Table Of Content

Astronomy&Astrophysicsmanuscriptno.popping˙WVFS˙II c ESO2011 (cid:13) January13,2011 i The WSRT Virgo H filament survey II Cross Correlation Data 1 A.Popping12 andR.Braun3 1 0 1 Laboratoired’AstrophysiquedeMarseille,38RueFre´de´riqueJoliot-Curie,13388MarseilleCedex13,France 2 2 KapteynAstronomicalInstitute,P.O.Box800,9700AVGroningen,theNetherlands n 3 CSIRO–AstronomyandSpaceScience,P.O.Box76,Epping,NSW1710,Australia a J 2 ABSTRACT 1 Context.The extended environment of galaxies contains a wealth of information about the formation and life cycle of galaxies which are ] O regulated by accretion and feedback processes. Observations of neutral hydrogen are routinely used to image the high brightness disks of C galaxiesandtostudytheirkinematics.Deeperobservationswillgivemoreinsightintothedistributionofdiffusegasintheextendedhaloof thegalaxiesandtheinter-galacticmedium,wherenumericalsimulationspredictacosmicwebofextendedstructuresandgaseousfilaments. . h Aims.Toobservetheextendedenvironmentofgalaxies,columndensitysensitivitieshavetobeachievedthatprobetheregimeofLymanlimit p systems.HiobservationsaretypicallylimitedtoabrightnesssensitivityofN 1019cm 2,butthismustbeimproveduponby 2ordersof - HI ∼ − ∼ magnitude. o r Methods.InthispaperwepresenttheinterferometricdataoftheWesterborkVirgoHiFilamentSurvey(WVFS)–thetotalpowerproductof t thissurveyhasbeenpublishedinanearlierpaper.Byobservingatextremehourangles,afilledapertureissimulatedof300 25metersinsize, s × a thathasthetypicalcollectingpowerandsensitivityofasingledishtelescope,butthewelldefinedbandpasscharacteristicsofaninterferometer. [ Withtheverygoodsurfacebrightnesssensitivityofthedata,wehopetomakenewHidetectionsofdiffusesystemswithmoderateangular resolution. 1 Results.Thesurveymaps135degreesinRightAscensionbetween8and17hoursand11degreesinDeclinationbetween 1and10degrees, v − 4 includingthegalaxyfilamentconnecting theLocalGroupwiththeVirgoCluster.Onlypositivedeclinationscouldbecompletelyprocessed 0 andanalysedduetoprojectioneffects.Atypicalfluxsensitivityof6mJybeam−1 over16kms−1isachieved,thatcorrespondstoabrightness 23 sliemnsititbivuittyreoqfuiNriHnIg∼an10o1p8ticcmal−c2.ouAnnteurnpbairat.seIndtsoetaarlc,h19h9asobbejeecntsdohnaveewbiethenadheigtehctseidg,niofifcwanhcicehth1r7esahroelndeawsHweildleatescetairocnhs.withalowersignificance . Conclusions.ByobservingatextremehourangleswiththeWSRT,afilledaperturecanbesimulatedinprojection,withaverygoodbrightness 1 0 sensitivity, comparable to that of asingle dish telescope. Despite some technical challenges, the data provide valuable constraints on faint, 1 circum-galacticHifeatures. 1 : Keywords.galaxies:formation–galaxies:intergalacticmedium v i X r 1. Introduction the intergalacticmediumin emission,usingthe 21-cmline of a neutralhydrogen.Mostof the gasin the Cosmic Web will be In the current epoch, numerical simulations predict that most highlyionised,duetothemoderatelyhightemperaturesabove of the baryons are not in galaxies, but in extended gaseous 104 Kelvin, resulting in a low neutral fraction and relatively filaments, forming a Cosmic Web (e.g. Daveétal., 1999; lowneutralcolumndensities.Amoredetailedbackgroundand Cen&Ostriker,1999)Galaxiesarejustthebrightestpearlsin introduction on this topic is outlined in Popping (2010) and thisweb,asthebaryonsarealmostequallydistributedamongst Popping&Braun(2010b) threecomponents:(1)galacticconcentrations,(2)a warm-hot To investigate column densities that probe the Lyman intergalactic medium (WHIM) and (3) a diffuse intergalactic Limit System regime,verydeep Hi observationsare required medium.Directdetectionoftheintergalacticgasisverydiffi- with a brightness sensitivity significantly better than N cultatUV,EUVorX-raywavelengths(Cen&Ostriker,1999) HI ∼ 1019 cm 2. Reaching these column densities is important to andsofartheclearestdetectionshavebeenmadeinabsorption − learn more about the distribution of neutral hydrogen in the (e.g.Lehneretal.,2007;Trippetal., 2008). Inthisandprevi- inter-galactic medium and to have a better understanding of ouspapersinthisseries, wemakeanefforttodetecttracesof feedback processes that fuel star formation in galaxies. In Send offprint requests to: A. Popping e-mail: Popping&Braun (2010b) and Popping&Braun (2010a) two [email protected] Hi surveys have been presented that reach these low column 2 A.PoppingandR.Braun:TheWSRTVirgoHifilamentsurveyII densities in a region of 1500 square degrees. The first ∼ data productdescribedin Popping&Braun(2010b) is the total power data of the Westerbork Virgo Filament Survey, an Hi survey mapping the galaxy filament connecting the Virgo