High-resolution X-ray spectroscopy and imaging of the nuclear outflow of the starburst galaxy NGC 253

0.69 MB·English

by  M. Bauer

Astronomy&Astrophysicsmanuscriptno.ngc253RGS (cid:13)c ESO2008
February5,2008
High-resolution X-ray spectroscopy and imaging of the nuclear
⋆
outflow of the starburst galaxy NGC 253
M.Bauer1,W.Pietsch1,G.Trinchieri2,D.Breitschwerdt3,M.Ehle4 andA.Read5
1 Max-Planck-Institutfu¨rextraterrestrischePhysik,Giessenbachstraße,85741Garching,Germany
2 INAFOsservatorioAstronomicodiBrera,viaBrera28,20121Milano,Italy
7 3 Institutfu¨rAstronomiederUniversita¨tWien,Tu¨rkenschanzstr.17,A-1180Wien,Austria
0 4 EuropeanSpaceAstronomyCentre,ESA,P.O.Box50727,28080Madrid,Spain
0 5 DepartmentofPhysics&Astronomy,UniversityofLeicester,LeicesterLE17RH,UK
2
Received4September2006;accepted18January2007.
n
a ABSTRACT
J

9 Aims.UsingXMM-Newtondata,wehaveaimedtostudythenuclearoutflowofthenearbystarburstgalaxyNGC253inX-rayswith
1 respecttoitsmorphologyandtospectralvariationsalongtheoutflow.
Methods.WeanalysedXMM-NewtonRGSspectra,RGSbrightnessprofilesincross-dispersiondirection,narrowbandRGSandEPIC

2 imagesandEPICPNbrightnessprofilesofthenuclearregionandoftheoutflowofNGC253.
v Results.WedetectadiversityofemissionlinesalongtheoutflowofNGC253.ThisincludestheHe-likeionsofSi,Mg,NeandO
2 andtheircorrespondingionsinthenexthigherionisationstate.AdditionallytransitionsfromFeXVIIandFeXVIIIareprominent.
0 Thederivedtemperaturesfromlineratiosalongtheoutflowrangefrom0.21±0.01to0.79±0.06keVandtheratioofFeXVIIlines
3 indicatesapredominantlycollisionallyionisedplasma.Additionallyweseeindicationsofarecombiningorunderionizedplasmain
0 theFeXVIIlineratio.Derivedelectrondensitiesare0.106±0.018cm−3forthenuclearregionand0.025±0.003cm−3fortheoutflow
1 regionclosesttothecentre.TheRGSimageintheOVIIIlineenergyclearlyshowsthemorphologyofanoutflowextendingoutto
6 ∼750pcalongthesouth-eastminoraxis,whilethenorth-westpartoftheoutflowisnotseeninOVIIIduetotheheavyabsorption
0 bythegalacticdisc.Thisisthefirsttimethatthehotwindfluidhasbeendetecteddirectly.ThelimbbrighteningseeninChandra
/ andXMM-NewtonEPICobservationsisonlyseenintheenergyrangecontainingtheFeXVIIlines(550–750eV).Inallotherenergy
h
rangesbetween400and2000eVnoclearevidenceoflimbbrighteningcouldbedetected.
p
- Keywords.X-rays:galaxies–Galaxies:individual:NGC253–Galaxies:spiral–Galaxies:starburst–ISM:jetsandoutflows
o
r
t
s
a 1. Introduction conestructure.Temperaturesofthebestfitthinthermalplasma
: modelsareintherange0.15-0.94keVfromXMM-NewtonEPIC
v
Starburstgalaxiesareknowntoshowverycomplexemissionin and0.46-0.66keV fromChandra.Stricklandetal. (2000) con-
i
X X-rays.Thisemission originateson the onehandfromsources cludedthatthedetectedemissionoriginatesfromtheshockedre-
r that appear to be point-like sources, like X-ray binaries, su- gionattheborderoftheoutflowwherethewindcollideswithin-
a pernovae and supernova remnants. On the other hand emis- terstellarmedium.Thewinditselfthoughwasthoughttobetoo
sion comes from the diffuse hot component of the interstel- hot and too thin to be detected directly. This picture, however,
lar medium like diffuse emission in the disc and gaseous out- disregardsthepossibilitythatthewindmaybemass-loaded,en-
flows driven out of the disc by massive stellar winds and core trainingambientinterstellar medium(ISM)as well as infalling
collapse supernovae, also called superwinds. The latter phe- material.If turbulentmixingproceedson a time scale whichis
nomenon can be quite spectacular in M82 (e.g. Stevensetal. largerthantheflowtimewithinagivenregion,suchasthebase
2003) and NGC 253 (e.g. Stricklandetal. 2000), for example, of the outflow studied here, we expectsome clumpinessin the
where these superwinds emerge from a starburst nucleus. In outflow, imprinted on an overall less dense wind. As we will
NGC253,Fabbiano&Trinchieri(1984)firstdetectedthesouth- show later, this is confirmedby our analysis, which showsthat
east part of this outflow in X-rays with Einstein and called it e.g.OVIIIisnotlimbbrightened.
