High redshift FRII radio sources large-scale X-ray environment

a r X i v :a s t r o -p h /9907397v 1 28 J u l 1999Draft version February 1,2008Preprint typeset using L A T E X style emulateapj v.04/03/99MOLECULAR GAS IN THE POWERFUL RADIO NUCLEUS OF THE ULTRALUMINOUS INFRARED GALAXY PKS 1345+12A.S.Evans 1,2,D.C.Kim 3,4,J.M.Mazzarella 3,N.Z.Scoville 1,andD.B.Sanders 6Draft version February 1,2008ABSTRACT Millimeter CO(1→0)interferometry and high resolution,Hubble Space Telescope (HST)1.1,1.6,and 2.2µm imaging of the radio compact galaxy PKS 1345+12are presented.With an infrared luminosity of ∼2×1012L ⊙,PKS 1345+12is a prime candidate for studying the link between the ultraluminous infrared galaxy phenomenon and radio galaxies.These new observations probe the molecular gas distribution and obscured nuclear regions of PKS 1345+12and provide morphological support for the idea that the radio activity in powerful radio galaxies is triggered by the merger of gas rich galaxies.Two nuclei separated by 2′′(4.0kpc)are observed in the near-infrared;the extended southeastern nucleus has colors consistent with reddened starlight,and the compact northwestern nucleus has extremely red colors indicative of an optical quasar with a warm dust component.Further,the molecular gas,3mm continuum,and radio emission are coincident with the redder nucleus,confirming that the northwestern nucleus is the site of the AGN and that the molecular gas is the likely fuel source.Subject headings:galaxies:ISM—infrared:galaxies—ISM:molecules—radio lines:galaxies—galaxies:active—individual:PKS 1345+121.INTRODUCTION Almost all ultraluminous infrared galaxies [ULIGs:L IR (8−1000µm)≥1×1012L ⊙]have optical/near-infrared morphologies indicative of galaxy-galaxy merg-ers (e.g.Joseph &Wright 1985;Armus,Heckman,&Miley 1987;Sanders et al.1988a;Murphy et al.1996;Kim 1995).Their large quantities of dust and molecu-lar gas (e.g.,Sanders,Scoville,&Soifer 1991;Solomon et al.1997),as well as evidence of abundant young star clusters in many ULIGs (e.g.Surace et al.1998),are strong evidence of recent/ongoing star formation.Molec-ular gas is also a likely source of fuel for active galactic nuclei (AGN)such as quasars and radio galaxies,many of which have high L IR and disturbed optical morphologies (e.g.,Stockton &MacKenty 1983;MacKenty &Stockton 1984;Heckman et al.1986;Smith &Heckman 1989a,b),thus providing a possible connection between mergers and the building of supermassive nuclear black holes.PKS 1345+12(IRAS 13451+1232:L IR =1.7×1012L ⊙)is a prime candidate for the link between the ULIG phe-nomenon and radio galaxies.With a radio luminosity of P 408MHz =2.4×1026W Hz −1,it is the most powerful ra-dio galaxy detected in CO(1→0)to date.It also belongs to a family of “warm”(f 25µm /f 60µm ≥0.2,similar to the colors of Seyfert galaxies:de Grijp,Miley,&Lub 1987)in-frared galaxies believed to be in a transition state betweenthe “cold”(f 25µm /f 60µm <0.2)ULIG phenomenon,when rampant star-formation is occuring and the accretion disk is forming around the nuclear black hole,and the optical quasar phase (Sanders et al 1988a,b).It is observed to have two nuclei with a projected separation of ∼2′′(∼4kpc:Heckman et al.1986;Smith &Heckman 1989a,b;Kim 1995),a very compact radio jet (0.1′′∼200pc),and an extremely high molecular gas mass (4.4×1010M ⊙:7Mirabel,Sanders,&Kazes 1989).The ratios of the nar-row optical emission lines in PKS 1345+12indicate that it contains a Seyfert 2nucleus (Sanders et al.1988b;Veilleux et al.1995).Further,recent near-infrared spectroscopic observations have detected broad (∆v FWHM ∼2600km s −1)Pa αemission,indicating the presence of a quasar nu-cleus which is obscured at optical wavelengths (Veilleux,Sanders,&Kim 1997).The near-infrared spectroscopy of PKS 1345+12illus-trates the importance of probing the nature and distribu-tion of obscured nuclear energy sources in ULIGs at near-infrared wavelengths.Such studies also complement CO interferometry,which provides spatial and kinematic in-formation of the molecular gas reservoirs in these dusty systems.In this Letter ,the superior resolution of the HST 8Near-Infrared Camera and Multi-Object Spectrom-eter (NICMOS:0.1–0.2′′resolution at 1–2µm)and the 1Divisionof Physics,Math,&Astronomy,California Institute of Technology,Pasadena,CA 911252Email Address:ase@3Infrared Processiong &Analysis Center,California Institute of Technology,MS 100-22,Pasadena,CA 911254Present Address:Academia Sinica Institute of Astronomy &Astrophysics P.O.Box 1-87,NanKang Taipei 115,Taiwan,R.O.C.5UCO/Lick Observatory,University of California,Santa Cruz,CA 950646Institute for Astronomy,2680Woodlawn Drive,Honolulu,HI 