ClusterwiththeLocalGroup.Thesurveyspans11degreesin Declinationfrom 1to+10degreesand135degreesinRight − Ascension between 8 and 17 hours. This survey has a point source sensitivity of 16 mJy beam 1 over 16 km s 1 corre- − − spondingtoacolumndensityofN 3.5 1016cm 2. HI ∼ · − Fig.1. Observing mode of the WSRT dishes: 12 of the 14 The second data product presented in Popping&Braun dishesareplacedataregularintervalof144meters.Whenob- (2010a) is reprocessed data, using originaldata that has been serving at large hour angles, an approximatelyfilled aperture observedfortheHiParkesAllSkySurvey(Barnesetal.,2001; of300by25meterscanbesimulatedinprojection. Wongetal., 2006). The 1500 square degree region overlap- pingtheWVFSwasreprocessedtopermitcomparisonbetween these data productsand detections. The pointsource sensitiv- 2.1.Observations ityofthereprocessedHIPASSdatais10mJybeam 1 over26 − The galaxy filament connecting the Virgo Cluster with the kms 1,correspondingtoacolumndensityofN 3.5 1017 − HI ∼ · LocalGrouphasbeenobservedusingtheWesterborkSynthesis cm 2. − Radio Telescope (WSRT) in drift scan mode. Data was ac- In this paper, a third data product is presented: the cross- quiredintwo20MHzIFbandscentredat1416and1398MHz. correlationdata of the WesterborkVirgoFilament Survey.As In total 1500degreeshas beenobservedfrom 1 to 10 de- explainedinPopping&Braun(2010b),theaimoftheWVFS greesin∼Declinationandfrom8to17hoursinRigh−tAscension. wastoachieveveryhighbrightnesssensitivityinalargeregion Forty-five strips have been observed at fixed Declinations of of the sky, to permit detection of Hi features that probe the Dec= 1, 0.75, 0.5...10degrees.Thecorrelateddatawas neutral component of the Cosmic Web. The configuration of average−d in−Right−Ascension every 60 seconds, correspond- thearraywaschosensuchthatthedishesoftheinterferometer ing to an angular drift of about 15 arcmin, to yield Nyquist formafilledapertureof 300metersinprojectionbyobserv- samplinginthescandirection.Allregionshavebeenobserved ∼ ingatextremehourangles.Becauseofthemuchsmallerbeam twice,oncewhenthesourceswererisingandoncewhenthey size comparedto the WVFS total-poweror HIPASS observa- weresetting. tions,wewillbeabletoidentifybrighterclumpswithindiffuse featuresifthesearepresent. 2.2.Observingtechnique Thespecialobservingconfigurationcreatessometechnical challenges itself. This novel observing strategy requires non- The data has been obtained at very extreme hour angles be- standarddata-reductionprocedures,whichwillbeexplainedin tween 80 and 90 degrees, to be able to achieve a filled- ± ± section2.Insection3wewillpresenttheresults,startingwith aperture projected geometry. The WSRT has 14 dishes of 25 a list ofdetectedfeatures.Objectsare soughtbothblindly,by meterdiameter,ofwhich12canbeplacedatregularintervals usingahighsignal-to-noisethreshold,andinconjunctionwith of144meter. Atthese extremehourangles,the dishesdonot a known optical counterpartby using a lower threshold. New shadoweachother,howevertheseparationissosmallthatthere Hidetectionsanddiffusestructuresarebrieflydiscussed,how- arenogapsintheUV-plane.Usingthe12dishesatregularin- ever detailed analysis of these features will be presented in a tervals, a filled aperture is created in projection of 300 25 × follow up paper, also discussing new and tentative detections meter in extentas demonstratedin Fig. 1. When observingin obtainedfromtheWVFStotal-powerdataandthere-processed thismode,wecanachievethehighsensitivityofasingledish HIPASS data as described in Popping&Braun (2010b) and telescope,butbenefitfromtheexcellentspectralbaselineprop- Popping&Braun(2010a).Wewillendwithashortdiscussion erties and PSF of an interferometer. Although each pointing andconclusion,summarizingthemainresults. is only observedone minute at a time and two minutesin total, the expected sensitivity after one minute of observing is ∆N 5 1017cm 2over20kms 1ina 35 3arcminbeam HI − − ∼ × ∼ × overa 35 35arcmininstantaneousfield-of-view.Itis im- ∼ × 2. ObservationsanddataReduction portanttonotethattheshapeofthebeaminasinglesnap-shot isextremelyelongated.Thehighcolumndensitysensitivityis The basic observations have already been described in onlyachievedinpractiseforsourceswhichcompletelyfillthe Popping&Braun (2010b), where the total power product of beam,whichisonlylikelyforthenearestsources,orwhenthey the Westerbork Virgo Filament Survey is presented. Cross- arefortuitouslyalignedwiththeellipticalbeam.However,each correlation data were acquired simultaneously with that total pointing is observed at two complementary hour-angles; one powerdata.Wewillonlysummarisetheobservationalparam- positiveandtheothernegative,implyingthattheorientationof