MNRAS000,000–000(2017) Preprint31January2017 CompiledusingMNRASLATEXstylefilev3.0 A gravitationally-boosted MUSE survey for emission-line galaxies at z (cid:38) 5 behind the massive cluster RCS 0224(cid:63) Renske Smit,1 A.M. Swinbank,1,2 Richard Massey,1,2 Johan Richard,3 Ian Smail,1,2 and J.-P. Kneib4,5 1Centre for Extragalactic Astronomy, Durham University, South Road, Durham DH1 3LE UK 2Institute for Computational Cosmology, Durham University, South Road, Durham DH1 3LE UK 3Univ Lyon, Univ Lyon1, Ens de Lyon, CNRS, Centre de Recherche Astrophysique de Lyon UMR5574, F-69230, Saint-Genis-Laval, France 7 4Institute of Physics, Laboratory of Astrophysics, Ecole Polytechnique F´ed´erale de Lausanne (EPFL), Observatoire de Sauverny, 1290 Versoix, Switzerland 1 5Aix Marseille Universit´e, CNRS, LAM (Laboratoire d’Astrophysique de Marseille) UMR 7326, 13388, Marseille, France 0 2 n a Accepted2017January26.Received2017January16;inoriginalform2016September4 J 7 2 ABSTRACT ] We present a VLT/MUSE survey of lensed high-redshift galaxies behind the z = A 0.77 cluster RCS0224−0002. We study the detailed internal properties of a highly G magnified (µ ∼ 29) z = 4.88 galaxy seen through the cluster. We detect wide-spread . nebular Civλλ1548,1551˚A emission from this galaxy as well as a bright Lyα halo h p withaspatially-uniformwindandabsorptionprofileacross12kpcintheimageplane. - Blueshifted high- and low-ionisation interstellar absorption indicate the presence of a o high-velocity outflow (∆v ∼300kms−1) from the galaxy. Unlike similar observations r t of galaxies at z ∼2−3, the Lyα emission from the halo emerges close to the systemic s velocity - an order of magnitude lower in velocity offset than predicted in“shell”-like a outflow models. To explain these observations we favour a model of an outflow with a [ strong velocity gradient, which changes the effective column density seen by the Lyα 1 photons. We also search for high-redshift Lyα emitters and identify 14 candidates v between z =4.8−6.6, including an over-density at z =4.88, of which only one has a 0 detected counterpart in HST/ACS+WFC3 imaging. 6 1 Key words: galaxies: high-redshift – galaxies: formation – galaxies: evolution 8 0 . 1 0 1 INTRODUCTION cal emission-line tracers such as Hα and [Oiii], which are 7 traditionallyusedtomeasurethepropertiesoftheISM,are 1 Over the last decade, deep observations of blank fields, in : shifted to observed mid-infrared wavelengths for sources at v particularwiththeHubbleSpaceTelescope(HST)haveiden- z(cid:38)3−4.Thesmallphysicalsizesofgalaxiesatz>3com- i tifiedasubstantialpoplationofgalaxiesbeyondz>3,using X pared to typical ground-based seeing also makes spatially broadbandphotometry(e.g.Madauetal.1996;Steideletal. resolvedobservationsdifficulttoobtain,inhibitingmeasure- r 1996,1999;Sawickietal.1997;Lehnert&Bremer2003;Gi- a mentsofdynamicalmasses,star-formationdistributionsand avalisco et al. 2004; Ouchi et al. 2004; McLure et al. 2009; wind energetics. van der Burg et al. 2010; Bowler et al. 2015; Bouwens et Recently,thecommissioningoftheMultiUnitSpectro- al. 2015a; Finkelstein et al. 2015). Despite the progress in scopicExplorer(MUSE)ontheVeryLargeTelescope(VLT) identifyinglargenumbersofgalaxies,itremainschallenging has led to an advance in the identification and characteri- to obtain spectroscopic redshifts and determine the physi- sation of z ∼ 3 − 6 galaxies though wide-field and deep cal properties of these systems. This is largely due to their spectroscopy of the rest-frame ultraviolet (UV) spectra of inherent faintness and the fact that bright rest-frame opti- these sources. For example, MUSE is starting to probe the physical properties of Hii regions within galaxies by ex- ploiting gravitationally lensing through their faint UV neb- (cid:63) Partially based on observations made with the NASA/ESA ularemissionlinessuchasCivλλ1548,1551˚A,Heiiλ1640˚A, HubbleSpaceTelescope,obtainedattheSpaceTelescopeScience Institute,whichisoperatedbytheAssociationofUniversitiesfor Oiii]λλ1661,1666˚AandCiii]λλ1907,1909˚A(Karmanetal. ResearchinAstronomy,Inc.,underNASAcontractNAS5-26555. 2015; Caminha et al. 2016; Patr´ıcio et al. 2016; Vanzella et Theseobservationsareassociatedwithprogram#14497. al. 2016), lines which are rarely seen in local star-forming (cid:13)c 2017TheAuthors 2 R. Smit et al. galaxies(e.g.Hainlineetal.2011;Rigbyetal.2015).These (10.5 ks), the Advanced Camera for Surveys (ACS) using lines are produced either by young, metal-poor stellar pop- theF814W(I )filter(2.2ks)andtheWideFieldCamera 814 ulations with high-ionization parameters (e.g. Christensen 3 (WFC3) using the F125W (J ) and F160W (H ) fil- 125 160 et al. 2012b; Stark et al. 2014; Rigby et al. 2015), or gas ters(2.6kseach).TheACSandWFC3imageswerereduced photo-ionisation by faint active galactic nuclei (AGN; e.g. with Drizzlepac v2.1.3 to 0.05 and 0.128 arcsec pixel−1 res- Stark et al. 2015; Feltre et al. 2016). Furthermore, MUSE olution respectively. The depth of the I , J and H 814 125 160 has enabled the detailed modelling of extended Lyα emis- band images is 26.3, 26.8 and 26.7 mag respectively (5σ in sion, gaining insights into the inflowing neutral gas and/or a 0.5(cid:48)(cid:48)-diameter aperture). The WFPC2 data was reduced wind energetics in the circum-galactic medium (CGM) of withtheSTSDASpackagefromIRAFto∼0.1arcsecpixel−1 galaxies(Swinbanketal.2015;Binaetal.2016;Gullberget resolutionasdescribedbyS07.Afalse-colorimageusingthe al. 2016; Wisotzki et al. 2016; Vanzella et al. 2017). I , J and H bands are shown in Figure 