Resolved Spectroscopy of a Gravitationally Lensed L* Lyman-break Galaxy at z~5

0.65 MB·English

by  A. M. Swinbank

Mon.Not.R.Astron.Soc.000,000–000 (0000) Printed5February2008 (MNLATEXstylefilev2.2)
∗
Resolved Spectroscopy of a Gravitationally Lensed L
∼
Lyman-break Galaxy at z 5
A.M. Swinbank1∗, R.G. Bower1, Graham P. Smith2,4, R.J. Wilman1,
Ian Smail1, R.S. Ellis2, S.L. Morris1, & J.-P. Kneib3
7
1Institute for Computational Cosmology, Department of Physics, Durham University,South Road, Durham DH13LE, UK
0
2California Institute of Technology, MC105-24, Pasadena, CA 91125, USA
0
3Laboratoire d’Astrophysique de Marseille, Traverse du Siphon - B.P.8 13376, Marseille Cedec 12, France
2
4School of Physics and Astronomy, Universityof Birmingham, Edgbaston, Birmingham, B15 2TT

n ∗Email: [email protected]
a
J

9

5February2008
1
v
1
ABSTRACT
2
We exploit the gravitational potential of a massive, rich cluster at z = 0.9 to study
2
the internal properties of a gravitationally lensed galaxy at z=4.88. Using high res-
1
0 olution HST imaging together with optical (VIMOS) and near-infrared (SINFONI)
7 IntegralField Spectroscopywe have studiedthe rest-frameUV andopticalproperties
0 of the lensed galaxy seen through the cluster RCS0224-002.Using a detailed gravita-
/ tionallens modelofthe cluster wereconstructthe source-framemorphologyon200pc
h
scalesandfindan∼L∗ Lyman-breakgalaxywithanintrinsicsizeofonly2.0×0.8kpc,
p
- a velocity gradient of <∼60kms−1 and an implied dynamical mass of 1.0×1010M⊙
o within 2kpc. We infer an integrated star-formation rate of just 12±2M⊙yr−1 from
r the intrinsic [Oii]λ3727 emission line flux. The Lyα emission appears redshifted by
t
s +200±40kms−1 withrespecttothe[Oii]emission.TheLyαisalsosignificantlymore
a extended than the nebular emission, extending over 11.9×2.4kpc. Over this area,the
:
v Lyα centroid varies by less than 10kms−1. We model the asymmetric Lyα emission
i with an underlying Gaussian profile with an absorber in the blue wing and find that
X
theunderlyingLyαemissionlinecentroidisinexcellentagreementwiththe[Oii]emis-
r sion line redshift. By examining the spatially resolved structure of the [Oii] and Lyα
a
emission lines we investigate the nature of this system. The model for local starburst
galaxiessuggestedbyMass-Hesseetal.(2003)providesagooddescriptionofourdata,
and suggests that the galaxy is surrounded by a galactic-scale bi-polar outflow which
has recently burst out of the system. The outflow, which appears to be currently lo-
cated >30kpc from the galaxy,is escaping at a speed of upto ∼500 kms−1. Although
∼
themassoftheoutflowisuncertain,thegeometryandvelocityoftheoutflowsuggests
that the ejected material is travelling far faster than escape velocity and will travel
more than 1Mpc (comoving) before eventually stalling.
Key words: galaxies:highredshift; galaxies:starburst;gravitationallensing;galaxy
clusters; Integral Field Spectroscopy; Gravitational Arcs: Individual: RCS0224-002
1 INTRODUCTION theorists as the missing link in galaxy formation models
which are otherwise unable to match the shape and nor-
One of the most important observational breakthroughs in malisation of the luminosity function (Benson et al. 2003;
recent years was the discovery that a significant fraction Baugh et al.2005).Thesefeedbackprocessesmayalsooffer
of high-redshift galaxies are surrounded by “superwinds” natural explanation as to why only 10% of baryons cool to
(e.g. Pettini et al. 2002; Shapley et al. 2003; Bower et al. form stars (the Cosmic Cooling Crisis; White & Rees 1978;
2004; Wilman et al. 2005) – starburst and/or AGN driven Balogh et al. 2001).
outflows which expel gas from the galaxy potential, hence
playing no further role in the star-formation history of the However, important questions remain unanswered. Ev-
galaxy. This phenomenon is beginning to be understood by idenceforthesesuperwindsisusuallybasedonobservations
2 Swinbank et al.
Figure 1. Left: True colour HST VI-band image of the core of the lensing cluster RCS0224-002 at z=0.78. The contours mark the
high-redshiftcriticalcurves(curvesofinfinitemagnification)fromthegravitationallensmodeldescribedin§3.Wealsooverlaythefield
of view ofthe SINFONI IFU (shown bythe whitebox) whichwas used tomapthe [Oii]λ3727 emission.Thecluster galaxies whichwe
areable to spectroscopically identify arelabelledCG1-6. R1 isthe radial counter-image of the z=4.88 arcand R2is the z=1.05radial
arcfromSandetal.(2005).SerendipitousbackgroundgalaxiesarelabelledA1andA2(VIMOS)and1–4(SINFONI)(seeAppendixA).
Right: VR(I+Lyα) colour image of the cluster core generated from the VIMOS IFU datacube. The inner and outer curves show the
z=4.88causticandcriticalcurvesrespectively.Thecenterofthecluster(0,0)isatα=02:24:34.255δ=-00:02:32.39(J2000)andwehave
rotatedandalignedtheHSTandVIMOSdatasuchthatNorthisupandEastisleftinbothpanels.
whichcomparethenebularemissionlinepropertieswiththe massive) systems (Bower et al. 2004; Wilman et al. 2005;
rest-frame UV emission and absorption lines (such as Lyα, Swinbank et al. 2005), with very few “normal” galaxies
HαandUVISMabsorption lines;e.g. Erb et al.2003).Ve- being studied in detail beyond z = 2 (Pettini et al.
locity offsets of several hundred km/s have been measured, 2002,2000;Teplitz et al.2000;F¨orster Schreiberet al.2006;
suggestive of large scale outflows comparable to starburst Genzel et al.2006).Onewaytoovercomethisproblemisto
drivenwindsoftenobservedinlow-redshift Ultra-Luminous usethedeeppotentialofmassivegalaxyclusterstoboostthe
Infra-RedGalaxies (ULIRGs)in thelocal Universe(Martin fluxand sizes of images of distant galaxies which serendipi-
2005). However, the current data lack spatial information, touslyliebehindthem.Thisnaturalmagnification provides
which is vital if we are to understand if material escapes the opportunity to study intrinsically faint high redshift
into the Inter-Galactic Medium (IGM) or whether the out- galaxieswithaspatialresolutionthatcannotbeattainedvia
flowingmaterialeventuallystalls,fragmentsanddrainsback conventional observations (Smail et al. 1996; Franx et al.
ontothegalaxy,potentiallydisruptingthediskandcausing 1997;Teplitz et al.2000;Ellis et al.2001;Campusano et al.
furtherbursts of star-formation. 2001; Smith et al. 2002; Swinbanket al. 2003; Kneib et al.
