Slit and integral-field optical spectroscopy of the enigmatic quasar HE0450-2958

0.73 MB·English

by  G. Letawe

Astronomy&Astrophysicsmanuscriptno.aa0450˙rev1 (cid:13)c ESO2008
February2,2008
Slit and integral-field optical spectroscopy of the enigmatic quasar
⋆
HE 0450–2958.
G.Letawe1,P.Magain1 andF.Courbin2
1 Institutd’AstrophysiqueetGe´ophysique,Universite´deLie`ge,Alle´edu6Aouˆt,17BatB5C,B-4000Lie`ge,Belgium
e-mail:[email protected]
8 2 Laboratoired’Astrophysique,EcolePolytechniqueFe´de´raledeLausanne(EPFL),Observatoire,CH-1290Sauverny,Switzerland
0
0
2 Received;accepted

n
ABSTRACT
a
J
Context.InterestinthequasarHE0450-2958arosefollowingthepublicationofthenon-detectionofitsexpectedmassivehost,leading
4 tovariousinterpretations.
Aims.ThisarticleinvestigatesthegaseousandstellarcontentsofthesystemthroughadditionalVLT/FORSslitspectraandintegral
] fieldspectroscopyfromVLT/VIMOS.
h
Methods.Weapply our MCS deconvolution algorithm on slitspectra for the separation of theQSO and diffuse components, and
p
developanewmethodtoremovethepointsourcesinIntegralFieldSpectra,allowingextractionofvelocitymaps,narrow-lineimages,
-
o spatiallyresolvedspectraorionizationdiagramsofthesurroundingsofHE0450–2958.
r Results.Thewholesystemisembeddedingas,mostlyionizedbytheQSOradiationfieldandshocksassociatedwithradiojets.The
t observedgasandstardynamicsareunrelated,revealingastronglyperturbedsystem.Despitelongerspectroscopicobservations,the
s
a hostgalaxyremainsundetected.
[
Keywords.Galaxies:actives–Quasars:individual(HE0450-2958)

2
v
3
1. Introduction
4
7 Intheframeworkofasystematicspectroscopicstudyof20QSO Galaxy
3
hostgalaxiesatlowredshift(Letaweetal.2007),wediscovered
.
9 a particularly interesting object with no detected host galaxy:
0 HE0450–2958.WhilealltheotherQSOsinthesampleshowa
7 host galaxy even in the short acquisition exposures taken with
0 the ESO Very Large Telescope (VLT), that of HE 0450–2958
:
v remains undetected, both on our deep VLT optical spectra and Quasar
i on high resolution ACS images taken with the Hubble Space
X
Telescope(HST).FromtheseobservationsMagainetal.(2005,
r hereafterM05) set a deep upperlimit, implying that, given the
a
luminosityofthe QSO, the hostgalaxyis underluminousbyat Blob
E
leastafactorofsix.
HE 0450–2958, at z = 0.285, lies in a rather complex en-
1"
vironment, as shown in Fig. 1. Aside from a foreground star
which happens to be close to the line of sight, about 1.7′′ to N Star
the North-West of the QSO, two other objects are at the same
redshiftas the QSO: 1- a stronglydistorted companiongalaxy,
locatedat1.5′′(6.5kpc)totheSouth-EastoftheQSO,2-acom-
Fig.1. HST/ACS image of HE 0450–2958, obtained with the
pactblobwhosespectrumshowsprominentemissionlines,typ-
HighResolutionChanneloftheinstrument,throughtheF606W
icalforhighlyionizedgas,butwithnotraceofcontinuumlight.
filter. The image has been spatially deconvoledusing the MCS
ThisgasismostprobablyionizedbytheQSOitself,assuggested
algorithm(adaptedfromM05).
bytheemissionlinesfluxratios(M05).
Since the publication by Magain et al. (M05), several pa-
pers have proposed alternative explanations to the fact that
no stellar light is detected around HE 0450–2958. In par-
ticular, the QSO could have been ejected during a 3-body
Sendoffprintrequeststo:G.Letawe
⋆ based on observations made with VLT/UT1 (program 76.B- interaction, where a binary black hole kicked the QSO off
0693(B)), and with the VLT/UT3 (program 72.B-0268(B)) at ESO- during a merger event (Hoffman&Loeb2006). An alterna-
Paranal observatory in Chile, in addition with NASA/ESA Hubble tive ejection mechanism would be recoil due to gravitational
SpaceTelescopeobservationsfromCycle13,proposal#10238. wave emission (Haehneltetal.2006). Although this scenario
2 G.Letaweetal.:HE0450-2958:2Dand3Dspectroscopy
is supported by the recent discovery of a genuine triple QSO HE0450−2958 + Neighbourhood
(Djorgovskietal.2007), further simulations tend to weaken
these hypotheses (Merrittetal.2006). In their analysis Merritt
etal.notethatHE0450–2958 emissionlinesarenarrowerthan Galaxy
Quasar
forthemajorityofTypeIQSOs.TheysuggestHE0450–2958to
Star
beahighluminosityanalogtotheso-callednarrow-lineSeyfert
1 galaxies (NLS1s), where a lower-mass supermassive black E
hole(SMBH)accretesmatterata highrate,possiblyabovethe
Eddington limit. This is indeed supported by X-ray analyses 2.5" G1 N G1
(Zhou et al. 2007). If the empirical relation between the mass
oftheSMBHandthemassofthehostgalaxybulgeholds(e.g. Fig.2.ImmediatesurroundingofHE0450–2958.Theimageis
Marconi & Hunt 2003), one would then expect a lower lumi- the combinationof the 30-secondexposuresin B, V and R ob-
nosity host galaxy for HE 0450–2958, possibly just below the tainedattheVLT/FORS2todesigntheslitmasks.Twodifferent
detection limit of M05. Estimates of the host magnitudeupper intensityscalesareshowntodisplaythefulldynamicalrangeof
limits were also performedusinga differentmethodby Kim et theimage.
al.(2007),confirmingthevalueofM05.Theseauthorsargue,in
agreementwithMerrittetal.(2006),thatthebulgemightbejust
toofaintfordetectionwiththeobservationsmadesofar.