1. The color 814 125 160 Moreover, MUSE is a promising new instrument for imageshowstwobrightarcsatz=2.40(lensedimagesB1– undertaking unbiased spectroscopic surveys. Bacon et al. B6) and z=4.88 (lensed images 1–4). (2015) used a 27 hour MUSE pointing of the Hubble Deep FieldsSouth(HDF-S)todetect89Lyman-αemittersinthe redshiftrangez∼3−6.Remarkably,66%oftheLyαemit- tersabovez(cid:38)5havenocounterpartintheHST broadband imaging (to a limiting magnitude of m ∼29.5), i In this paper, we extend current work on characteris- 2.2 MUSE spectroscopy ingtheUVspectraofintrinsicallyfainthigh-redshiftgalax- WeobservedtheclusterRCS0224−0002withasinglepoint- ies out to z ∼ 5 through the analysis of VLT/MUSE ob- ing (∼ 1 × 1 arcmin) of the VLT/MUSE IFU spectro- servations of one of the most strongly magnified galaxies graph (Bacon et al. 2010) between November 13, 2014, and known above a redshift of z > 3; the highly magnified September 16, 2015, programme 094.A-0141. Each individ- (µ=13−145×) z =4.88 lensed arc seen through the core ual exposure was 1500 seconds, with spatial dithers of ∼15 of the compact z = 0.77 cluster RCS0224−0002 (Gladders arcsec to account for cosmic rays and defects. One observ- et al. 2002; Swinbank et al. 2007, hereafter S07). ingblockwaspartlytakenintwilightandthereforeomitted S07 observed nebular [Oii] emission and an extended from the final data-cube, resulting in a co-added exposure Lyαhalointhisz=4.88sourceandtheyhypothesizedthat time of 13.5 ks. All the observations we use were taken in a galactic-scale bipolar outflow has recently bursted out of dark time with <0.8”V-band seeing and clear atmospheric this system and into the intergalactic medium (IGM). Our conditions. new observations obtain significantly higher signal-to-noise We reduced the data with the public MUSE ESOREX ratio(S/N)intheUVemissionandcontinuum,allowingus pipelineversion1.2.1,includingbias,dark,flat-fielding,sky to resolve the shape of the Lyα profile and detect the UV- subtraction, wavelength and flux calibrations. For each in- interstellar medium (ISM) lines. Furthermore, our MUSE dividualexposureweusedthelampflat-fieldtakenadjacent pointingcoversthecompletez∼6criticalcurves,whichal- in time to the observation for illumination correction. The lowsforanefficientsurveyforfainthigh-redshiftLyαemit- reduced data-cubes were registered and stacked using the ters. These sources are important targets to study in order EXP COMBINE routine. The seeing measured on the com- tounderstandthepropertiesoftheultra-faintgalaxypopu- binedexposureis∼0.68”fullwidthhalfmax(FWHM),with lation that could have contributed significantly to reionisa- aspectralresolutionof94kms−1(2.2 ˚A)FWHMat7000˚A. tion. A false-color image constructed from the final MUSE This paper is organised as follows: we describe our cubeisshowninFigure1.Weusemedianimagescenteredon MUSE dataset and we summarize the complementary data 5375˚A,6125˚Aand8275˚Aasbroadbandinputsandweadd presented by S07 in §2. We analyse the spectral properties a8˚Awidemeanimagecenteredon7146˚Atotheredchannel of the main z = 4.88 arc in §3. We present the results of a toemphasizetheLyαemissioninthez=4.88arc.Allbright blindsearchforLyαemittersin§4andfinallywesummarise HST sources are detected in the MUSE continuum, while our findings in §5. the z =4.88 arc is clearly detected with spatially extended For ease of comparison with previous studies we take Lyαemission.AnumberofotherLyαsourcesareidentified H = 70kms−1Mpc−1, Ω = 0.3,andΩ = 0.7, resulting 0 m Λ at the same redshift (see §4). in an angular scale of 6.4 kpc per arcsecond at z = 4.88. Magnitudes are quoted in the AB system (Oke & Gunn 1983). 2.3 SINFONI spectroscopy 2 DATA TocomplementtheMUSEdatasetweexploittheSINFONI 2.1 HST imaging IFU spectroscopy presented by S07. Briefly, the SINFONI We obtained HST imaging from the Space Telescope Sci- data was taken in the HK grating (λ/∆λ=1700) covering enceInstituteMASTdataarchive(GO:14497,PI:Smitand the [Oii]λλ3726.1,3728.8˚A doublet redshifted to ∼ 2.2µm. GO:9135,PI:Gladdders).RCS0224−0002(α=02:24:34.26, The ∼ 8×8 arcsec field-of-view (with a spatial resolution δ = −00:02:32.4) was observed with the Wide Field Plan- 0.25arcsecpixel−1)coversthelensedimages2and3ofthe etary Camera 2 (WFPC2) using the F666W (V ) filter z=4.88 arc . 666 MNRAS000,000–000(2017) A MUSE study of RCS 0224 3 Figure 1. HST I814, J125 and H160 (left panel) and MUSE B,V,I+7143˚A (right panel) colour images of the cluster core of RCS0224−0002 at z=0.77. The four images of the lensed galaxy at z = 4.88 are numbered 1–4 in the left panel and are recon- structed in the source plane in the inset panels. We also indicate the coverage of the SINFONI K-band spectroscopy and the images B1-B6ofazCiii] =2.4galaxyandimagesC1-C2ofaLyαemitteratzLyα =5.5thatareusedasconstraintsonthelensmodel(§3.1). Thez=4.8−6.5Lyαcandidatesselectedin§4aremarkedwithsquaresontherightpanel(see§4andAppendixD).FortheMUSEred channelwecombineabroad-bandcenteredon6125˚Awithan8˚A-widenarrowbandcenteredontheLyαhaloaroundthez∼4.88arc (Gladders et al. 2002); the Lyα emission is clearly extended beyond the continuum. Three sources (L2, L4 and L5) in our blind search forLyαemitters(§4)arefoundatthesameredshiftasthez=4.88arc,suggestingagalaxyover-density. Figure2.Leftpanel:TheMUSEone-dimensionalspectrumextractedoverthestellarcontinuumofthez=4.88arc(redline),smoothed with a Gaussian filter with σ =3˚A (the grey filled region shows the sky spectrum, offset from the spectrum for clarity). Regions with