Gemini Spectroscopy of Galactic Bulge Sources: A Population of Hidden Accreting Binaries Revealed?

1.8 MB·English

by  Jianfeng Wu

Mon.Not.R.Astron.Soc.000,000–000(0000) Printed30March2015 (MNLATEXstylefilev2.2)
Gemini spectroscopy of Galactic Bulge Sources: a population of
hidden accreting binaries revealed?⋆
5
1
0 Jianfeng Wu,1† P. G. Jonker,1,2,3 M. A. P. Torres,2,3 C. T. Britt,4,5‡ C. B. Johnson,4‡
2
R. I. Hynes,4‡ S. Greiss,6 D. T. H. Steeghs,1,6 T. J. Maccarone,5 C. O. Heinke,7,8

r
a T. Wevers3
M
1Harvard-SmithsonianCenterforAstrophysics,60GardenStreet,Cambridge,MA02138,USA
7 2SRON,NetherlandsInstituteforSpaceResearch,Sorbonnelaan2,3584CA,Utrecht,TheNetherlands
2 3DepartmentofAstrophysics/IMAPP,RadboudUniversity,Heyendaalseweg135,6525AJ,Nijmegen,TheNetherlands
4DepartmentofPhysicsandAstronomy,LouisianaStateUniversity,BatonRouge,LA70803-4001,USA

5DepartmentofPhysics,TexasTechUniversity,Box41051,LubbockTX,79409-1051,USA
]
E 6DepartmentofPhysics,UniversityofWarwick,Coventry,CV47AL,UK
H 7DepartmentofPhysics,UniversityofAlberta,CCIS4-183,Edmonton,ABT6G2E1,Canada
8MaxPlanckInstituteforRadioAstronomy,AufdemHugel69,53121Bonn,Germany
.
h
p
-
o
30March2015
r
t
s
a
ABSTRACT
[

2 We presentGeminispectroscopyfor21candidateopticalcounterpartstoX-raysources
v discoveredintheGalacticBulgeSurvey(GBS).Forthemajorityofthe21sources,theoptical
7
spectroscopyestablishes that they are indeed the likely counterparts.One of the criteria we
1 used for the identification was the presence of an Ha emission line. The spectra of several
0 sources revealed an Ha emission line only after careful subtraction of the F or G stellar
2
spectralabsorptionlines.Inasub-classofthreeofthesesourcestheresidualHa emissionline
0
isbroad(>400kms−1)whichsuggeststhatitisformedinanaccretiondisk,whereasinother
. ∼
1 casesthelinewidthis suchthatwe currentlycannotdeterminewhethertheline emissionis
0
formedinanactivestar/binaryorinanaccretiondisk.GBSsourceCX377showsthishidden-
5 accretionbehaviourmostdramatically.The previously-identifiedbroadHa emissionofthis
1
source is notpresentin its Gemini spectra taken ∼1 year later. However,broademission is
:
v revealedaftersubtractinganF6templatestarspectrum.TheGeminispectraofthreesources
i (CX446,CX1004,andCXB2)aswellasthepresenceofpossibleeclipsesinlightcurvesof
X
thesesourcessuggestthatthesesourcesareaccretingbinariesviewedunderahighinclination.
r
a
Keywords: binaries:close—stars:emissionline,Be—Galaxy:Bulge—X-rays:binaries
1 INTRODUCTION
PrevioussurveysoffaintX-raysourceshavebeenfocusedonthe
GalacticCenterorglobularclusters.WhiletheGalacticCenterSur-
vey (e.g., Muno et al. 2003) benefits from a high source density,
⋆ Based on observations obtained at the Gemini Observatory, which is the crowding and significant extinction make the optical/infrared
follow-upnecessaryforclassificationdifficult(e.g.,Mauerhanetal.
operated by the Association of Universities for Research in Astronomy,
Inc.,underacooperativeagreementwiththeNSFonbehalfoftheGemini 2009).
partnership:theNationalScienceFoundation(UnitedStates),theNational TheGalacticBulgeSurvey(GBS;Jonker etal.2011;Jonker
Research Council (Canada), CONICYT (Chile), the Australian Research etal.2014;PaperI&IIhereafter)isamultiwavelengthprojectthat
Council(Australia),MinistériodaCiência,TecnologiaeInovação(Brazil) isdesignedtoallowoptical/infraredclassificationofX-raysources
andMinisteriodeCiencia,TecnologíaeInnovaciónProductiva(Argentina).
detectedintheGalacticBulge.TheGBSconsistsofChandraand
† E-mail:[email protected]
multiwavelength observations of two 6◦×1◦ stripscentered 1.5◦
‡ Visitingastronomer,CerroTololoInter-AmericanObservatory,National
OpticalAstronomyObservatory,whichareoperatedbytheAssociationof above and below the Galactic plane (see Fig. 1 of Paper I), thus
Universities forResearchinAstronomy,undercontractwiththeNational avoidingthe|b|<1◦regionswithseriouscrowdingandextinction
ScienceFoundation. problems,whilestillmaintainingarelativelyhighsourcedensity.
2 JianfengWuet al.
TheGBSutilizesChandra observationswithanexposureof2ks Therearealsoseveralextensivespectroscopicstudiesonindividual
foreachpointing;theexposuretimeischosentomaximizetherel- GBSsources.Rattietal.(2013)presentedadynamicalanalysisof
ativenumbersoflow-massX-raybinaries(LMXBs)tocataclysmic GBSsourceCX93inwhichtheymeasuredthemassofthecompact
variables (CVs). The completed Chandra observations have de- primaryandthecompanionstar,andconcludedthatthesourceisa
tected 1640 unique X-ray sources (Paper II), agreeing well with longorbitalperiodCV.Hynesetal.(2014)identifiedasymbiotic
theestimationinPaperIwhichalsogaveabreak-downoftheex- X-raybinarywithacarbon starcompanion intheGalacticBulge
pectednumbersofvariouskindsofobjectsbasedonsourcedensity, basedonthespectraofitsopticalcounterpart.
