A Comprehensive Archival Chandra Search for X-ray Emission from Ultracompact Dwarf Galaxies

6.9 MB·

by  Viraj Pandya

Submitted to ApJ: November 30, 2015
PreprinttypesetusingLATEXstyleemulateapjv.01/23/15
A COMPREHENSIVE ARCHIVAL CHANDRA SEARCH FOR X-RAY EMISSION FROM ULTRACOMPACT
DWARF GALAXIES
Viraj Pandya1,2, John Mulchaey2, and Jenny E. Greene1
1DepartmentofAstrophysicalSciences,PeytonHall,PrincetonUniversity,Princeton,NJand
2TheObservatoriesoftheCarnegieInstitutionforScience,813SantaBarbaraStreet,Pasadena,CA91101,USA
(Dated: January 11, 2016)
Submitted to ApJ: November 30, 2015
ABSTRACT
WepresentthefirstcomprehensivearchivalstudyoftheX-raypropertiesofultracompactdwarf(UCD)
6
1 galaxies, with the goal of identifying weakly-accreting central black holes in UCDs. Our study spans
0 578 UCDs distributed across thirteen different host systems, including clusters, groups, fossil groups,
2 and isolated galaxies. Of the 336 spectroscopically-confirmed UCDs with usable archival Chandra
n sim−1a,gitnhge ogblosbearvlaXti-ornays,d2e1teacrteioXn-frraayc-tdioentecfoterdt.heImUpCoDsinpgopauclaotmiopnleitsen∼es3s%li.mOitfothfeL2X1>X-2ra×y-d1e0t3e8cteerdg
a
UCDs, seven show evidence of long-term X-ray time variability on the order of months to years. X-
J
ray-detected UCDs tend to be more compact than non-X-ray-detected UCDs, and we find tentative
7
evidence that the X-ray detection fraction increases with surface luminosity density and global stellar

velocity dispersion. The X-ray emission of UCDs is fully consistent with arising from a population
]
A of low-mass X-ray binaries (LMXBs). In fact, there are fewer X-ray sources than expected using a
naive extrapolation from globular clusters. Invoking the fundamental plane of black hole activity for
G
SUCD1 near the Sombrero galaxy, for which archival Jansky Very Large Array imaging at 5 GHz is
. publiclyavailable, wesetanupperlimitonthemassofahypotheticalcentralblackholeinthatUCD
h
p to be ∼< 105M(cid:12). While the majority of our sources are likely LMXBs, we cannot rule out central
- black holes in some UCDs based on X-rays alone, and so we address the utility of follow-up radio
o observations to find weakly-accreting central black holes.
r Keywords: accretion, galaxies: active, galaxies: dwarf, galaxies: supermassive black holes, X-rays:
t
s binaries, X-rays: galaxies
a
[

1 1. INTRODUCTION nations for this dark mass include a bottom-heavy IMF
v The formation mechanisms and central black hole oc- (Mieske & Kroupa 2008), a top-heavy IMF leading to
0 cupation fraction of ultracompact dwarf (UCD) galax- excessstellarremnants(Dabringhausenetal.2012),and
9 ies are unknown. First discovered more than fifteen supermassive black holes (SMBHs; Mieske et al. 2013).
6 years ago in the nearby Fornax cluster (Hilker et al. Recently, Seth et al. (2014) presented evidence for a
1 SMBH in M60-UCD1 that accounts for ∼ 10% of the
1999;Drinkwateretal.2000;Phillippsetal.2001),Hub-
0 total mass of that UCD, based on dynamical modeling
ble Space Telescope imaging and high-resolution spec-
. of spatially-resolved spectroscopy. The exciting detec-
1 troscopy have shown that UCDs are extremely dense
tion of a SMBH via dynamical modeling motivates us to
0 stellarsystemsratherthanmerelyunresolvedforeground
search for alternative evidence of central black holes in
6 stars in our own Galaxy. Although numerous stud-
UCDs, in this case using accretion. X-rays provide one
1 ies have established the distinctiveness of UCDs com-
: pared to considerably smaller and fainter globular clus- of the cleanest ways to identify accretion onto SMBHs
v because they are insensitive to dust and can probe the
ters(GCs)andconsiderablylargerandbrighterdwarfel-
Xi liptical(dE)galaxies(e.g.,Drinkwateretal.2003;Mieske low bolometric Eddington ratio regime (e.g., Gallo et al.
2010; Miller et al. 2015).
r et al. 2008b; Norris & Kannappan 2011; Brodie et al.
a 2011), the origin of UCDs is still not clear. There are AlthoughtheopticalandUV(e.g.,Mieskeetal.2008a)
properties of UCDs have been systematically explored,
thoughttobetwoprimaryformationchannelsforUCDs:
X-rays from UCDs as a whole remain largely unex-
(1) the merging of several smaller star clusters that pro-
plored. A handful of UCD discovery papers report
ducesasupermassivestarcluster,and(2)thetidalstrip-
merely whether or not the newly-classified UCDs host
ping of a larger galaxy that gives rise to a naked nucleus
an X-ray point source (e.g., the first bona fide X-ray-
(e.g., Brodie et al. 2011; Norris et al. 2014).
detected UCD was SUCD1 near the Sombrero galaxy;
One hint about the origin of UCDs comes from study-
Hau et al. 2009). Phillipps et al. (2013) present an
ingtheirdynamicalmasses. Interestingly,themoremas-
archival study of the X-ray properties of UCDs limited
sive UCDs show evidence for higher dynamical mass-to-
to Fornax alone, finding an X-ray detection fraction of
light ratios than those expected from stellar population
∼8%(downtoL ≈6×1037 ergs−1). Inthispaper,we
modelingalone,suggestingthatatleastsomeUCDshost X
presentthefirstcomprehensivearchivalsearchforX-rays
a large amount of “dark mass” (Ha¸segan et al. 2005;
from UCDs in a variety of host systems, with the goal
Mieske et al. 2008b, 2013). The most prominent expla-
of identifying accreting central black holes. As we will
Electronicaddress: [email protected] show,manyarchivalobservationsaredeepenoughtoalso
2 Pandya, Mulchaey & Greene
probe low-mass X-ray binaries (LMXBs), high-mass X- shows how many UCDs in each host system have usable
raybinaries(HMXBs),andultra-luminousX-raysources archival Chandra observations (see also subsection 3.1).