strongskyemissionlinesareomittedforclarity.ThesolidblacklineindicatesthestackedspectrumofthestrongestLyαemittingLyman break galaxies in the Shapley et al. (2003) sample. The spectrum is normalised to match the z =4.88 arc continuum at ∼7500 ˚A and redshiftedtoz[Oii] =4.8757.Theinsetpanelsshowthespectrawithrespecttothe[Oii]redshiftfortheSiiiλ1304˚A,Siivλ1394˚Aand Civλ1548˚A lines (indicated with black squares on the main spectrum). The spectra are binned to two times lower spectral sampling (2.5˚A/pix)andthepositionoftheOiλ1302˚A,Siiiλ1304˚A,Siivλλ1394,1403˚AandCivλλ1548,1551˚Alinesareindicatedwithdashed lines. The parts of the spectra strongly affected by skylines are indicated with a dashed line in the insets. We detect a single-peaked, strongly asymmetric Lyα line, narrow and strong Civ emission and narrow blueshifted Civ and Siiv absorption lines. All absorption featuresaresignficantlyblueshiftedwithrespecttothe[Oii]redshift.Rightpanel:Zoom-inoftheobservedLyαline(blueline,toppanel) andaGaussianwithVoigt-profileabsorberfittothedata(blacklines).ThenarrowLyαpeakincombinationwiththehigh-velocitytail arenotwelldescribedbythissimplemodelasshownbytheresiduals(bottompanel). MNRAS000,000–000(2017) 4 R. Smit et al. 3 ANALYSIS AND DISCUSSION Table 1. Detected spectral features of the z = 4.88 arc (inte- gratedovergalaxyimages1,2and3) 3.1 Lens model To constrain the intrinsic properties of the emission-line line z ∆v [kms−1]a EW0 [˚A] galaxies in this study we require an accurate lens-model. S07 constructed a simple mass-model of RCS0224−0002 Emissionlines with the two main elliptical galaxies in the centre of the Lyαhalo 4.8770±0.0005b 68±37 278±55 cluster and the dark matter component approximated by Lyαcont 4.8770±0.0005b 68±37 135±27 single truncated pseudo-isothermal elliptical mass distri- Siiiλ1304˚A 4.8761±0.0006 21±41 1.1±0.1 butions. Their primary observational constraints on the Siivλ1394˚A 4.8738±0.0003 -96±29 0.6±0.1 mass configuration are the four lensed images of the Siivλ1403˚A 4.8752±0.0006 -27±39 0.6±0.1 z = 4.88 arc. However, our MUSE observations also Civλ1548˚A 4.8753±0.0001 -23±26 5.6±0.4 cover the other arcs in the cluster. We extract spectra Civλ1551˚A 4.8755±0.0001 -12±26 3.7±0.3 over the multiply-imaged central blue arcs (B1–B6 in Fig- ure 1) and detect Ciii]λλ1907,1909˚A, Oiii]λλ1661,1666˚A Absorptionlines emission and Siiiλ1403˚A, Siivλλ1394,1403˚A, Siiiλ1526˚A, Siiiλ1304˚Ac 4.8710±0.0001 -240±25 -0.2±0.1 Civλλ1548,1551˚A,Feiiλ1608˚A,Aliiλ1671˚Aabsorptionin Siivλ1394˚A 4.8694±0.0001 -322±26 -1.1±0.1 images B1–B6 and measure a redshift zCiii] =2.396±0.001 Siivλ1403˚A 4.8693±0.0002 -327±28 -0.5±0.1 from the integrated light of these images (Smit et al., in Civλ1548˚A 4.8694±0.0001 -322±26 -0.8±0.1 preparation). Weusethesenewconstraintstoupdatethelensmodel a Velocityoffsetwithrespecttothethesystemicredshift byS07.AsinS07,thelensmodellingisperformedusingthe z[Oii]=4.8757±0.0005.Uncertaintiescombinetheuncertainty LENSTOOL software (Kneib et al. 1996; Jullo et al. 2007; onthelineredshiftwiththeuncertaintyonthe[Oii]redshift.b Jullo&Kneib2009).LENSTOOLisaparametricmethodfor UsingthepeakoftheLyαline.c BlendedwiththeOiλ1302˚A modelling galaxy clusters that uses a Markov Chain Monte line. Carlo (MCMC) fit for a specified number of mass peaks. droppingtheprior(andsimultaneouslyimposingconstraints Each mass peak corresponds to a dark matter halo mod- fromnewlyidentifiedlenssystems),acluster-scalemassdis- elled with a truncated pseudo-isothermal elliptical that is tributionwithellipticityε=0.63achievesrmsA =0.52(cid:48)(cid:48),or characterisedbyaposition(RA,dec),velocitydispersionσV, i rms = 1.03(cid:48)(cid:48) for all image systems. However, we achieve ellipticity ε, truncation radius r and core radius r . i cut core a still-better fit (rmsA = 0.48(cid:48)(cid:48), rms = 0.88(cid:48)(cid:48)) using two For our updated mass model we include mass compo- i i cluster-scale halos. These were given a Gaussian prior cen- nents for the brightest 22 cluster members and two compo- tred on the two BCGs. The first gets asymmetrically offset nents for the cluster halo. We include constraints from the totheNorth-West;thesecondremainsnearCG2.Thistwo- six images B1–B6 of the z = 2.396 galaxy arc, including halomodelachievesasuperiorlog(Likelihood)of–26.57and the de-magnified image in the centre. Another faint arc is χ2 = 61.3 in 11 degrees of freedom compared to the best- identified in the HST imaging (D1–D3 in Table A1), just fittingone-halomodel,whichhaslog(Likelihood)of–142.92 inside the z = 4.88 arc, but we do not detect any emis- and χ2 =294 in 17 degrees of freedom. The best-fit param- sionlinesfromthissourceintheMUSEdata-cube.Further- eters are listed in Table A2 in Appendix A. more, we search the MUSE cube for bright multiply lensed line emitters and find a Lyα emitter without an HST con- tinuumcounterpartatz =5.500±0.002(labelledC1at Lyα 3.2 The z=4.88 arc α=02:24:34.86,δ=−00:02:16.2andC2atα=02:24:34.02, δ=−00:02:36.3inFigure1).ThelocationsoftheLyαemit- The z = 4.88 arc was first discovered in the Red-Sequence terimagesarewellpredictedbythelens-modelthatusesall Cluster (RCS) by Gladders et al. (2002). Gladders et al. otherconstraintsandthereforeweincludethisdoublylensed (2002) detected the bright Lyα emission in galaxy images image as an additional constraint. In Figure 1 we show the 1–3 at z =4.8786 with VLT/FORS-2 spectroscopy. S07 Lyα criticalcurveofournewmodelataredshiftofz=4.88and targetedthearcwithVLT/VIMOS(galaxyimages1–4)and we list all multiple images used to constrain the model in VLT/SINFONI (galaxy images 2–3) spectroscopy