The key to resolving these issues is to identify similar 2004; Swinbank et al. 2006).
features in the spatially resolved spectra of distant, young The natural magnification caused by the gravitational
galaxies.However,attheredshiftswherethesefeedbackpro- lens allows us to spatially resolve the morphologies and in-
cessesareattheirpeakactivity(abovez∼2,i.e.whengalax- ternal dynamics of galaxies in a level of detail far greater
ies were most rapidly forming stars), theimmense luminos- thanotherwisepossible.Forthegalaxydiscussedinthispa-
itydistancesmeanthattheemissionlinefluxesofthesedis- per, the amplification factor of sixteen corresponds to an
tantgalaxiesaredramaticallyreduced.Thisproblemiscom- increase of over three magnitudes and makes it possible to
pounded by the fact that the nebular emission lines (such study the galaxy on independent spatial scales of ∼200pc
as[Oii],[Oiii]andHα)areredshiftedintothenear-infrared, at z =5. In contrast, without a gravitational lens, 1′′ corre-
wheretheskybackgroundisanorderofmagnitudebrighter spondsto 6.5kpcat z=4.88.
than in the optical. Coupled with the fact that galaxies at In this paper we present a VLT/IFU study of the z =
these redshifts typically have disk scale lengths of only a 4.88gravitationallylensedgalaxybehindthecoreoftherich
few kpc,even with adaptiveoptics assisted observations on lensing cluster RCS0224-002. This galaxy cluster was iden-
eighttotenmetertelescopes,onlyahandfulofindependent tifiedashavingasignificantconcentrationofredgalaxiesin
pixels can berecovered from a distant galaxy. theinnerregions byGladders et al.(2002).Follow-up spec-
By necessity therefore, above z ∼ 2 most detailed troscopyconfirmedaredshiftfortheclusterof0.773±0.0021.
studies of individual galaxies have concentrated on the The ground based imaging from Gladders et al. (2002)
most luminous (and therefore usually very active and shows several tangential arc-like structures, many of which
Optical & Near-IR IFU Spectroscopy of a z=5 Lensed Proto-Galaxy 3
Figure 2. Collapsed, one-dimensional spectrum of the arc from
our SINFONI IFU observations. The [Oii] emission line has a
redshiftof4.8757±0.0005andalinewidthofσ=180±30kms−1.
Theinsetshowstherenormalised[Oii]emissionlinesforthetwo
regions (A&B)showninFigure4.We alsoshow asky-spectrum
(below)forcomparison(scaledinfluxforclarity).
resemble images of lensed, background galaxies. One of
the most striking of these is a multiply imaged arc ap-
proximately 15′′ to the North-East of the Brightest Clus-
ter Galaxy (BCG) (see Fig. 1). Follow-up spectroscopy by
Gladders et al. (2002) yielded a redshift of z=4.88 from
identification of a strong Lyαemission line at 7148˚A.
Weconcentrateonthedynamicsandstar-formingprop-
Figure3.Image-planeobservationsofthez=4.88arc:HSTcon-
erties of the z=4.88 arc observed through the rest-frame
tinuum (greyscale), Lyα emission(red contours) and [Oii] emis-
UV spectra with the VIMOS IFU and through the nebu-
sion (white contours). The image shows that the Lyα emission
lar ([Oii]λλ3726.1,3728.8˚A) emission line doublet observed is much more extended than either the [Oii] or continuum mor-
with the SINFONIIFU in the near-infrared. The IFU data phology. The solid bar in the top right hand corner represents
provide a two-dimensional map of the galaxy’s properties the 0.8′′ seeing for both the VIMOS and SINFONI IFU obser-
in sky co-ordinates. In order to interpret this map we must vations.Inset:Continuumsubtracted,narrow-bandimagearound
correct for the magnification and distortion caused by the the redshifted[Oii]λ3727 emissionof the arcfromour SINFONI
lensing potential using a mass model for the cluster lens. IFU observations. Thehighest surfacebrightness components in
Thismodelis constrained byrequiringit toaccountfor the both the optical image and the [Oii] emission line map are well
matched, thoughthe[Oii]emissionlinemapalsoshowsthatthe
positions and redshifts of the gravitationally-lensed arcs in
highlyshearedcomponenthasmorestructurethanevidentinthe
the cluster. The lensing model then allows us to determine
opticalimaging.Right:
thesourceplanemorphologyandgeometryofthecontinuum
and line emission we use.
The structure of this paper is as follows. In §2, we lensing cluster RCS0224-002 were obtained from the HST
present the data on which this paper is based: a combina- public archive1. The I and V -band observations were
814 555
tion of Hubble Space Telescope (HST), ground-based imag- 10.5and8.4ksrespectivelyandthedatawerereducedusing
ing, and integral field spectroscopy obtained with the SIN- the standard stsdas package in iraf. The resulting image
FONI and VIMOS IFUs on the VLT. In §3 we present the (Fig.1)coversthebrightestcentralclustergalaxiesandthe
source-planepropertiesofthearcandthespatially resolved target arc at a resolution of 0.0996′′/pixel. From the HST
spectroscopy and in §4 and §5 we present our discussion imagingthez=4.88arcisred(V −I =1.7±0.1)withan
606 814
andconclusionsrespectively.Through-outthispaperweuse average surface brightness of µI814=25.0 and an integrated
the Vega magnitude system and assume a cosmology with magnitudeof I=22.2.
H =72kms−1, Ω =0.3 and Λ =0.7.
0 0 0
1 PID: 9135; Obtained from the Multimission Archive at the
SpaceTelescopeScienceInstitute(MAST).STScIisoperatedby
2 OBSERVATIONS AND DATA REDUCTION
the Association of Universities for Research in Astronomy, Inc.,
2.1 HST imaging under NASA contract NAS5-26555. Support forMAST for non-
HST data is provided by the NASA Office of Space Science via
HST WFPC2 I814− and V555−band observations of the grantNAG5-7584andbyothergrantsandcontracts.