etersasthese have beendiscussedpreviouslyandconcentrate the snap-shotbeams is also complimentary.When combining more on the observing technique and data reduction as these the observations,the resulting beam has a well defined circu- areverynon-standardforthisdataset. lar/ellipticalshape,eventhoughthe orientationof each ofthe A.PoppingandR.Braun:TheWSRTVirgoHifilamentsurveyII 3 (RR,LL,RLandLR).Redefinitionofpolarizationproductshas no effect for sources that are not polarised. However the calibrator, 3C286, is known to be about 10% linearly polarised. The“apparentStokes”valuesthathavebeenusedforthiscali- bratorare(14.75,-1.27,0,-0.56)JyforthefirstIFand(14.83, -1.28,0.-0.57)JyforthesecondIF.Gainsolutionsdetermined withthecontinuumdataareappliedtothelinedataaswell.The calibrateddataisexportedfromAIPSintotheuvfitsformatand importedintotheMiriad(Saultetal.,1995)softwarepackage. Fig.2.Duetothefilledapertureof 300 25meters,avery ∼ × Miriad has been used to further reduce the data. The mosaic elongated beam is created. By observing each pointing twice scansaresplitintoindividualpointings.Continuumemissionis atcomplementaryorientations,thecombinedbeamofthetwo subtractedfromthelinedatausingafirstorderpolynomial,ex- observationsisnearlycircular,althoughwithasignificantside- cludingtheedgeofthebandpassandregionscontaininggalac- lobelevel. ticemission. Althoughthereareonlyscansat45differentDeclinations, twobeamsisonlyafewdegreesoneithersideofvertical.The eachscancontains540pointingsinRightAscension.Because conceptisdemonstratedinFig.2wherethecombinationofthe the complete survey cannot be imaged simultaneously due to individualelongatedbeamsformsasymmetric,approximately computer memory and image projection limitations, all the circularresponse. scans have been split into individualpointings. The complete survey is separated in 18 blocks of 40 pointings in Right Ascension and 45 pointings in Declination, with a ten point- 2.3.DataReduction ingoverlapinRightAscensionbetweenneighbouringblocks. Boththeautoandcross-correlationdatahavebeenobtainedsi- Thiscorrespondstosub-regionsof 10 11degreesinsize. ∼ × multaneously, but were separated before importing them into The centralpositionsof eachof the sub-regioncubesis listed Classic AIPS (Fomalont, 1981). The reduction process of the in table 1. When invertingthe datafromthe UV to the image auto-correlationsortotalpowerdatahasbeendescribedinde- domainauniformweightingschemewasapplied,asthisisthe tail in Popping&Braun (2010b). Here we will describe the mostoptimalweightingforthe12innerantennasofthearray. cross-correlationdataproduct. Because of the regular antenna spacing, many baselines have A typical observation consisted of a calibrator source the same length but the uv-plane is fully sampled at all spac- (3C48 or 3C286), the actual drift scan at a fixed Declination ings between one and eleven dish diameters. To suppress the and another calibrator source. After importing the data into side-lobesduetoincompletesamplingofthelongerbaselines AIPS,thecalibrationanddriftscandatawereconcatenatedinto involving antennas 13 and 14, a Gaussian taper has been ap- asinglefile. pliedtothevisibilitydatawithaFWHMof200arcsec;asthis The total dataset consists of 90 drift-scans at fixed decli- isapproximatelythesizeofthefinalbeam.The250individual nation, containing 91 baselines in two polarisations; half ob- velocitychannelswereimagedbetween150and2000kms 1 − servedatanextremenegativehourangleandtheotherhalfat withasamplingof8.24kms 1.Anexampleofasinglechannel − an extreme positive hour angle. Each baseline was inspected inasub-regionisshowninFig.3,theinner10degreesinRight manuallyinClassicAIPS,usingtheSPFLGutility.Suspicious AscensionandDeclinationhaveuniformmosaicsamplingand features appearing in the frequency and time domain of each an approximatelyconstant noise level. Fig. 4 shows the same baselinewereinspectedcritically.Featuresthatwerenotcon- field,overlaidwithalltheindividualpointingpatternsthathave firmedinspectraacquiredsimultaneouslywereflaggedasradio beenusedtomosaicthisfield. frequencyinterference(RFI). Unfortunately the data from pointings with negative Thecalibratorsourceswereusedtodeterminethebandpass Declinationsaremissingfromthecubes,aswellastheaxisla- calibration and the bandpass solutions were inspected by eye belsinthetwoplotsthatareshown.Thedatahasbeeninverted beforeapplication.Continuumdatafileswerecreatedfromthe usingtheMiriadsoftwarepackage,whichgridsandinvertsthe linedatabyaveragingthecentral75%ofthefrequencyband- data using the natural NCP (North Celestial Pole) projection width. These were