The Megasecond Chandra X-Ray Visionary Project Observation of NGC 3115 (II): properties of point sources
Dacheng Lin, Jimmy A. Irwin, Ka-wah Wong, Zachary G. Jennings, Jeroen Homan, Aaron J. Romanowsky, Jay Strader, Gregory R. Sivakoff, Jean P. Brodie, Ronald A. Remillard
aa r X i v : . [ a s t r o - ph . H E ] J un Draft version July 16, 2018
Preprint typeset using L A TEX style emulateapj v. 5/2/11
THE MEGASECOND
CHANDRA
X-RAY VISIONARY PROJECT OBSERVATION OF NGC 3115 (II):PROPERTIES OF POINT SOURCES
Dacheng Lin , Jimmy A. Irwin , Ka-wah Wong , Zachary G. Jennings , Jeroen Homan , Aaron J.Romanowsky , Jay Strader , Gregory R. Sivakoff , Jean P. Brodie , Ronald A. Remillard Draft version July 16, 2018
ABSTRACTWe have carried out an in-depth study of low-mass X-ray binaries (LMXBs) detected in the nearbylenticular galaxy NGC 3115, using the Megasecond
Chandra
X-Ray Visionary Project observation(total exposure time 1.1 Ms). In total we found 136 candidate LMXBs in the field and 49 in globularclusters (GCs) above 2 σ detection, with 0.3–8 keV luminosity L X ∼ –10 erg s − . Other than13 transient candidates, the sources overall have less long-term variability at higher luminosity, atleast at L X & × erg s − . In order to identify the nature and spectral state of our sources, wecompared their collective spectral properties based on single-component models (a simple power lawor a multicolor disk) with the spectral evolution seen in representative Galactic LMXBs. We foundthat in the L X versus photon index Γ PL and L X versus disk temperature kT MCD plots, most of oursources fall on a narrow track in which the spectral shape hardens with increasing luminosity below L X ∼ × erg s − but is relatively constant (Γ PL ∼ kT MCD ∼ Chandra bandpass. Therefore we identified the track as the NS LMXB soft-state track and suggested sourceswith L X . × erg s − as atolls in the soft state and those with L X & × erg s − as Zsources. Ten other sources (five are transients) displayed significantly softer spectra and are probablyblack hole X-ray binaries in the thermal state. One of them (persistent) is in a metal-poor GC. Subject headings:
X-rays: binaries — globular clusters: general — Galaxy:stellar content — X-rays:individual (NGC 3115) INTRODUCTION
In a low-mass X-ray binary (LMXB), a neutron star(NS) or a stellar-mass black hole (BH) accretes matterfrom a Roche-lobe filling, low-mass companion star viaan accretion disk. Copious X-rays are produced in theinner disk (Shakura & Sunyaev 1973). In the case of NSLMXBs, there is also strong X-ray emission from theboundary layer formed by the settling of the accretionflow onto the NS surface (Inogamov & Sunyaev 1999;Popham & Sunyaev 2001). Our knowledge of LMXBswas revolutionized thanks to 16 years of intensive obser-vations of such objects in our Galaxy by the
Rossi X-rayTiming Explorer ( RXTE , Bradt et al. 1993).BH LMXBs constitute the majority of BH X-ray bi-naries (BHBs) known in our Galaxy, though a fewBHBs with a high-mass stellar companion also ex-ist. BHBs exhibit three main X-ray spectral states:the hard state, the thermal state, and the steep Space Science Center, University of New Hampshire,Durham, NH 03824, USA, email: [email protected] Department of Physics and Astronomy, University of Al-abama, Box 870324, Tuscaloosa, AL 35487, USA Eureka Scientific, Inc., 2452 Delmer Street Suite 100, Oak-land, CA 94602-3017 University of California Observatories, Santa Cruz, CA95064, USA MIT Kavli Institute for Astrophysics and Space Research,MIT, 70 Vassar Street, Cambridge, MA 02139-4307, USA Department of Physics and Astronomy, San Jos´e State Uni-versity, One Washington Square, San Jos´e, CA 95192, USA Department of Physics and Astronomy, Michigan State Uni-versity, East Lansing, Michigan, MI 48824, USA Department of Physics, University of Alberta, Edmonton,Alberta, T6G 2E1, Canada power law (SPL) state (Remillard & McClintock 2006;McClintock & Remillard 2006). These states are nor-mally described with spectral models consisting of twomain continuum components: a standard thermal multi-color disk (MCD, diskbb in XSPEC, Arnaud 1996) anda Comptonized component (often modeled with a singlepower law, PL, or the Comptonization model comptt inXSPEC). The hard state, which tends to occur belowseveral percent of the Eddington luminosity ( L Edd ), ischaracterized by a strong Comptonized component dom-inating the spectra at least above ∼ kT MCD ≪ kT MCD ∼ >
75% in 2–10 keV (as defined inRemillard & McClintock 2006). A weak Comptonizedcomponent is also often present but only dominates athigh energies. The SPL state tends to occur at veryhigh luminosities ( & L Edd ), and the hallmark of thisstate is a strong Comptonizaton component with pho-ton index Γ PL ∼ . Fig. 1.— : Chandra
X-ray image of NGC 3115. The im-age is false-colored using adaptively smoothed (with theCIAO task csmooth), exposure corrected images in 0.5–1.2 keV (red), 1.2–2 keV (green), and 2–7 keV (blue).The D ellipse of the galaxy is also shown. The dif-fuse emission near the galaxy center comes from diffusehot gas, unresolved LMXBs, and other unresolved stellaremission (Wong et al. 2011, 2014).ior is observed in transient NS LMXBs.There are two main classes of NS LMXBs, atoll andZ sources (Hasinger & van der Klis 1989; van der Klis2006), named after the patterns that they trace out inX-ray color-color diagrams (CDs) or hardness-intensitydiagrams (HIDs). Atolls radiate at ∼ L Edd andtrace out their patterns in CDs/HIDs on timescales ofweeks to months. They have two main distinct spectralstates, i.e., hard (Γ PL .
2, extending to 100 keV or above)and soft states (most emission .
20 keV). The hard statetends to be observed at low luminosity ( . L Edd ), whilethe soft state is normally observed at high luminosity(larger than a few percent L Edd ). Spectra in the “tran-sitional” state between these two are also occasionallyseen. Z sources are more luminous than atolls, at near orabove Eddington luminosity, and they trace out roughlyZ-shaped tracks in CDs/HIDs within a few days (i.e.,faster than atolls), with X-ray spectra that are generallysoft. Thanks to a recent transient Z source XTE J1701-462, which exhibited the Z-source characteristics whenit was accreting at near or above Eddington luminosityand transitioned to an atoll during the decay of its 2006-2007 outburst, we now know that Z and atoll sourcesare essentially the same type of objects at different massaccretion rates (Lin et al. 2009; Homan et al. 2010). Un-like BHBs, whose tracks in the CDs/HIDs tend to showlarge scatter, mainly due to the presence of the SPL state,NS LMXBs tend to trace out clear narrow tracks in the CDs/HIDs, which depend mostly on the accretion rate(Done & Gierli´nski 2003; Remillard & McClintock 2006;Fridriksson et al. 2015).The spectral modeling for NS LMXBs is complicatedby the presence of the boundary layer emission and isrelatively controversial, compared to BHBs. Lin et al.(2007) used a similar spectral model to the one used forBHBs (described above) except for an additional single-temperature blackbody (BB) to describe the boundarylayer and were able to infer L ∝ T trends for both thedisk and the boundary layer in the soft state of two atolls.Such a trend for the disk is often observed for the thermalstate of BHBs and is expected if the disk is essentiallythermal and is truncated at the innermost stable circu-lar orbit (ISCO). Lin et al. (2009) applied this model toXTE J1701-462 and also observed the L ∝ T trendsfor both the disk and the boundary layer in its atollstage. However, such trends were not observed in itsZ-source stage, because the inner disk and the boundarylayer both reach the local Eddington limit in this stageso that the increase in the accretion rate tends to lead toan increase in the emission area with relatively constanttemperature (increasing the inner disk radius at constantinner disk temperature for the accretion disk).Our knowledge of X-ray binaries in nearby galaxies hassignificantly increased since the launch of XMM-Newton and
Chandra X-ray Observatory in 1999. In particu-lar,
Chandra ’s superb spatial resolution and excellentsensitivity (Weisskopf et al. 2002) allow for populationstudies of X-ray binaries in a single galaxy with snap-shots, which is important for understanding the originand evolution of such sources. To fully understand thesesources, one key task is to identify their nature (i.e.,BH versus NS X-ray binaries) and the X-ray spectralstate. To achieve this, early studies tried to stack sourcesin luminosity ranges in order to improve the statistics(Maccarone et al. 2003; Irwin et al. 2003). Recently, rel-atively detailed studies of individual sources in nearbygalaxies were carried out (e.g., Brassington et al. 2010,2012; Burke et al. 2013; Barnard et al. 2013, 2014), butlimited by statistics, these studies have been mostly fo-cused on the few most luminous sources (above several10 erg s − ) in each galaxy. It turns out that the dif-ferentiation between BH and NS X-ray binaries is gen-erally very difficult. This is because most sources haveX-ray spectra that are relatively hard (Γ PL < .