4 Swinbank et al.
Figure 4.Source-plane observations ofthe z=4.88galaxy fromthe HST,VIMOSandSINFONIobservations. Top Left:[Oii]emission
lineintensity of the galaxy (dark regions represent regions of highest intensity). Top Right: [Oii] emissionlinevelocity structure of the
galaxywhichshowsamaximumvelocityshiftof60±20kms−1 alongthelongaxisofthegalaxy.BottomLeft:Reconstructedtruecolour
HST VI image of the z=4.88 arc. Inset: Reconstructed HST image after a smoothing scale of 0.8′′ has been applied to the sky-plane
image.BottomRight:ReconstructedLyαemissionlinemapofthez=4.88arc.TheboxesshowregionsfromwhichthespectrainFigure6
wereextracted.
2.2 VIMOS Integral Field Spectroscopy weconstructabroad-bandimagebycollapsingthedatacube
in wavelength between 6000˚A and 7500˚A and mask out the
The z=4.88 arc was observed with the VIMOS IFU thebrightestobjects.Wethenapplythismaskateachwave-
(LeFevreet al.2003)foratotalof43.2ks(splitinto16×2700 lengthandstacktheresultingdatacube(withoutspatialoff-
secondexposures)between2004December16and2005De- sets and ignoring masked pixels). This producesa “master-
cember 12 in <∼0.6′′ seeing and photometric conditions. We flatfield” which is then divided into each of the science ob-
used the MR-orange grism which results in a field of view servations to reduce the effects of the fringing. These indi-
of 27′′×27′′ at 0.67′′ per pixel and a spectra resolution of vidual datacubes are then sky-subtracted by masking the
λ/∆λ ∼1100 at 7000˚A. To reduce the data we used the brightest objects in the individual frames and constructing
VIMOS esorex pipeline which extracts, wavelength cali- a sky-spectrum by collapsing the datacube in both spatial
brates and flat-fields the data. The final datacubes cover a dimensions (since the four quadrants of the VIMOS IFU
wavelength range of 5000-11000˚A and has a resolution of have slightly different throughput’s, we note that the sky-
6.6˚A FWHM (measured from the width of the sky-lines at subtractionwasperformedquadrant-by-quadrant).Tobuild
a wavelength of 7150˚A). In all future sections, line widths the final datacube, we create broad band images from each
are deconvolved for instrumental resolution. Flux calibra- datacubeandcrosscorrelatetheseinordertospatiallyalign
tion was achieved by using observations of ESO standard andcreatethefinalmosaic.Thefinaldatacubeiscreatedus-
stars.Theseobservationsweretakeneitherimmediatelybe- ing an average with a 3σ clip to reject cosmic rays.
fore,orimmediatelyafterthescienceobservationsandwere In Fig. 1 we show a colour image of the cluster and
reduced in an identical manner. Since there are no point arc,generatedfromthedatacube.Wegeneratethreecolours
sources in our datacubes, we measure the seeing from the fromthedatacubebycollapsingthedatacubebetween5000-
standardstarsinthecollapseddatacubesandderivetypical 6000˚A, 6000-7000˚A and 7000-8000˚A as blue, green and red
seeing measurement of ∼0.8′′ at a typicalairmass 1.1. Dur- respectively. Alternatively, the data-cube can be viewed as
ing the standard star observation, the ESO seeing monitor a sequence of wavelength slices to make an animation (this
registered median seeing of 0.7′′ – hence we conservatively can beviewed at:
assumethattheVIMOSIFUobservationsweretakenin0.8′′ http://star-www.dur.ac.uk/∼ams/lensing/RCS0224 z5/ )
seeing.
After reducing each of the observations to a datacube,
2.3 SINFONI IFU Observations
we construct and apply a master flat-field in order to cor-
rect for the fringing effects which are particular prominent To map the nebular emission line properties of the z=4.88
above 8000˚A. In each of the (spatially dithered) datacubes arc, we used the SINFONI IFU (Eisenhauer et al. 2003).
Optical & Near-IR IFU Spectroscopy of a z=5 Lensed Proto-Galaxy 5
agewitha3σcliptorejectcosmicrays.Forfluxcalibration,
standard stars were observed each night during either im-
mediately before or after thescience exposures. These were
reduced in an identical manner to the science observations.
In Figure 2 we show the one-dimensional spectrum of the
arc (generated by collapsing the datacube over the object
in both spatial domains) and in Figure 3 we also show a
narrow-band image of the arc generated by collapsing the
datacubeinthewavelengthdirectionoverthe[Oii]emission
line.Figure3alsoshowstheHSTimageofthearcwith the
contours from the[Oii]and Lyαoverlaid.
3 ANALYSIS & RESULTS
3.1 Mass Modelling
Reconstruction of the intrinsic properties of the galaxy at
z=4.88 requires removal of the gravitational magnification
Figure 5. Collapsed, one-dimensional spectrum of the arc from fromtheobservables.Wethereforeusethefourspectroscop-
ourVIMOSIFUobservations.Thedashedverticallinesshowthe
ically confirmedimages of thez=4.88 galaxy toconstrain a
expected position of the UV emission and absorption lines for
modelofthemassdistributioninthecluster.Thesefourim-
the systemic redshift of z =4.8757 (as measured from the [Oii]
ages comprise thethreeimages in thetangential arc identi-
emissionline).TheLyαemissionappearsredshiftedwithrespect
fiedbyGladders et al.(2002)andthefourthimageidentified
to the systemic velocity, and has anasymmetric lineprofile. We
also detect the weak UV-ISM lines of Siiv. These lie ina region via our integral field spectroscopy (Appendix A). The can-
devoid of strong sky emission. The hashed regions show regions didate radial arc at z=1.05 (A3 in Fig. 1; Sand et al. 2005)
of strong sky emission and the horizontal dashed line marks a and the object identified at z=3.66 adjacent to the z=4.88
continuum level of zero. Top Inset: The Lyα emission line with arc (A1 in Fig. 1) appear not to be multiply-imaged, and
the model described in §3.2.2. The black (solid) line shows the so are not included as constraints. The configuration of the
observedspectra. Theasymmetriclineisbestfitwithanunder- fourobservedimagesandthelackofstrongradialamplifica-
lying Gaussian emission line profile (blue dotted line) combined
tionofthefourthimageimpliesthatthefifthimageliesvery
with a Voigt profile absorber (green dashed line) and a third
closetothecentreoftheclustermassdistribution.Wethere-
Gaussian which accounts for the red wing of emission. Whilst
fore expect the fifth image to be strongly de-magnified and
the observedLyαemissionappears redshiftedfromthe systemic
thuslikelyundetectableinthecurrentdata.Weidentifythe
velocity,thebest-fitunderlyingemissionprofilehasacentroidin
excellentagreementwiththe[Oii]emission.LowerInset:Spectra dense knot of emission in each of the four observed images
aroundtherest-frameUVabsorptionlinesSiivλλ1393.76,1402.8. asbeingimagesofthesameunderlyingregionofthegalaxy,
Weoverlaythebest-fittemplate(blue)fromShapleyetal.(2003) and use the position of these knots to constrain the model.
andusethistoinferavelocityshiftof-400±100kms−1 fromthe FollowingAppendixAofSmith et al.(2005),thenumberof
systemicredshift. constraints is therefore n =6.