the”minor-axiscomponent”.WithROSAT,Pietschetal.(2000)
also detected the part of the outflow pointing into the opposite HighresolutionspectraofNGC253andM82,takenwiththe
direction.HoweverthespatialresolutionofROSATwasnotyet XMM-NewtonReflectionGratingSpectrometer(RGS)werefirst
goodenoughtolearnmoreaboutthemorphologyofthisoutflow. published by Pietschetal. (2001) and Read&Stevens (2002),
Later on, observations with XMM-Newton (Pietschetal. 2001) respectively.BothspectrashowtheLy emissionlinesfromSi,
α
and especially with Chandra (Stricklandetal. 2000) showed Mg, Ne, O, N and also their helium-like charge states. Also
thattheoutflowcanbeexplainedwithalimbbrightenedhollow bothgalaxiesshowemission lines fromFeXVIIand FeXVIII,
and M82 additionally shows lines from Fe XX, Fe XXIII and
Sendoffprintrequeststo:M.Bauer,e-mail:[email protected] Fe XXIV. In M82the line ratiosforneon,ironand oxygenare
⋆ BasedonobservationsobtainedwithXMM-Newton, anESAsci- quitedifferentcomparedtoNGC253.IngeneraltheM82spec-
ence mission with instruments and contributions directly funded by trumappearstobehotterwithtemperaturesintherangeof∼0.3–
ESAMemberStatesandNASA 1.5keV,withitscontinuummoreconfinedtohigherenergies.Its
2 M.Baueretal.:High-resolutionX-rayspectroscopyandimagingofthenuclearoutflowofthestarburstgalaxyNGC253
X-ray flux, as well as its X-ray luminosity, in the RGS energy
band(0.35–2.5keV)ishigherthanthatofNGC253.
Howeverthesespectraonlygiveacombinedspectrumofthe
nuclearsourceandtheoutflow.Inthispaperwepresentananal-
ysis where we decomposethe total spectrum of NGC 253 into
regionscontainingthenucleusanddifferentpartsoftheoutflow,
whilemaintainingthehighspectroscopicresolution.
2. Observationsanddatareduction
The nuclear region of NGC 253 was observed with XMM-
Newton (Jansenetal. 2001) during two orbits in June 2000
and June 2003 using all of the European Photon Imaging
Camera (EPIC) instruments (Stru¨deretal. 2001; Turneretal.
2001) and the two co-aligned RGS spectrometers, RGS1 and
RGS2(denHerderetal.2001),foratotalofabout190ks.The
detailsoftheseobservationsareshowninTable1.Anadditional
archival observation (Obs. id. 0110900101) could not be used
for this analysis, since the pointing of this observation was in
thenorth-westhaloofNGC253withtheresultthattheoutflow
wasnotinthefieldofview(FOV)oftheRGS.
Beforewestarttodescribetheanalysisprocedures,wewant
to place emphasis on why it is possible at all to perform the
following spectroscopic analysis. First, the nuclear outflow of
NGC 253 is an extended X-ray object, which can be spatially
resolvedbyXMM-NewtonEPICandRGS,embeddedinaneven
largerregionofX-rayemissionfrompoint-likesourcesanddif-
fuseemissioninthediscandhaloofthegalaxy.Thismeansthe
RGSdataforthecentralregionsareinprincipleaffectedbythe Fig.1.EPICPNbrightnessprofilesalongtheRGSdispersiondi-
contaminationfromthesurroundingemission,but,asshownin rectionofeachextractionregion.Thestrongpeakat0′′iscaused
Fig.1,boththenucleusandtheoutflowinNGC253aresignif- bytheoutflowemission.Inthe Centreregion,shownin Fig.2,
icantlybrighterandwellabovethe galaxyemission,so we can this is superimposedby the nuclearsourceof the galaxy.Point
expectonlyaminorcontamination.Moreoverasweshowlater, sourcesintheextractionregionsareseenassharpspikesinthe
wecanidentifyand”remove”effectsduetothediscemission. profile.A positivedistance pointsparallelto the major axisto-
Second,sincetheRGSisaslitlessspectrometer,thespectra wardsthesouth-west.Thedistancefromthegalacticmajoraxis
ofallsourcesinthefieldofviewaresuperimposedoneachother isgivenbythevaluezinthecaptionsoftheindividualextraction
onthedetector.Aspatialdisplacementofasourcealongthedis- regions.
persiondirectioncorrespondstoawavelengthshiftinthespec-
trum of 2.31×10−3 Å arcsec−1 with respectto a not-displaced
source.Sincethe outflowhasanextentofupto1.4′ in thedis-
the eventfiles. The thresholdswere 0.25ct s−1 for observations
persiondirection,thespectralresolutionislimitedto∼0.19Åat
0125960101 and 0125960201, and 0.20 ct s−1 for observation
15Å.Thisisstillconsiderablybetterthantheenergyresolution
0152020101,where the count rate was more stable during the
fromCCDdetectors.
non-flaringtimeintervals.
Due to the superposition of all sources, other bright point
sourcesintheFOVcouldcontaminatethespectrumoftheout-
flow.Howevertheeffectiveareadecreasessignificantlyforoff- 2.1.RGSspectra
axissourcesandevenacontributionfromthebrightestoff-axis
source at the bottom of Fig. 2 (X21 from Pietschetal. 2000; SincethedispersiondirectionoftheRGSwasalignedapproxi-
Tanakaetal. 2005) can be neglected. The source at the south- matelywiththemajoraxisofthegalaxywewereabletoextract
westedgeoftheoutflowinregionSE1(cf.Fig.1and2)however spectra for different adjacent regions along the minor axis, i.e.
does affect the outflow spectrum. The spectrum of this source thecross-dispersiondirection,ofthegalaxy(seeFig.2).Theex-
does not show line features (X33 in Pietschetal. 2001), so its tentofthe extractionregionsare30′′ in cross-dispersiondirec-
contribution to the RGS spectrum from this region is an in- tion.AssumingadistancetoNGC253of2.58Mpc(Pucheetal.