968227Assuming an L ′CO to M H 2conversion factor of 4M ⊙(K km s−1pc 2)−1;see §3.8The NASA/ESA Hubble Space Telescope is operated by the Space Telescope Science Institute managed by the Association of Universities for Research in Astronomy Inc.under NASA contract NAS5-26555.12CO IN PKS1345+12Owens Valley Millimeter Array9(∼2′′resolution at3mm) are used to show for thefirst time the spatial coincidence of the molecular gas and the radio-loud nucleus of PKS 1345+12.These observations provide further support for the relationship between mergers,molecular gas,and AGN activity.An H0=75km s−1Mpc−1and q0=0.0are as-sumed throughout such that1′′subtends∼2.0kpc at the redshift of the galaxy(z=0.1224).2.OBSERVATIONS AND DATA REDUCTION2.1.NICMOS ObservationsHST NICMOS observations of PKS1345+12were ob-tained as part of a larger program to image infrared-luminous galaxies(Scoville et al.1999;Evans1999a).Ob-servations were obtained in a single orbit on1997Decem-ber5using camera2,which consists of a256×256array with pixel scales of0.0762′′and0.0755′′per pixel in x and y,respectively,providing a∼19.5′′×19.3′′field of view (Thompson et al.1998).Images were obtained using the wide-bandfilters F110W(1.10µm,∆λFWHM∼0.6µm), F160W(1.60µm,∆λFWHM∼0.4µm),and F222M(2.22µm,∆λFWHM∼0.14µm),which provide a resolution (FWHM)of0.11′′,0.16′′,and0.22′′,respectively.The ba-sic observation and data reduction procedures are the same as those described in Scoville et al.(1999);the total inte-gration times perfilter setting for these observations were 480sec(1.1and1.6µm)and600sec(2.2µm).Flux cal-ibration of the images were done using the scaling factors 2.03×10−6,2.19×10−6,and5.49×10−6Jy(ADU/sec)−1 at1.10,1.60,and2.22µm,respectively.The correspond-ing magnitudes were calculated using the zeropoints1775, 1083,and668Jy(Rieke1999).2.2.Interferometric ObservationsAperture synthesis maps of CO(1→0)and2.7mm continuum emission in PKS1345+12were made with the Owens Valley Radio Observatory(OVRO)Millimeter Ar-ray duringfive observing periods from1996September to1997May.The array consists of six10.4m tele-scopes,and the longest observed baseline was242m. Each telescope was configured with120×4MHz dig-ital correlators.During the observations,the nearby quasar HB891413+135(1.5Jy at103GHz;14h13m33.92s +13o:34′17.51′′[B1950.0])was observed every25minutes to monitor phase and gain variations,and3C273and3C 454.3were observed to determine the passband structure. Finally,flux calibration observations of Uranus were ob-tained.The OVRO data were reduced and calibrated using the standard Owens Valley data reduction package MMA (Scoville et al.1992).The data were then exported to the mapping program DIFMAP(Shepherd,Pearson,&Taylor 1995).3.RESULTSThe reduced1.1,1.6,and2.2µm images are shown in Figure1a–c.The galaxy consists of two nuclei with a pro-jected separation of2.0′′(4kpc).Low level surface bright-ness emission envelops both nuclei,with an east-west ex-tent of11′′(22kpc;full width at0.5%the maximumflux density at1.1and1.6µm),and a5.5′′(11kpc)southern extent beyond the southeastern nucleus(hereafter PKS 1345+12SE).The radial surface brightness profile does not constrain the nature of the progenitor galaxies or the type of galaxy they are evolving into;both an r0.25law and an exponential disk give reasonablefits to the profile(see also Scoville et al.1999).PKS1345+12SE is observed to be extended at all three wavelengths with a FWHM of0.15′′(300pc)at1.1µm. The measured1.1′′aperture magnitudes are16.86,15.85, and15.31at1.1,1.6,and2.2µm,respectively,and the derived colors are thus m1.1−1.6=1.01and m1.6−2.2= 0.54.In contrast,the northwestern nucleus(hereafter PKS1345+12NW)is unresolved with a FWHM of0.11′′(220pc)at1.1µm and magnitudes of16.66,15.45,and 13.96at at1.1,1.6,and2.2µm,respectively.The near-infrared colors of PKS1345+12NW are extremely red; m1.1−1.6=1.20and m1.6−2.2=1.49.The red nature of PKS1345+12NW relative to PKS1345+12SE is also ev-ident in the1.1′′-apertureflux density ratios of the two nuclei;f(SE)/f(NW)decreases from a value of0.82at1.1µm to0.29at2.2µm.Both the CO(1→0)emission and underlying2.7mm continuum in PKS1345+12are unresolved.The con-tinuumflux density is0.31Jy and is consistent with a power-law extrapolation of the radioflux density(e.g. Steppe et al.1995).The CO emission(Figure2)has a ∆v FWHM∼600km s−1,aflux density of14±4Jy km s−1, and a CO luminosity of L′CO=8.2×109K km s−1pc2. Thus,the line profile and luminosity are consistent with those derived from NRAO12m Telescope observations of PKS1345+12(i.e.,L′CO=1.1×1010K km s−1pc2: Mirabel et al.1989),and confirms that theflux measured in the single-dish observations(i.e.,81′′FWHM