A Chandra Study of the Dense Globular Cluster Terzan 5
a r X i v :a s t r o -p h /0303141v 1 6 M a r 2003
D RAFT
VERSION
F EBRUARY 5,2008
Preprint typeset using L A T E X style emulateapj v.11/12/01
A Chandra X-RAY STUDY OF THE DENSE GLOBULAR CLUSTER TERZAN 5
C.O.H EINKE ,P.
D.E DMONDS ,J.
E.G RINDLAY ,D.A.L LOYD
Harvard College Observatory,60Garden Street,Cambridge,MA 02138;
cheinke@https://www.sodocs.net/doc/365993790.html,,pedmonds@https://www.sodocs.net/doc/365993790.html,,josh@https://www.sodocs.net/doc/365993790.html,,dlloyd@https://www.sodocs.net/doc/365993790.html,
AND
H.N.C OHN AND P.M.L UGGER
Department of Astronomy,Indiana University,Swain West 319,Bloomington,IN 47405;cohn@https://www.sodocs.net/doc/365993790.html,,
lugger@https://www.sodocs.net/doc/365993790.html,
Draft version February 5,2008
ABSTRACT
We report a Chandra ACIS-I observation of the dense globular cluster Terzan 5.The previously known transient low-mass x-ray binary (LMXB)EXO 1745-248in the cluster entered a rare high state during our August 2000observation,complicating the analysis.Nevertheless nine additional sources clearly associated with the cluster are also detected,ranging from L X (0.5?2.5keV)=5.6×1032down to 8.6×1031ergs s ?1.Their X-ray colors and luminosities,and spectral ?tting,indicate that ?ve of them are probably cataclysmic variables,and four are likely quiescent LMXBs containing neutron stars.We estimate the total number of sources between L X (0.5?2.5keV)=
1032and 1033ergs s ?1as 11.4+4.7
?1.8by the use of arti?cial point source tests,and note that the numbers of X-ray sources are similar to those detected in NGC 6440.The improved X-ray position allowed us to identify a plausible infrared counterpart to EXO 1745-248on our 1998Hubble Space Telescope NICMOS images.This blue star (F110W=18.48,F187W=17.30)lies within 0.2”of the boresighted LMXB position.Simultaneous Rossi X-ray Timing Explorer (RXTE )spectra,combined with the Chandra spectrum,indicate that EXO 1745-248is an ultracompact binary system,and show a strong broad 6.55keV iron line and an 8keV smeared re?ection edge.Subject headings:X-rays:individual (EXO 1745-248)—X-rays :binaries —novae,cataclysmic variables —
globular clusters:individual (Terzan 5)—stars:neutron
1.INTRODUCTION
The high resolution of the Chandra X-ray Observatory has enabled astronomers to study the low-luminosity X-ray source populations in globular clusters in great https://www.sodocs.net/doc/365993790.html,bined X-ray,radio,and optical Hubble Space Telescope (HST )stud-ies of the globular cluster 47Tucanae have revealed quies-cent low-mass X-ray binaries (qLMXBs)that have not expe-rienced X-ray outbursts in the history of X-ray astronomy,cat-aclysmic variables (CVs)as X-ray luminous as any known in the ?eld,?aring behavior from coronally active stellar binary systems,and predominantly thermal X-ray emission from mil-lisecond pulsars (MSPs)(Grindlay et al.2001a,2002).Similar populations have been uncovered in the globular clusters NGC 6397(Grindlay et al.2001b),NGC 6752(Pooley et al.2002a),and ωCen (Rutledge et al.2001,Cool,Haggard &Carlin 2002),while the luminosities and broad X-ray spectral types of these sources have allowed classi?cation of sources in the more obscured cluster NGC 6440(Pooley et al.2002b)and M28(Becker et al.2003).
The globular cluster Terzan 5contains a known transient LMXB,EXO 1745-248,which was ?rst detected in a burst-ing state by Hakucho (Makishima et al.1981),and has been irregularly active since then (Johnston et al.1995and refs.therein).EXO 1745-248is one of the few luminous globular cluster LMXBs not analyzed by Sidoli et al.(2001;hereafter SPO01)or Parmar et al.(2001),who identify spectral distinc-tions between normal and ultracompact LMXBs.Ultracompact LMXBs,de?ned as having periods less than 1hour,are thought to possess a degenerate helium white dwarf secondary.Deutsch et al.(2000)remark upon the overabundance of ultracompact LMXBs in globular clusters,and speculate that the short peri-ods may be due to dynamical effects in globular clusters.
Terzan 5also contains two identi?ed MSPs,with many addi-tional MSPs probably making up the extended steep-spectrum radio source at the cluster core (Lyne et al.2000;Fruchter &Goss 2000).Terzan 5’s high central density and large mass make it a rich target for studies of binary systems,but its high reddening and severe crowding make optical and even infrared observations extremely dif?cult.The deepest infrared survey of Terzan 5was performed with the HST NICMOS camera by Cohn et al.(2002;hereafter CLG02).The extreme reddening indicates that the cluster parameters are best determined in the infrared.NICMOS observations produced the ?rst CMDs of Terzan 5to reach the main-sequence turnoff (CLG02;see also Ortolani et al.2001).Therefore we utilize the new cluster pa-rameters derived by CLG02in our analysis,particularly the red-dening,distance,core radius,and radial star-count pro?les.The attempt by Edmonds et al.(2001,hereafter EGC01)to iden-tify EXO 1745-248and the eclipsing MSP through time-series variability analysis and color information did not uncover any promising candidates due to the crowding and overlapping Airy pro?les,although it did identify two truly variable stars,one of which was shown to be an RR Lyrae variable.
This paper is organized as follows.Section 2.1describes the observations we used.Section 2.2explains our methods for de-tecting sources and performing an astrometric correction based upon identi?cation of serendipitous sources.Section 2.3de-scribes our search for an infrared counterpart for EXO 1745-248,extending the work of EGC01using our Chandra posi-tion.Section 2.4classi?es the faint X-ray sources,and quanti-?es our detection incompleteness (due to the outburst of EXO 1745-248).Section 2.5attempts simple spectral ?ts to the faint sources,while section 2.6examines the simultaneous RXTE
1
2Heinke et al.
and Chandra spectra of EXO1745-248.Section3.1compares our spectral analysis of EXO1745-248to other observations of LMXBs.Section3.2recalculates the central density and col-lision frequency of Terzan5,while section4summarizes our conclusions.