Otherpossibleexplanationsforthenon-detectionofthehost
are internal extinction by dust and/or the absence of a genuine
young stellar population. In both cases, the host galaxy might
be at the limit of detection of the M05 ACS images. To date, E
theonlyevidencesuggestingthepresenceofahostcomesfrom
N 1"
theradioobservationsofFeainetal.(2007).Theseauthorspoint
outthattheobservedfluxesindifferentradiobandsareconsis-
tent with star formation-induced emission and, moreover, that Fig.3. VIMOSreduceddata integratedoverthe wholespectral
they obey the Far Infrared/Radio Continuum correlation found range, for the HRblue (left) and HRred grisms (middle). For
forstarforminggalaxies. comparison we show the HST/ACS image on the right panel,
Thepresentpaperfocusesonnewlongslitandintegralfield rebinned at the same orientation and resolution as the VIMOS
spectroscopycarriedoutattheESO-VLT,withtheaimofchar- data.Intensityscalesarelogarithmic.
acterizing the physical state of the gas surrounding HE 0450–
2958 and also to attempt to detect the stellar continuumof the
hostgalaxy,ifitexists. exposureswere taken for this purpose,throughthe B, V and R
filters.Figure2showsthecombinationofthesethreeexposures,
with two differentintensity contraststo enhanceboth faint and
2. Observationsandreductions bright structures. They reveal many faint objects within a few
arcsecondsofHE0450–2958.However,asthequasarandmost
2.1.Slit(2D)spectroscopywithVLT/FORS2MXU
pointsourcesofsimilarbrightnessaresaturated,nofurtherdata
Ournewmulti-slitobservationsofHE0450–2958wereobtained processingiscarriedoutontheseimages.
with the ESO-VLT and the FORS2 instrument in the MXU
(maskexchangeunit)mode.WeusedtheG600RIgrism,cover-
2.2.Integralfieldspectroscopy(3D)withVLT/VIMOS
ingthespectralrangefrom5145to8470Å,i.e.,themajorstellar
andgasemission linesinthe redshiftedspectrumofHE 0450– Theintegralfieldunit(IFU)oftheVIMOSinstrumentwasused
2958.TheobservationsweretakenbetweenFebruaryandApril to obtain3D spectraofHE 0450–2958.Inthisconfiguration,a
2006,as5setsof6differentexposures,foratotalexposuretime spectrum is obtained for each position on a 13′′ ×13′′ field of
of 3.75 hours in dark time. Their spectral resolving power is view,every0.33′′.ThespectralresolutionisR∼2500.
R=1000. Two settings (HRblue and HRred) are used to cover the
TheslitletsoftheMXUmaskareall1′′wide,buthavevari- wavelengthrange4200Å<λ<8600Å.Unfortunately,thegap
ablelengths.AsinM05andLetaweetal.(2007)weplaceslitlets inwavelengthbetweenthetwogrisms(6150to6350Å)contains
on the quasar and on severalsurroundingstars of similar mag- the Hβ emission line, at the redshift (z = 0.285) of HE0450-
nitude. The spectra of the stars are used as PSF templates to 2958.Thespectraprovidedbytheopticalfibersfedbythetele-
remove the quasar light using the version of the MCS algo- scopeareorganizedinfourquadrantscontaining20×20fibers
rithm (Magainetal.1998) for spatial deconvolution of spectra each.
(Courbinetal.2000). The orientation of the slits is selected to The observationsconsist of 7×40 min exposureswith the
allow the simultaneousobservation of the quasar, the compan- bluegrisminDecember2003,andof11×19minwiththered
iongalaxyandtheforegroundstar.Theremainingslitsareused grism, taken between December 2003 and February 2004.The
to observethe surroundinggalaxies. The seeing duringthe ob- seeingvariedfrom0.45to0.9′′intheblue(0.63′′onaverage)and
servationsvariedfrom0.5to1.3′′. from0.53to1.2′′ intheredgrism(0.76′′onaverage).
Thebasicreductionsareperformedwithtasks,leading At the time of observation,the HRred grism was notavail-
toflatfielded,wavelengthcalibratedandskysubtractedspectra, ableonthefourthquadrantandwasreplacedbyHRorange,cov-
with1.5Å/pixelinthespectraldirectionand0.126′′/pixelinthe ering a bluer wavelength range and with a different response
spatialdirection. curve.Thiscomplicatestheanalysisof the fourthquadrant.As
Inadditiontotheslitspectra,severalimagesofthefieldwere aconsequence,only7outofthe11HRredexposureswereuse-
taken with FORS2 to design the slit masks. Three 30-seconds fulduetoprobleminthematchingofHRredandHRorangeon
G.Letaweetal.:HE0450-2958:2Dand3Dspectroscopy 3
the edges of the fourth quadrant, located at the NW corner of
thefield(seeFig.3).Inthisfigure,weshowtheVIMOSfieldof
view in the two settingsand we compareit with the HST/ACS GQalso Blob
imagedegradedtotheresolutionandpixelsizeofVIMOS.The Star
VIMOSPSFishighlyelongated.Sinceourreductionprocedure
(see below)correctsfor atmosphericrefraction,this elongation
isprobablyduetoatrackingproblem. Hβ [OIII] [OIII] New
The observations are reduced using various pipelines and detections
routines, taking advantage of the specific capabilities of each R1
package. The main steps of the reduction can be summarized
Gal
asfollows: Qso
Star R2
– The cosmic-ray rejection is carried out on the raw data by
applyingthefilter/cosmictaskoftothe2Dspectra
ineachofthefourquadrants. 1" λ R3
– Thebiassubtraction,flatfielding,wavelengthcalibration,ex-
tractionofthespectraforallfibers,andfluxcalibrationare Fig.4.CombinedFORS2spectrumofHE0450–2958,inthere-
doneusingtherecipesfromtheVIMOSPipelinepro- gion where the deconvolution shows acceptable residuals. The
videdbyESO.ThereduceddatawiththeHRbluegrismhave point sources (QSO+star) have been removed. The spectra are
a spectral sampling of 0.54Å/pixeland the data taken with shown with different intensity scales in the two panels. The 3
theHRredgrismhave0.58Å/pixel. newemissionlineregionsareindicated.Thesourceoftheirion-
– A basic 2D sky subtraction is carried out with the izationistheQSOradiation.
