Table Of Content

Astronomy&Astrophysicsmanuscriptno.9193main (cid:13)c ESO2008 February4,2008 The first XMM-Newton slew survey catalogue: XMMSL1 R.D.Saxton1,A.M.Read2,P.Esquej3,1,M.J.Freyberg3,B.Altieri1,andD.Bermejo1,4 1 XMMSOC,ESAC,Apartado78,28691VillanuevadelaCanãda,Madrid,Spain e-mail:[email protected] 2 Dept.ofPhysicsandAstronomy,UniversityofLeicester,LeicesterLE17RH,U.K. 8 3 Max-Planck-Institutfu¨rextraterrestrischePhysik,D-85748Garching,Germany 0 4 InstitutodeAstrofsicadeAndaluca,CSIC,Granada,Spain 0 2 ReceivedSeptember15,1996;acceptedMarch16,1997 n ABSTRACT a J Aims.Wereportontheproductionofalargearea,shallow,skysurvey,fromXMM-Newtonslews.Thegreatcollectingareaofthe 4 mirrorscoupledwiththehighquantumefficiencyoftheEPICdetectorshavemadeXMM-NewtonthemostsensitiveX-rayobservatory 2 flowntodate.WeusedatatakenwiththeEPIC-pncameraduringslewingmanoeuvrestoperformanX-raysurveyofthesky. Methods. Datafrom218slewshavebeensubdividedintosmallimagesandsourcesearched.Thishasbeendoneinthreedistinct ] energybands;asoft(0.2-2keV)band,ahard(2-12keV)bandandatotalXMM-Newtonband(0.2-12keV).Detectedsources,have h beenqualitycontrolledtoremoveartifactsandacataloguehasbeendrawnfromtheremainingsources. p Results.A’full’catalogue,containing4710detectionsanda’clean’cataloguecontaining2692sourceshavebeenproduced,from o- 14%ofthesky.InthehardX-rayband(2-12keV)257sourcesaredetectedinthecleancataloguetoafluxlimitof4×10−12ergss−1 r cm−2.Thefluxlimitforthesoft(0.2-2keV)bandis6×10−13ergss−1cm−2andforthetotal(0.2-12keV)bandis1.2×10−12ergss−1 t cm−2.Thesourcepositionsareshowntohaveanuncertaintyof8′′(1σconfidence). s a Keywords.X-rays:general–surveys– [ 1 v 1. Introduction Table1.PropertiesoflargeareaX-raysurveys 2 3 There is a strong tradition in X-ray astronomy of using 7 data taken during slewing manoeuvres to perform shallow 3 surveys of the sky. The Einstein (Elvisetal.,1992), Exosat Satellite Energyrange Coveragea Fluxlim. Position . 1 (Reynoldsetal.,1999) and RXTE (Revnivtsevetal.,2004) (keV) %ofsky b error 0 slew surveys all provide a useful complement to dedi- RASS 0.2-2.4 92 0.03 12′′ 8 cated all-sky surveys such as ROSAT (Vogesetal.,1999) Einsteinslew 0.2-3.5 50 0.3 1.2′ 0 and HEAO-1 (Piccinottietal.,1982) and the smaller area, XMMslew(soft) 0.2-2 14 0.06 8′′ : medium sensitivity ASCA survey (Uedaetal.,1999) and EXOSATslew 1-8 98 3 20′ v HEAO-1/A2 2-10 100 3 60′ Xi (pBenracnild-tbeetamal.X,2M00M0-)Ndeewetponlo(Hokass.inIgterweatsall.o,n2g00r1e)coangdniCsehdantdhraat RXTEslew 3-20 95 1.8 60′ XMMslew(hard) 2-12 14 0.4 8′′ r XMM-Newton (Jansenetal.,2001) with its great collecting a area, efficient CCDs, wide energy band and tight point-spread a TheXMM-Newtonslewskycoveragehasbeencomputedbyadding function(PSF)hasthepotentialtomakeanimportantcontribu- theareacontainedinalloftheimagesusedinsourcesearchingwithan tion to our knowledgeof the local universe from its slew data. exposuretimegreaterthan1second. Early estimates (Lumb,1998; Jones&Lumb,1998), based on bFluxlimit,unitsof10−11ergss−1cm−2 anexpectedslewingspeedof30degreesperhour,andaslightly lower backgroundlevelthan that actuallyencounteredin orbit, predicteda0.5–2keVfluxlimitof2×10−13ergcm−2s−1.While the chosen in-orbit slew speed of 90 degrees per hour reduces longstreaksintheMOScamerasbutarewellformedinthefast the sensitivity, initial assessments of the data showed that the observingmodesofthepncamera(Stru¨deretal.,2001).Forthis quality of the data is good and that many sources are detected reason,onlytheEPIC-pndatahavebeenanalysed. (Freybergetal.,2005). A review of properties shows that the Inthispaperwepresentacataloguedrawnfromslewstaken XMM-Newtonslewsurveycomparesfavourablywithotherlarge betweenrevolutions314and978coveringaskyregionof6240 areasurveysintermsofdepthandpositionalaccuracy(Table1). squaredegrees.Themainpropertiesoftheslewsurveydiscussed Duringslews all the three imagingEPICcamerastake data inthispaperaregiveninTable2. in the observing mode set in the previous pointed observation Slews have been source searched down to a likelihood and with the Medium filter in place. The slew speed of 90 de- thresholdof 8, whichafter manualrejectionof false detections greesperhourcombinedwiththeslowreadouttimeoftheMOS gives4710candidatesources. Using simulationswe have been detectors(2.6s;Turneretal.,2001)meansthatsourcesappearas able to identify a subset of high significance sources, with a likelihoodthresholddependentuponthebackgroundconditions, Sendoffprintrequeststo:R.Saxton that gives 2692 sources with a spurious fraction of ∼ 4% (the 2 R.D.Saxtonetal.:ThefirstXMM-Newtonslewsurveycatalogue:XMMSL1 Table2.PropertiesoftheXMM-Newtonslewsurvey Property Range(keV) 0.2–2 2–12 0.2–12 00 1 Observingtime(s) 5.4×105 5.4×105 5.4×105 s Meanexposuretimea 6.2s 6.0s 6.1s ew Totalnumberofphotons 1.6×106 2.2×106 3.8×106 of sl Meanbackgroundb 0.09 0.14 0.23 er Mediansourcecountratec 0.68 0.90 0.81 mb Limitingsourcefluxd 6.0×10−13 4.0×10−12 1.2×10−12 Nu 0 5 Num.sources(fullcat)e 2606 692 3863 Num.sources(cleancat)e 1874 257 2364 a The mean exposure time, after correcting for the energy-dependent vignetting. bcnts/arcmin2. 