(ULXs). In addition to Table 1, we also refer the reader to the
Thispaperisorganizedasfollows. Insection2,wede- following compilations of physical properties of compact
scribeoursampleofUCDsandtheirphysicalproperties. stellar systems, including UCDs: Norris et al. (2014);
In section 3, we derive the X-ray properties of UCDs. In Norris & Kannappan (2011); Forbes et al. (2013); Mis-
section 4, we present correlations between X-ray emis- geld & Hilker (2011); Hilker (2011).
sion and physical properties, and address the origin of
the X-ray emission. We discuss our results in section 5 2.2. The Physical Properties of UCDs
and summarize in section 6. Appendix A gives the IDs We adopt the optically-derived physical properties of
ofarchivalChandra datasetsused,AppendixBdiscusses our UCDs from the literature (i.e., from one of the ref-
borderline detections, Appendix C comments on UCDs erences given in Table 1). A substantial number of our
in Perseus and NGC 3115, and Appendix D gives the UCDsdonothavefundamentalphysicalpropertiesavail-
optical properties and X-ray upper limits for non-X-ray- able. Here we briefly address the homogeneity of our
detected UCDs. literature-based physical properties.
Whenever possible, we adopt the global stellar veloc-
2. SAMPLE
itydispersion(σ)andr fromMieskeetal.(2008b)and
hl
In this section, we introduce our sample of UCDs and Mieske et al. (2013), since they accounted for aperture-
describe their optically-derived properties. and seeing-related effects in a homogeneous way. Oth-
erwise, we adopt the values from the original papers
2.1. Optically-Defined UCDs in the Literature if available. Note that different studies used different
We began our study by searching the literature for all structural models to derive sizes (Sersic, Nuker, or King
samples of optically-defined UCDs. We find a total of profiles). In addition, we adopt the dynamical masses
578 known UCDs distributed among thirteen different (Mdyn), and dynamical mass-to-light ratios in the V
host systems and four different environment types (clus- band (Mdyn/LV), derived by Mieske et al. (2013) and
ters, groups, fossil groups, and isolated galaxies).1 Of the original papers for ∼50 UCDs. We verified that the
those 578 UCDs, 195 are classified on the basis of their Mdyn definitionsusedindifferentstudiesareindeedcon-
photometry alone, and so are considered UCD “candi- sistent with each other by recomputing the virial mass,
dates.” It is likely that some of these 195 UCD can- which we estimate following Spitzer (1987) and Hilker
didates are actually background galaxies or foreground et al. (2007): Mvir ≈9.75× rhGlσ2.
stars. The remaining 383 UCDs are spectroscopically-
confirmedmembersoftheirrespectivehostsystems, and 3. X-RAYANALYSIS
socanbeconsideredbonafideUCDs.2 Tomaximizethe
Inthissection,wedescribeallworkrelatedtoderiving
sample size considered in this work, we consider both
the X-ray properties of UCDs.
candidate and spectroscopically-confirmed UCDs.
Although we do not attempt to homogenize the def- 3.1. Chandra X-ray Data Retrieval
inition of UCDs across different studies, we have tried
For each UCD, we used the CIAO3 tasks
to limit our sample to objects that were unambiguously
find chandra obsid and download chandra obsid
classifiedasUCDsintheoriginalstudies. Thedefinition
to search for archival ACIS4 imaging (excluding grating,
ofanunambiguousUCDvariesbetweendifferentstudies,
continuous clocking, and interleaved mode observations)
andcandependonM ,opticalhalf-lightradius(r ),or
V hl with off-axis angle5 θ < 10 arcmin. Although there was
acombinationofthetwo. Typically,MV ∼< −11magand no constraint on the depth of the Chandra data, the
rhl ∼> 10 pc for an unambiguous UCD, but there are ex- shortest exposure time for an individual observation is
ceptions(e.g.,Greggetal.2009;Brodieetal.2011). The
generally 5 ksec. The restriction on θ is crucial because
M distributionofourUCDshasmean−11.26magand
V the Chandra point spread function (PSF) is highly
standarddeviation0.97mag. Ther distributionofour
hl variable as a function of θ (and effective energy): the
UCDs has mean 18.2 pc and standard deviation 14.9 pc,
PSFstartstobecomeellipticalatθ >1.0arcmin,andas
with the largest value being ∼ 100 pc. Given that the
θ increases, the radius of a circular aperture needed to
boundary between GCs and faint UCDs is imprecisely
enclose 90% of the PSF also increases. If a point source
defined, our sample size would greatly increase if we ex-
has θ > 10 arcmin, it will look like locally-enhanced
tended our study to encompass borderline objects (e.g.,
background, have an artificially-extended structure, and
the ω Centauri-like objects described in Hilker 2011).
is unlikely to be found by most detection algorithms6.
InTable1,weorganizethelistofpreviousstudiesfrom
After all observations were downloaded, we dis-
which we initially drew our UCD samples and proper-
carded any observations whose Validation and Verifi-
ties, as a function of UCD host system. Table 1 also
cation (V&V) reports indicated severe problems. We
then reprocessed every observation with the CIAO task
1 Recently, Liu et al. (2015) presented a sample of several new
UCDs around M87, M49, and M60 in Virgo (a small fraction of
which are spectroscopically-confirmed). Although we do not in- 3 The Chandra Interactive Analysis of Observations software
clude these new UCDs in our study, we do not expect our results providedbytheChandraX-rayCenter(Fruscioneetal.2006).
tochangeifwedidincorporatethem. 4 TheAdvancedCCDImagingSpectrometerinstrument.
2ThereareafewmorehostsystemswhichcontainmostlyUCD 5 The off-axis angle is defined as the separation between the
candidates(noradialvelocities)butwhichhavenousablearchival positionofasourceandtheaimpointofanobservation.
