A Chandra X-ray study of the interacting binaries in the old open cluster NGC6791

3.8 MB·English

by  Maureen van den Berg

DraftversionJanuary14,2013
PreprinttypesetusingLATEXstyleemulateapjv.5/2/11
ACHANDRAX-RAYSTUDYOFTHEINTERACTINGBINARIESINTHEOLDOPENCLUSTERNGC6791
MaureenvandenBerg1,2,FrankVerbunt3,4,GianpieroTagliaferri5,TomasoBelloni5,LuigiR.Bedin6,andImantsPlatais7
1AstronomicalInstitute“AntonPannekoek”,UniversityofAmsterdam,SciencePark904,1098XHAmsterdam,TheNetherlands
2Harvard-SmithsonianCenterforAstrophysics,60GardenStreet,Cambridge,02138MA,USA;[email protected]
3DepartmentofAstrophysics/IMAPP,RadboudUniversityNijmegen,POBox9010,6500GLNijmegen,TheNetherlands
4SRON,NetherlandsInstituteforSpaceResearch,Sorbonnelaan2,3584CAUtrecht,TheNetherlands
5INAF/OsservatorioAstronomicodiBrera,ViaE.Bianchi46,I-23807Merate(LC),Italy
6INAF/OsservatorioAstronomicodiPadova,Vicolodell’Osservatorio5,I-35122Padova,Italyand
7DepartmentofPhysicsandAstronomy,TheJohnsHopkinsUniversity,Baltimore,MD21218,USA
DraftversionJanuary14,2013
3
1 ABSTRACT
0
WepresentthefirstX-raystudyofNGC6791,oneoftheoldestopenclustersknown(8Gyr). OurChandra
2
observationisaimedatuncoveringthepopulationofcloseinteractingbinariesdowntoL ≈ 1×1030 ergs−1
X
n (0.3–7keV).Wedetect86sourceswithin8(cid:48)oftheclustercenter,including59insidethehalf-massradius. We
a identifytwentysourceswithproper-motionclustermembers,whichareamixofcataclysmicvariables(CVs),
J
activebinaries(ABs),andbinariescontainingsub-subgiants. Withfollow-upopticalspectroscopyweconfirm

0 thenatureofoneCV.Wediscoveronenew,X-rayvariablecandidateCVwithBalmerandHeIIemissionlines
1 initsopticalspectrum; thisisthefirstX-ray–selectedCVconfirmedinanopencluster. ThenumberofCVs
perunitmassisconsistentwiththefield,suggestingthatthe3–4CVsobservedinNGC6791areprimordial.
]
WecomparetheX-raypropertiesofNGC6791withthoseofafewoldopen(NGC6819,M67)andglobular
E
clusters(47Tuc,NGC6397). ItispuzzlingthatthenumberofABsbrighterthan1×1030 ergs−1 normalized
H
byclustermassislowerinNGC6791thaninM67byafactor∼3–7. CVs, ABs, andsub-subgiantsbrighter
. than1×1030 ergs−1 areunder-representedperunitmassintheglobularclusterscomparedtotheoldestopen
h
clusters,andthisaccountsforthelowertotalX-rayluminosityperunitmassoftheformer. Thisindicatesthat
p
- theneteffectofdynamicalencountersmaybethedestructionofevensomeofthehardest(i.e.X-ray–emitting)
o binaries.
r Subject headings: open clusters and associations: individual (NGC6791); X-rays: binaries; stars: activity;
t
s binaries: close;cataclysmicvariables
a
[

1 1. INTRODUCTION enable a detailed comparison with theoretical models of the
evolution of stellar clusters and of the binary population in
v X-rayemissionoflate-typestarsarisesinhotgascontained
them. To explain the large number of blue stragglers ob-
1 bymagneticloopsabovethestellarsurface. Theseloopsare
3 produced by the interaction of convective motion and differ- served in M67 (4 Gyr), the model by Hurley et al. (2005)
3 entialrotationinthestellarenvelope;hencethemagneticac- requires a large initial number of short-period binaries that
2 tivity of a late-type main-sequence star—and with it the X- act as seeds from which blue stragglers are formed. As the
1. ray emission—is found to be higher in rapidly rotating stars simulatedhard-binaryfractionandperioddistributiondonot
evolve much, the close binaries remain abundant throughout
0 (see, for example, Pallavicini 1989). As stars age their rota-
thelifetimeofthecluster. Geller&Mathieu(2012)pointout
3 tionslowsdown,anditwasasurprisethataROSATstudyof
thatcomparisonwiththeobservedbinariesintheoldcluster
1 theoldopenclusterM67discoveredalargenumberofX-ray
NGC188(6.5Gyr;Meibometal.2009)findsnoevidenceof
: sources(Bellonietal.1993). Comparisonwithopticalobser-
v such a high fraction of close binaries. The number of clus-
vationsprovidedtheexplanation: inanoldclusterstarsmay
i tersstudiedisstillsmall,asisthenumberofdetailedmodels,
X rotate rapidly when tidal forces spin them up towards coro-
andmorestudyisrequiredbeforeafinalverdictontheseand
tationwiththebinaryrevolution(Bellonietal.1998). X-ray
r otherdiscrepancies(seeGeller&Mathieu(2012)fordetails)
a studiesofoldopenclustershelpinidentifyingsuch,otherwise
canbemade.
inconspicuous,activebinaries(ABs)inwhichtidalforcesare
orhavebeeneffective,andcontributetotheunderstandingof Bynow,observationsofglobularclusterswiththeChandra
X-ray Observatory have uncovered hundreds of faint (L (cid:46)
tidalinteraction. Sincethetypicaltimescalefortidalinterac- X
1033ergs−1)closebinaries,onlypartofwhichhavebeenclas-
tion rapidly increases with the ratio of the orbital separation
sifiedsofar. Asthesebinariesareanimportantdriverofthe
to stellar radius (Zahn 1989), only stars in binaries with rel-
long-termevolutionofacluster,understandingtheproperties
atively short periods or relatively large radii experience the
effects of tidal locking. X-ray observations of ABs in old and frequency of the various binary classes is a major goal
ofinvestigating(globular)clusterX-raysources. Sometimes
clusters thus probe the populations of hard binaries, and the
the outcome of a dynamical encounter is a binary or multi-
relative number of such systems found in clusters with dif-
ple system in a very unusual configuration, and this allows
ferent properties is an important tool for understanding the
effectsoftheclusterenvironmentonthebinarycontent. ustostudysourcetypesthatarenot,oronlyrarely,foundin
thefield. BesidesABs,whicharethedominantX-raysource
Thankstoprolongedopticalstudies,e.g.intheWIYNOpen
class in old open clusters, globular-cluster X-ray sources in-
Cluster Study (Mathieu 2000), our knowledge of the binary
clude cataclysmic variables (CVs), low-mass X-ray binaries
population in clusters is reaching completeness levels which
2 vandenBergetal.
in quiescence (qLMXBs), and milli-second pulsars (MSPs).