C
The central region of RCS0224 contains two bright
elliptical galaxies: CG1 and CG2 (Fig. 1) – we there-
Atz=4.88 the[Oii]λλ3726.1,3728.8 emission linedoublet is fore construct a three component model comprising the
redshifted to 2.19µm and into a stable part of the K-band cluster-scale mass distribution (dark matter and gas),
relatively free from strong OH airglow emission. The SIN- CG1 and CG2. We parametrise all three mass com-
FONI IFU uses an image slicer and mirrors to reformat a ponents as truncated pseudo-isothermal elliptical mass
field of 10×10′′ at a spatial resolution of 0.25′′/pixel. We distributions, each described by the following parame-
used the HK grating which results in a spectral resolution ters: {x,y,ǫ,θ,r ,r ,v } (Kassiola & Kovner 1993;
core cut disp
of λ/∆λ=1700 at 2.20µm (the sky-lines have a resolution Kneib et al. 1996) – i.e. 21 parameters in total, comprising
13˚A FWHM at this wavelength) and covers a wavelength thecenter,ellipticity,position angle,coreradiusandcut-off
range of 1.451 to 2.463µm. In all future sections, emission radiusofeachmasscomponent.Toobtainawell-constrained
linewidthsaredeconvolvedfortheinstrumentalresolution. model, (and given the number of constraints above: n =6)
C
To observethetarget we used ABBA chop sequences,how- we prefer no more than 6 of these 21 parameters to be free
ever since the arc only filled one half of the datacube, we parameters in the lens model. The geometrical parameters
chopped 2′′ West to sky (see Fig. 1), thus keeping the ob- (x,y,ǫ,θ)describingCG1andCG2arethereforematchedto
ject inside the IFU. We observed the target for a total of their 2-dimensional light distributions, and their respective
43.2ks(split into16×2700 second exposures) between 2005 r and r parameters are matched to the shape of the
core cut
August10and2005September28in<∼0.6′′ seeingandpho- light profile of each galaxy. Finally, the mass-to-light ratio
tometric conditions. Individual exposures were reduced us- ofeachgalaxyisfixedbymatchingtheirvelocitydispersions
ingtheSINFONIesorexdatareductionpipelinewhichex- to the measured velocity dispersion (Table B). Turning to
tracts, flatfields, wavelength calibrates and forms the dat- theparametersdescribingthecluster-scalemass,wefixboth
acube. The final datacube was generated by aligning the r and r at 75kpc and 1Mpc respectively. The former
core cut
individualdata-cubesandthencombiningtheusinganaver- valueis representative of massive clusters, and our decision
6 Swinbank et al.
to fix this parameter is caused by the non-detection of the grated emission line flux of 5±1×10−20Wm−2 suggests a
central fifth image of the z=4.88 arc. The lack of the fifth star-formation rate(corrected for lensingamplification, but
image reduces our ability to constrain the radial shape of uncorrected for reddening) of 12±2M⊙yr−1 (assuming the
the mass distribution within the tangential critical curve. calibration of L([Oii]) to SFR given in Kennicutt1998).
We also run models with r =50kpc and 100kpc to con- In Figure 2 we show the intensity distribution and col-
core
firm that the final results are not altered by the adopted lapsedspectrumofthe[Oii]emission.Intensityandvelocity
value of r . For completeness, we include the result of maps were derived by fitting the [Oii] emission line dou-
core
this check in the error on the lens magnification calculated blet in each 0.25′′ pixel using thesame techniquedescribed
below. Fixingr =1Mpc is motivated by thesmall field of above. We used a χ2 minimisation procedure, taking into
cut
view of the HST data and within those data the small an- accountthegreaternoiseofatthepositionsofthesky-lines.
gular scale subtendedbythestrong lensing constraints. We Incaseswherethefitfailedtodetectthelinetheregionwas
are therefore unable to constrain this parameter, but our increased to2×2pixels(0.5′′×0.5′′).Usingacontinuumfit,
resultsareinsensitivetoitsvalue.Insummary,wetherefore we required a minimum signal-to-noise of 3 to detect the
fit a model with five free parameters to the data. The free line, and when this criterion is met, we fit the [Oii] emis-
and fixed parameters are listed in Table 1. sion linedoubletwithaGaussian profileof fixedseparation
All of the lens modelling is performed using the allowing the central wavelength and normalisation to vary.
lenstool software (Kneib 1993; Kneib et al. 1996), incor- In Figure 4 we show the intensity distribution and velocity
porating a Markov Chain Monte Carlo (MCMC) sampler structureof the [Oii]emission line flux.
(Jullo et al. 2006, in preparation). The model parameters It is clear that the brightest components in the rest-
with the lowest χ2 found with this MCMC approach and frame UV are also the brightest in the [Oii] emission. In
68%confidenceintervalsaroundtheseparametersvaluesaf- the source plane we find that the eastern most compo-
ter marginalising over the other four parameters in each nent is marginally redshifted relative to the western com-
case, are listed in Table 1, and we show the critical curve ponent.Relative to the combined spectrum, theeastern re-
of the best fit model in Fig. 1. The lens model predicts the gion has v = −20±20kms−1 (with a velocity dispersion
fifth image to lie at the center of CG1 and to be very faint of σ=80±18kms−1), while the western component is at
–R>∼36,i.e.de-magnifiedby∼14magnitudesrelativetothe +20±15kms−1(andavelocitydispersionof85±12kms−1).
tangentialarc.Althoughthemodelmakesaclearprediction The quoted uncertainty here is dominated by possible vari-
for the location of this image, it is clearly too faint to de- ations in the[Oii]doublet line ratios, which was left free in
tectwiththeavailabledata.Weray-tracethez=4.88galaxy thesefits.