backgroundtask,bylinearinterpolationoneachspec-
tralelementatpositionsdevoidofsources.
– Whenever a strong sky emission line is not removed well
enoughbythetask,aroutineisappliedtoclean
thespectrumbylinearinterpolationbetweentheregionsun-
affectedbytheseskyemissionlines.
– routinesarethenusedtore-organizethespectrainto
3D data cubes of 40 x 40 spatial resolution elements each
containinga1Dspectrum.Weusetheword“spaxels”inthe
following,whenreferringtoeachofthe40×40elements,to
avoidconfusionwithpixels.
– Thereducedexposuresarethenalignedusingatask,
whichshiftsthespatiallocationofthequasarforeachspec-
tralelementtothecenteroftheIFUfield.Thisensuresthat
atmosphericrefractioniscorrectedineachexposure.
– The aligned exposures are co-added with appropriate
weighting,takinginto accountthe differentcoverageofthe
skyduetooffsetsbetweenexposures.
– The resulting combined data cubes are smoothed using
a Gaussian kernel of 2.5Å full width at half maximum Fig.5. FORS2 spectrum of the late-type galaxy G1 shown in
(FWHM). Fig.2.Themeasuredredshiftisz=0.2866
3. Slitspectroscopy of the QSO and host cannot be achieved in this case: the see-
ing during the observations, worse than originally planned, is
BecauseoftheveryhighcontrastbetweentheQSOanditspu- of the same order as the slit width, resulting in a loss of infor-
tativehostgalaxy,particularcaremustbepaidtothedecompo- mationonthe wingsof thePSF. Thiseffectis enhancedbythe
sitionofthedataintopoint-likeandextendedcomponents,both combined effects of a slight slit misalignment with respect to
intheslitandIFUspectra. thepointsources,andofdifferentialatmosphericrefraction.The
The FORS2/MXU spectra of HE 0450–2958 are all consequence is that the QSO and the PSF stars are clipped in
deconvolved using the MCS algorithm for slit spectra different ways, making it impossible to build reliable PSFs. In
(Magainetal.1998; Courbinetal.2000), allowing a sepa- addition, the overallquality of the PSF does not allow simula-
ration of the QSO light from the underlying extended objects. tionstobeperformed(seeM05)inordertoplaceausefullimit
This method has been used successfully on many occasions, onthemagnitudeofthecontinuum.
either on data of QSO host galaxies (Courbinetal.2002; We show in Fig. 4 a representative portion of the data, be-
Letaweetal.2004; M05; Letaweetal.2007) or on data of tween Hβ and [OIII], where the deconvolution process is ac-
gravitationally lensed QSOs (e.g., Eigenbrodetal.2006; ceptable, i.e., a region where the remaining flux after removal
Eigenbrodetal.2007). ofthepointsourcesisnon-negative,andwheretheresidualsof
The goalof the presentobservationswas to go deeperthan the deconvolutionprocess do not present obviousfeatures of a
in M05, to attempt the detection of a stellar continuum in the misadapted PSF. The image is the combination of all 5 of our
putativehostgalaxyandtodeterminetheionizationstateofthe individual, deconvolved,spectra. Even with bad deconvolution
gas far away from the QSO. Unfortunately,optimal separation residuals, as shown by the consequent noise at the position of
4 G.Letaweetal.:HE0450-2958:2Dand3Dspectroscopy
the point sources, and thanks to the deep exposures,three new fibers,theweightingisallowedtovarylinearlywithwavelength,
emission-line regions are detected at large distances from the foreachspaxel.
QSO(seeFig.4).
The relative weighting of the QSO and of the stellar tem-
The three regions R1, R2, and R3, are located at 2.6, 1.5
platesareconsideredassatisfactorywhenthereisnotraceleft,in
and5.9arcsecondsfromtheQSO.Thiscorrespondstoprojected
agivenspaxel,oftheBalmerbroademissionlinesfromtheQSO
distances of 12, 7 and 27 kpc respectively, at the redshift of
andofthe(broad)absorptionlinesforthestar.Visualinspection
HE0450–2958.WhileregionsR1andR3hadnotbeendetected
oftheresidualsiscarriedoutforeveryspaxelcontaminatedbya
previously,regionR2wasreportedinFeainetal.(2007),onthe
pointsource,allowingustomakesurethattheresultofthesub-
basis on our FORS1/MOS analysis (M05). It is almost super-
tractiondoesnotshowsignsofover-orundersubtraction,atany
imposed on the foregroundstar lying to the North-West of the
wavelength.Thisprocedureleavesuswithadatacubecontain-
QSO.
ing only the companiongalaxyand any other extendedobjects
Only emission lines are seen in the 3 regions, despite the
suchasregionsofionizedgas.