ThecorrelationofthenumbersofqLMXBsandCVswiththe
stellarencounterfrequencyrevealstheimportantrolethatdy-
namicalencountersplayinthecreationofsuchbinaries(Poo-
ley&Hut2006;Heinkeetal.2006), butthenetbalancebe-
tweenformationanddestructionislessclear. Foralargepart
thisisduetothelackofareferencepointthatquantifiestheir
numberdensitiesinadynamicallyinactiveenvironment. Ide-
ally,thecomparisonshouldbemadeagainsttheGalacticfield
wherestellardensitiesarelow,butthisiscomplicatedbecause
of the generally limited information on distance and age for
stars in the field, and the intrinsic scarcity of sources such
asqLMXBs. Oldopenclustersprovideanalternativethatis
worthwhile to explore, but so far only few have been stud-
ied in detail in X-rays. Moreover, it should be kept in mind
that dynamical encounters cannot be completely ignored in
open clusters. Observational indications for the occurrence
ofdynamicalinteractionshavecomefromindividualsystems
whosepropertiesaredifficulttoexplainbybinaryevolutionas
itwouldtakeplaceoutsideaclusterenvironment(e.g.vanden
Bergetal.2001).
Figure1. V-bandimageofNGC6791fromStetsonetal.(2003)showingin
In this paper we present the results of the first X-ray ob- blacktheoutlineoftheACISchips(labelled)asinourChandraobservation.
servation of NGC6791, which, at an age of about 8 Gyr, is Thesmallwhite,innermostcircleindicatestheclustercenterfromdeMarchi
one of the oldest known open clusters in our Galaxy. Be- et al. (2007). The larger white circles mark the area inside the half-mass
sidestheChandradata,weobtainedopticalspectratoclassify radiusrh (4.(cid:48)42), andtheareainside8(cid:48), whichwasanalyzedinthispaper.
Northisup,easttotheleft.
candidate optical counterparts to the detected X-ray sources.
WeobservedNGC6791withtheAdvancedCCDImaging
NGC6791 has been studied extensively in the optical. The
Spectrometer (ACIS)1 on Chandra from 2004 July 1 20:51
richbodyofavailableliteraturehasproventobeveryuseful
UTCuntilJuly210:49UTCforatotalexposuretimeof48.2
for the classification of our Chandra sources, and for sepa-
ks (ObsID 4510). The observation was done in very faint,
ratingcluster membersfromnon-members. Forthestudy of
timed exposure mode with a single-frame exposure time of
interactingbinariesitisapromisingtarget: beforethiswork
3.2 s. In order to achieve optimal sensitivity below ∼1 keV,
it was known to harbor two of only three spectroscopically-
we placed the central part of the cluster on the backside-
confirmedCVsfoundinopenclusters(Kaluznyetal.1997),
illuminated S3 chip; the neighboring S1, S2, S4, I2, and
and many optical-variability studies have revealed dozens of
I3 chips were used to cover the outer parts of the clus-
close binaries in the field of the NGC6791 (see de Marchi
ter. The cluster center lies at approximately α=19h20m53s,
et al. (2007) and references therein). It is therefore interest-
δ=+37◦46(cid:48)18(cid:48)(cid:48) (J2000;deMarchietal.2007);weshiftedthe
ingtoinvestigateiftheX-raysourcecontentofNGC6791is
center 0.(cid:48)5 away from the S3 aimpoint in the −Y (approxi-
asrichandvariedaswasfoundforthefewoldopenclusters
matelysouth-east)directionsothatalargerpartofthecluster
studied previously, viz. M67 (Belloni et al. 1993, 1998; van
couldbeimagedonasinglechip. Plataisetal.(2011)derive
denBergetal.2004)andNGC188(Bellonietal.1998;Gon-
doin2005). Theclusterliesatadistanceof∼4.0–4.3kpcand aclusterhalf-massradiusrh of4.(cid:48)42±0.(cid:48)02. Ourobservation
is reddened by E(B−V)=0.09–0.16; see e.g. Carraro et al. coversalmosttheentireareainsiderh ((cid:38) 98.5%)witheither
theS3ortheS2chip;onlythesouthernmosttipfallsoffthe
(2006),Basuetal.(2011),andBrogaardetal.(2011)forre-
detectorarea.Fig.1showstheoutlineoftheACISchipsover-
centdeterminationsoftheclusterparameters.Throughoutthe
layedonanopticalimageofthecluster.
paperweadoptadistanceof4.1kpcandaconstantreddening
E(B−V) = 0.14, which corresponds to a neutral-hydrogen Our data reduction started with the level-1 event file pro-
column density N = 7.8 × 1020 cm−2 (Predehl & Schmitt duced by version 7.6.7.2 of the Chandra X-ray Center pro-
H
cessing pipeline. We used the CIAO 3.4 package with the
1995).Thereddeningcouldvaryacrosstheareaofthecluster
(Plataisetal.2011;Brogaardetal.2012)buttheeffectistoo CALDB 3.3.0calibration filesfor furtherprocessing follow-
ingstandardprocedures2,includingthevery-faintmodeback-
small to have a significant impact on our results. NGC6791
ground cleaning. In order to improve the precision of the
stands out in several ways, including its high age and large
source positions we removed the randomization of event co-
mass. Remarkably for such an old cluster, the metallicity is
higherthansolar([Fe/H]≈+0.4). ordinatesthatisappliedinstandardprocessing. Towardsthe
endoftheobservation,thebackgroundlevelbetween0.3and
The X-ray and optical observations and data reduction are
7.0keVincreasedbyafactorof∼5asaresultofabackground
describedinSect.2,followedbythedataanalysisinSect.3.
flare. After removing this interval from the observation, the
Wedetectedvarioustypesofsourcesintheclusterwhichare
exposuretimeisreducedto42.9ks.
presentedinSect.4. Ourresultsarediscussedinthecontext
of the populations of interacting binaries in star clusters in
2.2. Opticalspectroscopy
Sect.5. WefinishwithourconclusionsinSect.6.
We obtained low-resolution spectra of candidate optical
counterparts to guide the classification of the X-ray sources.
2. OBSERVATIONSANDDATAREDUCTION
1http://cxc.harvard.edu/proposer/POG/html/ACIS.html
2.1. X-rayobservations 2http://cxc.cfa.harvard.edu/ciao3.4/guides/acisdata.html
AChandraX-raystudyofNGC6791 3
A total of 16 candidate counterparts brighter than V ≈ 18.3 tistical uncertainty in centroiding the spatial distribution of
were observed with the FAST long-slit spectrograph on the thedetectedeventsofagivensource,butdonotincludesys-
1.5-m Tillinghast telescope on Mt. Hopkins on 9 nights be- tematicerrorsthatresultfromdataprocessingandPSFasym-
tween 2005 June 7 to September 2. We used the 300 lines metries at large offset angles. To compute positional uncer-
mm−1 grating,resultinginawavelengthcoveragefrom3480 taintieswethereforeadoptEq.5inHongetal.(2005). This
to 7400 Å and a 3-Å resolution. Exposure time was chosen formularelatesthe95%confidenceradiusonthesourceposi-
toachieveasignal-to-noiseratioS/N (cid:38)20forV (cid:46)17.5,and tionr95tothewavdetectcountsandoffsetanglefromtheaim-
10 (cid:46) S/N (cid:46) 20 for fainter sources. FAST spectra were ex- point, andisbasedonextensivesimulateddetectionsofarti-
tractedandwavelength-calibratedwithadedicatedreduction ficialsources. Combinationofthebroad,soft,andhard-band
pipeline3. Flux standards were observed on the same nights source lists results in a master catalog of 86 distinct sources
asthesciencetargets. within8(cid:48) ofthe clustercenter, ofwhich 59lieinside rh. Ta-
Candidate optical counterparts fainter than V ≈ 17 were ble 1 summarizes their basic properties. To investigate the
observed with the fiber-fed multi-object spectrograph Hec- validity of the sources, we also ran wavdetect with a thresh-
tospeconthe6.5-mMulti-MirrorTelescope. Useofthe270 oldof10−7oranexpectednumberofspurioussourcesof1.6.