2.ANALYSIS
2.1.Description of Observations
The Chandra X-ray Observatory observed the globular clus-ter Terzan5on July24,2000,for45ksec(05:46to18:22TT), and on July29,2000for5ksec(00:56to09:20TT),with the ACIS-I instrument at the focus.Due to an error in the obser-vation upload,both exposures were performed in1/8subar-ray mode(as intended for the shorter observation only),with frame times of0.841s(longer exposure with more chips)and 0.541s.EXO1745-248entered an outburst during July2000, its2-10keV?ux varying between54and600mCrab during July and August,approaching its Eddington limit at maximum ?ux(Markwardt et al.2000a,2000b).Rossi X-ray Timing Ex-plorer(RXTE)All-Sky Monitor(ASM)observations(results provided by the ASM/RXTE team1)show that the2-10keV countrate from EXO1745-248on July24,2000,was5.3±0.9 cts s?1(～72mCrab)during our longer Chandra observation, rising to9.8±0.9cts s?1in the second observation(Fig.1). Unfortunately,the Chandra observation of EXO1745-248 was not optimized to study such a bright object(Fig.2).The intense photon?ux led to severe pileup,which occurs when two or more photons landing in the same or adjacent pixels be-tween frames are recorded as a single event.Pileup can in-crease the energy of recorded events,or by altering the grade of the recorded event to a“bad”grade,cause the rejection of the event either before or after telemetry to the ground(See the Chandra Proposer’s Observatory Guide,v.5,chapter6).In this case the pileup was severe enough to cause the pre-telemetry re-jection of all events recorded within～1”of the LMXB position. The LMXB’s X-ray halo,by increasing the local background, greatly degraded our sensitivity to faint cluster sources.(The halo is due to a combination of dust grain scattering and the in-trinsic breadth of the Chandra mirrors’point spread function.) Nevertheless,the spectacular resolution of Chandra did allow us to identify additional point sources up to105times fainter within10”of EXO1745-248.The readout streak(caused by photons from EXO1745-248arriving during the readout of the CCD)also degraded our survey,but we were able to extract a useful spectrum of EXO1745-248from the readout streak. We also analyzed a simultaneous Rossi X-ray Timing Explorer (RXTE)pointed observation from the HEASARC archive2(on July24,2000,15:15to16:16TT),for broad spectral coverage of the outburst of EXO1745-248.The HST NICMOS data we used to search for a possible infrared counterpart to EXO1745-248are described in EGC01and CLG02.
2.2.Detection and astrometry
We used the CIAO software package3to search for point sources,produce hardness ratios,and extract spectra and lightcurves.We reprocessed the two observations to remove the pixel randomization added in standard processing,and merged the two observations.No periods of high background?aring were observed.EXO1745-248’s X-ray halo displays an x-ray color(de?ned here as2.5log[0.5-1.5keV counts/1.5-6keV counts],following Grindlay et al.(2001a))of-2.9,harder than most known faint globular cluster sources.Therefore,we se-lected a soft band,0.5-2keV,to search for point sources.We selected a2.5×104pixel(1.7arcmin2)square region including the cluster and ran W A VDETECT with the signi?cance thresh-old set to give false positives at the rate of10?4.W A VDETECT indeed found two spurious sources(identi?ed as spurious by eye,and by not appearing in more than one energy band)far from the cluster,as expected.Within three optical core radii (24”)we identify nine point sources besides EXO1745-248, each con?rmed by visual inspection(see Figure2)and with signi?cance above2.8σ.We name the nine additional real sources in Table1with both positional names and simple ref-erence names W2-W10(used in the rest of this paper).Outside the globular cluster,we binned the remaining data into1arc-second pixels and searched for back-or foreground serendip-itous sources(setting the signi?cance threshold to a false rate of10?6).We identi?ed four sources that also pass visual in-spection on the active portions(36arcminutes2)of the ACIS-I array.As they are all more than2’(2.5half-mass radii)from the cluster center,we hereafter identify them as“serendipitous sources”.The ROSAT source S2from Johnston et al.(1995) was not included in the?eld of view.Inspection of the locations of the known millisecond pulsars(A and C;Lyne et al.2000) show no evidence of X-ray emission(using the astrometry be-low).The upper limits are3and5counts respectively in the 0.5-2keV band,giving L X≤2×1031and≤3×1031ergs s?1 for a power law of photon index2.These are well above the X-ray luminosities of all identi?ed millisecond pulsars in47Tuc (Grindlay et al.2002),NGC6397(Grindlay et al.2001b),and NGC6752(D’Amico et al.2002),but below the luminosity (L X=1.1×1033ergs s?1)of the MSP B1821-24in M28(Dan-ner et al.1997).Soft thermal spectra give even less constraining limits,due to the high extinction.The two variables identi?ed by EGC01are also not seen,as expected based on their RR Lyrae and probable eclipsing blue straggler identi?cations. Inspection of a Digital Sky Survey image reveals that one of the serendipitous sources has a probable optical counter-part.According to the Guide Star Catalog2.2,a star with V=13.56,R=13.22is located only0.′′36from the position de-rived by W A VDETECT for the X-ray source that we name CXOU J174803.3-244854.This is consistent with the0.′′6ab-solute astrometry(90%conf.radius)reported by Aldcroft et al. (2000)for Chandra.We estimate a probability of a positional coincidence within0.′′5of one of the15brightest stars on the 7’by7’survey plate with any of the four serendipitous X-ray sources of8×10?5.Another X-ray source(CXOU J174812.6-244811.1)has a faint star(R=16.8)0.′′14away(when the frame is shifted to match star1).We estimate the chances of a star this bright or brighter landing within0.′′5of one of four Chandra sources as4×10?3.The other two X-ray sources show no stars in the GSC2.2Catalog within3”.We shift our Chandra frame by+0.s004in RA,and+0.′′396in Dec,to match the weighted
1See https://www.sodocs.net/doc/365993790.html,/.
2Available at https://www.sodocs.net/doc/365993790.html,/docs/corp/data.html.
3Available at https://www.sodocs.net/doc/365993790.html,/ciao/.
4GSC2.2absolute astrometric errors,when compared to the international celestial reference frame,are of order0.3-0.75arcseconds;see https://www.sodocs.net/doc/365993790.html,/gsc/gsc2/GSC2home.htm.
Terzan53 GSC2.2star positions.4
The V and R magnitudes of star1suggest an F star,and
the X-ray(0.5-2.5keV)to V-band?ux ratio(V-band?ux de-
?ned as10?0.4V?5.43ergs cm?2s?1)is2×10?4,on the high end
of values for nearby F stars(Hünsch et al.1999).Star2has
no color information in the GSC2.2Catalog.Assuming V-
R～1.5(appropriate for M2stars)gives a?ux ratio of1×10?2,
which is common among M stars(Hünsch et al.1999).The
other two sources may be background AGN,or(perhaps more
likely)CVs in the galactic bulge.Both show an Xcolor(2.5
log[0.5-1.5keV cts/1.5-6keV cts])of-1.5,indicating strong
absorption.Simple spectral?tting of the brighter source gives
N H=2.8+2.4
?1.2×1022cm?2for a powerlaw of photon index1.7,
as typical for AGN.A10keV bremsstrahlung spectrum(as for
a bulge CV)would have N H=2.7+2.5
?1.2×1022cm?2.The2-10
keV log N–log S relation of Giacconi et al.(2001)suggests that1-3AGN may be expected at these observed?ux levels in our ACIS-I subarray?eld(although the location of detected sources near to the aimpoint suggests that our sensitivity is not uniform across the?eld).Recent results from galactic bulge surveys(Grindlay et al.2003)indicate that bulge CVs or com-pact binaries outnumber AGN at these?ux levels,suggesting that these two sources may be bulge CVs.
We found the position of EXO1745-248by centering the symmetric“hole”caused by extreme pileup and the readout streak.We estimate our error in this determination at2/5of a pixel,or0.′′2.We present the positions(in the GSC2.2 frame)of EXO1745-248,the nine additional globular cluster sources,and the four serendipitous sources(with relative po-sitional errors from W A VDETECT)in Table1,along with the background-subtracted counts in three bands.