depth of the observations. In the 3 cases, the [OIII] doublet is
detected while the Hβ line is absent. For region R1, this con- Themethoddescribedabovereliesontheimportantassump-
verts into a lower limit of [OIII]/Hβ > 7.7, and for region R2, tionthattheQSOtemplateactuallycontainsonlylightfromthe
[OIII]/Hβ > 28.8. The third region is so faint that the lower QSO,whichmightnotbethecase.Ourtemplateprobablycon-
limitderivedisirrelevant.Followingtheionizationdiagramsof tainsasmallbutnon-zerocontributionof(narrow)gasemission
Veilleux&Osterbrock(1987),aratioabove6cansafelybeas- linesnotbelongingtotheQSOitself.Wemaythereforesystem-
sociated with ionization by QSO radiation. Our FORS2 obser- atically oversubtract the emission lines in each spaxel, in par-
vations therefore indicate that the source of ionization in these ticular for the “blob”, which lies angularly close to the QSO.
regions is the QSO itself and, as a consequence, that ionizing Moreover,thesubtractionsignificantlyincreasesthenoiseinthe
photonscan escape theQSO surroundingsto ionizegasclouds regionswhere the QSO or the star dominatethe flux, i.e. close
asfarasatleast7and12kpc.Thisgivesfurthersupporttothe tothecentersoftheirPSFs.
ideathat,ifdustispresentintheimmediatesurroundingsofthe
The adopted procedure for point source subtraction being
QSO,itsgeometryissuchthatUVradiationcanescapefreelyin
based on visual inspection of the removal of spectral point
severaldirections:towardsus(nodetectedreddeningoftheQSO
sourcecharacteristics,theerrorbarsaredifficulttoquantifypre-
andblobspectra,M05)andtowardstheseemissionlineregions.
cisely. However, as we aim to subtract the broad components
There is still no detection of a host galaxy in these spectra
withnoconditiononnarrowones,thepresenceofnarrowemis-
but, as previously mentioned, the quality of the deconvolution
sionlinesinthesubtractedspectraisreliable,andmightjustbe
processisnotoptimal.However,acomparisonofthefluxinthe
slightlyunderestimatedbythepossiblesubtractionofpartofthe
companion galaxy and the remaining flux under the QSO (i.e.
blob spectrum along with the quasar template. The continuum
the putative host) reveals that the host should at least be two
slope mightalso be affected by a wrongweighting of the tem-
magnitudefainterthanthecompaniongalaxy,inthewavelength
plate spectra, so we did not use these slopes to determine any
range between [OIII] and Hα (equivalentto the observed filter
propertiesoftheemittingregions.
R). This would give a magnitude for the host of M ∼ −21,
V
compatible with M05 and Kim et al. (2007) estimates. Given
thatmostofthefluxdetectedundertheQSOislikelytobedueto
deconvolutionartefactscausedbyaninaccuratePSF,theputative 4.2.Extractedspectra
hostmightbemuchfainterthanthatupperlimit.
Finally, some of the slits used for these observations were TheresultoftheQSO+starsubtractionisdisplayedinFig.6.In
placed on galaxies close to the line of sight to HE 0450–2958. thisfigure,wealsodividethefieldofviewintodifferentregions
One of them is galaxy G1 of Fig. 2, for which we obtain the whereweintegratethesignal.Threemainareasaredefinedand
spectrum shown in Fig. 5. The redshiftof this late-typegalaxy thensubdividedintosmallerareas:theregionofthecompanion
is close to that of the quasar, z = 0.2866. The projected lin- galaxy,themorecentralregionofthe“blob”,andseveralregions
eardistancebetweenG1andthequasaris44kpc.Nootherre- furtherawayfromtheQSO,whichalsoturnouttocontainion-
latedgalaxywasfoundintheseandthepreviousFORS1spectra izedgas.
(Letaweetal.2007).
4.2.1. Thecompaniongalaxy
4. Integralfield(3D)spectroscopy
Thecompaniongalaxycanbesubdividedintothreeareaswhich
4.1.Subtractionofthepointsources
we name Gal1 to Gal3 (Fig. 6). The corresponding 1D spec-
Integralfieldspectroscopy(IFS)allowsinvestigationoftheim- tra are displayed in Fig. 7. While the spectrum of region Gal1
mediate surroundings of the QSO, provided its own spectrum isdominatedbyyoungstars(importantcontinuum,weak[OIII]
canbesubtractedfromthedata.Becauseofthelackofisolated lines,broadHydrogenabsorptions),regionGal2isamixofstar
starsintheVIMOSfieldofview,andbecauseoftheelongation andgas,andregionGal3isalmostexclusivelymadeofionized
ofthePSF,deconvolutionofthedata,asdonewithslitspectra,is gas. As the smearing of light is of consequencein the VIMOS
notpossible.Instead,weuseanotherapproach,takingadvantage data, we deliberately integrated the signal over large regions,
ofthe3Dinformationwehaveatourdisposal. which have about the length-scale of the seeing. The match-
First,weselectareferencespectrumatacentralpositionfor ing between the regions and the HST/ACS broad-band image
eachpointsource,i.e.,theQSOandtheforegroundstarofFig.1. isshowninFig.6.Clearly,regionsofhighlyionizedgasextend
The two spectra are used as templates to be removedfrom the beyondthebroad-bandimageofthegalaxy:thereisbasicallyno
otherspaxels,afteradequateweighting.Asthereisaclearvaria- light in the HST/ACS image of the Gal3 region, as well as no
tionofthegrism+detectorresponsefunctionaccrossthearrayof continuuminitsspectrum.
G.Letaweetal.:HE0450-2958:2Dand3Dspectroscopy 5
Fig.6. Spatial regions over which we integrate individual 1D spectra. The regions are shown in overlay on the deconvolved
HST/ACS image (left panel), and on the VIMOS image reconstructed from the full datacube by integrating over the full wave-
lengthrange(middlepanel).TheimageontherightistheresultofthesubtractionoftheQSO andofnearbystar spectra.Inthis
case, the wavelength rangearoundthe redshifted [OII]emission line has been selected. The central30 spaxelsof the QSO, very
noisyaftersubtraction,arenotusedwhenintegratingthesignalovertheregions.