expectedChandrafluxlimit,etc.Amongthe1640X-raysources, In this work, we present Gemini spectroscopy of 21 GBS
∼600areexpectedtobeCVs,includingbothintermediatepolars X-raysourceswithabetterX-raypositionalaccuracythantheme-
(IPs)andnon-magnetic CVs,whilethenumber ofLMXBsisex- dian (due to low off-axis angles). These 21 sources are listed in
pectedtobe∼250(seeTable2ofPaperI).Wealsoexpect∼600 Table1.Wewillrefertothesesourceswiththeirlabels,i.e.CX(or
chromosphericallyactivestarsorbinaries,e.g.,RSCanumVenati- CXB)IDs,introducedintheGBSsourcecatalogue(PaperIandII).
corumvariables(RSCVnsystems;Walteretal.1980). GBSsources listedinPaper Ihave theprefixof “CX”,whilethe
TheGBScombinesalargeskycoveragewiththegoodsensi- remainingGBSsources, detected inthelastquarter of theChan-
tivitytofaint X-raysourcesandtheexcellentpositional accuracy dracoveragepublishedinPaperII,havetheprefixof“CXB”.The
possible with Chandra. There are two main science goals of the sources in each catalogue were ranked by their Chandra counts,
GBS(see§1ofPaperIformoredetails):1)constrainingthenature whereCX1hasthemost countsamongtheCXsources. Thema-
ofthecommon-envelope phase inbinary evolutionbycomparing jority of the sources in this work are CXB sources, while previ-
the observed number of sources with model predictions in each ous works were focused on CX sources. This paper isstructured
class, e.g.,CVsand LMXBs; 2) measuring themass of thecom- asfollows.In§2wedescribetheGeminiobservationsanddatare-
pactobjectsinX-raybinaries,e.g.,eclipsingquiescentblackhole duction.In§3wepresenttheanalysisoftheGeminispectroscopy,
(BH)andneutronstar(NS)LMXBs,toinvestigatetheGalacticBH includingspectralclassificationandradialvelocityanalysis.In§4
mass distribution(e.g., Özel et al. 2010) and to constrain the NS wegiveresultsanddiscusseachinterestingobject.Overallconclu-
equationofstate(EoS). sionsaresummarizedin§5.
Bothofthesciencegoalsrelyonthemultiwavelengthidentifi-
cationandclassificationofthislargesampleoffaintX-raysources.
Avarietyofoptical/infraredfollow-upcampaignshavebeencon-
ducted.Hynesetal.(2012)identified69X-raysourcesintheGBS 2 OBSERVATION&DATAREDUCTION
usingtheTycho-2catalogue.Thesesourcesarecoincidentwithor
2.1 GeminiSpectroscopy
veryclosetothebrightstarsinthatcatalogue, mostofwhichare
likelytobetherealopticalcounterpartstotheX-raysources.This The list of 21 GBS sources of which we obtained Gemini spec-
sampleisamixofobjectswithabroadrangeofspectraltypes,in- troscopyconsistsofsixCXsourcesand15CXBsources.Thesix
cludingbothlate-typestarswithcoronalX-rayemissionandearly- CX sources were proven to be interesting on the basis of earlier
typestarswithwindX-rayemission.Manysourcesareforeground spectra and/or photometric variability. For example, three of the
objectsinsteadofresidingintheGalacticBulge.Brittetal.(2014) CXsources(CX377,CX446,andCX1004)haveshownHa emis-
reportedonanopticalphotometricsurveyofthreequartersofthe sionlinesintheirpreviousspectra(Torresetal.2014)obtainedby
sky area covered by the Chandra GBS, and presented the light theVIsibleMulti-Object Spectrograph (VIMOS) mounted on the
curves of variable objects consistent with the X-ray positions of VeryLargeTelescope(VLT).Basedonourfollow-upstrategy(i.e.,
GBSsourcescataloguedinPaperI.Aboutaquarteroftheoptical prioritizing sources with higher positional accuracy and brighter
counterpartsarevariable,andtheyareexpectedtobeamixofIPs, in optical/infrared), the 15 CXB sources are chosen to have off-
non-magnetic CVs, LMXBs, and RS CVns. Greiss et al. (2014) axisangles less than 5arcmin intheir Chandra observations and
provided likelynear-infraredidentificationofGBSX-raysources alsohavesufficientcountstoallowforanaccurateX-rayposition
using current near-infrared sky surveys. Maccarone et al. (2012) (<1′′). Optical/infrared brightness, colour and photometric vari-
found12candidateradiomatchestotheGBSX-raysourcesusing abilityarealsoamongthefactorsofsampleselection.
thearchivalNRAOVLASkySurvey(NVSS;Condonetal.1998). The finding charts of our sources are shown in Appendix A
Themajorityofthemappeartobebackgroundactivegalacticnu- (seeFig.12–13).ThecoordinateslistedinTable1areforthecandi-
clei. dateopticalcounterparts,i.e.,theobjectsforwhichwetookGem-
Optical/infraredspectroscopyofthedetectedX-raysourcesis ini spectroscopy. The astrometry was performed on images from
anessentialtooltoinvestigatetheirnature.Accretingbinariescan VLT/VIMOS (CX377, CX446, and CX1004; Torres et al. 2014),
beidentifiedbytheemissionfeaturesintheiropticalspectra.The Gemini(CXB117),andMosaic-II/DarkEnergyCamera(DECam;
onlyfirmwaytodistinguishwhitedwarf(WD),NS,andBHasthe other sources; see §2.2). All of them have a RMS accuracy of
primariesoftheX-raybinariesisviameasurementsoftheaccretor <0.2′′,whilesomesources(CX377,CX1004,andCXB117)have
masses,whichrequireshigh-qualityoptical/infraredspectroscopy. <0.1′′positionalaccuracy(Brittetal.2014;Torresetal.2014).