5) in the
Chandra and
XMM-Newton bandpass (about 0.3–8 keV)and it is difficult to determine whether they are BHBs inthe hard state or NS LMXBs in the soft state (they areunlikely to be NS LMXBs in the hard state due to thehigh luminosities of the sources studied). NS LMXBs inthe soft state can appear hard in such an energy bandbecause of the presence of the hot boundary layer com-ponent in the X-ray spectra (e.g., Lin et al. 2010, 2012a).NGC 3115 was selected as the target of a 1 Megasec-ond
Chandra
X-ray Visionary Project (XVP) in Cycle13. The goals were to study the gas flow inside theBondi radius of the central supermassive black hole andobtain a deep look at the X-ray binary population ofa normal early-type galaxy. The former has been pre-sented in Wong et al. (2014). For the X-ray binaries, wehave presented the X-ray luminosity function in Lin etal. (2014, Paper I hereafter), and here we concentrate
TABLE 1Observation Log
Notation Obs. ID Date Exposure Offset a (ks) (arcmin)1 2040 2001-06-14 35.8 1.52 11268 2010-01-27 40.6 0.13 12095 2010-01-29 75.6 0.14 13817 2012-01-18 171.9 0.05 13822 2012-01-21 156.6 0.06 13819 2012-01-26 72.9 0.07 13820 2012-01-31 184.1 0.08 13821 2012-02-03 157.9 0.09 14383 2012-04-04 119.4 0.310 14419 2012-04-05 46.3 0.311 14384 2012-04-06 69.7 0.3 a Aim point offset from observation 13820. on the detailed properties of discrete sources, such as thelong-term spectral and flux variability and the spectralcharacteristics. One main goal of our study is to iden-tify their nature, but different from previous studies, wewill achieve this by systematic comparison of our sourcesat various luminosity levels with Galactic X-ray binaries.NGC 3115 is a lenticular (S0) galaxy with an age of ∼ Chandra on this galaxyis ∼ ∼ erg s − , making it one of Chandra ’s best observed normalearly-type galaxies.In Section 2, we describe the source detection, the cal-culation of flux, the simple spectral fits, the measurementof long-term and short-term variability, and multiwave-length cross-correlation. In Section 3, we present the var-ious properties of X-ray binaries in NGC 3115, includingthe long-term variability and spectral characteristics thatare used to classify the sources. We further discuss thepossible nature of our sources in Section 4. We presentour conclusions in Section 5. DATA ANALYSIS
Observations and Source Detection
The eleven
Chandra observations of NGC 3115 arelisted in Table 1. They were made during three epochs:one in 2001, two in 2010, and nine in 2012. We hereafterrefer to them as Obs 1–11 in chronological order (Ta-ble 1). All observations used the imaging array of theAXAF CCD Imaging Spectrometer (ACIS; Bautz et al.1998). We analyzed the data with the
Chandra
Interac-tive Analysis of Observations (CIAO, version 4.6) pack-age. We reprocessed the data to apply the latest calibra-tion (CALDB 4.5.9) and the subpixel algorithm (Li et al.2004) using the CIAO script chandra repro . Backgroundflares are only clearly seen in Obs 1, 5, and 6, for onlyvery short durations. We excluded them if they arehigher than 4 σ above the mean background level. Inthis way, we excluded 1.2 ks, 3.6 ks, and 2.7 ks data forObs 1, 5, and 6, respectively. The final exposure used foreach observation is given in Table 1.To increase the detection sensitivity, we combined all11 observations to create a deep merged observation (Obs Esum hereafter) using the CIAO script merge obs . Tocorrect for relative astrometry between different obser-vations, we created new aspect solution files by compar-ing the source list from each single observation to the source list from a single reference observation, which wechose to be the longest observation (13820). We onlyused bright sources ( > σ ) with off-axis angles < ′ in thecross-correlation. The average separation residuals of thesource matches are 0.1 ′′ after relative astrometry correc-tion. The new aspect solution files were then applied tothe event files and subsequent analysis.We used the CIAO wavdetect wavelet-based source de-tection algorithm (Freeman et al. 2002) to search for dis-crete X-ray sources. The search was done twice, firstover the single observations to determine the relative as-trometry correction described above and the second timeover the (relative astrometry corrected) single observa-tions and the merged one. The count images were madein the broad ( b ) energy band 0.5–7.0 keV adopted inthe Chandra
Source Catalog (CSC), while the exposuremaps were constructed at the corresponding monochro-matic effective energy, (i.e., 2.3 keV, Evans et al. 2010).The point spread function (PSF) maps used correspondto the 50% enclosed counts fraction (ECF) at 2.3 keV.For the merged observation, the PSF map was obtainedby averaging those from single observations weighted bythe exposure. We used two different resolutions: one atsingle sky pixel resolution (0 . ′′ / ′ × ′ centering at the center of NGC3115. The subpixel binning images were used to improvethe spatial resolution in the crowded field near the cen-ter of the galaxy. The limiting significance level was setto 10 − , which formally corresponds to ∼ × σ ) statistical positional uncertainty or if their 50% PSFcircular regions overlap with each other across differ-ent observations. The statistical positional uncertaintiesthat we used are based on Equation (12) of Kim et al.(2007), which provides the 95% statistical positional un-certainty as a function of the source net counts and theoff-axis angle based on a large number of simulationsusing the Chandra simulation tool MARX. The 95% po-sitional uncertainties are converted to the 99.73% errorsby multiplying by a factor of 1.405 (this assumed a two-dimensional, circularly symmetric Gaussian distributionfor the source position from wavdetect ). As a check on theabove matching criteria, we also tested a smaller search-ing radius by using the 95% statistical positional uncer-tainty only (the 50% PSF overlapping is not used). In theend 95 more sources were found. From visual inspection,we found that six of them lie in the central crowded fieldand two at the CCD edge, thus probably having largesystematic positional uncertainties and explaining theirrelatively large offsets from detections from the merged Lin et al.observation. The remaining sources generally have theirseparations from those of the merged observation muchless than the size of their region ellipse from wavdetect .Thus in the end we did not treat any of these 95 sourcesas new sources and used the source list obtained aboveusing the 99.73% statistical positional uncertainty andthe 50% PSF circular region.
Flux and Spectral Characterization
After obtaining a merged source list, we extracted thespectrum for each source for each single observation. Thesource region was set to be a circle enclosing 90% of thePSF at 2.3 keV. The background region was set to bea concentric annulus, with inner and outer radii of twoand five times the source radius, respectively. Nearbysources, if present, were excluded from the source andbackground regions, but the inner circular source regionenclosing 50% of the PSF was not excluded. We usedthe CIAO task mkacisrmf to create the response matrixfiles and the CIAO tasks mkarf and arfcorr to createthe point-source aperture corrected auxiliary responsefiles. The spectral and response files corresponding to themerged observation were created using the CIAO task combine spectra .The background-subtracted count rates (but notaperture-corrected) in different energy bands were ob-tained from the spectral files. To correctly determineconfidence bounds for low count limits, we used theCIAO task aprates . The conversion from the countrates to the fluxes were based on the response files andassumed an absorbed PL spectral shape with Γ PL =1 . N H = 4 . × cm − (Kalberla et al. 2005). Throughout the paper, all fluxesand luminosities quoted (including those obtained fromspectral fits described below) are corrected for Galacticabsorption (but not intrinsic absorption, unless indicatedotherwise).To characterize the spectral properties of our sources,we calculated the hardness ratios HR = ( H − S ) / ( H + S ),where S and H are the energy fluxes in the soft and hardenergy bands, respectively, using the method of Bayesianestimation (Park et al. 2006). We also carried out sim-ple spectral fits to spectra above 4 σ using two single-component models: a PL and a MCD. Due to the lowstatistics of most sources, we binned the spectra to havea minimum of one count per bin and used the C statisticin the fits. Both models included absorption (we usedthe wabs model in XSPEC; we found no significant ef-fect on our results if we chose other absorption modelssuch as tbabs , due to little absorption of most of oursources), with the minimum set to be the Galactic valueof N H = 4 . × cm − . Long-term and Short-term Variability
The variability of the source was measured in severalaspects. We defined the long-term flux variability as V var = F max /F min and the significance of the differenceas S var = F max − F min ( σ + σ ) / , (1)where F max and F min are the maximum and minimum0.5–7.0 keV fluxes of a unique source among the single ob-servations, with the corresponding errors σ max and σ min , respectively. We only used detections with the flux abovetwice the error ( σ ) when calculating F max (if no detec-tions above 2 σ , the one with the highest significance wasused as F max ), while we used 2 σ as the flux for detectionswith the flux less than 2 σ when calculating F min .We measured the short-term variability using theGregory-Loredo algorithm (Gregory & Loredo 1992) im-plemented by the CIAO tool glvary (Evans et al. 2010).It splits the events into multiple time bins and looks forsignificant deviations. The variation of the effective areawith time was taken into account and was obtained byanother CIAO tool dither region . The different degreesof confidence are indicated by the parameter of “variabil-ity index”, which spans values within [0, 10] and is largerfor variability of higher confidence (Evans et al. 2010). Multiwavelength Cross-correlation
Accompanying the
Chandra
XVP observation, a sixpointing
Hubble Space Telescope ( HST ) mosaic obser-vation in the F475W and F850LP filters (hereafter g and z ) using the Advanced Camera for Surveys (ACS)was also acquired in the field of NGC 3115. The to-tal field of view of this mosaic observation is slightlylarger than the D region of NGC 3115, which has asemi-major axis of a = 3 . ′ (10.2 kpc), a semi-minoraxis of b = 1 . ′ (3.5 kpc) and a position angle of 40 ◦ (de Vaucouleurs et al. 1991). The galaxy was also im-aged in g , r , i -band filters on 2008 January 4th usingSuprime-Cam on the 8.2-m Subaru telescope. In PaperI, we have cross-correlated our X-ray sources with the360 globular clusters (GCs) from the HST /ACS mosaicimaging and the 421 ones from the Subaru/Suprime-Camimaging (Arnold et al. 2011; Jennings et al. 2014). Thematch was identified if the separation is less than the99.73% positional uncertainty (combining the X-ray andoptical components). The maximum separation allowedis 2 ′′ in order to limit the spurious rate. We note that be-fore the cross-correlation, the systematic offset betweendifferent source lists has been corrected through multi-ple steps: the Subaru/Suprime-Cam astrometry was reg-istered to the USNO-B1.0 Catalog (Monet et al. 2003)first, then the HST /ACS astrometry was registered to theSubaru/Suprime-Cam one, and in the end the astrometryof our X-ray sources was registered to the
HST /ACS one(therefore the absolute astrometry of our X-ray sourceshas been corrected).Table 2 lists the 37 matches with
HST /ACS GCs (23have the g − z color > .
13 and thus are red/metal-rich,while the other 14 have g − z < .
13 and are blue/metal-poor, following the division in Jennings et al. 2014) andthe 7 matches with Subaru/Suprime-Cam GCs identifiedin Paper I. In Paper I, we also identified five other sourceswhose optical counterparts were not classified as GCsby Jennings et al. (2014) but were assumed to be GCcandidates by us (Table 2). Four of them (S12, S53,S65, and S79) are within 0.25 D , thus very unlikely tobe AGNs (Paper I). The other one (S92), at an outerregion, has an optical counterpart with the size and thecolor typical of GCs, though it has a radial velocity fromthe spectroscopic measurement (238 km s − , Table 2)lower than typical values seen in other GCs ( >
350 kms − ).Adopting the same matching criteria, we also searchedfor the non-GC counterparts to our X-ray sources outside0.25 D from these optical observations. Such matchesare most probably cosmic X-ray background sources(CXBs), especially active galactic nuclei (AGNs), in-stead of the optical counterparts to field LMXBs. Theoptical emission of field LMXBs can achieve the max-imum when the accretion rate is near the Eddingtonlimit and the companion is an evolved star (so brightoptical emission from both a large disk and a largecompanion), as seen in the Galactic Z source Cyg X-2 (van Paradijs & McClintock 1995). However, suchsources are still below the detection limit of our opticalimages, by ∼ ± ◦ , ± ◦ ,and 180 ◦ and using X-ray sources above 4 σ , we estimatedthe rate of spurious matches to be about 3% and 5% forthe HST /ACS counterparts and Subaru/Suprime-Cammatches, respectively.From visual inspection, we found that some very faintsources in the
HST /ACS observation were not detectedby the tool SExtractor used by Jennings et al. (2014).Concentrating on the region within (0.25–1) D , we vi-sually identified five X-ray sources coincident with suchfaint sources and assume them to be AGNs. Within0.25 D , we also visually found one source, i.e., S65,with a faint counterpart not detected by Jennings et al.(2014), and we have assumed it to be a GC, as mentionedabove. RESULTS AND DISCUSSION
The Source List and Identification
Figure 1 shows the false-colored
Chandra
X-ray im-age of NGC 3115. We detected 525 unique sources fromthe merged and single observations. After eliminatingsources below 2 σ (i.e., the net counts within the 90% PSFregion divided by the error less than 2), we are left with490 sources. We found that the ACIS-S1 chip, whichis well outside the galaxy ( > D ), shows some brightstreaks, especially at energies . . D ,they are expected to be dominated by LMXBs and arethus assumed to be such objects, except those with non-GC optical counterparts, which we identify as AGNs.The LMXBs identified in this way are expected to becontaminated by some AGNs, because our optical im-ages are not very deep. Limiting to the 137 relativelybright sources with L X , max & × erg s − (i.e., flux & . × − erg s − cm − ) within D , which arethe main targets of our study of this paper, we iden-tify 9 AGNs. As we found in Paper I that the CXBdensity in our field is consistent with the average valuefrom Georgakakis et al. (2008), to within 20%, we expect < .