4.2.2. Theblob WeshowinFig.10acolor-coded2Ddiagnosticdiagramfor
theregionsinthevicinityofHE0450–2958.The3-colorscalein
Theregionoftheblobisprobablythemostdifficulttointerpret
therightpanelgivestheionizationsourceasafunctionofcolor.
on the VIMOSdata, as it is locatedveryclose to the reference
The three main sources of ionization are labeled in the figure
fibertakentomodelourQSOtemplatespectrum.Althoughthe
as “AGN” for ionization by strong power-law source (here the
subtraction process may bias the spectrum of the blob (essen-
QSO),“HII”whentheionizationisduetostars,and“shock”for
tially by removingpart of it together with the QSO spectrum),
shock-inducedionization.
thepresentresultssupporttheM05findingsthattheblobismade
In order to be accurate, the line ratio must be corrected for
onlyof gasand showsno trace of a stellar continuum.The 1D
absorptionbydust.ThelackofanHβlineinthesespectramakes
spectrumofthisregionisshowninFig.8.
it hard to carry out this correction and hampers the use of the
Notethatevenifoursubtractionproceduredimmedthecon-
line ratio proposedby Veilleux&Osterbrock(1987). However,
tinuumof the blob,it hasalso dimmedthe emission lines. The
our diagnostic diagram is constructed using the [NII]/Hα
fact that the latter are still very bright after subtraction of the
and[OII]/[OIII]lineratios,aspresentedinBaldwinetal.(1981).
QSOspectrumshowsthatwehavenotcompletelyremovedthe
Withtheseratios,anyreddeningleavesthe[NII]/Hαunchanged,
signal.
asthetwolinesarealmostatthesamewavelength,butdecreases
the [OII]/[OIII] ratio: [OII] being at shorter wavelength than
4.2.3. Emissionlineclouds [OIII],itismuchmoreaffectedbydustabsorption..
ThefourregionslabelledEm1toEm4inFig.6aredominatedby The vastmajority of spaxelsare foundin the AGN-ionized
emissionlines.Theirintegrated1DspectraareshowninFig.9. region (blue dominant). They concentrate in the blob, and in
All emission lines are real, as no appropriate weighting of the theregionsGal2/3andEm3/4.TheSouthernpartofthesystem
subtractednuclearspectracanmakethemdisappear,evenwhen (Gal1/Em1) is ionized by shocks and stars. Following Balmer
trying to oversubtractthe templates. At the depth of the obser- decrements evaluated on FORS1 slit spectra (M05), the com-
vations, only Em1 contains both gas and stars, as can be seen panion galaxy is the only place where dust reddening is non
negligible. However, a correction for this extinction, increas-
from the non-negligible4000Åbreak. This regionmightbe an
ing [OII]/[OIII],would not affect the dominance of the shock-
extensionofthecompaniongalaxy.Thethreeotherregionsare
inducedionization.
compatiblewithgasemissiononly.TheweirdcontinuumofEm
3iscausedbyitsproximitytothecenteroftheforegroundstar, As the ionization source of the blob is clearly the quasar
thesespaxelsbeingmoreaffectedbythePSFvariationwithpo- light, the possible oversubtraction of a part of the blob (nar-
sition and wavelength. Nothing can be seen at the position of row component) along with the quasar template spectrum (as
Em2, Em3 and Em4 in the HST/ACSimage (Fig. 6, left). Em3 explainedinSect.4.1)possiblylowersthehardionizationacross
probably corresponds to the faint gaseous emission R2 in the the entire system. The dominance of AGN-caused ionization
VLTMXUspectraofFig.4. measured in the whole system may thus be even stronger than
suggestedbyFig.10.
Theonlyregionswheretheionizationbystardominates,are
4.2.4. A2Ddiagnosticdiagram
the region Em2b and the outer edge of region Gal1, where the
Diagnostic diagrams such as the ones introduced by signal is very weak. Even if we find almost no continuum in
Veilleux&Osterbrock(1987) and Baldwinetal.(1981) al- theweakEm2region,thismightindicatethepresenceofyoung
low identification of the nature of the source of ionization in a starsevenoutsidethecompaniongalaxy.Thedistributionofthe
gascloud.Thesediagramsinvolveemissionlineratiossensitive greenshock-inducedionizationregionsinFig.10,oneachside
to the hardnessof the ionizingflux.Our VIMOSdata allow us oftheQSO,suggeststhatthestarswereformedduringaviolent
tomapsomeoftheseratiosaccrossthefieldofHE0450–2958. event,probablytriggeredbytheQSO(ortogetherwiththeQSO)
6 G.Letaweetal.:HE0450-2958:2Dand3Dspectroscopy
Fig.7. Spectra of the companiongalaxyin the HRblue and HRred VIMOS grisms. For the two settings, the spectrum in the top
panelisthesumofthespectraforthe3regionslabeledGal1toGal3inFig.6,andshowninthelowerpanels.
Fig.8. Spectra correspondingto the “blob” region in the HRblue and HRred grisms. The spectra are shown before and after the
QSOisremoved,intheupperandlowerpanelsrespectively.
Fig.9.SpectracorrespondingtotheregionslabelledEm1toEm4inFig.6,alldominatedbyionizedgas.
