Table Of Content

MNRAS000,1–12(2015) Preprint12January2016 CompiledusingMNRASLATEXstylefilev3.0 An extensive radial velocity survey toward NGC 6253 M. Montalto,1⋆ C. H. F. Melo,2 N. C. Santos,1,3 D. Queloz,4 G. Piotto,5 S. Desidera,6 L. R. Bedin,6 Y. Momany,6 I. Saviane2 1Institutode Astrof´ısica e Ciências do Espaço, Universidade do Porto, CAUP, Rua das Estrelas, PT4150-762 Porto, Portugal 2European Southern Observatory, Alonso de Cordova 3107, Vitacura Casilla 19001, Santiago 19, Chile 3Departamento de Fisica e Astronomia, Faculdade de Ciências, Universidade do Porto, Portugal 6 4Observatoire de Genéve,51 Ch. des Mailettes,1290 Sauverny, Switzerland 1 5Dipartimento di Fisica e Astronomia Galileo Galilei, Universita´di Padova, Vicolo dellOsservatorio3, Padova IT-35122 0 6INAF - Osservatorio Astronomico di Padova, Vicolo dellOsservatorio5, Padova, IT-35122 2 n a AcceptedXXX.ReceivedYYY;inoriginalformZZZ J 0 1 ABSTRACT The old and metal rich open cluster NGC 6253 was observed with the FLAMES ] multi-object spectrograph during an extensive radial velocity campaign monitoring P 317 stars with a median of 15 epochs per object. All the targeted stars are located E along the upper main sequence of the cluster between 14.8 < V < 16.5. Fifty nine . h starsareconfirmedcluster membersboth byradialvelocitiesandpropermotions and p do not show evidence of variability. We detected 45 variable stars among which 25 - belongtoNGC6253.Wewereabletoderiveanorbitalsolutionfor4clustermembers o (and for 2 field stars) yielding minimum masses in between ∼90 MJ and ∼460 MJ r t and periods between 3 and 220 days. Simulations demonstrated that this survey was s a sensitivetoobjectsdownto30MJ at10daysorbitalperiodswithadetectionefficiency [ equal to 50%. On the basis of these results we concluded that the observedfrequency of binaries down to the hydrogen burning limit and up to 20 days orbital period is 1 around (1.5±1.3)% in NGC 6253. The overall observed frequency of binaries around v the sample of cluster stars is (13±3)%. The median radial velocity precision achieved 6 bytheGIRAFFEspectrographinthismagnituderangewasaround∼240ms−1(∼180 5 ms−1 for UVES). Based on a limited follow-up analysis of 7 stars in our sample with 2 2 theHARPSspectrographwedeterminedthataprecisionof35ms−1canbereachedin 0 thismagnituderange,offeringthepossibilitytofurtherextendthevariabilityanalysis . into the substellar domain. Prospects are even more favourable once considering the 1 upcoming ESPRESSO spectrographat VLT. 0 6 Key words: open clusters: general – open clusters: individual NGC 6253 1 : v i X 1 INTRODUCTION al. (2012) where C,N,O abundances and carbon isotope ra- r a tiosforfourredclumpstarsofNGC6253wereanalyzedand N33G5◦.C5,6b25=3 (α6◦2.030)0is=a1n6hol5d9man0d5sm,δe2t0a0l0-ri=ch−o5p2e◦n42cl′u3s0t′e′,rllo=- in Cummings et al. (2012), focused on lithium abundance. − In the context of extrasolar planet searches NGC 6253 catedinsidethesolarringandprojectedtowardarichGalac- plays a major role. Given its high metallicity and the well tic stellar field in the direction of the Galactic center. The known fact that the frequency of jupiter planets found metal-rich nature of this cluster and its importance in the aroundfieldstarsisknowntostronglycorrelatewithmetal- contextofGalacticstudieswereearlyrecognizedinthesem- licity (e.g. Gonzalez 1997; Santos et al. 2001; Santos et inalworkofBragagliaetal.(1997)andlaterondiscussedby al. 2004; Fischer & Valenti 2005; Mortier et al. 2012) we several other authors in the past (Piatti et al. 1998; Sagar, mayexpectalargepopulationoftheseobjectstobepresent Munari, & de Boer 2001; Twarog, Anthony-Twarog & De inthiscluster,unlessotherevolutionaryandenvironmental Lee 2003; Anthony-Twarog, Twarog, & Mayer 2007; Car- factors alter theirformation histories. retta et al. 2000; Carretta, Bragaglia & Gratton 2007; Ses- Our previous efforts to detect this population and to tito et al. 2007; Anthony-Twarog et al. 2010). Novel chem- better characterize cluster properties were presented in a ical studies have been recently presented in Mikolaitis et sequenceofworksstartingwithMontaltoetal.(2009,hereafter M09) where we delivered a photometric and astro- ⋆ E-mail:[email protected] metric catalog including proper motions membership prob- (cid:13)c 2015TheAuthors 2 M. Montalto et al. abilities. De Marchi et al. (2010) studied the photometric cluster members, and an important step toward the detec- variability properties of cluster members and thesurround- tionoflowmassobjects.Thepreparationoftheobservations ing field producing a catalog of 595 variables, 35 of which presented in this work preceeded in time the photometric wereproposedasclustermembers.InMontaltoetal.(2011, andastrometricanalysisdiscussedinM11.Sampleselection hereafter M11) we presented the results of a radial veloc- was based on the work of Twarog, Anthony-Twarog & De ity survey conducted with the UVES/GIRAFFE spectro- Lee(2003).Themagnituderangeofthesamplestarsiscom- graphsat VLT,mostlyconcentratedontheevolvedportion prised in between 14.8< V <16.5. All the targeted stars lie ofthecolor-magnitudediagram.Inthatwork,wepresented on theuppermain sequenceof NGC 6253. a follow-up analysis of three planetary candidateswe found InSection 2,we describetheobservationsweacquired. inourphotometricstudiesandhighlightthediscoveryofthe In Section 3, we discuss data reduction. In Section 4, we firstclusterdoublelinedeclipsing binarysystem,located in present the methods used to identify variables and in Sec- the turn-off region (star 45368 in our catalog) which mem- tion 5 our classification criteria. In Section 6, we describe bershipwasconfirmedbothbypropermotionsandbyradial theanalysisofasampleof spectroscopicbinaries.InSec.7, velocities. We analyzed a sample of 139 stars in the cluster wecompareourresultswiththeonesobtainedduringprevi- region concludingthat35wherelikelyclustermembersand ous surveys.InSection 8, we calculate thesurveydetection 12 likely close binary systems. efficiency.FinallyinSection9,wesummarizeandconclude. InMontaltoetal(2012),weperformedequivalentwidth analysis and derived Fe, Si, Ca, Ti, Cr and Ni abundances as well as abundance ratios of a main-sequence 2 OBSERVATIONS star, twored-clumpstars andabluestraggler clustermembers. For our main-sequence star, we obtained a metal- Theobservationsdescribedinthisworkhavebeenacquired licity of [Fe/H]=+0.26 0.11 (rms), whereas for the two with FLAMES (Pasquini et al. 2002) and