8, which is negligibly small for our purposes.Sources outside D should be dominated by AGNs, aswe only expect 1.5 field LMXBs above 4 × erg s − based on the IR light in the K s band outside D (Paper Fig. 2.— : The long-term variability versus the maximum0.5–7 keV luminosity for all LMXBs excluding those inthe central a = 10 ′′ ellipse. The top panel is for GCLMXBs, with the blue squares and red diamonds forthe HST /ACS blue/metal-poor and red/metal-rich GCsfrom Jennings et al. (2014), respectively, and the greencircles for other GCs (the Subaru/Suprime-Cam GCs andthe extra five
HST /ACS GC candidates identified by us;Section 2.4). Two transients, both in red GCs, are plot-ted with red triangles. The bottom panel is for fieldLMXBs. The blue triangles denote the transient candi-dates. In both panels, the filled symbols denote BHCs(see text for details).I). Therefore, we assumed all sources outside D to beAGNs, except those coincident with GCs, the supersoftX-ray source (SSS) S109, three coronally active stars inour Galaxy (they are coincident with stars and show softX-ray spectra and possible stellar flares), two galaxies(i.e., due to hot gas emission in galaxies; they are coinci-dent with galaxies in the optical images and show soft X-ray spectra), and the BHC S179 (Sections 3.2 and 3.3.2;it is slightly outside D but is also identified as a fieldLMXB, considering its possible transient nature and softspectra). In total we have 136 candidate field LMXBsand 49 candiate GC LMXBs (Table 2). We also markedthe 13 transients and 10 BH X-ray binary candidates(BHCs) in the source type column. Their identificationwill be described in the following sections.Table 4 gives the counts, fluxes and hardness ratios ofour sources in various energy bands in the merged obser-vation (also in the high-state and low-state observationsfor transients identified in the next section). The countsand fluxes of our sources in various energy bands in singleobservations are given in Table 5. Lin et al. (a) Transient candidates with the outburst detected in the first epoch.(b) Transient candidates with the outburst detected in the second epoch. Fig. 3.— : Long-term light curve (left panels, with the source number and long-term variability V (i.e., V var inSection 2.3)) and sample spectra (right panels, with annotations for the observations and spectral models used) forspecial sources. Long-term Variability
The observations of NGC 3115 by
Chandra span morethan a decade, which is ideal for investigation of the long-term variability of LMXBs. Figure 2 shows the depen-dence of the long-term variability V var on the maximum0.5–7 keV luminosity L X , max for all candidate LMXBs ex-cept those (24) within the central elliptical region withsemi-major axis a = 10 ′′ and eccentricity and positionangle following the D ellipse (this region is too crowdedand the source extraction is subject to large systematicerrors). We find that except for some transients (trian-gles, see below), V var generally decreases with L X , max forboth bright GC and field LMXBs with L X , max & × erg s − . At lower luminosities, V var seems to increase with L X , max , especially for field LMXBs. However, thisis most probably artificial due to the detection limit ofthe observations; most of these sources have the min-imum luminosity less than 2 σ and we calculated their V var using the 2 σ upper limit of the minimum luminos-ity. Some GCs might contain multiple LMXBs (PaperI), but considering that we detect variability for all GCLMXBs, such source blending effects should not be sig-nificant. The long-term stability of the most luminoussources that we see in NGC 3115 is also seen in othergalaxies (e.g., Irwin 2006).To search for transients, i.e., quiescent sources with asingle outburst, we concentrated on the 152 sources thatare either within 2 D (including those within the cen-tral a = 10 ′′ eclipse) or coincide with GCs and have at (c) Transient candidates with the outburst detected in the third epoch. Fig. 3.— : (continue) Lin et al. (d) Transient candidates with the outburst detected in the second and third epochs.
Fig. 3.— : (continue) (e) A SSS with kT BB = 86 +4 − eV(f) A very hard X-ray source (Γ PL = 0 . ± .
1) in a GC, which is a candidate NS LMXB with a strong magneticfield or a high inclination.
Fig. 3.— : (continue) (g) Persistent BH X-ray binary candidates. S96 is in a GC while others are in the field. S108 has only fivedetections because it is in the CCD gap in other observations.
Fig. 3.— : (continue)0 Lin et al.least one detection above 4 σ . If we required transientsto have V var ≥ < σ ) in all observations in at leastone epoch, and have the long-term light curve consis-tent with a global outburst (instead of an irregular, largevariation), we are left with 13 candidates. We note thatwe did not use a very strict condition on the variabilityto select transients, as some persistent Galactic LMXBsare known to vary by factors of >
10 (e.g., Homan et al.2009; Maccarone et al. 2010). Therefore we cannot ruleout that some transient candidates that we found mightbe just highly variable persistent sources. We plot thelong-term light curves of the transient candidates in theleft panels in Figure 3 a–d. There are three active in thefirst epoch (Figure 3a), one in the second epoch (Fig-ure 3b), six in the third epoch (Figure 3c), and three ac-tive in both the second and the third epoch (Figure 3d).We note that S36 (Figure 3c), which was bright in thethird epoch, seems to show some emission in the first twoepochs, making it not formally a good transient candi-date, but this might be due to extended emission nearthe galactic center. We also note that only S8 and S103are coincident with GCs.As most of our observations were made within threemonths in the third epoch, only outbursts in this epochare relatively well monitored. Among the six transientcandidates that were active only in the third epoch (Fig-ure 3c), three (S103, S125, and S198) clearly showed fluxevolution during the outburst. We probably detected thedecay of the outburst for S103 and S125. For S198 wehad relatively good coverage of the outburst, includingthe rise, the peak and the decay. Moreover, for S198,we fortunately caught a fast rise ( < ∼ X-ray Spectral Properties
The Hardness Ratios
Figure 4 shows the luminosity versus hardness ratiodiagram for our candidate LMXBs, obtained from themerged observation (or the merged high-state and low-state observations for transients). We separate the GC(left panel) and field (right panel) sources. For the for-mer, we further differentiate different subgroups, i.e.,the
HST /ACS blue/metal-poor GCs in blue squares, the
HST /ACS red/metal-rich GCs in red diamonds, and theSubaru/Suprime-Cam GCs and the extra five
HST /ACSGC candidates identified by us (Section 2.4) in green cir-cles. We observe no clear spectral differences betweenthese different groups of GC LMXBs, agreeing with pre-vious findings (e.g., Kim et al. 2006). For both GC andfield populations, we find that other than a few very softor very hard outliers, our sources seem to follow a globaltrend that the luminous sources ( ≥ × erg s − ) havehard spectra consistent with Γ PL ∼ .
5, while the faintersources have systematically softer spectra. In the follow-ing sections we will present more detailed source spectralproperties based on simple spectral fits and provide sys-tematic comparison with Galactic LMXBs, which willallow us to shed more light on the nature of our sourcesand the cause of their spectral evolution.
Spectral fits of field LMXBs
The results of simple PL and MCD fits to the mergedspectra (for transients, see Figure 3 for the spectra used)of candidate LMXBs are given in Table 6. The C statisticthat we adopted in the fits does not indicate the fittingquality, but based on bright spectra that were rebinnedto have a minimum of 20 counts per bin and fitted withthe χ statistic, we found that the fits are mostly accept-able with the reduced χ < & χ of 1.4–1.7 andthe null hypothesis probability of 10 − –10 − . Becausewe are mostly interested in using the PL and MCD fitsto roughly characterize the spectral shapes, the fittingquality overall is sufficient for our purposes.The PL fits are shown in Figure 5, with the 0.3–8 keV L X versus Γ PL in the top panels and N H versus Γ PL in thebottom panels, while the MCD fits are shown in Figure 6,with L X versus kT MCD in the top panels and N H versus kT MCD in the bottom panels. The GC and field LMXBsare plotted separately, with GCs in the left panels andfield LMXBs in the middle panels. In Figure 6, somedotted reference lines are included to show the expecteddependence of the 0.3–8 keV L X luminosity on kT MCD from the standard thermal disk truncated at the ISCOof compact objects of several masses ( M = 1.4 M ⊙ , 3 M ⊙ , 5 M ⊙ , 10 M ⊙ , and 20 M ⊙ ). We used the empiricalrelation between the real inner disk radius r in and thenormalization of the MCD model N MCD (Kubota et al.1998; Makishima et al. 2000): r in = 1 . s N MCD d cos θ (2)where d is the source distance in units of 10 kpc and θ is the inclination angle. The relation takes into accountthe spectral hardening effect, with the hardening factor1 Fig. 4.— : The 0.5–7 keV luminosity versus the hardness ratio for candidate LMXBs above 4 σ (excluding those in thecentral a = 10 ′′ ellipse). The hardness ratio is defined as HR = ( H − S ) / ( H + S ), where S and H are the energyfluxes in the 0.5–2 keV and 2–7 keV energy bands, respectively. The luminosities and hardness ratios were calculatedfrom the merged observation, except for the 13 transients, for which the merged high-state and low-state observations(Figure 3 a–d) were used. The left panel is for GC LMXBs, with the blue squares and red diamonds for the HST /ACSblue/metal-poor and red/metal-rich GCs from Jennings et al. (2014), respectively, and the green circles for other GCs(the Subaru/Suprime-Cam GCs and the extra five
HST /ACS GC candidates identified by us; Section 2.4). Twotransients, both in red GCs, are plotted with red triangles. The filled square marks the BHC S96. The right panel isfor field LMXBs. The blue triangles denote the transient candidates. The filled circles and triangles denote the BHCs.The error bars in both panels are at the 1 σ confidence level. The vertical dotted lines from right to left in both panelscorrespond to hardness ratios for an unabsorbed PL with Γ =1.5, 2.0, and 2.5, respectively.assumed to be 1.7, and the fact that the disk temperaturedoes not peak at the inner radius. The dotted referencelines in Figure 6 assume θ = 60 ◦ .Here we focus on field LMXBs first, with GC LMXBsto be presented in the next section. The most strikingresult of the PL fits to field LMXBs is the strong depen-dence of the photon index on the luminosity, as shown inthe top middle panel in Figure 5. Most sources fall on anarrow track (the light gray region), with Γ PL decreasing(thus the sources becoming harder) with increasing L X up to L X ∼ × erg s − and then remaining at avalue around 1.5 above this luminosity. Such a trend hasbeen indicated, though with larger scatters, in the lumi-nosity versus hardness diagram in Figure 4. A few verysoft outliers lying on the right of the light gray track canalso be seen. Five of them are persistent sources (filledcircles), and the other four are transients (filled triangles;S198 has two data points, corresponding to its high andlow states). At L X . × erg s − , some hard sourceswith Γ PL . . PL . . L X . × erg s − is due to selection bias because we only fitted sourcesabove 4 σ and harder sources tend to have lower signifi-cance levels at a given luminosity. The column densitywas inferred to be . . × cm − for most sources.It seems to increase with the photon index, which couldbe caused by the use of a (non-physical) PL model tofit the spectra that are mainly thermal and are softer atlower luminosity (more discussion on the column densityis given at the end of the section).The MCD fits give results consistent with the PL fits. Most sources reside in the light gray region below the M = 3 M ⊙ dotted reference line (thus corresponding tolower mass) in the top middle panel plotting L X verus kT MCD in Figure 6, with kT MCD increasing with L X be-low ∼ × erg s − and then remaining at around 1.5keV above this luminosity. The very soft outliers identi-fied from the PL fits above fall in the region correspond-ing to M & M ⊙ . The column density of most sourcesfrom the MCD fits is at the minimum value allowed inthe fits, i.e., the Galactic value.To shed light on the possible nature of the sources, inthe right panels in both Figures 5 and 6, we plot the ex-pected single-component (PL and MCD) fitting resultsof three representative Galactic X-ray binaries, the atollsource 4U 1705-44 (filled red diamonds), the Z sourceGX 17+2 (green crosses) and the BHC XTE J1817-330 (filled blue squares), based on the spectral fits ofthese sources by Lin et al. (2010), Lin et al. (2012a), andRykoff et al. (2007), respectively, and assuming them tobe at the distance of NGC 3115 with absorption at theGalactic value in the direction of NGC 3115 (i.e., as-suming zero intrinsic absorption). One main reason forus to choose these sources to compare with our sampleis that they had high-quality broad-band spectra span-ning large luminosity ranges and were carefully mod-eled in the above studies. The spectra of 4U 1705-44studied by Lin et al. (2010) used broad-band observa-tions by Suzaku (1.2–40 keV) and
BeppoSAX (1–150keV) and included two hard-state spectra and seven soft-state spectra over a large dynamical range ( ∼ L Edd ). For GX 17+2, Lin et al. (2012a) compiled spec-2 Lin et al.
