be due to the quasar, I
compared the photometric properties of this source to that of other quasars at X-37's
redshift of z=0.26. I chose U and V-band photometry from the HST images and Ks-band
photometry from the WIRC images. This choice in filters covers the full wavelength range
used in my studies of the Antennae (C'h! Ilter 5).
I performed aperture photometry in each of these bands following the procedures
discussed in C'!s Ilters 2 and 5. Next, I converted these photometric measurements to
reddening-corrected absolute magfnitudes. At the redshift of X-37, the distance modulus
is 40.1 mag. I derived reddening in V, Av = 1.3 mag, using the X-ray derived column
density, NVH, liSted in Table 5 of Zezas et al. (2002a) and the relationship Av = 5 x 10-22
cm-2 1H. I used the NVH value provided by the power-law spectral fits of Zezas et al.
(2002a). These authors computed NVH aSSuming the absorption is at zero redshift. This
should not he a problem for me since the broad absorption lines observed in the X-:37
spectrum -II- -- -1 there is little extinction along the line of sight to it.
To find reddening in other filters, I used the extinction law defined in Cardelli et al.
(1989, CC11L). I also computed separate reddenings in each filter using the Rieke-Lehofsky
(RL) Law (Rieke and Lehofsky, 1985). I then expressed errors in reddening, aA, as the
difference in Aue, A44 and A4x derived from each law. Adding aA and a, in quadrature, I
computed a total error in absolute magnitude.
In Table 5-1, I compare the absolute magnitudes, 1Mr. AA-, and A~K,, of the X-:37
counterpart to typical magnitudes for quasars at a similar distance. I obtained Ah, and
AA- from the Sloan Digital Sky Survey (SDSS) quasar catalog (Schneider et al., 200:3)
for :33 quasars at a redshift of 0.26. Using color transformations listed in Fukugfita et al.
(1996), I converted SDSS magnitudes to the Johnson photometric system. I used the
catalog described in Barkhouse and Hall (2001) to derive A~Ks for 17 quasars at a redshift
of 0.26. Considering the absolute magnitude of X-:37 in the U, V and Ks hands falls
within the range of catalog quasar magnitudes, X-:37 has the luminosity of a typical
The identification of X-:37 with a background quasar further demonstrates the
importance of spectroscopic followup for the study of ITLX sources. While the overabundance
of such sources near galaxies clearly demonstrates that a physical connection does exist for
many ITLX sources as a population (e.g. Colbert and Ptak, 2002), the strong possibility
of background quasar contamination makes such identification for any particular source
perilous (see also Gutiidrrez and Lopez-Corredoira, 2005).
A specific example involving X-:37 is the recent work of 1\iller et al. (2004). In their
Figure 2, 1\iller et al. (2004) compare the black hole mass to inner disk temperature
for both ITLX sources and "-I I.1,1 1.1 stellar-mass black holes, findings that the ITLX
sources have cooler disk temperatures. They use this to conclude that ITLX are likely to
he powered by accretion onto Intermediate-Mass Black Holes (IMBH). While some of the
sources in this diagram are almost certainly ITLXs (i.e. NGC 1:31:3 X-1 and NGC 1:31:3
X-2), and thus possible IMBH, it is somewhat surprising to note that the "ULX" X-:37 is
grouped with these sources in the diagram-despite the fact that its actual black hole mass
is at least 4 orders of magnitude higher than assumed by Miller et al. (2004). Thus, X-:37
illustrates the importance of follow-up spectroscopic confirmation of ITLXs, and sounds a
cautionary note regarding conclusions based on observations without such confirmation.
Table 6-1: Absolute Magfnitude Comparison to Quasars at z m 0.26
Observed Magfnitudes *
Typical Catalog Range
-23.9 to -20.8
-23.4 to -21.0
-25.6 to -23.8
43.7 to 44.1
" Absolute magfnitudes include uncertainties in each value.
Catalogf values were taken from separate sources: U and V
from Schneider et al. (2003), Ks from Barkhouse and Hall (2001),
and L, from Schartel et al. (1996).
Figure 6-1. U hand image (left) and Ks hand image (right) showing a 1-arcsecond
positional error circle centered on the X-ray source, X-37. Notice the bright
source well with in this circle.
6000 6200 6400 6600
Wa vele an gt h(A)
- -on source
- --off source
-~ ~ -
7800 8000 8200 8400 8600 8800
Figure 6-2. X-37 spectrum. Red text denotes quasar lines, blue text denotes lines from the
Antennae background emission.