Chandra imaging (e.g., the Antlia cluster; Caso et al. 2013). We 6 See section 4.2.3 of The Chandra Proposers’ Observatory
donotconsiderthosesystemshere. Guide: http://cxc.harvard.edu/proposer/POG/html/chap4.html
X-ray Emission from Ultracompact Dwarf Galaxies 3
Table 1
UCDSample
System Ntot Nspec Nchandra References
(1) (2) (3) (4) (5)
Centaurus 30 30 27 Mieskeetal.(2007a,2009)
Coma 87 41 35 Chiboucasetal.(2011);Madridetal.(2010)
Fornax 78 78 62 Bergondetal.(2007);Firthetal.(2008);Greggetal.(2009);Hilkeretal.(2007);Mieskeetal.(2008b,2013)
HCG22+90 21 21 21 DaRochaetal.(2011)
Hydra 83 54 67 Misgeldetal.(2008,2011);Wehner&Harris(2007)
NGC1023 15 0 1 Mieskeetal.(2007b)
NGC1132 6 6 6 Madrid&Donzelli(2013);Madrid(2011)
NGC3115 31 6 31 Jenningsetal.(2014)
NGC5128 27 27 23 Rejkubaetal.(2007);Mieskeetal.(2008b,2013);Pengetal.(2004);Martini&Ho(2004)
NGC7252 1 1 1 Schweizer&Seitzer(1998);Fellhauer&Kroupa(2005);Marastonetal.(2004)
Perseus 84 14 28 Pennyetal.(2012,2014)
Sombrero 1 1 1 Hauetal.(2009)
Virgo 114 104 70 Brodieetal.(2011);Francisetal.(2012);Ha¸seganetal.(2005);Sethetal.(2014);Norrisetal.(2014);Sandoval
etal.(2015);Straderetal.(2013);Zhangetal.(2015);Evstigneevaetal.(2008,2007);Paudeletal.(2010);Janz
etal.(2015)
Total 578 383 373 −
Note. —Referencesforthe13UCDhostsystemsprobedinthispaperaswellasthenumberofUCDsineachhostsystemsatisfyingourselectioncriteria. Col(1):
UCDhostsystem. Col(2): TotalnumberofUCDs. Col(3): Totalnumberofspectroscopically-confirmedUCDs. Col(4): TotalnumberofUCDswithatleastoneusable
Chandraobservation. Col(5): ReferencesfromwhichwedrewoursampleofUCDs.
chandra reprotousethelatestsoftwareandcalibration of the distances from each observation’s aim point to its
updates (CALDB version 4.6.7). Forgoing astrometric assignedcentroidposition(therearek centroidpositions
corrections,weadoptedthestandardChandra astromet- initialized by the user) via an iterative procedure. This
ricreferenceframebecauseithasanoverall90%absolute hastheadvantagethateachUCDisthenassignedtoonly
positionaluncertaintyof0.6arcsec,whichisgoodenough one of the k merged images – the image whose k-means
for our purposes. centroidpositionisclosesttotheUCD’sownopticalpo-
Table 1 gives the number of UCDs with at least one sition.
usableChandra observationasafunctionofhostsystem, We used the CIAO task merge obs to combine the
and Appendix A gives the list of identifiers for all obser- individual observations associated with each of the k
vations used in this work. We find that 373/578 ≈ 65% groups, resulting in k merged images. No binning of
of UCDs have at least one usable Chandra observation. the data was done so as to maintain the native resolu-
tion of Chandra (0.492 arcsec px−1). For each of the
3.2. Detecting X-ray Counterparts merged images, we constructed a merged PSF map by
doing an exposure map-weighted average of the individ-
We take a two-tiered approach to searching for X-ray
ual contributing observations’ PSF maps. The exposure
point sources at the known optical positions of UCDs.
map-weightedaverageaccountsforsignificantlydifferent
The first tier involves the use of merged exposures (in-
exposuretimesandeffectiveareasbetweentheindividual
dividually spanning several years) to gain the maximum
contributing observations.
possible sensitivity to faint sources. The second tier in-
On each of the k merged images, we ran the CIAO
volves the use of the individual exposures to account for
possible variable sources that may have been missed us- detectiontoolwavdetectinthreedifferentenergybands:
fullband(0.5-7.0keV),softband(0.5-2.0keV),andhard
ing the merged-observation approach.
band(2.0-7.0keV).Weusedeightdifferentwaveletscales
3.2.1. Tier I: Merged Observations (1, 2, 4, 8, 16, 24, 32, and 48 pixels), and required a
significance threshold of 10−6 which corresponds to one
The process of merging Chandra observations requires
false positive detection per ACIS CCD.
great care because of the significant degradation of the
To cross-match the UCDs to the X-ray sources found
PSF at θ >1 arcmin and the non-uniformity of the PSF
by wavdetect, we adopted a circular matching radius of
between different observations. Within most of the thir-
1.5 arcsec. If a wavdetect X-ray source is found within
teenUCDhostsystems,therearetypicallyseveraldiffer-
1.5 arcsec of a known UCD optical position, then that
ent observations, some of whose aim points can be sepa-
X-ray source is said to be coincident with the UCD. Al-
ratedbyafewdegrees. Combiningsuchhighly-separated
though the optical radii of UCDs are typically less than
observations into a single merged observation can con-
one arcsec, our choice of 1.5 arcsec for the X-ray cross-
taminate point sources (from low-θ observations) with
matching radius helps to account for the instability and
background counts (from high-θ observations), resulting
broadening of the Chandra PSF at θ > 1 arcmin. We
in a noisy and artificially-extended source which is un-
repeated this cross-matching for all three energy bands
likely to be significantly detected. Therefore, we typi-
butwebaseourdetectionclassificationsonresultsinthe
cally created multiple merged observations for each host
full band. There were no instances of a source being
systembyseparatingindividualobservationsintogroups
detected in the hard and/or soft bands but not in the
with similar aim points.
full band. We also did not have multiple unique X-ray
Wechosetomergeobservationswhoseaimpointswere
sources cross-matched to a single UCD.
within10arcminofeachotherfortworeasons: (1)thatis
We found nineteen X-ray point sources coincident
typically the maximum θ that a bright point source can
with the optical positions of UCDs using this merged-
have before it looks merely like locally-enhanced back-
observation approach. The full-band postage stamps of
ground, and (2) that was the search radius used to as-
thesenineteenX-raypointsourcesareshowninFigure1
sociate archival Chandra observations with each UCD.
and their optical properties are given in Table 3. In Ap-
We accomplished this task using the machine learning
pendix B, we discuss five borderline detections, while in
k-means clustering algorithm which minimizes the sum
4 Pandya, Mulchaey & Greene
Appendix C we discuss non-detections in Perseus and side the chosen source aperture by creating a circularly-
NGC 3115. symmetric PSF model at the UCD location using the
task arfcorr.