G.Letaweetal.:HE0450-2958:2Dand3Dspectroscopy 7
1.5
Shock
AGN
) 0
a
H
/
I
I
N
(
g
o
l
HII
−1.5
1" −1.6 0 1.6
log(OII/OIII)
Fig.10.Color-codeddiagnosticdiagramconstructedfromtheVIMOSemissionlineimagesaftersubtractionoftheQSOandofthe
nearbystar,asintherightpanelofFig.6.ThepositionsoftheQSOandofthenearbystarareindicatedbywhitediamonds,while
thecompaniongalaxyandthe“blob”areindicatedbywhiteellipses.Thecolorcodeintheimageisdefinedintherightpanel.The
fieldofviewis7×7′′,i.e.,azoomonthecentralpartoftheVIMOSfield.Theimagehasbeenslightlysmoothed,andweightedat
eachpositionbythecorrespondingemissionintensity.
andnotinaglobal,continuousstarformationinthecompanion
galaxy. Continuum [OII]
Finally, note that the spectra of the different Em zones are
very faint, and that Hα and [NII] are not well separated in
thoseregions(seeFig.9),resultinginlargeuncertaintiesonthe
measuredratios.Still,themeasured[NII]/Hαwouldhavetobe
wrongbyanorderofmagnitudetochangeourmainconclusion
thatthegasisionizedbyahardradiationfield,probablythatof
theQSO.
4.3.Emissionlineimages [OIII] Hα α + [NII]
We use our VIMOS data to produce narrow-band images of
HE0450–2958,centeredonseveralemissionlines:[OII]3727Å,
[OIII]4959+5007Å, and Hα6563Å + [NII]6583Å. For each
line, we subtract an image in the continuum around the emis-
sionline,afterappropriatenormalization.We alsoshowanim-
ageofthecontinuumbetween[OII]andHβ,awavelengthrange E
dtoeuvrosidofofemanisysieomnislisnioesnilminaegs.esWoenotvheerldaeycoonnvFoilgve.d11HtShTe/AcoCnS- (cid:0)(cid:0)(cid:0)(cid:0)(cid:0)(cid:0)(cid:0)(cid:0)(cid:1)(cid:1)(cid:1)(cid:1)(cid:1)(cid:1)(cid:1)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1) N
data, where pointsourcesare indicatedby dots. We set to zero
the noisy pattern near the position of the QSO, because of the
Fig.11. Contours plot for the continuum VIMOS image (see
largeuncertaintyduetotheremovalofitsspectrumatthisloca-
text) and for different emission lines. The contours are shown
tion.
inoverlayoftheHST/ACSimage,aftersubtractionoftheQSO
Weobserveaslightshiftbetweenthepositionoftheblobin andofthenearbystar.Wealternatesolidanddottedcontoursfor
theHSTimageandintheVIMOScontourplots.Thereasonfor clarityofdisplay.Thestepsbetweencontoursareevery20%of
thisistwofold:first,aswetakeareferencespectrumatthecen- themaximumintensityofeachimage.
tralpositionforPSFsubtraction,weremovepartoftheunderly-
ingextendedcomponent(i.e.theblob)togetherwiththequasar
lightatthosepositions.Second,thespatialresolutionofVIMOS
ismuchlowerthanthatofACS,sosmallemissionregionsinthe
HST data are spread on muchlarger areasin the VIMOSdata. of the blob. Only small and faint structures are seen in the
OnespaxelintheVIMOSdataisaboutthesizeoftheblob. vicinityoftheQSOandareallcompatiblewiththenoise.
– The [OII] emission is presentin the companiongalaxy,the
Themoststrikingdifferencesbetweenthespatialdistribution
blob,andextendsallaroundHE0450–2958.Thislinearises
inthevariousspectralbandsaregivenbelow:
both in star forming regions and in gas clouds ionized by
AGNlight.Itsextendeddistributionindicatesthatthesystem
– The continuum light nicely follows the ACS image of the seemsgloballyembeddedin gas. Note thattheeffectis not
companion galaxy. No continuum is found at the position due to an arbitrary choice of the intensity scales: although
8 G.Letaweetal.:HE0450-2958:2Dand3Dspectroscopy
the [OII] envelope is fainter than the [OIII] emission (see,
e.g.,Fig.9),itextendsmuchfurtherawayfromtheQSO.
– The[OIII]doubletneedsharderionizingphotonsthan[OII],
suchasionizationbyanAGN.InFig.11,[OIII]isessentially
present in the blob, in the Northern part of the companion
galaxy,aswellasclosetothepositionoftheforegroundstar.
– TheHαdistributionismoreextendedthan[OIII]inthecom-
paniongalaxy,probablybecauseofongoingstarformation.
Theblobisalsovisibleinthisline.
4.4.Velocitymaps
The velocity maps of Fig. 12 are constructed by fitting a
Gaussian profile to the emission lines in each spaxel. The ve-
locitiesarethengivenrelativetothecompaniongalaxyaverage
velocity,asestimatedfromtheGal1emissionlines.Thevelocity
contoursobtainedin thisway are then slightlysmoothed.With
thesettingoftheVIMOSobservations,wecanmapboththegas
andstellarvelocityfields.
The bestlinestotrace the gasin ourobservationsare [OII]
and [OIII]. The Hα velocity map is noisier because of the
high nucleus-to-host ratio in this line. Spaxels in the immedi-
Fig.12.Velocitymapsforthestellarandgascomponents,over-
ate surroundingof the quasar position are masked. The veloci- laidontheHST/ACSdeconvolvedimage.Thevelocitycontours
ties derived from these lines range from –400 to +150kms−1,
arerelativetothecompaniongalaxysystemicvelocity.Theblue
with respect to the companion galaxy (or from about –300 to
dashedcontoursarenegative,andtheredsolidcontoursarepos-
+300kms−1, with respect to the QSO). The gas can be traced
itive.Topleft:CaIIabsorptionlinevelocitycontours,from–200
veryfarawayfromtheQSO,evenwherenolightisseeninthe to +200kms−1 with steps of 55kms−1. Top right: [OII] emis-
HST image, as already seen in Fig. 11. Even with discrepan-
sion line velocities, where the contours are drawn from –190
ciesbetween[OII]andHαvelocityfields,theglobaltrendisto to +160kms−1 by steps of 50kms−1. Bottom left: the [OIII]
showpositivevelocitiesonbothsidesofthegalaxy,andnegative emission line, with contours from –290 to +160kms−1 with
elsewhere.The[OIII]velocityfieldisslightlydifferent,globally steps of 50kms−1. Bottom right: Hα emission, from –290 to
positiveonthe whole upperpartofthe field and negativeWest +160kms−1withstepsof60kms−1.Seetextforinterpretations.