Brittetal.(2013)presentedfiveaccretingbinariesidentifiedinthe Opticalspectroscopyofthe21GBSsourceswasobtainedwith
GBSbasedonthestrongemissionlinesintheirspectra,including GMOS (Gemini Multi-Object Spectrograph; Davies et al. 1997)
threelikelyIPs,oneCVundergoingadwarfnovaoutburst,andone mountedontheGemini-SouthTelescopeinChile.Alltheobserva-
likely quiescent LMXB (qLMXB). Torres et al. (2014) identified tionsweretakenbetween2012Apr20and2013May4underpro-
22newaccretingbinariesviatheHa emissionlinesintheiropti- grammes GS-2012A-Q-44 and GS-2012A-Q-67 (see Table 1 for
calspectra.Theydevelopedcriteriaofaccretingbinariesbasedon anobservation log). Nineobjects have multi-epoch spectroscopy,
theequivalent width(EW)ofHa emissionline(EW>18Å),the while other objects have one epoch; each epoch has 900–3600 s
breadth of theHa emission line(FWHM>400 km s−1), or the integrationtime.Theseeingofeachspectroscopicobservationwas
strengthofHeIl 5876,6678incaseofnarrowandweakHa lines. measured from the corresponding acquisition image (see the last
Geminispectroscopyof GalacticBulgeSources 3
columnofTable1).Typicalseeingwasaround0.7′′ (witharange thefulldescriptionandresultsofthisvariabilitycampaign). Typ-
of0.5–1.3′′).GMOSwasoperatedinlong-slitmode.Weusedthe icalseeing fortheseimaging observations wasaround 1′′ (witha
R400_G5325 grating(400linemm−1)anda0.75′′ slit.Thetwo- rangeof 0.8–3.0′′).Thedatawerereduced using theNOAO Mo-
dimensional spectra were binned by a factor of two in both spa- saicPipeline(Shaw2009). Weperformeddifferentialphotometry
tial and spectral dimensions, resulting in a spatial dispersion of usingAlard’simagesubtractionISIS(Alard&Lupton1998;Alard
0.15′′/pixel and a spectral dispersion of 1.36 Å/pixel. The spec- 2000) to obtain the changes in flux with respect to reference im-
tral resolution is estimated to be ≈5 Å FWHM for the sources ages. The zero-point flux in the reference images was measured
that had filledthewhole slitduring theobservation (i.e.,thesee- with either aperture photometry or DAOPHOT-II (Stetson 1987).
ing was greater than the slit width 0.75′′), while spectral resolu- Thenumberofvariableinterloperswithinthe95%errorcirclesof
tionshouldbebetterthan≈5ÅFWHMforthesourceswithsee- theX-raypositionis∼40(∼3%;see§3.1ofBrittetal.2014for
inglessthan0.75′′.AlltheGemini/GMOSspectraweresplitinto details).
three equal parts in the wavelength dimension by detector gaps. The optical photometry information of four CX sources
Thebluestpartofthespectrawasignoredintheanalysisbecause (CX84,CX377,CX446,CX1004)inthispaperisshowninTable2,
ofthelowsignal-to-noiseratio(S/N)inthespectraduetotheex- whiletheirlightcurveshavebeenpresentedinliterature(CX84in
tinctiontowardsoursourcesandthelackofarclinesinthispartof Fig.6ofBrittetal.2014;CX377,CX446andCX1004inFig.A2
the spectra. The middle part (wavelength range ∼4800–6100 Å) of Torreset al.2014). For CX138, therearetwoblended sources
andtheredpart(wavelengthrange∼6200–7600Å)ofthespectra intheMosaic-IIimageatitsX-rayposition; neither ofthepossi-
werereducedandanalyzedseparately.Theresultspresentedinthis blecounterpartsisvariable.ThecounterparttoCX139issaturated
workaremainlybasedontheanalysesoftheredpartofthespectra intheMosaic-II imaging. Among thefour CX sources withlight
becauseitisleastaffectedbyextinction. curves, only CX84 shows possibly periodic behaviour with a pe-
The Gemini/GMOS data were reduced using the FIGARO riod of 4.67 days. However, it isworth noting that it is currently
package implemented in the STARLINK software suite and the not possible to confirm the periodic nature of this variability be-
packagesofPAMELAandMOLLYdevelopedbyT.Marsh.Thetwo- causethebaselineofourmonitoringwasonly8days,whichisless
dimensionalspectrawerebias-correctedandflat-fielded.Thebias than two full cycles. The light curves of the other three sources
was corrected using the overscan areas of the detectors. We uti- onlyhaverandomflickeringwithanRMS∼0.05–0.1 magnitude,
lizedtheflatfieldstakendirectlyfollowingeachtargetobservation althoughCX446possiblyexperienceseclipseevents.
forflatfielding.Wefitthebackgroundonbothsidesofthespectra Optical photometry for fiveof theCXB sources inour sam-
withasecond-orderpolynomialanddeterminedthebackgroundat ple(CXB2,CXB64,CXB82,CXB99,andCXB113)wasobtained
thepositionofthespectra.Theobject spectrawereoptimallyex- usingtheDarkEnergyCamera(DECam)instrument mounted on
tractedusingtheoptexpprocedureinthePAMELApackage(Marsh theBlanco4-metertelescopeatCTIOonthenightsofJune10and
1989).Thespectrawerewavelength-calibratedinMOLLYusingthe 11 of 2013. The average seeing for both nights was 1.3′′ with a
helium-argon arc spectra which were taken either right after ob- rangeof0.9–1.9′′.DECamprovidesa2.2×2.2squaredegreefield
servingthetargetorattheendofthenight.Theaveragearcspec- ofview combining 62scienceCCDs,8focusCCDs,and4guid-
trum was used in case of multiple arc spectra in one night. The ing CCDs with a scale of 0.27′′ per pixel. For all of our images,
resultingrootmeansquare(RMS)scatteronthewavelengthcali- theSDSSr′ filterwasusedwithexposuretimesofeither2×90s
brationis <0.3Å.Thewavelengthcalibrationwasexaminedusing or 2×1 s for the faint and bright sources, respectively. The DE-
∼
the skyline O I l 6300.303. Offsets relative to the wavelength of Campipelinereductionprovidedtheresampledimageswithcross-
thislinehavebeencorrected.Forsourceswithseeinglessthanthe talkcorrections,overscan,trimmedsections,biassubtraction,flat-
slitwidth(0.75′′),thecentroidinguncertainty(i.e.,ifthesourceis fieldingandsaturationmasks.1WethenusedstandardIRAFtasks
not placedinthemiddleof theslit)mayintroduce asmallwave- (including wcsctran, digiphot, and apphot) to extract magni-
lengthshift(Bassaetal.2006).Thiswavelengthshiftcannotbecor- tudesandfluxesandtogeneratelightcurvesthroughtheuseofdif-
rectedbyexaminingthesky-linewavelengthasthosefillthewhole ferential photometry. Calibration of the target stars was achieved
slit.However,itcanpotentiallybeassessedbycheckingthewave- byusingreferencestarsinthefieldofviewthatwerecontainedin
lengths of diffuse interstellar bands (DIBs; Herbig 1995). Three theCarlsbergMeridian14catalogue(Evansetal.2002)andVizieR
suchsources(CX84,CXB149,andCXB174)havestrongDIBsat catalogue:I/304.