used for determiningthe gain calibrations. thatappliestodataobtainedwithanEast-Westinterferometer. AsweareusingtheAIPSpackagewhichwasoriginallydevel- In most cases this is not a problem, howeverthe NCP defini- opedforVLAdata,are-definitionofthepolarisationproducts tion breaks down at a Declination of zero degrees. The NCP is necessary for correct gain calibration. The VLA measures projectionisdefinedinBrouw(1971)by: rightandleftcircularpolarisations(RR = I +V, LL = I V, RL = Q + iU and LR = Q iU), while the WSRT−data L=cosδsin∆α (1) − consists of the two perpendicular linear polarisation products and (XX =I Q,YY = I+Q,XY = U+iV andYX = U iV). − − − − Both definitionsarein termsof thesame true Stokesparame- M =(cosδ cosδcos∆α)/sinδ (2) 0 0 ters(I,Q,U,V).Byredefiningacalibrator’sparametersas(I, − ′ Q, U, V) = (I, -U, V, -Q) it is possible to successfully cali- At Declinations approaching zero, the M-value diverges ′ ′ ′ bratethe(XX,YY,XY andYX) databytreatingitasif itwere to infinity, making any projection impossible. The commonly 4 A.PoppingandR.Braun:TheWSRTVirgoHifilamentsurveyII Fig.3. Example of a channel map from one of the processed Fig.4.SameimageasFig.3,butnowwiththepositionsofthe sub-region cubes (Cube 9) from the WVFS cross-correlation pointingsoverlaid.Eachcubecontainsdatafrom45pointings data.Theinnerregioniswellsampledbythepointingsandthe in Declination, by 40 pointings in Right Ascension covering noiseisuniform,whileemissionattheedgesisnoisier.There 11 by 10 degrees. Due to the NCP projection of the data, the is 25% overlap in Right Ascension between adjacent cubes. Declinations close to zero degrees are increasingly distorted, Unfortunatelythe coordinatescouldnotbe shown in thisplot as canbe seen inthe shape ofthe pointingsthatare squeezed duetoprojectioneffects,asexplainedinthetext. toward the lowest declinations. Unfortunately, pointings with both positive and negative Declinations could not gridded si- multaneouslyduetolimitationsoftheNCPprojection. usedimagingtasks inbothMiriad andAIPShavebeenfound to work effectively when given either all positive or all negative Declinations, but failed when given both simultaneously. elongated beams, a circular main-lobe is formed, although Various methods of circumventing this problem have been substantialX-shapedside-lobesarealsoapparent. tested, but none of them was successful in re-projecting all of the data to a more useful grid. We have therefore Afterthecubeshavebeenimaged,thedata wereHanning chosen to neglect the negative Declination pointings of the smoothed using a width of 3 channels, to eliminate spectral cross-correlation data, as they contain less than 10% percent sidelobes and lower the RMS noise. Although the channel ofthetotal.Otherprojectionsforthenorthernpartofthedata samplingofthecubesisunchanged,thevelocityresolutionof have been considered, but these are not favorable since they thedataisdecreasedto 16kms 1. − ∼ yield a highly position dependant PSF. The data is gridded in NCP-projection, for which a well defined beam applies. After Hanningsmoothinghasbeen applied, the beam and When re-projecting the 10 10 degree field to e.g. a smoothed dirty maps were deconvolved using the MOSSDI ∼ × SIN-projection, the response to point sources, particularly task within Miriad to create clean cubes. This task is similar at the lowest Declinations, is severely distorted. The effect to the single-field CLEAN algorithm, but can be applied to of the NCP-projection can also be seen in the shape and mosaic-data.OnlyonepassofCLEANdeconvolutionhasbeen distribution of the pointings in Fig. 4 as the pointings at low donewithoutanymasking.Becauseoftherelativelylargesize Declination are squeezed into narrow ellipses. Quite apart of the18 cubes,the cleaningstep takes significantprocessing from the NCP-projection, it is an inherent property of an power. For the cleaning step a cutoff level of 50 mJy beam 1 − east-westinterferometer,thatthenorth-southspatialresolution was used. This cutoff level was determined empirically to thatcan be achievedat Dec=0is onlyas goodas the primary be optimal in cleaning the data as deeply as possible while beam,whichis 35arcmin. not creating false components. Although on the whole the ∼ data quality was significantly improved by deconvolution, Fig. 5 shows the synthesized beam for one of the central the brightest sources are still suffering from some residual pointingsinthestripatDec=10degrees.Theleftpanelshows side-lobe artifacts. A second cleaning pass using a clean the beam due to a single observation, which is extremely mask was not undertaken, since residual sidelobes prevented elongated.The rightpaneldisplays the beam after combining effective mask definition in an automatic way. The survey both complimentary observations. When averaging the two volumewasdeemedtoolarge,todeterminea maskmanually. A.PoppingandR.Braun:TheWSRTVirgoHifilamentsurveyII 5 Cube RAmin RAmax rms[Jybeam−1] Because the”cleaning”step duringthedata reductionhasnot 1 07:57:04 08:41:35 6.4 beenperfect,thereareartifactsinthevicinityofbrightgroups 2 08:26:10 09:11:42 5.5 of sourcesthatarecaused byside-lobesin thebeam.In some 3 08:56:19 09:41:44 5.3 cases the residual side-lobes of a bright source are similar in 4 09:26:25 10:11:49 6.3 strengthtofaintersourcesinit’svicinity. 