ofthequasar.Thisreinforcesthehypothesisofahighlyionized
gas cloud traced by [OIII], detached from the galaxy to which
[OII]andHαaremorerelated.
large distances from the nucleus because of the very irregular
Theextractionofthestellarvelocityfield,fromtheCaIIab-
gas(andpossiblydust)distributionaroundtheQSO.
sorption lines, is restricted to the companiongalaxy.Our mea-
Theimmediatesurroundingsof the QSO obviouslycontain
surementsarepresentedinFig.12(topleft),rangingfrom–150
somegas(suchastheblob),butarealsosufficientlytransparent
to300kms−1.Thedifferencewiththegasvelocityfieldisstrik-
tolettheQSOradiationfieldescapetolargedistances.
ing, with almost reverse velocities, positive between the posi-
The detection of an extended gas distribution in the [OII]
tionofthegalaxyandthequasar,andnegativeelsewhere.Taking
3727Å line tends to weaken the hypothesis that the QSO host
intoaccounttheelongationoftheVIMOSPSF,therotationaxis
galaxy might remain undetected because of strong absorption
probablymatchesthemajoraxisofthegalaxy.
by dust. Indeed,if the whole hostwas embeddedin (orhidden
Clearly, gas and stellar dynamics in the companion galaxy
behind)ahugedustcloud,itisdifficulttounderstandhowradia-
arenotrelated.Eitherthegasandthestarsfromwhichwemea-
tionatsuchshortwavelengthsmightbedetectedthroughoutthe
surethedynamicsarelocatedinphysicallydifferentregionsseen
system.
atthesamepositionontheplaneoftheskyduetoprojectionef-
Deepinfraredobservations,aimedattestingthishypothesis
fects,orthegasandthestellarspectralfeaturesareformedinthe
arecurrentlyunderanalysis.Howeverthepresenceofdustcan-
same location. The first hypothesiswould fit with the idea that
not be ruled out for the moment, and in this case the radiation
HE 0450–2958is embeddedin a largecloud of gasionizedby
of the quasar mightevaporatethe dust in some loci, producing
theQSO radiationfieldatverylargedistances.Thesecondhy-
patchyandhighlyionizedgasclouds.
pothesiswouldratherindicateaviolentshockbetweengalaxies
withthepossibledisruptionofthehostgalaxy. Whatever the origin of the gas patches, a possible source
ofionizationin additionto the quasarradiationitself is a radio
jet/cloudinteraction,asalreadyproposedbyFeainetal.(2007).
5. Ionizedgasandradiojets InFig.13wereproducetheir6208MHzATCAradiodata,over-
laidontheHSTimageandontheVIMOS[OIII]image.Ifdust
IonizedgasispresentinalargevolumearoundHE0450–2958, was present on the path of the radio jets and lobes, appropri-
as shownbothby the slit andIFU observations.Two questions ate correction for reddening might shift the ionization source
thusarise:(1)wheredoesthegascomefrom?(2)howcanitbe forthese regionsfromAGN to shock-inducedionization,these
sohighlyionizedasfarawayas30kpcfromtheQSO? shocksthenbeingcausedbythejet.AsseenfromFig.13,there
Bothquestionsmayfindananswerinthecontextofacolli- isaclearalignmentbetweentheradiolobesandthe[OIII]emis-
sionbetweenthecompaniongalaxyandtheputativequasarhost, sion. The alignment is found only between the radio data and
dispersingthegasandallowingthehardradiationfieldtoreach the [OIII] data, rather than with [OII] and Hα, suggesting that
G.Letaweetal.:HE0450-2958:2Dand3Dspectroscopy 9
8. Severaloftheseobservationsmayfindanexplanationinthe
contextofacollisionwhichwouldhavedispersedmatterin
alargevolumearoundtheQSOandcompaniongalaxy.
Since HE0450-2958 is a strong infrared emitter ( IRAS
pointsourcecatalog,deGrijpetal.1987;Lowetal.1988),dust
shouldbe presentin its surroundings,but the spatialresolution
oftheIRASobservationsusedforthisidentificationdoesnotal-
lowustolocatepreciselyitsorigin.Kimetal.(2007)explained
whythisstronginfraredemissioncannotbecausedbystar for-
mationinthecompaniongalaxy:IftheIRASmeasuredinfrared
fluxesweretobeassociatedwithstarformation,theywouldim-
plyastar formationrateofalmostathousandM /yr,whilethe
⊙
strength of the [OII]3727λline in the optical FORS spectra of
Fig.13. Radio map obtained from the ATCA observations of
the companion galaxy only gives a maximum of ∼ 10 M /yr.
⊙
Feainetal.(2007), overlaid on the deconvolved HST image
Accordingly,thestronginfraredemissionisthusmostprobably
(left) and on the VIMOS [OIII] image (right). Note the strong
not associated with the companion galaxy, but rather with the
correlationbetweentheradiostructuresandthe[OIII]imageand
quasaritself.Asimpleexplanationwouldbeadusttorusaround
themisalignmentwithrespecttothecompaniongalaxy.
theaccretiondisk.Thefactthattheblobisunreddenedsuggests
thateitherthistorusisquitecompact(atmostafewhundredpar-
secsinordernottocovertheblob)ortheblobliesinfrontofit.
theradiojetsarespatiallyrelatedtohighlevelionizationclouds, The dusttorusshouldalso beinclined atan intermediateangle
andnottostarformingregions.However,asthefluxusedinthe withrespecttotheline-of-sight.Indeed,anedge-ontoruscanbe
radio-to-infraredfluxratioestimatedbyFeainetal.(2007)con- ruledoutasitwouldmasktheQSO.Similarly,anearlyface-on
cerns only the central peak of radio emission, this leaves their toruswouldhardlybecompatiblewiththefactthattheQSOUV
conclusionsbasicallyunchanged. radiationreachesprojecteddistancesaslargeas30kpc.