HARPS (Mayor, ± giants we found that our metallicities were on average M.etal.2003)spectrographs.FLAMESisthemulti-object, [Fe/H]=+0.19 0.13(rms),lowerthanwhatwasdetermined intermediate and high resolution spectrograph of the VLT ± in previousstudies(Sestitoet al. 2007, [Fe/H]=+0.36 0.07 installed at the UT2 (Kueyen telescope), in Paranal, Chile. ± rms, Carretta et al. 2007, [Fe/H]=+0.46 0.03 rms). FLAMESisacomplexsystemthatfeedstwodifferentspec- ± NGC-6253 fundamental parameters are not yet firmly trographs, UVES and GIRAFFE. While UVES provides set. It is well known that the different calibrated photome- the maximum resolution (R = 47 000), but can access up triespresentedsofar donotagree well, andthatsignificant to eight targets at a time, GIRAFFE has an intermedi- differencesexistamongthem.Theclusteragewassetto3.5 ate resolution permitting to target up to 132 objects at Gyr in M09 and the reddening to E(B V)=0.15. These thetimeortoperformintegralfieldspectroscopy.FLAMES − estimates were respectively the upper bound age and the data were obtained in two different observing seasons be- lowerboundreddeningderivedfromisochronefitsofar,once tween April- July, 2004 and March-July 2005 respectively. compared with the other results in the literature. However For UVES we used the standard setup centered at 580nm. recentlyRozyczkaetal.(2014)presentedadetailedanalysis For GIRAFFE we used different high resolution settings: of the eclipsing binary 45368 (renamed in their work V15). HR8, HR9B, HR11, HR12, HR13, HR14A, HR14B overall Their results indicate that the age of NGC 6253 should be covering the wavelength range between 491.7nm and 670.1 comprisedbetween3.80-4.25Gyrfromthemass-radiusdi- nmwitharesolutioninbetweenR=17740andR=28800.In agram and should be even older (3.9 - 4.6 Gyr) from color- total 79 epochs were acquired obtaining 6558 spectra (553 magnitude diagram (CMD) fitting. They also report a red- withUVESand5955withGIRAFFE)correspondingto317 dening equal to E(B V)=0.113 mag, and therefore lower stars. In order to allow high accuracy in the radial velocity − thanourownvalue.ThesameauthorspresentedinKaluzny measurements for both spectrographs simultaneous Th-Ar et al. (2014) theresultsof anovelphotometriccampaign in lamps observations were acquired together with the scien- the cluster region, conducted with the 1.0m Swope Tele- tifictargets.However,for21outofthe79GIRAFFEplates scope in Las Campanas, focusing primarily on the bright no simultaneous calibrations were acquired. For the rest of portion of the CMD, in a domain only partially covered by the plates 5 fibers were allocated to the calibration lamps, ourpreviousphotometricsurveys.Amongasampleof25ad- while for UVES 1 fiber was always allocated to the simu- ditional variablestheauthorsdetectedthreenoveleclipsing taneous calibration lamp. For both spectrographs no fibers binaries members of the cluster. The analysis of these ob- were allocated tothe sky. jectswilllikelypermittoreachmorestringentconstraintson HARPS is the High Accuracy Radial velocity Planet clusterpropertiesandtofurthercomplementtheresultsob- Searcher at the ESO La Silla 3.6m telescope. We used the tainedfor45368.Ithasbeenalsodemonstratedthattheoret- High Efficiency mode (EGGS). HARPSdata havebeen ob- icalmodels failtoaccurately reproducetheobservedCMD, tained between 28-30 May, 2011. The observations were particularly for evolved stars (M09, Anthony-Twarog 2010, much more limited with respect to the FLAMES runs and Rozyczkaet al. 2014). in general obtained under non optimal conditions. We ob- In this work, we present the results of another exten- served a total of 7 stars and a maximum of 5 epochs sive radial velocity campaign that was performed on this per object, resulting in 20 measurements. These targets clusterinthepastyears.Thissurveysuperseedsbyfarboth were selected among the sample of stars earlier observed innumberofmonitoredobjectsandinnumberofepochsthe with UVES/GIRAFFE with the purpose to follow-up po- analysispresentedinM11andalsothesimilaronediscussed tential planetary candidates. The early analysis of the in Anthony-Twarog (2010). It represents a more ambitious UVES/GIRAFFE data relied upon strong assumptions on efforttobettercharacterizeradialvelocityvariabilityamong theprecisionofthemeasurements,whichwasclaimedatthe MNRAS000,1–12(2015) An extensive radial velocity survey toward NGC 6253 3 levelof40-50m/s.Furthermore,suchanalysisneglectedthe was calculated with the DER SNR algorithm presented in presence of potential and very poorly understood sources Stoehr et al. (2008). Using the ESO exposure time calcula- of systematics that could have compromised the accuracy tor we found a reasonable agreement between the observed of the measurements. In this work, we will study in detail S/N and theexpected one. this problem, presenting for the first time a complete and troughout analysis of theentiredataset. Table 1 offers an overview of all theobserving runs. 3.1 Systematics analysis 3.1.1 GIRAFFE 3 DATA ANALYSIS Beforemergingthedatasetswecheckedandcorrectforsys- tematiceffects.FirstweanalyzedtheGIRAFFEdataset.We UVES data were reduced using the REFLEX (Freudling et selected a sample of reference stars. To dothat we imposed al. 2013) UVES-FIBER workflow. GIRAFFE data were re- an RMS treshold limit in radial velocity RMS< 1 kms−1. ducedusingGASGANO.ForHARPSdataweretrievedfrom Wefurtherlimited thesampletoonlystarswith atleast 10 theESOarchivetheScienceDataProductsviathePhase3 measurements resulting in a list of 161 stars. We checked spectral queryform1. then for plate to plate zero point offsets in radial velocities After the data reduction step we cross-correlate all the as illustrated in Fig.2 (top panel). This Figure represents FLAMES spectra with a reference spectrum. Because all of the average subtracted radial velocities (RVs) for all refer- the monitored objects were located on the upper main se- encestarGIRAFFEmeasurements,plottedasafunctionof quenceoftheCMD(seebelow)weadoptedasolarspectrum the plate number. A clear pattern of systematic variations asatemplate.InparticularweusedtheatlasofFLAMESso- is present which amounts to maximum shifts of up to 1 larspectra2.Thissetofmeasurementscompletelycoversall km s−1. To correct for this effect, we calculated a robu∼st high-resolution and low-resolution settings of the FLAMES average of the residual RVs for each plate and subtracted fibersystem.Wetookcaretoselectthespectracorrespond- this