2.3.Identi?cation of plausible infrared counterpart to EXO
1745-248
Our re?ned position for EXO1745-248allowed us to under-take a careful search of the small Chandra error circle in the June1998Hubble Space Telescope NIC2F110W and F187W data,for which the photometric analysis is described in EGC01. We estimate an error of0.′′1in the shift to the GSC2.2frame, based on the scatter of the two stars used in the boresighting.An additional error of0.′′37(following EGC01)is incurred in the shift between the GSC2.2and HST NICMOS https://www.sodocs.net/doc/365993790.html,-bining these with the0.′′2centering positional uncertainty gives an error for the position in the NICMOS frame of0.′′4.This er-ror circle is fortuitously free of bright red giants.Error circles for six other X-ray sources in the cluster lie wholly or partially within the NIC2?eld of view.
Comparison of the F110W and F187W images with a ratio image(Figure3)shows a signi?cantly blue faint star(hereafter star A)within the Chandra error circle,only0.′′2from the cor-rected position of EXO1745-248.A combination of this star and another star just above it(Star B)was shown in the cleaned CMD of EGC01(their?gure8)as one of only two stars that lie more than3σto the blue from the distribution of stars in that color-magnitude diagram.Our attempts at separate PSF-?tting using ALLSTAR in DAOPHOT encountered two prob-lems:(1)we were unable to?t stars A and B simultaneously and(2)the sky-?tting component in PHOT and ALLSTAR gave local sky(background)values that were too high,because stars A and B are surrounded by a ring of bright stars.The latter is a formidable problem in this crowded?eld,especially in F187W where the background contribution of neighboring red giants is greater than in F110W.
We therefore adopted a customized procedure to estimate the F110W and F187W magnitudes for stars A and B.For sky de-
termination we adopted the mean of the two2-by-2pixel re-
gions having the lowest intensity values within an8pixel radius of star https://www.sodocs.net/doc/365993790.html,ing the F110W and F187W images,and the ratio
image,we determined the positions of stars A and https://www.sodocs.net/doc/365993790.html,ing
PSF models and the DAOPHOT program ADDSTAR to create models for stars A and B we subtracted these models from the
two images to produce residual images,while keeping the po-
sitions of stars A and B?xed.Since no other close-by stars are obviously present,stars A and B are accurately modeled
if the mean aperture counts(after subtracting the sky)at their
positions in the residual image are equal to zero.We iteratively varied the magnitudes of these two stars until this goal was met.
The results of these?ts are shown in the lower right-hand cor-
ner of Figure3.Clearly the subtractions of stars A and B were successful,given the relatively smooth residuals.
We performed aperture photometry on the residual images to
produce photometry that is consistent with Figure8of EGC01. Photometry for stars A and B were measured by subtracting
off stars B and A respectively,and the results are shown in
Figure4.Variables1and2from EGC01are labeled with crosses,and stars within1"of X-ray sources are circled.The
main-sequence turnoff(not reached at the cluster core)is at
F110W≈20(CLG02).Note the apparently blue color of star A(～2.5σfrom the subgiant sequence)lying near the top of
the blue straggler sequence.The error in the color of star A
is dominated by the error in the F187W magnitude.The high sky background in F187W caused the original F187W magni-
tude determination for the combination of stars A and B to be
too low,and thus shifted the combination bluewards.Our cus-tomized procedure thus gives redder colors for both stars A and
B than the original determination for the combination.
Given the blue appearance of star A in both the normal and
residual images,the reasonable evidence for blue colors in our
photometric analysis,and its proximity to the center of the X-ray error circle,star A is a reasonable candidate for the IR coun-
terpart to EXO1745-248.No evidence is seen for statistically
signi?cant variability of star A,but the faintness of star A and its strong blending with star B limits the depth of variability
searches.Only the F110W images are useful for time series
analysis,and only sparse coverage was obtained here over1.16 days.Images in the V and I passbands with the Advanced Cam-
era(in High Resolution mode)on HST should do a much better
job of separating stars A and B and would provide much bet-ter constraints on variability if taken at several different epochs
near outburst and quiescence.
Since the F187W band does not correspond to any ground-based?lter,standard calibrations for it do not exist.We ap-
ply rough transformations to estimate interesting quantities.
We estimate the unreddened J magnitude from the F110W and F187W magnitudes by the equation
J0=F110W+0.033?0.178(F110W?F187W)+A J
derived assuming that the spectral slope is constant over this https://www.sodocs.net/doc/365993790.html,ing A J/A V=0.282and A V=6.75(CLG02),we?nd J0=16.40,M J=1.7.We calculate the color(F110W-F187W)0=
?0.17±0.3.Assuming that the photometry is accurate,three alternatives to an LMXB nature remain for this star:it could be a blue horizontal branch star,it could be a blue straggler,or it could be a foreground star.We plot simple interpolated blue straggler and horizontal branch sequences in Figure4.Star A’s
4Heinke et al.
position may be consistent with that of a blue horizontal branch star,although a blue horizontal branch has only been hinted at in Terzan5(CLG02).It is also consistent with an extremely heavy(more than twice the turnoff mass)blue straggler,or a blue straggler binary(as variable2from EGC01,plotted as an “X”next to star A in?gure4,may be).The other extremely blue star seen in Figure4is probably a foreground star,which also cannot be excluded as a possibility for star A.However, the chance of one of the?ve bluest stars in the19.′′2square NIC2FOV landing in the0.′′4error circle by chance is only 7×10?3.None of the stars in the other X-ray error circles(cir-cled in Figure4)display any unusual colors or variability.This is expected based upon the magnitudes of CVs and quiescent LMXBs,in the?eld and other globular clusters,which very rarely reach the turnoff magnitude(Grindlay et al.2001a,b, Pooley et al.2002a).
We compare this suggested counterpart to other low-mass X-ray binaries observed in the infrared by calculating the ratio of X-ray to infrared https://www.sodocs.net/doc/365993790.html,ing the ASM lightcurve in?g-ure1,we see that the average ASM countrate for the11days leading up to the HST observation was1.5±0.2cts s?1,com-pared to5.3±0.9cts s?1at the time of the Chandra observa-tion.The high source confusion near the galactic center for the ASM makes this data relatively noisy,so we take1.5cts s?1as an upper limit on the X-ray?ux.Assuming the same spectrum as we?nd in2000(see Sect.2.5),we derive an un-absorbed0.5-10keV X-ray?ux of6.1×10?10ergs cm?2s?1 and an unabsorbed J-band?ux of1.2×10?13ergs cm?2s?1, thus log(L X/L J)≤3.7.We note that van Paradijs and McClin-tock(1995)?nd log(L X(2-11keV)/L Opt(3000-7000?))～2.7 for typical bright LMXBs.