beam)is entirely recovered with OVRO(∼2.2′′synthesized beam). Assuming a standard ratio(α)of CO luminosity to H2 mass of4M⊙(K km s−1pc2)−1,which is similar to the value determined for the bulk of the molecular gas in the disk of the Milky Way(Scoville&Sanders1987;Strong et al.1988),the molecular gas mass is calculated to be 3.3×1010M⊙,or14times the molecular gas mass of the Milky Way.10Finally,using the FWHM beam of the CO map(2.2′′),the molecular gas concentration is calculated to be>2000M⊙pc−2,or>15–1500times that observed in local early-type spiral galaxies(e.g.,Young&Scoville 1991)and comparable to the concentrations observed in a sample of luminous infrared galaxies observed by Scoville et al.(1991)and Bryant&Scoville(1999).4.ASTROMETRY OF THE NEAR-INFRARED IMAGES The interpretation of the data presented in this Letter depends on accurate astrometry of the multiwavelength images.To determine the coordinates of the two nuclei of PKS1345+12,the positions of stars within3.5′of the galaxy werefirst retrieved from the USNO-A1.0database.A plate solution(world coordinate system)was then de-rived for a7′×7′R-band image(Kim1995),which also9The Owens Valley Millimeter Array is a radio telescope facility operated by the California Institute of Technology and is supported by NSF grants AST93–14079and AST96–13717.10Radford,Solomon,&Downes(1991)have used theoretical models to determine thatαranges from2–5M⊙for a reasonable range of temperatures and densities,thus the molecular gas mass of PKS1345+12may actually be as low as1.6×1010M⊙.EVANS ET AL.3shows the two nuclei of PKS1345+12,using the IRAF task PLTSOL.The coordinates of the two nuclei,along with the posi-tions of the CO and radio emission from the galaxy,are listed in Table1.The radio and CO emission appear to be spatially coincident with PKS1345+12NW-the measured near-infrared peak of PKS1345+12NW is displaced0.3′′NE of the radio emission,but is within the uncertainties associated with the positions of the stars used to derive the position of the PKS1345+12nuclei.Likewise,the mea-sured near-infrared peak of PKS1345+12NW is displaced 0.44′′NE of the CO emission and continuum centroids, consistent with the measured OVRO beamsize(see Figure 1d).5.DISCUSSIONThe bulk of the activity in PKS1345+12is associated with PKS1345+12NW.The derived NICMOS colors of the two nuclei provides further support that the AGN resides in PKS1345+12NW.While the colors of PKS 1345+12SE are consistent with starlight reddened by1–5 magnitudes of dust,PKS1345+12NW has colors similar to other warm ULIGs observed with NICMOS-i.e.,sim-ilar to optically-selected quasars with a500–1000K dust component(see also Scoville et al.1999and the sum-mary in Evans1999a).Similar results are derived from near-infrared ground-based observations of PKS1345+12 (e.g.,Surace&Sanders1999),and recent,high-resolution near-infrared spectroscopy(Veilleux&Sanders1999)con-firm that PKS1345+12is the source of the broad-lines detected by Veilleux,Sanders,&Kim(1997).The presence of a large and concentrated reservoir of molecular gas in PKS1345+12NW is consistent with the notion that this gas is a source of fuel for the radio phe-nomenon(e.g.,Mirabel et al.1989).In this scenario, the molecular gas in the western galaxy is driven inward via gravitational instabilities induced by interactions with PKS1345+12SE.Of all of the radio galaxies detected in CO(1→0)to date(Phillips et al.1987;Mirabel et al.1989;Maz-zarella et al.1993;Evans1999b;Evans et al.1999), PKS1345+12is the most molecular gas-rich,and the only one that clearly has two nuclei.Thus,while PKS 1345+12is an advanced merger in terms of its relatively small nuclear separation,the fact that the stellar nuclei are4kpc apart implies that this system is dynamically younger than the other radio galaxies observed.If the assumption is made that the nuclei of PKS1345+12are gravitationally bound,the relative velocity of the merging galaxies is|v|∼<(GM gal/R sep)1/2∼<300km s−1,where M gal=1011M⊙and R sep∼4kpc,and thus the merger has at least an additional∼107years before the nuclei coalesce.From the extent of the radio emission,it is also clear that the jet activity commenced fairly recently.In the2.3 and8.5GHz radio maps shown in Fey,Clegg,&Foma-lont(1996),the radio jet has a maximum extent of0.10′′(∼200pc).Thus,if the jet propagates at a speed of0.1c, it can be no more than7000years old.For comparison, the linear extent of the jets associated with single-nuclei radio galaxies detected in CO are10–200kpc,but their estimated jet ages(∼<107years)are significantly less than the timescale of the merger process(109years).The cre-ation of the radio jets so late in the life