linesmm−1gratingresultedinspectrathatcover3700to9150 Thefourteensourcesnotdetectedinthisrunaremarkedwith
anasteriskinTable1. Inthispaperweadoptashortversion
Å with a resolution of 6 Å. A total of 16 candidate counter-
of the source names to refer to the sources instead of using
parts were observed on the nights of 2005 May 13 and July
their official CXO names; see columns 1 and 2 in Table 1.
4–6. Each setup was repeated 4 to 5 times with individual
Theshorthandnamesareassignedbyfirstsortingthesources
exposuresof900s. Exposureswerecombined,andthespec-
withinr onnetcounts,andthenthesourcesbetweenr and
tra extracted and wavelength-calibrated, with the Hectospec h h
reduction pipeline4. Sky background spectra for each setup 8(cid:48).
Weusedtheacis extractpackage(version3.107.2; Broos
were created by combining the spectra of fibers positioned
at off-source locations. Hectospec observations did not in- et al. 2002) to measure net source counts. Events between
0.3and7.0keVwereextractedfromaregioncorresponding
cludefluxstandards. Tocorrectfortheinstrumentalresponse
to 90% of the PSF at 1.5 keV; for a few sources this region
weconstructedsensitivitycurvesusingthespectraofthesub-
wasreducedtoavoidcontaminationbyacloseneighbor. The
dwarfB (sdB)starsB4, B5andB3in NGC6791 (Kaluzny
backgroundwasdeterminedfromasource-freeannuluscen-
&Udalski1992),includedintheMay13,July4andJuly5–
tered on the source position. We convert net count rates to
6 setups, respectively. We assumed that their intrinsic spec-
tracanbedescribedbyblackbodieswithtemperaturesasde- fluxeswithinSherpa5 witharfandrmfresponsefilesappro-
termined for these sdB stars by Liebert et al. (1994). The priateforthechiplocationandsource-extractionareaofeach
blackbodyspectrawerethennormalizedtoreproducetheob- source. Column 7 of Table 1 lists absorption-corrected (u)
servedV magnitudes(Stetsonetal.2003)aftercorrectionfor fluxes FX,u computed under the assumption that the under-
theclusterreddening. Thiswayweachievedanapproximate lying spectrum is an optically thin plasma (described by the
absolute-flux calibration of the target spectra after applying xsmekalmodel)withkT = 2keV.ForsourceCX29thatlies
theresultingsensitivitycurves. neartheaimpoint1counts−1 correspondsto6.8×10−12 erg
Flux-calibrated target spectra were assigned spectral types cm−2 s−1. This temperature is appropriate for the most ac-
bycomparisontostandardspectrawithsimilarresolution(e.g. tive ABs in the cluster (see e.g. van den Berg et al. 2004),
Jacobyetal.1984). while too high for the least active ABs and too low for
most CVs. For a 1-keV Mekal model, a power-law spec-
3. ANALYSIS trum (powlaw1d) with photon index Γ = 2, and a 10-keV
3.1. X-raysourcedetectionandextraction thermal-bremsstrahlungmodel(xsbremss)theconversionfac-
tor is 24% smaller, and 30% and 42% larger, respectively.
Werestricttheanalysistotheareawithin8(cid:48) ofthecluster
In each case we account for a column density equal to the
center. Sources further away have relatively large positional
cluster value using the xsphabs model. For the adopted dis-
errors, which complicates the optical identification. We fo-
tancetoNGC6791of4.1kpc, a3-countdetectionlimitcor-
cusontheareainsider whichisclosetothelargestcircular
h responds to a limiting unabsorbed X-ray luminosity L of
area around the cluster center that is entirely covered by the X,u
(0.7−1.4)×1030 ergs−1 cm−2 (0.3–7keV)wheretherange
observation.
correspondstothechoiceofmodelsspecifiedabove.
We performed source detection in a broad (0.3–7.0 keV),
Acis extract performs a Kolmogorov-Smirnov test on the
soft(0.3–2.0keV)andhard(2.0–7.0keV)energyband, also
eventarrivaltimestotestforvariability. Twosourcesarethus
usedinourChandrastudyofM67(vandenBergetal.2004),
found to be potentially variable, with probabilities that their
tofacilitatecomparison. TheCIAOdetectionroutinewavde-
count rates are constant (p ) smaller than 2.5%: CX19
tectwasrunforscalesof1.0to11.3pixels,instepsincreasing K−S
√ (a candidate CV, see Sect. 4.2) for which 15 of a total of
byafactor 2,withthelargerscalesappropriateforlargeoff- 19 counts are detected within 5.5 hours (p = 0.018),
K−S
axis angles where the point-spread function (PSF) becomes and CX37 (no optical counterpart found) for which all 9
significantly broader. We computed exposure maps for the counts are detected in the first 7 hours of the observation
responseat1keVtoaccountforspatialvariationsofthesen- (p =0.024).
K−S
sitivity. The wavdetect detection threshold was set to 10−6,
from which we expect two spurious detections per detection
3.2. X-rayspectralproperties
scale(sosixteenspuriousdetectionsintotal)intheareathat
Only CX1 has sufficient counts to allow a constraining
weconsiderhere. Wavdetectpositionalerrorsreflectthesta-
spectral fit. A spectrum was extracted from the acis extract
3http://tdc-www.harvard.edu/instruments/fast
4 http://tdc-www.harvard.edu/instruments/hectospec/ 5http://cxc.harvard.edu/sherpa
4 vandenBergetal.