0 ccnts/second. 0 5 10 15 20 Mean count rate (7.5−12 keV) d ergs/s/cm2.Basedonadetectionof4photonsfromasourcepassing nearthedetectorcentre(exp.time=10s),withapower-lawspectrum ofslope1.7andgalacticabsorptionof3×1020atomscm−2. Fig.1. The background level distribution measured from the eThenumberofsourcesflaggedasgoodinthefullandcleancatalogues mean7.5–12keVcountrateineachslew.Thedashedlineshows (seetext). thecut-offusedtoexcludehighbackgroundslewsinthetailof thedistribution. ”clean” catalogue). Of these, 2621 are from unique sources. 199.19ms, and 47.66ms, whichconvertsto a scanneddistance In the hard (2–12 keV) band the clean catalogue contains 257 of 6.6 arcseconds, 17.9arcseconds, and 4.3 arcsecondsper cy- sources(253unique)ofwhich ∼ 9%are expectedto be dueto cle time, respectively for a slew speed of 90 degrees per hour. statisticalfluctuations. In the Small Window mode only the central CCD is operated The slew catalogue and accompanying images and expo- anda windowof 64×64pixelsis read out,i.e. only about1/3 suremapshavebeenmadeavailablethroughtheXMMScience ofthesingle,primeCCD. Inthefastmodes,TimingandBurst, Archive (XSA) as a queryable database and as FITS files1. A only 1-dimensional spatial information for the central CCD is summaryofscientifichighlightsfromtheslewsurveyhasbeen available and thus these modes are not well suited for source publishedinReadetal.,2006. detection. It was discovered in initial tests that slews with a high background gave a large number of false detections and resultedinextremelylongexecutiontimesforthesourcesearch- 2. Dataselection ingsoftware.Forthisreason,slewswithanaverage7.5–12keV (FLAG=0,PATTERN=0-4)countrateinexcessof5.5c/s(25% TheXMM-Newtonsatellitemovesbetweentargetsbyperform- of all slews), were excluded from the analysis (Fig. 1). This inganopen-loopslewalongtherollandpitchaxesandaclosed- leavesanominal312slewspotentiallyusefulforscientificanal- loop slew, where measurements from the star tracker are used ysis.Inpracticeasignificantnumberofslewswerenotabletobe in addition to the Sun–sensor measurements to provide a con- processedforavarietyofreasonsincludingfailuretocreateex- trolledslewaboutallthreeaxes,tocorrectforresidualerrorsin posuremaps,unreasonablyhighexposuretimes,attituderecon- thelongopen-loopphase.Theopen-loopslewisperformedata struction problems, missing keywords and the processing pro- steadyrateofabout90degreesperhouranditisdatafromthis ducingtoolargeimages.Whileitisstronglyhopedthatsomeof phasewhichmaybeusedtogiveauniformsurveyoftheX-ray thesedatasetsmayberecoverableinthefuturebyimprovements sky. totheprocessingsystem,forthepurposesofthisfirstcatalogue Slew Data Files (SDF) have been stored in the XSA from theyhavebeenleftoutandthecatalogueconstructedfrom218 revolution314onwards(beforethisdateslewswereperformed slews.Alistoftheobservationnumbersoftheseslewsisavail- withtheCLOSEDfilterinplaceandnoscientificallyusefuldata ableathttp://xmm.esac.esa.int/external/xmm science/slew survey was taken). Data from revolutions 314 to 978 have been used /obsid tab.html and the slew paths are shown in Fig. 2. About for this first catalogue but slews continue to be accumulated 4% of the covered area has been slewed over more than once and increments to the catalogue are planned to be released on and eventuallya deeper survey will be available by combining aregularbasis.Onlysciencedatafromslewswithadurationof overlappingslewdata,especiallyneartheeclipticpoles. greaterthan30minutesweredownlinkedduringthemajorityof these revolutions2. For the Slew Survey catalogue we have se- TheEPIC-pndetectorpassesoverasourceinabout14sec- ondsdependingonthepositionofthesourceandtheanglesub- lected only EPIC-pn exposures performed in Full Frame (FF), tendedbytheY-axisofthedetectortotheslewpath(theimpact Extended Full Frame (eFF), and Large Window (LW) modes, angle).Normallythisisclosetozero,butanimpactangleofup i.e.modeswhereall12CCDsareintegrating(inLWmodeonly to 20 degrees is possible. If the source passes through the de- halfofeachCCD).Thecorrespondingcycletimesare73.36ms, tector optical-axis an effective on-axis exposure time of ∼ 11 1 ThecataloguewasinitiallyreleasedonMay32006.Inthispaper secondsisachieved.InFig.3weshowtheexposuretimeofthe wediscusstheupdatedversionreleasedinOctober2006. slewsurveyasafunctionofskycoverage.Duetothevignetting 2 Thispolicy waschanged fromrevolution 921 onwards toinclude function,themeanexposuretimeisenergy-dependentbeing6.2 allslewslongerthan15minutes secondsinthesoft(0.2–2keV)bandand6.0secondsinthehard R.D.Saxtonetal.:ThefirstXMM-Newtonslewsurveycatalogue:XMMSL1 3 – anincreaseinthemaximumnumberofattitudepointswhich maybeusedbyeexpmap;releasedwithSAS v6.5. 3.1.Initialreduction The XMM-Newton Slew Data Files (SDFs) for EPIC-pn were processedusingtheepchainpackageoftheSAS.Fordiagnos- ticreasonsafewparametersweresettonon-defaultvalues(e.g. eventsbelow150eVwerekept). 