creasedcontinuumflux.Thisdoesnotaffectourconclusions. 1991) this corresponds to a width of the extraction regions of
We analysed the data using the Science Analysis System 375 pc. The events in these regions were additionally filtered
(SAS), version 6.5.0, together with the most recent calibration withaCCDpulseheightfiltertoselectonlythem=−1spectral
filesavailableatthetimeoftheanalysis.Themetataskrgsproc order.Thereferencepointsfortheoriginsoftheenergyscalesof
1.19.6wasusedtoprocesstheRGSdata.We firstfilteredthe the spectrawere set onthe minoraxisof the galaxy,wherethe
datasetsfortimeperiodswithlowcontaminationbylowenergy outflowhasitspeakemission(seeFig.1).
protons. Then we extracted a light curve from the background NGC253isanextendedsourceandcoversmostofthearea
regiononCCD9,thechipclosesttotheopticalaxisofthetele- of the RGS detectors. To prevent contamination in the back-
scope, thereforethe most affected by backgroundflares, to de- ground spectra we used the task rgsbkgmodel 1.1.5, which
termine the threshold count rate, which we then used to filter computesbackgroundspectrafromRGSbackgroundtemplates.
M.Baueretal.:High-resolutionX-rayspectroscopyandimagingofthenuclearoutflowofthestarburstgalaxyNGC253 3
Table1.XMM-NewtonNGC253observationlog.Observationidentification,observingdate,pointingsandorientationofthesatel-
lite,totalexposuretime(T )andexposuretimeafterscreeningforhighbackground(T )aregiven.
exp exp,clean
Nr. Obs.id. Obs.dates Pointingdirection P.A. T T
exp exp,clean
RA/DEC(J2000) (deg) (ks) (ks)
(1) (2) (3) (4) (5) (6) (7) (8)
1 0125960101 2000-06-03 00:47:36.74 -25:17:49.2 56.9 60.8 45.1
2 0125960201 2000-06-04 00:47:36.57 -25:17:48.7 57.0 17.5 7.0
3 0152020101 2003-06-19 00:47:36.89 -25:17:57.3 53.8 113.0 75.9
Fig.2.LogarithmicallyscaledEPICPN+MOSimageofthecentralregionofNGC253intheenergyband0.5−2.0keV.Thecentre
ofthegalaxyismarkedwiththeblackcross.TheD25ellipseisoverlaidinblack,brightnesscontoursandtheextractionregionsfor
theRGSspectraareoverlaidinwhite.TheblackXatthebottomoftheimagemarksthesourceX21.
To increase statistics we combined the spectra of the two rgsfluxer. The task’s description states that the fluxed spec-
RGS detectors and added up the spectra from the three obser- trumproducedbyrgsfluxershouldnotbeusedforanyserious
vations. As the position angles of XMM-Newton did not differ analysisofthedata.Ifweuseitinspiteofthiswarning,wehave
verymuchinthethreeobservations,theregionsfromwhichthe to consider the following effect that will add to uncertainties:
spectra were extracted are only slightly tilted relative to each The task rgsfluxer neglects the redistribution of monochro-
other. A difference in position angle between observations has maticresponseintothedispersionchannels,sotheintrinsicline
the effectof degradingthespatialcoincidencein the extraction broadening of the detector is not removed from the spectrum.
regions.Howeverinourcasepositionangledifferencesaresmall As we donotdetermineline positionsor line widthsand since
enoughthatwecanneglectthiserror.Acorrectionofthespec- we integrate over the whole line including its wings to derive
tra for effective area and the combination of the spectra from fluxesforindividuallines,thiseffectdoesnotrestrictouranal-
different observations and instruments was done with the task
4 M.Baueretal.:High-resolutionX-rayspectroscopyandimagingofthenuclearoutflowofthestarburstgalaxyNGC253
ysis. Additionally we are only interested in relative line fluxes then combinedinto one image. In a final step the images were
andnotinabsolutevalues. smoothedwithagaussianfilter.
To obtain acceptable statistics (> 3σ) for most of the Ingeneralthemethodisaffectedbytwodifferenteffects:(i)
emission lines while maintaining a high spectral resolution of ADopplershiftduetotheradialvelocitycomponentofanemit-
∼0.39Åinthespectrawecombinedsixchannelsatatime. tingsourcechangesitspositionintheimagealongthedispersion
For the observations the dispersion direction of the spec- direction axis. A radial velocity of 1000 km s−1 would corre-
trometerswasapproximatelyalignedalongthemajoraxisofthe spondto21.6′′atawavelengthof15.0Å.(ii)Assumingthereis
galaxy.Thereforethespectrafordifferentcross-dispersionareas notonlyonebuttwolinesinthewavelengthextractioninterval,
correspondtoregionswithdifferentdistancezfromthegalactic wewouldhavetwoimagesoftheobjectintheresultingimage,
plane (cf. Fig. 2). The spectra obtained from these regions are superimposed with an offset along the dispersion direction of
showninFig.3. 7.2′ Å−1. Forexamplethe emissioninthe two FeXVII linesat
16.780Åand17.055Åwouldbesuperimposedwithanoffsetof
2.2.RGSimages ∼2.0′.TheFeXVIIat17Åimageistheonlycasewhereweac-
tuallyhavetoconsiderthatwehavecreatedanimageusingtwo
TheRGSisaslitlessspectrometer.Sothewholeobservedtarget
lines,i.e.theFeXVIIlinesat17.055Åandat17.100Å.Inthe
withintheFOVisimagedonthedetectorplane,withanoffsetin
spectrathelinesarenotseparatedandtheyappearaboutequally
dispersiondirectionforeverywavelengthitemitsin.Thisaspect
strong, which is most likely an effect due to the low statistics.