Table1
Chandrasourceswithin8(cid:48)ofthecenterofNGC6791
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
CX CXOUJ α(J2000) δ(J2000) r95 Offset Counts FX,u E50 Opt
(deg) (deg) ((cid:48)(cid:48)) ((cid:48)) (ergcm−2s−1) (keV)
Sourcesinsidethehalf-massradiusrh
1 192044.9+374640 290.187175 +37.777870 0.32 1.64 213±15 318.0 1.3±0.1 +
2 192039.8+374354 290.165914 +37.731752 0.39 3.54 151±13 242.9 1.00±0.03 +
3 192035.7+374452 290.148996 +37.747852 0.39 3.70 125±12 260.1 1.4±0.1 +
4 192056.3+374613 290.234628 +37.770331 0.34 0.66 116±11 170.7 1.4±0.1 +
5 192047.3+374318 290.197161 +37.721768 0.43 3.20 72±9 118.0 1.4±0.1 −
6 192114.4+374530 290.310342 +37.758603 0.63 4.32 72±9 112.0 1.4±0.1 −
7 192052.3+374550 290.218110 +37.764148 0.35 0.47 53±8 77.9 1.2±0.1 +
8 192038.2+374441 290.159518 +37.744838 0.46 3.33 48±7b 221.7 1.6±0.2 +
9 192058.4+375008 290.243732 +37.835590 0.56 3.98 37±7 59.8 1.6±0.3 +
10 192037.3+374612 290.155726 +37.770073 0.43 3.09 34±6 69.3 1.4±0.2 −
Note. —Columns: 1)sourcenumber; 2)sourcename; 3and4)sourcecoordinatesindecimaldegreesafterapplyingtheboresightcorrectionof∆α =
−0(cid:48).(cid:48)06±0(cid:48).(cid:48)06,∆δ=−0(cid:48).(cid:48)21±0(cid:48).(cid:48)04(Chandraminusoptical);5)95%uncertaintyradiusonthesourceposition;6)angularoffsetfromtheclustercenter;7)net
countsinthe0.3–7keVband;8)unabsorbedflux(×10−16ergs−1cm−2)inthe0.3–7keVbandfortheassumptionthatthesourcespectrumisa2-keVMekal
plasmaseenthroughaneutral-hydrogencolumnofdensity7.8×1020cm−2;9)medianenergyinthe0.3–7keVband(onlyforsourceswith>5counts);10)flag
forthedetectionofacandidateopticalcounterpart.Table1isavailableinitsentiretyinamachine-readableformintheonlinejournal.Aportionisshownhere
forguidanceregardingitsformandcontent.
sourceregionwiththeCIAOtoolpsextract,andwasgrouped trometric reference frame of any given Chandra observation
tohaveatleast20countsbin−1towarrantuseoftheχ2statis- as a whole is aligned to the ICRS with a typical 0(cid:48).(cid:48)6 accu-
tic; the background contribution (<1 count) can be ignored. racy(90%uncertainty6).Wemeasuretheboresightcorrection
GiventheclassificationofCX1asanactivegalacticnucleus using matches with optical variables only. This minimizes
(AGN)basedonitsopticalspectrum(Sect.4.6),wetrytofit thenumberofchancealignmentsasthelightcurvesofmany
the data with an absorbed power-law and find an acceptable variablesrevealthemasclosebinariesandthereforeasplau-
result for a photon index Γ = 1.9±0.3 and a column den- sible X-ray emitters. A list of variables was compiled from
sity NH < 1.2×1021 cm−2 (1-σupperlimit; χ2 = 9.5, 6de- the studies of Kaluzny & Rucinski (1993), Rucinski et al.
greesoffreedom).ThelimitonNHisconsistentwiththeinte- (1996), Mochejska et al. (2002, 2003, 2005), Bruntt et al.
gratedGalacticcolumndensityinthedirectionofNGC6791 (2003), Hartman et al. (2005), and de Marchi et al. (2007);
(9×1020 cm−2,Schlegeletal.1998)whilethevalueforΓis we only selected variables that we could identify with ob-
typicalforanAGN(e.g.Tozzietal.2006). jects in the S03 catalog so that we could use the S03 posi-
Theremainingsourceshavetoofewcountsforusefulcon- tions for the boresighting. The boresight correction is com-
straintsbyspectralfitting.Insteadweusethemethodofquan- puted iteratively. We looked for matches to X-ray sources
tileanalysisdevelopedbyHongetal.(2004),wheretheme- withmorethan10netcountsandr ≤ 1(cid:48)(cid:48). Foreachsource,
95
dianenergy(E50)andthe25%and75%energyquartiles(E25 the match radius is set to be the quadratic sum of the error
andE75)ofthesourcephotonsareusedtocharacterizespec- on its X-ray position, and the typical 1-σ error on the opti-
tral properties. The advantage of using energy quantiles as calpositions(0(cid:48).(cid:48)27)scaledtoa95%errorradiusassuminga
opposedtocomparingcountsinpre-definedenergybandsby 2Dgaussianerrordistribution. Theboresightissettobethe
meansofhardnessratiosisthattheerrorsonthediagnostics weighted (with r−2) mean of the X-ray–optical offsets of all
95
arelesssensitivetotheunderlyingspectralshape matches found. Next, the X-ray positions are corrected for
the boresight, and the matching is repeated, now including
3.3. Opticalidentification thestatisticalerroronthemeanboresightinthematchradius.
Thisprocedurequicklyconvergestoaboresightcorrectionof
3.3.1. Cross-identificationagainsttheStetsoncatalog
∆α=−0(cid:48).(cid:48)06±0(cid:48).(cid:48)06,∆δ=−0(cid:48).(cid:48)21±0(cid:48).(cid:48)04(Chandraminusop-
Welookedforopticalcounterpartsinthedeep BVI photo-
tical)basedontenmatcheswithopticalvariables. Matching
metriccatalogofNGC6791compiledbyStetsonetal.(2003;
the boresight-corrected X-ray catalog to the clean S03 cata-
S03hereafter),whichcoverstheentireareastudiedhere. The
logresultsin51opticalmatches(including26variables)for
limiting magnitude is V ≈ 24 for the central area but the
47sources,outofthe86X-raysourcesintotal.
sensitivity is lower for the outer regions. In an attempt to
TheprobabilitytofindamatchintheS03catalogbychance
eliminateartifactsfromthecatalog,weremovedentrieswith
depends on the projected star density (which decreases with
photometric-quality indicators that flag them as suspicious, distance from the cluster center r) and on the match radius
viz. sources with |sharpness| > 1, and separation index < 0; (which typically increases with r because r increases). To
95
see S03 for an explanation of these indicators and a motiva-
estimatetheexpectednumberofspuriousmatches,wedivide
tionforthesecriteria. the cluster in a central area defined by r ≤ r , and an outer
h
Wefirstmeasureandcorrectfortheboresight,i.e.apossi- areaofr < r ≤ 8(cid:48). ThecleanS03cataloggivesaprojected
blesystematicoffsetbetweentheastrometricreferenceframes densityohf0.030opticalsourcesarcsec−2 and0.0092sources
of the Chandra and optical positions. The optical posi-
tionsaretiedtotheInternationalCelestialReferenceSystem
(ICRS)withanrmsaccuracyofabout0(cid:48).(cid:48)27(S03),buttheas- 6http://cxc.harvard.edu/cal/ASPECT/celmon/
AChandraX-raystudyofNGC6791 5
arcsec−2 for the central and outer area, respectively. If we
multiplythisbythetotalareacoveredbythematchcirclesof
the X-ray sources, we find that the expected number of ran-
dom matches is 5.9 (center), and 3.8 (outer annulus). This
amountsto15%and29%ofthematchesfoundintherespec-
tive regions. On the other hand, a similar calculation shows
thatallmatcheswithvariablestarsarelikelytobereal,with
the average number of chance alignments < 0.1 both for the
centralandouterregions.