3.2.Slewdivision Photon events are recorded initially in RAW or detector coordinates and have to be transformed, using the satellite attitude history, into sky coordinates. The tangential plane geom- etry commonlyused to define a coordinategrid for flat images is only valid for distances of 1–2 degreesfrom a reference position, usually placed at the centre of the image. To avoid this limitation, slew datasets have been divided into roughly one Fig.2. The paths of the slews used in the construction of degree by half a degree event files and attitude corrected us- XMMSL1inGalacticcoordinates. ing the task attcalc. Images and exposure maps were then extracted from the event files using the tasks evselect and eexpmap. This procedure relies on the attitude history of the satellitebeingaccuratelyknownduringtheslew;apointwhich 0 is addressed in section 4 . Software, based on SAS and ftools 0 60 (http://heasarc.gsfc.nasa.gov/ftools ; Blackburn,1995), to per- formtheprocedureofdividingandattitude-correctingslewdata eg2) hasbeenmadeavailableviatheXMM-Newtonweb-site3. d d sky area ( 4000 b[paantdTteshr:en=fpu0rl−olc4be]da),nudreso(h0fat.s2−bb0ea.e5nndkreeVp(e0a[.pt2ea−dt0tei.n5rnkt=eh0Vre]e+s[epp0aat.t5rea−rtn1e=2e.00n]ekregV+y e grat 0.5−2.0keV [pattern=0−4]), and hard band (2.0−12.0keV e Int 000 [pattern=0−4]). During the data processing stage severe prob- 2 lemswiththetransferandstorageoffilesfromlongslewswere encountered. To alleviate this, only exposure maps for the full energy band were produced. This leads to an approximation 0 being needed for exposure times in the different energy bands 0 5 10 whichisaddressedinsection6.3. Effective on−axis exposure (s) Fig.3. A histogram of the cumulative sky area covered for a 3.3.Sourcesearching given effective on-axis exposure time in the total (0.2-12 keV) band. Pilot studies were performed to investigate the optimum processing and source-search strategies. Uneven (and heightened) slew exposure is observed at the end of some slews (the (2–12keV) band.Smallripplesin the histogramare causedby ’closed-loop’phase) and imageswith an exposuretime greater gaps between the EPIC-pn CCDs and also by the different ob- than 20 seconds have been removed to ensure the uniformity serving modes; in LW mode only half of the CCD is exposed of the survey. We tested a number of source-searching tech- andthemaximumeffectiveexposuretimeis 6seconds. niques and found that the optimum strategy was the usage of a semi-standard ‘eboxdetect (local) + esplinemap + eboxdetect (map) + emldetect’ method, tuned to ∼zero background, 3. Dataprocessing and performed on a single image containing just the single Thedatahavebeenusedtoperformthreeindependentsurveys; events (pattern=0) in the 0.2−0.5keV band, plus single and asoftband(0.2–2keV)X-raysurveywithstrongparallelstothe double events (pattern=0−4) in the 0.5−12.0keV band. This ROSATall–skysurvey(Vogesetal.,1999;RASS),ahardband is similar to the technique used for producing the RASS cat- (2–12keV)surveyandanXMM-Newtonfull-band(0.2–12keV) alogue (Cruddace,HasingerandSchmitt,1988) and resulted in survey. the largest numbersof detected sources, whilst minimising the Data reduction was performed as detailed below, with the number of spurious sources due to detector anomalies (usu- publicXMM-NewtonScienceAnalysisSoftware(SAS),version ally caused by non-single, very soft (<0.5keV) events). The 6.1plusthefollowingmodifications: sourcedensitywas foundto be ≈0.5sourcespersquaredegree toanemldetectdetectionlikelihoodthreshold(DET ML)of10 – a modification for the oal library to handle the relatively (∼3.9σ). largetimegapsintheRawAttitudeFile(RAF);subsequently releasedwithSAS v7.0. 3 http://xmm.esac.esa.int/sas 4 R.D.Saxtonetal.:ThefirstXMM-Newtonslewsurveycatalogue:XMMSL1 4. Attitudereconstructionandpositionalaccuracy Attitude reconstruction for observation 9073300002 The good point spread function of the X-ray telescopes 67 (Aschenbachetal.,2000)shouldallowslewsourcepositionsto bedeterminedtoanaccuracyofaround4arcseconds,similarto 66 thatfoundforfaintobjectsinthe1XMMcatalogueofserendip- itous sources detected in pointed observations 4. Any errors in the attitudereconstructionforthe slew couldseriouslydegrade 65 g) thisperformanceandamajortechnicalchallengeofthedatapro- de cessinghasbeentoachievethenominalaccuracy. dec ( 64 The attitude information of the XMM-Newton satellite is providedbytheAttitudeandOrbitControlSubsystem(AOCS). Astartrackerco-alignedwiththetelescopesallowsuptoamax- 63 imumoffivestarstobecontinuouslytrackedgivingaccuratestar positiondataevery0.5seconds,whichoperatesinadditiontothe 62 SunsensorthatprovidesapreciseSun-linedetermination.Such 210 220 230 240 250 ra (deg) informationisprocessedresultinginanabsoluteaccuracyofthe reconstructedastrometryoftypically1arcsecond(1sigma)for pointedobservations.Forthe open-loopslews, largeslews out- sidethestar-trackerfieldofviewof3x4degrees,theon-board softwaregeneratesathreeaxismomentumreferenceprofileand Fig.4. A zoom into the problematic region of the attitude file atwo-axis(rollandpitch)Sun-sensorprofile,bothbasedonthe intheslew9073300002.Thepointsshowthegeneratedattitude groundslew telecommanding.Duringslewmanoeuvringa mo- information. mentumcorrectionissuperimposedontothereferencemomen- tum profile and, as there are no absolute measurementsfor the yawaxis,aresidualyawattitudeerrorexistsattheendofeach thisdiscrepancyhasnopracticaleffectonnormalstable-pointed slewthatmaybecorrectedinthefinalclosed-loopslew(Elfving observationaldata. 