5 09:56:29 10:41:53 5.8 Atsomelocations,thequalityofthedatawasnotoptimal 6 10:26:34 11:11:59 5.1 and parts of several Declination strips had to be flagged. As 7 10:46:39 11:42:04 5.8 a result, the restoringbeam is notwell matchedto the data at 8 11:26:45 12:12:08 5.5 theselocationsandthesensitivityis √2worse.Duetotheex- 9 11:56:49 12:42:15 5.5 tendednatureofthemosaicpointingpattern,suchinstancesare 10 12:26:56 13:12:18 6.5 11 12:56:59 13:42:24 6.4 often compensated by adjacent pointings. The effect of poor 12 13:27:04 14:12:28 7.0 data qualityis mostapparentaroundadeclinationof 9.25de- 13 13:57:09 14:42:34 6.6 grees (dueto solar interference),where the noise is enhanced 14 14:27:14 15:12:38 6.2 throughouttheentiresurvey. 15 14:57:19 15:42:43 6.6 Althoughanattemptwasmadetoobserveonlyduringnight 16 15:27:24 16:12:48 6.1 time hours, portions of scans are suffering from solar inter- 17 15:57:29 16:42:53 6.1 ference. The regionswhere this occursare relatively isolated, 18 16:27:35 17:01:56 5.9 however the data quality in these regions is significantly im- paired. Table 1. RightAscensionrangeandnoiselevelsofthe18in- GalacticHiemissionisanotherpotentialcauseoffalsede- dividual sub-cubes of the WVFS cross-correlation data. All tectionsatradialvelocitiesbelow +400kms 2. − cubesarecentredatadeclinationof5degrees. ∼ 2.6.Fluxdetermination In the case of single dish observationswith a largebeam, the Neverthelessthe improvementin dynamicrangeis significant fluxdensityofanobjectcanbe determinedbyintegratingthe ascanbeseenintheexampleshowninFig.6.Intheleftpanel spectrum over the line-profile of an object. When multiplied a channelis shown before cleaning and in the rightpanel the with the velocityresolutionof the observations,thisgivesthe same field is shown after cleaning. In the left panel a very linestrengthin[Jykms 1].Thismethodcanbeusedwhenthe strong X-shaped sidelobe pattern can be seen at the location − telescopebeamismoreextendedthanthespatialsizeofanob- ofbrightsources.Intherightpaneltheside-lobeshavealmost jectimplyingthatthesourcesareunresolved.Thisassumption completelydisappeared. has been used for the HIPASS data and for the WVFS total power data. Although the WVFS cross-correlation data has a relativelylargebeam size comparedto typicalinterferometric 2.4.Sensitivity observations,someobjectsareapparentlyresolved. Thefluxofeachobjectisthereforedeterminedintwodif- Wereachanalmostuniformnoiselevelthroughoutthesurvey ferent ways. The first method simply employs the integral of area of 6 mJy beam 1 over 16 km s 1 which corresponds to − − thelinestrengthinthespectrumwiththehighestpeakbright- a column density sensitivity of N 1.1 1018 cm 2. The HI − ness.Theerrorinthisestimateisgivenby: ∼ · noise level in each of the individual 18 cubes is listed in table 1. Althoughwe reach veryhighsensitivities, we notethat 1.5 W the data is affected by some residual side-lobe contamination σ= · 20δv rms (3) inthevicinityofbrightsources. s vres × Typicalfeaturesthatoccurinthedatacubesareillustrated wherev isthevelocityresolutionofthedata,whileδvisthe res inFig.3.Attheedgeofthefieldthereisanincreaseinthenoise channelseparationin[kms 1]. − dueto the finite numberof mosaicpointingsincludedin each Thesecondmethodisamoresophisticatedone,thatisbet- cube. The large dark and light structure in the upper part of ter suited to extended sources. Ideally the flux would be de- thefieldiscausedbysolarinterferenceduringtheobservation. terminedinteractivelyforeachobject,byselectingtheregions Thisinstanceofsolarinterferenceisveryextremeandcertainly that contain significant emission in a moment map integrated nottypical.Distributedoverthefieldaremanygalaxiesthatap- over the velocity extent of the source. While this is possible pearaspointsources.TheX-shapedresidualside-lobepattern when observing a modest number of individual galaxies, the is still apparentwithin the noise aroundthe brightestof these total area of the WVFS survey and the