and de- Table A1 in Appendix A. tected Lyα at z = 4.8760 and [Oii]λλ3726.1,3728.8˚A Lyα OurmassmodeldiffersfromthatofS07inthreeways. at z[Oii] = 4.8757. S07 measured a star-formation rate of First, owing to the different assumed cosmology, our model 12±2M yr−1, a velocity gradient of (cid:46) 60kms−1, and an (cid:12) is ∼ 35% less massive: M = (3.8±0.2)×1014M com- estimated dynamical mass of ∼ 1010M within 2kpc from (cid:12) (cid:12) pared to M = (5.9±0.4)×1014M . We recover the S07 the [Oii] emission lines. (cid:12) mass if we switch back to their cosmology. Second, the in- For our MUSE study of the z = 4.88 arc we will as- clusion of mass from all cluster member galaxies makes the sume the systemic velocity of the galaxy is best estimated z = 2.4 critical line better match the observed features of by zsys = z[Oii] = 4.8757±0.0005 (integrated over galaxy lensed system B. Third, our distribution of mass is more images 2–3). Furthermore, from our lens-model we find lu- elongatedtowardtheNorth-West.TheS07modelhadclose minosity weighted amplifications of µ = 29+9 , µ = 21+12, −11 −8 to circular symmetry, forced by a prior on the ellipticity µ = 138+7 and µ = 1.30+0.01 for images 1, 2, 3 and 4 re- −74 −0.01 of the cluster-scale dark matter. This resulted in a scatter spectively(notethatimage3hasaveryhighamplification, of rmsA = 1.21(cid:48)(cid:48) between the predicted and observed posi- but also a very large uncertainty, because the arc crosses i tions of images A1-A4 (G. Smith et al. private comm.). By thecriticalcurve).Thesevaluesareslightlyhigherthanthe MNRAS000,000–000(2017) A MUSE study of RCS 0224 5 mean,luminosity-weightedmagnificationofµ=16±2found theforegroundcontamination.Wemaskanyremainingflux by S07 for images 1,2 and 3 integrated (though within the from foreground sources by hand. Furthermore, we extract uncertainties for images 1 and 2). The uncertainty on our the PSF from a nearby star and place this at the position numbers is largely due to the fact that a small shift of the of the Lyα peak in our lens model in order to construct criticalcurvecanchangetheluminosityweightedamplifica- the source-plane image of the PSF and measure its spatial tionsignificantly.Inparticular,wenotethatthehighmagni- profile. ficationofimage3isdominatedbyafewpixelsthatoverlap TheLyαhaloappearsroughlyisotropic,withlittlesub- with the critical curve, while the estimated magnification structure,exceptforanextendedlowerluminosityregionin for any modelling method is most uncertain near the criti- the South-East. Comparing the Lyα halowiththe UVcon- cal curves (see Meneghetti et al. 2016). tinuum image in Figure 3 indicates the extended nature of TomeasurethedetailedpropertiesoftheUVspectrum thefaintLyαprofilebeyondthecontinuum.TheLyαhalois of this galaxy we first construct a one-dimensional spec- consistentwithanexponentialprofile.Forcomparisonwith trum (up to ∼ 1600˚A in the rest-frame) of the z = 4.88 Lyα halos in the literature, we measure the Petrosian ra- arcfromtheMUSEcubebymeasuringtheintegrated(non- dius (Petrosian 1976) of the halo, defined as the annulus weighted) spectrum extracted from pixels in the lensed im- wheretheLyαfluxisequaltoη timesthemeanfluxwithin ages1,2and3withaS/N>2σinthecontinuumimageofthe theannulus.ThePetrosianradiusisausefulmeasure,since MUSE data-cube. The resulting spectrum is shown in Fig- it is only weakly dependent on the seeing of the observa- ure2.AswellasbrightLyαemission,whichhasanobserved tions. For η = 20% we find R = 8.1±0.4kpc and p20,Lyα equivalent width (EW) of 793±159˚A (rest-frame EW = R = 2.9 ± 0.8kpc, which is somewhat lower than 0 p20,UV 135 ± 27˚A), we clearly detect the absorption line dou- the range of Petrosian radii R ∼ 10 − 30kpc (for p20,Lyα bletSiivλλ1394,1403˚A,whichoriginatesintheISMand/or R ∼ 1.3−3.5kpc) found by Wisotzki et al. (2016), p20,UV CGMandtheemissionlinedoubletCivλλ1548,1551˚A,with but similar to some of the largest Lyα halos around z ∼ 0 someevidenceforanabsorptioncomponentaswell(seeinset analogues in the LARS sample (Hayes et al. 2013), which panels), which is likely to arise from a combination of stel- typically show radii R (cid:46)8kpc (Figure 4). p20,Lyα lar,nebularandISM/CGMcomponents.Theobservational Faucher-Gigu`ere et al. (2010) and Rosdahl & Blaizot parameters of the UV spectroscopic features in the MUSE (2012) use radiative transfer simulations to investigate the dataarelistedinTable1(seeAppendixBformeasurements expectedmorphologyofLyαhalosgeneratedbycoolingra- on the individual lensed galaxy images). diation and they predict concentrated emission, which can Inthenextsectionswewillfirstdiscussthemorphology extend out to 10−30kpc. While stacked Lyα halos extend oftheemissionlines,beforemovingtoadetailedanalysisof out to ∼ 100kpc (Steidel et al. 2011; Matsuda et al. 2012; thespectralpropertiesofthez=4.88arc,thekinematicsof Momose et al. 2014), indicating cooling is not the origin of thesystemandthephysicalpicturethatemergesfromthese Lyα emission in typical galaxies, our observations do not observations. have sufficient S/N to trace the z = 4.88 Lyα halo be- yond 10kpc, necessary to rule out a gas cooling scenario. We will therefore further investigate whether the faint, ex- 3.2.1 Lyα morphology tendedLyαemissionisproducedbyresonantscatteringfrom The Lyα emission in the z = 4.88 arc (see Figure 1) ap- a central source or by cooling radiation in §3.2.2 based on pears to be significantly extended. Lyman Break galaxies the spectral properties of the line. and Lyα emitters at z ∼ 2−6 often exhibit extended Lyα halosaroundthestellarcontinuumofthegalaxies,following 3.2.2 Spectral properties of the Lyα line anexponentialsurfacebrightnessdistribution(Steideletal. 