canbeusedtoextractnarrowbandimagesforvariousemission
Accordingto theory(Meweetal. 1985)theline strengthofthe
lines.Incross-dispersiondirectiontheimageinaselectedemis-
latterlineshouldbe∼52–85%oftheformerone,dependingon
sionlineisdirectlymappedontothedetectorCCD,whileindis-
the temperatureof the plasma (1.08keV and 0.11keV, respec-
persion directionthe observedobjectis compressedinto a nar-
tively).Theseparationofthelinesis0.046Åwhichcorresponds
rowregion.Thetechniquewasfirstappliedto RGSdata ofthe
toashiftindispersiondirectionof19.9′′.Bysmoothingtheim-
supernovaremnantDEML71byvanderHeydenetal.(2003).
agewithalargergaussianwithaFWHMof20′′wecanaccount
WiththehelpofK.J.vanderHeyden(privatecommunications)
fortheerrorwemakebyusingbothlines.
we developedour own code to produce these narrow band im-
ages.Theprocedureisasfollows:Thelowbackgroundeventlists Images in the NeX, FeXVII and OVIII lines are shown
arefilteredforthewavelengthrangeofthedesiredlineandthe in Fig. 5. The number of photons that were extracted from all
”bananaregion”inwavelength-energy-spacetoexcludesecond three observationsare 806,1077,816 and 1231for the images
orderspectraandnoise.Bysettingthewavelengthrangenarrow intheNeX(11.98–12.35Å),FeXVIIat15Å(14.86–15.13Å),
enoughwemadesurethatnoneighbouringlineswouldbeinthe FeXVII at 17 Å (16.90–17.21Å) and OVIII (18.80–19.17Å)
extractionregion.Thedatasetthusobtainedhastobeconverted linesrespectively.TheFWHMofthegaussianfilterwas12′′for
intospatialcoordinatesandithastobeuncompressedalongthe NeX,FeXVIIat15Å,OVIII,and20′′forFeXVIIat17Å.
dispersionaxisusingthefollowingequationasdescribedinthe
SAStaskrgsangles:
2.3.RGScross-dispersionprofiles
sinα F
∆β= ∆φ (1)
sinβ L Togetadditionalspatialinformationofthelinedistribution,we
producedemissionlineprofilesinthecross-dispersiondirection.
with the change in the grating exit angle ∆β due to the offset ThereforeweextractedeventsfromtheRGSeventfilesbyapply-
in the angular component ∆φ of an off-axis source parallel to ingthesamefiltersinwavelengthandwavelength-energy-space
the dispersion direction, the angle of incidence α, the grating asforgeneratingtheRGSimages,butthenwebinnedthecounts
exit angle β, the focal length F, and the distance between the into 30′′ binsto matchthe extractionregionsthatwere applied
ReflectionGratingAssemblyandtheprimefocusL. forthespectra.Backgroundcountsintherespectivewavelength
The images were corrected for exposure and binned to a rangesweretakenfromthe spectrathatwereobtainedwiththe
pixel size of 0.4′′. In a next step we included a RA-DEC co- RGS backgroundmodeltask and subtractedfrom the emission
ordinatesystem. Asreferencecoordinate,we chosethe coordi- lineprofiles.FouroftheseprofilesareshowninFig.4.
nateof thecentreof NGC253(thepositionofthe brightestIR
sourceinthe galaxy,α = 0h47m 33s.3,δ = −25◦17′18′′,
2000 2000
Forbesetal.2000).Incross-dispersiondirectionthepositionof 2.4.EPIC-PNimages
our reference coordinate on the CCD could be taken directly
fromthesourcelistfilewhichwasproducedintheprocessingof To verify the results from the RGS images we also extracted
theRGSdata.Indispersiondirectionthepositionisgiveninthe EPIC PN narrow band images in approximately the same en-
above procedureby the Doppler shift corrected line centre po- ergy ranges. Therefore we filtered the PN eventfiles in the en-
sition.Thevelocitythathadtobeaccountedforisthegalaxy’s ergybandsaroundNeX(992–1052eV),FeXVIIat15Å(795–
systemic velocity of 243 km s−1 (Koribalskietal. 2004). This 844 eV), OVIII (625–690 eV) and FeXVII at 17 Å (694–
shiftsthereferencecoordinatealongthedispersiondirectionby 734 eV). The spectral resolution of the EPIC PN detector is
∼0.35′′ Å−1 times the centre wavelengthin which the image is ∼70eV,soitispossiblethatphotonswithhigherorlowerener-
calculated,e.g.6.6′′fortheOVIIIimage.Theeffectsoftheve- giescontributetotheenergybandofinterest.Additionallythere
locitiesoftheearthwithrespecttothesun,andXMM-Newton’s is contaminationfrom higherenergiesdue to the redistribution
orbitalvelocitycanbeneglectedsincetheyareonlyoftheorder inthedetector.Photonscanloseupto60%oftheirenergyinthe
of 0.7′′ and 4′′×10−4, respectively in the OVIII image where CCDbeforetheyaredetected.Thismeansthatbrightfeaturesin
theeffectwouldbelargest,andthereforemuchsmallerthanthe someenergyrangecanshowuptosomedegreeinlowerenergy
widthofthepointspreadfunction.Theimagesforthelineswere bands. The filtered eventfiles of the different observations and
created separately for each of the three observations and were instrumentsweremergedusingtheSAStaskmerge.Wecreated
M.Baueretal.:High-resolutionX-rayspectroscopyandimagingofthenuclearoutflowofthestarburstgalaxyNGC253 5
imagesofthe eventfilesandsmoothedthem with a gaussianof weak or not detectable at all. If we assume that the north-west
6′′.TheresultingimagesareshowninFig.6. and south-east outflow have a similar intrinsic spectrum, then
absorptioncanfullyaccountforthedifferencebetweenregions
NW and SE 1, at the same projected distance from the centre.