value for all the measurements of that plate. In Fig.3 ing to our scientific setups as described in the previus Sec- (left panel) we show the histograms of the residuals RVs. tion.Foreachgiveninstrumentandsettingwealsoconstruct We fit a Gaussian function to these distributions to judge amodelforthetelluriccontamination.Thiswasdonebyus- theimprovementoftheresidualscatterduringthedifferent ing the ESO sky correction tools and in particular the Sky stepsofpost-processing.Inthetoppaneltheobservedresid- Model Calculator3. The model spectra were analyzed and uals(beforeanycorrection)arevisible.Thedispersionofthe the regions with the strongest contamination were then ex- best-fitGaussian isequalto631ms−1.Inthesecondpanel cluded from the analysis of the scientific spectra. For the fromthetoptheeffectoftheplatetoplatecorrectionisvis- cross-correlation analysis we developed ourown tools. First ible.Suchacorrectionsubstantiallyimprovetheprecisionof anaccuratemodelofthecontinuumwascreated,byaverag- themeasurementsbringingthedispersionto212ms−1 and ing the spectra in 20 spectral subregions, strongly down- it is therefore of paramount importance. The fiber system, weighting absorption and emission features. Then a five- however, may be subjected also to more subtle mechanical order polynomial was fit and used to normalize the spec- drifts which may depend on a variety of factors. Such sys- trum. We calculated the cross-correlation in between -200 tematic drifts may in turn induce intra-plate systematics. kms−1and200kms−1instepsof1kms−1.Thecentroidpo- To check for that we decorrelate the radial velocity residu- sitionwascalculatedadoptingagaussianfunctionusingthe als against the fiber positioning number. The result of this upper 30% of the CCF measured from the peak. From the correctionisshowninthethirdpanelfromthetopinFig.3. CCF function we also calculated the bisector as described The corresponding dispersion improved further being equal in Queloz et al. (2001). The obtained radial velocities were to 197 m s−1 after this correction demostrating that intra- corrected to put them in the heliocentric reference system platesystematicsareindeedpresent,albeittheirmagnitude and the heliocentric julian dates at mid-exposure were cal- seem to be smaller than the plate to plate systematics. It culated with the task rvcorrect of IRAF. We then applied is possible that the resulting radial velocities are affected a quality cut, dropping from the list of measurements all also byvariableobservingconditionsduringeach night.We those for which the CCF peak was found lower than 0.3 or checked for this effect by regrouping all measurements as a theassociated error in theradial velocity was larger than 5 functionoftheobservingnightanddecorrelatingtheresidu- kms−1orforwhichtheresultingspectrahadaverylowS/N alsagainstairmass (calculatedforeachstar),color(derived (tipicallylowerthan3).Asampleof5334GIRAFFEand445 from our catalog) and allowing for a systematic zero point UVESmeasurementsremainedafterthesestepscorrespond- offset. No improvementis observed in this case (thedisper- ing to a total number of 313 stars. All the stars monitored sionisequalto199ms−1).Wedecidedinanycasetoapply with UVES were monitored also with GIRAFFE. The me- this epoch-to-epoch correction to the GIRAFFE data. The diannumberofmeasurementsperstarinthefinalcatalogis distributions in Fig.3 appear close to Gaussian. The distri- equalto15.Fig.1,presentstheS/Nofthespectra.TheS/N bution of the observed residuals in the top panel appears slightly asymmetric. Because the GIRAFFE measurements have been ac- 1 http://archive.eso.org/wdb/wdb/adp/ quiredwith 7different setups,we checkedalso for thepres- phase3 spectral/form?phase3 collection=HARPS 2 http://www.eso.org/observing/dfo/quality/GIRAFFE/ enceofsetup-dependentsystematics.InFig.4,weregrouped all the residual radial velocities as a function of the used pipeline/solar.html 3 http://www.eso.org/observing/etc/bin/gen/ setting.Nosignificant offsetisobservedafter thecorrection form?INS.MODE=swspectr+INS.NAME=SKYCALC stepsreportedabove.However,wedidnotethefactthatthe MNRAS000,1–12(2015) 4 M. Montalto et al. different setups do not appear to achieve the same perfor- GIRAFFEmeasurementswasadoptedastheradialvelocity mances. The setting HR09B delivered the highest precision reference system to correct the UVES measurements. With (155 ms−1 as measured from the RMS of the residuals), a median value of 15 measurements per star with an RMS while HR12 was found the lowest performant (287 ms−1). down toaround200 ms−1,and an overlap of 3-4references This is likely due to the strongest telluric contamination perUVESplatetheprecisionofsuchareferencesystemcan present in the spectral window of this setting. It appears be expected at the level of 20 ms−1. Fig.2, illustrates the therefore that beyond the systematics discussed above, the plate-to-plate systematics of the UVES spectrograph with choice of the instrumental setup is of great importance to respecttosuchareferencesystem.Oncecomparedwiththe achievethehighest radial velocity precisions. GIRAFFEobservedresiduals(illustratedinthesamefigure) The internal error of the individual measurements has itappearsthatthethetwospectrographsdidnotfollowthe been initially assigned by our Gaussian fitting algorithm same pattern of systematics. The upper and lower UVES once performing the CCF analysis, being the error of the spectra delivered instead very similar results. Importantly, centroidpositionobtainedbytheLevemberg-Marquardtal- wecanclearlyseethepresenceofanoffsetbetweentheradial gorithm.Afterthecorrectionforsystematicswethencalcu- velocityscaleofthetwospectrographs.Weselectedthenthe lated the ratio of the median RMS to the median internal starsincommonbetweentheUVESandGIRAFFEdatasets error for UVES and GIRAFFE separately. The individual whichwereconsideredGIRAFFEreferencestars(asdefined errors were then multiplied by this factor. This approach above) and having also at least 5 UVES measurements. By allows to match the average error levels based on the RMS using this sample (in total 34 stars) we calculated, for each analysis, but preserves the relative errors among different star, the difference between the average radial velocity cal- measurements. culatedusingonlyGIRAFFEmeasurementsandonlyUVES It is interesting to compare the results obtained by using measurements (this time considered in the UVES reference the plates which were observed with simultaneous