l 5780 for whichthelineprofilesdonot deviatefromGaussians. ThephotometryofthesefivesourcesareincludedinTable2
Wecheckedthesefeaturesandfoundtheyhaveminorshiftsrelative whiletheirlightcurvesareshowninFig.1.AmongthefiveGBS
totherest-framewavelength(65kms−1forCX84,−30kms−1for sources, CXB2 shows possible eclipsing/dipping events and an
CXB149,and−75kms−1forCXB174inheliocentricframe).We “outburst”.Althoughtheoutburstwastowardstheendofthenight
corrected the wavelength scale for these small shifts. Each spec- withhigherairmass,visualinspectionoftheimagesconfirmsthat
trum was normalized by dividing it by the result of fittinga 5th- the brightening is real. This outburst is possibly the reprocessed
orderpolynomialfittothecontinuum. X-rayemissionfromaTypeIX-rayburst,whichissimilartothe
caseofNSLMXBEXO0748−676(Hynesetal.2006).Thelight
curve shows a likely periodic modulation of P≈0.447 day (see
2.2 OpticalPhotometryfromMosaic-II&DECam
thephase-folded lightcurve).CXB82andCXB99bothappear to
Time-resolvedoptical photometry for allsixCXsources wasob- belongperiodvariableswithperiodslonger thanour2-night ob-
tainedwiththeMosaic-II imagermountedontheBlanco4-meter servingrun.Forbothsources,thereisasteadyincreaseinbright-
telescopeattheCerroTololoInter-AmericanObservatory(CTIO) ness by about ∼0.1 and ∼0.25 magnitude, respectively. Due to
in2010 July8–15. Nineteen exposures in theSloanr′-band with
an integration time of 120 s were taken on 45 overlapping fields
tocoveraninesquaredegreearea,whichcontainallbutsevenof 1 See DECam Data Handbook at http://www.noao.edu/meetings/decam/
the X-ray sources identified in Paper I (see Britt et al. 2014 for media/DECam_Data_Handbook.pdf.
4 JianfengWuet al.
Figure1.TheDECamr′-bandlightcurvesforCXBsources(CXB2,CXB64,CXB82,CXB99,andCXB113).CXB2showsaperiodicmodulationof0.447
days.CXB113hasasinusoidalmodulationof0.588dayinitslightcurve.Thephase-foldedlightcurvesofthesetwoobjectsarealsoincluded.
Geminispectroscopyof GalacticBulgeSources 5
Figure2.Gemini/GMOS spectraforsixsourceswithHa emissionlines Figure4.Gemini/GMOSspectrafortheremaining6sourcespresentedin
(CX446,CX1004,CXB2,CXB64,CXB99,andCXB113).ThebroadHa thispaper.Allthespectraarenormalizedtounity.ThepositionsoftheHa
emissionofCX446,CX1004,andCXB2likelyoriginate intheaccretion andtheHb emissionlinesarelabeledbythedottedlines.
disk,whilethenarrowHa andHb emissionofCXB64andCXB113are
an indication of chromospherically active binaries. The Ha emission of
CXB99isnarrowandweak.Allthespectraarenormalizedtounity.The
positions oftheHa andtheHb emissionlines arelabeled bythedotted
depth. Complementary coverage from the Two-Micron All Sky
lines.
Survey (2MASS; Skrutskie et al. 2006) and UKIDSS Galactic
PlaneSurvey(GPS;Lucasetal.2008)werealsoutilizedforbright
the lack of r′ standard stars in their fields, CXB64 and CXB113 (Ks<12.5)andfaint(Ks>16)sources,respectively.Greissetal.
only have approximate magnitudes without absolute calibration. (2014)developedamethodtoestimatethelikelihood(bycalculat-
CXB64 has a counterpart USNO B1.0 star 0578−0732346 with ingthefalsealarmprobability)ofanear-infraredsourcetobethe
the optical magnitude of R∼18.7 and I =16.9. CXB113 shows counterpart of theX-raysource. However, for theoptical sources
asinusoidalmodulationwithperiodofP=0.58791(12) day.The where we have obtained Gemini/GMOS spectra, we are able to
phase-foldedlightcurveforthissourceisalsoshowninFig.1.The search for the near-infrared counterparts that matches our optical
opticalcounterparttoCXB113wasalsoidentifiedbyOGLE(Opti- sources by comparing the optical and near-infrared images. We
calGravitationalLensingExperiment;Field:BUL_SC37,StarID: set a 0.2′′ astrometric error circle, and searched for any matches
9614; Udalski etal.1992). TheOGLEsource hasoptical magni- withinthat error circlebetween theoptical images and VVVim-
tudesofV =19.135(154)andI=15.847(49).Italsoshowsape- ages.Thenwevisuallyexaminedbothimagesandselectedthetrue
riodicmodulationofP≈0.588day. near-infraredcounterparts.Avisualinspectioniscrucialgiventhat
theGBSfieldsarecrowded.
Table 3 lists the near-infrared counterparts to 13 of the 21
2.3 Near-InfraredPhotometry
opticalsourceswithGemini/GMOSspectroscopy.Theothereight
The likely near-infrared matches to the GBS X-ray sources de- sources do not have reliable near-infrared counterparts. Three of
tectedbyChandra(Greissetal.2014)weremainlyobtainedfrom them (CX446, CXB26, and CXB137) are too faint in the VVV
the public variability survey VISTA Variables in the Vía Láctea framesandtherearenonear-infraredsourcesattheopticalposition.