Fig. 5.— : The PL fit results of LMXBs excluding those in the central a = 10 ′′ ellipse, with the top panels plottingthe 0.3–8 keV luminosity versus the photon index (symbol size proportional to column density logarithm) and thebottom panels plotting the column density (including the Galactic absorption) versus the photon index (symbol sizeproportional to luminosity logarithm, using data with L X > × erg s − for clarity). The left and middle panelsare for GC and field LMXBs, respectively (the meanings of the symbols are the same as in Figure 4). The right panelsplot the PL fits to the atoll source 4U 1705 −
44 (filled diamonds), the Z source GX 17+2 (crosses) and the BHCXTE J1817 −
330 (filled squares), based on the spectral fits of these sources by Lin et al. (2010), Lin et al. (2012a), andRykoff et al. (2007), respectively, and assuming them to be at the distance of NGC 3115 with only Galactic absorption( N H = 4 . × cm − ). In these panels, we also plot the data shown in the middle panels for field LMXBs, but ina gray color. The error bars (sometimes smaller than the symbol size) in all panels correspond to the 90% confidencelevel. The light gray region in the top panels marks the possible NS LMXB soft state track, where atolls in the softstate and Z sources reside.tra over the whole Z track in the HID and used RXTE data ( ∼ Swift /XRT in the softX-ray energy band of 0.5–10 keV, close to the
Chan-dra bandpass, and spanned a large dynamical range,while the source is also one of the Galactic BHCs withthe lowest absorption ( N H = 6 × cm − ). Severalspectral models were applied to these three sources in the above three studies, but we used the fits of themodels that carefully addressed Comptonization, i.e.,the SIMPL(MCD)+BB model, the MCD+BB+nthcompmodel and the MCD+comptt model in Lin et al. (2010),Lin et al. (2012a), and Rykoff et al. (2007), respectively.This is important because Comptonization with a steepslope can significantly overestimate the low-energy emis-sion if it is fitted with models like a simple PL.To convert the source flux to that at the distanceof NGC 3115, we have adopted the source distance of d = 7 . d = 8 . d = 103 Fig. 6.— : Similar to Figure 5 but for the MCD fits. The dotted lines in the upper panels show the expected dependenceof the 0.3–8 keV luminosity L X on kT MCD from the standard thermal disk truncated at the ISCO of compact objectsof several masses ( M = 1.4 M ⊙ , 3 M ⊙ , 5 M ⊙ , 10 M ⊙ , and 20 M ⊙ corresponding to lines from the right to the left,respectively; see text for details. The light gray region in the top panels marks the possible NS LMXB soft state track,where atolls in the soft state and Z sources reside.kpc for XTE J1817-330 (as assumed in Rykoff et al.2007). The distance of 4U 1705-44 was derived fromtype I X-ray bursts and has a relatively small uncertainty( ∼ ∼ M ⊙ , based onEquation 2 (Rykoff et al. 2007). As long as the centralBH is not too small and the Eddington ratio of the ther-mal state in this source is not too low ( & erg s − to 7 × erg s − . Such trends are very similarto those traced out by a majority of our field LMXBsin the light gray region in Figures 5 and 6 in a similarluminosity range. GX 17+2 shows similarly hard spectrato the brightest observation of 4U 1705-44 in 0.3–8 keV,with Γ PL ∼ . kT MCD ∼ L X ∼ × erg s − being atolls in the soft state and brighter sources be-ing Z sources. Then our observation that most of oursources appear harder at higher luminosity in 0.3–8 keVcan be easily explained because for the atoll soft state,the temperatures of the disk and boundary layer ther-mal emission increase with luminosity at relatively con-stant emission areas (Lin et al. 2007, 2009, 2010). Suchtrends are not seen in Z sources because the inner diskand the boundary layer reach the local Eddington limit,in which case the change in the accretion rate tends tolead to change in the emission area (the inner radius forthe disk), instead of the temperature (Lin et al. 2009).We note that our sources seem systematically slightlysofter than the simulated spectra of Galactic NS LMXBs,and one explanation for this is the possible presence ofsoft excess of the real disk spectrum compared with thesimple MCD description (i.e., diskbb in XSPEC) usedin Lin et al. (2010) and Lin et al. (2012a), which will bediscussed in Section 4.2.Although the soft-state observations of 4U 1705-44 inLin et al. (2010) are above L X ∼ erg s − , atollsin the soft state can be fainter. For example, the softstate of 4U 1608-52 has Eddington ratios as low as 0.01(Lin et al. 2007), four times lower than 4U 1705-44 inLin et al. (2010). Therefore, the atoll soft-state explana-tion could apply to our relatively soft (Γ PL & .
0) faintsources with L X as low as a few 10 erg s − .As mentioned above, we also have nine very soft out-liers (filled circles and triangles in the middle panels inFigures 5 and 6). They are much softer than NS LMXBsat corresponding luminosities, but are similar to the BHCXTE J1817-330 in the thermal state. Therefore, we iden-tified them as BHCs. Their long-term luminosity curvesand sample spectra are shown in Figure 3 (those fittedwith a MCD or a MCD+PL model), and the PL andMCD fit results are given in Table 6. The spectra ofmost of these BHCs can be described with a MCD, ex-cept S100. This source seems to show a hard tail, andwe tried to fit it with a MCD plus a PL, with the pho-ton index fixed at a value of 2.5. The fitting result isincluded in Table 7, indicating the presence of a PL atthe 6 σ confidence level. S198 is an interesting BHC thatwill be described separately in Section 3.4.1. It is theonly source with two spectra (one in the high state andthe other in the low state, Section 3.4.1, Figure 3c) inFigures 5 and 6.The faint hard sources with L X . × erg s − andΓ PL . . × erg s − in Figures 5 and 6).We cannot rule out the possibility that there may besome BHBs in the hard state. For example, simulatingthe hard-state spectrum of the BHB GX 339-4 studiedby Miller et al. (2006) and fitting with a PL, as we didabove for other three Galactic X-ray binaries, gave L X ∼ × erg s − and Γ PL ∼ .
6. However, consideringthat NS LMXBs are expected to be more common thanBHBs, the possibility being BHBs should be low.Our PL fits to simulated spectra of three Galactic X-ray binaries in the soft/thermal state systematically gave column densities larger than the value assumed in thesimulation (i.e., the Galactic value toward NGC 3115; seethe bottom right panel in Figure 5). Therefore, the col-umn densities from the PL fits to our sources are system-atically overestimated if our identification of our sources(i.e., most sources are NS LMXBs in the soft state withthe very soft outliers being BHBs in the thermal state) iscorrect. On the contrary, the MCD fits tend to underes-timate the column density, due to the presence of a hardcomponent (weak Comptonization in the case of BHBs inthe thermal state or the boundary layer emission in thecase of NS LMXBs in the soft state) and/or possible ex-tra soft emission of the real disk spectrum relative to theMCD description, which will be discussed in Section 4.There is one main caveat of the above comparison ofour sources with Galactic sources to be kept in mind.To have enough statistics, we have to use the spectraof our sources accumulated over long exposures ( ∼ L X , max & × erg s − , most sources have the long-term variability V var .
3, which is not large enough toaffect our comparison with Galactic sources significantly.However, for fainter sources, their long-term variabilityis uncertain, and there might be relatively large system-atic errors produced by the dependence of the spectralproperties on the luminosity obtained.
Spectral fits of GC LMXBs
On the whole, the simple PL and MCD fits indicatethat GC LMXBs have spectral properties very similarto those of field LMXBs (compare the left panels for GCLMXBs and the right panels for field LMXBs in Figures 5and 6). In particular, most GC sources are also in theNS LMXB soft state track (the light gray region), whichexhibits a trend of decreasing Γ PL or increasing kT MCD with increase in L X below L X ∼ × erg s − andrelatively constant Γ PL ( ∼ .
5) or kT MCD ( ∼ . L X ∼ × erg s − in this track could be atollsin the soft state and Z sources, respectively. Some hardsources with Γ < . L X ∼ × erg s − are alsopresent, and they could be NS or BH LMXBs in the hardstate. It seems that GC LMXBs have more hard sourcesat luminosities around 10 erg s − than field LMXBs.For example, over the range L X = (1–2) × erg s − ,there are six out of 10 GC LMXBs with Γ < .
6, butthere is only one out of 14 field LMXBs with Γ < . L X ∼ × erg s − ).Its long-term luminosity curve and merged spectrum areshown in Figure 3g. We identified it as a BHC. Thereis another source S223 much harder than other sourcesat similar luminosity ( L X ∼ erg s − ), which is acandidate NS LMXB with a high magnetic field or ahigh inclination and will be presented separately in Sec-tion 3.4.3.5 CXB sources
The results of PL fits to CXB sources are shown inFigure 7. Sources inside and outside 2 D are plotted inthe left and right panels, respectively. For both groups ofsources, we see no clear dependence of the photon indexon the flux and column density. Using sources outside2 D and excluding extreme sources with Γ PL > . PL < . L X > × − erg s − cm − , we obtained a median of Γ PL = 1 .
77 with68.3% within 0.38 around it, consistent with the resultsfrom Lin et al. (2012b).In Figure 7, we include three sources (triangles; S317,S362, and S372) that are candidate coronally active starsbecause they all have Γ PL > .
5, coincide with brightstars in the Subaru/Suprime-Cam imaging, and havepossible stellar flares detected, as shown in the bot-tom three panels in Figure 8. Two faint soft sources(Γ PL & .
6; S288 and S290) are plotted with squares.Since they are coincident with bright extended galaxiesin the Subaru/Suprime-Cam imaging, they might be dueto hot gas emission in the galaxies.
Special Sources
S198: A transient BH LMXB candidate with aprominent disk at a very low state
As shown in Section 3.2, S198 is the only source whoseoutburst is covered relatively well by our observations(Figure 3c). It also has the most clear spectral evolu-tion caught among our sources. To accumulate enoughstatistics, we combined all the data in the rising (ob-servations 4–5) and decay (observations 9–11) stages tocreate a low-state spectrum (the spectra in these twostages seem to be consistently soft) and the others (6–8) for a high-state spectrum. When we adopted the PLmodel, we obtained Γ PL = 2 . ± . PL > . kT MCD = 0 . ± . kT MCD = 0 . ± .