calibra- system). torsornot.Inparticulartheplateswithoutcalibratorswere Thehistogramofthesedifferencesareillustratedinthe those obtained with theHR14B setting. A look at Fig.4 in- bottom panels of Fig.2. Both the upper and lower spectra dicates that the final RMS obtained for this setting is in of UVES deliver offset radial velocities with respect to GI- generalnotworser theoneobtainedwithothersettings.By RAFFE.The average offset is found equalto: comparinginsteadthedistributionoftheobservedresiduals (withoutanypost-correctionapplied,thereforetheresultof thepipeline) we obtainedfor thissettingadispersion equal RVUVES−RVGIRAFFE =(−0.87±0.15)kms−1. (1) to 690 m s−1, worser than the 631 m s−1 quoted above for Subsequently we compared the radial velocity scale of the theentiresample.Usingonlytheplateswithcalibrators we two spectrographs with the one of HARPS (for the stars obtain 632 m s−1,very close tothevalueof theentiresam- in common as reported below), and we concluded that the ple,becauseplateswith calibrators aredominant.Itresults measurements in the reference system of GIRAFFE are in therefore that the correction applied by the pipeline based thereference system of HARPSwith nonoticable offset. on the simultaneous calibrators does not improve substan- Fig.3reportstheresultsobtainedapplyingtheplate-to-plate tially the precision, especially in comparison with the final correction toUVESdata(middleandrightpanels).Alsoin valuesthatweobtainedbymeansofouranalysis.Inarecent thiscaseweobserveasubstantialimprovementofthepreci- work Malavota et al. (2015) illustrated that thewavelength sion.ThedispersionofthebestGaussianfittotheobserved calibration solutionmaybeinaccurateinthefirstplacedue distribution is in this case 598 m s−1 (average of the upper to the presence of non linear distorsion terms between the andlowerUVESspectra),whileaftertheplate-to-platecor- wavelength solution of the morning calibration and the si- rection the dispersion is equal to 156 m s−1. Thanks to the multaneous calibrations. This in turn can produceartificial adoption of the GIRAFFE reference system we were then zero points offsets. It may be also possible that the simul- abletocheckalsoforintra-platesystematicsforUVES.Note taneous calibrators are not tracking perfectly the radial ve- thatsuchacorrectionisnotpossibleusingonlythesimulta- locityzeropointsoffsetsduetothepresenceofadifferential neousfibers,giventhatonlyonesimultaneousfiberperplate offset between the calibrators and the scientific fibers. In- is allowed for UVES. The adoption of this correction im- accurate fiber scrambling could be a possible reason. If the provedfurtherthedispersion (119ms−1)althoughwenote light is not uniformly spread across the fiber, spurious ra- that the corresponding residuals distribution presents evi- dial velocity variations can be produced, as a consequence dent non-Gaussian tails. Likely intra-plate systematics are of thermal, pressure or other environmental variables (e. g. present although the smaller number of stars used for the seeing) variations. Our analysis seems to indicate that the correction maynotbeabletoperfectly capturethesystem- plate-to-platesystematicsarethemostprominentsourcesof atic trend.. We proceeded further and analyzed theepoch- noise. to-epoch UVES systematics, decorellating all residuals ac- quiredduringthesameobservingnightasafunctionofcolor, 3.1.2 UVES airmass and allowing for thepresenceof azero point offset. Inthiscasesuchacorrection appearstodeteriorate(127m TheFLAMES-UVESdetectoristhemosaicoftwodifferent s−1) theprecisions, and we therefore did not apply it. CCDs coveringthebluerand redderportion of thespectral After the above reported correction steps, the upper and rangeofthissetup.Thepipelineprovidesanoutputforeach lower UVES spectra were merged, and the results finally oneofthemandweanalyzedthemindependently.Inthefol- merged to the one of GIRAFFE to produce a single com- lowingwewillindicatethemasthelower(L)andupper(U) bined catalog. For what has been said, the radial velocity UVESspectra.Theaveragevalueofthesystematiccorrected scale of this catalog is the one of GIRAFFE (that is the MNRAS000,1–12(2015) An extensive radial velocity survey toward NGC 6253 5 one of HARPS, see below). The GIRAFFE and combined usedtheGeneralizedLombScargle(GLS)algorithmofZech- catalogs contain 313 stars in total, while theUVES catalog meister&Ku¯rster(2009)andinparticulartheirEq.5.This contains 75 stars. The total number of stars with at least algorithm allows to fit a sinusoid to the data together with two measurements is equal to 300 stars for GIRAFFE(and a constant offset. To properly set the detection treshold of combined catalog) and 67 stars for UVES. the GLS we constructed a mock sample of constant artifi- Fig.5 shows the final RMS obtained for the GIRAFFE, cial radial velocities. The measurements for each star were UVESandcombinedsampleforallthestarshavingatleast simulatedassumingagaussiannoisewithadispersionequal two measurements respectively and with RMS<1 kms−1 toσref.Weconsideredthetruedistributionofepochsofour (250starsforGIRAFFEandcombinedcatalogsand61stars stars, each simulated star being a constant mock copy of a for UVES catalog). real observed one. We created around 105 simulated stars The median precision (denoted by the vertical dashed and applied the GLS to all of them. We searched for pe- lines in Fig.5) of GIRAFFE is 240 m s−1 while the one of riodic signals considering periods in between 0.5 and 1000 UVESis180ms−1.ThecombinedsamplemedianRMS(250 dayssubdividingthisintervalin10000equalfrequencysteps. ms−1)reflectstheGIRAFFEsampleRMSduetothemuch From the distribution of GLS powers (pGLS) we obtained, larger number of measurements acquired with GIRAFFE. weimposedaFalseAlarmProbability(FAP)tresholdequal Fig.6,representsthehistogramofalltheradialvelocitymea- to0.1%, which was found equal topGLS,FAP=0.956. surementsinthefinalcatalog.Thepeakduetoclustermem- IntheexactlysamewayweappliedtheGLStoallrealmea- bers is clearly visible. surementsflaggingoutallobjectshavingpGLS > pGLS,FAP. We found that this condition was met for 28 objects. We considered however as reliable GLS variables only those for 3.1.3 HARPS which an orbital solution could be found as described in Sect.6.MostoftheGLSsolutionsimpliedlongtermtrends The HARPS radial velocities measurements were obtained or in any case the phase coverage was not considered suf- directly from the headerof thepipeline processed files, and