throughthefamilyofmodelswithinthe68%confidencesur- Since the [Oii] velocity dispersion (σ) reflects the dy-
face in the 5-dimensional parameter space to compute the namicsofthegasinthegalaxies’potentialwellweestimate
mean,luminosityweightedmagnificationofµ=16±2(which the dynamical mass of the galaxy. Following the same pre-
corresponds to ∆m=3.0±0.2 magnitudes). scriptionofErb et al.(2006),weusethevelocitydispersion
Accounting for the lensing magnification, the intrinsic andspatialextentofthe[Oii]fluxtoinferadynamicalmass
magnitude of this galaxy is I=25.2. In comparison, deep of ∼ 1× 1010 M⊙ within a radius of 2kpc. Clearly this
imaging surveys of Lyman-break galaxies at z∼5 have de- mass has large uncertainties since the mass dependson the
rived an L∗ of i ∼25.3 (Ouchiet al. 2004). This suggests massdensityprofile,velocityanisotropyandrelativecontri-
that the z =4.88 galaxy is typical of the UV continuum se- butionsto σ from random motions or rotation and possible
lected LBG population at this redshift. In Fig. 4 we show differences between the total mass and that of the tracer
the reconstructed HST image of the galaxy. In the source- particlesusedtomeasureit.Nevertheless,itisworthnoting
plane thegalaxy has aFWHM of only 2.0×0.8kpc, and in- thatthemasswederiveisafactorof∼5×smallerthanthe
dicatesaflattened(orbar-like)geometry.Suchanelongated mediandynamicalmassofz∼3LBGsfromErb et al.(2006)
morphologyisnotunusualforgalaxiesattheseredshifts:in measured in exactly thesame way.
a recent high-resolution imaging survey of LBGs at z>2.5
with HST, Navindranath et al. (2006) concluded that upto
3.2.2 Lyα
50%ofLBGsshowevidenceforbar-likemorphologies(with
meanscalelengthsof1.7–2.0kpc).Thusthegalaxyappears In Fig 4 we show the reconstructed source-plane Lyα mor-
typical in terms of its source plane morphology, brightness phology of the arc. It is immediately apparent that in the
and size of the population at z ∼5. source-plane, the Lyα is more extended than the [Oii], or
thecontinuumlight,extendingover11.9×2.4kpc(FWHM).
Figure 5 shows the collapsed spectrum of the arc overlaid
3.2 Spatially Resolved Spectroscopy with the expected position of the Lyα emission line for
the systemic redshift (we also mark the expected position
3.2.1 Nebular emission
of the UV ISM lines which lie in regions of the sky free
ByfittingadoubleGaussianprofileoffixedseparation(cor- from strong OH airglow emission). The integrated Lyα line
respondingto theredshifted separation of thedoublet) and fluxis3.7±0.5×10−15ergs−1cm−2,corresponding to astar-
intensity but variable width, redshift and intensities of the formation rate of 5±2M⊙yr−1 (unlensed). In comparison
doublet to thecollapsed spectrum, the[Oii]emission yields the1500˚A continuum flux suggests a star-formation rate of
asystemicredshiftof4.8757±0.0010andandintrinsicwidth 3±2M⊙yr−1 (Kennicutt 1998).
of σ=100±20kms−1 (in the rest-frame of the galaxy). We From the collapsed spectrum of the galaxy we mea-
assume that the [Oii] maps the systemic redshift and all sure a redshift from the centroid of the Lyα emission line
velocities are given with respect to this value. The inte- and measure a velocity offset from the systemic redshift of
Optical & Near-IR IFU Spectroscopy of a z=5 Lensed Proto-Galaxy 7
Table1.
GravitationalLensModelParameters
∆RA ∆Dec ǫ θ rcore rcut vdisp
′′ ′′
( ) ( ) (deg) (kpc) (kpc) (km/s)
DMhalo −0.6±1.6 2.1±1.4 0.18±0.02 −49±5 [75] [1000] 935±30
CG1 [0.0] [0.0] [0.07] [169] [0.2] [42] [230]
CG2 [2.4] [−1.2] [0.11] [202] [0.2] [38] [220]
Table 1.Note:Numbersinsquarebrackets arefixedinthefit.Positionangles areclockwisefromNorth.
+200±40kms−1.Velocityoffsetsofthismagnitudebetween ular emission line redshift as measured from [Oii]. This
Lyα and nebular emission lines (such as [Oii] or Hα) are a underlying Gaussian profile also has a width of σ=4.8˚A
common feature of galaxies at z ∼3 (Shapley et al. 2003; (σ=200kms−1 in the rest-frame of the galaxy). The ab-
Erb et al. 2003). In both the collapsed spectrum (Fig. 5) sorber in the blue wing of the Lyα emission (green dashed
and the spatially resolved spectra (Fig. 6) the strong Lyα line) has a centroid of 7138.0±0.1˚A (z=4.8717±0.0001, or
from the galaxy shows an asymmetric profile in which the ∆v = −500kms−1 relative to the intrinsic emission) and
bluewingoftheLyαisabsorbed,presumablygivingriseto a width of ∼100kms−1. Since the flux in the Voigt pro-
theapparentredshiftoffset betweenthe[Oii]andLyα.Itis file reflects the column density of the neutral gas, we infer
also clear that there is an extended wing of emission in the N =1.6+2.5×1019cm−2andσ=110±50kms−1.Thelarge
HI −1.1
red side of theline. uncertaintiesinthevalueofN andσ reflectthatfactthat
HI
Asymmetry in the Lyα emission line spectra of LBGs theinferredvalueofNHIliesbetweentheflatandlinearpart
atz ∼2–3arecommonandoftengiverisetovelocityoffsets of thecurveof growth for Hydrogen.
of several hundred km/s. These velocity offsets are usually With this two component model we are able to fit the
attributed to galactic-scale outflows, presumably powered asymmetryinboththecollapsedandspatiallyresolvedspec-
bysupernovae.However,sincetheopticaldepthoftheLyα tra. However, even with two components there remains ex-
forestisanorderofmagnitudelargeratz=5thanatz =3, cessfluxintheredwingoftheemissionline.Tocompensate
(wheremostspectroscopicstudieshavetakenplace)wefirst forthisemission,wealsofitathirdGaussianprofilecompo-
investigate whether the IGM could cause the absorption in nentintheredwing,againallowing theredshift,widthand
theblue wing of theemission line. intensity to vary. The improvement between the three fit
At z = 5 the mean optical depth of the Lyα forest is and two fit models is ∆χ2 =12−38 (∼3−6σ; depending
1.7±0.2 (Fan et al. 2006) which corresponds to a transmit- on whether we fit the individual spectra or the combined
ted flux of 20%. We therefore investigate whether the Lyα spectrum). The best fit model therefore includes a third
profile could be explained as the result of average IGM ab- componentwithredshiftcentroidat7153.44(z=4.8860)and
sorption,orwhethertheabsorptionneededtobeduetoma- a width of σ=260kms−1. The velocity offsets of the three
terial associated with the lensed galaxy. The mean optical components are summarised in Table 2. As we discuss in
depthismeasuredbyaveragingtheabsorptionofindividual §4,thisthirdcomponentmaynaturallyarisefromLyαpho-
clouds on scales of 6000kms−1 (6Mpc comoving). We fit tonsback-scatterontheinteriorofanoutflowingshell.Thus
theLyαemissionlinewithanunderlyingGaussianemission thebest-fitmodeltotheasymmetricLyαemission isonein
line profile (centred on the systemic redshift of the galaxy whichtheLyαphotonsaregeneratedinsimilarregions(and
asmeasuredfromthe[Oii]emission),convolvedwithastep at the same redshift) as the [Oii], but the emission profile
functionwhichmodelstheaverageIGM(allprofilesarethen ismodifiedbyforegroundneutralmaterial.Thisforeground
convolvedwiththeinstrumentalresolution).Wealsoinclude materialabsorbs/scattersphotonsfromthebluewingofthe
a broad Gaussian emission line profile (also centred at the line,causinganapparentredshiftoffsetbetweenthenebular
systemic redshift) to account for the extended red wing of emission and Lyα.