In order not to overlook any matches, we repeated the
matching using the entire S03 catalog, and inspected the re-
gions around the Chandra sources in the optical fits image
from S03 to discard matches with image artifacts or dubi-
ous detections. We thus found nine extra candidate coun-
terparts for eight Chandra sources, which are not included
in the clean catalog because the values of the quality flags Figure2. ACS/WFCfindingcharts,eachmeasuring3(cid:48).(cid:48)4×3(cid:48).(cid:48)4insize,made
slightly exceeded the adopted cutoff limits, the object is re- fromthestackedGO-9815F814Wimagerepresenting7024sofexposure
timeintotal. ThecombinedX-ray/optical95%errorcirclesareshownas
allyextended,orbecausetheobjectisfaintandliesclosetoa
circles centered around the boresighted X-ray positions, while tick marks
relativelybrightstar.Sinceselectionoftheseadditionalcoun- indicatetheopticalmatches. Exceptforthebrightcandidatecounterpartto
terparts was not done in a very systematic way but relies on CX21ontheedgeoftheerrorcircle,thesematchesarefoundintheHSTdata
visual inspection, it is not trivial to estimate the number of only,andnotintheStetsonetal.(2003)catalog.Northisup,easttotheleft.
randomcoincidencesamongthenewmatches. Twonewcan- thesevenHSTcandidatecounterpartsarematchedtoChandra
didate counterparts are matched to a single Chandra source sourceswithanothercandidatecounterpart, weestimatethat
outside rh, so at least one of the two must be spurious. The anappreciablefractionofHSTmatchescouldberandom,on
other seven are uniquely matched to seven Chandra sources theorderof2/7(≈29%)orevenmore.Astheastrometrically-
insiderh.BasedonthehighersourcedensityintheentireS03 calibrated image is not readily available, we show the find-
catalogcomparedtothecleanS03catalog,onewouldexpect ing charts of these additional identifications in Fig. 2. This
tofind∼1.5extraspuriousmatchesinthecentralarea. Keep- bringsthefinaltallyoftheopticalidentificationto53candi-
inginmindthatsomeoftheseoptical“sources”arenotreal, datecounterpartsfor48ofthe57X-raysourcesinsider ,and
h
weestimatethatatmostoneortwoofthesevennewmatches 14 candidate counterparts for 12 of the 29 sources between
insiderharespurious. rh and 8(cid:48). The properties of the candidate counterparts are
In total the S03 optical survey provides 60 astrometric summarizedinTables2and3,andtheirlocationsintheopti-
matchesfor55sources. calcolor-magnitudediagrams(CMDs)areshowninFigures3
and4. Moreinformationonthelight-curvepropertiesofthe
3.3.2. HSTimaging
opticalvariables(column10ofTable2)canbefoundbycon-
Bedin et al. (2006) used Hubble Space Telescope (HST) sultingtheoriginalreferences. Variableswithnameslistedin
multi-epochimagingwiththeWideFieldChannelontheAd- theformatnnnnn *werefirstdiscoveredbydeMarchietal.
vanced Camera for Surveys (ACS) to measure proper mo- (2007); Table1indeMarchietal.(2007)givesanoverview
tions of objects in a central 3.(cid:48)4 × 3.(cid:48)4 region of NGC6791. ofthediscoverypapersforvariableswithnamesstartingwith
We make use of their results to establish cluster member- a’V’or’B’,althoughupdatedlightcurvescansometimesbe
ship of candidate counterparts included in the ACS images foundinmorerecentpapersmentionedinthattable.
(Sect. 4.1), and refer to Bedin et al. (2006) for details of
the data and the proper-motion analysis. We also use these 3.3.3. Sourceswithmultiplecounterparts
deep images to look for additional faint candidate counter-
For seven X-ray sources we find two candidate counter-
parts. The data set consists of F606W and F814W images
parts. In four cases—CX27, CX57, CX78, CX84—one of
takenon2003July16and17(GO-9815),andon2005Jul13
the two is an optical variable, and given the low probability
(GO-10471). Anastrometricreferenceframewascreatedby
for a chance coincidence we consider the variables to be the
combiningtheGO-9815F814WimageswiththeSTScIMul-
truematches. ThevariablesalllieclosertotheX-raysource
tidrizzle software, which removes artifacts like cosmic rays
than the alternative match. We also tentatively list the clos-
and bad pixels, and corrects for the geometric distortion of
estmatchtotheotherthreeX-raysourcesasthelikelycoun-
ACS images. In order to align the coordinate system to the
terpart in Table 2, but more information is needed to firmly
ICRS,wecomputedanastrometricsolutionbasedontheposi-
establish which, if any, of the two is the true match. For
tionsof273unsaturatedstarsfromtheS03catalogthatcould
completeness we list the properties of the alternative identi-
beidentifiedinthestackedimage.Fittingforzeropoint,plate
ficationsinTable3. CX34islikelyanextra-galacticsource,
scale,androtationresultedinasolutionwherethermsresidu-
asbothHSTcounterpartsareextended. Theclosestmatchto
alsofindividualstarsare0(cid:48).(cid:48)035inrightascensionand0(cid:48).(cid:48)032
CX21 is a faint point source that is unrelated to the cluster;
indeclination; thisisnegligiblecomparedtotheX-rayposi-
itsHSTpropermotionistypicalforabackgroundgalaxy.The
tional errors. The boresighted X-ray positions and error cir-
othermatchisaproper-motionmemberonthemainsequence
cleswereoverlayedonthisimage,andphotometryandproper
(see also Fig. 2). CX82 matches with two bright (V < 15)
motionsfortheastrometricmatcheswereextractedfromthe
stars.Thenearestisaproper-motionnon-memberandalikely
sourcecatalogcreatedfromtheentireHSTdataset. Wefind
foregroundKdwarf.
sevennewcandidatecounterpartsforsixX-raysources. The
new matches are all likely extra-galactic, given their proper
3.4. Sourceclassification
motion or extended morphology. From the fact that two of
6 vandenBergetal.