Rough(J?K)0colors have been reported for the soft X-ray transients in outburst QX Nor,or X1608-52(0.5<(J?K)0
2.4.X-ray Classi?cation and Completeness
We extracted the numbers of counts within1”radius of the sources(containing90%of source?ux at1.5keV5)in the0.5-1.5,0.5-4.5,and1.5-6keV bands.We then subtract1/3(ratio of areas)of the counts from background1”-2”annuli around these sources for each band(except where the annuli would overlap with other sources,where we offset the extraction re-gion).The results are listed in Table1.Constructing an X-ray CMD(as in GHE01a;Xcolor is de?ned as2.5log[0.5-1.5keV counts/1.5-6keV counts],while the vertical axis is log[0.5-4.5 keV counts]),we are able to identify two populations of faint sources with distinct X-ray https://www.sodocs.net/doc/365993790.html,ing an N H of1.2×1022 cm?2(derived from CLG02’s estimate of A V=6.75,using Pre-dehl&Schmitt’s(1995)conversion factor),we apply a uniform shift to the coordinates of the CMD as Pooley et al.(2002b)do for the moderately reddened globular cluster NGC6440.The identi?ed dim sources(and the color of the LMXB halo,which determines the background)are plotted in the upper panel of Figure5.The upper and right axes show the raw(absorbed) Xcolor and counts,while the lower and left axes represent the corrected(unabsorbed)Xcolor and counts.(We note that the in-terstellar medium decreases soft?uxes more than hard?uxes. Thus this simple shift of axes should be used with caution.) Another version of the same diagram,but plotting luminosities derived from spectral?ts(section2.5below)against corrected Xcolor as above,is shown in the lower panel of?gure5.Errors on the vertical axis are derived from errors in the counts,and do not include spectral uncertainties.This version of the diagram (inspired by Becker et al.2003),is more useful for comparing different clusters,but is dependent on spectral and reddening assumptions.The luminosities of the two populations roughly agree with the range observed for their respective counterparts in47Tuc and NGC6440.This suggests that the soft population (W2,W3,W4,W8)should be identi?ed with qLMXBs contain-ing neutron stars and that the harder population is composed of bright CVs.
We estimate our incompleteness,due to the X-ray halo and streak from the LMXB EXO1745-248,by the use of arti?cial-star tests.We generated uniform grids of arti?cial point sources at different luminosities using the MARX soft-ware6and merged the simulated data with our0.5-2.0keV mas-ter image to see if we could detect the fake sources.We scaled the average(Poisson-distributed)counts of the arti?cial sources to the average detected counts of real sources for which XSPEC ?tting gave the target luminosity(see section2.5below).E.g., for L X=1032erg s?1sources,the average number of counts per simulated source was20,and for L X=3×1032erg s?1it was 45counts.(The counts-luminosity relation varied depending on the kind of sources prevalent at different luminosities.)We performed the tests for arti?cial source0.5-2.5keV luminosi-ties of1032,1.7×1032,3×1032,and1033ergs s?1,assuming spectra similar to the average of our detected sources.We sep-arated the arti?cial sources by5”so that each source would suffer only the crowding that exists in the real data.We merged grids of point sources with the Chandra image four times for the 1032and1.7×1032tests,and three times for the other two tests, shifting the grids by several arcseconds between each trial.We then applied W A VDETECT to each image,keeping the same parameters as in section2.2.We counted the number of arti?-cial sources that landed within one optical core radius,between one and two core radii,and between two and three core radii for each trial,and the number of sources detected in each region, for a total of～1000point-source trials.We note that this ap-proach takes crowding into account,in that the test sources will be crowded by any real undetected sources.Parts of the1-2core radii annulus and2-3core radii annulus were imaged only in the 5ksec exposure and not in the42ksec exposure.The2-3r c bin also includes a small area that is completely unexposed(due to the use of the subarray).We scaled the MARX point source luminosities appropriately and performed separate trials for the low-exposure regions(where sources below3×1032ergs s?1 were never detected),and then included these results in the to-tal completeness calculations.This allowed an incompleteness factor to be determined for each radial annulus for each of four luminosity bins,which we created by equally dividing(in a log-arithmic sense)the ranges between our test luminosities.
5CXC Proposer’s Observatory Guide,https://www.sodocs.net/doc/365993790.html,/udocs/docs/docs.html 6Available at https://www.sodocs.net/doc/365993790.html,/ASC/MARX.
Terzan55 The number of real sources detected was too small to ef-
fectively determine both a radial distribution and a luminosity
function(1source within1r c,5within1-2r c,3within2-3
r c).Therefore we assumed a luminosity function of the form
dN∝L?γd ln L,L>L min following Johnston&Verbunt(1996).
We?xedγat0.5?rst,as found for sources between1032and
1033ergs s?1in NGC6440(Pooley et al.2002b)and the gen-
eral globular cluster system(Johnston&Verbunt1996),and
then allowedγto vary within the range0.29to0.78found by
Pooley et al.(2002b).The uncertainty in our incompleteness
due to different possible values ofγwas small compared with
errors due to small-number statistics.For each annulus,we de-
termined an average incompleteness factor using the results of
our testing and the weights derived from the luminosity func-
tion for the luminosity bins as speci?ed above.
To calculate how many sources were missed,we take the
number of sources detected in each annulus and apply the1σ
errors for small numbers given by Gehrels(1986).These num-
bers and errors are multiplied by1/f,where f is the complete-
ness fraction,to obtain the average density of sources in each
annulus.Then we calculated the expected numbers of missed
sources,using the estimated density and again applying small-
number errors.We ignore possible systematics in the method,
since the random errors are so large.Thus,above1032ergs s?1,
we?nd that within1r c we predict0.8+2.8
?0.8additional sources,
within1-2r c we predict1.5+2.8
?1.3additional sources,and within
2-3r c we predict1.1+2.6
?1.0sources,giving a total probable num-
ber of11.4+4.7
?1.8sources between1032and1033ergs s?1in Terzan
5.We note that this source population is very similar to the11 sources in this?ux range identi?ed in NGC6440(Pooley et al.2002b),as expected considering the similar cluster parame-ters(as derived in section3.2below).
The radial distribution we infer can be compared with that predicted by a“generalized King model”(Lugger,Cohn,& Grindlay1996)where the surface density of sources varies as S x(r)∝[1+(r2
whereαX=?(3q?1)and q=
6Heinke et al.
2.6.Spectral?tting of EXO1745-248
When observed by the ROSAT All-Sky Survey at a luminos-
ity of log(L X)=35.3,the0.5-2.5keV spectrum of EXO1745-
248was well?t by a powerlaw of photon indexα=1.2,using a hydrogen column of N H=2×1022cm?2(Verbunt et al.1995).
A ROSAT HRI observation in March1991found it at a lumi-
nosity of log(L X)=34.6,assuming the same spectrum(Johnston
et al.1995).Only a broad-band X-ray spectrum can provide detailed information about the nature of the LMXB,by simulta-neously constraining the photoelectric absorption column,any
iron lines or edges,and the overall spectral shape.Fortuitously,
there exist RXTE data simultaneous with our Chandra data,
which we also analyze.
To study the Chandra spectrum of EXO1745-248in out-
burst,we?rst selected an annulus from5to9arcseconds from
our best-determined position,and removed the portion affected
by the readout streak.This annulus is not affected by pileup,
and includes only one faint source(W4)which should not sig-
ni?cantly affect the spectral?tting(contributing～17of24739 counts).We extracted the spectrum using bins of width75eV (oversampling the energy resolution)and using a background
region on the same chip2’away.We also extract a lightcurve
from the same region.We see a gradual rise from0.4counts s?1
to0.5counts s?1during the?rst observation,while the second
(5ksec)observation?ve days later shows a rate of0.9counts
s?1(consistent with the rise in the ASM countrate).Flicker-
ing is clearly seen,but we put off detailed study of temporal variability for future work.An absorbed powerlaw of photon
index0.24provides a good?t to the continuum,with a strong
broad iron line(EW440+190
?230eV,σ=0.5±0.1keV)required
at6.7keV.However,we note that the spectrum will be altered both by the effects of dust scattering and by the energy depen-dence of the Chandra mirror psf(e.g.,Smith,Edgar&Shafer 2002).Therefore we extracted the spectrum from the readout streak,which does not suffer this hardening since the counts are recorded in the image core.