2011; Matsuda et al. 2012; Momose et al. 2014; Wisotzki High-redshiftLyαemitterscanexhibitawiderangeofspec- et al. 2016). These Lyα halos are thought to be generated tral properties, such as blueshifted and redshifted emission, either by cooling radiation (e.g., Dijkstra et al. 2006; Dijk- singleanddoublepeakedlinesanddifferentlinewidthsand stra & Loeb 2009; Faucher-Gigu`ere et al. 2010; Rosdahl & velocity offsets, which gives insight into the emission mech- Blaizot2012)orbyresonantscatteringfromacentralpower- anismofLyαandthecolumndensityandvelocitydistribu- ingsource,suchasstar-formationorAGN(e.g.,Verhamme tionoftheISMandCGMneutralgas(e.g.Verhammeetal. etal.2006;Gronkeetal.2015).First,weinvestigatethemor- 2006; Gronke et al. 2015). phologyofLyαinthez=4.88arcbehindRCS0224−0002. For the z = 4.88 arc, the Lyα emission line profile is Figure 3 shows the source-plane reconstruction of the very asymmetric and we find a single redshifted Lyα line, continuum subtracted Lyα halo. We use image 4, since the with a peak at z = 4.8770±0.0005 (using the wave- Lyα Lyα halos of images 1 to 3 are incomplete and merged to- lengthandwidthofthespectralelementwhereLyαpeaks), gether(seetherightpanelofFigure1).Forthespatialpro- ∼40−90kms−1 redshifted with respect to the [Oii] emis- file we use bins of 0.1 arcsec in concentric circles around sion which marks the systemic redshift, and FWHM = red the peak flux of the Lyα emission. The MUSE Lyα halo 285kms−1, with very little flux bluewards of the [Oii] red- has an observed FWHM of 2.2 kpc, while the HST con- shift (see Figures 2 and 5). We set an upper limit on the tinuum has a FWHM of 0.2 kpc. A number of foreground presence of a weaker blue line; at −v we find an upper red clustergalaxiescontaminatethemeasurementofthestellar limit on the flux ratio of any blue peak to the red peak spatialprofiledirectlyfromthecontinuumimage.Therefore F /F <0.027. Furthermore, we detect a faint peak,blue peak,red weconstructabroadbandimageredwardsofLyα,centered tailofredshiftedLyαemissionoutto∼1000kms−1.Asim- on7400˚A,andwesubtractacontinuumimagebluewardsof plified model for this asymmetric line shape is that of a theLyαbreakcenteredon6975˚Ainordertoremovemostof GaussianemissionlineprofileconvolvedwithaVoigtprofile, MNRAS000,000–000(2017) 6 R. Smit et al. Figure 3. Left: The source-plane reconstruction of the Lyα halo of the lensed galaxy image 4. White contours indicate 50% and 10% ofthepeakfluxintheLyαhalo,whiletheyellowellipseinthecornerindicatesthe50%peakfluxforthede-lensedPSFoftheMUSE continuum image. The black contour indicates 50% of the peak flux of the UV continuum observed in the HST I814 band. The Lyα fluxisdominatedbyasmoothandlargelycircularlysymmetrichalothatisextendedbeyondthestellarcontinuum,butwealsoidentify an extended region to the South-East (North-West in the image plane). Right: The exponential surface-brightness profile of the Lyα halo,comparedtotheMUSEcontinuum(usingafore-groundsubtractedbroad-bandimageat7400˚A)surface-brightnessprofileandthe MUSEPSFmeasuredfromthesource-planereconstructionofgalaxyimage4.TheLyαhaloisextendedbeyondthestellarcomponent measuredfromtheUVcontinuum. shape; instead we observe an exponential profile as a func- tionofvelocityovertwoordersofmagnitudeinflux(Figure 5),remarkablysimilartotheexponentialsurfacebrightness profile (Figure 3). S07 observed the modest redshifted narrow Lyα line in combination with the high velocity tail, and interpreted this as a combination of emission from the central source combinedwithredshiftedemissionfromanoutflow.Totest this model, in Figure 5 we show the spatial variation of the spectralLyαprofileintheimage-planeofthelensedgalaxy image1(thehighestS/Nimage).Whileweusedthesource- planereconstructionofgalaxyimage4forderivingthespa- tial properties of the Lyα halo, we use the brightest galaxy image for spectral analysis to obtain higher signal-to-noise information. First, we partition the halo along contours of constant observed Lyα flux. While the Lyα flux in the halo drops by more than an order of magnitude compared to the emission over the stellar continuum, the shape of the Lyαprofile,afternormalisingtothepeakflux,changesonly marginally from the Lyα profile extracted over the stellar continuum. Across the lensed image, the wavelength of the Figure4.AcomparisonoftheLyαandUVcontinuumPetrosian peakoftheLyαlinechangesbylessthan∼50kms−1,while scale radii for the z = 4.88 arc and including measurements at thewidthandtheshapeofthehighvelocitytailstaysnearly z = 0−6 from the Lyα reference sample (LARS; Hayes et al. constant (see also Patr´ıcio et al. 2016, for a similar pattern 2013)andMUSEHDF-Sobservations(Wisotzkietal.2016). in a Lyα halo at z=3.51). Theseresultsdifferstronglyfromthescenariodescribed describing the collisional and Doppler broadening of inter- inS07,wherethemainpeakoftheLyαprofilecomesdirectly stellar absorption lines, as shown in Figure 2, where we fix from the star-forming regions and the high velocity wing is theredshiftoftheunderlyingGaussianemissiontothe[Oii] re-scattered in an expanding shell of gas within the CGM. redshiftz=4.8757.Thebest-fitmodelinFigure2indicates For this model to hold we would expect the star-formation a Hi absorber with a column density of 1019cm−2, how- component (the peak of Lyα) to drop off rapidly with in- ever,thefitfailstoreproduceboththenarrowpeakandthe creasing radius, while the back-scattered CGM component high-velocity tail of the Lyα line. In fact, the emission-line changeslittlewithradius,andthereforethepeakfluxwould component of the Lyα line shows a strongly non-gaussian shifttohighervelocitiesandtheshapewouldchangesignif- MNRAS000,000–000(2017) A MUSE study of RCS 0224 7 Figure 6. Comparison of the emission line ratio Civ/Lyα as a function of rest-frame Lyα equivalent width with similar de- Figure 5. The spatial dependence of the Lyα velocity profile tections in the literature. We show the measurements along the (withrespecttothe[Oii]redshift)indifferentbinsintheimage lensed image 1 of the z = 4.88 arc. We also show data from planeofgalaxyimage1.Theblacklineindicatesthefluxwithin lensedCivemittersatz∼2foundbyChristensenetal.