number of objects is sourcesinthefield. too large to treat each objectmanually.An automatic method is used, which consists of integrating the flux of the moment mapsinthevicinityofeachpeakouttoacertainradius.Ideally, 2.5.FalsePositives beyondacertainradiustheintegratedfluxdensityremainsap- Although the reduced data has a good flux sensitivity, there proximately constant apart from fluctuations due to the noise aremanyartifactsinthedata,thatareimportanttounderstand. and local background. Because of side-lobe and large-scale 6 A.PoppingandR.Braun:TheWSRTVirgoHifilamentsurveyII Fig.5.BeamshapeofacentralpointinginthestripatDec=10degrees.Theleftpanelshowsthesynthesizedbeamofa single observation,whiletherightpanelshowsthebeamforthecombinationoftwocomplimentaryscans.Bothpanelshavethesame intensityscaleandcontoursaredrawnat70,80and90percentofthepeak. Fig.6.Theleftpanelshowsaregionofthedirtymap;datathathasonlybeeninvertedfromtheu,vplanetotheimageplane.The rightpanelshowsexactlythesameregionafteronepassofcleaning.Whileintheleftpanelthecrosspatternofthebeamisvery apparent,intherightpaneltheside-lobesarealmostcompletelysubtracted. backgroundeffectsthisisoftennotthecase;theintegratedflux Foreachobjectazerothmomentmapiscreatedandthera- dropsagainorkeepsincreasing.Aradiushastobedetermined dialprofileofthefluxvaluesisdetermined.AGaussianfunc- thatyieldsthebestestimateoftheintegral,whilerestrictingthe tionisfittedtotheradialprofile,theσvalueofthefitisanin- effectsofconfusingfeaturesasmuchaspossible dicationofhowextendedthesourceis.Thepixelbrightnessis integratedforallpixelswithinaradiusof3.5σandconverted A.PoppingandR.Braun:TheWSRTVirgoHifilamentsurveyII 7 from [Jy beam 1 km s 1] to [Jy km s 1] by division with the In the projected frame the sum of the beam responses − − − beamarea.Theerroroftheintegratedfluxiscalculatedfrom: changes dramatically with declination if the projection is not appliedtotheshapeoftheprimarybeam. error= 1.5·W20 1.5·Asδv rms (4) As a result, fluxes appear lower in the projected frame at s vres × sAbeam × low declinations, as the sum of the beam responsesby which thefluxisweightedislarger.Athighdeclinationstheopposite where A is the surface area of the source and A is the s beam isthecasewherethefluxvaluesintheprojectedframebecome surface area of the synthesised beam. The integrated value of enhanced. thepixelvaluesisused,ratherthantheintegraloftheGaussian This effect is clearly visible in the data, as the measured fit,tobemoresensitivetopossibleextendedemissionwhichis fluxofobjectsatlowdeclinationissystematicallytolow.This likelytobesuppressedbythewingsoftheGaussian. is demonstrated for a sample of objects in the right panel of Fig. 8. The data points show the ratio between the fluxes ob- tainedfromtheWVFStotalpowerdataasdescribedinchapter 2.7.Fluxcorrection 4andthefluxesfromthecross-correlationdata,withoutaflux correction. The flux ratios are plotted on a logarithmic scale Wehavenoticedashortcomingintheinverttaskwithinmiriad againstdeclination.Atlowdeclinationsthetotalpowerfluxes when the data is converted from the u,v plane to the image aremuchhigher,whileathighdeclinationstheoppositeisthe plane. When creating the mosaiced image, each individual case. pointing is inverted and then the set of images are combined Thecorrectionthathastobeappliedisgivenby: usingtheprimarybeammodelfortherelativeweightings. WhengriddingthedatausingtheNCPprojectiontheoffset ΣB(M) C(δ)= i (7) indeclinationwithrespecttothetangentialpointiscalculated ΣB(δ) i usingequation2.Inthisfunctionδ isthecentraldeclination, 0 δ is the observed declination that has to be gridded and ∆α This ratio is plotted in the left panel of Fig. 8 as function is the differencebetween the central RightAscension and the of declination. The cross indicates the central position of the Right Ascension to be gridded. At declinations close to zero, projected grid at 5 degrees, where the ratio is exactly 1 as no thedifferencesinMintheprojectedframebecomeverysmall: correctionhastobeapplied.Fordeclinationsbelow5degrees δ = 0+ǫ andδ = 0+2ǫ aregriddedtothesamepixelinthe thefluxeshavetobescaledup,whileforhighdeclinationsthe projectedframeandforδ = 0thereisasingularityasthereis fluxeshave to be scaled down.At bothendsof the correction nosolutionatall. function there is a bump, as the edges of the mosaic are not Intheobservedframe,themosaicisNyquistsampledand Nyquist sampled in the observations, so the integrated beam there is a pointing every 15 arcmin between 0.25 and 10 de- responsesaredifferenthere.Fordeclinationsapproachingzero greesin declinations.Thisis shownin the toppanelofFig. 