05 keV forthe low state. Therefore, the spectrum of S198 in thelow state in the 2012 outburst is much softer than in thehigh state. The 0.3–8 keV L X is about 0.12 and 0.008Eddington luminosity (assuming a BH of mass M BH =6 M ⊙ (see below) and a disk inclination of 60 ◦ ) in the highand low states, respectively.The disk temperature and luminosity in the highstate are typical for BH X-ray binaries in the thermalstate (Figure 6; also refer to Figure 16 in Done et al.2007), in which the thermal disk emission dominates(Remillard & McClintock 2006). Based on Equation 2and the MCD fit in Table 6 and assuming a disk inclina-tion of 60 ◦ , the BH has a mass about 6 M ⊙ .The low state is more difficult to understand, andto determine the corresponding spectral state, we com-pare our source with Swift J1753.5-0127 in Rykoff et al.(2007). We first checked for the possible presence of a PLcomponent in the low state by adding a PL to the MCDmodel in the fit to the spectrum. We fixed Γ PL = 1 . PL = 2 .
5, typical values seen in BHBs, and found thatthe PL component contributes less than 23% and 37%(90% upper limit) of the 0.3–8 keV unabsorbed flux, re-spectively. Thus the MCD still dominates in this energy range in such fits. In comparison, Swift J1753.5-0127 hasa disk thermal flux fraction of about 71%, 62%, and 56%in observations 15, 16, and 17 (those with L X ∼ . × erg s − , 2 . × erg s − , and 3 . × erg s − in Fig-ures 5 and 6) in Rykoff et al. (2007), respectively. Thedisk temperatures are about 0.19 keV, 0.25 keV, and 0.33keV, and the Eddington ratios are about 0.010, 0.019,and 0.026, assuming a BH mass of 10 M ⊙ and a sourcedistance of 10 kpc, in these three observations, respec-tively (Rykoff et al. 2007). The source was probably inthe transitional state in observations 15 and 16 and inthe hard state in observation 17, based on the broad-band data from the Proportional Counter Array onboardthe Rossi X-ray Timing Explorer (Gierli´nski et al. 2008).While Rykoff et al. (2007) concluded that the inner diskradius was consistent with the ISCO in all the three andother brighter
Swift observations, Gierli´nski et al. (2008)demonstrated that the inner disk receded from the ISCOin observations 15–17, after taking into account the effectof irradiation from the hot corona on the disk. The innerdisk radius in the low state of S198 is about three timeslarger than that in the high/thermal state, though onlyat the 2.2 σ level (Table 6). Therefore, based on the disktemperature, the Eddington ratio and the larger innerdisk radius than in the high state, S198 in the low statecould be similar to Swift J1753.5-0127 in observation 17,i.e., in the hard state. Alternatively and probably morelikely, it could be in the transitional state, consideringits more prominent disk than Swift J1753.5-0127 in ob-servation 17. S109: A SSS
SSSs have a characteristic temperature . V var = 3. The spectrum can be fitted witha BB with kT BB = 86 +4 − eV, apparent emission radius R BB = 3 . +5 . − . × km (Table 7). The unabsorbed bolo-metric luminosity is (0 . +2 . − . ) × erg s − . SSSs withluminosity around the Eddington limit for a white dwarf(WD), i.e., ∼ erg s − , could be nuclear burning ofmaterial accreted by a WD (Greiner 2000). Thereforeour SSS is probably one such source. S223: A very hard luminous X-ray source in a GC
As shown in Section 3.3.3, S223 has a spectrum sig-nificantly harder (Γ PL = 0 . ± .
1) than other sources(Γ PL & . V var = 1 . σ signifi-cance level and a mean 0.3–8 keV luminosity of 0 . × erg s − (Figure 3f). We detected short-term variabilityin observation 11, with a Gregory-Loredo variability in-dex of 6 (Figure 8), but not in other observations, withGregory-Loredo variability index of ≤
1. The source iscoincident with a spectroscopically confirmed GC, whose g − z color (1.1) is close to the boundary value of 1.13used to separate the blue and red GCs (Paper I). There-fore it is very unlikely to be a background AGN or aforeground star. It is probably not composed of multi-ple sources in the same GC, considering the detection of6 Lin et al. Fig. 7.— : Similar to Figure 5 but for the PL fits to candidate CXB sources. Sources inside and outside 2 D areplotted in the left and right panels, respectively. The bottom panels for N H versus Γ PL only show sources with L X > × − erg s − cm − . Three candidate coronally active stars (triangles in the right panels) are also included.Two faint soft sources (squares in the right panels) could be from hot gas emission of optically bright galaxies.both long-term and short-term variability.There are other X-ray sources that were found to showvery hard spectra and high luminosities and reside inGCs in nearby galaxies. Burke et al. (2013) reporteda source (their S48) in a GC in Centaurus A havingΓ PL ∼ . ∼ × ergs − . Trudolyubov & Priedhorsky (2004) reported twosources (their S22 and S32) in two GCs (Bo91 and Bo158,respectively) in M31 having Γ PL ∼ . ∼ (0.6–1.8) × erg s − . These twosources showed a spectral cutoff of ∼ PL .
1, cut-off energies around 20 keV, andluminosities & erg s − (White et al. 1983). Con-sidering that such objects normally show coherent pul-sations, we searched for them for our source by creat- ing power density spectra for each observation (refer toLin et al. (2014) for the procedure). We found no detec-tion above the 99% significance level in any observationfrom the timescale of the whole observation length upto the Nyquist frequencies (0.62 Hz for observations 1–3and 0.64 Hz for observations 4–11). It is possible that itspulsation period is shorter than the readout frame timeor that the pulsation is not strong enough to be detectedwith current data. Most normal accreting X-ray pulsarsare high-mass X-ray binaries, with only a few known tobe LMXBs (Bildsten et al. 1997). Given the coincidencewith a GC, S223 is more likely to be a LMXB if it is anormal accreting X-ray pulsar.We note that the hard X-ray source in Bo158 inM31 showed periodic dips occasionally and thus prob-ably has a high inclination (Trudolyubov et al. 2002;Trudolyubov & Priedhorsky 2004). Therefore we specu-late an alternative explanation for the hard spectra of theabove sources: they might have high inclination angles,which suppress the observed disk emission and enhancethe observed boundary layer emission and thus cause the7 Fig. 8.— : The light curves of some sources with in-teresting variability, with Gehrels errors (Gehrels 1986)shown. The top panel is for S223, which is a candi-date NS LMXB with a strong magnetic field or a highinclination in a GC, in observation 11, showing signifi-cant short-term variability. The bottom three panels arefor three coronally active star candidates in observationsshowing sign of flares. The bin size ∆ t of each light curveis annotated in each panel.hard X-ray spectra (but the inclination should not be toohigh for the boundary layer to be strongly obscured bythe disk). Limited by statistics, we cannot determinewhether the short-term variability of our source S223 inobservation 11 is due to dipping. DISCUSSION
The NS LMXB Soft State Track and The NewSource Identification Scheme
Except some very soft outliers, which should be strongBHCs in the thermal state, the majority of our brightsources appear hard in the
Chandra bandpass (0.3–8keV). They could be BHBs in the hard state or NS LMXBs in the soft state. The spectra of NS LMXBsin the soft state can appear hard in this energy banddue to the hot boundary layer emission. Differentiatingbetween the above two scenarios is nontrivial based onthe narrow-band spectra in hand. The method that weadopted is to compare the collective spectral propertiesof our sources with the spectral evolution of three repre-sentative Galactic X-ray binaries based on simple PL andMCD fits. We found that most of our sources fall on anarrow track in the L X versus Γ PL and L X versus kT MCD plots, exhibiting harder spectra at higher luminosity be-low L X ∼ × erg s − but relatively constant spectralshape above this luminosity. Such spectral evolution isclose to that expected for NS LMXBs in the soft state inthe Chandra bandpass. Therefore, we identify the trackas the NS LMXB soft state track, in which sources below L X ∼ × erg s − are most likely atolls in the softstate and sources above this luminosity are Z sources.Our sample of candidate LMXBs in NGC 3115 also in-cludes some hard sources at low luminosities. Althoughwe believe that they should be dominated by NS LMXBs,the possibility of some being BHBs (even AGNs) cannotbe ruled out. Therefore our list of BHBs should be con-servative, only including the ten strong BHCs that wereidentified based on their very soft spectra. Restrictingour search to (0.046–1.0) D and above 10 erg s − , wefound one persistent BHC out of 23 GC LMXBs andnine BHCs (five persistent and four transient) out of59 field LMXBs. Therefore, we found a much largerfraction of BHBs in the field than in GCs, as seen inour Galaxy. We have identified four BHCs out of 11field transients. This is more abundant than the simula-tions by Fragos et al. (2008) ( <
10% for transients above10 erg s − ). Our discovery of five persistent BHCs inthe field with L X & erg is also not predicted byFragos et al. (2008).X-ray sources in the old populations in other galax-ies seem to show similar spectral properties to those ofour sources and can thus be identified in the same way.Trudolyubov & Priedhorsky (2004) fitted the XMM-Newton (using the 0.3–10 keV band) and
Chandra (usingthe 0.5–7 keV band) spectra of 31 bright GC LMXBs inM31 with an absorbed PL. Their sources had the 0.3–10 keV luminosity L X in the range of ∼ erg s − to10 erg s − . In the L X versus Γ PL plot (their Figure6), we can see that most of their sources reside in theNS LMXB soft state track and thus should be atolls inthe soft state and Z sources following our identificationscheme, except some hard sources at low luminosities,which could be NS (most likely) or BH LMXBs in thehard state. Trudolyubov & Priedhorsky (2004) also ar-gued that most of their sources, almost all being per-sistent, should be NS LMXBs, based on the similarityof their spectral properties and long-term variability tothose of luminous persistent X-ray binaries in our Galaxy(they are mostly NS LMXBs). However, in their fits tothe three brightest ( L X > erg s − ) sources withthe MCD+BB model, two have kT MCD . Chandra , Burke et al. (2013) also obtained the L X versus Γ PL plot and the L X versus kT MCD plot (theirFigures 3 and 2, respectively), but for only some bright8 Lin et al.sources (0.5–10 keV luminosity L X in the range of ∼ erg s − to 4 × erg s − ). They tried to determinewhether the source spectra were dominated by a MCD ora PL based on the behavior of the inferred column den-sity. Their method followed Brassington et al. (2010),who came up with this method based on simulations oftypical BHB spectra. We prefer our source identificationscheme, which is based on the spectral behavior of bothNS and BH LMXBs. Almost all the 17 sources in the L X versus Γ PL plot in Burke et al. (2013) that they found tobe PL dominated are in the NS LMXB soft state track.Among the 16 sources in the L X versus kT MCD plot inBurke et al. (2013) that they found to be MCD domi-nated, 12 are in the NS LMXB soft state track, whilethe other 4 have very soft spectra ( kT MCD . Comparison with The Double Thermal ModelSource Identification Method
Given that Lin et al. (2007, 2009, 2010, 2012a) suc-cessfully fitted thousands of NS LMXB soft-state spectrawith the double thermal model (MCD+BB), plus a weakComptonized component when necessary, Barnard et al.(2013) applied the double thermal model to 35 sourcesin M31 showing bright hard spectra and found them tobe BHCs because their best-fitting parameters lie sig-nificantly outside the space occupied by NS LMXBs. Inparticular, they found that the disk temperatures of theirBHCs from the fits with the double thermal model aresystematically lower ( kT MCD . . kT MCD & . >