1999). Otherissuesaffectingtheastrometryperformanceappeared Two types of attitude data may be used as the primary after a careful visual examination of the RAF files, where two source of spacecraft positioning during event file processing. types of peculiarities appeared in some of the slews affecting TheyaretheRawAttitudeFile(RAF)andtheAttitudeHistory either a localised region or the totality of the slew. Five of the File(AHF).Forpointedobservations,theRAFprovidestheat- slewspresentedsharpdiscontinuitiesintheattitudereconstruc- titudeinformationatthemaximumpossiblerate,withoneentry tion, revealing the existence of single bad RAF point. As an every0.5secondswhiletheAHFisasmoothedandfilteredver- example a source in the slew 9073300002 was discovered to sion of the RAF, with times rounded to the nearest second. In have a closest ROSAT counterpart at a distance of 8 arcmin- slew datasets the RAF stores attitude information every 40–60 utes. Investigation showed that the source was observed at a seconds while the AHF contains the same records as the RAF timecoincidentwithalargeerrorintheattitudefile(Fig.4).A with identical positions and again with timing information in testinvolvingtheremovalofthebadpointandrecalculationof integer seconds. The user can select which one to use for data theattitude,whileshowinganimprovementinsourcepositions processingbysettinganenvironmentvariable. didn’t improve the astrometry to the level of the other slews. In a pilot study where the AHF was used for attitude re- Therefore,sourcesfallingina regionofbadattitudehavebeen construction, source detection was performed and correlations flagged in the catalogue with the ’Position Suspect’ flag (see with the ROSAT and 2MASS catalogues indicated a slew rel- section5.1).Sectionsofsevenotherslewsdisplayedanattitude ative pointing accuracy of ∼ 10 arcseconds . However, an ab- reconstruction that can best be described as turbulent (Fig. 5). solute accuracy of 0-60 arcseconds (30 arcseconds mean) was Again,sourcesfallinginthese slew sectionshavebeenmarked obtainedintheslewdirection,resultinginathin,slew-oriented withthe’PositionSuspect’flag. errorellipsearoundeachsource.Thiserrorappearstobeconsis- Asubsampleof1260non-extendedsources(definedashav- tentwiththeerrorintroducedbythequantisationofthetimeto ing an extent parameter < 2 from the emldetect source fit- 1secondintheattitudefileandledustochangetheprocessing ting) with DET ML > 10, have been correlated with several softwareasa better accuracyshouldbeobtained.As a test, the catalogues within a 60 arcseconds offset. The correlation with RAFwasusedtocomputetheastrometryforsomeobservations. the RASS reveals that 63% of the slew sources have an X- Here,anoffsetof∼1arcminutefromtheROSATpositionswas ray counterpart of which 68% (90%) lie within 16 (31) arc- found,butwithasmallerscattercomparedwiththepositionsre- seconds (Fig. 6). This gives confidence that the majority of turnedbytheAHFprocessing.Theconsistencyoftheseoffsets slews have well reconstructed attitude. To form a sample of suggestedthattheycouldbeduetoatimingissue.Afterdiscus- catalogues with highly accurate positions but which minimise sionswiththeflightdynamicsgroupitwasrealisedthatthestar thenumberoffalsematches,weusedtheAstronomicalVirtual tracker CCD integration time of 0.75 seconds is not included Observatory(AVO)tocorrelatetheslewpositionsagainstnon-X in the times in the attitude history. When this 0.75 seconds is raySIMBADcatalogues.Thisgave508matchesofwhich68% subtractedfromeveryentryintheRAFweobtainanoptimalat- (90%)werecontainedwithin8(17)arcseconds(Fig.7). titudefilefortheprocessing.Notethatthisoffsetremainsinall XMMrawattitudefilesbutanautomaticcorrectionhasbeenap- pliedintheSASsoftwarefromversion7onwards.Alsonotethat 5. Thecatalogue 4 TheFirstXMM-Newton SerendipitousSourceCatalogue, XMM- Source lists were produced by searching each slew down to a NewtonSurveyScienceCentre(SSC),2003. likelihood DET ML > 8 and combined to produce an initial R.D.Saxtonetal.:ThefirstXMM-Newtonslewsurveycatalogue:XMMSL1 5 Attitude reconstruction for observation 9084100002 Slew sources - RASS positional offsets 5 80 0 60 g) dec (de -5 N 40 -10 20 -15 0 194.0 194.1 194.2 194.3 194.4 0 10 20 30 40 50 60 ra (deg) angular separation (arcsec) Offsets in ra and dec between slew and Rosat sources Fig.5. Aplotshowingnon-smooth,orturbulent,attituderecon- 100 structionintherevolution0841attitudefile. 