3.2.2. Tier II: Individual Observations We then did an exposure-weighted sum of the indi-
While the first tier search is sensitive to very faint vidualcontributingobservations’sourceandbackground
sources thanks to the significantly increased exposure region spectra, using the CIAO task combine spectra.
times of merged observations, there are several instru- After computing the background-subtracted spectrum8,
mental and astrophysical effects that could conspire to we assumed an absorbed power-law spectral model with
give a non-detection in a merged image. During the first the absorption fixed to the Galactic value at the optical
tier search, we assumed the following: (1) no intrinsic position of the UCD (Stark et al. 1992). In order to de-
variabilityinasource,(2)nosignificantbackgroundvari- terminethebest-fitphotonindex(γ)andnormalization,
ationsamongobservationsspanningmorethantenyears, we followed a two-step process.
and (3) no significant background contamination effects First, since 11 of the 21 X-ray-detected UCDs have
induced by merging observations whose aim points can sufficiently high net counts (> 100 net counts) to do
be separated by up to 10 arcmin. spectral fitting, we kept both γ and the normalization
For the above reasons, we also carried out a search on as free parameters for these 11 UCDs. We binned the
the individual (non-merged) observations. For each host spectra to ensure at least 15 counts in each energy bin.
system, we first ran wavdetect on all individual obser- UsingtheMonteCarlofittingalgorithm(moncar)imple-
vations in the same three energy bands (full, soft, and mented in Sherpa (Freeman et al. 2001) along with the
hard) and with the same parameters mentioned previ- Gehrels(1986)χ2 statistic,wefoundreasonablespectral
ously. We then cross-matched the UCDs to the X-ray fits; these are shown in Figure 2. We investigated the
point sources found in each individual observation and reason why χ2 < 1 for all fits as follows. For the five
red
looked for new matches missed during the first tier. Of UCDswith>300netcounts,weincreasedtheminimum
course, a UCD detected in a merged image may not be bin size from 15 to 40, 80, or 100 counts and re-fit the
detected in every individual observation contributing to spectra. When the error per energy bin decreased due
that merged image because the individual observations’ to the larger number of counts per bin, χ2 indeed in-
red
exposure times (and the UCD’s θ) may vary. In general, creased to ∼1, but our fits did not significantly change.
however, this unlocks time variability studies, which we We settled on a bin size of 15 counts because some of
will discuss in subsection 3.5. our sources do not have enough counts to justify a big-
We found two additional UCDs hosting X-ray point ger number and we want to apply a consistent bin size
sources using this individual-observation approach: for all 11 sources. Altogether, χ2 and visual inspection
red
gregg45 in Fornax and HGHH92-C12=R281 in NGC of the fits suggests that additional components beyond
5128. Both gregg45 and HGHH92-C12=R281 were de- the fixed-absorption power law (e.g., soft X-ray thermal
tected in only a single exposure. Their Chandra postage component or intrinsic absorption) are not necessarily
stampsareshowninFigure1andtheiropticalproperties justified.
are given in Table 3. For the remaining ten X-ray-detected UCDs, we fixed
γ to1.5(themedianofthebest-fitvaluesderivedduring
3.3. Deriving X-ray Luminosities the first iteration), and then we solved for the best-fit
Sincewavdetectisonlyadetectiontoolandnotappro- normalization using the same fitting approach described
priate for X-ray photometry, here we describe our algo- above. Thefluxandits90%confidenceintervalineachof
rithmforderivingthefluxandluminosityofeachX-ray- thethreeenergybands(full,soft,andhard)werederived
detected UCD in the three energy bands (full, soft, and by drawing 1000 random values from a Gaussian distri-
hard). Thereisneitherastandardpracticeadvocatedby butionwithmeansettothebest-fitmodelfluxandstan-
theChandraX-rayCenternoramethodwidelyadopted dard deviation based on the uncertainty of that best-fit
throughouttheliteratureforderivingX-rayluminosities modelflux. ThiswasaccomplishedwiththeSherpa task
from a merged Chandra dataset. It is crucial that we sample flux. Finally, we converted the fluxes to lumi-
derive X-ray luminosities using the merged datasets be- nosities with the luminosity-distance formula, assuming
cause most of our detections are based on the merged the distance to each UCD’s host system (the distances
images. are given in Table 3).
For each X-ray-detected UCD, we extracted its spec- The X-ray properties of the X-ray-detected UCDs are
trumfromeachoftheindividualobservationsusedforits given in Table 4.
detection. We used the CIAO task srcflux which auto-
matically defines a circular source aperture centered on UCD. In some cases, there was a bright source nearby and so we
manually resized or moved the background apertures to prevent
the optical position of each UCD with the radius chosen
contamination.
to enclose 90% of the PSF (the radius depends on θ and 8ThemajorityofourX-ray-detectedUCDs’sourceregionshave
the effective energy of the full-band: 2.3 keV). A back- total (i.e., including the underlying background) counts (cid:29)100 so
groundannulusisalsoautomaticallydefinedbysrcflux subtractingthebackgroundshouldbevalid. Intheremaininglow-
count cases, a standard practice advocated in the literature is to
withinnerradiusequaltothesourceapertureradiusand
simultaneously fit the source and background spectra. However,
outer radius equal to five times the source aperture ra- modeling the Chandra X-ray background is not trivial, especially
dius.7 srcflux automatically also applies PSF correc- with few background counts, and we do not want to introduce
tions to account for the fraction of counts falling out- additionalsystematicuncertainties. Thecrucialpointisthat,after
subtractingthebackgroundforthefewlow-countUCDs,wedonot
endupinthe(non-Poissonian)negativecountcase.
7 Wevisuallyinspectedthesourceandbackgroundaperturesin
every individual observation associated with each X-ray-detected
X-ray Emission from Ultracompact Dwarf Galaxies 5
Figure 1. Approximately 30”×30” postage stamps from the full band (0.5-7.0 keV) merged image for each of the 21 X-ray-detected
UCDs. The green cross marks the optical position of each UCD. The green circle is the exposure time-weighted average of the circular
apertures adopted to enclose 90% of the PSF in the individual observations (see subsection 3.3). The color-coding represents the count
rate,whereineachimageblueisfaintest,redisintermediate,andwhiteisbrightest.