7 degreesthecorrectiongoestoinfinity,becauseofthesingular- wheretheresponseofeachbeamisplottedasfunctionofdec- ityintheNCPprojectionhere. lination.Ateachpositionthesumofallthebeamresponsesis IntherightpanelofFig.8thesamecorrectingratioisplot- equalandtheweightedsumisalwaysunity. tedonalogarithmicscale,togetherwithasampleofdatapoints Theseparationbetweenthe pointingswith respectto each asdescribedbefore.Althoughthescatterislargethedatapoints other changes, when they are convertedto the projectedNCP followthecorrectingfunctionreasonablywell.Thefluxesob- frame.Notonlythepositionofpointingschanges,thecomplete tainedfromthetotalpowerandcross-correlationdataactually shape of the primarybeam becomesdifferent, and a distorted canbedifferentasthetotalpowerdataismoresensitivetoex- beamshouldbeappliedwhendoingtheweighting.Wesuspect tended emission, but to get the cross-correlation data on the thatthisisnothappeningandthattheundistortedbeamisused. rightlevel,thecorrectionhastobeapplied. This is demonstratedin the bottom panel of Fig. 7 where the undistortedbeamresponseofthepointingsis plottedasfunc- 3. Results tionoftheprojectedMvalue.TheMvaluegivestheoffsetwith respecttothereferencepixel,so M = 0correspondstoδ = 5 3.1.SourceDetection degrees. BecauseofthelargeextentoftheWVFS,andthelargenumber Inanimagethefluxisdeterminedbytheweightedsumof of independentpixels, an automated source finding algorithm thebeamresponseofallthecontributingpointings.: isessentialtoobtainalistofcandidatedetections.Althoughthe ΣS (δ) B(δ) sensitivityofthedataisgood,sourcedetectionisnotstraight- Sδ = iΣB(·δ)i (5) forwardbecauseoftheartifactsthatareapparentinthedataas i describedinsection2.5.Twostrategieshavebeenemployedto and circumventthesecomplicationsandtoobtainalistofcandidate ΣS (δ) B(δ) S = i · i (6) detectionsthatisascompleteaspossible.Thefirstmethodisa M ΣBi(M) blindsearchthatusesaclippinglevelof8σ.Thisconservative WhereS (δ)isthemeasuredfluxbyapointing,B(δ)isthe clippinglevelwillprovideareliablelistofbrightfeatures.The i i beamresponseintheobservedframe,whileB(M)isthebeam second method uses a less conservative clipping level of 5σ i responseintheprojectedframe. thatwillalsoyieldmanyfalsepositives.Anextraconstrainton 8 A.PoppingandR.Braun:TheWSRTVirgoHifilamentsurveyII 1 0.8 0.6 0.4 0.2 0 0 2 4 6 8 10 12 δ [degrees] 1 0.8 0.6 0.4 0.2 0 −2 0 2 4 6 8 M [degrees] Fig.7.Beam responseofpointingsinthemosaicasfunctionoftheobserveddeclination(δ)(toppanel)andtheNCP projected declination(M).TheobservedmosaicisNyquistsampled,witharegularintervalbetweenbeamsof15arcmin.Whenprojecting thedata,thesamplingbecomesdifferent. 5 1 4 0.5 n) n3 o ctio ecti orre Corr C2 g( 0 o l 1 −0.5 0 0 2 4 6 8 10 0 2 4 6 8 10 Declination [degrees] Declination [degrees] Fig.8.Leftpanel:Correctionfactorasfunctionofdeclination,tobeappliedontheobtainedfluxes.Rightpanel:Correctingfactor onalogarithmicscaleswithdatapointsrepresentingthefluxratiobetweenWVFStotal-powerandcross-correlationdata,before applyingthecorrectiononthecross-correlationdata. thesefeaturesisthatanopticalcounterpartisrequiredwithina pixelabovea declinationof9 degreeshasbeenblanked.This suitablesearchradiusandatacomparableradialvelocity.The regioncouldnotbeusedduetosolarinterference,whichheav- search radius is variable, dependingon the declination of the ilyaffectedthequalityofthedata.Theinterferenceissostrong candidatefeature,aswillbeexplainedbelow. herethatitincreasesthenoiseandcausesmanyfalseandunre- liabledetections.Asaconsequence,anyobjectsbetween 172 ForbothapproachestheDuchamp(Whiting,2008)source ∼ findingalgorithmhasbeenused,withdifferentcontrolparam- and 190degreesinRightAscensionandabove9degreesin ∼ Declinationwillnotbedetected. eters. The two methodsand their results are described below. Whensearchingforobjects,Duchampusestheprocessed,three dimensionaldatacubes.Thesourcefinderhasbeenrunoneach ofthe18cubesindividually,althoughforCubes8and9every A.PoppingandR.Braun:TheWSRTVirgoHifilamentsurveyII 9 V rms clip grow size res 7 [kms 1] [mJybeam 1] [σ] [σ] [pixels] − − Setup1 16 6.7 8 3 3 6 Setup2 32 5.0 8 3 5 Setup3 64 4.1 8 3 9 5 Setup4 128 3.1 8 3 17 e) c Table 2. Duchamp parameters for the blind Hi search in the n4 e d WVFScross-correlations. ci n g(I3 o l 2 1 0 −10 −5 0 5 10 rms [σ] Fig.9. Histogram of incidence of brightness in units of σ in Cube13ofthe WVFS. Ata levelof 5σthereare stillmany − pixels,indicatingthatthenumberoffalsepositivesatthe+5σ levelisalsolikelytoberelativelyhigh.Atalevelof 8σthere − arenonegativepixels,soaboveaclippinglevelof+8σnofalse detectionsareexpected. 