Con72655856493628251918 868179776857a54504441393330...23221715943 CX(1)
tin
u11 1 1 1 1 1 1 1 111 1
edon3270460259134436591067510837562721166697 477311853737746556173971278800123907011588306113626...1111269526526371005093152893 OID(2)
n
extpa1.311.110.350.070.230.900.560.240.400.13 1.930.790.211.200.920.200.390.280.620.230.381.300.31...0.210.750.260.160.340.090.23 ()(cid:48)(cid:48)OXd(3)
g
e
99 9
............45............... 9999909,m9,m776,mm9999999999...9899999999m... µp(4)
2212222222 1121111211111 1111122
2.9:3.058.380.830.820.142.714.093.713.99 7.557.741.067.967.659.279.912.727.898.277.209.577.96...7.137.807.867.276.512.700.64 V(5)
...0.4:−0.931.291.470.53...0.4:0.212.4:− 1.040.971.451.35...1.151.071.310.931.060.891.041.15...0.990.940.271.151.210.240.82− BV−(6)
0.5:0.540.931.691.790.920.382.121.301.45 1.030.961.931.391.531.321.352.481.011.120.941.201.29...1.071.020.261.411.220.880.63 VI−(7)
(1
0
5.49.01.11.11.83.23.83.85.95.9 3.53.52.74.512.71.11.41.72.22.32.63.34.8...4.65.46.47.712.034.352.3 erg30Xu,L(8)
So s−
u 1
rc ) O
0.1:...−0.10.10067010±2.60.3...−±1.70.3017249−±1.50.2...−±1.50.2...−±0.40.2...−±0.20.1...±0.20.1...±0.30.1...± eswithuncertainornomemb 2.50.3V12−±2.40.2V7−±1.20.2...−±2.20.2V59−±1.90.10143110−±2.30.3V41−±1.90.3...−±0.70.2...−±2.50.2V80−±2.40.2V76−±2.80.2V5−±1.70.2V42−±2.80.2V17−±......2.50.1V100−±2.20.1V16−±2.10.1B7−±2.240.09V9−±2.350.08...−±0.580.04062899±0.060.04B8−± Proper-motionclusterm XuVlog()Var/FF(9)(10) pticalpropertiesofcandidateTable2
.........1.613.................. ershipinform 1.5230.488...13.8337.6400.482......4.8864.0920.3130.5066.366or23.94712.5222.266...3.187......... embers (days)P(11) counterparts
............midK,Hemissionα.........BalmerandHeIIemission... ation midGearlymidG/...midK,weakHemissionα......K;filled-inHαlateKearlyM;Hemissionα/...lateG,earlyKmidG...lateG,filled-inH?α...lateGearlyK/...Balmerabs.lineswithems.cearlymidK,filled-inHα/early-Kgiantemissionlines... Opticalspectrum(12)
o
re
s
FBFBAB?ABAB?FBFBFBCV?FB ABABABSSGSSGABAB?ABABSSGABABSSG...ABABCVSSGRGCVCV Clas(13)
s
...long-termvariableonmainsequencespottedvariableonbinarymainseque............... eclipsingbinaryeclipsingbinaryonbinarymainseque...spottedbinaryeclipsingorspottedbonbinarymainVI−onbinarymainsequeeclipsingbinary...contactbinaryspottedbinary......spottedvariable?eclipsingorspottedb...eclipsingbinary......... Comments(14)
nce nce inaryseq.nce inary
AChandraX-raystudyofNGC6791 7
(13)(14)ClassComments EG...ABeclipsingbinarySirregularvariableEG...Slong-periodvariableEGextendedEG?propermotionconsistent...withbackgroundgalaxyABforegroundcontactbinaryEGextendedABforegroundcontactbinaryS...EG...EGextendedS...S...EGextendedStooredtobeamemberEGextendedS...EGextendedABeclipsingbinaryS...EGextendedEGextendedEG?extendedABspottedvariable?ABforegroundcontactbinaryS...S...Slong-termvariable magesor(forCX55only)byvisualinspectionnumericvaluesarebasedontheproper-motionmbers”)and“nm”(“(non)-members”)arebased−heF606WmagnitudeandtheF606WF814WV0.3–7keV)tobandfluxesaftercorrectingfor=CVandCV?(candidate)cataclysmicvariable,==onofbinarity,EGextra-galacticsource,FB(1993),Rucinskietal.(1996),Mochejskaetal.
Table2(Continued) (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)−−dpVBVVILFFP/CXOIDlog()VarOpticalspectrumVµ,OXXuXu(cid:48)(cid:48)−301()(10ergs)(days) Sourceslikelynotassociatedwiththecluster −±≈z.153610.195920.360.650.76...0.090.03......AGN,116−±α238860.14016.221.091.35...1.860.04V331.173midK,Hemission−±778780.17015.391.211.49...2.680.06V19...late-G,early-Kforegrounddwarf−±≈z.834720.17...20.220.240.69...0.300.07......AGN,236−±12129010.92016.061.231.38...2.650.09V6650.498early-Kforegrounddwarf>13...0.25...............2.1.........a±21...0.08nm23.96...1.12...0.20.1.............................................−±2462710.292716.320.891.01...2.90.1V10.268...±26...0.22...28.5...1.5...1.90.2.........a−±27113760.04015.970.660.81...3.00.1V60.279...−±2973280.18013.600.871.26...4.00.2......midK−±≈z.31112120.25...22.270.510.75...0.60.2......AGN,153a>34...0.33...............1.6.........−±3545710.28014.621.381.35...3.70.2......mid-Kgiant−±3875340.1211,nm21.041.572.29...1.20.2.........±42...0.15...26.6...0.7...1.00.3.........−±4336290.789822.111.752.94...0.60.2.........−±≈z.4543530.37018.981.451.57...1.70.2......0092−±46100880.21011.541.101.10...5.10.2......mid-Gforegroundgiant−−±47109670.32nm23.880.22.09...0.20.2.........−±4824260.66019.531.041.15...1.80.2V110.883earlyG/−±51120370.41014.010.710.74...4.20.2......late-Fearly-Gdwarf±52...0.18...27.68...2.32...1.20.3.........−±5387050.20nm23.70.31.3...0.40.3.........b−±55...0.80...18.76.........2.40.3.........−±76148332.37016.481.001.05...2.90.2V548.314mid-G,early-Kforegrounddwarfa−±785532.58016.851.031.10...2.70.2V230.272...−±8010312.05020.360.850.94...1.30.2.........a−±82134040.03014.470.870.89...3.90.2......early-Kforegrounddwarfa−±10......84154341.49021.2:...2.9:...1.20.301225 aThissourcehasanothercandidatecounterpart,seeTable3.b(cid:48)gThecounterpartisnotincludedintheS03catalog.Thelistedvalueisthemagnitudeintheband(Plataisetal.2011). /NoteHST.—Columns:1)sourcenumber;2)numberofcandidateopticalcounterpartintheS03catalog,sourceswithoutanumberwerefoundintheACSiaftercorrectingforboresightoftheS03clusterimage;3)angularseparationbetweentheX-rayandopticalsource;4)probabilityforclustermembership,wherestudiesofPlataisetal.(2011)(thecaseofCX3isdescribedinmoredetailinSect.4.2)andK.Cudworthetal.(privatecommunication),andthelabels“m”(“me/HSTHSTontheACSstudydescribedinBedinetal.(2006);5–7)opticalmagnitudesandcolorsfromS03,exceptforthe(unnumbered)counterpartswherewelistt−301colorifavailable;8)unabsorbedX-rayluminosity(0.3–7keV)inunitsof10ergsassumingadistanceof4.1kpcandthefluxesinTable1;9)ratioofX-ray(extinction;10)nameoftheassociatedopticalvariable;11)periodofvariability;12)spectroscopicpropertiesofthecandidateopticalcounterpart;13)sourceclass:====ABandAB?likelyorcandidateactivebinary,SSGsub-subgiant,RGredgiant,SlikelyforegroundstarnotassociatedwiththeclusterwithoutanyindicatiFaintobjecttotheBlueofthemainsequence;14)comments.Informationonopticalvariabilitywasgatheredfromthefollowingreferences:Kaluzny&Rucinski(2002,2003,2005),Brunttetal.(2003),Hartmanetal.(2005),deMarchietal.(2007),Brogaardetal.(2011)
8 vandenBergetal.