We used the CIAO task acisreadcorr to identify and repo-sition events in the EXO1745-248readout streak.We use a strip of width2pixels and omit a region of25pixels radius around the source location,to avoid photons from the X-ray halo and regions that suffered pileup.Pileup is not an issue in the selected region,as the8750counts selected give0.15 cts(0.841s CCD frame)?1,spread along a78×2pixel column. We extracted those events using the CIAO script psextract,and chose a background region adjacent to the strip(adjusting the BACKSCAL parameter by hand).We ignore data above10keV and below0.7keV(which is almost entirely background pho-tons,considering the high absorption to this source).We?t this Chandra readout streak spectrum along with the simultaneous RXTE data to understand the full spectral shape.
From the RXTE data,only the PCA STANDARD2data and the HEXTE Archive data(from both HEXTE clusters separately)are analyzed in this paper.We choose time in-tervals when three of the PCA units were on(the maxi-mum during the observation)and the elevation above the earth’s limb was greater than10degrees.Selecting only the top xenon layer data from the PCA,we use PCABACKEST version3.0(released Feb.1,2002)and PCARSP version 8.0,and correct the PCA data manually for deadtime.We used the HEXTE response matrices hexte_97mar20c_pwa.rmf and hexte_97mar20c_pwb013.rmf,and corrected the HEXTE data for background and deadtime using the HXTBACK and HXTDEAD(version2.0.0)scripts.We eliminate200seconds around a possible Type I X-ray burst,which we do not investi-gate in this paper(Type I bursts are common from EXO1745-248;see Inoue et al.1984).This gave total corrected exposure times of2060,731,and777seconds for the PCA and HEXTE clusters A and B respectively.We analyze PCA data from3to 25keV,and HEXTE data from25to125keV.Following Bar-ret et al.(2001),we add systematic errors of0.5%to the PCA data below15keV,and1%to PCA data above15keV using the FTOOLS GRPPHA.We leave an overall normalization free between the Chandra,PCA and HEXTE data,but link all other parameters between them in a joint?t.(The PCA and HEXTE relative normalizations are generally not well-calibrated;we ?nd the HEXTE normalization to be40%lower than the PCA normalization.)
The standard models for?tting neutron star LMXB spec-tra are an absorbed multicolor blackbody,or a simple black-body,with an additional hard component due to comptonization of soft photons by hot electrons(generally assuming,as here, a spherical geometry);see Barret et al.(2000),and Sidoli et al.(2001).Fitting a simpler model to the RXTE data,consist-ing of DISKBB plus a gaussian and powerlaw,gave a photon index of1.5,but failed to?t the HEXTE data due to a dropoff of?ux at high energies(χ2ν=8.1for230degrees of freedom). We begin by?tting our RXTE spectra with a model consisting of an absorbed multicolor blackbody(DISKBB;Makishima et al.2000),a comptonization model(COMPTT;Titarchuk1994), and the Fe-line gaussian,and with another model consisting of a blackbody,COMPTT,and the gaussian.The RXTE?ts for both models are signi?cantly improved by the addition of a smeared iron edge(“smedge”;Ebisawa et al.1994)near8keV, as expected for re?ection of Comptonized hard X-rays from a disk.(An F-test gives a probability of10?4that the smeared edge is not needed.)For the rest of this analysis,the“standard model”shall refer to PHABS(DISKBB+COMPTT+GAUS-SIAN)*SMEDGE*CONSTANT.The data/model ratio for PCA data,?t with the standard model with the normalizations of the gaussian and smeared edge set to zero,are shown in Figure6. The Chandra spectrum extracted displays what appear to be emission lines at～1.95and2.1keV.These features are not apparent in either the adjacent background,or the annu-lus spectrum described above.Similar features are also seen in the readout streak spectrum of GX13+1(Smith et al.2002), and in the high count rate continuous-clocking spectrum of RX J170930.2-263927(Jonker et al.2003).The best-?tting stan-dard model to the RXTE data does not give an acceptable?t to the Chandra data above5keV,where they overlap,nor does any other model?t both the Chandra and RXTE data;see Fig-ure7.For the best-?t standard model,the reducedχ2=1.41 for225degrees of freedom(null hypothesis prob.=5.7×10?5). Note that the Chandra spectrum appears to have been shifted in energy compared to the model predictions.We understand this effect as being due to a gain shift between the calibrated re-sponse of the timed exposure mode,and the actual response to events occurring during the readout period,when the voltages in the CCD are expected to be different.By measuring the dif-ference between the model and data edges near2keV,and the model(derived from RXTE?ts)and data between5and9keV, we estimate the gain shift at7±1%.
We attempt to compensate for this gain shift by increasing the size of the energy bins in the events?le by7%,to15.6 eV from14.6eV,while keeping the same response,ancillary
Terzan57 response,and background?les and extracting the altered-gain
spectrum in the same way.Fitting this shifted-energy spectrum,
along with the RXTE spectra,gives a greatly improved?t(see
Figure8).The standard model?t to these spectra gives a re-
ducedχ2=1.06for225dof(prob=0.250).The derived param-
eter ranges are robust;spectral?ts to RXTE data alone give
similar parameter values(with larger uncertainties for the soft
components,and N H virtually unconstrained).Although this
simple calibration must be used with a great amount of caution,
we believe that the quality of the resulting spectral?ts supports
our decision to utilize it for this analysis.
We list the best-?tting parameters(and90%con?dence er-
rors)for the two models in Tables3and4,along with the de-
rived values for the radii of the thermal components and the
ratio of luminosities in the two components f.We have not
corrected the multicolor disk-blackbody spectral parameters for
the spectral hardening factor expected in high accretion rate
systems(see Shimura&Takahara1995,Merloni et al.2000),
for ease of comparison to the work of SPO01and Parmar et
al.(2001)below.The averages of the best-?t luminosities(from
the two?ts’PCA normalizations)are L X(0.5-10)=2.1×1037,
and L X(0.1-100)=6.6×1037ergs s?1.Several authors(e.g.
Bloser et al.2000)have commented upon an absolute uncer-
tainty in the PCA?ux normalizations of order15%,which we
do not attempt to compensate for.We note that the ranges of N H
required by these?ts(1.3–1.9×1022cm?2)are slightly larger
than the infrared-derived estimate of CLG02,1.2×1022,indi-
cating probable internal absorption in the LMXB system.
3.DISCUSSION
3.1.Ultra-compact nature of EXO1745-248
We compare our results from?tting combined Chandra and
RXTE data with our“standard model”spectrum,to the spec-
tral?ts of SPO01and Parmar et al.(2001)for the other glob-
ular cluster LMXBs,observed with BeppoSAX.Five compar-
isons using the DISKBB+COMPTT model in SPO01separate
the“normal”(not ultra-compact)globular cluster LMXBs in
Terzan2,NGC6440,NGC6441,and Terzan6from the ultra-
compact(binary periods<1hour)LMXBs in NGC6624,NGC
1851,NGC6712,and(probably)NGC6652.SPO01note that
these comparisons suggest that the DISKBB model is physi-
cally meaningful only for the ultracompact LMXBs.