50 osat _r caalotagluoeguwehoerfe5k1n8o0wdnetsepcutiroionus.sTdheitsecitsioanvsaihlaabvleebaesetnhefla’gfuglel’dcoaut-t w - dec 0 accordingtoasetofcriterialaidoutinthenextsection. dec_sle -50 5.1.Causesofspuriousdetections 5.1.1. Multipledetectionsofanextendedsource -100 -100 -50 0 50 100 ra_slew - ra_rosat The source detection software attempts to parameterise source extents up to 20 pixels (82 arcseconds here) in radius. Larger sources, or sources with discontinuous or lumpy emission are reported as multiple small sources. This is particularly evident Fig.6. Acomparisonofslew sourcepositionswiththosefrom forthebrightsupernovaeremnants(SNR),e.g.Puppis-Awhich the RASS catalogue; 68% of the sources lie within 16 arcsec- resultsin81separate,confused,detections(Fig.8).Allaffected onds.Theupperpanelshowsahistogramoftheoffsetmagnitude sourceshavetheVER INEXTflagsettrue. whilethelowerpanelgivestheabsoluteoffsetinarcsecondsof theslewsourcefromtheROSATposition. 5.1.2. ThewingsofthePSFofabrightsource It was noticed during the construction of the 1XMM serendip- itous source catalogue that, due to the imperfect modelling of the PSF, a halo of false detections is often seen around bright 5.1.4. Brightsourcesoutsidedetectorfield-of-view sources.Thesameeffectisseeninslewexposuresbutduetothe reducedexposuretimeisonlyimportantforverybrightsources Reflectionsfrom the CRAB SNR, about10 arcminutesoutside ≫10c/s.AllaffectedsourceshavetheVER HALOflagsettrue. theEPIC-pnfield-of-viewduringtheslew9041000004,andthe TYCHOSNR,5arcminutesoutsidethefield-of-viewduringthe slew 9058600002,caused 5 false detections; VER HALO flag 5.1.3. Highbackground settrue. Flares in the background, due to solar protons (Lumbetal.,2002, Carter&Read,2007), cause a sudden 5.1.5. Badposition increaseinthenumberofeventsseeninaslewwhichmimicthe effect of slewing over an SNR (Fig. 9). These flares typically Sevensourcesarefoundneartheedgeofthedetectorortheedge last between 10 and 40 seconds and hence affect between 15 ofanimagewhichleavesanuncertaintyastothetruecountrate arcminutesand1degreeofaslew.Noautomaticflarescreening of the source and where its centre lies. These sourceshave the has been performed on the data but light curves of all slews VER NREDG flag set true and also the VER PSUSP flag set have been manually inspected and 175 sources, falling within truetoindicatethatthepositionissuspect.Inaddition,sources flare-affected sections have been flagged as bad. All affected in sections of slew with bad attitude (see section 4) have their sourceshavetheVER HIBGNDflagsettrue. VER PSUSPflagsettrue. 6 R.D.Saxtonetal.:ThefirstXMM-Newtonslewsurveycatalogue:XMMSL1 Slew sources - Simbad positional offsets 60 50 40 N 30 20 10 0 Fig.9. A heavily smoothed section of the slew 9097100002. 0 10 20 30 40 angular separation (arcsec) Thebrightpatchofeventsinthecentrehasbeenproducedbya backgroundflare. Fig.7. Ahistogramofthedistributionoftheangularseparation of the slew sources fromtheir Simbad counterpart;68%of the matchesliewithin8arcseconds. Fig.10. Left: A raw image of gamma Cru, an M star with m =1.6,detectedinslew90130900002.Right:Thesameimage v afterapplyingthefilter(FLAG==0,PATTERN==0,PI>200). 5.2.Statisticalfluctuations Fig.8. TheslewimageofthelargeSNR,Puppis-A.Itisdetected The numberof detections, after removingthe spurioussources asmanysmallsources(circles). highlightedintheprevioussection,risessteeplywithdecreasing detectionlikelihood(Fig.11)asexpected.Theyarealsoseento havea dependencyonthe backgroundrate withinthe image in whichthesourcewasfound(Fig.12);wherethebackgroundis definedastheeventcountrateabove10keV(PI >10000). 5.1.6. Zeroexposure Simulationshavebeenconductedtoinvestigatetherelation- ship between the number of spurious sources expected from Twosourcesareflaggedasfalsebecausetheylieattheveryedge backgroundfluctuationsandthenumberofeventsinan image. oftheslewandarereportedashavingzeroexposuretimeinone Simulated slew images have been created by inserting events ormoreof theenergybands.Thesehavethe VER FALSE flag into a template slew image, of 842 by 600 pixels and area 0.5 settrue. square degrees, at random positions. A flat spatial distribution ofeventshasbeenusedbecausethebackgroundis likelydom- inated by charged particle induced events, which show little 5.1.7. OpticalLoading variation across the detector, and internal flourescent emission lineswhichmapthe distributionofmetalsin the detectoritself Despite initial concerns that the lack of a detector offset map (Lumbetal.,2002). The resultant simulated images have been duringslews,wouldleadtoopticalloadingproblems,inpractice source searched and reveal a strong increase in the number of littleornoeffectwasfound.InFig.10weshowafirstmagnitude spurious detections with image counts, that rises steeply until starwhichisbrilliantinrawslewdatabutcompletelydisappears around500eventsandthenflattensout(Fig.13).Wehaveover- when the default event filtering is applied. None of the source laidthedistributionofcountsintherealslewimagesinfigure13 