Figure3. (V,B−V)and(V,V−I)color-magnitudediagramsofNGC6791withthecandidateopticalcounterpartsthatareassociated,orpossiblyassociated,
withtheclustermarkedinred.Candidatecounterpartsthatareproper-motionmembersareindicatedwithlargefilledsymbols,andsourceswithoutmembership
informationwithlargeopensymbolsorcrosses:diamondsare(candidate)CVs,circlesare(candidate)ABs,trianglesaresub-subgiants,squaresarestarswithout
anyindicationofbinarity,andcrossesmarkunclassifiedfaintsourcesawayfromthemainsequence. StarsinthefieldofNGC6791areplottedasdots,while
starswithaproper-motionmembershipprobability>50%(K.Cudworth,privatecommunication)areshownassmallfilledblackcircles.Theinsetszoominon
thecrowdedregionsofthediagrams.PhotometryisfromStetsonetal.(2003).SeetheelectroniceditionoftheJournalforacolorversionofthisfigure.
light curve constrain the nature and period of a close bi-
Table3
nary. Proper-motion information, and to some extent opti-
Alternativeopticalcounterparts
cal colors, can establish cluster membership and puts limits
onthedistance. Todistinguishbetweenclustermembersand
CX OID dOX V B−V V−I comment non-memberswemainlyrelyontheproper-motionstudyby
((cid:48)(cid:48)) Plataisetal.(2011),whichincludes41ofourcandidatecoun-
21 11098 0.78 18.48 0.93 0.97 pµ=90 terparts. Fornineadditionalcandidatecounterpartsmember-
27 11365 0.68 16.58 0.76 0.94 pµ=0,Fstar ship could be established from the HST study by Bedin et
34 ... 0.68 ... ... ... extended al.(2006,seeSect.3.3.2)andfromtheresultsoftheproper-
57 7395 0.76 21.02 1.47 1.96 non-member,Kstar
motionstudybyK.Cudworth(privatecommunication).
78 569 3.12 16.50 0.95 1.03 pµ=0 WeusetheenergyquantilestoconstraintheunderlyingX-
82 13395 0.70 14.66 0.96 0.98 ...
rayspectrumforsourceswithmorethan15netcounts(0.3–7
84 15442 3.01 20.64 ... 1.7 ...
keV). For fainter sources the errors on the quantiles are too
Note.—SeeTable2fortheclosestastrometricopticalmatchtotheX-ray
largetomeaningfullysayanythingabouttheirspectra. Quan-
source,andforthemeaningofthecolumns.
tilecolor-colordiagrams(QCCDs)areshowninFig.5. Each
panelshowstwotypesofmodelgrids,whichindicatetheex-
Many factors contribute to the classification of the X-ray pectedlocationsofasourceintheQCCDforemissionfrom
sources. An extended morphology of the candidate opti- aMekalplasma(blue/yellow)andforapower-lawspectrum
calcounterpartseparatesextra-galacticfromgalacticsources. (black/gray). Theformerisappropriatefortheemissionfrom
Based on optical spectra one can immediately distinguish hotcoronaearoundactivestarsforwhichtheplasmatemper-
between AGN, accreting binaries with substantial accretion ature kT can reach up to several keV. CVs can be anywhere
disks, and ordinary stars. Period and shape of the optical
AChandraX-raystudyofNGC6791 9
Atotalof30sourcesarenot,orlikelynot,associatedwith
thecluster. Thesearestarsandbinarieswithpropermotions
that clearly set them apart from the members, background
galaxies,andstarswhoseveryredcolorsclassifythemasun-
likely members. They are shown in Fig. 4 and briefly dis-
cussedinSects.4.5,4.6,and4.7.
For ten candidate counterparts membership information is
lacking or inconclusive. These sources include (candidate)
ABs, a new candidate CV, and faint optical sources that lie
to the blue of the main sequence; the latter are discussed in
Sect.4.7. OpensymbolsandcrossesmarktheminFig.3.
In the following we discuss the sources by type of X-ray
emitter. Unless stated otherwise, X-ray luminosities quoted
inthetextrefertothe0.3–7keVband.
4.2. Cataclysmicvariables
We have detected three CVs that belong to the cluster and
discovered one CV candidate without membership informa-
tion. Forallfour,theX-raycolorsandluminositiesareinthe
expected rangefor CVs. While the X-ray–to-opticalflux ra-
tios in Table 2 are also typical for CVs, their values can be
misleading: theX-rayandopticaldataarenotcontemporane-
Figure4. Color-magnitudediagramofNGC6791withcandidatecounter- ous,whileCVsmayshowlargevariationsinbrightnessona
partsthatarelikelynotassociatedwiththeclustermarkedinred.Downward- timescaleofweekstomonths.
pointingtrianglesarebackgroundgalaxies,circlesare(candidate)ABs,and CX4 is matched to the faint cluster member and optical
squaresarelikelyforegroundstarswithoutanyindicationofbinarity. HST
variable 06289 9 that lies to the blue of the cluster main se-
counterparts,forwhichphotometryinVandIisnotavailable,arenotshown.
SeetheelectroniceditionoftheJournalforacolorversionofthisfigure. quence. Based on its optical colors and the detection of an
.
outburst-likeevent,deMarchietal.(2007)alreadysuggested
between soft or very hard; the spectrum of a typical dwarf thatthisstarisadwarfnova.ToourknowledgetheHectospec
novahaskT ≈2−10keV(e.g.Bycklingetal.2010). Power- spectrum in Fig. 6, which shows the Balmer lines clearly in
lawspectraareappropriatefortheharderemissionfromAGN emission,isthefirstspectroscopicconfirmationofitsCVna-
(Γ=1−2).GridsarecomputedwithSherpausingtheenergy ture. The X-ray quantiles suggest that the X-ray emission
response of the source closest to the aimpoint. Variations in arises in a kT ≈ 4 − 10 keV plasma (Fig. 5). Thus L in
X
theshapeofthegridsasfunctionofchippositionaretypically Table2, whichwascomputedfortheassumptionofa2-keV
smallerthantheerrorsonthemeasurements,andareignored. plasma,canbeupto∼40%toolow(seeSect.3.2).
The ratio of the fluxes in the X-ray and optical (V) band, ThenewcandidateCVistheX-rayvariableCX19,which
or limits thereon, can help to classify a source, even in the is matched to a faint blue object with a Hectospec spectrum
absence of an optical match down to the limit of the obser-
that shows HeII 4686 Å and Hβ emission lines (Fig. 6),
vations. We compute this ratio as follows: log(F /F ) =
X opt u and hints of Hα and Hδ in emission. The X-ray luminosity
log(Fx)u + (V − AV)/2.5 + 5.44. The zeropoint for the V- (L ≈ 6×1030 ergs−1),thehardspectrumassuggestedby
bandfluxistakenfromBesseletal.(1998). Thetotalextinc- X,u
its quantile values (Fig. 5), and the high (F /F ) ratio, are
tion A was assumed to be equal to the cluster value, which X V
V all consistent with a classification as CV. HeII 4686 Å is of
underestimates (overestimates) the flux ratio for foreground
comparablestrengthasHβ. Thisisseeninhighmass-transfer
(background) objects. Typically, stars and active binaries
rate systems (nova-like CVs), and in CVs containing white
havelog(F /F ) (cid:46) −1,whileAGN,CVsorotheraccreting
X V u dwarfswithmuchstrongermagneticfields(i.e.(cid:38)1MG)than
binaries with unevolved late-type companions, and very ac-
in dwarf novae systems. If we assume cluster membership
tivelate-typedwarfscanhavehigherratios(e.g.Stockeetal.
wefindanabsolutemagnitude M ≈ 10.2, whichfavorsthe
1991). V
explanationasamagneticCV.