In addition,the strongspatialcorrelationbetweenthe radio ThefactthattheUVradiationfieldoftheQSO,afterhaving
jet and the [OIII] emission suggests that the jet is either the left its immediate surroundings,can escape freely to large dis-
source of ionization or an efficient cleaner of the surroundings tances suggests that, at the kpc scale, the QSO neighbourhood
oftheQSO,allowingitsradiationtoescapefromthecentralre- isbasicallytransparentinseveraldirections.Evenifwehaveno
gionandtoionizemuchmoredistantgasclouds. evidence for the complete absence of a host galaxy, this trans-
parency is much easier to explain if the QSO does not lie in a
massive host. Indeed, if such a galaxy was present in an envi-
6. Discussion ronmentsostronglyperturbedbygravitationalinteractions,one
would expect starburst regions with large amounts of gas and
Our new VLT 2D and 3D spectroscopic observations provide
dust, as can be seen in the companion galaxy. Such a medium
furtherevidencethattheenigmaticQSOHE0450-2958liesina
wouldnotlikelybetransparentenoughtoexplainthelongrange
stronglyperturbedenvironment.
ionization.Thepresentobservationsmightthusreinforcethehy-
pothesisofanundermassivehost,ifanyatall.
1. Theputativehostremainsundetected,thecompaniongalaxy
Ontheotherhand,theanalysisofradiobyFeainetal.(2007)
isstronglydistorted,withagasvelocityfieldapparentlyun-
seems to pointtowardssome star formationaroundthe quasar,
relatedtothestellarone.
leavingthepossibilityofadustenshroudedhostgalaxy,explain-
2. The whole system seems embedded in ionized gas, as best
ingitsnon-detectionintheoptical,adustmediumwhichcould
shownbytheextendeddistributionofthe[OII]emission.
havebeensweptinseveraldirectionsbytheradiojets,allowing
3. RadioemissionshowsapeakatthepositionoftheQSO as
theobservedlongrangeionizationbythepowerfulquasar.
well as two roughly symmetrical lobes, one of them being
locatedclosetothecompaniongalaxy.
4. Severalcloudsofhighlyionizedgas(thosewiththehighest Acknowledgements. G.L.isateachingassistantsupportedbytheUniversityof
Liege.GLandPMacknowledgesupportfromProdex90195(ESA/PPSScience
[OIII]/[OII]ratio)arefoundalongtheradioaxis,extending
Policy, Belgium). FCis supported bythe Swiss National Science Foundation
fromtheimmediatevicinityoftheQSO(i.e.the“blob”)up (SNSF).
tomorethan30kpc.
5. These gas clouds appear to be ionized either by the QSO
radiationitselforbyshocksassociatedwithjets. References
6. Thislongrangeionizationsuggeststhatthesurroundingsof
Baldwin,J.,Phillips,M.,Terlevich,R.1981,PASP,93,5
the QSO are basically transparent to UV radiation, at least
CourbinF.,MagainP.,KirkoveM.,SohyS.,2000,ApJ,539,1136
in two directions(alongthe radio axis and towardsus - we CourbinF.,LetaweG.,MagainP.,WisotzkiL.,JablonkaP.etal.,2002,A&A,
recallthattheblobspectrumshowsnoreddening-M05). 394,863
7. The facts that the radio lobe and long range ionization are deGrijpM.H.K.,LubJ.,MileyG.K.,1987,AAS70,95
DjorgovskiS.G.,CourbinF.,MeylanG.,Sluse,D.,Thompson,D.etal.,2007,
stronger in the S-W direction (i.e. towards the companion
ApJ662,L1
galaxy) than in the opposite direction might be due to the Eigenbrod, A.,Courbin, F.,Meylan, G.,Vuissoz, C.,Magain, P.,2006, A&A
fact that the gas cloud immediately N-E of the QSO (the 451,759
blob) absorbs most of the energy emitted in that direction. Eigenbrod,A.,Courbin,F.,Meylan,G.,2007,A&A465,51
FeainI.,PapadopoulosP.,EkersR.,MiddelbergE.,2007,ApJ,662,872
Itmayalsobeexplainedbyalargeramountofgasinthedi-
HaehneltM.,DaviesM.,ReesM.,2006,MNRAS,366,22
rectionofthecompaniongalaxythanontheothersideofthe
HoffmanL.,LoebA.,2006,ApJ,638,75
QSO/blob. KimM.,HoL.,PengC.,ImM.,2007,ApJ,658,107
10 G.Letaweetal.:HE0450-2958:2Dand3Dspectroscopy
LetaweG.,CourbinF.,MagainP.,HilkerM.,JablonkaP.etal.,2004,A&A,424,
455-464
LetaweG.,MagainP.,CourbinF.,JablonkaP.etal.,2007,MNRAS,378,83
LowF.J.,CutriR.M.,HuchraJ.P.,KleinmanS.G.,ApJ327,L41
Marconi&Hunt,2003,ApJ589,21
MagainP.,CourbinF.,SohyS.,1998,ApJ,494,452
MagainP.,LetaweG.,CourbinF.,JablonkaP.etal.,2005,Nature,437,381-384
(M05)
Merritt D., Storchi-Bergmann T., Robinson A., Batcheldor D., Axon D., Cid
FernandesR.2006,MNRAS367,1746;
Theys&Spiegel,1976,ApJ,208,650and1977,ApJ,212,616
Veilleux,S.,Osterbrock,D.1987,ApJS,63,295
ZhouX.,YangF.,LX.,WangJ.-M.,2007,AJ,133,432