of the merger can be understood in terms of merger dynamics-there will be a natural offset in the time at which the merger begins and the AGN activity occurs because of the time it takes for the molecular gas to agglomerate in the nuclear regions of the galaxy.Further,depending on how typical such a de-lay is in radio galaxies,the consumption of molecular gas by extended star formation may be well underway prior to the onset of the radio activity,and may continue for another107years or so.This provides a natural expla-nation of why radio galaxies with older,extended jets are not observed to have large reservoirs of molecular gas(i.e., M(H2)∼<109M⊙:Mazzarella et al.1993;Evans1999b).We thank the staffof the Owens Valley Millimeter array and the NICMOS GTO team for their support both during and after the observations were obtained,and the referee Sylvain Veilleux for many useful suggestions.ASE also thanks M.Shepherd,D.Frayer,and J.Surace for useful discussion and assistance.ASE was supported by NASA grant NAG5-3042.J.M.M.and D.-C.K.were supported by the Jet Propulsion Laboratory,California Institute of Technology,under contract with NASA.REFERENCESArmus,L.,Heckman,T.M.&Miley,G.H.1987,AJ,94,831 Bryant,P.M.&Scoville,N.Z.,ApJ,in pressEvans,A.S.1999a,Ap&SS,in press(astro-ph/9903272)Evans,A.S.1999b,in Highly Redshifted Radio Lines,ed.C.Carilli, S.J. E.Radford,K.Menten,&ngston(San Francisco: PASP),156,74Evans,A.S.,Sanders,D.B.,Mazzarella,J.M.,&Surace,J.A.1999, ApJ,511,730Fey,A.L.,Clegg,A.W.,&Fomalont,E.B.1996,ApJS,105,299 de Grijp,M.H.K.,Miley,G.K.,&Lub,J.1987,A&AS,70,95 Heckman,T.M.et al.1986,ApJ,311,525Joseph,R.D.&Wright,G.S.1985,MNRAS,214,87Kim,D.-C.1995,PhD Thesis,University of Hawaii at Manoa Ma,C.1998,AJ,116,516MacKenty,J.W.&Stockton,A.1984,ApJ,283,64Mazzarella,J.M.,Graham,J.R.,Sanders,D.B.,&Djorgovski,S. 1993,ApJ,409,170Mirabel,I.F.,Sanders,D.B.,&Kaz`e s,I.1989,ApJ,340,l9 Murphy,T.W.,Armus,L.,Matthews,K.,Soifer,B.T.,Mazzarella, J.M.&Neugebauer,G.1996,AJ,111,1025Phillips,T.G.et al.1987,ApJ,322,L33Radford,S.J.E.,Solomon,P.M.,&Downes,D.1991,ApJ,368, L15Rieke,M.1999,in prep.Sanders,D.B.,Scoville,N.Z.,&Soifer,B.T.1991,ApJ,370,158 Sanders,D.B.,Soifer,B.T.,Elias,J.H.,Madore,B.F.,Matthews,K., Neugebauer,G.,&Scoville,N.Z.1988a,ApJ,325,74 Sanders,D.B.,Soifer,B.T.,Elias,J.H.,Neugebauer,G.&Matthews, K.1988b,ApJ,328,L35Scoville,N.Z.et al.1999,AJ,submittedScoville,N.Z.,Carlstrom,J.C.,Chandler,C.J.,Phillips,J.A., Scott,S.L.,Tilanus,R.P.,&Wang,Z.1992,PASP,105,1482 Scoville,N.Z.&Sanders,D.B.1987,in Interstellar Processes,ed.D.Hollenbach&H.Thronson(Dordrecht:Reidel),21 Scoville,N.Z.,Sargent,A.I.,Sanders,D.B.,&Soifer,B.T.1991, ApJ,366,L5Shepherd,M.C.,Pearson,T.J.,&Taylor,G.B.1995,BAAS,27, 903Smith,E.P.&Heckman,T.M.1989a,ApJ,341,658Smith,E.P.&Heckman,T.M.1989b ApJS,69,365Solomon,P.M.,Downes,D.,Radford,S.,&Barrett,J.W.1997, ApJ,478,1444CO IN PKS1345+12Steppe,H.,Jeyakumar,S.,Saikia,D.J.&Salter,C.J.1995,A&AS, 113,409Stockton,A.&MacKenty,J.W.1983,Nature,305,678Strong,A.W.et al.1988,A&A,207,1Surace,J.A.&Sanders,D.B.1999,ApJ,512,162Surace,J. A.,Sanders, D. B.,Vacca,W. D.,Veilleux,S.,& Mazzarella,J.M.1998,ApJ,492,116Thompson,R.I,Rieke,M.,Schneider,G.,Hines,D.C.,Corbin,M. R.1998,ApJ,492,L95Veilleux,S.,Kim,D.-C.,Sanders,D.B.,Mazzarella,J.M.,&Soifer, B.T.1995,ApJS,98,171Veilleux,S.&Sanders,D.B.1999,in preparationVeilleux,S.,Sanders,D.B.,&Kim,D.-C.1997,ApJ,484,92 Young,J.S.&Scoville,N.Z.1991,ARAA,29,581Figure CaptionsFigure1.(a–c)HST NICMOS1.1,1.6,and2.2µm images of PKS1345+12.The images have peak intensities of7.5,11, and29µJy for the1.1,1.6,and2.2µm images,respectively.The speckle pattern surrounding PKS1345+12NW at2.2µm is a PSF artifact.(d)Continuum-subtracted CO(1→0)emission superimposed on the false-color NICMOS image. The NICMOS data are displayed with blue as1.1µm,green as1.6µm,and red as2.2µm.The CO data are plotted as 50%,60%,70%,80%,90%,and99%,where50%corresponds to3.9σrms and100%corresponds to a peakflux of0.0339 Jy/beam.The CO emission is unresolved,with a beam FWHM of2.46×1.95at a position angle of-70.5o.Figure2.Extracted CO(1→0)spectrum of PKS1345+12.The spectrum is smoothed with a∼80km s−1filter and sampling of40km s−1intervals(S rms∼0.005Jy).EVANS ET AL.5Table1Peaks of Optical-to-Radio Emission in PKS1345+12Source RA Dec(B1950.0)1The radio coordinates are taken from Ma(1998).This figure "asefig1.jpg" is available in "jpg" format from: /ps/astro-ph/9907397v1。