fluxesintheslewcataloguearebelievedtobecontaminatedby toshowthatthemajorityofrealimagesshouldgeneratefewspu- opticalphotons. rioussourcesforDET ML>12. R.D.Saxtonetal.:ThefirstXMM-Newtonslewsurveycatalogue:XMMSL1 7 Table 3. The number of sources found in slew images and in 800 simulatedslew imagesforcombinationsofminimumdetection likelihoodandbackgroundcountrate 600 mber of sources 400 M88 La cb-3/k.s0gb C4374o15m01b//in94e25d96 T3380o61ta38l//5384B08andcH6492a27rd//29732 S2169o08ft61//16896 Nu 10 - 2998/195 2580/118 312/61 2031/46 10 3.0 2419/106 2155/86 239/24 1638/13 200 12 - 2161/40 1875/28 185/12 1661/10 12 3.0 1782/21 1587/20 157/5 1354/2 14 - 1700/9 1470/7 139/4 1361/1 14 3.0 1423/6 1257/4 119/3 1122/1 0 10 20 50 100 10/14d 3.0/- 2696/109 2368/89 259/25 1877/13 DET_ML aTheminimumsourcelikelihood. Fig.11. Numberofdetectionsasafunctionofdetectionlikeli- bThemaximumbackgroundrateacceptedwithinanimage,definedas hood(totalband;0.2–12keV). thecountrateofeventswithenergygreaterthan10keV(PI>10000). c The number of detected sources (first number) and the number of expected false sources (second number) from simulations, in this energy band, for this combination of minimum detection likelihood 5 1. and maximum background count rate. The”combined” band is made from the unique distinct sources detected in any of the total, soft or hardbands. d A selection of all sources with DET ML>14 and sources with eg−2) 1 DET ML>10 from images where the background count rate is less d nsity ( than3c/s. e d e urc o S 5 0. Toassesstheabsolutenumberoffalsesourcesfoundbythe sourcedetectionchainasafunctionofDET ML,thepositions ofthe photonsin all11137slew imageshavebeenrandomised andthe imagessource-searchedagain.Results show thata sig- 0 1 10 100 nificantnumberofsourceswith low DET ML can be expected Background count rate (c/s) to be false. From Fig. 13 we know that the false source rate is influencedbythenumberofeventsintheimagewhichisinturn Fig.12. The mean density of total band (0.2-12keV) sources, related to the backgroundrate. Selections of DET ML and im- flagged good in the full catalogue, plotted as a function of the agebackgroundratecanbemadefromthesimulationresultsto backgroundcountrate(PI>10000). chooseaparticularspurioussourcefractionforagivenpurpose (Table3).Itisclearthatthehardband,havingtypicallyahigher backgroundthanthesoftbandandalowersignaltonoiseratio than either the soft or total band,is most affected by statistical fluctuations. A source selection [DET ML > 14 or (DET ML > 10andimage bckgnd rate<=3.0)]givesanexpected25false mages 200 DET_ML>8 d(0e.t7e%cti)ofnosrtfhroemso2ft5b9anhdaradnbda1n0d9sforoumrce2s6(9∼6(94%%)),f1o3raflrlosmou1rc8e7s7. No. of false sources per 1000 i 100 DET_ML>10 iAtlNlneaorrogsgmltuieeesecetptdtaihooontafhsndtiest4w8oi’o0uchenl%rarecedeaernrostreh’ofmhecriaisamor,stvVadabelguEoedeegeRsfnutrooehPp.matSrFhvoUieetdnhSauaaePtlctl’beiyflctda,ulaecfdgfoakernugsoe’firrmotslceoutartunthupardierlcsooeirbgnsasleutetwelehme.hec<Tsit’c=ifhdohuiinssls3lch’.lau0eoncaswdacsvtee/eaisddss-. atotalof2692sources(257inthehardband)inthe’clean’catalogue. DET_ML>12 0 100 1000 Counts per image 5.3.ReleasedCatalogues Boththe’full’catalogue,with5180detectionswithDET ML> Fig.13. The false source rate as determinedfrom simulations. 8andthe’clean’catalogue,with2692sourceshavingDET ML The histogram represents the distribution of counts in the real > 14 or (DET ML > 10 and image bckgnd rate<=3.0) slewimages. and the VER INEXT, VER HIBGND, VER HALO and VER FALSE flags set false (see section 5.1) are avail- 8 R.D.Saxtonetal.:ThefirstXMM-Newtonslewsurveycatalogue:XMMSL1 able from http://xmm.esac.esa.int/external/xmm data acc/xsa. Table4.Observingmodestatistics The catalogues contain columns with the detection threshold, backgroundrateandspurioussourceflagssufficienttoallowthe usertoselectasubsampleforaparticularscientificpurpose.The Mode FrameTime Extensiona Fractionb ’clean’ catalogue is conservativefor the soft band but may not (ms) (arcseconds) % bestrictenoughforsomeapplicationsinthehardband. FF 73 6.6 69 eFF 199 17.9 21 From hereon we discuss the properties of sources drawn LW 48 4.3 10 fromthe’clean’catalogueunlessotherwisestated. a The extension of the PSF along the slew direction caused by the satellitemovementduringCCDintegration. bThepercentageoftheslewskyareacoveredinthismode. 6. Sourceproperties 6.1.Namingconvention 0 2 The adopted name for sources detected in the XMM-Newton slewsurveystartswiththeprefix,XMMSL1,andthenencodes theJ2000skyposition,e.g.XMMSL1J010537.6+364858.The 15 name is assigned in two passes. When the three independent els) energy band source lists are combined to form one catalogue pix n ( thesourcenameissetusingthepositioninthebandwherethe nsio 10 DET ML likelihood is the highest. A second pass is then per- Exte formedsuchthatsourceswhichhavebeenobservedinmorethan one slew are giventhe same name.Again,priorityis givende- 5 pending on the detection likelihood. Detections are deemed to befromthesamesourceiftheircentresliewithin30arcseconds ofeachother.Giventhescarcityofslewsources(0.8detections 0 inthe’full’cataloguepersquaredegree)onthesky,30arcsec- 1 10 100 1000 104 Extension likelihood ondswasfoundtobeareasonablyrobustmatchradiusforpoint sources.Itisnotsogoodforextendedsourcesandthecatalogue insomecasescontainsmultipledetectionsofthesameextended 20 sourcewithdifferentnames. 