ficient to perform a fit. All of these objects lie beyond the no post-correction was applied. This sample is composed 5σref treshold and are therefore also RMS variables. by a set of 7 stars that were all included in the UVES- ThefinallistofreliableGLSvariablescontains6objects GIRAFFE dataset observed in 2004-2005. While four of whicharedenotedbythegreendotsinFigure7andFigure8. theseobjects appearedpossibleplanetarycandidatesat the Thetotal list of variables contains 45 stars. time of the follow-up, the HARPS observations have dis- missed all of them as false positives. Onthecontrary,we notethatthetransitingplanetary can- didate (star 171895) we found in our previous photometric 5 CLASSIFICATION OF THE STARS campaigns, described in M11, has not been targeted with HARPSanditremainsthereforeaprimarytargetforfuture We proceeded by calculating the average radial velocity of observational efforts. thecluster.Todothatweusedboththepropermotionprob- To estimate the average precision obtained by HARPS on abilities derived in M09 and the radial velocities calculated this sample of stars we subtracted the mean radial velocity inthiswork.Wefirstconsideredthoseobjectsforwhichthe and calculated the RMS of all combined residuals. In this propermotionmembershipprobabilitywaslargerthan90% way we obtained a valueof 35 m s−1. atmagnitudeV =12.5andlargerthan50%atV =18,inter- BycomparingtheaveragevelocityobtainedwithGIRAFFE polating linearly between these extremes. Furthermore, we andwithHARPSforthesestarsweobtainedanaveragedif- impose thatthecorrespondingx and ycoordinates of these ference between the two spectrographs radial velocity scale stars should have been contained within 800 < x < 1300, equal to (-0.094 0.091) kms−1, which demonstrates that 1000 < y < 3000 in the reference system of CCD51 of the ± theradial velocity scale ofGIRAFFEisconsistent with the WideFieldImagerdetectorwheretheclusterwaslocatedin one of HARPS. M09.Beyondtheselimitsweconsideredourpropermotions doubtful as reported in earlier works. These tresholds are the same adopted in the past to isolate cluster members. We also limited the sample to stars that were not flagged 4 SEARCH FOR VARIABLES out as variables accordingly to the analysis performed in the previous Section. In this way we identify 66 stars. The Figure 7 shows the RMS distribution for all the catalog stars. We adopted a magnitude dependent treshold for the average cluster radial velocity RVcl we obtained is detection of variables as visible in Fig. 8. This RMS treshold (σref) was calculated by medianing the RMS values of RVcl =( 28.81 0.003)kms−1 (2) the stars found in 0.2 mag bins and fitting a linear model − ± totheresultingmedianvalues.Weadopteda5σref treshold where the error quoted is the error of the mean. The RMS denotedbythedotted linein theabovementionedFigures. was σcl=1.05 km s−1. To obtain this result we used 1531 All objects beyond this treshold were considered bona-fide individualmeasurements. variables and are denoted in particular by thebluedots. We proceeded as follows to classify all our stars into likely We also searched for periodic variables given the extent of clustermembersandlikelyfieldstars.Clustermemberswere our survey and the fact that many objects have multiple identifiedasstarswhich propermotion probabilitywassat- measurements in our catalog. In particular we limited the isfying the constraints reported above and which average search for periodicities to stars with at least 10 epochs. We radial velocity was in between 3σcl from thecluster mean ± MNRAS000,1–12(2015) 6 M. Montalto et al. average velocity 4. In absence of proper motions (or in the thisinitialsolutionisrecordedandcomparedwiththeχ2 new case proper motions were considered doubtful) we consid- obtained in the following step. The following step is ob- ered as bona fide members those stars satisfying only the tained jumping from the initial position to another one in radial velocity criterium. The total sample of cluster stars the multidimensional parameter space randomly selecting is equal to 192 stars, 25 of which were flagged out as vari- one of the free parameters and changing its value by an ables. Consequently, the sample of field stars contains 121 arbitrary amount which is dependent on a jump constant stars, 20 of which are variables. On the basis of this classi- and theuncertainty σ of the parameter itself. Stepsare ac- fication we obtain an observed frequency of variables equal cepted or rejected accordingly to the Metropolis-Hastings to 13% for the cluster, and equal to 17% for the field. We criterium. If χ2 is lower than χ2 the step is executed, new old also obtain that the overall field contamination is around otherwise the execution probability is P = e−∆χ2/2 where 38%.Somecaveatsshouldbekeptinmindhowever.Among ∆χ2 =χ2 χ2 . In this latter situation a random num- new− old allclustermembersonly59satisfy thestringentconstraints ber between 0 and 1 is drawn from a uniform probability on proper motions and radial velocity, the remaining ones distribution. If this number is lower than P then the step are only radial velocity members. Therefore a certain field is executed, otherwise the step is rejected and the previous contamination can be expected from field stars sharing the step is repeated instead in the chain. In any case the value samevelocityofclustermembers.Equivalently,amongfield oftheχ2 ofthelaststepisrecordedandcomparedwiththe variables some fail to satisfy the radial velocity criterium oneofthefollowing stepuptotheendofthechain.Wead- but have high proper motion probabilities. Since they are justed thejump constants (one for each parameter) in such variables, their average radial velocity is not necessarily co- a way that the step acceptance rate for all the parameters incident with the systemic velocity. Figure 9 and Figure 10 was around 25 per cent. We then excluded the first 20 per showthecolormagnitudediagramsforclusterandfieldstars cent steps of each chain to avoid the initial burn-in phase, respectively.Inthebackgroundwedisplayedalsopropermo- andforeachparameterwemergedtheremainingpartofthe tion likely members from our catalog in the first case and chains together. Then we derived the mode of the resulting allotherobjectsinthesecond.Fromthesefiguresitisclear distributions,andthe68.3percentconfidencelimitsdefined that field stars are much more dispersed in the CMD, and bythe15.85thandthe84.15thpercentilesinthecumulative that cluster members