a r X i v :a s t r o -p h /0701539v 1 18 J a n 2007Draft version February 5,2008Preprint typeset using L A T E X style emulateapj v.10/09/06THE NATURE OF OPTICAL FEATURES IN THE INNER REGION OF THE 3C 48HOST GALAXY 1Alan Stockton 2,3,Gabriela Canalizo 4,Hai Fu 2,and William Keel 5Draft version February 5,2008ABSTRACTThe well-known quasar 3C 48is the most powerful compact steep-spectrum radio-loud QSO at low redshifts.It also has two unusual optical features within the radius of the radio jet (∼1′′):(1)an anomalous,high-velocity narrow-line component,having several times as much flux as does the narrow-line component coinciding with the broad-line redshift;and (2)a bright continuum peak (3C 48A)∼1′′northeast of the quasar.Both of these optical features have been conjectured to be related to the radio jet.Here we explore these suggestions.We have obtained Gemini North GMOS integral-field-unit (IFU)spectroscopy of the central region around 3C 48.We use the unique features of the IFU data to remove unresolved emission at the position of the quasar.The resolved emission at the wavelength of the high-velocity component is peaked 0.′′25north of the quasar,at virtually the same position angle as the base of the radio jet.These observations appear to confirm that this high-velocity gas is connected with the radio jet.However,most of the emission comes from a region where the jet is still well collimated,rather than from the regions where the radio maps indicate strong interaction with an external medium.We also present the results of HST STIS spectroscopy of 3C 48A.We show that 3C 48A is dominated by stars with a luminosity-weighted age of ∼1.4×108years,substantially older than any reasonable estimate for the age of the radio source.Our IFU data indicate a similar age.Thus,3C 48A almost certainly cannot be attributed to jet-induced star formation.The host galaxy of 3C 48is clearly the result of a merger,and 3C 48A seems much more likely to be the distorted nucleus of the merging partner,in which star formation was induced during the previous close passage.Subject headings:galaxies:high-redshift—galaxies:formation—galaxies:evolution1.INTRODUCTION3C 48(z =0.369)is a remarkable quasar quite aside from the fact that it was the first to be identified (Matthews et al.1961;Matthews &Sandage 1963).As seems to be the case with a number of “prototypes,”it is far from typical.The original identification of 3C 48de-pended in part on the small size of its radio source,and 3C 48and 3C 277.1remain the only examples of com-pact steep-spectrum (CSS)radio sources among power-ful quasars at redshifts <0.5.The host galaxy of 3C 48is unusually large and bright in comparison with those of other low-redshift quasars (Kristian 1973).It is now rec-ognized that these host galaxy properties are largely a result of a major merger (Stockton &MacKenty 1987;Canalizo &Stockton 2000;Scharw¨a chter et al.2004).3C 48was the first quasar for which an extended dis-tribution of ionized gas was observed (Wampler et al.1Based in part on observations made with the NASA/ESA Hub-ble Space Telescope,obtained at the Space Telescope Science In-stitute,which is operated by the Association of Universities for Research in Astronomy,Inc.,under NASA contract NAS 5-26555.These observations are associated with program #GO-09365.Also based in part on observations obtained at the Gemini Observatory,which is operated by the Association of Universities for Research in Astronomy,Inc.,under a cooperative agreement with the NSF on behalf of the Gemini partnership:the National Science Foun-dation (United States),the Particle Physics and Astronomy Re-search Council (United Kingdom),the National Research Coun-cil (Canada),CONICYT (Chile),the Australian Research Council (Australia),CNPq (Brazil)and CONICET (Argentina).2Institute for Astronomy,University of Hawaii 3Cerro Tololo Inter-American Observatory4Department of Physics and Astronomy and Institute of Geo-physics and Planetary Physics,University of California at Riverside 5Department of Physics and Astronomy,University of Alabama1975),and it still is one of the most luminous ex-amples of extended emission among low-redshift QSOs (Stockton &MacKenty 1987).It is one of a handful of previously identified QSOs to have been shown by IRAS also to be ultraluminous IR galaxies.Star for-mation is currently underway at a prodigious rate in the inner part of the host galaxy (Canalizo &Stockton 2000),and there are still massive reserves of molecu-lar gas (Scoville et al.1993;Wink,Guilloteau,&Wilson 1997).All things taken together indicate that we are witnessing a brief but important phase in the history of a quasar,a time when host-galaxy star formation is near its peak,the radio jet is breaking through the dense ma-terial in the inner part of the host galaxy,and UV radia-tion from the central continuum source has just recently become visible along many lines of sight.Wampler et al.