(2012b), acontouroverthestellarcontinuumfluxandthegreenlinesare Starketal.(2014),Starketal.(2015),Caminhaetal.(2016)and emittedwithincontoursofconstantLyαflux.Thelightblueline Vanzella et al. (2016). Narrow-line AGN have been observed to indicatesaregionofhigh-velocityexcessflux.Thepixelsusedto showaemissionlineratioofCiv/Lyα(cid:38)0.2(Shapleyetal.2003; generatethefourspectraareshownintheinsetpanel.Thespatial Erbetal.2010;Hainlineetal.2011).Alongthez=4.88arcthis variation of the Lyα halo is surprisingly small, with the peak of emission line ratio is steady and always below Civ/Lyα < 0.1 theLyαlinevaryinglessthanthewidthofonespectralelement (see§3.2.3). correspondingto∼60kms−1.Thesmalldifferenceinthevelocity profilebetweentheinnerandouterhalo,wouldsuggestthatthe bulk of the Lyα photons are CGM generated/rescattered, since thereisnoevidenceforanISMcomponentthatfallsawayaswe et al. (2010) model such scenarios using radiative transfer observe Lyα further away from the stellar continuum. Spatially simulations and find that Lyα should typically be double offsetfromthesphericalhalo,wedetectexcessfluxoutto1000km s−1 (corresponding to the extended region to the South-East in peakedandblueshiftedwithrespecttothesystemicvelocity of the galaxy. Assuming these models provide a reasonable the source-plane image shown in Figure 3), possibly indicating a collimated high velocity outflow superimposed on an isotropic descriptionofthesystem,thesingleredshiftedLyαpeakwe componentthatdominatesthetotalLyαemission(see§3.2.1). observe excludes cooling as a source of Lyα photons in the z=4.88 arc. To reproduce the Lyα line profile for the z = 4.88 arc icantly. The spatially-uniform Lyα spectral profile instead we thus favour a picture where a central powering source is suggests that the Lyα peak is also produced or resonantly surroundedbyalargelyisotropichaloofneutralgas,which scattered within the CGM, which generates Lyα emission dampens Lyα bluewards of the systemic velocity and reso- with a wide range in velocities. nantlyscattersthevastmajorityofphotonstowardshigher Wecantestthisfurtherbysearchingforanydeviations velocities within the expanding gas behind the galaxy (e.g. fromtheaverageLyαprofile.Weusethespectrumextracted Dijkstra et al. 2006; Verhamme et al. 2006). overthestellarcontinuumasamodelforfittingtheLyαline The strong similarity between the Lyα surface bright- ineachindividualpixelofthez=4.88arc,leavingthenor- ness profile (Fig. 3) and the spectral profile in the z =4.88 malisation as the only free parameter and considering only arccouldsuggestthepresenceofasmoothlyvaryingvelocity the peak of the Lyα line as a model constraint. After sub- gradient in the CGM gas that resonantly scatters the Lyα tracting our one-parameter model we detect only a weak photons into our line of sight. This scenario is qualitatively residual. In Figure 5 we show the spectrum extracted over in good agreement with the Lyα profile considered in Ver- the region with the largest residual, which shows a slightly hamme et al. (2006, see their Fig. 7), shown in Appendix offset peak compared to the Lyα extracted over the stel- C, where the relatively low column density (at a given ve- lar continuum and a broadened profile out to 1000kms−1, locity) created by the strong velocity gradient causes the indicating a collimated high-velocity outflow on top of the escape of Lyα photons predominantly at low velocity (and isotropic Lyα halo component that is described above. small radii), while a weak high-velocity tail is still observed GiventheextendednatureoftheLyαemission(§3.2.1) due to the photons that are resonantly scattered through wenowconsidervariousgenerationmechanismsfortheemis- theacceleratingoutflow.Amodelwithagasvelocitygradi- sion. Given the spatially invariant Lyα line profile, which ent furthermore predicts the absence of a blue peak, which indicates that only a minor fraction of the Lyα emission is difficult to reproduce in a shell model, since a low ve- reaches us directly from the galaxy, it is reasonable to con- locity shell with low covering fraction which gives rise to a siderwhetherthehalocanbeproducedbycoolingradiation red Lyα peak close to the systemic velocity also produces a from the CGM. Dijkstra et al. (2006) and Faucher-Gigu`ere nearlysymmetricbluepeak.In§3.3wewilldiscusshowthis MNRAS000,000–000(2017) 8 R. Smit et al. Figure 7. The spatial distribution of the Civ emission lines along the lensed galaxy images 1, 2 and 3 of the z = 4.88 arc (green contours,leftpanel),comparedtothe[Oii]distributionoverthelensedimages2and3fromtheSINFONIdata(S07,browncontours, rightpanel).TheHST (V666+I814)-bandimageisshowningreyscale.Thehigh-ionisationemissionlinesarespatiallyextendedalong thearcandtracetheUV-continuumlight.ThemorphologiesofCivand[Oii]showsimilaritiesthatmightimplyanebularoriginofCiv