?The ultracompact LMXBs show an inner DISKBB tem-
perature below1keV,while the normal LMXBs show kT in be-
tween1.9and3.5keV.EXO1745-248shows kT in=0.80+.38
?.16
keV.
?The seed photon temperature kT0for the ultracompact bi-
naries is roughly equal to the inner DISKBB edge temperature,
while for the normal LMXBs the seed photon temperature is
4–5×lower.EXO1745-248seems to have similar values for
kT0(1.29+.40
?.13keV)and kT in(0.80+.38
?.16
keV).
?The inferred inner radius R in(cos i)0.5of the normal LMXBs is very small(0.3to1km),while for the ultracom-pact LMXBs it is3to45km.EXO1745-248’s implied R in
(cos i)0.5is8.7+5.0
?4.4km.
?The seed photon emission radius R W=3×104d
8Heinke et al. N H values are inconsistent with the optically derived extinction
value(CLG02;see?gure1in Kuulkers et al.2002),requir-
ing an absorption column within the LMXB system ten times
larger than shown in any other globular cluster LMXB.If this
were the case,we would expect a high inclination and signi?-
cant dips due to variable absorption.While these are not seen in
our Chandra and single-observation RXTE lightcurves,R.Wi-
jnands notes(https://www.sodocs.net/doc/365993790.html,m.)that such dips are indeed seen in
the full RXTE lightcurves of EXO1745-248’s2000outburst
(J.Homan,in prep.).This may imply that the system is at high inclination,though the observed disk re?ection component sug-
gests a low inclination.We also note that their spectral param-
eters,while not agreeing with ours in every detail,support our
claim that EXO1745-248is ultracompact,particularly in their
small value of kT in for their DISKBB+COMPTT?t.
3.2.Terzan5cluster parameters
It has long been suspected that Terzan5has one of the high-
est rates of close encounters between stars of any Galactic glob-
ular cluster.The“collision rate”,or rate of close encounters
given byΓ∝ρ20r3c/σ(whereρ0is the central density,r c is the
core radius,andσis the central velocity dispersion),is ex-
pected to predict the relative rates of formation of accreting
binary neutron star systems by two-body encounters(Verbunt
&Hut1987).Thus,the similar numbers of X-ray sources in
NGC6440and Terzan5might seem a surprise,as Terzan5
has been predicted to show three times as many collision prod-
ucts as NGC6440and17%of the total of such objects in the Galactic globular system(Verbunt2002).
Our calculation of the central density of Terzan5uses the extinction-corrected central surface brightnessμV(0)=20.5, combining the star-count pro?le of CLG02for the inner core
with the surface brightness pro?le of Trager et al.(1995)for normalization beyond10”.We use the central concentration pa-rameter c=2.0,core radius r c=7.′′9,heliocentric distance of8.7
kpc,and A V=6.75from CLG02,as well as M V=?7.91from
the updated Harris(1996;rev.1999)catalog.Following the prescription of Djorgovski(1993),our result is that Terzan5’s
central density is1.7×105L⊙pc?3,signi?cantly less than given
by Djorgovski(1993),Harris(1996),or CLG02.The former
two studies used a larger value of A V,which produces a larger correction of the surface brightness.CLG02had scaled the cen-
tral surface brightness value of Djorgovski(1993)for changes
in the pro?le,without updating the value of A V.We note that
this calculation is consistent with the lower limit of5.0×105
M⊙pc?3derived by Lyne et al.(2000)from the acceleration-induced˙E of pulsar Terzan5C,provided that M/L≥https://www.sodocs.net/doc/365993790.html,-
ing this central density,and the distance and core radius from
CLG02,the collision rate in Terzan5is similar to that in NGC
6440(5.9%of the total collision rate in the globular system),
instead of three times larger.This is consistent with the re-
sults of our arti?cial star tests(section2.3),which suggest that
Terzan5may contain～11sources in the range1032?33ergs s?1, compared to11in NGC6440.A full analysis of the luminosity function and density weighting of quiescent LMXBs in globular clusters will be presented in Heinke et al.(2003,in prep.)We
also revise the estimate of the central relaxation time of Terzan
5to2×108years from CLG02’s value of4×107years,follow-
ing Djorgovski’s(1993)scaling of t rc∝ρ0.5
0r3c.This indicates
that the cluster is not as close to the verge of core collapse as suggested by CLG02.The high density of compact binaries in the core seems to be due primarily to the high density of the cluster’s massive core.The large numbers of millisecond pul-
sars(60-200)in the core of Terzan5estimated by Fruchter and Goss(2000)would require a long period of high core density to form the millisecond pulsar progenitors.MSPs may be most ef?ciently formed early in the cluster history by intermediate-mass X-ray binaries(e.g.Davies&Hansen1998).Therefore it seems likely that Terzan5’s MSPs were formed early in its his-tory,as were those in47Tuc(Grindlay et al.2002),which has a roughly similar inferred ratio(～15?50)of MSPs to qLMXBs.
4.CONCLUSIONS
We have presented a reasonable infrared candidate to EXO 1745-248in Terzan5,identi?ed by its blue color and posi-tional coincidence with the boresighted Chandra position.We have assembled a broad X-ray spectrum using a simultaneous RXTE observation and the Chandra spectrum from the readout streak,slightly altering the energy scale of the Chandra read-out spectrum to account for observed gain variation during the readout.We utilized the empirical comparisons of SPO01to indicate that this LMXB appears similar to other ultracompact LMXBs in globular clusters,suggesting that EXO1745-248is the?fth ultracompact LMXB known in a globular cluster.We also identify a broad,strong6.55keV iron line,the strongest (EW=188+86
?83
eV)yet discovered in a globular cluster LMXB, with an accompanying smeared～8.1keV iron edge.
The superb resolution of Chandra has allowed us to identify nine faint X-ray sources within30”of an LMXB in outburst in Terzan5.Spectral?tting with bremsstrahlung and power-law models,and a neutron star hydrogen atmosphere model (Lloyd2003),suggests that four of these sources are qLMXBs, while?ve are candidate CVs.Arti?cial point source testing suggests that we are missing～30%of the sources in the range L X(0.5?2.5keV)=1032?33ergs s?1due to the presence of the LMXB in outburst.This implies a total cluster population of
11.4+4.7
?1.8
(1σ)sources with L X>1032ergs s?1(excluding the LMXB).A recalculation of the central density of Terzan5from updated cluster parameters gives log(ρ0)=5.23,suggesting that Terzan5is not the richest of the globular clusters in stellar en-counter products and is not as dynamically unstable as previ-ously thought(CLG02).Thus we?nd that the numbers of X-ray sources in Terzan5are consistent with the numbers discov-ered in other globular clusters and the currently favored forma-tion methods.
Upcoming Chandra observations of Terzan5and NGC6440 (PI:R.Wijnands)will allow us to better constrain the variability and spectra of the qLMXBs in those clusters.V and I observa-tions of Terzan5,with the HST Advanced Camera for Surveys in HRC mode at times when EXO1745-248is in outburst vs. quiescence,could unambiguously verify the identi?cation pro-posed here.