closely follow the main sequence of distributions. the cluster. Blue and green dots have the same meaning The radial velocity measurements after subtraction of the than previous figures, denoting respectively RMS and GLS barycentricvelocities,alongwiththebestfitmodel,bisector variables. Red open dots denoteconstant stars. diagram and theperiodogram are shown in Fig. 12-Fig. 17. Table2andTable3showthecompletelist ofvariables Our best-fit parameters are given in Table 4. The best-fit along with theirbasic parameters. models correspond to values of the reduced χ squared between χ2 =0.8 2.5. p r − Thebisectorerrorwascalculated fromthedispersionofthe 6 ANALYSIS OF PERIODIC VARIABLES distributionofbisectorvaluesofallthespectra,andassumed identicalforallstars.Wealsocheckedforlinearcorrelations For six variables we were able to derive an orbital solution. between the bisector and radial velocity measurements cal- TheperiodwasfixedtotheGLSperiodcorrespondingtothe culatingthePearsoncorrelationcoefficient.Forthesixcases peak with the strongest power. We modeled the Keplerian consideredthecoefficient iscomprisedinbetween -0.47and motion as: 0.56 which denoted that nostrong correlations are found. Forthosevariablesthatwereconsideredclustermembers(4 RV = K˜ cosu+k +γ (3) out of 6 as shown in Table 4) we also calculated the mass √1 h2 k2 function and, by assuming for the primary the mass (M∗) − − obtainedbyisochronefit5 wereportinTable4alowerlimit where K˜ is the radial velocity semi-amplitude without the on the minimum mass of the companion which is given by contribution of the eccentricity e, k = e cosω, h = e sinω, thefollowing equation γ is the barycentric radial velocity and u = ν +ω is the true argument of latitude with ν the true anomaly and ω theargument of thepericenter. msini> M2K˜3P 13 (4) Weconsideredasfreeparameters:K˜,h,k,thetimeofmaxi- (cid:16) ∗2πG(cid:17) mumradialvelocity(t )andγ.Theconvergencetoward the value given by the expression on the right represents MAX thebest-fitsolutionwasobtainedbymeansofaLevenberg- a lower limit on the minimum mass because we neglected Marquardt algorithm (Press 1992), and the uncertainties thetermonthemassratioq=m/M∗,whichcontributesas and the best fit values of the parameters by means of a (1+q)2/3tothenumeratorofthetermontheright.Theval- Markov Chain Monte Carlo analysis. For each radial veloc- ueswe obtainedarein between 90MJ and 460MJ.The ∼ ∼ ity curve we created 20 chains of 105 steps. Each chain is periodsinstead are inbetween 3daysand 220 days.We ∼ ∼ started from a point 5-σ away (in one randomly selected notethattheattributionoftheexactperiodisstillambigu- free parameter) from the best-fitting solution obtained by ousforsomeobjects.Amongthissampleofobjectsstar478 the Levenberg-Marquardt algorithm. The χ2 of the fit of seems to have a non negligible eccentricity e=(0.16 0.03). old ± For star 29531 instead the χ of the fit is 2.5 and this may r suggest thepresence of an additional companion. 4 Notethat forstarsforwhichwewereabletoderiveanorbital solutionthesystemicvelocitywasconsideredratherthantheav- eragevelocity 5 Weconsideredthe3.5GyrisochronethatwasdiscussedinM09. MNRAS000,1–12(2015) An extensive radial velocity survey toward NGC 6253 7 7 COMPARISON WITH PREVIOUS SURVEYS Star8247(7303inAnthony-Twarogetal.2010)wasconsid- eredasalikelymemberclustervariablebytheauthorswhich So far four main surveys have been performed toward gave an RMS equal to 16.4 kms−1. In our data this object NGC 6253 to search for variables. As reported in theintro- is instead not variable, but we note that our mean radial duciton De Marchi et al. (2010) and Kaluzny et al. (2014) velocity differs from the one of the authors by 13 kms−1. searched for photometric variables. Anthony-Twarog et This object is likely a long period cluster variable. al. (2010) and M11 searched for radial velocity variables. On the contrary star 7470 in Anthony-Twarog et al. (2010, We observed the detached eclipsing binary V23 (39883 in which does not havean entry in our own catalog), is in the our catalog) reported in Kaluzny et al. (2014). This object listofourlikelyclustervariablesinTable2,havinganRMS is a proper motion member of NGC 6253 and sits close to equalto4.519kms−1.Anthony-Twarogetal.(2010)didnot the turn-off. We obtained only 1 measurement on it which consider this object as a variable, but their RMS is similar gave aradial velocity equal to(-52.96 0.18) kms−1. to us (3.24 kms−1), quite larger than their average level, ± The star 16649 was classified in De Marchi et al. (2010) as although probably at thelimit of detectability. alongperiodvariable,likelyclustermemberandlocatedat around 7 arcmin from the cluster center. It showed a photometric variability at the 1% level. This object has been extensively observed here. A set of 32 measurements have 8 DISCUSSION been acquired. It results that star 16649 is also a radial ve- The survey we presented in this work is one of the deepest locity variable and it is listed in Table 2 as a likely cluster searchesforbinariesinanoldopenclustereverperformedso membervariable.Themeanradialvelocityis-28.446kms−1 far.Millimanetal.(2014),basedontheirongoinglong-term andtheRMSis7.7kms−1.Membershipisestabilishedsolely radialvelocitymonitoringoftheoldopenclusterNGC6819 on the basis of radial velocity. The GLS algorithm flagged concludedthatthefractionofbinarieswithperiodslessthan out this star as a variable, but at the nominal period (∼63 104 daysisequalto22% 3%.Ourtotalfrequencyislower days) it presentsa verypoor phasecoverage and no orbital ± as reported below, but it is also based on a 2 year survey solution was possible. These observations support the idea whileMillimanetal.