ENVIRONMENTS OF X-RAY POINT SOURCES IN THE DWARF STARBURST
GALAXY NGC 1569
NGC 1569 is a dwarf, starburst galaxy vastly different from the large, merging
system of the Antennae. It contains two super star clusters (SSCs) (Arp and Sandage,
1985) and 16 X-ray sources, in the line of site, which range in luminosity from Lx=
~-5x1033-3x1036 eTgS S-1 (\l Iortin et al., 2002). At a distance of 2.2 Mpc (Israel, 1988), it
is closer than the Antennae at 19.3 Mpc (for Ho=75 km s-l Mpc-l), allowing me to probe
star cluster masses, down to ~-1000 Me-
My work presented in OsI Ilpter 4 defined the quantity, rl, relating the observed number
of XRBs per star cluster mass as a function of mass in the Antennae. I found that the
functional form of rl remained constant, demonstrating that the number of observed XRBs
simply scaled up linearly with cluster mass. In this chapter, I will compute rl for NGC
1569. The large contrast in galactic environments between the Antennae and NGC 1569
raises several fundamental questions regarding rl. For instance, does the functional form of
rl in NGC 1569 remain consistent with a single value, as it does for the Antennae? If so,
how does this value compare between these two, disparate galaxies.
In this chapter, I extend my study of X-ray sources and their IR counterparts to NGC
1569. In 97.1, I discuss my IR, J and Ks observations and the subsequent data reduction.
I identify IR counterparts to X-ray sources and analyze their photometric properties to
characterize theses sources in 97.2. I summarize results and give conclusions in 97.3.
7. 1 Observations and Data Analysis
7.1.1 Infrared Imaging
On January 17-19, 2005, I acquired IR observations of NGC 1569 using FLAMINGOS
on the K(PNO 4-meter telescope. FLAMINGOS has a field-of-view (FOV) of 10'x10' with
a plate scale of 0.32". The total exposure time of my final science image was ~-2 hours
in J and ~1 hour in Ks. These composite science images consisted of many, randomly
dithered exposures with minimum integration times of 60 seconds in J and 15 seconds
in Ks. To correct for the high sky background in the IR, I imaged the sky using similar
dithered exposures every 40 minutes in J and every 30 minutes in K,. The typical seeing
throughout the observing was ~1.2" in both filters.
I produced science quality images using the pipeline, the Florida A-i 1..1...11.; Tool
Born Of Yearning for high scientific quality data (FATBOY). FATBOY was developed by
C. Warner, A. H. Gonzalez, S.S. Eikenberry and other members of the FLAMINGOS-2
team. This software calibrates each image using flat fields and dark frames and corrects
for cosmic rays, bad pixels and image distortion. This software produced final images
for each dither run, which I drizzled together to produce master images in the J and K,
bands. In Figure 7-1, I show a 4x4-arcminute color image of NGC 1569.
7.1.2 Astrometric Frame Ties
Before performing a frame tie between Chandra coordinates and IR positions, I
used the astrometry routine bundled with FATBOY to fit Two Micron All Sky Survey
(2MASS) coordinates to the J and K, master images. This routine cross-correlates sources
between 2MASS and FLAMINGOS and then applies a 4th order polynomial fit to match
2MASS coordinates to the IR images. The fit produced an rms positional uncertainty of
As I demonstrated in OsI Ilpter 3, it is essential to produce a relative astrometric fit
between the X-ray images and images at other wavelengths to search for counterparts
to X-ray point sources. As I did in OsI Ilpter 3, I chose IR images as my initial link to
X-ray source positions in NGC 1569. This choice of wavelengths offers many benefits.
While extinction intrinsic to this dwarf galaxy is low, it is pII ll. v, ranging from Av =
0.0-0.8 (Relaflo et al., 2006). Considering that the near-IR will reduce these values by
a factor of 10, making extinction almost negligible, this wavelength regime facilitates
counterpart identification. Furthermore, since my IR observations are less confused than
HST observations of NGC 1569 (Anders et al., 2004) my chances of finding accurate
matches to X-ray sources are significantly improved. Finally, I note that IR and X-ray
wavelengths penetrate dust in similar r-wsi~, making it easier to correlate sources.