3.4. Deriving Upper Limits ManyofourUCDsarecoveredinmultiplearchivalob-
servationsspanningnearlyfifteenyears,andherewetake
The majority of our UCDs are not directly detected
advantageofthatfacttosearchfortimevariabilityusing
by wavdetect, and so for those UCDs, we derive upper
the data products created during the X-ray luminosity
limits on the full-band (0.5-7.0 keV) flux and luminos-
derivation process (see subsection 3.3). We restrict the
ity.9 In many cases, we have extraordinarily deep imag-
followingtimevariabilityanalysisonlytoX-ray-detected
ingandcanplaceanimportantconstraintonX-rayemis-
UCDs because we know that: (1) they are definitely X-
sion from UCDs thought to host central BHs. However,
ray emitters, and (2) they are securely detected in at
defining and deriving upper limits is a highly non-trivial
least one archival observation. Individual archival ob-
and controversial topic (e.g., Kashyap et al. 2010).
servations in which an otherwise X-ray-detected UCD
It is even more difficult in our case because we want
is not “re-detected” by wavdetect are still included in
to get the deepest possible upper limits by leveraging
our analysis to test the possibility that the UCD’s X-ray
the summed exposure times of merged observations, but
source “turned off.” However, we discard any observa-
the individual contributing observations each have dif-
tions in which srcflux can only measure a 90% upper
ferent effective exposures and aim points, and can span
limit instead of a net count rate with 90% confidence
across more than 10 years. Tools such as WebPIMMS10
bounds; this is typically a problem only for observations
are not appropriate because they make two assump-
tions that are generally flawed for large archival stud- with exposure times ∼< 10 ksec.
As is standard practice in active galactic nucleus
ies such as ours: (1) the source is perfectly on-axis, and
(AGN)variabilitystudies(e.g.,Nandraetal.1997;Ptak
(2) the sensitivity of ACIS for only one observing cy-
et al. 1998; Vaughan et al. 2003; Thornton et al. 2009;
cle is used. In practice, the source is not perfectly on-
Young et al. 2012), we searched for variability in net
axis (θ > 0 arcmin), and merged observations comprise
count rate space rather than flux or luminosity space.
datasets taken in many different observing cycles. Due
Following Luo et al. (2013), we took a two-pronged ap-
tothesecomplications, andbecausethereisnostandard
proachtotestforlong-termX-raytimevariability. First,
algorithm recommended for deriving upper limits based
we fit each UCD’s “X-ray light curve” with a zero-slope
on merged Chandra datasets, we describe our method in
straight-line model and computed χ2 . Second, we
Appendix D. In short, we derived upper limits for the red
computed the maximal time variability indicator, σ ,
non-X-ray-detected UCDs by computing the number of max
based on the maximum difference between any two net
net counts that would correspond to a local background
count rates within a given UCD’s light curve, normal-
fluctuation with Poisson probability 1%.
ized by their uncertainties (again, see Luo et al. 2013,
TheopticalpropertiesandX-rayupperlimitsfornon-
and references therein):
X-ray-detected UCDs are given in Appendix D.
|R −R |
σ =max i j , (1)
3.5. X-ray Time Variability max i,j(cid:113)
σ2 +σ2
Ri Rj
9 We did not derive upper limits for UCD candidates because: where Ri and Rj are the net count rates in the ith and
(1) all X-ray-detected UCDs are spectroscopically-confirmed, and jth observations, and σ and σ are the uncertainties
(2)UCDcandidatesarenotusedwhencomputingdetectionfrac- Ri Rj
in those net count rates, respectively. The uncertainties
tionsordeterminingcorrelations.
10Seehttps://heasarc.gsfc.nasa.gov/docs/software/tools/ σR include the contribution from the background and
pimms.html. were derived using the Gehrels (1986) approximation.
6 Pandya, Mulchaey & Greene
Figure 2. Spectra and best-fit models for the subset of our X-ray-detected UCDs with > 100 net counts. The best-fit photon index
(γ)with68%(1σ)confidencebounds,assumedGalacticabsorptionvalueinthedirectionoftheUCD,andreducedχ2 areshownineach
subplot.
X-ray Emission from Ultracompact Dwarf Galaxies 7
We present the results of the two approaches in Ta-
ble 2. Adopting slightly more restrictive criteria than
Luo et al. (2013), we consider a UCD to be a variabil-
ity candidate if χ2red ∼> 2 and σmax ∼> 3. The following
five UCDs clearly satisfy both criteria: F-17, HGHH92-
C6, HGHH92-C7, HGHH92-C23, and HCH99-18. M60-
UCD1 also exhibits evidence for variability, albeit with
somewhat lower values of χ2 and σ . HGHH92-C21
red max
is a borderline case due to a single spike in net count
rate. Thedatasuggestlong-termvariabilityontheorder
of months to years for these seven UCDs, and their me-
dian σ ≈ 5.3. Many of the non-variable UCDs have
max
fewerobservationsavailableforcharacterizingtheirlong-
term variability as compared to the variable sources (see
the N column in Table 2), but their observations do
obs
span several years (with the exception of ∼3 months for
gregg25). The long-term X-ray light curves of the seven
variability candidates and the non-variable SUCD1 are
shown in Figure 3.
Lastly,wealsolookedforshort-termvariability(within
a single observation – on the order of ksec) from each
of the X-ray-detected UCDs using only observations in
which the source region yielded > 50 net counts. First,
we used the CIAO task dither region to account for
any time periods where our sources of interest were
ditheredoverbadregions(off-chip, badpixel, etc.) since
that would induce artificial time variability due to in-
strumental effects. Then, we employed the CIAO task
glvary which uses the Gregory & Loredo (1992) algo-
rithmto: (1)sliceanobservation’seventfileintodifferent
timebins,and(2)lookforsignificantdeviationsbetween
the time-binned events. We did not find any compelling
evidenceforshort-termtimevariabilitywithinanyofthe
UCDs’ observations.
Ideally, we would like to derive a characteristic long- Figure 3. Thelong-termX-raylightcurvesforthesevenvariabil-
term variability timescale and relate that to a hypothet- itycandidatesandthe non-variable SUCD1. The modifiedJulian
ical black hole mass. However, we do not have a suffi- dates (MJD) shown are relative to the Chandra-specific reference
MJD:MJD =50814days(January1,1998).
cient number of data points in the light curves to detect ref
a characteristic long-term variability timescale or am-
sioninUCDs: LMXBs(seediscussioninsubsection4.4),
plitude. We also do not detect compelling evidence for
and weakly-accreting central black holes (see discussion
short-termvariabilityandthuscannotderiveanestimate
in subsection 5.2). Here, we briefly discuss the expected
of,e.g.,thecrossingtime,whichdirectlyprobestheblack
properties of the X-ray emission if dominated by one or
hole mass (e.g., see section 2 of Peterson 2001). Never-
the other.
theless, the long-term variability of F-17, HGHH92-C6,
At luminosities below ∼ 5×1038 erg s−1, the LMXB
HGHH92-C7, HGHH92-C21, HGHH92-C23, HCH99-18,
X-rayluminosityfunctionofGCssteeplyrises(e.g.,Fab-
and M60-UCD1 warrants further investigation.