65% versus < L X > × erg s − . There are33 (14 in GCs and 19 in the field), excluding the softBHCs that we have identified. We found that they allhave kT MCD < kT MCD at the 90% confidence level < L X > × erg s − in NGC 3115 are BHCs, whichseems very unlikely, considering that BHBs are expectedto be much rarer than NS LMXBs in an early-type galaxylike NGC 3115 (Fragos et al. 2008).We note that we are using a narrower and softer energyband and have sources subject to much less absorptionthan Lin et al. (2007, 2009, 2010, 2012a). These differ-ences could cause problems if the double thermal modelis directly applied to our sources. We estimated the pos-sible systematic errors arising from the use of a narrow Fig. 9.— : The residuals of fits to sample LMXBs withthe double thermal model with the temperatures of thethermal components fixed at values typical of Z sources( kT MCD = 1 . kT BB = 2 . kT MCD = 1 . kT MCD = 1 . +0 . − . keV(the error bar corresponds to the 90% confidence levelin the sense that the upper and lower error bars eachinclude 45% of the fits). We also found that 77% and46% of the simulated spectra have the best-fitting BBfraction in 2–10 keV larger than 30% (the BB fractionof suz4) and 50%, respectively. The above simulationresults indicate that the use of the limited energy bandcould systematically infer a lower disk temperature andmore BB contribution in 2–10 keV than expected.We investigated whether the sources could be fittedwell if the temperatures of the double thermal compo-nents are fixed at values typically seen in NS LMXBs.We concentrate on six bright sources (the two bright-est persistent field LMXBs (S25 and S68), the brightesttransient field LMXB (S35), the two brightest persistentLMXBs in GCs (S200 in a blue GC and S212 in a red9GC) and the brightest transient GC LMXB (S8); see Ta-ble 7). Because they all have near or super-Eddingtonluminosity ( L X > erg s − , brighter than the bright-est observation (suz4) of 4U 1705-44 in Lin et al. (2010)),we expect them to be Z sources if they are NS LMXBs.Then they should have kT MCD around 1.7 keV and kT BB around 2.5 keV based on the double thermal model(Lin et al. 2009, 2012a). Therefore we also fitted themwith the temperatures of the thermal components fixedat these values. The fit results are given in Table 7, andthe residuals are shown in Figure 9. The fits overall seemfine, but below around 1 keV, we can see a clear system-atic soft excess in all sources.One possible explanation of such a soft excess is thatthe real disk spectrum is broader than the simple MCDdescription. A detailed disk model should carefully cal-culate radiative transfer through the vertical structureof the disk and account for the relativistic smearing.This is implemented in the relativistic disk model bh-spec , which indicates that the real disk spectrum shouldhave excess emission relative to the MCD descriptionby more than 10% at low energies (Davis et al. 2005;Davis & Hubeny 2006; Davis et al. 2006; Kubota et al.2010). Because NSs tend to have a hotter disk, resultingin more of the low-energy part of the disk spectrum tobe observed in the Chandra bandpass than would be thecase for BHs, the soft excess of the real disk spectrumrelative to the MCD description is expected to be moreobvious for NSs than for BHs. Such soft excess was notseen by Lin et al. (2007, 2009, 2010, 2012a) because theyused spectra above 1 keV and the sources that they stud-ied had much higher absorption than the sources studiedhere. However, bhspec has two limitations, preventing usfrom applying it to our sources. One is that it is a tablemodel for the BH accretion disk with the minimum BHmass M BH of 3 M ⊙ , not suitable for the accretion diskaround NSs. Besides, the model assumes that the innerdisk radius is at the ISCO, which is not necessarily thecase for our luminous sources, whose inner disk couldreach the local Eddington limit and thus be truncatedoutside the ISCO (Lin et al. 2009).Given the above possible soft excess problem, we didnot use the MCD+BB fits to find more BHCs, other thanthose identified in Section 3.3 based on their very softspectra, a characteristic that is not seen in NS LMXBs. CONCLUSIONS
We have studied LMXBs detected in NGC 3115 usingthe Megasecond
Chandra
XVP observations. Includingthree previous observations, the total exposure time is 1.1Ms. Thus NGC 3115 is one of
Chandra ’s best observedgalaxies. In total we have 136 candidate LMXBs in thefield and 49 in GCs detected above 2 σ , with L X in therange of ∼ erg s − to 10 erg s − . We calculatedthe long-term variability for all sources and identified 13transient candidates, whose long-term variability factorsare >
5. Excluding these transients, the sources have long-term variability overall decreasing with increase inluminosity, at least at L X & × erg s − .We carried out simple fits to our sources using single-component models (a simple PL or a MCD). We foundthat in the L X versus Γ PL and L X versus kT MCD plots amajority of our sources fall on a narrow track, showingharder spectra at higher luminosity below L X ∼ × erg s − but relatively constant spectral shape (Γ PL ∼ kT MCD ∼ . Chandra bandpass, we identified the track as the NS LMXB softstate track and suggested sources below L X ∼ × erg s − as atolls in the soft state and sources above thisluminosity as Z sources. However, the spectra of oursources seem to show systematic soft excess relative tothe double thermal (MCD+BB) modeling by Lin et al.(2007, 2009, 2010, 2012a) of Galactic NS LMXBs. Oneexplanation for this is that the real disk spectrum hasexcess soft emission relative to the MCD description.This is expected from detailed simulations involving care-ful calculation of radiative transfer through the verticalstructure of the disk and accounting for the relativisticsmearing (Davis et al. 2005; Davis & Hubeny 2006).Ten sources are significantly softer than others at sim-ilar luminosities and are strong BHCs in the thermalstate. Five of them are persistent (one in a blue GC),and the other five are transient.Some special objects were discovered. S198 is the onlytransient BHC whose outburst was covered relatively wellby our observations. The source displayed clear spectralchange during the outburst. In the peak, it is consistentwith a BHC of M BH ∼ M ⊙ in the thermal state with kT MCD = 0 . ± .
07 keV and L X at 0.12 Eddingtonluminosity. The spectrum during the rise and decay wasmuch softer, with kT MCD = 0 . ± .
05 keV and the diskflux fraction >
63% (the 90% upper limit), and L X wasat ∼ kT BB = 86 +4 − eV and L BB , bol = 0 . +2 . − . × erg s − , and can be explained as due to steady nuclearburning on the surface of a WD. S223 is a persistentluminous source in a GC with very hard (Γ = 0 . ± . Chandra
XVP grantGO2-13104X. This material is based upon work sup-ported in part by the National Science Foundation un-der Grants AST-1211995 and AST-1308124. This mate-rial is based upon work supported in part by HST-GO-12759.02-A and HST-GO-12759.12-A. GRS acknowl-edges support from an NSERC Discovery Grant.
REFERENCESArnaud, K. A. 1996, in Astronomical Society of the PacificConference Series, Vol. 101, Astronomical Data AnalysisSoftware and Systems V, ed. G. H. Jacoby & J. Barnes, 17–+Arnold, J. A., Romanowsky, A. J., Brodie, J. P., et al. 2011, ApJ,736, L26 Barnard, R., Garcia, M. R., & Murray, S. S. 2013, ApJ, 770, 148Barnard, R., Garcia, M. R., Primini, F., & Murray, S. S. 2014,ApJ, 791, 33
Bautz, M. W., Pivovaroff, M., Baganoff, F., et al. 1998, in Societyof Photo-Optical Instrumentation Engineers (SPIE) ConferenceSeries, Vol. 3444, Society of Photo-Optical InstrumentationEngineers (SPIE) Conference Series, ed. R. B. Hoover &A. B. Walker, 210–224Bildsten, L., Chakrabarty, D., Chiu, J., et al. 1997, ApJS, 113,367Bradt, H. V., Rothschild, R. E., & Swank, J. H. 1993, A&AS, 97,355Brassington, N. J., Fabbiano, G., Blake, S., et al. 2010, ApJ, 725,1805Brassington, N. J., Fabbiano, G., Zezas, A., et al. 2012, ApJ, 755,162Burke, M. J., Raychaudhury, S., Kraft, R. P., et al. 2013, ApJ,766, 88Chen, W., Shrader, C. R., & Livio, M. 1997, ApJ, 491, 312Davis, S. W., Blaes, O. M., Hubeny, I., & Turner, N. J. 2005,ApJ, 621, 372Davis, S. W., Done, C., & Blaes, O. M. 2006, ApJ, 647, 525Davis, S. W. & Hubeny, I. 2006, ApJS, 164, 530de Vaucouleurs, G., de Vaucouleurs, A., Corwin, Jr., H. G., et al.1991, Third Reference Catalogue of Bright Galaxies, ed. deVaucouleurs, G., de Vaucouleurs, A., Corwin, H. G., Jr., Buta,R. J., Paturel, G., & Fouque, P.Di Stefano, R., Kong, A., & Primini, F. A. 2010, New A Rev., 54,72Done, C. & Gierli´nski, M. 2003, MNRAS, 342, 1041Done, C., Gierli´nski, M., & Kubota, A. 2007, A&A Rev., 15, 1Evans, I. N., Primini, F. A., Glotfelty, K. J., et al. 2010, ApJS,189, 37Fragos, T., Kalogera, V., Belczynski, K., et al. 2008, ApJ, 683,346Freeman, P. E., Kashyap, V., Rosner, R., & Lamb, D. Q. 2002,ApJS, 138, 185Fridriksson, J. K., Homan, J., & Remillard, R. A. 2015,arXiv:1504.00022Galloway, D. K., Morgan, E. H., Krauss, M. I., Kaaret, P., &Chakrabarty, D. 2007, ApJ, 654, L73Galloway, D. K., Muno, M. P., Hartman, J. M., Psaltis, D., &Chakrabarty, D. 2008, ApJS, 179, 360Gehrels, N. 1986, ApJ, 303, 336Georgakakis, A., Nandra, K., Laird, E. S., Aird, J., & Trichas, M.2008, MNRAS, 388, 1205Gierli´nski, M., Done, C., & Page, K. 2008, MNRAS, 388, 753Gregory, P. C. & Loredo, T. J. 1992, ApJ, 398, 146Greiner, J. 2000, New Astronomy, 5, 137Haberl, F. & Titarchuk, L. 1995, A&A, 299, 414Hasinger, G. & van der Klis, M. 1989, A&A, 225, 79Homan, J. & Belloni, T. 2005, Ap&SS, 300, 107Homan, J., Kaplan, D. L., van den Berg, M., & Young, A. J.2009, ApJ, 692, 73Homan, J., van der Klis, M., Fridriksson, J. K., et al. 2010, ApJ,719, 201Inogamov, N. A. & Sunyaev, R. A. 1999, Astronomy Letters, 25,269Irwin, J. A. 2006, MNRAS, 371, 1903Irwin, J. A., Athey, A. E., & Bregman, J. N. 2003, ApJ, 587, 356 Jennings, Z. G., Strader, J., Romanowsky, A. J., et al. 2014, AJ,148, 32Kalberla, P. M. W., Burton, W. B., Hartmann, D., et al. 2005,A&A, 440, 775Kim, E., Kim, D.-W., Fabbiano, G., et al. 2006, ApJ, 647, 276Kim, M., Kim, D.-W., Wilkes, B. J., et al. 2007, ApJS, 169, 401Kubota, A., Done, C., Davis, S. W., et al. 2010, ApJ, 714, 860Kubota, A., Tanaka, Y., Makishima, K., et al. 1998, PASJ, 50,667Li, J., Kastner, J. H., Prigozhin, G. Y., et al. 2004, ApJ, 610, 1204Lin, D., Irwin, J. A., Wong, K.-W., et al. 2015, arXivLin, D., Remillard, R. A., & Homan, J. 2007, ApJ, 667, 1073—. 2009, ApJ, 696, 1257—. 2010, ApJ, 719, 1350Lin, D., Remillard, R. A., Homan, J., & Barret, D. 2012a, ApJ,756, 34Lin, D., Webb, N. A., & Barret, D. 2012b, ApJ, 756, 27—. 2014, ApJ, 780, 39Maccarone, T. J. 2003, A&A, 409, 697Maccarone, T. J. & Coppi, P. S. 2003, MNRAS, 338, 189Maccarone, T. J., Kundu, A., & Zepf, S. E. 2003, ApJ, 586, 814Maccarone, T. J., Long, K. S., Knigge, C., Dieball, A., & Zurek,D. R. 2010, MNRAS, 406, 2087Makishima, K., Kubota, A., Mizuno, T., et al. 2000, ApJ, 535,632McClintock, J. E. & Remillard, R. A. 2006, Compact StellarX-ray Sources, ed. W. Lewin and M. van der Klis (Cambridge:Cambridge Univ. Press), 157–213Miller, J. M., Homan, J., Steeghs, D., et al. 2006, ApJ, 653, 525Miyamoto, S., Kitamoto, S., Hayashida, K., & Egoshi, W. 1995,ApJ, 442, L13Monet, D. G., Levine, S. E., Canzian, B., et al. 2003, AJ, 125, 984Park, T., Kashyap, V. L., Siemiginowska, A., et al. 2006, ApJ,652, 610Popham, R. & Sunyaev, R. 2001, ApJ, 547, 355Pota, V., Forbes, D. A., Romanowsky, A. J., et al. 2013,MNRAS, 428, 389Remillard, R. A. & McClintock, J. E. 2006, ARA&A, 44, 49Rykoff, E. S., Miller, J. M., Steeghs, D., & Torres, M. A. P. 2007,ApJ, 666, 1129S´anchez-Bl´azquez, P., Gorgas, J., Cardiel, N., & Gonz´alez, J. J.2006, A&A, 457, 809Shakura, N. I. & Sunyaev, R. A. 1973, A&A, 24, 337Soleri, P., Fender, R., Tudose, V., et al. 2010, MNRAS, 406, 1471Tonry, J. L., Dressler, A., Blakeslee, J. P., et al. 2001, ApJ, 546,681Trudolyubov, S. & Priedhorsky, W. 2004, ApJ, 616, 821Trudolyubov, S. P., Borozdin, K. N., Priedhorsky, W. C., et al.2002, ApJ, 581, L27van der Klis, M. 2006, in Compact Stellar X-ray Sources (eds. W.Lewin & M. van der Klis, Cambridge University Press), 39–112van Paradijs, J. & McClintock, J. E. 1995, X-ray Binaries, 58Weisskopf, M. C., Brinkman, B., Canizares, C., et al. 2002,PASP, 114, 1White, N. E., Swank, J. H., & Holt, S. S. 1983, ApJ, 270, 711Wong, K.-W., Irwin, J. A., Shcherbakov, R. V., et al. 2014, ApJ,780, 9Wong, K.-W., Irwin, J. A., Yukita, M., et al. 2011, ApJ, 736, L23 TABLE 2The GC LMXBs and Candidates
Source GC δ XO g ACS z ACS g SCam r SCam i SCam R h Vel(arcsec) (mag) (mag) (mag) (mag) (mag) (pc) (km s − )(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)4 A4 0.06 20 . ± .