My astrometric frame tie method follows the procedure described in detail in OsI Ilpter
3. In summary, I identified seven bright, isolated sources in common between the IR and
X-ray images (Table 7-1). Applying a two-dimensional linear mapping function to the
Chandra right ascension and declination coordinates of these source (see Table 1 of Martin
et al., 2002) and the corresponding x and y pixel centroids in the FLAMINGOS K, image,
I produced a frame tie between the images in these two wavelengths. I achieved a fit with
a rms positional uncertainty of 0.2".
7.1.3 Infrared Photometry
I restricted photometry to an elliptical region centered on NGC 1569 with a long
axis of 2.4' and short axis of 1.0' (see Figure 7-3). This ellipse surrounds most of NGC
1569 and I was confident that most of the sources contained within it are star clusters.
Due to the numerous stars surrounding this galaxy, I estimated the stellar density in the
field outside of the ellipse encircling it and found there were 7 x 10-3 StasS per square
arcsecond. Multiplying this density by the area of the ellipse, implies that 11 of the
clusters in my sample are stars, assuming an isotropic distribution. In 22.214.171.124 below, I
will examine what effect this will have on my results.
I then performed J- and K,-band aperture photometry on 462 sources within this
ellipsoidal region using a self-written IDL program previously developed for the Antennae
(see C'!s Ilter 2 and Brandl et al., 2005, for additional details). I found a full width at
half maximum (FWHM) of 3.5 pixels (1.1") for both the Ks- and J-bands. I used a
photometric aperture with a 4.4-pixel radius in each band, corresponding to ~ So of the
I compensated for the crowded field in NGC 1569 by measuring sky background
in two separate annuli, from ~6-100- of the Gaussian PSF. In addition, I divided the
annuli into arcs for exceptionally crowded regions. Multiplying the mean and median sky
background in each annulus by the area of the central aperture and then subtracting this
from the central aperture flux yielded four separate flux measurements for the source.
I then averaged these four values to give me the source flux. When computing errors, I
considered variations in the sky background, asky, and Poisson noise, wasd. To calculate
asky I found the standard deviation of the four flux values for each source. Dividing the
mean source flux by the instrument gain and then taking the square root of these, I found
sasd. The known gain for FLAMINGOS is 4.9 e-l DN (Elston, 1998). Both asky and wasd
were added in quadrature to yield a total error in flux, afl,,
I then calibrated my photometry using the bright 2MASS star '2MASS 0 :11.877+645163'
at 04 30m58.77", +64 51m56.33" which is listed in the 2MASS database with J = 16.197,
aJ = 0.09 1!! I_ and Ks = 15.768, aKs = 0.1 mag. This star was in the FLAMINGOS
FOV for the NGC 1569 observations.
7.2 Results and Discussion
7.2.1 Identification of IR Counterparts to Chandra Sources
I restricted my analyzes to clusters with signal-to-noise (SNR) > 10 (see 126.96.36.199
below). I searched for IR counterparts to X-ray sources in NGC 1569 using the same
technique developed for the Antennae and discussed in detail in OsI Ilpter 3. In summary, I
used the 0.2" rms positional uncertainty to define circles with 2a and 3a radii. Those IR
sources lying within 0.4" (2a) of a Chandra X-ray source I defined as a -1I ning!" match.
If an IR source was between 0.4" and 0.6"1 (2-3a) of an X-ray source position, I defined
this as a p.. --h!I. ~" match. Including all 45 X-ray sources listed in Table 1 of Martin
et al. (2002), I found 19 -1ining~" matches and eight p.. --!Iub matches to IR sources.
Restricting X-ray sources to the 16 within the ellipse defined in 97.1.3 above, I found three
possible and four strong matches to IR star cluster counterparts in NGC 1569.
Following the method discussed in OsI Ilpter 3, I attempted to estimate the source
contamination by measuring the source density in a 0.8"-1.2" (4-6a) annular region
around the X-ray sources. Including all the X-ray sources in the 9.0' x 9.0' field, I expect
two with la uncertainty of +1.2/-0.7 of the -I in ng! matches and two with a la
uncertainty of +1.2/-0.4 of the p.. --!Iub-" matches are chance alignments of unrelated
objects. For those X-ray sources most likely associated with IR clusters in NGC 1569, I
expect one with la uncertainty of +0.8/-0.3 of the -I in ng! matches and two with a la
uncertainty of +0.7/-0.3 of the p.. --!Iub-" matches to be false associations.
These statistics indicate that most of the IR counterparts found in the field of NGC
1569 are real-of all 27 matches only four are chance alignments. But, the story is different
for IR counterparts identified on the face of NGC 1569, where three of the seven IR
matches to X-ray sources are most likely false identifications. This is not surprising,
considering the high IR source density in this galaxy. Nevertheless, the n 1..! i ~r~y of
the matches are real and source contamination is not significant enough to prevent a
meaningful photonietric study of them.