(VVV; Minniti et al. 2010). The VVV survey provides the most Forthe other fivesources (CX1004, CXB64, CXB117, CXB189,
complete near-infrared coverage of the GBS area with consistent and CXB201), there could be near-infrared matches but they are
6 JianfengWuet al.
Figure3.Gemini/GMOSspectraforninesourceswithHa absorptionfeatures(CX138,CX377,CXB26,CXB73,CXB117,CXB149,CXB174,CXB189,
andCXB201),manyofwhichalsohaveHb absorption.Allthespectraarenormalizedtounity.ThepositionsoftheHa andtheHb emissionlinesarelabeled
bythedottedlines.
blendedwithnearbysources,forwhichthecurrentVVVdatare- 3 DATAANALYSIS
leasedoesnotprovidephotometry.
The Gemini/GMOS spectra of the 21 GBS sources are shown in
Fig.2–4.Threesources(CX446,CX1004,andCXB2)showappar-
entbroadHa emissionlines,whileanotherthree(CXB64,CXB99,
andCXB113)havenarrowHa emission(seeFig.2).WefitGaus-
sian profiles to these broad Ha emission lines to measure their
Theinfraredcolourscanbeusedtoestimatethedistanceofthe widthandvelocityseparations(ifdouble-peaked)usingthemgfit
sourcebycalculatingtheneededextinctiontomatchthesourcein- procedure in the MOLLY package. The equivalent widths (EWs)
fraredcolourstothoseofastandardstarwiththesamespectraltype of these Ha lines were measured using the light procedure in
(e.g.,Rattietal.2013).SomeoftheGBSsourceshaveaninfrared MOLLY.NineotherobjectshaveHa absorptionfeatures(Fig.3).
excess, i.e., their infrared colours are redder than expected based ThestellarabsorptionfeaturesshownintheGemini/GMOSspectra
onopticalcolours(see§4ofHynesetal.2012),whichispossibly canbeutilizedtoperformspectralclassification,andradialveloc-
forcaseswheretheopticalclassificationmayrepresentthehotter ityanalysis.Thespectraoftheremainingsixsourcesareshownin
component ofabinary(orofablend,ifanotherstarisserendipi- Fig.4;theyappearatfirstsighttohaveneitherHa emissionfea-
touslyalongthesamelineofsight)whiletheinfraredcoloursare turesnorHa absorptionfeatures.
fromthecoolercomponent.Accretingsystems(CVsandLMXBs) SomeoftheobjectsinoursamplehavestrongDIBsintheir
andBestarswithacircumstellardiskmayalsohavecoloursthatdo spectra.Weestimatedthereddeningforthesesourcesviatheequiv-
notmatchthoseofsinglestars.Bestarscouldalsohavesignificant alentwidth(EW)oftheDIBatl 5780withthecalibrationinTa-
colour variations due to the formation/dissolution of the circum- ble3ofHerbig(1993).Wealsocomparedthemeasuredreddening
stellardisks. totheBulgereddeningalongthelineofsightprovidedbyGonza-
Geminispectroscopyof GalacticBulgeSources 7
lez et al. (2011,2012), which utilized the Red Clump stars in the
Bulge.2
3.1 SpectralClassification
We utilize the optimal subtraction technique following previous
works (e.g., Marsh et al. 1994; Ratti et al. 2013) to classify the
spectra.Asetoftemplatestarspectrawaschosenfromthelibrary
oftheUltravioletandVisualEchelleSpectrographParanalObser-
vatory Project (UVES POP; Jehin et al. 2005), covering spectral
typesfromA0toM6withluminosityclassofV(seeTable4).The
UVESPOPtemplatespectraprovidecoverageoverthewavelength
rangeof3000–10000Åwithaspectralresolutionof∼80,000.The
templates were re-sampled and Gaussian-smoothed to match the
spectral resolution of the object spectra. The object spectra were
Doppler-shiftedintothesamerestframeandaveraged.Eachstel-
lar template was optimally subtracted from the object spectrum, Figure 5. The Ha region of both epochs of Gemini/GMOS spectra for
whilea c 2 test was performed on the residuals. All the emission CX446,acquired on2012Jun22(blueline) and2013May4(redline),
lines,DIBs,andtelluricfeatures(e.g.,Kurucz2006;Wallaceetal. respectively,showingevidenceforstrongvariabilityintheEWoftheHa
2011)weremaskedduringtheprocedure.Theresultingc 2values betweenthetwoepochs.TheVLT/VIMOSspectrumofCX446isalsoover-
laid(blackline).
for each template were compared with each other. The template
withtheminimumc 2valueprovidesourbestestimateofthespec-
tralclassificationofeachGBSsourceinoursample.Thesespectral
standardstartemplate.Threesources(CX84,CX138,CX139)have
classification procedures were first performed on the red part of radialvelocityvariationsof∼100kms−1ontimescalesofdays.
theGeminispectra.Weverifiedourresultsbyperformingthesame
procedurestothemiddlepartoftheGeminispectraandtheresults
areconsistentwitheachother.Theresultsofspectralclassifications
arelistedinTable5.Notewearenotcontrollingluminosityclass 4 RESULTS&DISCUSSION
in the spectral classification. Nine of the 21 GBS sources in our
4.1 ObjectswithHa EmissionLines
samplewerespectrallyclassified.Theuncertaintyofspectralclas-
sification is estimated to be one or two spectral sub-classes. The There are six sources with apparent Ha emission in their Gem-
otherthreesourceslistedinTable5areclassifiedasearly–midM- ini/GMOSspectra(Fig.2).Threeof them(CX446,CX1004, and
typebasedontheirspectralfeatures. CXB2) have broad Ha emission lines (FWHM>800 km s−1),
∼
We have also tried to measure the rotational broadening of while those of the other three sources (CXB64, CXB99, and
theabsorptionfeatures.However,ourlimitedspectralresolutionof CXB113) are narrow (FWHM<200 km s−1). The strong, broad
∼
∼200kms−1precludedthedeterminationofrotationalvelocities. Ha emission likely originates from an accretion disk, indicating
CX446,CX1004,andCXB2arelikelyaccretingbinaries(§4.1.1).