(2014)aremonitoringNGC6819since that this object is a long period binary system. It is cu- 17years.Binariessurveysinthefield(albeitincludingavari- rious to note that in fact, this object was also observed etyofdetectionmethods)suggestthatthetotalmultiplicity by Anthony-Twarog et al. (2010) during the Hydra radial fractionaroundfieldstarsisaround54%(e.g.Raghavanet velocity survey on NGC6253. Also in that case star 16649 al. 2010). (correspondingtostar 7592 in theirnumeration) was found Tounderstandourdetectionlimitsandbinaryfractions a radial velocity variable star. The authors report a mean we performed a set of simulations. Similarly to what was radial velocityequalto-27.38 kms−1 and anRMS equalto doneinSect.4,wecreatedamocksampleofmeasurements 10 kms−1 measured out of a set of 3 measurements. foreachobservedstar,assumingthebaselinenoiselevelσref The interesting star 55053 was also listed in De Marchi et andinjectinganartificialsignalintothedatatomimickthe al. (2010) as a long period variable, likely cluster member. presenceofacompanionstar.Werandomizedboththephase Itislocatedataround5.2arcminfromtheclustercenter.In and the inclination of the orbits and considered 25 bins in our study we obtained a set of 15 radial velocity measure- true mass (between 20 and 500 MJ) and an orbital period mentsonit.Interestingly,thisobjectisnotaradialvelocity equaleitherto10daysorto200days.Wethereforeapplied variable, at least down to 0.4 km s−1, the value we calcu- the5σref tresholdcalculatingthedetectionefficiencyineach lated for the RMS. In our study it is classified as a radial massbin.Figure11,showsthedetectionefficiencycurvesfor velocity non variable cluster member. Membership is esta- the two periods considered as a function of the companion bilishedonthebasisofradialvelocitygiventhattheaverage mass. Those curves are essentially average efficiencies over velocity is -28.613 kms−1. the whole sample of stars monitored. The simulations indi- No matches were found with our previous radial velocity cate that at a period of 10 days, objects with masses down surveypresentedinMontaltoetal.(2011),becausetherewe to 30MJcouldhavebeendetectedwitha 50%efficiency. focused on brighterobjects. At∼90%confidencelimitweexpecttobeabl∼etodetectstars The overlap with the Anthony-Twarog et al. (2010) survey with at least 0.15 M⊙. Considering a period of 200 days, isinsteadsubstantial.Wecountedatotalof32starsincom- a 50% detection efficiency corresponds to 0.1 M⊙, and a ∼ monamongthetwosurveyswhichgivesustheopportunity 90% detection efficiency to 0.4 M⊙. ∼ to further check for long-term variability for these objects. Looking at the results we obtained in Tab. 4, we con- Comparingourradialvelocitieswiththeonesoftheauthors cludethatessentiallyforallthedetectedobjects,thedetec- we found in general an excellent agreement. tion efficiency should be beyond 80%. The only exception Star21659 isan interesting variablethat was discoveredby is star 30324 which has a long period (223 days) and a low Anthony-Twaroget al. (2010). Theauthorsreport an RMS minimummass(92MJ)andmayindeedlieina 50% effi- equal to 19.5 kms−1 for this object (their star 7495) and ciency region. If the true mass of this object is c∼lose to the cosidered it among their list of five likely cluster variable limit indicated by its minimum mass, this would indicate stars. In our study this object was classified as well as a thatacorrectingfactoratleastequalto2shouldbeapplied clustervariable.Inadditiontothatitisamongthelistofob- to derive the true frequency of stars with long period com- jectsforwhichwederivedtheorbitalsolution,asreportedin panions. This may also be supported by the fact that for Table 4. The radial velocity semi-amplitude we found (41.8 most ofthevariableobjectswedetectedit was notpossible kms−1) is consistent with the RMS reported by Anthony- toderive an orbital solution. Twarog et al. (2010). Considering therefore only the sample of close-in cluster MNRAS000,1–12(2015) 8 M. Montalto et al. members variables for which an orbital solution was found, from FCT through thegrant and SFRH/BDP/71230/2010. we conclude that the frequency of binaries down to the hy- This wark is based on observations made with ESO Tele- drogen burning limit and with periods up to 20 days is scopes at the La Silla Paranal Observatory under pro- ∼ around (1.5 1.3)%, that is 3 out 200 members, in the up- gramme ID 073.C-0251, 075.C-0245 and on data products ± permain sequenceofNGC6253, whilethetotalbinaryfre- fromobservationsmadewithESOTelescopesattheLaSilla quencyis equal to (13 3)%. The errors were obtained con- ParanalObservatoryunderprogrammeID087.C-0497. The ± sidering a binomial distribution. anonymous referee is also aknowledged for the useful com- Comparatively, in M11 we obtained a frequency of binaries mentsandsuggestionswhichhelpustofurtherimprovethis equal to (29 9)%. These binaries were all flagged out with manuscript. ± a 5σ detection treshold like in this work and were consid- eredlikely shortperiod binariesgiventhattheobservations spanned only a few days. The sample of cluster stars (35) REFERENCES wasmuchsmaller thantheoneanalyzedinthiswork which givesalargeruncertaintyfortheestimatedfrequency.Even Anthony-Twarog,B.J.,Twarog,B.A.&MayerL.2007,AJ,133, accounting for that, it appears however that the frequency 1585 among evolved stars is larger (at a 2-σ level) than the to- Anthony-Twarog, B.J., Deliyannis,C. P.,Twarog, B. A.,Cum- tal frequency of binaries on the main sequencewe obtained mings,J.D.,Maderak,R.M.2010, AJ,139,2034 Bragaglia, A., Tessicini, G., Tosi, M., Marconi, G., Munari, U. above. 1997,MNRAS,284,477 Carretta, E., Bragaglia, A., Tosi, M., Marconi, G. 2000, ASPC, 198,273 9 CONCLUSIONS Carretta, E., Bragaglia, A. & Gratton, R. G. 2007, A&A, 473, 129 Wedescribedanextensiveradialvelocitysurveytowardthe Cummings, J. D., Deliyannis, C. P., Anthony-Twarog, B., old andmetal rich open clusterNGC 6253. Thesurvey was Twarog,B.,Maderak,R.M.2012,AJ,144,137 performed duringtwo seasons between April2004 and July DeMarchi,F.,Poretti,E.,Montalto,M.,Desidera,S.,Piotto,G. 2005 using the FLAMES spectrograph monitoring a total 2010,A&A,509,17 Fischer,D.A.&Valenti,J.2005,ApJ,622,1102 numberof317starswithamediannumberof15epochsper Freudling, W., Romaniello, M., Bramich, D. M., Ballester, P., star. A more limited follow up of 7 objects was performed Forchi,V.,Garc´ıa-Dablo´,C.E.,Moehler,S.,Neeser,M.J.2013, with theHARPSspectrograph in June2011. A&A,559,96 We obtained a median precision equalto 240 ms−1 for Gonzalez, G.1997,MNRAS,285,403 the GIRAFFE spectrograph, 180 ms−1 for UVES and 35 Kaluzny,J.,Rozyczka,M.,Pych,W.,Thompson,I.B.2014,AcA, ms−1forHARPS,workinginamagnituderangeinbetween 64,77 14.8< V <16.5. Among the sample of monitored stars, 59 Mayor,M.,Pepe, F.,Queloz, D., Bouchy, F.,Rupprecht, G., Lo arenowclassifiedasradialvelocityandpropermotionclus- Curto, G., Avila,G., Benz, W. et al. 2003, The Messenger, 114, ter members and do not show any evidence of variability. 