biano 2006; Kim et al. 2009)). Thus, our X-ray detec-
4. ORIGINOFTHEX-RAYEMISSION tions at these low L are likely dominated by accretion
X
The most immediate conclusion to be drawn from our ontostellar-massobjectsinLMXBs. AthigherLX,how-
archival study is that across thirteen different host sys- ever, we can hope to be sensitive to accretion both onto
temsthatcollectivelyhost578UCDs,wefindonly21X- central black holes and in LMXBs. For a given stellar
ray-emitting UCDs (all coincidentally spectroscopically- mass,theLMXBfractionshouldrisewithstellardensity
confirmed). Considering only the 336 spectroscopically- duetotheincreasedprobabilityofproducinginteracting
confirmedUCDswithusablearchivalX-raydataandim- stellar binaries (Jord´an et al. 2004). In fact, as we will
posingacompletenesslimitofL >2×1038 ergs−1 (to quantify in subsection 4.4, given the optical properties
X
account for the variable depths in the different host sys- of the UCDs, we would expect many more LMXBs than
tems), we measure a global X-ray detection fraction of our number of X-ray detections.
4/149 ≈ 3%. In this section, we explore links between To estimate the likely emission from a central accret-
theX-rayandphysicalpropertiesofUCDswiththegoal ingblackhole, considerthattheEddingtonluminosity11
of determining the origin of the X-ray emission. of a 106M(cid:12) black hole is ∼1.3×1044 erg s−1.12 Adopt-
4.1. LMXBs versus Central Black Holes 11 Ledd=1.3×1038(MBH/M(cid:12))ergs−1.
Our main concern is to quantify the expected contri- 12 A106M(cid:12) blackholewouldaccountfor∼10%ofthetypical
UCD’s dynamical mass, similar to the black hole mass fraction
butionfromthetwomostrelevantsourcesofX-rayemis- foundinM60-UCD1(Sethetal.2014).
8 Pandya, Mulchaey & Greene
Table 2
Long-termX-rayTimeVariability
UCD System N σ χ2 Variable?
obs max red
(1) (2) (3) (4) (5) (6)
F-17 Fornax 7 3.99 3.86 Yes
F-22 Fornax 4 1.20 0.56 No
F-34 Fornax 9 1.47 0.34 No
F-3 Fornax 3 0.88 0.40 No
F-18 Fornax 9 1.05 0.21 No
gregg25 Fornax 2 0.58 0.33 No
gregg26 Fornax 4 0.84 0.32 No
HGHH92-C6 NGC 5128 19 4.16 2.22 Yes
HGHH92-C7 NGC 5128 12 6.22 3.69 Yes
HHH86-C18 NGC 5128 21 1.77 0.51 No
HGHH92-C21 NGC 5128 19 3.04 1.16 Maybe
HGHH92-C23 NGC 5128 24 9.82 11.79 Yes
HGHH92-C36=R113 NGC 5128 4 0.24 0.03 No
HGHH92-C37=R116 NGC 5128 5 0.94 0.31 No
HCH99-16 NGC 5128 23 1.16 0.27 No
HCH99-18 NGC 5128 24 7.17 5.68 Yes
SUCD1 Sombrero 3 0.99 0.50 No
M60-UCD1 Virgo 6 2.54 1.52 Yes
HGHH92-C12=R281 NGC 5128 6 0.68 0.16 No
gregg45 Fornax 8 0.98 0.38 No
Note. —Thelong-termX-raytimevariabilitypropertiesofX-ray-detectedUCDswithmultipleobservations. Col(1): UCD
name. Col(2): UCDhostsystem. Col(3): Numberofindividualobservationsinwhichsrcfluxcouldmeasureanetcountrate
with 90% confidence bounds (regardless of detection). Col (4): The maximal time variability indicator. Col (5): The reduced
chi-squared from assuming a zero-slope straight-line fit model. Col (6): Whether we consider the UCD to exhibit long-term
X-ray time variability.
ing the relation from Ho (2008) that L ≈16×L for We now ask whether the X-ray detection fraction de-
bol X
thetypicallow-luminosityAGN,ourtypicalX-rayupper pends on the physical properties of the UCDs. If the
limitsof1037and1039ergs−1allowustoprobebolomet- most massive and compact UCDs with the highest stel-
ric Eddington ratios (≡ L /L ) down to ∼ 7×10−8 lar velocity dispersions hosted either central accreting
bol edd
and ∼ 7×10−6, respectively, assuming a 106M cen- black holes (e.g., Mieske et al. 2013) or a large popula-
(cid:12)
tral black hole. Given that the X-ray-detected M60- tion of dark stellar remnants (e.g., Dabringhausen et al.
UCD1’s dynamically-confirmed SMBH is radiating at 2012), thenwewouldexpectpositivephysicaltrendsbe-
L /L ≈ 4×10−7, we certainly have the ability to tween the X-ray detection fraction and various physical
bol edd
detect some accreting central black holes if they exist. properties. It is a bit complicated to address this vital
The X-ray spectral and variability properties of UCDs questionbecausewehavesuchvariableX-rayupperlim-
canprovidesomeadditionalcluesaboutthenatureofthe its. Therefore, when computing the detection fraction
X-rayemission. However,themedianderivedphotonin- at a given LX completeness limit, we include all X-ray-
dex (assuming an absorbed power law spectral model; detectedUCDs(abovethatcompletenesslimit)aswellas
γ ≈ 1.5) is consistent with both a low-luminosity AGN non-X-ray-detectedUCDswhoseupperlimitsaregreater
(Ho2008)andanLMXBinthehard/lowstate(Fabbiano than the completeness limit. We bin the detections as a
2006;Remillard&McClintock2006). Itisnotclearwhat function of four different physical properties: absolute
conditions would be required to produce the hard X-ray V-band magnitude, dynamical mass-to-light ratio in the
spectra(γ ≈1)ofsomeUCDs(F-34,HGHH92-C21,and V band,globalstellarvelocitydispersion,andsurfacelu-
SUCD1). In terms of variability, the greatest argument minositydensity(Σ= LV ). Asaconcretelowerbound
2πr2
infavorofLMXBswouldcomefromshort-termvariabil- hl
on the detection fraction upper limits, we also show the
ity which would imply short crossing times. Alas, we
actual detection fraction in each bin (without including
find no evidence for short-term variability (on the order
thenon-X-ray-detectedUCDs),assumingacompleteness
of ksec), and instead detect only long-term variability
limit of L >1038 erg s−1.