C.H.thanks https://www.sodocs.net/doc/365993790.html,ler,R.Wijnands,T.Gaetz,R.K.Smith, R.Edgar,and the anonymous referee for useful comments that have improved this paper.CH also thanks S.Wachter for ac-cess to unpublished data.This work was supported in part by Chandra grant GO0-1098A and HST grant GO-7889.01-96A. RXTE data and results provided by the ASM/RXTE teams at MIT and at the RXTE SOF and GOF at NASA’s GSFC.The Guide Star Catalogue-II is a joint project of the Space Tele-scope Science Institute and the Osservatorio Astronomico di Torino.This research has made use of the data and resources obtained through the HEASARC on-line service,provided by
Terzan59 NASA-GSFC,the VizieR catalogue access tool,CDS,Stras-bourg,France,and NASA’s Astrophysics Data System.
REFERENCES
Aldcroft,T.L.,Karovska,M.,Cresitello-Ditmar,M.L.,Cameron,R.A., Markevitch,M.L.2000,Proc.SPIE,4012,650
Arnaud,K.A.1996,in G.Jacoby&J.Barnes,(eds.)ASP Conf.Series Astronomical Data Analysis Software and Systems V.,vol.101,17
Asai,K.,Dotani,T.,Nagase,F.,&Mitsuda,K.2000,ApJS131,571 Barret,D.,Olive,J.F.,Boirin,L.,Done,C.,Skinner,G.K.,&Grindlay,J.E. 2000,ApJ533,329
Becker,W.et al.2003,ApJ submitted(astro-ph/0211468)
Bloser,P.F.,Grindlay,J.E.,Kaaret,P.,Zhang,W.,Smale,A.P.,&Barret,D. 2000,ApJ542,1000
Cohn,H.N.,Lugger,P.M.,Grindlay,J.E.,&Edmonds,P.D.2002,ApJ571, 818
Cool,A.M.,Haggard,D.,Carlin,J.L.2002,in F.van Leeuwen,J.D.Hughes, G.Piotto(eds.),Omega Centauri,A Unique Window into Astrophysics,V ol. 265of ASP Conference Series,ASP,p.277
D’Amico,N.,Possenti,A.,Fici,L.,Manchester,R.N.,Lyne,A.G.,Camilo, F.,&Sarkissian,J.2002,ApJ570,L89
Daniel,W.W.1990,Applied Nonparametric Statistics,2d ed.,PWS-Kent Danner,R.,Kulkarni,S.R.,Saito,Y.,&Kawai,N.1997,Nature388,751 Davies,M.B.,&Hansen,B.M.S.1998,MNRAS301,15
Deutsch,E.W.,Margon,B.,&Anderson,S.F.2000,ApJ,530,L21 Djorgovski,S.1993,in S.Djorgovski,G.Meylan(eds.),Structure and Dynamics of Globular Clusters,V ol.50of ASP Conference Series,ASP,p. 373
Ebisawa,K.et al.1994,PASJ46,375
Edmonds,P.D.,Grindlay,J.E.,Cohn,H.N.,&Lugger,P.M.2001,ApJ547, 829
Fruchter,A.S.&Goss,W.M.2000,ApJ536,865
Gehrels,N.1986,ApJ303,336
Giacconi,R.et al.2001,ApJ551,624
Grindlay,J.E.,Heinke,C.O.,Edmonds,P.D.,&Murray,S.S.2001a,Science 292,2290
Grindlay,J.E.,Heinke,C.O.,Edmonds,P.D.,Murray,S.S.,&Cool,A.M. 2001b,ApJ563,L53
Grindlay,J.E.,Camilo,F.,Heinke,C.O.,Edmonds,P.D.,Cohn,H.,&Lugger, P.2002,ApJ in press(available at astro-ph/0208280)
Grindlay,J.E.et al.2003,AN in press(available at astro-ph/0211527) Harris,W.E.1996,AJ,112,1487
Heinke,C.O.,Edmonds,P.D.,Grindlay,J.E.2001,ApJ562,363
Heinke,C.O.,Grindlay,J.E.,Lloyd,D.A.,&Edmonds,P.D.2002,ApJ, submitted
Hertz,P.&Grindlay,J.E.1984,ApJ282,118
Hünsch,M.,Schmitt,J.H.M.M,Sterzik,M.F.,&V oges,W.1999,A&AS, 135,319
Inoue,H.et al.1984,PASJ36,855Johnston,H.M.,Verbunt,F.,&Hasinger,G.1995,A&A298,L21 Johnston,H.M.&Verbunt,F.1996,A&A312,80
Jonker,P.G.,Mendez,M.,Nelemans,G.,Wijnands,R.,&van der Klis,M 2003,MNRAS(in press;astro-ph/0301475)
Kurucz,R.1992,Rev.Mex.A.A.23,181
Kuulkers,E.,den Hartog,P.R.,in’t Zand,J.J.M.,Verbunt,F.W.M.,Harris, W.E.,&Cocchi,M.2002,A&A(in press;astro-ph/0212028)
Lloyd,D.A.2003,MNRAS(submitted)
Lugger,P.M.,Cohn,H.N.,&Grindlay,J.E.1996,ApJ439,191
Lyne,A.G.,Mankelow,S.H.,Bell,J.F.,&Manchester,R.N.2000,MNRAS, 316,491
Makishima,K.et al.1981,ApJ247,L23
Makishima,K.et al.2000,ApJ535,632
Markwardt,C.B.&Swank,J.H.2000a,IAU Circ.7454
Markwardt,C.B.,Strohmayer,T.E.,Swank,J.H.,&Zhang,W.2000b,IAU Circ.7482
Merloni,A.,Fabian,A.C.,&Ross,R.R.2000,MNRAS313,193 Ortolani,S.,Barbuy,B.,&Bica,E.1996,A&A308,733
Ortolani,S.,Barbuy,B.,Bica,E.,Renzini,A.,Zoccali,M.,Rich,R.M.,& Cassisi,S.,A&A,2001,376,878
Parmar,A.N.,Oosterbroek,T.,Sidoli,L,Stella,L.,&Frontera,F.2001,A&A 380,490
Pooley,D.,et al.2002,ApJ569,405
Pooley,D.,et al.2002,ApJ573,184
Predehl,P.&Schmitt,J.H.M.M.1995,A&A293,889
Rutledge,R.E.,Bildsten,L.,Brown,E.F.,Pavlov,G.G.,&Zavlin,V.E.2001c, ApJ578,405
Shimura,T.,&Takahara,F.1995,ApJ445,780
Sidoli,L,Parmar,A.N.,Oosterbroek,T.,Stella,L.,Verbunt,F.,Masetti,N.,& Dal Fiume,D.2001,A&A368,451
Smale,A.P.1995AJ,110,1292
Smith,R.K.,Edgar,R.J.,&Shafer,R.A.2002,ApJ581,562
Titarchuk,L.1994,ApJ434,570
Trager,S.C.,King,I.R.,&Djorgovski,S.1995,AJ,109,218
van Paradijs,J.&McClintock,J.E.1994,A&A,290,133
van Paradijs,J.&McClintock,J.E.1995,in X-ray Binaries,ed.Lewin,van Paradijs,van den Heuvel(Cambridge U.Press),p.58
Verbunt,F.,&Hut,P.1987,IAU Symp.125,187
Verbunt,F.,Bunk,W.,Hasinger,G.&Johnston,H.M.,1995,A&A,300,732 Verbunt, F.2002,in ASP Conf.Ser.ωCentauri,a unique window in astrophysics,ed.van Leeuwen,Piotto,Hughes(available at astro-ph/0111441)
Wachter,S.1997,ApJ485,839
Wachter,S.1998,Ph.D.thesis,University of Washington
Wang,Z.et al.2001,ApJ563,L61
10Heinke et al.