CXB64 and CXB113 are possibly chromospherically active
3.2 RadialVelocityAnalysis stars or binaries because of their narrow and weak Ha emission
lines(Torresetal.2014;alsoseeFig.2).TheH−Kcolour(0.300±
TheradialvelocitiesoftheopticalcounterpartofeachGBSsource
0.033; see Table3) of CXB113 agrees well withthat of a M4V–
aremeasuredbycross-correlatingtheobjectspectrausingthexcor
M5Vstar(H−K≈0.30;seeTable5ofPecaut&Mamajek2013),
procedureinMOLLY.Foreachsource,alltheobjectspectraandthe
whileitsJ−H colour(0.509±0.030) appearsslightlybluerthan
template spectrum were rebinned to the same velocity dispersion
thatofaM4V–M5Vtypestar(J−H≈0.57).
withthe vbin procedure. We also masked all the emission lines,
CXB99 has aweak, narrow Ha emission. Afterthe optimal
interstellar/telluricfeaturesduringthecross-correlationprocedure. subtraction with the best-fit K2V template, the Ha emission ap-
Weperformedtwosetsofthecross-correlationanalysis.Thefirst pearstobestronger,indicatingthatitpartiallyfillsintheHa ab-
setusesthefirstsourcespectrumasthecross-correlationtemplate,
sorptionlineinthespectrum.
i.e.,wederivetheradialvelocity(RV1)relativetothefirstsource
spectrum.FortheGBSsourcesthatwerespectrallyclassifiedusing
the procedure in §3.1, we also performed a second set of cross-
4.1.1 PotentialQuiescentAccretingBinaries:CX446,CX1004,
correlationanalyses,takingtheUVESPOPstandardstarwiththe
andCXB2
best-fitspectraltypeasthetemplate.Thusthesecondsetofradial
velocity(RV2)isrelativetothetemplatestar.Allspectrahadbeen BothepochsoftheGemini/GMOSspectraofCX446showbroad
shiftedtoaheliocentricframe. Ha emission (seeFig. 5). The strength of the Ha line isweaker
RadialvelocityvalueswerederivedbyfittingaGaussianpro- inthespectratakenon2013May4thanthaton2012Jun22.The
fileto the cross-correlation function. The results are listed in Ta- linewidth(1250±50kms−1)issmallerthanthatoftheHa emis-
ble6.Sevensourceshavenon-zeroradialvelocitiesrelativetothe sionlinesintheVLT/VIMOSspectraofCX446(2200±50kms−1;
Torresetal.2014),whilethelineEWisbigger.Noabsorptionlines
fromthecompanionstararevisibleintheGeminispectra.Wealso
2 See the Bulge Extinction And Metalicity (BEAM) calculator at see no evidence of He I l 6678. Thelight curve of CX446 does
http://mill.astro.puc.cl/BEAM/calculator.php. show apossibleeclipseevent withadepth of0.4magnitude (see
8 JianfengWuet al.
(−170±20kms−1;Torresetal.2014).Onepossiblescenariofor
thisisthattheaccretiondiskofCX1004isprecessing.Thedouble-
peaked profile of the Ha emission lineis asymmetric inthree of
thefourepochs;therelativestrengthofthebluepeakandthered
peakvarieswithtime(seeFig.6).Thisbehaviourcanbeexplained
bythepresenceofanS-waveoriginatinginahot-spotorthedonor
star(see§IVofJohnstonetal.1989).
Theprominentmolecularabsorptionsinthewavelengthrange
of6300–7300Å aresignaturesofanM-typecompanionstar.How-
ever,thedetailedrotationalbroadeninganalysisandspectralclassi-
ficationarenotfeasiblefortheGemini/GMOSspectraofCX1004.
AsdiscussedinTorresetal.(2014),thelinewidthandthedouble-
peakvelocityseparationareconsistentwiththatofaneclipsingqui-
escentCVorqLMXB.However,possiblyowingtothefaintnessof
thesource,thelightcurveofCX1004(seeFig.A2ofTorresetal.
2014) does not show significant periodic behaviour. Torres et al.
(2014)suggestedCX1004tobeanearbysourceduetothelackof
Figure6.TheHa regionofthefourepochsofGemini/GMOSspectrafor
diffuseinterstellarbands.
CX1004(fromtoptobottominchronologicalorder;seeTable1).Allfour
The four epochs of Gemini/GMOS spectra of CXB2 are
spectrashowabroad,double-peaked Ha profilewhichvariedinstrength
showninFig.7.Thespectraofthefirstandthirdepochsappearto
betweentheobservations.
haveweak,broadHa emission.Bothofthelineshavelargeveloc-
ityoffsets(whicharealsovisibleinthefigure)withvpeak=−260±
60kms−1and−200±30kms−1,respectively.Thewidthsofthe
twolinesareFWHM=825±175kms−1 and725±100kms−1,
respectively.ThespectrumofthefourthepochshowsanarrowHa
absorption feature (FWHM = 180±40 km s−1), which also has
a significant velocity offset (vpeak =−175±15 km s−1). These
large negative offsets could originate in a high space velocity of
this source, which may imply a NS or a BH as the primary star.
CXB2 also shows the Ha emission variation as we have seen in
CX377.Nophotosphericlinefromthecompanionstarisdetected
fromthespectraofCXB2.
CXB2 is brighter in X-rays than most of the GBS sources.