(spanning years). Since the data do not suggest any pe- X
AscanbeseeninFigure4,theupperlimitsonthede-
riodicity, the variability could arise from either a central
tection fractions show an increasing trend between some
black hole or a population of LMXBs.
bins for each measured property. However, the trends
The X-ray properties alone are therefore inconclusive
are not consistent for the different completeness limits
in addressing the possible domination of LMXBs over
that we explore, and they are also not consistent with
central black holes. In contrast, considering both the
the trends seen for the actual detection fractions assum-
X-ray and optical properties together may help reveal
ing a completeness limit of L > 1038 erg s−1. More
concretedifferencesbetweentheLMXBandcentralblack X
observations are needed, particularly at the high end of
hole scenarios.
each physical property’s distribution where the number
of UCDs is lower. Tentatively, there does appear to be
4.2. X-ray Detection Fraction
an increasing trend in the actual detection fractions (as-
X-ray Emission from Ultracompact Dwarf Galaxies 9
alal of
Ref(13)Mieske+13Mieske+13Mieske+13Mieske+08aMieske+08aGregg+09Gregg+09Gregg+09Mieske+13Mieske+08aMieske+08aMieske+08aMieske+08aMieske+08aMieske+08aMieske+08aMieske+08aRejkuba+07Hau+09Strader+13Sandoval+15m.Col(3):Optic.Col(8):Optic(cid:12)Expectednumber
σM/LnVLMXBdyn(10)(11)(12)28.0±1.42.1±0.40.90129.1±1.53.2±0.90.29015.3±1.51.2±0.40.31831.3±1.5−−19.1±1.4−−−−−−−−−−−20.7±1.51.5±0.40.36919.12.70.11814.70.9±0.30.26417.23.9±1.40.06229.51.7±0.61.39314.72.2±0.80.15612.01.5±0.50.1738.42.2±0.80.01318.73.7±1.40.04313.13.10.02725.0±5.63.00.10668±54.9±0.70.302−−13.875mergedobservations.Col(2):UCDhostsyste[mag].Col(7):logof(dynamicalmass)/M10ghtratiointheV-band[M/L].Col(12):(cid:12)(cid:12)
Table3OpticalPropertiesofX-ray-detectedUCDs SystemRADecDMlog(M/M)rv(cid:12)Vhlrad(2)(3)(4)(5)(6)(7)(8)(9)Fornax54.64083−35.4325520.9−11.376.8±5.83.461390.0Fornax54.82375−35.4250320.9−11.226.9±5.95.341054.0Fornax54.55291−35.4825020.9−10.836.3±5.64.191639.0Fornax54.54333−35.4017220.9−11.70−−1626.0Fornax54.67500−35.5536120.9−11.70−−2024.0Fornax54.55008−35.6658920.9−−−1307.0Fornax54.55937−35.4455020.9−−−1377.0Fornax54.67975−35.4670820.9−−−1574.0NGC5128201.34245−43.046003.42−11.016.5±5.74.3854.5NGC5128201.52254−42.942333.42−11.096.8±6.37.5594.9NGC5128201.41616−43.083863.42−10.146.2±5.73.2479.8NGC5128201.46975−43.096223.42−10.396.7±6.27.0461.3NGC5128201.47741−42.990393.42−11.666.8±6.33.3673.7NGC5128201.53220−42.866753.42−9.916.3±5.83.6702.7NGC5128201.54408−42.895193.42−9.836.0±5.63.3611.7NGC5128201.37620−42.993003.42−10.086.3±5.812.1458.2NGC5128201.38167−43.000783.42−11.387.0±6.613.7455.0NGC5128201.27383−43.175193.42−10.526.6310.8±1.4440.4Sombrero190.01304−11.667869.95−12.317.314.7±1.41293.1Virgo190.8999011.5347016.7−14.208.3±7.524.2±0.51290.0Virgo186.3451618.1815816.7−12.55−1.85±0.9658.0optedfromtheliterature.∗meansthattheUCDwasdiscoveredduringthesecondtiersearchusingindividualratherthanCol(4):OpticaldeclinationofUCD[deg].Col(5):DistanceadoptedtoUCD[Mpc].Col(6):AbsoluteV-bandmagnitude−1−1Heliocentricradialvelocity[kms].Col(10):Globalstellarvelocitydispersion[kms].Col(11):Dynamicalmass-to-li17)ofSivakoffetal.(2007).Col(13):PrimaryreferencefortheUCD’sopticalproperties.
UCD(1)F-17F-22F-34F-3F-18gregg25gregg26∗gregg45HGHH92-C6HGHH92-C7HHH86-C18HGHH92-C21HGHH92-C23HGHH92-C36=R113HGHH92-C37=R116HCH99-16HCH99-18∗HGHH92-C12=R281SUCD1M60-UCD1M85-HCC1Notes:Col(1):UCDnameadrightascensionofUCD[deg]. half-lightradius[pc].Col(9):LMXBsaccordingtoequation(
10 Pandya, Mulchaey & Greene
Table4X-rayPropertiesofX-ray-detectedUCDs UCDSystemEXPθsRNCNCNCFFFLLLHRγxFHSFHSFHS(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)+0.17+0.35+0.17+0.17+0.35+0.17F-17Fornax2403.60.374.9169.61±44.4926.16±35.66144.36±26.71−14.30−14.44−14.7538.4238.2737.97−0.691.49−0.27−0.76−0.30−0.27−0.76−0.30+0.20+0.36+0.18+0.20+0.36+0.18F-22Fornax935.30.598.0126.79±42.6029.12±35.5699.67±23.60−14.05−14.11−14.5038.6738.6138.22−0.551.49−0.35−0.94−0.35−0.35−0.94−0.35+0.12+0.21+0.13+0.12+0.21+0.13+0.14F-34Fornax2795.30.138.8272.86±47.4786.16±37.62186.37±29.02−13.98−14.11−14.5538.7438.6138.17−0.371.19−0.17−0.48−0.19−0.17−0.48−0.19−0.14+0.11+0.21+0.13+0.11+0.21+0.13+0.14F-3Fornax4144.50.897.0288.53±47.8592.87±37.84195.32±29.36−14.15−14.29