emission. Allowing the redshift of the IGM component to Thedescriptionofthelineprofilegivenaboveistypical
vary we find that an acceptable match emission line shape of the emission profile seen in other Lyman break galaxies.
is best matched when the IGM step function is placed be- The key advantage of observing the lensed system is that
tween-80and+10kms−1 fromthegalaxy(between-85kpc weareabletospatially resolve theemission line,andhence
and +10kpc comoving for out adopted cosmology). This is investigatethespatialvariationsintheabsorptionandemis-
inconsistentwithourassumptionofalarge-scaleaverageab- sion. We divide the IFU data cube into the spatial regions
sorption,implyingthattheabsorptionweseeresultsfroman showninFig.4.Forreferencethesehavebeenlabelled0-15.
individualcloud(orclouds)lyingclosetothetargetgalaxy. TheresultingspectraareshowninFig.6.Themoststriking
Moreover,asimplestepfunctionfailstoadequatelyre- featureofthisfigureisthat(apartfrom variations ininten-
produce the line shape of the blue wing. A better fit is sity)thelineprofileappearsremarkablyconstantacrossthe
obtained if the line shape is instead modified by a single galaxy. For each spatial element, we measure the centroid
absorbing cloud. We model the emission profile by Gaus- oftheemissionandfindthattheintrinsicemission linecen-
sian profile combined with a Voigt profile absorber in the troid varies by less than 10kms−1 (rest-frame) across the
blue wing. In this fit the wavelength of the underlying whole galaxy image (this value was derived by both fitting
best fit Gaussian profile (blue dotted line in Fig 5) was the 2 and 3 component models above). Indeed, as Fig. 6
allowed to vary, but the fitted centroid at 7143.0±0.5˚A shows, the asymmetry in the Lyα profile is evident across
(z=4.8758±0.0001) is in excellent agreement with theneb- the whole galaxy. Applying a free fit, each of the individ-
8 Swinbank et al.
Table2.
Propertiesof theemission andabsorption lines
Component z ∆v σ vshear
kms−1 kms−1 kms−1
[Oii] 4.8757[5] 0 100±30 −30<v<30
Lyα(Principle) 4.8760[5] +20±50 230±40 <∼20
Lyα(redwing) 4.8860[15] +520±150 260±80 <∼50
Lyα(blueabs.) 4.8680[10] −500±100 110±30 <∼20
Siivabsorption 4.8700[10] −400±100 ∼480 <∼150
Table 2. Note the value given in the [] z column is the error
in the last decimal place. vshear denotes the maximum velocity
gradient/shear inthegivenemission/absorptionline.
manbreakgalaxiesfromShapley et al.(2003)andfindthat
theseabsorption lines are blue-shiftedby−400±100kms−1
(z=4.870±0.0007) from the systemic redshift as defined by
the [Oii] emission line (Figure 5). It is worth noting that
in the LBG composite spectrum from Shapley et al. (2003)
theinterstellarabsorptionfeatures(Oiλ1303+Siiiλ1260)are
detected with similar significance as the Siiv. However, at
z =4.88 these are redshifted to 7656˚A and into the middle
of the Frauenhoffer A-band and therefore are therefore not
expected to bedetected.
The observed velocity offsets between the nebular
([Oii]),LyαandUVISMlinesarethereforeconsistentwith
thesituationseeninmosthigh-redshiftLymanBreakGalax-
ieswherevelocityoffsetsofseveralhundredkm/shavebeen
measured (e.g. Erb et al. 2003; Steidel et al. 2004). These
velocityoffsetsareattributedtostarburstdrivensuperwinds
Figure 6. One-dimensional spectra around the redshifted Lyα and we base much of our interpretation on thismodel.
emission from the z=4.88 galaxy from the fifteen regions shown
inFigure4(black). Each ofthe spectra has thecomposite spec-
tra (scaled) and overlaid for comparison (red dashed). The Lyα
emission in all of the spectra show an asymmetric profile, with 4 DISCUSSION
the blue wing of the Lyα emission truncated making it appear
redshiftedby +200±40kms−1 fromthe nebular ([Oii]) emission Before attempting to interpret the observations, we briefly
line.Furthermore,thecentroidoftheLyαemissionvariesbyless review the results of the observations. The [Oii] emission
than10kms−1 (rest)acrossthewholegalaxyimage. mapsthestar-formingregions,whichappeartohaveanelon-
gated (bar-like) morphology with a spatial extent of only
2.0×0.8kpc,whichappearstocontainallofthestarsinthe
ualspectraisconsistent with themean,with themaximum galaxy. We use the [Oii] emission to infer a systemic red-
∆χ2=1.3 (and an average ∆χ2 ∼0.5). From the position of shift of 4.8757±0.0005. Using the[Oii]emission weinfer an
the Voigt absorber, we also note that the blue-edge of the integrated star-formation rate of 12±2M⊙yr−1 (Kennicutt
emissionlinecutsoffatawavelengthcorrespondingtovari- 1998)andthespatialextentandlinewidthofthe[Oii]sug-
ationsoflessthan20kms−1 ataredshiftof4.8717±0.0001. gestadynamicalmassof∼1×1010 M⊙within2kpc.Turning
This constancy in the line shape and centroid across the
to the rest-frame UV spectrum: the Lyα emission is much
galaxy image is discussed further below. more extended than the [Oii], extending over 11.9×2.4kpc
(i.e. at least a factor of 4× the spatial extent of the [Oii]).