It has 147 Chandra ACIS-I counts in 0.3–8.0 keV band in
a 2 ks exposure. CXB2 was serendipitously detected by an
archival Suzaku observation with 53 ks exposure time (ObsID:
507031010). Full details of the X-ray analysis will be presented
elsewhere, but we briefly summarize relevant results here. The
Suzaku XIS0 and XIS1 data were reduced and analysed using
Figure7.TheHa regionofthefourepochsofGemini/GMOSspectrafor standard FTOOLS. A good spectral fit (c 2/n =1.0) was found for
CXB2(fromtoptobottominchronologicalorder;seeTable1).Thespectra
ofthefirstandthirdepochsshowbroad,weakHa emission,whereasitis CXB2 with an absorbed power-law model (photon index G =
absentinthebottomspectrum.Boththeemissionandabsorptionfeatures 0.96±0.11,NH=1.5−+11..04×1021 cm−2).TheX-rayfluxis1.5×
havesubstantialnegativevelocityoffsetwhichmayindicatethehighspace 10−12ergcm−2s−1,whichisconsistentwiththatfromourChan-
velocityofthisobject. dra observation (1.3×10−12 erg cm−2 s−1, assuming the same
spectralmodel).ThissourceisprobablyassociatedwiththeASCA
source AX J1754.0−2929 which is 7.2′′ away (see catalogue in
Fig.A2ofTorresetal.2014; onHJD=2455387.82). Nosignifi- Sakano et al. 2002). The X-ray flux in the ASCA observation is
cant periodicityisfound inthelightcurve. CX446isacandidate 1.4×10−12cm−2s−1in0.7–10keVband,whichissimilartothe
eclipsingCVorqLMXB. X-rayfluxintheChandraobservation.Thissourcewas,however,
TheGemini/GMOSspectraofCX1004show double-peaked not detected in ROSAT and is classified as a transient (Paper II).
Ha emissioninallfour epochs (seeFig.6).WemeasuretheHa TheN valueissmallerthanthatpredictedtotheGalacticBulge
H
emissionlinepropertiesusingthespectraofthelastepochbecause (∼1022 cm−2),henceweinferthatCXB2probablyliesinthe1–
of itsbest S/N. Thelinewidth isFWHM= 2500±100 kms−1, 4 kpc distance range. Its near-infrared colours also indicate low
whichisbroaderthanthatintheVLT/VIMOSspectraofCX1004 reddening.TheX-rayluminositywouldbe1–40×1032ergs−1(as-
(FWHM = 2100±20 km s−1; see Torres et al. 2014). The line sumingtheabovedistancerange),whichisconsistentwitheithera
strength (EW=38.0±0.6 Å) has also slightly increased com- high-inclinationCVoraqLMXB.
pared to that in the VLT/VIMOS spectra (EW=32.9±0.4 Å).
The velocity separation between the red and blue peaks is D v=
1160±30 km s−1, which is consistent with the result in Tor- 4.2 ObjectswithHa inAbsorption
res et al. (2014). The centroid of the Ha line does not have sig-
nificant radial velocity (velocity offset −15±20 km s−1), while Nine sources (CX138, CX377, CXB26, CXB73, CXB117,
that of theVLT/VIMOSspectra shows substantial radial velocity CXB149, CXB174, CXB189, and CXB201) in our sample only
Geminispectroscopyof GalacticBulgeSources 9
Figure8.Theresidualspectrafortheninespectrallyclassifiedsources(CX84,CX138,CX139,CX377,CXB73,CXB82,CXB99,CXB149,andCXB174)
afteroptimallysubtractingthestandardstartemplate. ThedottedlineineachpanellabelsthepositionofHa .AlltheresidualspectrashowHa emission
features,exceptCXB149andCXB174.Forallpanelsweplottheobservedspectrumontop,thebest-fittemplatestarspectruminthemiddleandtheresidual
spectrum,obtainedaftersubtractingthebest-fittemplatefromtheobservedspectrum,atthebottom.Theoffsetbetweenthethreespectraperpanelischosen
sothatthespectralfeaturesarediscernible.Theinterstellarabsorptionfeatureat∼6280Åismaskedoutwhenoptimalsubtractionwasperformed.
showHa absorptionlinesintheirGemini/GMOSspectra(Fig.3). dial velocity variation of ∼ 100 km s−1 between its Gemini
We are able to obtain a spectral classification for five of them spectra. Therefore, CX138 is a likely candidate of the hidden
(CX138, CX377, CXB73, CXB149, and CXB174). The residual accreting binaries. The reddening measured via the strength of
spectra of these sources after optimally subtracting their best-fit DIB l 5780 (EW = 1971±131 mÅ) is AK =1.29±0.08, which
stellar templates are shown in Fig. 8. Three of them (CX138, also puts the object in the Galactic Bulge, if not farther. For the
CX377, and CXB73) have unambiguous Ha emission features, GalacticBulgedistance(∼8kpc),theabsoluteV-bandmagnitude
which partially fill in the stellar Ha absorption features in our ofCXB138wouldbeMV =−2,suggestingagiantcompanionstar.
Gemini/GMOSspectra.Wemeasuredthestrength(EW)andwidth
(FWHM)oftheHa emissionshownintheseresidualspectra(see
(b)CX377isalsoacandidatehiddenaccretingbinary.TheGem-
Table7).ThebroadHa emission(FWHM>400 kms−1;seeTa-
∼ ini/GMOS spectra of CX377 do not show any notable emission
ble7andFig.8)ismostlikelyduetotheaccretiondisk.Therefore,
features (see Fig. 3 and Fig. 9). However, the optical spectra of
for the first time, we may have discovered a hidden population
CX377taken∼1yearbefore(2011May28and2011July23)by
of accreting binaries where the accretion disk contribution to the VLT/VIMOSshowstrongbroadHa emissionwithanasymmetric
opticallightissmall,andtheHa emissionlinesbecomeapparent
double-peakedprofile(Torresetal.2014;seeFig.9).Theintrinsic
onlyaftersubtractingthestellarcontributiontothespectra.Inwhat width of the Ha emission is ∼1200 km s−1, while the double-
follows, we provide further details on the spectroscopic and/or peakseparation is∼700kms−1.Thisapparent dramaticchange
photometricpropertiesofseveralindividualobjects.
of the spectrum is not caused by incorrect targeting. The finding
chartsofCX377forourGemini/GMOSspectroscopy andforthe
(a) CX138 has a broad Ha emission line (FWHM= VLT/VIMOSspectroscopy(Fig.10;alsoseeFig.9inTorresetal.
1350±75 km s−1) in its residual spectrum. It also has a ra- 2014)showtheexactsametargets.Theslitwasplacedduringour
10 JianfengWuet al.

Figure 10. The finding charts of CX377 for our Gemini/GMOS spec-
troscopy (left panel) and the VLT/VIMOS spectroscopy in Torres et al.