3.2.BlindDetections AlthoughthenoiseinthedataisapproximatelyGaussian,itis clear froman inspectionof the histogramofbrightnessesthat there are excess negative outliers from this ideal distribution. Due to the purely relative nature of interferometric data (the Fig.10. Example of a source with a bright residual side-lobe absence of the auto-correlation), artifacts caused by e.g. pattern.AnXpatternisclearlyvisibleandthoseregionswithin side-lobes,continuumsourcesandsolarinterferencearesym- theside-lobeswithmultiplecontoursmightbemis-identifiedas metricintheirpositiveandnegativeexcursionswithrespectto individualobjectsbytheDuchampsource-findingalgorithm. zero. For this reason, the number of noise and artifact pixels withinanegativebrightnessintervalisequaltothenumberof likelyfalse positivepixelswithinthe same brightnessinterval of 7 arcmin. A radius of 7 arcmin was chosen both to acco- on the positive side of the histogram. This is illustrated in modatesomeintrinsicoffsetofthe Hiandopticalcentroidas Fig.9whereahistogramoftheobservedbrightnessisplotted well as those instances of limited positional accuracy, as will from the central portion of Cube 13. The incidence of each beexplainedbelow. brightnessscaledbythelocalRMSfluctuationlevelisplotted. At 4 or 5 σ the number of negative pixels is still significant, All spectra and moment maps were inspected visually, to but these rapidly drop to zero for a brightness below 8σ. A lookforartifacts.Continuumartifacts,orsolarinterferencecan clipping level of +8σ was therefore chosen since it−should resultinafalsedetection,butareeasilyrecognisedbyeye. yieldnofalsepositivedetectionsduetoeithernoiseorartifacts. Themomentmapswereusedtoeliminatethosecandidate sourcescoincidentwithresidualside-lobeartifacts.Whencan- FortheactualsourcefindingDuchamphasbeenusedwith didate sources are coincident with one of the arms of the X- four different settings, that are summarised in table 2. The shapedresidualside-lobeof a brightnearbysource,thenthey cubes were first Hanning smoothed to velocity resolutions of areverylikelyunreliable.Anexampleofaclearresidualside- 16,32,64and128kms 1,tobemoresensitivetosourceswith lobestructureisshowninFig.10. − different line widths. Although an initial clipping level of 8σ Furthermore, all multiple detections of the same source hasbeenusedforpeakidentification,thedetectedfeaturesare were excludedfrom the list of detections. Featuresthat are at ”grown”toalevelof3σ.Furthermoreacertainnumberofpix- theedgeofacubearelikelytobedetectedtwiceintwoadja- elsinthevelocitydomainisrequired,thatisrepresentativefor centcubes.Inthecubeswithdifferentvelocityresolution,the thevelocityresolution. peakfluxofthecandidateisnotalwaysatthesamespatialposi- Foreachcandidatedetection,thespectrumwasdetermined tion,resultinginaslightlydifferentapparentcentroid.Finally, and a moment map has been created by integrating the data sourcesaresometimescountedtwicewhentherearetwobright cubeoverthevelocitywidthofthedetection.Furthermore,an spectral components that do not connect. For example in the opticalcounterpartwassoughtforeachsourcewithinaradius caseofadoublehornedprofile,wheretheregionbetweenthe 10 A.PoppingandR.Braun:TheWSRTVirgoHifilamentsurveyII two peaks does not exceed the 3σ level which was set as the “growing”limit. A total of 135 sourceshave been detected using the blind detection approach, which have a peak brightness exceeding 8σ. The properties of each detection are listed in the online appendix, where the spectrum of each detection is shown as well. The first column gives the name of the source, consist- ingofthecharacters”WVFSCC”(WesterborkVirgoFilament SurveyCross-Correlation), followed by six plus six digits for theRightAscensionandDeclinationrespectivelyinhh:mm:ss anddd:mm:ssformat.ThesecondcolumngivestheopticalID if present, followed by the Right Ascension, Declination and Velocity.The sixth and seventhcolumngive the line-width at 50 and 20% of the peak (W and W ). The following three 50 20 columns give the integrated line-strength (S ), the integrated l moment map flux density (S ) and the error in the integrated i flux density (σ ). The differences between the different flux S measurementsandthemethodofestimatingtheuncertaintyare explainedinsection3.4.Thelastcolumninthetableindicates whether the object has been found in the blind search (B) or in the search where an optical identification is required (O). Whenobjectsappearinbothsearchmethods,thisisindicated withtwoletters. Of the 135 detected sources in the blind search method, onlyWVFSCC120929+080730doesnothaveanopticalcoun- terpart, the properties of this objects are given in table A.1, whichhasthesamecolumndescriptionastheonlineappendix, butonlyshowingnewHidetectionsinthesurvey.