2.5.EPIC-PNbrightnessprofiles
The strongest lines in the NW region are from NeX, NeIX,
To detect the limb brightening of the outflow, as found by FeXVII at ∼15 Å and ∼17 Å, OVIII and from the forbidden
Stricklandetal. (2000) and Pietschetal. (2001), we also ex- lineoftheOVIItriplet,butmostoftheemissionoriginatesfrom
tracted brightness profiles perpendicular to the outflowing di- the discandis smoothedoutintothe broadbase, asmentioned
rection from the merged PN eventfiles of all observations. above.
Furthermore to check for an energy dependence in the limb The regions south-east (SE 1 to SE 3) of the centre do not
brighteningwesubdividedthemintoenergybinswithawidthof suffer from this absorption by the disc. One can easily follow
∼150eVstartingfrom400eVupto2000eV.Theemissionlines how the lines increase or decrease in strength as one is going
ofOVIIIandtheFeXVIIlinesareincludedintheenergyranges awayfromthegalacticdiscsouth-eastalongtheminoraxis.The
550–700eVand700–850eV,respectively.NeXismostlyinthe lines at short wavelengthsfrom silicon and magnesium are the
1000-1150eVbin.Theextractionregionsmatchtheregionswe first to disappear with distance from the centre. The NeX line
usedfortheRGSspectra.FurthermorewesplittheregionSE1 is seen to decrease considerably in strength south-east of the
intothetworegions’SE1(1)’and’SE1(2)’withawidthof15′′ Centre region, whereas the line from the lower ionised NeIX
incross-dispersiondirectionoftheRGS.Thebrightnessprofiles, is notaffected as muchandis evenstrongerthan the NeX line
sortedforenergyband,areshowninFig.7.Sincewearenotin- in regions SE 1 and SE 2. All the lines from iron decrease in
terested in the emission from the nucleus nor from the bright
strength exceptfor the FeXVII lines at 17 Å which grow by a
sourceX33south-westofthenucleus,theprofilesarelimitedto
factor of ∼1.5. Also the OVIII line increases in strength. The
amaximumof250counts.Thisstillshowsthemainfeaturesin
OVII triplet has about the same strength in region SE 1 as in
the Centre and SE 1 (1) region, while cutting off the peaks in
the Centre region. The lines from NVII and CVI increases in
somecases.
strengthcomparedtothecentralregion.
Further away from the centre, in region SE 2, the FeXVII
3. Results lines at ∼15 Å are still detectable and the OVIII line is the
strongest line in the spectrum. Also the OVII triplet is still
3.1.RGSspectra strong. The detection of all the other lines is below 2σ, even
TheRGSspectrashowemissioninmanydifferentlines(Fig.3). though the lines at wavelengths longer than 12 Å can still be
Especially in the region including the major axis of NGC 253 identified.
(Centre) a large variety of lines from different elements can In regionSE 3 only very weak lines from OVIII and from
be identified. The spectra extend from the SiXIV line at the theNeIXtripletremain.
highest energy down to the CVI line at the low energy end
Unfortunately the statistics in the spectra are not good
(λ = 6...34 Å). All the He-like ions in this range (SiXIII,
enough to allow a quantitative spectral analysis with XSPEC.
MgXI,NeIX,OVIIandNVI)andtheircorrespondingionsin
However,severalconclusionscanbedrawn.
the next higher ionisation state (SiXIV, MgXII, NeX, OVIII
and NVII) can be resolved. The iron 3d-2p transitions around Temperatureestimatescanbeinferredfromlineratiosofdif-
ferent elements or of the same element in different ionisation
15 Å as well as the 3s-2p lines around 17 Å are detected, i.e.
statesusingmodelcalculations.Assumingcollisionalionisation
their peak heights are more than twice the error in the wave-
equilibrium(CIE),Meweetal. (1985) calculated line strengths
length bin. Also the lines at ∼16.0and ∼16.1Å can be clearly
fordifferentelementsandtransitionsdependingonthetemper-
detected,butitisnotclearwhethertheline at∼16.0Å isfrom
aturesoftheplasma.Bymeasuringthefluxesoftwotransitions
FeXVIIIorfromOVIII.Intherange10Åto11.5Åthereisan
in a spectrum and taking the ratio of these, the obtained value
indicationofironlinesfromFeXXIIIandFeXXIV.
can be comparedwith the tables in Meweetal. (1985) and the
Aspectralfeaturethatisprominentinthespectraisabroad
temperatureoftheplasmacanbederived.
base at the FeXVII lines at 15 Å and 17 Å and at the OVIII
andOVIIlinepositionswithawidthofupto∼0.5Å.Ifthisline We used the line strength ratio between the Lyα state of a
broadeningiscausedbythevelocitydispersionoftheoutflowing givenelementanditshelium-likechargestatethatmatchedthe
gas,itwouldimplydeprojectedvelocitiesofupto∼40000km/s, morphologyof the Lyα state in the RGS images best. The line
correspondingto temperatures of ∼ 1011 K. Such high veloci- strengthswerederivedbyintegratingthefluxofthelineoverthe
ties are in disagreementwith outflow velocities in other galax- wavelength(cf.Table2).Wefoundthatthederivedtemperature
ies. Extreme cases show values of up to ∼3600 km/s (e.g. in valueanditserrorshowonlyaweakdependenceonhowmuch
NGC3079,Veilleuxetal.1994).A moreplausibleexplanation of the wings of the line we include in the flux integration.The
for the broad base is contributions from the disc emission of resultingtemperaturesforSi,Mg,NeandOandtheirvariation
NGC253:InFig.1discemissionextendsabout200′′indisper- along the outflow direction are shown in Table 3. Using only
siondirection(majoraxisofthegalaxy)inbothdirectionsfrom the peakheightof the line howevergivestemperaturesthatare
lower,exceptfor oxygenin the regionsCentre andSE 1, com-
thecentre.Thiscorrespondstoa∼0.5Åshiftinwavelengthand
paredtothevaluesshowninTable3byupto50%.
thereforecanexplaintheobservedeffect.