20. Field contaminations was equalto 38% of thetotal sample. Malavolta,L.,Piotto, G.,Bedin,L.R.,Sneden, C.,Nascimbeni, V.,Sommariva,V.2015,MNRAS,454,2621 Intotalwefound45variablestars,amongwhich25are Mikolaitis, S˘., Tautvai˘siene˙, G., Gratton R., Bragaglia, A., Car- consideredclustermembersand20fieldobjects.For6spec- retta,E.2012,A&A,541,137 troscopic binaries we obtained the orbital solution, which Milliman,K. E., Mathieu, R. D., Geller, A. M., Gosnell, N. M., impliedminimummasses inbetween∼90MJ-∼460MJ and Meibom,S.,Platais,I.2014,AJ,148,38 orbital periods in between 3 daysand 220 days. Montalto, M., Piotto G., Desidera, S., Platais, I., Carraro, G., ∼ ∼ Thefrequencyofbinariesdowntothehydrogenburning Momany, Y., De Marchi, F., Recio-Blanco, A. 2009, A&A, 505, limit and up to 20 days orbital period is found equal to 1129,(M09) (1.5 1.3)%.intheuppermainsequenceofNGC6253,while Montalto, M., Villanova, S., Koppenhoefer, J., Piotto, G., ± thetotal binary frequency is equalto (13 3)%. Desidera,S.,DeMarchi,F.,Poretti,E.,Bedin,L.R.etal.2011, The precisions achieved by the HAR±PS spectrograph A&A,535,39,(M11) Montalto, M.,Santos, N. C.,Villanova, S., Pace, G., Piotto, G., aresufficienttofurtherextendtheexplorationofthebinary Desidera, S., De Marchi, F., Pasquini, L. et al. 2012, MNRAS, frequency well within the substellar domain, and to ana- 423,3039 lyze the high mass end of the planetary domain (down to Mortier,A., Santos, N. C., Sozzetti, A., Mayor, M. Latham, D., 4-5 MJ). The upcoming ESPRESSO spectrograph at VLT Bonfils,X.,Udry,S.2012,A&A,543,45 will represent an even more powerful tool to study plane- Pasquini,L.,Avila,G.,Blecha,A.,Cacciari,C.,Cayatte,V.,Col- tary frequencies in stellar clusters. A precision of around 5 less,M.,Damiani,F.,DePropris,R.etal.2002,TheMessenger, ms−1 may be expected with similar exposure times on our 110,1. targets. This will push further our exploration well within Piatti,A.E.,Clariá,J.J.,Bica,E.,Geisler,D.,Minniti,D.,1998, theplanetary domain, down toaround half ajupiter mass. AJ,116,801 Press W. H., Teukolsky S. A., Vetterling W. T., Flannery B. P. 1992,Cambridge:UniversityPress,|c1992,2nded. Queloz,D.,Henry,G.W.,Sivan,J.P.,Baliunas,S.L.,Beuzit,J. ACKNOWLEDGEMENTS L., Donahue, R. A., Mayor, M., Naef, D. et al. 2001, A&A 379, 279 WeacknowledgethesupportfromFundaçãoparaaCiênciae Raghavan, D., McAlister, H. A., Henry, T. J., Latham, D. W., aTecnologia(FCT,Portugal)intheformofgrantreference Marcy, G. W., Mason, B. D., Gies, D. R., White, R. J. 2010, PTDC/FIS-AST/1526/2014.MMacknowledgesthesupport ApJS,190,1 MNRAS000,1–12(2015) An extensive radial velocity survey toward NGC 6253 9 Rozyczka, M., Kaluzny, J., Thompson, I. B., Dotter, A., Pych, W.,Narloch,W.2014,AcA,64,233 Sagar,R.,Munari,U.,&deBoer,K.S.2001,MNRAS,327,23 Santos,N.C.,Israelian,G.,Mayor,M.2001,A&A,373,1019 Santos,N.C.,Israelian,G.&Mayor,M.2004, A&A,415,1153 Sestito,P.,Randich,S.&Bragaglia,A.2007, A&A,465,185 Stoehr, F.,White, R.,Smith, M.,Kamp, I., Thompson, R.,Du- rand, D., Freudling, W., Fraquelli, D. et al. 2008, ASPC, 394, 505 Twarog, B. A.,Anthony-Twarog, B. J., & DeLee, N. 2003, AJ, 125,1383 Zechmeister,M.&Ku¯rster,M.2009,A&A,496,577 MNRAS000,1–12(2015) 10 M. Montalto et al. Table 1.Observations Observingtime Exposuretime Instrument Dates 58hr 900s-3600s UVES/GIRAFFE 2/4/2004-15/7/2004 38.5hr 2020s-2700s UVES/GIRAFFE 3/5/2005-1/7/2005 13.5hr 680s-3600s HARPS 28-30/05/2011 Table 2.Clustervariablesa ID RA(J2000) DEC(J2000) B-V V N.obs. RV(kms−1) RMS(kms−1) P(%)b 39667 254.732063568 -52.736025614 0.845 15.137 42 -29.087 4.030 88. 29531 254.699632781 -52.704366462 0.828 15.628 28 -27.300 24.509 89. 33955 254.700423295 -52.630001667 0.832 15.990 26 -29.945 13.486 90. 31617 254.689311198 -52.669417089 0.831 15.797 11 -28.108 1.725 91. 39854 254.781010467 -52.723246777 0.828 15.189 3 -30.762 10.677 94. 43081 254.767887749 -52.693853688 0.835 15.119 17 -30.744 11.849 93. 31834 254.734726224 -52.665662838 0.810 16.068 25 -26.430 3.467 82. 30324 254.746004485 -52.690480237 0.814 15.940 41 -28.595 1.455 − 21659 254.876860986 -52.797707300 0.825 15.814 37 -29.166 30.154 − 39831 254.854747288 -52.724276741 0.789 15.464 36 -26.177 29.672 93. − 254.871791667 -52.737444444 0.822 15.010 19 -29.012 2.035 − 35905 254.658158455 -52.596792279 0.866 16.213 6 -28.896 7.217 24. − 254.979041667 -52.729833333 0.844 15.769 7 -28.434 4.519 − 24648 254.756000212 -52.791118012 0.913 16.428 39 -26.037 6.109 0. 49148 254.607856534 -52.777415622 0.846 15.550 7 -25.939 2.451 − 34667 254.708667366 -52.617760801 0.865 16.456 31 -29.062 13.748 85. 16649 254.892499238 -52.618200382 0.856 16.040 32 -28.446 7.678 − 30290 254.850999373 -52.690813347 0.825 15.544 41 -26.183 19.692 90. − 254.794625000 -52.807416667 0.847 16.324 39 -29.067 3.014 − 14466 254.902801435 -52.714731958 0.854 14.774 28 -26.381 1.467 − 27965 254.847398993 -52.732799296 0.812 15.812 19 -28.740 5.294 70. 27341 254.863111261 -52.743533585 0.819 15.658 16 -26.352 9.431 85. 49667 254.600056580 -52.710400816 0.828 15.794 6 -30.035 6.405 − 41026 254.645602194 -52.641141858 0.811 14.869 22 -31.695 1.350 6. − 254.977791667 -52.770055556 0.825 15.617 6 -28.714 2.009 − a– Belowthehorizontal linemembershipisestabilishedonlyonthebasisofradialvelocities. b– Propermotionmembershipprobability. Table 3.Fieldvariables ID RA(J2000) DEC(J2000) B-V V N.obs. RV(kms−1) RMS(kms−1) P(%) 31505 254.687118376 -52.671306636 0.864 15.728 41 -25.258 4.116 91. 478 254.923022901 -52.809584139 0.879 14.783 35 -55.207 4.216 − 40696 254.778223506 -52.662991026 0.807 15.255 23 -24.845 5.866 95. 28889 254.782822852 -52.715868624 0.824 15.701 29 -24.585 1.415 90. 27108 254.786303150 -52.747918476 0.821 15.426 41 -23.120 3.147 89. 28017 254.789168050 -52.731954093 0.814 15.436 25 -36.817 17.897 92. 9156 254.897722920 -52.620682013 0.875 16.213 6 -2.398 1.442 − − 254.956791667 -52.780388889 0.851 16.056 25 -24.514 7.795 − 39659 254.842084425 -52.736277043 0.768 15.266 30 -28.756 10.886 28. 49446 254.621343512 -52.739416289 0.905 15.163 40 -12.317 4.569 − 9756 254.913463429 -52.608067153 0.910 16.263 8 -71.134 66.779 − 28733 254.790227260 -52.718374337 0.795 15.830 27 -26.188 7.492 56. 23507 254.713571036 -52.811072103 0.808 15.779 36 -25.407 5.046 − 31512 254.810969168 -52.670869985 0.822 15.562 36 -32.037 1.814 90. − 254.933625000 -52.587555556 0.889 16.047 2 -21.799 2.256 − 10638 254.903502113 -52.587731891 0.946 15.490 8 -11.044 5.990 − − 254.865708333 -52.674027778 0.885 16.024 37 -61.851 8.402 − 23848 254.736362325 -52.805171881 0.835 16.104 14 -100.238 6.578 − 24548 254.742155375 -52.792685121 0.800 16.005 23 -93.670 1.698 0. − 254.769458333 -52.707944444 0.833 15.427 5 -23.596 1.624 − MNRAS000,1–12(2015)