Moreover, the peak of the Lyα emission is redshifted by
3.2.3 UV absorption
+200±40kms−1 from the [Oii]. Across the whole spatial
Intheintegratedrest-frameUVcontinuumspectrumofthe extent, the Lyα emission line centroid varies by less than
galaxy, we also detect the weak Siiv(λλ1393,1402.77) ab- 10kms−1(rest-frame)andthelineprofileisremarkablycon-
sorption features. Whilst these are very weak in our spec- stantoverthesamespatial extent.TheLyαemission (both
tra (the S/N in continuum in regions free from strong sky in the collapsed spectrum and spatially resolved) has an
emission is only ∼3), the separation between the doublet asymmetric profile which is best explained by a three com-
is 9.80±0.15˚A (rest), is in excellent agreement with the ponentmodel.TheunderlyingGaussianemissionlineprofile
expected value of 9.77. To calculate the redshift of these hasacentroidinexcellentagreementwiththenebularemis-
absorption lines we cross correlate the spectrum around sion,anabsorberwhichisblue-shiftedby−500±100kms−1
8200˚A with the rest-frame UV composite spectrum of Ly- from the systemic redshift and an extended “red wing” of
Optical & Near-IR IFU Spectroscopy of a z=5 Lensed Proto-Galaxy 9
emissionisredshiftedby+520±150kms−1.Finally,theUV being an example of a system in these later stages, and the
ISM lines of Siiv are blue-shifted by -400±150kms−1 from z=5 arc fits intothe paradigm well.
the systemic redshift, although it has not been possible to The model interprets the spatially integrated proper-
derive any spatial information across thegalaxy from these tiesoftheemission andabsorption lines seenin Figure6as
weak lines. follows: themainpeakofLyαemission comesfromphotons
emittedfromthestarformingregions.Toreachtheobserver
thesemustpassthroughpartoftheforeground(blueshifted)
shell. This absorption causes the peak emission wavelength
4.1 A starburst with a bi-conical outflow?
to appear redshifted relative to the nebular emission lines.
Theobservedvelocityoffsetsbetweenthe[Oii],LyαandUV In addition, photons may be scattered or emitted from the
ISM lines in RCS0224arc are consistent with situation seen receding shell. Photons that are either created on theinner
in most high redshift Lyman Break Galaxies in which the surface of the shell (e.g. by UV irradiation from the star-
presenceof “superwinds”(drivenbythecollective effectsof burst), or multiply scattered within the receding shell so
star-formationandsupernovae)naturallyexplaintheobser- thattheyacquirethemeanvelocityoftheshell(seethedis-
vations.Ourspatiallyresolvedspectrathereforeofferusthe cussion of Hansen & Oh2006) will beseen as redshifted by
opportunitytostudythespatialstructureandenergetics of theobserver.
theoutflowing material in this galaxy. However,whiletheintegratedspectrumiscertainlycon-
The velocity offset between the [Oii] and transmit- sistent with the presence of an outflowing bubble, it gives
ted Lyα emission profile are usually interpreted in terms noindication of its physical size, or whether thebubblege-
of a starburst driven outflow. This model provides a ometry is appropriate. For example, the integrated spectra
good description of the integrated line emission profiles could be equally well explained by a large scale outflow, or
of high redshift galaxies and more detailed observations a small bubble only just surrounding the starburst region.
of local star bursts (eg., Pettini et al. 2002; Shapley et al. To investigate the energy of the outflow, and its ability to
2003; Tenorio-Tagle et al. 1999; Heckman et al. 2000; escapethehostgalaxy’sgravitationalpotential,wemustuse
Grimes et al. 2006). Herewe investigate theimplication for spatial information in our spectra.
this model from the spatially resolved profile that we have We begin by considering the dominant contribution to
observed.Webaseourinterpretationonthespecificmodels the Lyα line from photons travelling directly towards the
of Mas-Hesse et al. (2003) which assume that the spatially observer. The wavelength of the blue cut-off in the line is
compact [Oii] nebular emission traces the star forming re- determined by the outflow velocity seen by the observer. If
gions in thenucleus of a star-bursting galaxy. thebubbleissphericalandclosetothegalaxywewouldex-
In these models, the velocity offsets between [Oii] and pecttoseesignificantvariationinthebluecut-offacrossthe
Lyα is generated by a star forming region embedded in an emissionregionsincethelineofsightvelocityoftheoutflow
outflowingbubble.Themodelshavebeenextensivelyinves- depends on p1−(b/Rs)2, where b is the impact parame-
tigated in local starburst galaxies where similar spatially ter and Rs is the shell radius. Clearly the consistency of
resolved data is available. Indeed, there is strong similar- the blue cut-off in Lyα across the galaxy implies that the
ity between thedata we present,and thelocal galaxy Haro shellismuchlarger thanthegalaxy sothattheilluminated
2 studied by Mas-Hesse et al. (2003). The spatial extend portion of the absorbing screen has a planar geometry. We
(500pc)andvelocityshift(v ∼200 kms−1)seeninHaro illustrate the likely viewing geometry in Fig. 7 (cf., Fig. 18
shell
2aresmallerthanthez =4.88galaxybutotherwisethesys- of Mas-Hesse et al. 2003) for an outflow in the shell recom-
tems are remarkably similar. Both show Lyα emission that bination phase (phase 4 in their classification scheme). In
extendsbeyond the nebularemission region and both show Haro2,Legrand et al.(1997)areabletoestimatethesizeof
ablueabsorptioncut-offtotheLyαemissionlineandasim- theshell from its low surface brightness Hαemission. They
ilar red-wing. Crucially, both data sets show no detectable estimate that the projected diameter of theshell is 2.5kpc,
variation in the shape of the Lyα emission profile across around five times the diameter of the emission region. A
the emission region. It is remarkable that this galaxy, seen similar ratio would seem appropriate for the arc, although
1GyraftertheBig-Bangissosimilartothelocal systemat theshell could obviously belarger.
z=0.0049. Our data also require an additional contribution from
The basis of the interpretation is that the star-burst photonsscatteredorcreatedintherecedingshell.Foralarge
generates a large over pressurised region which surrounds spherical shell, we would expect to see a widespread low
the nebula. Within the ISM, the flow is Reighly-Taylor un- surface brightness emission at this wavelength (e.g. Fig. 7).
stable so that the initial shell breaks up leaving behind a However,nosuchdistributedemissionisseeninourdata,or
densergasclumpsthatmayberesponsiblefortheextended inMass-Hesseetal.’sobservationsoflocalsystems.Instead,
Lyαemission region. The outflow tendsto break out of the by collapsing the datacube over a region dominated by the
galaxy along the axis of least resistance. In the case of a excessfluxintheredwing,wefindthattheredshiftedwing
disk (or flattened) galaxy this generates an approximately seems to have a similar angular distribution to that of the
bi-conical outflow. Whentheoutflow reachesthesmoother, Lyαemission comingdirectlyfromthegalaxy.However,we
lower density gas distribution in the halo of the galaxy, a can extend the model of Mass-Hesse et al. to explain the
new shell develops as the outflow sweeps up material from limited spatial extent of this emission as a result of the bi-
the galaxy’s halo, creating a screen between the observer conical outflow geometry. Redshifted photons from most of
and the source. As the shell travels away from the galaxy, thereceding shell must pass through thearea of thegalaxy
it becomes dense enough to recombine and its Lyα opacity thathasnotyetbeenionisedbythestarburst.Eventhough
becomes large. Mas-Hesse et al. (2003) interpret Haro 2 as theyareredshiftedby+v relativetothegalaxy,theywill
shell
10 Swinbank et al.
still be absorbed since the absorbing column is likely to be foreground absorber. This suggests a linear size for the end
extremely high. Only in the ionised region surrounding the of the cone of >∼ 30kpc. Adopting an opening angle for the
starburst will the Lyα opacity be sufficiently low to allow outflowof∼60◦,weestimatethattheshellmustbelocated
photonstopassthoughtheotherwiseneutraldisk(photons >∼30kpc from thestarburst region.