Theregionnorth-westofthecentre(NW)isstronglyaffected Usingthelinefluxfromanemissionline(cf.Table2)andan
by absorption from the galactic disc that lies between the out- estimateofthesizeoftheemittingregionwecanderiveelectron
flowandtheobserver.Pietschetal.(2000)deriveanadditional densitiesforthe nuclearregionof NGC253andthe south-east
absorbing column N of 1 − 2 × 1021 cm−2 for this position outflow.Meweetal.(1985)givethelinepowerP′normalizedto
H
north-west of the centre. Therefore most of the lines are only the electron density for differenttemperaturesand X-ray emis-
6 M.Baueretal.:High-resolutionX-rayspectroscopyandimagingofthenuclearoutflowofthestarburstgalaxyNGC253
Fig.3.CombinedRGSspectraofNGC253extractedfromdifferentregionsalongtheoutflow.Thelabelgivesthesouth-eastposition
zoftheextractionregionalongtheminoraxisrelativetothecentreofthegalaxyinarcmin.
M.Baueretal.:High-resolutionX-rayspectroscopyandimagingofthenuclearoutflowofthestarburstgalaxyNGC253 7
Table 2.Fluxvaluesfordifferentemission linesin theextractionregions.The referencesforthe expectedwavelengths(λ )
expected
areMeweetal.(1985)andPhillipsetal.(1999).
Transition λ Flux
expected
(Å) (10−6s−1cm−2)
NW Centre SE1 SE2
SiXIV 6.18 5.6±4.3
SiXIIIw 6.65 5.8±3.4
SiXIIIx+y 6.69 4.0±2.8
SiXIIIz 6.74 9.2±4.3
MgXII 7.11 6.5±3.0 1.9±1.6
MgXIw 7.76 1.4±1.3 6.5±2.7 1.3±1.2
MgXIx+y 7.81 5.4±2.4 1.7±1.3
MgXIz 7.87 6.3±2.6 2.5±1.6
NeX 12.1 3.4±2.2 16±4.5 5.7±2.6 2.0±1.6
NeIXw 13.5 4.3±2.3 5.4±2.5
NeIXx+y 13.6 3.0±2.0 6.2±2.8 6.1±2.7 2.4±1.8
NeIXz 13.7 1.8±1.6 7.6±2.8 4.1±2.2
FeXVII15 15.0 4.5±1.9 18±3.8 13±3.1 3.3±1.6
FeXVII17 17.1 3.0±1.7 13±3.8 13±3.9
OVIII 19.0 6.7±2.5 16±3.8 18±4.0 6.9±2.5
OVIIw 21.6 2.3±2.2 3.0±2.4
OVIIx+y 21.8 3.4±2.6
OVIIz 22.1 5.1±3.3 9.5±4.4 13±5.2
NVII 24.8 2.2±1.5 6.3±2.8 1.5±1.3
NVIw 28.8 1.4±1.3 4.2±2.4
NVIx+y 29.1 2.4±1.8
NVIz 29.5 3.2±2.0 1.8±1.5
CVI 33.7 3.0±2.5 4.5±3.0 6.6±3.3 5.7±3.1
Table3.Temperaturesoftheplasmafordifferentregionsofthe ForthecentralregionweselectedtheNeXlineandassumed
outflow of NGC 253 derived from line ratios of different ele- a uniform, spherical emitting region with a radius of ∼160 pc.
ments.ForregionSE3notemperaturescouldbederivedasthe The outflow in region SE 1 is best represented in the OVIII
linesaretooweak. line and we assumed a uniform, cylindrical volume with a ra-
dius of ∼200 pc and a height of 375 pc, the latter being con-
fined by the extraction region. We selected these lines because
they are strong in the spectra and because we can get a good
Region TemperatureinkeV
estimatefortheemittingvolumefromtheRGSimages.There-
Si Mg Ne O
sultingelectrondensitiesaren =0.106±0.018cm−3and
e,nucleus
NW 0.61±0.08 0.51±0.08 0.21±0.01 n = 0.025±0.003cm−3 for thenucleusandthe outflow
e,outflow
Centre 0.79±0.06 0.66±0.04 0.43±0.02 0.22±0.01 region,respectively.
SE1 0.46±0.04 0.38±0.03 0.21±0.01 Apartfromthederivationoftemperaturesandelectronden-
SE2 0.25±0.02 0.31±0.04 sities, some selected emission lines can be used as diagnostic
lines.
The FeXVII lines at 15 Å and 17 Å can be used to de-
rivetheionisingmechanismintheplasma.IntheCentreregion
of NGC 253 as well as in the regions NW and SE 2 the line
sion lines. The electron density can then be derived using the strengths indicate a predominantly collisional ionised plasma.
formula Region SE 1, however, shows an inverted line ratio. Here the
linesat17Åarestrongerthanthelinesat15Å,whichpointsat
FE 4πd2 aphotoionisedplasma.