CX3 and CX17 are the known, spectroscopically-
4. RESULTS
confirmed CVs B8 and B7 (Kaluzny et al. 1997). With
4.1. Clustermembersversusnon-members L ≈ 5.2×1031 erg s−1, B8 is the brightest X-ray source
X,u
Twentycandidatecounterpartsarepropermotionmembers. in the cluster. The actual value of L is likely higher, as the
X
Of these, all but one (CX9) show signs of binarity. On this X-ray quantiles suggest a plasma temperature (∼4 keV) that
basisweconsideratleast19ofthemasthetruecounterparts is higher than our nominal value. B8 has been classified as
totheChandrasources. Wederivethebinarystatusandtype aSUUMadwarfnovabasedonthedetectionofseveralout-
ofbinaryfromtheopticalspectra,theopticallightcurves,ora bursts of 1–2 mag in amplitude, and a recent superoutburst
locationintheCMDonthebinarymainsequence,indicating (∼3 mag) in the Kepler light curve (Mochejska et al. 2003;
thatanunresolved multiplesystemisresponsiblefor thede- Garnavich et al. 2011). We include B8 among the cluster
tectedlight. TheclusterbinariesareamixofCVs,ABs,and members. Plataisetal.(2011)estimateda10%membership
binariesbelowthesub-giantbranch,andarediscussedfurther probability from proper motion, but closer inspection of the
inSects.4.2, 4.3, and4.4, respectively. CX9isdiscussedin data reveals that a few detections of this star were contam-
Sect.4.5. Clustermembersaremarkedwithfilledsymbolsin inated by a fainter close neighbor. When the affected data
theCMDsofFig.3. pointsarenotconsidered,thepropermotionofB8isconsis-
10 vandenBergetal.
Figure5. Quantilecolor-colordiagramswithmodelgridsrepresentingaMekalplasma(blue/yellow)andapower-lawspectrum(black/grey). Foracertain
choiceofmodel,theplasmatemperatureorspectralindex,andthecolumndensitycanbeestimatedfromthelocationofasourceinsidethegrid. Themedian
energyE50canbereadofffromthetopx-axis. SymbolsareasinFigures3and4,whileblackfilledcirclesmarksourceswithoutopticalcounterparts. Dueto
theirdifferentenergyresponses,sourcesonthebackground-illuminated(BI)S3chipandforeground-illuminated(FI)S2andS4chipsareshowninseparate
panels.Weincludesourceswithmorethan15netcounts(0.3–7keV);errorbarsareshownonlyforthesourceswiththehighestandlowestnumberofcounts.
tentwithclustermembership. troscopicbinariesinthe6.5-GyroldopenclusterNGC188it
CX17 or B7 is a solid cluster member. The X-ray quan- is expected that, for main-sequence binaries, orbits up to at
tiles indicate that kT ≈ 2 keV. Most reports in the literature least15dayshavebeencircularizedattheageofNGC6791
find B7 to be relatively bright with V ≈ 18, which suggests (Mathieuetal.2004). Therefore,sincetidalsynchronization
a high mass-transfer rate; occasional small (0.5–1 mag) out- operatesonashortertimescalethantidalcircularization(Hut
bursts, and a drop in brightness of ∼3 mag in V have also 1981; Zahn 1989), main-sequence binaries with periods be-
beenseen. Mochejskaetal.(2003)proposethatB7isanova- low15daysarealsoexpectedtobetidallylocked. Allofour
like variable of the VY Scl type, or perhaps a ZCam dwarf main-sequenceABswithmeasuredphotometricperiodshave
nova mostly seen during periods of standstill between mini- periods shorter than 4.9 d; in Sect. 5.2 we discuss how this
mum and maximum brightness. Our Hectospec spectrum is couldbeduetotheperiodversusX-rayluminosityrelationof
qualitativelysimilartothespectruminKaluznyetal.(1997), ABsandourdetectionsensitivity. CX50,CX54,andCX79
andlikelytakenwhenthemass-transferratewashighandthe arenotidentifiedwithknownvariables,buttheylieonthebi-
disk optically thick: the spectrum shows broad Balmer ab- nary main sequence or have Hα in emission in their spectra,
sorptionlineswithnarrowemissioncores,withtheemission whichisasignatureofenhancedmagneticactivity. Wenote
componentdominantinHα(Fig.6). that in the V versus B−V CMD, CX54 lies slightly to the
TheX-ray–to–opticalfluxratiosandblueopticalcolorsof blueofthemainsequence. Thisisnotexpectedforthecom-
CX18, CX25, CX28, CX36, CX65, and CX72 resemble binedlightoftwomain-sequencestars,soitsclassificationas
thoseofCVs, andthesesourcesareprimetargetsforoptical ABislesssecure. Perhapsthisstarisanon-memberafterall,
follow-upspectroscopy. withapropermotionthatissimilartothatoftheclusterstars.
Radial-velocitymeasurementscanprovidemoreclarity.
4.3. Activebinaries Stars that have evolved past the turnoff, i.e. have evolved
Weclassifyatotaloftwentysourcesaslikelyorcandidate to larger radii, can in principle circularize wider orbits than
ABs. Thanks to the extensive photometric variability stud- main-sequence stars within a given time span as the circu-
iesofNGC6791,theorbitalperiodsofmanyofthesecanbe larization time scale is very sensitive to radius. For the sub-
inferred from the light curves, either through eclipses, ellip- giant AB CX23 the eccentricity is unknown, and estimates
soidalvariations,orspotactivity.Amongtheclustermembers of the binary period are only known from photometric vari-
wefindelevenABs,includingeightwithphotometricperiods ability. Brunttetal.(2003)quoteaperiodof∼12.5d, while
(Table 2). Most have colors that place them near the main Mochejskaetal.(2005)findnearlytwicethatvalue,∼23.9d.
sequenceor,inthecaseoftheWUMacontactbinaryCX39, FromtheVandB−Vmeasurementsoftheopticalcounterpart,
rightattheturnoff.ForCX86,the1-sigmaerrorsonB−Vand weestimatethattheradiusoftheprimaryisatmost∼2.8R(cid:12)
V −I aresuchthatitcouldlieanywherefromjustbelowthe (Flower1996); asbothcomponentsofthebinarylikelycon-
turnofftothebaseofthered-giantbranch.CX23liessecurely tribute to the measured optical light, the actual radius of the
above the sub-giant branch, and we discuss it separately be- primaryissmaller. Ifwenowlookatthediagnosticdiagram
low. NGC6791isatleast8Gyrold;bycomparisontotheor- in Verbunt & Phinney (1995, Fig. 5a; appropriate for a pri-
bitalperiodversuseccentricitydistributionofsolar-typespec- mary mass of 1.25 M(cid:12), i.e. somewhat larger than the turnoff