inalargenumberofthestar-formingregionsofthez=4.88arc. Figure8. Left:PredictedCivequivalentwidthevolutionofnebularCivemissionduringthefirst100Myrofasinglestellarpopulation atZ=0.05,0.2,0.4Z(cid:12)metallicity(solid,dot-dashedanddashedlinesrespectively)usingtheBPASSstellarpopulationsynthesismodels including binary rotation (Eldridge & Stanway 2012). The green region indicates the rest-frame equivalent width observed (including uncertainty)intheintegratedspectrumofthez=4.88arc.ThenebularCivemissioncouldbeproducedbyaverylow-metallicitystellar population if the star clusters are observed during the first ∼ 3 Myr of their lifetime. Right: A comparison of the stellar Civ profiles (excluding nebular emission) predicted by the same three BPASS models at 3 Myr with our data. The binned points on either side of thenebularCivlinesshownoevidenceforthestrongP-Cygniprofilesproducedbystellarwinds,indicatingametallicityofthestellar populationlowerthantheZ=0.05Z(cid:12) BPASSmodel. MNRAS000,000–000(2017) A MUSE study of RCS 0224 9 model fits into a physical picture that can explain our full ual star-forming clumps along the arc show slopes between set of observations. β=-1.68−-2.64. Since high-redshift galaxies that host faint AGN have measured UV continuum slopes of β ∼-1.4−-0.3 (Hainlineetal.2011;Giallongoetal.2015)andwetherefore 3.2.3 CIV emission concludethatlow-massaccretingblackholesareunlikelyto The detection of narrow (FWHM = 156 ± 16kms−1) contributetotheradiationfieldgivingrisetotheCivemis- Civλλ1548,1551˚Aemissioninthez=4.88arcisinteresting sion. sincetheUVspectraoffieldgalaxiesgenerallyshowCivin absorption from ISM/CGM gas, or else exhibit a P-Cygni 3.2.4 Metal absorption lines profile from the stellar winds of O-stars, with Civ emission redshifted by a few hundred kms−1 (Leitherer et al. 2001; The high-ionisation Civλλ1548,1551˚A line is expected to Shapleyetal.2003;Jonesetal.2012).AGNcanalsoproduce beacombinationofnebular,stellarandISM/CGMcompo- Civ in emission, though with typical line-widths of at least nents. While no absorption by the ISM appears present at afewhundredkms−1.NarrowCivλλ1548,1551˚Ahassofar the systemic velocity of Civ, blueshifted Civ absorption is been observed in a handful of strongly lensed high-redshift observedat∆v=−322±26kms−1 whichalsohasanarrow galaxies (e.g. Holden et al. 2001; Christensen et al. 2012b; profile, with FWHM∼200kms−1. Starketal.2014,2015;Caminhaetal.2016;Vanzellaetal. The spectrum displays a strong similarity between the 2016). To date these galaxies have either been studied with Civλ1548˚A and Siivλ1394˚A absorption profiles (Figure slitspectroscopyortheyareunresolvedinground-basedob- 2), indicating the absorption of both lines is due to highly servations, inhibiting the study of the spatial distribution ionisedgascloudsintheISM/CGMofthegalaxymovingto- of the Civλλ1548,1551˚A. The MUSE observations of the wards us. Furthermore, both Siivλ1394˚A and Siiiλ1304˚A brightly lensed z = 4.88 arc therefore provides us with a show no evidence for absorption at the systemic velocity unique opportunity to investigate the origin of this line in (Figure 2) and we even find weak emission lines, possibly more detail. indicating a low covering fraction of gas in the ISM of the In the absence of rest-frame optical spectroscopy, a galaxy. Given the strong absoprtion of the high-ionisation common approach to assessing the possible presence of linesat∼−300kms−1,galacticfeedbackinthisgalaxyhas AGN is using UV emission line ratios (e.g., Stark et al. possibly ejected a large fraction of the interstellar gas into 2015; Feltre et al. 2016). At z = 4.88, this requires the CGM/IGM. near-infraredspectroscopytomeasuretheHeiiλ1640˚Aand The Siiiλ1304˚A absorption appears to be weaker than Ciii]λλ1907,1909˚A lines. With the current observations we the Civλ1548˚A and Siivλ1394˚A absorption lines. The ra- can only assess the Civ/Lyα ratio, which has a ratio of tio of the equivalent with of the Siiiλ1304˚A line over that (cid:38)0.2inthecompositespectraofAGN(Shapleyetal.2003; of Siivλ1394˚A is EW(Siiiλ1304˚A)/EW(Siivλ1394˚A)=0.2 Hainline et al. 2011). In contrast, this ratio is Civ/Lyα = (see Table 1). This is in contrast to the typical UV spectra 0.054±0.006inthez=4.88arc,withlittlevariationalong of high-redshift galaxies, where the low-ionisation lines are the images, consistent with the interpretation that this line stronger than the high-ionisation lines of the same species, is associated with star formation and not with a hidden with for example EW(Siiiλ1304˚A)/EW(Siivλ1394˚A)=1.2 AGN. Figure 6 shows the Civ/Lyα ratio as a function of inthecompositespectrumofLymanbreakgalaxiesbyShap- Lyα equivalent width for various lensed galaxies in the lit- ley et al. (2003). We also fit the low- and high-ionisation erature (Christensen et al. 2012b; Stark et al. 2014, 2015; lines with a Gaussian profile convolved with a Voigt-profile Caminha et al. 2016; Vanzella et al. 2016) and in 5 spatial absorber and estimate column densities of logN/cm−2 = bins along the lensed image 1 of the z = 4.88 arc (using 14.5±0.3andlogN/cm−2 =14.6±0.1respectively.Possibly both the 1548 and 1551˚A lines). The Civ/Lyα ratio and alargerfractionoftheoutflowinggasishighlyioniseddueto the observed equivalent width of Civ 55±2˚A (rest-frame ahardionisationfield,whichispresentgiventhewidespread EW =9.3±0.4˚A)donotchangesignificantlyasafunction nebularCivλλ1548,1551˚Aemissioninthegalaxy.Itispos- 0 ofpositionalonggalaxyimage1.Furthermore,weobserveno sible that the neutral gas swept up by the ionised outflow strongemissionfromtypicalAGNlinessuchasNvλ1240˚A, is optically thin because of this, or else that the covering Sivλλ1393,1402˚AandNivλλ1483,1486˚A(e.g.,Hainlineet fraction of neutral gas in the CGM is incomplete (e.g., Erb al. 2011). 