5 1 6.2.SourceExtent els) pix Thesourcesearchalgorithmattemptstoparameterisedetections n ( o 0 as a point source or as an extended source with a radius of up nsi 1 e to 20pixels(82arcseconds).A largenumberof sources(30%) Ext are detected as being extended (extension parameter> 1), The measuredextensionisrelatedtotheextensionlikelihoodandto 5 thenumberofsourcecountsasshowninFig.14.Hereitcanbe seen thatthere is an upperbranchwhere an increasing number of photonsare needed to detect larger source extensions and a 0 1 10 100 1000 104 lowerbranchwheresmallextensionsarefoundforalargerange Background subtracted counts ofsourcestrengthsandextensionlikelihoods.Thepointspread Fig.14. Sourceextension,inunitsof4.1arcsecondsimagepix- function(PSF)fortheEPIC-pndetectorisafunctionofoff-axis els,plottedagainsttheextensionlikelihood(upper)andnumber angle(thedistancebetweentheoptical-axisandthe source).In of source counts (lower) for extended sources detected in the thedataanalysiswehaveusedtheaverageoff-axisangleofthe totalenergyband.Sourcesidentifiedas clustersofgalaxiesare pathofasourcethroughthedetector,tocalculatetheappropri- markedwithcircles. ate PSF. This is reasonablyaccurate forthe LW andFF modes wheretheframetimeisshortandtheextensionofthePSFalong the slew direction small, but introduces an inaccuracy for the ∼ 20%ofobservationstakenineFFmode,wheretheextension 6.3.Countrates alongtheslewdirectionis∼18′′(Table4).Thelowerbranchis contaminatedbyfalsedetectionsofextensioncausedbythisef- The count rates are calculated from the backgroundsubtracted fectwhichcanbedemonstratedbythefractionofeFFsourcesin countswithina circle aboutthe source,correctedfortheencir- thisbranch(35%)comparedwiththeexpected20%intheupper cledenergyfractionanddividedbythePSF-weightedexposure branch.Atlargecountrates,photonpile-updepressesthecounts time within the source region(see the emldetectuser guide for inthecentralpixelsofthesourceprofilealsocausinganapparent moredetails5).Forreasonsoflimitedresources,sourcesearch- extension.AcorrecttreatmentoftheslewPSFisneededtoprop- ing wasperformedin allbandsusingthe exposuremap forthe erlyparameterisesourceextension,neverthelesssourcesfalling totalenergyband.Thisproducesincorrectexposuresinthesoft in the upper branch of Fig. 14 are considered to be genuinely extended. 5 http://xmm.esac.esa.int/sas/6.1.0/doc/emldetect/index.html R.D.Saxtonetal.:ThefirstXMM-Newtonslewsurveycatalogue:XMMSL1 9 Table5.Fluxconversionfactors 10 Band energyrange Conversionfactora (keV) 1 Total 0.2–12.0 3.16 S count rate HSaorfdt 20.0.2––122..00 91..1444 RAS 0.1 aConvertsfromsourcecountrate(c/s)tofluxinthegivenenergyband inunitsof10−12ergs/s/cm2 0.01 Thesoft-bandfluxesareparticularlydependentonthespec- tralmodelusedand canbe quitediscrepantforstars wherethe 0.1 1 10 100 absorbingcolumnmaybesmall. XMM soft band count rate At the fluxes probed here, source confusion within the 20 arcsecondsextractionradiusisalmostabsent. Fig.15. AcomparisonofRASSandslewsoftbandcountrates (c/s).Thesolidlinerepresentsaratioof10:1andthedottedlines 6.5.Hardnessratio representafactorof10variationfromthisvalue. In Fig. 16 we plot the positions in Galactic coordinates of the softandhardbandsourcesandseparatelyallthesources,colour- and hardbandsdue to the energy-dependentvignetting.A cor- codedbyhardnessratio.Thehardnessratioisdefinedas rectingfactor: H =(S −S )/(S +S ) (3) R H S H S e =0.0013806×e3+0.0085632×e2+0.84282×e (1) h t t t whereS isdefinedasthesourcecountrateinthehardband H es =0.0014723×e3t −0.058884×e2t +1.3509×et (2) andSS asthecountrateinthesoftband. As expected, hard sources, notably the bright LMXB, pre- wheree ,e ande arethesoft,hardandtotalbandexposure dominate in the Galactic plane and soft sources at higher s h t timesrespectively,hasbeenappliedtotheexposuretimestocor- Galacticlatitudes. rectforthiseffect.Thesefactorswerecalculatedbycomparing severalsets of total, softand hard-bandexposuremapsandfit- 6.6.Spectra tingafunctiontotherelationshipbetweenthebands.Thisintro- duces a systematic uncertainty of ∼ 5% into the soft and hard Spectralanalysisof slew data isfundamentallycomplicatedby bandexposuretimesandcountrates.Duetothelimitof82arc- thefactthatasourcebrightenoughtoproduceareasonablespec- secondsonsourceextentusedbythesearchalgorithm,sources trum will suffer from photon pile-up. A quantative analysis of extendedbeyondthisvaluewillhavetheircountrateunderesti- these spectra awaits the implementation of a realistic pile-up mated.Verybrightsourcesareaffectedbyphotonpile-upwhich correction and the development of an energy-dependent point tendstoreducethecountrate.Thesourcestrengthlimitsforthe