Figures 7-2 and 7-:3 show IR counterpart candidates in NGC 1569. Figure 7-2 is a
9.0' x 9.0' field around this galaxy showing all X-ray source positions identified by 1\artin
et al. (2002). Figure 7-3 is a 4.0' x 4.0' field with the ellipse used to designate those IR
sources that are most likely clusters associated with NGC 1569. In further analyzes, I only
consider the IR counterparts to X-ray sources in this region. I display subintages for all IR
counterparts to X-ray sources in Figures 7-4 and 7-5.
7.2.2 Photometric Properties of the IR Counterparts
188.8.131.52 Color magnitude diagrams
Using the 424 clusters with .7 and K. photometry, I produced a (.7 Ks) versus
K., color magnitude diagram (C'jl1); see Figure 7-6). Initially, I defined a magnitude
cutoff using the method discussed in C'!s Ilter :3. Specifically, I found all sources with
an error in magnitude, ma,, > 0.2 ning. I found median .7 and K., magnitudes for these
1 Found using confidence levels for small number statistics listed in Tables 1 and 2 of
sources and used these cutoffs; discarding all fainter objects. Unfortunately, due the
exceptional crowding and variable background flux of NGC 1569, a,no, was abnormally
high for luminous sources. Thus when I took the median magnitude of sources with a,n,,
>0.2 1! I_ the cutoff was unrealistically bright. Applying this cutoff excluded many
bright sources with high signal-to-noise. Therefore, I chose a different method to define
a magnitude cutoff; I measured the SNR for each source and selected only those sources
with SNR >10 to use in further analyzes
The low Galactic latitude of NGC 1569 (b = 110; Cotton et al., 1999) meant that
I could not ignore foreground extinction. Relaflo et al. (2006) estimated an optical
reddening of Az- = 1.64 nng. Using the extinction law defined in Cardelli et al.
(1989)(hereafter CC11 ), I found that this value corresponds to an IR reddening of A~s=
0.2 nmag in K., and A., = 0.4 ning in J. Subtracting these values front cluster magfnitudes,
I corrected for foreground reddening (see Figure 7-7).
I measured t;ood" photometry for six of the seven clusters associated with X-ray
sources. Here, as in ChI Ilpter 6, I defined t;ood" photometry as non-negative fluxes after
performing background sky subtraction. Comparing these six clusters to the general
population of clusters reveals a uniform spread in K. magnitude. This differs front
the Antennae, in which I observed that X-ray associated clusters are brighter in K.
Furthermore, there does not appear to be a significant trend in cluster color. The mean
(.7 Kis) color for all clusters is 0.81 ning with an unc~ertainty of a(J_ =) 0.021 wzhile
those clusters that are IR, c~ounterparlts have a mean of (.7 K~',) =0.74 with a .,_ =
I extended my analyzes of IR counterpart properties by computing K. luntinosities,
Ms-,, for all IR clusters. I corrected for extinction internal to NGC 1569 using the
reddening estimates presented in Relaflo et al. (2006). Using Ha/IH/ emission line ratios
to produce an extinction nmap of this galaxy, they found A44 ranges front 0.0-0.8 ning.
Using the CC11L reddening law, this implies a range in AKs of 0.0-0.1 mag, demonstrating
that the reddening within NGC 1569 in the IR is almost negligible. Subtracting the mean
intrinsic, AK,= 0.04 mag, and the distance modulus to NGC 1569, 26.7 mag, from the Ks
magnitude for each cluster, I computed M~K,. Here the K, magnitude has already been
corrected for foreground reddening (see 184.108.40.206 above). I defined the error in internal
reddening aS aA~s = 0.05, which covers the range in AKs for this galaxy. Adding aA~s in
quadrature with am, for each source, I computed errors in M~K,*
In Figure 7-8 I plot a histogram of M~Ks foT all CluSterS. This distribution has a mean
M~K, = -9.2 mag with an uncertainty of aM~ = 0.1 1!! I while the mean for those clusters
associated with X-ray sources is M~K, = -10.0 mag with a- = 0.7 mag. Performing
a K(-S test between the distribution of X-ray associated clusters and all clusters in
NGC 1569, I found a D-statistic of 0.32 and a probability of 0.50 that they are related
distributions. This so__~-1-;- that there is no distinction between these two groups.