−14.6838.5638.4238.04−0.361.30−0.16−0.49−0.18−0.16−0.49−0.18−0.14+0.36+0.43+0.39+0.36+0.43+0.39†F-18Fornax2045.50.519.732.87±43.220.96±36.9831.92±22.46−14.68−14.37−15.1138.0438.3537.61−0.941.49−0.87−1.00−0.98−0.87−1.00−0.98+0.41+0.42+0.39+0.41+0.42+0.39†gregg25Fornax744.70.828.718.12±40.128.00±35.0010.12±19.69−14.49−14.05−14.9238.2238.6737.80−0.121.49−1.04−0.95−0.91−1.04−0.95−0.91+0.44+0.44+0.44+0.44+0.44+0.44†gregg26Fornax1942.71.214.75.99±40.201.00±34.784.91±20.25−15.31−14.58−15.7437.4138.1436.97−0.661.49−1.05−1.01−1.02−1.05−1.01−1.02+0.34+0.42+0.39+0.34+0.42+0.39∗†gregg45Fornax2654.81.497.645.20±41.894.00±35.1741.04±22.83−14.87−14.70−15.3137.8538.0237.41−0.821.49−0.93−1.04−0.89−0.93−1.04−0.89+0.18+0.31+0.19+0.18+0.31+0.19+0.23HGHH92-C6NGC51286124.40.357.2175.11±46.0267.95±38.24108.00±25.70−14.61−14.72−15.0936.5336.4236.06−0.231.55−0.29−0.86−0.35−0.29−0.86−0.35−0.22+0.24+0.37+0.25+0.24+0.37+0.25+0.27HGHH92-C7NGC51284455.40.459.2106.16±43.3230.37±36.2176.71±23.89−14.70−14.72−15.1336.4536.4236.02−0.431.70−0.59−1.05−0.62−0.59−1.05−0.62−0.26+0.06+0.12+0.06+0.06+0.12+0.06+0.09HHH86-C18NGC51285935.00.577.9597.54±51.70210.46±40.53388.96±32.19−13.91−14.08−14.4037.2337.0636.75−0.31.52−0.07−0.16−0.08−0.07−0.16−0.08−0.09+0.06+0.10+0.08+0.06+0.10+0.08+0.10HGHH92-C21NGC51286396.10.2610.0602.25±52.19311.54±42.82293.46±29.98−13.86−13.95−14.5837.2937.2036.570.030.95−0.07−0.12−0.09−0.07−0.12−0.09−0.10+0.02+0.03+0.02+0.02+0.03+0.02+0.04HGHH92-C23NGC51286525.10.208.23586.70±82.611312.04±57.912286.62±59.09−13.15−13.31−13.6538.0037.8337.50−0.271.49−0.02−0.03−0.02−0.02−0.03−0.02−0.04+0.25+0.39+0.27+0.25+0.39+0.27HGHH92-C36=R113NGC51281067.1−12.969.42±41.415.67±35.0964.75±22.12−14.05−14.10−14.5637.0937.0536.59−0.841.49−0.61−0.94−0.68−0.61−0.94−0.68+0.18+0.32+0.20+0.18+0.32+0.20HGHH92-C37=R116NGC51282916.5−11.1124.92±42.9635.33±36.0789.46±23.40−14.28−14.39−14.7936.8636.7536.36−0.431.49−0.37−0.76−0.40−0.37−0.76−0.40+0.09+0.24+0.10+0.09+0.24+0.10+0.15HCH99-16NGC51288373.40.324.3324.79±47.66100.02±38.21226.65±28.61−14.46−14.68−14.8836.6836.4736.27−0.391.73−0.12−0.51−0.14−0.12−0.51−0.14−0.14+0.02+0.03+0.02+0.02+0.03+0.02+0.04HCH99-18NGC51288433.10.133.93386.95±81.371279.31±57.692119.43±57.56−13.26−13.42−13.7937.8837.7337.35−0.251.41−0.02−0.03−0.02−0.02−0.03−0.02−0.04+0.31+0.44+0.34+0.31+0.44+0.34∗HGHH92-C12=R281NGC51282168.21.4316.269.48±42.2425.31±36.0644.04±22.08−14.50−14.38−15.0136.6436.7636.13−0.271.49−0.79−0.95−0.86−0.79−0.95−0.86+0.25+0.33+0.23+0.25+0.33+0.23+0.23SUCD1Sombrero1923.10.294.2106.37±41.9948.12±36.0159.21±21.72−14.17−14.23−14.7937.9037.8537.28−0.11.05−0.51−0.88−0.57−0.51−0.88−0.57−0.22+0.21+0.42+0.20+0.21+0.42+0.20+0.20M60-UCD1Virgo3071.40.151.9107.77±42.0126.87±35.4081.90±22.74−14.69−14.62−15.1037.8437.9037.42−0.511.61−0.46−1.19−0.41−0.46−1.19−0.41−0.20+0.39+0.43+0.38+0.39+0.43+0.38†M85-HCC1Virgo390.70.153.122.68±40.460.54±35.0522.14±20.30−14.25−14.02−14.6838.2838.5137.85−0.951.49−0.95−1.00−0.94−0.95−1.00−0.94†Notes:Col(1):UCDnameadoptedfromtheliterature.∗meansthatX-rayemissionfromtheUCDwasdetectedduringthesecondtiersearchusingindividualratherthanmergedobservations.indicatesaborderlinedetection(netcountsareconsistentwithzero);seeAppendixBformoreinformation.Col(2):UCDhostsystem.Col(3):Totalexposuretime[ksec].Col(4):Exposuretime-weightedaverageoff-axisangleofallobservationsusedforfluxderivation[arcmin].Col(5):SeparationbetweentheopticalpositionoftheUCDandthecenterofthecross-matchedX-raysource’sapertureasdefinedby[arcsec].Col(6):Exposuretime-weightedwavdetectaverageradiusofthecircularaperturesusedtoenclose90%ofthePSFintheindividualobservations[arcsec].Col(7):Netcountsinfullband(0.5-7.0keV)with1σGehrels(1986)error[counts].Col(8):Netcountsinhard−1−2band(2.0-7.0keV)with1σGehrels(1986)error[counts].Col(9):Netcountsinsoftband(0.5-2.0keV)with1σGehrels(1986)error[counts].Col(10):Fluxinfullbandwith90%confidenceinterval[ergscm].Col−1−2−1−2−1(11):Fluxinhardbandwith90%confidenceinterval[ergscm].Col(12):Fluxinsoftbandwith90%confidenceinterval[ergscm].Col(13):Luminosityinfullbandwith90%confidenceinterval[ergs].Col−1−1(14):Luminosityinhardbandwith90%confidenceinterval[ergs].Col(15):Luminosityinsoftbandwith90%confidenceinterval[ergs].Col(16):Classicalhardnessratio:(C−C)/(C+C).Col(17):Best-fitHSHSpowerlawphotonindex(setto1.49–themedianofthe11derivedindices–ifγwasfixedduetoNC<100).F