2015). To distinguish between these explanations we would In Figure 7 we show the spatial distribution of needacleanobservationoftheSiiiλ1260˚AandSiiiλ1527˚A Civλλ1548,1551˚A, using a continuum subtracted narrow- absorption features, which are currently obscured by sky- band image of the Civ emission and overlaying the con- lines. tours on the HST continuum image. Civ clearly extends along the arc and shows a morphology that is consistent 3.2.5 Stellar population with the [Oii] emission. While we would expect a centrally concentrated source for Civ if it was originating from an GiventhattheCivemissionappearstobenebularinorigin AGN, we can distinguish at least four different ‘clumps’ in and powered by star formation (see §3.2.3), we investigate theCivmorphologywithsimilarbrightness,suggestingthat the properties of the stellar population that are needed to the Civ emission is nebular in origin and emerging from reproduce the ∼ 9˚A rest-frame equivalent width nebular multiple star-forming regions throughout the galaxy. Civ emission. Finally, we measure the UV-continuum slope β (f ∝ Figure8showstheevolutionofthenebularCivequiva- λ λβ) from the J − H color of galaxy image 1 (inte- lentwidthwithmetallicity,obtainedfromtheBinaryPopu- 125 160 grated flux) and find β =-2.19±0.14, while the individ- lationandSpectralSynthesis(BPASS;Eldridge&Stanway MNRAS000,000–000(2017) 10 R. Smit et al. Figure10.Acomparisonofthevelocityoffsetsfromthesystemic redshiftfortheLyαpeakandtheinterstellarabsorptionlineSiiv for the z =4.88 arc and including similar measurements at z = 0−2 from the Lyα reference sample (LARS Rivera-Thorsen et al. 2015) and UV-selected star-forming galaxies observed with the LRIS spectrograph on Keck (Erb et al. 2006). A“shell”-like outflow is expected to have the Lyα peak be shifted by ∼ 2× Figure 9. The spatial variation of the various line components along a tangential line through lensed image 1 of the z = 4.88 ∆vIS (dashedline),whilethez=4.88arcisamongtheveryfew arc (from North-East to South-West in the image plane, corre- galaxieswithaLyαpeakshiftbelow∆vLyα=∆vIS (solidline). spondingtoalinefromNorthtoSouthinthesourceplane)with respecttothe[Oii]redshiftofz=4.8757measuredfromgalaxy images 2 and 3. The Civ emission is close to the [Oii] redshift, dance of the stellar population of the z = 4.88 arc could while the Siiv interstellar absorption lines are blue-shifted by be even lower than that assumed in the lowest-metallicity ∼300−400kms−1 (thesolidpurplepointcorrespondstoamea- BPASS models available. The reasonable consistency be- surementcombininggalaxyimages1,2and3toobtaina>5σline tween the stellar and nebular components in the Civ line detection).ThepeakoftheLyαhaloisredshiftedby∼80kms−1, profile provides confidence that we are indeed witnessing with very little velocity structure along the arc. The width the the earlystar formationinagalaxywithavery metal-poor Lyα line increases on one side of the galaxy as indicated by the stellar population. positions of the half, 25%, and 10% of the max flux in the line. We find no evidence of a strong velocity gradient in the nebular lines,andthepeakoftheLyαemissionstaysnearlyconstanteven 3.2.6 Kinematics beyondthegalaxy.Thereis,however,awideningoftheLyαpro- fileandaweakvelocitygradientintheSiivabsorption,possibly To derive spatially resolved dynamics of the stars and gas, indicatingacollimatedhigh-velocityoutflow. we exploit the enhanced spatial resolution of the strongly lensed z = 4.88 arc and the IFU data to understand the spatial variation in the kinematics of the ISM/CGM. 2012)models,usingasinglestellarpopulationandincluding In Figure 9 we show the spatial variation of Lyα, binarystellarevolution.Thehighlyionisingphotonsneeded Civλλ1548,1551˚AandSiivλλ1394,1403˚Aalongthelensed togeneratethehigh-equivalentwidthnebularCivlinescan galaxyimage1runningfromtheNorth-EasttoSouth-West. be generated by a young stellar population (1−3 Myr old) We use galaxy image 1, since galaxy images 2 and 3 are in- with a low metallicity Z =0.05Z . complete images that cross the critical curve (see Figure 1) (cid:12) The low-metallicity BPASS models also predict signifi- andthestellarcontinuumofgalaxyimage4isnotspatially cantlyreducedequivalentwidthstellarP-Cygniprofiles,due resolved in the MUSE data. The velocities in Figure 9 are to the fact that the winds of hot stars are driven by metal given with respect to a redshift of z[Oii] = 4.8757 obtained lineabsorptionandthereforelow-metallicitystarsaremuch from galaxy images 2 and 3 (S07). lessefficientindrivingstellarwinds.Thisisconsistentwith Asnotedin§3.2.2,theLyαprofileisnotwelldescribed the observed profile of Civ in the z = 4.88 arc (see Fig- by a traditional Gaussian profile convolved with a Voigt- ure 8), where we see no evidence for any redshifted stel- profile absorber and we therefore choose a non-parametric laremission(>500kms−1 )orbroadblueshiftedabsorption description of the Lyα profile. We characterise the spatial (<−500kms−1) from the systemic velocity. In fact, a lower variationintheshapeoftheasymmetricLyαprofilebyfind- equivalentwidthstellarP-Cygniprofilecouldprovideanim- ing the wavelength that corresponds to 50%, 25% and 10% provedfittoourdata,suggestingthatthestellarironabun- of the peak flux redwards of the Lyα peak. MNRAS000,000–000(2017)