(1975)noticed that the [O III ]pro-file in 3C 48was significantly blueshifted relative to the broad H βline.Higher-resolution spectroscopy of the quasar (Chatzichristou,Vanderriest,&Jaffe 1999;Canalizo &Stockton 2000)shows the reason for this ap-parent shift:in addition to the usual nuclear narrow-line region,there is very luminous extended high-velocity gas within about 0.′′5of the nucleus.Chatzichristou et al.(1999)presented evidence from their integral-field-unit spectroscopy that the high-velocity gas has been acceler-ated by the interaction of the radio jet with the surround-ing medium.They discussed the variation of relative line ratios,velocities,and line widths of the high-velocity and systemic-velocity components.Canalizo &Stockton (2000)found that the high-velocity gas was resolved on a slit through the nucleus at position angle −28◦,with an extent of ∼0.′′5and a discernible velocity gradient.But2neither of these studies had sufficient spatial resolution to explore the detailed relation between the radio jet andthe high velocity gas.Boroson&Oke(1982,1984)long ago showed the presence of strong Balmer absorption in the3C48 host galaxy.Canalizo&Stockton(2000)examined the host galaxy of3C48in more detail with the Keck Low-Resolution Imaging Spectrograph(LRIS;Oke et al. 1995)and showed that the spectrum of the stellar pop-ulation had a classic“K+A”character.The fact that they saw Balmer absorption lines from a relatively young stellar population as well as evidence for an older popu-lation(e.g.,from the Mg I b feature at∼5180˚A)meant that they could do a decomposition of the two compo-nents(using stellar synthesis models,such as those of Bruzual&Charlot1993,2003)and estimate,at least in a relative sense,the time that had elapsed since the end of the major starburst phase.With sufficient signal-to-noise,it is possible to carry out such a decomposition at each point in the host galaxy,building up a map of starburst ages along lines of sight through the hostgalaxy(strongly weighted towards the time at which the most recent major episode of star formation ceased). Canalizo&Stockton(2000)were able to determine star-burst ages and stellar radial velocities in32distinct re-gions.But one of the most interesting features in the host galaxy was too close to the quasar to permit suffi-ciently low levels of contamination from quasar light to extract the galaxy spectrum.This is the lumi-nosity peak3C48A∼1′′northeast of the quasar,first noticed by Stockton&Ridgway(1991)and con-firmed by subsequent groundbased and HST observa-tions(Chatzichristou et al.1999;Kirhakos et al.1999; Boyce et al.1999;Canalizo&Stockton2000).One of our ground-based images is shown in Fig.1a,and HST PC images obtained from the archive are shown in Fig.1b and c.Stockton&Ridgway(1991)originally suggested that3C48A was the secondary nucleus from a merging companion;this suggestion was consistent with the pres-ence of the tidal tail to the northwest,as well as ev-idence that we are seeing the3C48host galaxy near the stage offinal merger(Canalizo&Stockton2000). The attempt to model the merger by Scharw¨a chter et al. (2004)shows that there is short period during which the two nuclei would have a separation and position angle roughly similar to that observed for3C48and 3C48A.On the other hand,the morphology of the CSS radio jet(Wilkinson et al.1991)and the presence of the high velocity emission-line components near the quasar (Chatzichristou et al.1999;Canalizo&Stockton2000) both suggest that there is a strong interaction between the radio jet and the ambient gas in the general region of 3C48A.3C48A is dominated by continuum radiation;i.e.,the offset peak is seen on continuum images but not on narrow-band[O III]images(Stockton&Ridgway 1991;Canalizo&Stockton2000).However,there is no detailed correspondence between the morphology of the jet and that of3C48A(the jet proceeds nearly north-wards rather than northeasterly),so the optical emission from3C48A cannot be due to synchrotron radiation. It clearly is unlikely to be nebular thermal continuum from an interaction of the jet and the gas,because we would expect fairly strong emission-line radiation at rea-Fig. 1.—a—Our deep line-free-continuum groundbased image of3C48,after removing the quasar nucleus with the STSDAS pro-cedure cplucy.The cross marks the position of the quasar.3C48A lies about1′′NE of the nucleus(North is up and East to the left).All insets show the central region of the corresponding main panel at lower contrast.b and c—HST PC images from the HST archive.The nucleus has been removed with synthetic Tiny Tim PSFs,so there are large residuals in the inner0.