spreadfunction,whichwillbeavignetting-weightedaverageof observingmodesare givenbyStru¨deretal.,2001for FF mode thePSFateachdetectorpositiontraversedbythemotionofthe as6c/sandforLWmodeas9c/s.ForeFFmodethepile-uplimit sourcethroughthedetector. isinprinciple2c/sforapointedobservationbutwillbehigher hereastheslewingmovement,togetherwiththerelativelylong 7. Sampleproperties frametime,willreducethecountrateonthecentralpixelofthe PSF. In Fig. 17 we show the cumulative number count distribu- A comparison of the soft band count rates against RASS tion for the full energy band for sources within (|b| ≤ 20◦) countratesispresentedinFig.15for894non-extendedsources and outside (|b| > 20◦) the Galactic plane. A linear fit to with a ROSAT counterpart.The countrate ratio, XMM/RASS, sources outside the Galactic plane, in the central part of the is typically ∼ 10 but varies considerablywith the source spec- distribution (log counts between -0.2 and 1.0) gives a slope of trum.Thecomparisonofthesetwosurveys,coupledwithupper- −1.41±0.01. This is a little flatter than the Euclidian slope of limits analysisrepresentsa powerfultoolforfindinghigh vari- 1.5foundbyASCA (Uedaetal.,1999)andearlierbyHEAO-1 abilityX-raysources.An initialanalysisofhighvariabilityex- A2(Piccinottietal.,1982)andGinga(Kondoetal.,1998).The tragalactic sources found in this way has been published in differenceinslopemaybeduetoincompletenes,whichappears Esquejetal.,2007. tobecomeevidentinthesourcedistributionbelow∼0.8c/s.The distributionofsourceswithintheGalacticplaneshowsabreakat about12c/s(F =3.6×10−11ergss−1cm−2foranabsorbed 6.4.Fluxconversion 0.2−12 power-lawwith N = 3×1020 cm−2 andphotonindex1.7).At H Source fluxes have been calculated from count rates based on fainterfluxestheslopeis−1.29±0.02whileforbrightersources energyconversionfactors assuming a spectral modelof an ab- itis−0.43±0.02.Thebrightsourcepopulationagreeswellwith sorbedpower-lawwithN =3×1020cm−2andphotonindex1.7 the previous result of UHURU (Formanetal.,1978), except at H (see XMM science survey centre memo, SSC-LUX TN-0059, fluxesaboveF ∼8×10−10ergss−1cm−2wheregrosspile- 0.2−12 for a generaldescriptionof the technique).The energyconver- upeffectssignificantlyreducetheobservedcountrate(seesec- sionfactorsaregiveninTable5. tion6.3).ThelogN–logSrelationbelowF = 3.6×10−11 0.2−12 10 R.D.Saxtonetal.:ThefirstXMM-Newtonslewsurveycatalogue:XMMSL1 FX (ergs s−1 cm−2) 10−13 10−12 10−11 10−10 10−9 10−8 | | | | | | 0 1 ASCA GIS (2−10 keV) 1 (|b|<0.3deg, |l|<40deg) g−2] 0.1 e d F) [ > N( 0.01 |b|>20 UHURU (|b|<20) (2−6 keV) 0−3 |b|<20 1 0−4 1 −1 0 1 2 3 Log count rate (c/s) Fig.17. The cumulative number count distribution of total band (0.2-12 keV) sources inside and outside the Galactic plane, plotted against count rate and absorbed flux. A spectral model of an absorbed power-law of photon index 1.7 and N = 3 × 1020 cm−2 has been used to convert the slew H count rate into flux. The UHURU Galactic plane (dotted line; Formanetal.,1978)andtheASCA Galacticplanesurvey(dot- dashline;Sugizakietal.,2001)logN–logSrelationshavebeen displayedforcomparison.Inbothofthesecasesthefluxeshave beenconvertedtoanenergyrange0.2–12keVandtotheabove spectralmodel. is steeper than the 0.79±0.07 observed by ASCA in a survey with |b| < 0.3◦ in the 2–10keV band(Sugizakietal.,2001). It is likely that the difference is at least partly due to the sky re- gionssampled.ThewiderXMM-Newtonslewsamplewill,for example, contain a higher fraction of extragalactic sources. A thoroughanalysisofthelogN–logSrelationwillrequireacare- fulanalysisoftheeffectsofpile-up,completeness,skycoverage andtheEddingtonbiasandwedeferthistoalaterdatewhena largerfractionoftheskyhasbeencovered. 8. Duplicatedetectionsandvariability Inthefullcatalogue96sourceshavebeendetectedinmorethan oneslew.Countratevariabilityuptoafactor5,5and2isseen inthe total,softandhardbandsrespectively.Thegreatestvari- ability is seen in the total and soft band fluxes of XMMSL1 J125616.0-114632(2MASXJ12561595-1146367);whichvaried by a factor 5 between slews 9037400004 and 9083600004 (a baseline of 2.5 years). An upper limits analysis will likely find greatervariability. Fig.16. An AITOFF projection in Galactic coordinates of sourcesfromthefirstXMM-Newtonslewsurvey,wherethecir- 9. Identifications clesizescaleslogarithmicallywiththecountrate.Toppanel:the Allsourcesdetectedinthesurveyhavebeencorrelatedwithdif- totalbandsources;the hardnessratiois colour-codedsuch that ferent catalogues in order to identify the XMM-Newton slew light red is soft and blue is hard. Middle panel: the soft band sources with previously known objects (See Table 6 for a sources. Bottom panel: the hard band sources. The two bright summary of the resources used). The catalogues used for this sourcesseeninthehard-bandplotatlatitude∼−30areLMCX- aim comprise two astronomicaldatabases, a catalogue of clus- 1andLMCX-3. ters of galaxies and nine other catalogues (some which have been queried through the HEASARC astronomical database).