(2014;rightpanel).Thetwoshortthickbarsineachpanelindicatethepo-
sitionofCX377,whilethethinlinesshowtheslitposition.Theskyarea
showninbothpanelsis15′′×15′′.Northisupandeastisleftinbothpan-
Figure9.ThecomparisonoftheHa regionofCX377opticalspectra,ac- els.
quired byVLT/VIMOS(upperline) andGemini/GMOS (lowerline), re-
spectively. The broad Ha emission line in the VLT/VIMOSspectrum is
absentintheGemini/GMOSspectrum,indicatingstrongspectralvariabil-
ityeventhoughthesourceremainsinquiescence. ofCX377obtainedinJuly2010onlyshowsnon-periodicflickering
withaRMSscatterof0.06magnitude(seeFig.A2ofTorresetal.
2014). ThestrengthofDIBl 5780(EW=1093±86 mÅ)inthe
Gemini/GMOSobservations(seeFig.10)tocontainboththetarget spectraofCX377correspondstoAK=0.73±0.06.Extinctionmap
andtheadjacentobjectsothatweareabletodistinguishtheirspec- from Gonzalez et al. (2012) gives AK =0.72±0.15 for CX377,
tra. The seeing during the Gemini/GMOS observations (∼0.6′′) whichisconsistentwiththeextinctionderivedabove.Therefore,it
wasbetterthanthatduringtheVLT/VIMOSobservations(∼1′′). islikelythatCX377residesintheGalacticBulge.
Thus, the contamination of the nearby bright sources is less of a Theneutralhydrogencolumndensity,NH =(1.44±0.14)×
problemfor theGemini/GMOSspectra. Therefore, theHa emis- 1022 cm−2,wasestimatedfromtheextinctionfollowingtherela-
sionfeatureofCX377indeedhassignificantvariationwithinone tioninGüver &Özel (2009). CX377 has 7counts intheACIS-I
year. Although the Gemini/GMOS spectra of CX377 only show 0.3–8 keV band in the 2 ks exposure. Assuming an absorbed
Ha absorptionlines,itsresidualspectrumafteroptimallysubtract- power-law spectrum with photon index of 1.6, the unabsorbed
ingtheF6VstandardstartemplateshowsbroadHa emission(see X-rayfluxofCX377is≈1×10−13 ergcm−2 s−1.Atadistance
Fig. 8). The Ha emission has not completely disappeared, but it of∼8kpc,theX-rayluminositywouldbeLX=8×1032ergs−1.
issignificantlyweakerduringourGemini/GMOSobservationthan Withtheestimatedextinctionandtheapparentr′-bandmagnitude
duringtheVLT/VIMOSobservationsofTorresetal.(2014).Ithas (seeTable2),wecalculatetheabsolutemagnitudeofMr′ =−1.3
also become narrower (FWHM = 660±30 km s−1 in the Gem- andtheX-raytoopticalfluxratio fX/fr′ ∼0.005;bothvaluesare
ini/GMOS residual spectrum). This rare behaviour of Ha emis- consistentwiththoseforagiantcompanionstar.However,thedisk
sion line variations indicates significant accretion disk variations luminosityatthetimeoftheVIMOSobservationshouldhavebeen
for CX377. The soft X-ray transient GRO J1655−40 has shown larger than that of the F6 giant (Mr′ =−1.3), implying that the
similarstrongvariabilityofitsHa emissionline(e.g.,Soriaetal. sourcemighthaveexperiencedafaintX-rayoutburst(cf.Wijnands
2000).GROJ1655−40hasaperiodof2.62days;thespectraltype &Degenaar2013).
ofitscompanionstarhadbeenclassifiedasF3–F6(Orosz&Bai-
lyn1997),similartothatofCX377(seebelow).However,theHa (c) CXB149 and CXB174 were spectrally classified as G6 and
emission line variability of GRO J1655−40 occurred when this F4, respectively. Their residual spectra after optimal subtraction
objectwasgoingthroughanoutburst cycle,whileCX377hasre- do not show unambiguous Ha emission or absorption features.
mainedinquiescence.ItispossiblethatCX377experiencedafaint The Gemini/GMOS spectra could possibly be from interlopers
X-rayoutburstendingbeforeJune1,2012whenourGeminispec- instead of thetrue optical counterparts of the X-raysources. The
traweretaken,aslongastheX-rayfluxremainedbelowthedetec- Ha absorption features of these two sources are significantly
tionthresholdofallskymonitorssuchasMonitorofAll-skyX-ray blueshifted from the laboratory wavelength (−109±7 km s−1
Image (MAXI; Matsuoka et al. 2009). Another possibility isthat for CXB149 and −179±8 km s−1 for CXB174). This shift
CX377hasvariableaccretionrateduringquiescence. is not introduced by the wavelength calibration process since
CX377isspectrallyclassifiedasF6type.Therelativeradial offsets to the sky lines were small and they have been corrected.
velocitiesbetweentheGeminispectraofCX377(seeTable6)are It cannot be explained by the wavelength shifts caused by the
consistentwithzeroorlessthan10kms−1,whichisnotsurpris- centroiding uncertainty within the slit (which are −30 km s−1
inggiventhatthethreeobjectspectrawereonly15minutesapart, and −75 km s−1, respectively; see §2.1). These velocities are
andtheorbitalperiodofCX377isexpectedtobemuchlongerthan consistent with that for starsresiding inthe Galactic Bulge (e.g.,
15 minutes for an F6 main sequence star filling its Roche lobe. Zoccali et al. 2014). Regarding the reddening measurements,
Then we cross-correlated the object spectra to the F6V star tem- CXB149hasEW(DIBl 5780)=1099±28mÅ;thecorresponding
plate(HD16673).Theradialvelocitiesrelativetothestartemplate reddening is E(B-V) = 2.11±0.05 and AK =0.73±0.02, which
rangefrom−50to−20kms−1.Theradialvelocityofthestartem- putthisobjectintheGalacticBulgeorfarther(AK =0.30±0.07
plateitselfisconsistentwithzero(−4±5kms−1).Thelightcurve from the map of Gonzalez et al. 2012). For CXB174, EW(DIB