002 18 . ± .
002 20 . ± .
003 19 . ± .
004 18 . ± .
003 2 .
49 11238 A238 0.03 23 . ± .
021 21 . ± .
013 23 . ± .
049 22 . ± .
045 22 . ± .
046 0 . · · ·
10 A41 0.19 21 . ± .
007 20 . ± .
010 21 . ± .
107 20 . ± .
122 20 . ± .
116 1 .
30 45611 A60 0.03 21 . ± .
013 20 . ± .
015 21 . ± .
165 20 . ± .
175 20 . ± .
183 1 . · · ·
22 A171 0.09 22 . ± .
018 21 . ± .
015 23 . ± .
032 22 . ± .
033 21 . ± .
031 0 . · · ·
23 A15 0.03 20 . ± .
003 19 . ± .
003 20 . ± .
005 19 . ± .
005 19 . ± .
005 1 .
91 69624 A62 0.02 22 . ± .
041 20 . ± . · · · · · · · · · · · · · · ·
70 A54 0.02 21 . ± .
005 20 . ± .
005 21 . ± .
008 20 . ± .
009 20 . ± .
010 2 .
23 40776 A46 0.02 21 . ± .
029 20 . ± .
028 20 . ± .
206 19 . ± .
196 19 . ± . · · · · · ·
96 A11 0.06 20 . ± .
004 19 . ± . · · · · · · · · · . · · ·
101 A8 0.05 20 . ± .
002 19 . ± .
002 20 . ± .
004 19 . ± .
004 19 . ± .
004 2 . · · ·
103 A35 0.03 21 . ± .
006 20 . ± .
005 21 . ± .
012 20 . ± .
010 20 . ± .
011 2 .
42 821106 A7 0.04 20 . ± .
002 18 . ± .
002 20 . ± .
004 19 . ± .
004 19 . ± .
004 1 .
76 697113 A16 0.12 20 . ± .
004 19 . ± . · · · · · · · · · . · · ·
114 A45 0.10 21 . ± .
006 20 . ± .
005 21 . ± .
008 20 . ± .
009 20 . ± .
009 2 . · · ·
121 A10 0.07 20 . ± .
003 19 . ± .
004 20 . ± .
028 19 . ± .
030 19 . ± .
023 2 . · · ·
129 A2 0.13 19 . ± .
002 18 . ± .
002 19 . ± .
003 19 . ± .
004 18 . ± .
003 1 . · · ·
135 A29 0.21 20 . ± .
005 20 . ± .
006 21 . ± .
008 20 . ± .
009 20 . ± .
009 1 . · · ·
145 A57 0.10 21 . ± .
006 20 . ± .
004 22 . ± .
010 21 . ± .
010 20 . ± .
009 1 .
08 798150 A33 0.13 21 . ± .
004 20 . ± .
004 21 . ± .
007 20 . ± .
008 20 . ± .
008 2 .
30 949153 A5 0.06 20 . ± .
002 18 . ± .
002 20 . ± .
004 19 . ± .
004 19 . ± .
003 1 .
74 682171 A14 0.03 20 . ± .
003 19 . ± .
002 20 . ± .
004 19 . ± .
005 19 . ± .
004 1 . · · ·
183 A63 0.06 21 . ± .
005 20 . ± .
005 21 . ± .
008 20 . ± .
009 20 . ± .
009 1 .
84 441187 A99 0.11 22 . ± .
010 20 . ± .
006 22 . ± .
013 21 . ± .
014 21 . ± .
013 2 .
12 609188 A13 0.03 20 . ± .
003 19 . ± .
003 20 . ± .
004 19 . ± .
005 19 . ± .
004 1 .
24 426192 A262 0.03 23 . ± .
022 22 . ± .
019 23 . ± .
034 22 . ± .
040 22 . ± .
041 1 . · · ·
200 A17 0.07 20 . ± .
003 19 . ± .
003 20 . ± .
005 20 . ± .
005 19 . ± .
005 2 .
54 439212 A102 0.01 21 . ± .
006 21 . ± .
006 21 . ± .
008 21 . ± .
010 21 . ± .
010 2 . · · ·
213 A232 0.15 22 . ± .
011 21 . ± .
011 22 . ± .
018 22 . ± .
022 22 . ± .
023 1 .
44 885214 A167 0.12 22 . ± .
011 21 . ± .
008 22 . ± .
016 22 . ± .
018 21 . ± .
017 1 .
10 750219 A339 0.09 24 . ± .
036 23 . ± .
035 24 . ± .
059 23 . ± .
058 23 . ± . · · · · · ·
223 A32 0.08 21 . ± .
004 20 . ± .
004 21 . ± .
006 20 . ± .
007 20 . ± .
007 1 .
44 706226 A24 0.09 20 . ± .
003 19 . ± .
003 20 . ± .
005 20 . ± .
006 20 . ± .
006 2 .
37 809229 A297 0.04 23 . ± .
026 22 . ± .
023 23 . ± .
035 23 . ± .
052 22 . ± . · · · · · ·
285 A249 0.37 22 . ± .
013 22 . ± .
016 22 . ± .
019 22 . ± .
024 22 . ± .
025 2 . · · ·
299 A18 0.15 20 . ± .
004 19 . ± .
003 20 . ± .
005 20 . ± .
005 19 . ± .
005 2 .
37 881358 A192 0.97 22 . ± .
008 21 . ± .
008 22 . ± .
012 22 . ± .
016 21 . ± .
015 4 .
78 618199 S624 0.12 · · · · · · . ± .
010 21 . ± .
010 21 . ± . · · · · · ·
322 S570 0.05 · · · · · · . ± .
031 23 . ± .
032 22 . ± . · · · · · ·
325 S454 0.64 · · · · · · . ± .
015 22 . ± .
018 21 . ± . · · · · · ·
332 S578 0.53 · · · · · · . ± .
014 21 . ± .
015 21 . ± . · · · · · ·
356 S364 0.64 · · · · · · . ± .
015 22 . ± .
016 21 . ± . · · · · · · · · · . ± .
010 21 . ± .
012 21 . ± . · · · · · ·
451 S638 0.87 · · · · · · . ± .
003 18 . ± .
003 18 . ± . · · · · · · · · · . ± .
045 21 . ± . · · · · · · · · · . · · · · · · . ± .
062 20 . ± . · · · · · · · · · . · · · · · · . ± .
033 19 . ± . · · · · · · · · · . · · · · · · . ± .
006 20 . ± .
005 21 . ± .
011 20 . ± .
011 20 . ± .
010 1 .