As discussed in 97.1.3, 11 of identified "clusters" could be stars. I tested the
possible repercussions of stellar contamination by randomly removing 11 of the sources
in my sample. Then, I recalculated the mean M~Ks for two groups: all remaining clusters
and for all remaining clusters associated with X-ray sources. Repeating this procedure
10000 times, I produced a simulated distribution of mean M~K, for the two groups
mentioned above. For the first group of all remaining clusters, I found a M~Ks of -9.2 mag
with an uncertainty of aM~ = 0.1 mag, while for those remaining clusters associated with
X-ray sources I found M~K, = -10.0 mag and eM~s = 0.1 mag. These values are in close
agreement with the observed value of M~K, that I derived using all sources, demonstrating
stellar contamination does not significantly affect these findings.
7.2.3 XRB-to-Cluster Mass Fraction
Using the function, rl(M ), presented in Cl. .pter 4, I explored the relationship
between XRBs and cluster mass in NGC 1569. As a reminder, this function relates the
number of X-ray source detections per unit mass as a function of cluster mass (see Eq.
4-1). As I did with the Antennae, I first computed ty(M l) using cluster Ks-band flux, Fxs
to approximate cluster mass. In Figure 7-9, I show the distribution of ty(Fx,), where the
hin size is FK, = 1.5x104 data numbers (DN). This hin size is small enough to show a
trend in ty(Fx,), but large enough to contain at least two clusters with X-ray sources,
allowing me to compute errors in each hin. The error hars plotted on the graph are the
measurement uncertainties in the three values of ty(Fx,) added in quadrature with the
Poisson uncertainty of the mean ty(Fx,) for each hin. Notice each value of ty(Fx,) is
roughly consistent with a mean ty(Fes) = 2.4x10-6 DN-1, with an uncertainty, op=-
1.0 x10-6 DN-1
Next, I converted ty(M~c) into the physical units of solar mass, At As demonstrated
in C'!s Ilter 4, this conversion requires an assumption about cluster ages. Unlike the
Antennae, I could not make the simplified assumption that all clusters have similar ages
as there is a wide range ages in NGC 1569, front 4 Myr to 10 Gyr (Anders et al., 2004).
Furthermore, Waller (1991) -11---- -1 this galaxy underwent six starbursts in its history.
Clearly, assigning a single age to all clusters is not an accurate representation of the
cluster environment in NGC 1569. Instead, I tried finding matches between my IR clusters
and optical clusters discussed in Anders et al. (2004). Due to the complex structure of
NGC 1569 and the difference in resolution between HST and FLAMINGOS, I was only
able to find matches to 28 sources. Therefore, I adopted a different approach to estimate
cluster ages in this dwarf galaxy to derive ty(A ).
Following the procedure developed for the Antennae (see (I Ilpter 4), I selected ages
at random front a uniform distribution between 1 Myr and 10 Gyr, and assigned an age
to each cluster in nly sample. Applying the BC models, I computed each cluster's mass
based on the assigned age, produced a mass distribution, and then calculated a mean ty.
Performing a Monte Carlo (1!C) simulation, I recreated cluster mass distributions 10,000
times, producing a large sample of ty's with a mean of if = 3.274 x10-6 A(- and o =
1.2x10-s AGI1 (Figure 7-10).
7.2.4 Comparison with Models
As I did with the Antennae, I used the models presented in Sepinsky et al. (2005)
to calculate what these authors would predict for rl(Me) in NGC 1569. As a reminder,
these authors used StarTrack (B.1.I. i-in-1:; et al., 2002) to investigate the rate of XRB
formation and ejection from young, massive clusters. As discussed above, the cluster
ages in NGC 1569 are not well defined and span a wide range. Therefore, I picked the
maximum number of XRBs predicted by the models for a given age and thus computed
upper limits for rl(Me). Given a cluster mass of 5x106 M,, the models predict the cluster
will yield 6 XRBs, while a 5x104 Mo Cluster will only produce 0.05 XRBs. Recall an XRB
is associated with a cluster if it is within 6", which is equivalent to 6.4 pc at the distance
to NGC 1569. Since these models use a limiting X-ray luminosity of 5.0 x1035 eTgS S-1
and the actual limiting Lx for NGC 1569 is 5.5x1033 eTgS S-1, I used the cumulative XLF
power law slope of a~ = -0.5 (11 Istin et al., 2002) to estimate the number of observed
XRBs predicted by these models for this galaxy. Therefore, the models predict 29 XRBs
should be observed in a 5 x104 Mo Cluster, While 38 should be observed in a 5 x106 M
cluster. Expressing these model results as a fraction of XRB-to-cluster mass, Sepinsky
et al. (2005) predict rl(Al ) ranges from 9.6-11.4x10-6 --1. These values are only within
a factor of two of my measured value of rl in NGC 1569. Interestingly, if I place the
Antennae at the distance to NGC 1569 and scale its measured value of rl, I find rl(Me)
2.5x10-6 --1, Which is well in agreement with my measured value for NGC 1569.