γ
n = (2)
e s VP′ In general the helium-like line triplet of OVII can provide
theelectrondensity,theelectrontemperatureaswellastheioni-
wheren istheelectrondensity,F theflux(incountss−1 cm−2) sationprocess(Porquetetal.2001).However,werefrainedfrom
e
inanemissionlinewithE fromanemittingregionwithvolume using the OVII triplet for the following reasons: In the com-
γ
V anddistanced.Theemittingregionisassumedtobeuniform binedandfluxedspectrathesignificanceisbelow3σformostof
inn . thespectralbins.Alsotheindividuallinesarenotclearlydistin-
e
8 M.Baueretal.:High-resolutionX-rayspectroscopyandimagingofthenuclearoutflowofthestarburstgalaxyNGC253
Fig.4.NeX(upperleft),FeXVIIat15Å(upperright),OVIII(lowerleft)andFeXVIIat17Å(lowerright)profilesofbackground-
subtracted counts against cross-dispersion distance for the combined NGC 253 RGS1 and RGS2 data. The bin at distance zero
corresponds to the extraction region ‘Centre’. Negative distances are towards the north-west, and positive values correspond to
areastothesouth-east.
guishablefromeachother.Thelatter isprobablyenhanceddue aredominatedbylowstatistics.Wethereforerestrictourresults
totheuseofthetaskrgsfluxerasdescribedabove.Wealsore- andconclusionsontheregionswithgoodstatistics.
frainedfromdoingasimultaneousfitofthesingleuncombined
spectrawithXSPEC,asthestatisticsinonesinglespectrumare The strong OVIII emission clearly traces the outflow. It
barelyabove2σforthestrongestbin. reaches out to ∼750 pc projected distance along the south-east
minor axis and has an extentperpendicularto that of ∼400pc.
Because of the high absorption,emission from the farther side
3.2.RGScross-dispersionprofiles ofthediscisnotdetected.WiththeRGSspatialresolution,the
outflowinOVIIIdoesnotshowlimbbrightening.Theemission
Inthecross-dispersionprofiles(Fig.4)lineemissionisstrongest
is strongest close to its central axis and its intensity decreases
in the Centre bin at a distance of 0′′ for NeX and FeXVII at
towardsthe border.Thissuggeststhattheemittingionsarenot
15 Å. In the OVIII and FeXVII at 17 Å profile this is not the
concentratedonthebordersoftheoutflowwheretheoutflowgas
case. Here the SE 1 region at a distance of +30′′ is the bright-
interacts with the surrounding material, but that this emission
est. However the north-west half of the central bin is already
comesdirectlyfromtheoutflowinggas.Theextenttothesouth-
affectedbyabsorptionfromthedisc.Whenonebinsthebright-
east is less than the one in the EPIC images where the bright
nessprofileintosmallerspatialregions,thestrengthdropsdra-
outflowemissioncanbefollowedto∼1.3kpc(Fig.6),dueboth
maticallyinthecentralbinandtowardsthenorth-west.Towards
to the sensitivity and to the smaller energy band (∆λ =1.87 Å
the south-east, on the other hand, it only dropsslowly. Further
vs. 0.37 Å) that was used to extract the images for the RGS.
tothesouth-eastwithdistanceslargerthan45′′theNeXandthe
TheemissionfromNeXisclearlyconcentratedinthediscand
FeXVII profiles show a large drop in brightness, whereas the
nucleusanditgivesnocontributiontothe outflow.Theimages
OVIIIprofiledeclineswithalesssteepgradient.Thereforethe
intheironlinesbothshowthestrongestemissionsouth-eastof
NeX and FeXVII emissions are more concentrated within the
thenucleusandtheemissionisslightlyextendedalongtheout-
disc,whereastheOVIIIemissionextendsfartherawayfromthe
flowdirection.Howevertheydonottracetheoutflowmorphol-
disc. The general behaviour that the emission from higher en-
ogy as seen in the OVIII line. This is not surprising, because
ergylinesismoreconcentratedandnotasextendedisalsoseen
theexcitationcrosssectionsfortheironandOVIIIlineshavea
inM82(Read&Stevens2002).
different temperature dependence.Due to the work that is per-
formed when the outflow expands against the pressure of the
3.3.RGSimages ambientmedium,aswellastothedivergenceoftheflowinthe
cone perpendicular to the disc, the temperature, or more pre-
TheRGSimagesallowustoderiveadditionalinformationabout cisely the kineticenergyof the electrons,shoulddecrease (e.g.
the spatial composition of the outflow. But before we describe Breitschwerdt&Schmutzler1999)withheightzabovethedisc,
theimagesinmoredetailweneedtopointoutthattheseimages explainingtherelativeincreaseinemissionofOVIIItoiron.
M.Baueretal.:High-resolutionX-rayspectroscopyandimagingofthenuclearoutflowofthestarburstgalaxyNGC253 9
Fig.5.RGSimagesofNGC253intheNeX,FeXVIIandOVIIIlines.Thewhitelinesmarkthemajor(north-easttosouth-west)
andminoraxis(north-westtosouth-east)ofNGC253.TheimageintheFeXVIIlineat17ÅwassmoothedwithFWHMof20′′,
allotherswith12′′.Whitecontoursindicate2σ,3σ,4σ,...abovethebackground.Thewhitecrosssouth-westofthenucleusmarks
thebrightsourceX33fromPietschetal.(2000).TheRGSdispersiondirectionissuchthatwavelengthincreasesfromsouth-west
tonorth-eastasindicatedbythearrow.
3.4.EPIC-PNimages tunatelyhastoolowstatisticstoconfirmthis.Theotherimages
givenoindicationoflimbbrightening.