66 238
Note . — Columns: (1) master source unique index; (2) GC in Jennings et al. (2014); (3) the offset between the GC center and our X-ray source;(4)
HST /ACS g -band magnitude; (5) HST /ACS z -band magnitude; (6) SCam g -band magnitude; (7) SCam r -band magnitude; (8) SCam i -bandmagnitude; (9) half-light radius; (10) heliocentric velocity, if available, from the Pota et al. (2013) catalog. The top group includes 37 LMXBscoincident with HST /ACS GCs. The middle group includes 7 LMXBs coincident with GCs detected/covered only in the Subaru/Suprime-Camimages, but not in the
HST /ACS images. The bottom group includes 4 GC LMXB candidates whose optical matches were not classified as GCsby Jennings et al. (2014) but were assumed to be GCs by us (there is another similar source, S65, which is not included in the table because thephotometry is not available due to being too close to the bright galaxy center; see text for details). Our GC LMXB list is slightly different fromthat in Jennings et al. (2014), mainly because of our exclusion of very faint X-ray sources ( < σ ) and update of calibration in our X-ray analysis. TABLE 3The master source catalog
Source CXOU Name PU α/R L X , max S/N V var G-L max
Type(1) (2) (3) (4) (5) (6) (7) (8) (9)78 J100515.4-074254 0.05 0.168 7.41e+37 20.3 9.2 1 F,BH96 J100516.2-074235 0.05 0.230 4.33e+37 19.5 1.7 2 GC,BH97 J100514.2-074233 0.05 0.279 5.41e+37 19.0 2.6 7 F,BH100 J100517.1-074217 0.07 0.323 2.67e+37 13.3 2.1 2 F,BH104 J100516.5-074207 0.06 0.344 6.40e+37 19.1 4.0 2 F,BH108 J100518.5-074138 0.06 0.518 1.27e+38 19.3 1.3 0 F,BH179 J100510.9-074533 0.18 1.040 4.83e+37 6.2 19.8 1 F,BH,T181 J100510.0-074529 0.15 0.936 1.71e+38 7.9 126.9 3 F,BH,T193 J100508.7-074443 0.28 0.571 3.99e+37 2.9 30.1 1 F,BH,T198 J100506.7-074433 0.08 0.711 1.17e+38 20.2 33.8 1 F,BH,T8 J100517.2-074352 0.05 0.913 1.60e+38 37.2 54.8 2 GC,T25 J100516.7-074317 0.04 0.532 1.88e+38 42.6 1.5 2 F35 J100515.8-074312 0.04 0.349 2.10e+38 37.1 40.4 2 F,T68 J100513.7-074300 0.03 0.078 2.28e+38 46.7 1.5 0 F200 J100506.0-074428 0.06 0.808 5.71e+38 69.2 1.4 0 GC212 J100527.3-074316 0.07 2.234 4.85e+38 57.7 1.4 2 GC
Note . — This table is published in its entirety in the electronic edition of the Astrophysical Journal. A portionis shown here, using 10 BHCs (the top group) and 6 bright candidate NS LMXBs (the bottom group), for guidanceregarding its form and content. Columns: (1) master source unique index; (2) IAU name (following the convention ofCXOU Jhhmmss.s+/-ddmmss); (3) 1- σ statistical positional uncertainty (in units of arcsec) in each coordinate, basedon Equation (12) of Kim et al. (2007); (4) The ratio of the angular offset from the galaxy center to the elliptical radiusof the D isophotal ellipse in the direction from the galaxy center to the source; (5) 0.5–7 keV maximum luminosity(in units of erg s − , assuming a source distance of 9.7 Mpc); (6) the signal to noise ratio (the 0.5–7 keV net countsdivided by the error); (7) long-term variability; (8) the maximum Gregory-Loredo short-term variability index amongdifferent observations; (9) the source type (field LMXB (“F”), GC LMXB (“GC”), transient (“T”), BHC (“BH”),star, AGN, galaxy (“G”), and SSS).3 TABLE 4The Source Counts, Flux, and Hardness Ratio in the Merged Observation
Source Obs Expo C b C lb C ub F b F lb F ub HR HR l3 HR u3 (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)78 0 1131.0 418.4 397.8 439.2 3.36e-15 3.20e-15 3.53e-15 − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − .
18 0 . − . − .
33 0 . − . − .
41 0 . − . − . − . − . − . − . − . − . − .
688 0 1131.0 1370.0 1333.2 1407.2 1.10e-14 1.07e-14 1.13e-14 − . − . − .
028 h 978.9 1370.8 1334.0 1407.9 1.30e-14 1.26e-14 1.33e-14 − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . . − .
02 0 . − . − . − . Note . — This table is published in its entirety in the electronic edition of the Astrophysical Journal. A portionis shown here, using 10 BHCs (the top group) and 6 bright candidate NS LMXBs (the bottom group), for guidanceregarding its form and content. Columns: (1) master source unique index; (2) observation (“0” refers the combinationof all available observations; “h” refers to the combination of the high-state observations for the 13 transients, asshown in Figure 3 a–d; and “l” refers to the combination of the low-state observations for source 198, as shown inFigure 3c); (3) exposure time, in units of ks; (4-6) the net counts in the broad band (0.5–7.0 keV) and the lower andupper limits; (7-9) the energy flux in the broad band and the lower and upper limits, in units of erg s − cm − ; and(10-12) the hardness ratio using the fluxes in the 0.5–2.0 keV and 2.0–7.0 keV energy bands and the lower and upperlimits. All limits are at the 68% confidence level.4 Lin et al. TABLE 5The Source Counts and Flux in Individual Observations
Source ObsID C b C lb C ub F b F lb F ub (1) (2) (3) (4) (5) (6) (7) (8)78 13820 86.9 77.6 96.4 4.37e −
15 3.90e −
15 4.84e − −
15 2.13e −
15 2.85e − −
15 2.44e −
15 3.20e − −
15 1.33e −
15 1.91e − −
15 1.63e −
15 2.26e − −
17 0 1.46e − − − −
14 9.60e −
15 1.12e −
148 13820 237.2 221.9 252.7 1.19e −
14 1.12e −
14 1.27e − −
14 1.39e −
14 1.56e − −
14 1.01e −
14 1.17e − −
14 1.75e −
14 1.94e − −
14 4.08e −
14 4.39e − −
14 3.02e −
14 3.31e − Note . — This table is published in its entirety in the electronic edition of the Astrophysical Journal. A portionis shown here, using 10 BHCs (the top group) and 6 bright candidate NS LMXBs (the bottom group) in observation13820, for guidance regarding its form and content. Columns: (1) master source unique index; (2) observation ID;(3-5) the net counts in the broad band (0.5–7.0 keV) and the lower and upper limits; and (6-8) the energy flux in thebroad band and the lower and upper limits, in units of erg s − cm − . All limits are at the 68% confidence level.5 TABLE 6Spectral fit results the PL Model the MCD ModelSource Obs N H Γ Norm
L N H kT R L (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)78 0 18 . +8 . − . . +0 . − . . +0 . − . . +0 . − . . +2 . . +0 . − . . +8 . − . . +0 . − .
96 0 14 . +7 . − . . +0 . − . . +0 . − . . +0 . − . . +1 . . +0 . − . . +9 . − . . +0 . − .
97 0 16 . +9 . − . . +0 . − . . +0 . − . . +0 . − . . +1 . . +0 . − . . +18 . − . . +0 . − .
100 0 0 . +3 . . +0 . − . . +0 . − . . +0 . − . . +1 . . +0 . − . . +60 . − . . +0 . − .
104 0 16 . +8 . − . . +0 . − . . +0 . − . . +0 . − . . +2 . . +0 . − . . +9 . − . . +0 . − .
108 0 24 . +8 . − . . +0 . − . . +1 . − . . +1 . − . . +3 . . +0 . − . . +24 . − . . +0 . − .
179 h 31 . +26 . − . . +0 . − . . +2 . − . . +1 . − . . +20 . . +0 . − . . +279 . − . . +0 . − .
181 h 10 . +13 . − . . +0 . − . . +2 . − . . +4 . − . . +3 . . +0 . − . . +92 . − . . +2 . − .
193 h 9 . +77 . − . . +2 . − . . +8 . − . . +2 . − . . +31 . . +0 . − . . +201 . − . . +1 . − .
198 h 20 . +9 . − . . +0 . − . . +1 . − . . +1 . − . . +3 . . +0 . − . . +11 . − . . +0 . − .
198 l 20 . . − . . +0 . − . . +0 . − . . . +0 . − . . +481 . − . . +0 . − . . +4 . − . . +0 . − . . +0 . − . . +1 . − . . +0 . . +0 . − . . +2 . − . . +1 . − .
25 0 3 . +3 . − . . +0 . − . . +0 . − . . +1 . − . . +0 . . +0 . − . . +2 . − . . +1 . − .
35 h 3 . +4 . − . . +0 . − . . +0 . − . . +1 . − . . +0 . . +0 . − . . +2 . − . . +1 . − .
68 0 1 . +3 . − . . +0 . − . . +0 . − . . +1 . − . . +0 . . +0 . − . . +2 . − . . +2 . − .
200 0 5 . +2 . − . . +0 . − . . +0 . − . . +3 . − . . +0 . . +0 . − . . +0 . − . . +2 . − .
212 0 9 . +3 . − . . +0 . − . . +0 . − . . +2 . − . . +0 . . +0 . − . . +0 . − . . +2 . − . Note . — This table is published in its entirety in the electronic edition of the Astrophysical Journal. A portionis shown here, using 10 BHCs (the top group) and 6 bright candidate NS LMXBs (the bottom group), for guidanceregarding its form and content. Columns: (1) master source unique index; (2) the observation (“0” refers the combi-nation of all available observations; “h” refers to the combination of the high-state observations for the 13 transients,as shown in Figure 3(a)–(d); and “l” refers to the combination of the low-state observations for source 198, as shownin Figure 3(c)); (3)–(6) the intrinsic column density (in units of 10 cm − , constrained to be ≤ cm − ), pho-ton index (constrained to be < . − photons keV − cm − s − at 1 keV), and 0.3-8keV luminosity (in units of 10 erg s − , assuming a source distance of 9.7 Mpc) corrected for Galactic absorptionfrom the fit with the PL model; (7)–(10) the intrinsic column density, the maximum disk temperature (in units ofkeV, constrained to be ≤ R MCD (in units of km) from the normalization N MCD ≡ (( R MCD / km) / ( D/
10 kpc)) cos θ , where D is the source distance 9.7 Mpc and θ is the disk inclination as-sumed to be 60 ◦ , and 0.3-8 keV luminosity corrected for Galactic absorption from the fit with the MCD model. Allfits used the C statistic. All errors are at the 90% confidence level. For S198, the fit to the low-state spectrum has thecolumn density fixed (marked with “f”) at the value obtained from the fit to the high-state spectrum.6 Lin et al. TABLE 7Spectral fit results of some special sources
Source model N H Other Parameters χ ν/ν L abs L unabs(1020 cm −
2) 1036 erg s − . . kT MCD = 0 . . − .
04 keV, R MCD = 660+1463 −
437 km, ΓPL = 2 . f ), N PL = (0 . . − . × − · · · − − . . kT BB = 0 . . − .
012 keV, R BB = 3376+5272 −
674 km · · · −
10 41+79 −
88 MCD+BB 0 . . kT MCD = 0 . . − .
20 keV, R MCD = 56+46 −
27 km, kT BB = 1 . . − .
35 keV, R BB = 14 . . − . . −
12 166+13 − . . kT MCD = 0 . . − .
18 keV, R MCD = 100+238 −
51 km, kT BB = 1 . . − .
16 keV, R BB = 26 . . − . . −
14 184+29 − . . kT MCD = 0 . . − .
18 keV, R MCD = 77+98 −
21 km, kT BB = 1 . . − .
22 keV, R BB = 18 . . − . . −
11 137+12 − . . kT MCD = 0 . . − .
09 keV, R MCD = 154+128 −
59 km, kT BB = 1 . . − .
15 keV, R BB = 24 . . − . . −
14 239+22 − . . kT MCD = 0 . . − .
15 keV, R MCD = 106+60 −
54 km, kT BB = 1 . . − .
14 keV, R BB = 41 . . − . . −
21 556+21 − . . kT MCD = 0 . . − .
15 keV, R MCD = 113+100 −
44 km, kT BB = 1 . . − .
13 keV, R BB = 36 . . − . . −
20 416+30 −
168 MCD+BB 0 . . kT MCD = 1 . R MCD = 13 . . − . kT BB = 2 . R BB = 0 . . . − −
725 MCD+BB 0 . . kT MCD = 1 . R MCD = 13 . . − . kT BB = 2 . R BB = 0 . . . − −
735 MCD+BB 0 . . kT MCD = 1 . R MCD = 12 . . − . kT BB = 2 . R BB = 0 . . . − −
668 MCD+BB 0 . . kT MCD = 1 . R MCD = 15 . . − . kT BB = 2 . R BB = 0 . . . − − . . kT MCD = 1 . R MCD = 24 . . − . kT BB = 2 . R BB = 0 . . . −
14 565+13 − . . kT MCD = 1 . R MCD = 21 . . − . kT BB = 2 . R BB = 0 . . . −
12 422+12 − Note . — Columns: (1) master source unique index; (2) the spectral model; (3) intrinsic column density (4) other spectral parameters ( R BB is the apparent source radiusfrom the BB normalization N BB ≡ (( R BB / km) / ( D/ χ χ ◦◦