′′5radius and in the diffraction spikes.The scale of these panels is enlarged by a factor of2from that of a.Thefilter bandpasses are indicated.d—Enlargement(by another factor of2)of the inner region of the HST PC F555W image,showing the position and width of the STIS slit.sonable temperatures.Chatzichristou et al.(1999)find colors for3C48A consistent with stars with ages simi-lar to the likely age of the radio jet( 107years).The HST images show that3C48A is curiously distorted;this distortion could result either from the shape of an over-pressured region resulting from shocks from the radio jet or from the tidal stretching of the nucleus of a merging companion.Here we use Gemini Multi-Object Spectrograph (GMOS)integral-field-unit(IFU)spectroscopy to ana-lyze more precisely the position and extent of the high-velocity emission,and spectroscopy with the Space Tele-scope Imaging Spectrograph(STIS)on the Hubble Space Telescope(HST)to attempt to determine the nature of 3C48A.Our principal goals for the STIS observations are (1)to determine whether the continuum radiation asso-ciated with3C48A is indeed due to stars;if so,(2)to de-termine the ages of the stellar populations involved and whether there is an age discontinuity;and(3)to mea-sure velocities of the stellar component:do these show any evidence for distinct kinematical signature,such as might be expected if3C48A is the distorted nucleus of a merging companion.We assume aflat cosmology with Ωm=0.3and H0=70km s−1Mpc−1.2.OBSERVATIONS AND DATA REDUCTION2.1.Gemini North GMOS Integral-Field-UnitSpectroscopyWe obtained two-dimensional spectroscopy of the cen-tral region of3C48with the integral-field mode of GMOS on the Gemini North Telescope(program ID no.GN-3 2003B-C-5).We used the half-field mode with the B600grating,which gave afield of3.′′3×4.′′9and a wavelengthrange of4318–7194˚A.The lenslets have a width of0.′′2,and the CCD pixels correspond to0.′′05.The total expo-sure was9000s,obtained as51800s exposures.Observa-tions of the spectrophotometric standard star G191B2Bprovided theflux calibration.The raw spectral images were processed with the stan-dard Gemini IRAF routines to give theflux calibrateddata cube.The6256spatial planes of the data cube havedimensions of66×98pixels(0.′′05square)and are spacedat intervals of0.46˚A.This data cube still suffers fromthe effects of atmospheric dispersion.Since the quasarwas included within thefield of view,we were able todetermine its centroid for each spatial plane of the datacube.Wefitted a low-order cubic spline curve to the plotof each of the coordinates as a function of wavelength andused these to generate correction offsets for each planeof the data cube.Applying these corrections generateda new data cube for which all of the image planes werecorrectly registered.Thefield of view for which the fullrange of wavelengths is available is,of course,somewhatsmaller than the nominalfield because of the atmosphericdispersion.This corrected data cube was used for anal-ysis when we had to deal with a large wavelength range.However,for our detailed analysis of the[O III]profile,we used the uncorrected data cube,as described in§3,in order to avoid unnecessary interpolation.Thefinalimage resolution at6800˚A(near redshifted[O III])was0.′′55(FWHM).2.2.HST STIS SpectroscopyThe STIS spectroscopy was carried out with the G430Lgrating,covering a nominal region from2900˚A to5700˚A.The slit was centered at a position0.′′97east and0.′′50north of the quasar,at a position angle of44.◦74.Theefficiency decreases quite drastically towards the UV endof the spectrum,and the useful spectral range for ourproject was restricted to3630˚A to5685˚A.Because ofthe low surface brightness of3C48A,we used the0.′′5slitand binned by2pixels in the dispersion direction,obtain-ing a total of21076s of integration over8orbits.Thespectra were dithered over4positions along the slit.Theslit position,superposed on the HST PC F555W image,is shown in Fig.1d.We also obtained626s of integra-tion on the quasar itself with the same configuration,butwithout binning.Each of the8individual spectra of3C48A wasfirstgiven a preliminary cleaning of cosmic rays with la-cos4to eliminate the broad Hβline in the residual.We show the region of Hβand[O III]λλ4959,5007lines after this subtraction for our6off-nuclear B apertures(centered 1.′′1from the quasar)in Fig.3(left panel),as well as the same region for the quasar.In the right panel of Fig.3,we show similarly subtracted spectra for the A apertures(centered0.′′4from the quasar).For the aper-tures at1.