WefindthattheEPICPNnarrowbandimages(Fig.6)areallaf-
fectedbytheredistributioneffectofthedetector,i.e.theimage
iscontaminatedbyeventswithhigherenergiesfromoutsidethe
3.5.EPIC-PNbrightnessprofiles
energyfilterboundaries.Especiallyverybrightsourceslikethe
centralsourceofthegalaxycontributestronglytothiseffectand
In the EPIC-PN brightness profiles (Fig. 7) the peak in region
allimageswillthereforeshowthesesources.Thebrightnuclear
SE1(1)thatiscausedbytheoutflowemissioncoincidesalways
sourceisclearlyvisibleinalloftheEPICPNimages.Alsothe
withthecentre.Adoublepeakwithacentraldepressionisvisi-
brightsource∼0.5′south-westofthenucleusisclearlyvisiblein
bleinSE1(2)andSE2at700–1000eV(coveringalsotheiron
allEPICPNimages.FortheRGSimagesthelatterisonlytrue
lines).Thismaypointatalimbbrightenedoutflowinthisregion.
for the FeXVII at 15 Å image. This clearly shows the advan-
An indication of this structure in region SE 2 at 550–850 eV
tage,namelythefarbetterenergyresolution,oftheRGSimages
could be a redistribution effect of the detector (see above). At
comparedtotheEPICPNnarrowbandimages.
other energies no indication of limb brightening can be seen.
With respect to the limb brightening of the outflow emis- Theprofilesshoweitheraflatplateauorahumppeakingaround
sion, the imagein the energyrange694–734eV (includingthe thecentreoftheoutflow.ThisconfirmsthefindingsoftheRGS
FeXVII at 17 Å line) indicatesa morphologythat could result images. Starting from 700 eV region NW comes out and it is
fromalimbbrightenedoutflowatadistanceof∼0.7’awayfrom stronger than the two SE 1 regions together in energies above
thegalacticcentre.Alsointheenergyrange625–690eV(includ- 1150 eV. The strong peak in the region SE 1 (1) (blue, dotted
ingtheOVIIIline)thereisanindicationofthismorphologyat line)at∼+20′′ distanceisthebrightpointsourcesouth-westof
a distance of about1.1’.The correspondingRGS image unfor- thenucleusandnotpartoftheoutflowregion.
10 M.Baueretal.:High-resolutionX-rayspectroscopyandimagingofthenuclearoutflowofthestarburstgalaxyNGC253
Fig.6.EPICPNimagesofNGC253intheenergybandsaroundtheNeX(922–1052eV),FeXVII(795–844eVand694–734eV)
and OVIII (625–690eV) lines. The images show the region of NGC 253 that is covered by the RGS images (Fig. 5). The thin
blacklinesmarkthemajor(north-eastto south-west)andminoraxis(north-westtosouth-east)ofNGC253.Theareawithinthe
thickblacklinesgivestheextractionregioninthecorrespondingRGSlineimage.Blackcontoursindicate2σ,3σ,4σ,...abovethe
background.TheimagesweresmoothedusingagaussianwithaFWHMof6′′.
ToruleoutthatourfindingsareaffectedbyXMM-Newton’s nucleus. This change in temperature would also affect the ex-
spatial resolution we compared brightness profiles, extracted citation of FeXVII. Since both elements, oxygenand iron, are
with the same regions and energy bands from Chandra obser- fromthesamekindofsources,i.e.typeIISNe,itseemsunlikely
vation3931,to the EPICPN profiles.Therethe doublepeakis that the distribution is different, unless there is a very different
clearly detected in region SE 1 (2) between 700 and 1000 eV, clumpinessinOandFeinitially.Thereforeachangeintheoxy-
possiblyalsoupto1150eV.ItisnotvisibleinregionsSE1(1) gentoironlinefluxismorelikelyduetothechangeintemper-
or SE 2. The profiles in region NW show the same behaviour aturethanduetoadifferentradialabundanceprofile.
asinthe EPIC-PNdata.Sotakingthedifferencesinthe instru- ThelinestrengthsoftheFeXVIIlinesat15Åcomparedto
mentsintoaccountXMM-NewtonandChandragiveaconsistent the one at 17 Å indicates a predominantly collisional ionised
picture. plasma. Region SE 1, however, shows an inverted line ratio.
There, the flux in the FeXVII lines at 17 Å is enhanced com-
paredtothelinesat15Å.Thispointsataphotoionisedplasma,
4. Discussion
butnostrongphotoionisingsourcesweredetectedinthevicinity.
Therearetwoadditionalalternativestocreatethislineratio:
4.1.Lineratiosandtemperatures
A) the plasma is highly underionisedcomparedto the ionising
TheextractedRGSspectraoftheoutflowalongtheminoraxisof electrons.Thetimesincetheheatingoftheplasmawastooshort
thegalaxyshowemissionlinesfrommanyionsindifferention- to reach an equilibrium state. In these so called underionised
isationstates:theLy linesfromSi,Mg,Ne,O,Nandalsotheir plasmas inner-shell ionisation is highly operational (Kosenko
α
helium-like charge states. Additionally we see emission lines 2006) and leads to an enhanced 17.10 Å flux from FeXVII
fromFeXVIIandFeXVIII. (Doron&Behar2002).Examplesthatshowunderionisedplas-
With increasing distance from the nuclear region, the rela- masaresupernovaremnantslikeN132D(Beharetal.2001)and
tive flux in the OVIII line intensifies compared to the flux in DemL71(vanderHeydenetal.2003).
FeXVII, i.e. the line flux at longerwavelengthsincreases. The B)theplasmaisoverionised.Itcanbeproducedinfastadiabatic
same effect can also be seen in the line flux ratio of O VII to coolingof hot(T ∼ 108 K) and almostcompletelyionised gas
O VIII. For the oxygen line ratio this implies that the temper- expanding out of a superbubble (Breitschwerdt&Schmutzler
ature decreases; the gas is cooling as it flows away from the 1999). The ionisation cross sections as well as the recombina-