that pass through the ionised region are unlikely to be ab- At a velocity of ∼500 kms−1 it takes 60Myr to travel
sorbed by the foreground shell since they are redshifted by outtoadistanceof30kpc;withinthistime,supernovaewill
+2v ).Thecolumndensityoftheresidualgasdiskisalso provide a total energy of ∼6×1057erg. Clearly there are a
shell
likely to be much larger than that of the expanding shell. number of uncertainties in this estimate: not least that the
The result of this geometry is that the redshift wing will shell may have decelerated from an initial higher velocity
havesimilar spatial extent to themain Lyαline. which would reduce the timescale. Nevertheless, we can es-
Theproposedgeometryexplainsthespatialdistribution timatethemassoftheoutflowvia:Moutflow =NHI×MH/x
of the spectra without needing to invoke a special viewing where NHI is the column density, A is the area of the cone
angle.Theonlyconstraintisthatthesystemisvieweddown and MH is the atomic mass of Hydrogen. Adopting the
through the outflow cone. As the inclination between the value of the observed column density we derived in §3.2
outflow axis and the viewer becomes significant, we might of N = 1.6 × 1019cm−2 and that the cone is uniform
HI
expected to detect some rotation from the galaxy (a small over an area of 700 kpc2, the total mass of the outflow is
shearissuggestedbyourSINFONIspectraofthe[Oii]emis- 1.8×108/xM⊙, where x is its Hi fraction. The kinetic en-
sion),however,ashiftofaseveral10 kms−1 intheintrinsic ergyoftheoutflowisthenE = 1mv2 =5×1056/xerg.Thus,
2
profile of theLyαline is unlikelyto beobservable since the theoutflowisenergetically feasible iftheneutralfraction is
line shape is so strongly shaped bythe blue-shifted absorp- greaterthan10%,orthecolumndensityoftheshellislower
tion(whichdoesnotsharethevelocityshearofthegalaxy). than we have assumed. (Note, however, that these two fac-
Ifthesystemis viewedat alarger angle, Lyαislikely tobe torswelltendtoplayoffagainsteachother:alowercolumn
completely absorbed since the expansion velocity perpen- density will tend to havea lower neutral fraction).
dicular to the cone is small. In this case, it is unlikely that It is interesting to note that the implied mass outflow
a redshift would have been obtained for this arc. It is also rate is therefore ∼3/x M⊙yr−1. If we assume x=0.1, then
worth noting that the outflow spends much of its time in the mass outflow rate is very efficient at ejecting baryons
this post-blow-out phase. In earlier phases, the foreground from the galaxy, with the mass loading of the wind being
shellwillhavelowercolumndensityormayevenshowemis- more than twice thestar-formation rate of thegalaxy.
sion at its leading edge; in later phases, as the shell slows Theestimatesaboveareclearlyuncertain.Nevertheless,
downandstalls,Lyαdoesnotescapeatall.Thismodeldoes itisinterestingtoseehowfartheexpandingmaterialmight
not require us to be observing the galaxy at a particularly beexpected to travel after the end of thestarburst. We as-
special phase of its evolution or at a special viewing angle. sumeahalocircularvelocityof150 kms−1andawindspeed
Whilst the proposed bi-polar outflow fits the observ- of 500 kms−1 located 30kpc from the center of the halo. If
ables, there are other possible scenarios which would ex- the wind flows freely out of the galaxy it is completely un-
plain the velocity offsets we measure: if the outflow were bound. However, the wind will slow down dramatically if
instead within the galaxy, and the Lyα emission was non- it sweeps up material. To make an estimate of how far the
Gaussian (e.g. a Lorentzian) then the red wing of emission wind will travel we assume that the wind expands into an
seen at z =4.8860 could be explained without a receding NFWhalowithabaryonfractionof18%andthatthewind
shell.Inthismodel,theblue-shiftedmaterialcouldarisedue has an opening angle of 60◦. Even though thewind rapidly
to an ISM outflow which is in the process of breaking out gains mass as it expands, it reaches a minimum of 160kpc
of the galaxy. However, the constancy of the Lyα emission (almost 1 co-moving Mpc) before stalling. Thus, this star-
line shape and centroid is difficult to explain – any velocity burstdrivenwindhasenoughenergytopolluteavolumeof
motion within the star-forming regions should be reflected approximately ∼3Mpc3.
by theLyα, which is not seen in ourdata.
4.1.1 Energetics of the outflow 5 CONCLUSIONS
Although there are uncertainties in the geometry and col- The issues of metal ejection and feedback are some of the
umn density of the outflowing shell, it is nevertheless in- most outstanding problems in galaxy formation. One of
teresting to consider the energetics of the outflow, in or- the most important recent observational breakthroughs is
derto see if thebi-polar outflow scenario is reasonable. For that most high-redshift proto-galaxies appear to be sur-
a normal stellar initial mass function, supernovae provide roundedby“superwinds”whichareexpellingmaterialfrom
∼1049erg per solar mass of stars (eg., Benson et al. 2003), thegalaxydisc.However,athighredshift(wherestarforma-
thusforagalaxywithastar-formationrateof12±2M⊙yr−1, tionefficiencywasatitspeakandthereforethisphenomenon
∼1050ergareavailableperyearandthemaximumextentof is likely at its peak activity), observational evidence is usu-
the Lyα emission is ∼ 12kpc diameter. We estimate the allybasedonmeasuringthevelocityoffsetsbetweentheneb-
distance of the shell from the galaxy by assuming that the ularemission (such as[Oii]orHα) and rest-frameUVlines
linear size of the region covered by the swept up shell is at suchasLyαorUVISMlines.Theseobservationsareusually
least q times the size of the maximum extent of Lyα emis- based on longslit observations which lack spatial informa-
sion region. ForHaro 2, the shell is at least threetimes the tion, which is crucial if we are to understand the dynamics
spatialextentofthestar-forming regions.Ifweadoptq<3 and fate of theoutflowing gas.
thenwe would expectto detectvariations in thevelocity of In this paper, we have spatially resolved and mapped