T ABLE1
N AMES,P OSITIONS AND C OUNTS OF D ETECTED S OURCES
Source Name(Label)RA Dec0.5-4.5keV0.5-1.5keV 1.5-6keV
(17:)(-24:)(counts)(counts)(counts)
Sources not associated with Terzan5
Note.—Sources detected in and around Terzan5.RA and Declination values for cluster members begin with17:and-24:, respectively.Errors are centroiding errors within the(shifted to match GSC2.2)Chandra frame,and do not include systematic errors incurred in matching to other frames(absolute errors perhaps0.′′5;see text).Counts in each energy band were determined by subtracting the average count rate in surrounding2”annulus from the counts in1”circle around source position(leading to one negative entry).
T ABLE2
S PECTRAL F ITS TO F AINT S OURCES
Source H-atmosphere Bremsstrahlung Powerlaw0.5-2.5L X
(kT,eV)(χ2ν/dof)(kT,keV)(χ2ν/dof)(α)(χ2ν/dof)(ergs s?1)
Note.—Spectral?ts to faint cluster sources,with background subtraction,in XSPEC.All?ts
include photoelectric absorption?xed at the cluster N H,1.2×1022cm?2.Hydrogen atmosphere?ts
are made with radius?xed to10km.X-ray luminosities are unabsorbed for the range0.5to2.5keV,
from hydrogen-atmosphere NS?ts(W2,W3,W4,W8)or thermal bremsstrahlung?ts.Errors in all
the tables are90%con?dence for a single parameter.
Terzan511
T ABLE3
S PECTRAL F ITS TO LMXB EXO1745-248C ONTINUUM
Model N H COMPTT BB or DiskBB aχ2ν/dof
(1022cm?2)(kT0)(kT e)(τp)(kT)(R,km)(f b)
Note.—Spectral?ts to the LMXB EXO1745-248from RXTE data and the Chandra readout streak
with energy scale alteration(see text),with background subtraction in XSPEC.A gaussian,a smeared edge, photoelectric absorption,and comptonization using the XSPEC model COMPTT is included in both?ts.
The?rst also includes a multicolor disk-blackbody(DBB),while the second includes a blackbody(BB).
Values are in keV unless otherwise noted(f andτp are unitless).Values from normalizations utilize the PCA normalizations;uncertainty in PCA absolute normalization is not included.
a kT and R refer to the inner edge of the diskb
b for model1(where R is R in cos(i)0.5),and to the blackbody
for model2.
b f is the ratio of the blackbody or diskbb?ux to COMPTT?ux,over0.1-100keV range.
T ABLE4
S PECTRAL F ITS TO LMXB EXO
1745-248F EATURES
Gaussian
Smeared edge
Note.—Spectral?ts to the
Fe-line gaussian and disk re?ection
smeared edge in the spectrum of
the LMXB EXO1745-248.The
spectral?ts in this table are to the
“standard model”including a multi-
color disk-blackbody(see text);line
and edge parameter values for the
other model considered in the text
are within the errors in this table.
Values are in keV unless otherwise
stated,exceptτmax which is unitless.
12Heinke et al.
50010001500
10
20
30
F I
G .1.—The RXTE All-Sky Monitor lightcurve of the LMXB EXO1748-25in Terzan 5.The dates of the Chandra and HST observations are marked.
Terzan 513
2
3
4
5
6
7
8
10
F I
G .2.—Chandra ACIS-I image of the globular cluster Terzan 5,energy range 0.5-2.0keV .The dominant feature is the piled-up halo from the LMXB EXO 1745-248in its high state.The streak is due to out-of-time LMXB events recorded during the frame transfer,and the position of the LMXB has no good events due to pulse saturation.Several other sources are visible,marked with 1”circles and indicated with their shorthand names.The cluster core is indicated with a 7.9”(1r c )circle.
14Heinke et al.
F187W
F110W
F187W F110W F110W A
F187W B
B
A
A
B
B subtracted
0.4"
B subtracted
A&B subtracted
F110W/F187W
A&B subtracted
A
A
F I
G .3.—Finding chart,showing stars A and B in F110W (upper left)and F187W (lower left)HST NICMOS data,and the ratio image F110W/F187W (upper right).The best estimate of the uncertainty in the position of EXO 1745-248is indicated by the 0.8”(2σ)error circle.Note that star A is the bluest object in this ?eld (some red giant cores appear blue due to saturation).The results of our manual psf ?tting of stars A and B are shown at lower right.
Terzan515
F IG.4.—Aperture photometry color-magnitude diagram of Terzan5,from the HST NIC2camera(using highly dithered imaging via the drizzle algorithm)on the core of Terzan5.Stars from the photometry of EGC01within1”of the seven X-ray error circles are circled.Xs mark the locations of variable stars from EGC01 (X2is the dimmer one).Two stars that are more than3σto the blue of the distribution are indicated by<.>.The lower is the blend of stars A and B(see text). The separately derived positions of stars A and B are indicated,with errors from the photometry.The horizontal branch and the expected blue straggler sequence, terminating in a Kurucz model(Kurucz1992)of a1.6M⊙star(twice the turnoff mass)are indicated.
16Heinke et al.
F IG.5.—X-ray color-magnitude diagrams of Terzan5.Upper panel produced from the absorption-corrected counts in the0.5-1.5,0.5-4.5,and1.5-6keV energy bands(lower and left axes;upper and right axes provide the observed colors and magnitudes).We correct roughly for photoelectric absorption by shifting the data +0.57units on the left axis and+1.86units on the bottom axis.However,note that the effects of absorption are not uniform.The probable qLMXBs are marked by crosses,and probable CVs by?lled triangles.We plot representative errors for a few points.Background subtraction leaves W4with-1counts in the hard band;we show its Xcolor lower limit and counts range(it is only clearly detected in the soft band).The Xcolor of EXO1745-248’s X-ray halo is indicated with a dotted line and labeled’LMXB Xcolor’.Lower panel produced using the same absorption-corrected color for x axis,but using the unabsorbed0.5-6keV luminosity for y axis. Luminosities derived from best-?tting hydrogen atmosphere?ts for W2,W3,W4,and W8,and best-?tting thermal bremsstrahlung?t for W5,W6,W7,W9,and W10.Luminosity errors are based on0.5-4.5keV counting statistics,and do not include uncertainties in spectral?tting.
Terzan517
F IG.6.—Ratio of EXO1745-248PCA data to standard model(see text),with normalizations of the gaussian iron line and smeared edge set to zero to show the relative contributions.
F IG.7.—EXO1745-248spectrum from Chandra readout streak data,RXTE PCA,and RXTE HEXTE data,simultaneously?tted with the standard model(see text).Note the poor?t to the Chandra data near2keV and the pronounced wave above5keV,suggesting an incorrect energy calibration.
18Heinke et al.
F IG.8.—EXO1745-248observed(top)and unfolded(bottom)spectra,with best-?t absorbed(DISKBB+COMPTT+GAUSSIAN)×SMEDGE model spectrum. Here the energies assigned to Chandra readout streak events have been increased by7%.Relative normalizations of the different instruments are allowed to vary (see text).The6.55keV iron line is visible in both the RXTE and Chandra unfolded spectra,along with the DISKBB and COMPTT components(the COMPTT is the harder component).