In this chapter I extended my study of XRB environments to the dwarf, starburst
galaxy, NGC 1569. I produced a frame tie between Chandra X-ray coordinates and
FLAMINGOS pixel positions with an rms positional uncertainty of 0.2"'. With the frame
tie in place, I found seven of the 16 X-ray sources associated with NGC 1569 had IR
cluster counterpart candidates. Of these seven matches, four were -I in isg" matches
(< 0.4"' separation) and three were "possible" matches (0.4"-0.6"1 separation). After
performing a photonietric study of these IR cluster counterparts, I reached the following
1. After estimating the IR cluster density around each X-ray source, I predicted one
of the -lI nisg!" counterparts and two of the p.. --h!I I:" counterparts are chance alignments
of unrelated objects. While this is a high number of false matches considering the small
sample size of IR counterparts, most are real and therefore did not hamper a photonietric
study of these X-rali-- I--aciated clusters.
2. A K. vs. (J K~) C111)> revealed no trend in clusters with X-ray sources compared
to the general population of clusters. They occupied a wide range in K., magnitude, with
no obvious trend in color.
3. I found those clusters with X-ray sources are not significantly more luminous
than the general population of clusters. This result was substantiated by a K(-S test,
which revealed a relatively high probability of 0.5 that the distribution of clusters with
X-ray sources are related to all clusters. This differs front the Antennae where I found a
preference for more luminous clusters to harbor X-ray sources.
4. I estimated a constant value for if = 3.274 x10-6 A(~, with op- = 1.2 x10-8 -1~
If I placed the Antennae at the distance to NGC 1569 and scale its measured value of ty, I
found if = 2.5x10-6 ACI1, which is consistent with the derived value for NGC 1569. This
demonstrates if is similar for two, very different galactic systems. A larger galactic sample
is necessary to confirm the universality of ty for all environments.
5. Lastly, the models of Sepinsky et al. (2005) predict ty is only a factor of two higher
than my measured value for ty. Considering I chose the nmaxiniun number of XR Bs
predicted by the models, the model-predicted if is an upper limit, -II---- -1;l,-_ the models
produced in Sepinsky et al. (2005) roughly agree with my results.
Table 7-1: Common Sources Used for the Chandra/FLAMINGOS Astrometric Frame Tie
Source ID 0
9 .. .. .. .. .
20. .. .. .. .
37 .. .. .
04 30 30.91
04 30 39.03
04 30 44.21
04 30 48.93
04 30 58.37
04 31 14.02
04 31 16.85
+64 52 05.30
+64 50 12.80
+64 49 24.90
+64 49 40.10
+64 48 52.30
+64 51 07.90
+64 49 50.00
" ID numbers follow the naming convention of Martin et al. (2002).
b Units are magnitudes, from WIRC photometry. Values in parentheses indicate
Table 7-2: Potential IR Counterparts to Chandra X-Ray Sources
04 30 39.59
04 30 48.61
04 30 52.34
04 30 48.57
04 30 49.85
04 30 57.41
04 30 50.23
Chandra Src ID 0
I was unable to measure good photometry for the IR counterpart to Chandra source 23 and
therefore exclude magnitudes for it,
" ID numbers follow the naming convention of Martin et al. (2002).
b Chandra coordinates with an uncertainty of 0.5" (11 Istin et al., 2002).
" Positional offsets in units of seconds of arc from the Chandra coordinates to the FLAMINGOS
coordinates of the proposed counterpart
SUnits are magnitudes, from FLAMINGOS photometry. Values in parentheses indicate uncertain-
ties in the final listed digit.
Table 7-3: Summary of Potential IR Counterpart Properties.
KoK a~ (J K)
" Uncertainties in each value.
Figure 7-1. False color image of NGC 1569. Red is Ks and green and blue are J. Notice
redder clusters reside on the edge of the galaxy and bluer clusters are in the
center. The field-of-view of this image is 4.0' x 4.0'.