′′1radius(left panel in Fig.3,the high-velocity component shows up most strongly in the B0and B60 apertures;it is essentially completely absent in the B180, B240,and B300apertures.There is a clear velocity offset of the high-velocity component between the B0and B60 apertures,as well as evidence for more than two velocity components in the former.The high velocity emission is much stronger in the inner apertures,particularly the A0aperture,where theflux has over10times the high-est value that it has in any of the outer apertures.It is absent from the A120,A180,and A240apertures,but present in the A60and A300apertures.For this inner ring of apertures,the aperture boundaries are adjacent to each other,so there will be unavoidable spill-over be-cause of the0.′′55FWHM seeing disk.Atfirst sight,it may appear that is a problem with the subtraction procedure because of negative[O III] residuals,particularly in the A180spectrum.Note,how-ever,that these negative excursions are only at the wave-length of the high-velocity component.The explanation is almost certainly that the quasar aperture encompasses high-velocity emission slightly extended to the north, which is then excess to the scattered nuclear quasar con-tribution to apertures on the south side.In order to explore the spatial distribution of the high-velocity material in more detail,we analyze the image planes of the data cube as we step across the[O III]line profile.Since we are here concentrating on a very small range of wavelengths around the[O III]λ5007emission complex,we can use the original data cube,uncorrected for atmospheric dispersion,thus avoiding any smearing from subpixel interpolation.The maximum shift due to dispersion was confirmed to be less than0.1pixel(= 0.′′005)over this wavelength range.The general approach is this:we use the continuum on either side of the[O III]λ5007line to precisely define the PSF of the quasar;we bin narrow wavelength intervals to form images from the data cube as we step across the line profile,and we then subtract a scaled PSF from each of these to show the distribution of extended emission close to the quasar for each part of the line profile.In more detail,we used narrow regions of continuum (∼20˚A wide)immediately on either side of the emission line to define the PSF centroid at the line,but we used wider(∼100˚A)continuum regions somewhat farther away to obtain better S/N for the PSF profile itself.This high S/N PSF was then carefully registered(at the sub-pixel level)to the PSF at the emission-line wavelength. We used114.6-˚A-wide slices to span the double line pro-file.For each of these images,we ran the two-component deconvolution task plucy(Hook et al.1994).Briefly, plucy allows separate treatment of point sources(which can be removed)and a variable background compo-nent,which is deconvolved via the Richardson-Lucy algo-rithm,subject to an adjustable entropy constraint.The result was that any unresolved nuclear emission(i.e.,that having a distribution indistinguishable from that of the quasar continuum)was removed,and only extended emission was retained in the residual images.These im-ages are shown in Fig.4,along with the location in the line profile each represents.We also show the radio jet (Feng et al.2005)at the same scale.What is immedi-ately clear is that the high-velocity emission is concen-trated to the north of the quasar,precisely in the initial direction of the radio jet,and mostly within the region where the jet appears relatively unperturbed.There is some slight variation in the position and shape of the distribution of emission across the high-velocity line pro-file,and there is a tail of emission at high velocities to the east,but the main emission peak remains closely aligned with the radio jet.As the wavelength window shifts towards the systemic redshift,the centroid of the extended emission moves more to the west,and more low-surface-brightness extended emission at greater distances becomes visible,including a possible ring of emission at about1′′radius from the quasar.Our decomposition of the profile indicates that the narrow-line emission com-ponent at the systemic velocity is dominated by the high-velocity component at all points in the composite profile. The contribution of the systemic component is greatest (almost50%)near its peak,but it drops offsharply at higher(as well as lower)velocities,because of the much broader velocity width of the high-velocity emission.We also expect the classical narrow-line region to be more closely centered on the nucleus and to be mostly unre-solved.Reconstruction of the line profile from the de-convolved,PSF-subtracted images tends to bear out this expectation,although some residual systemic narrow-line emission still remains,perhaps indicating that this region is slightly resolved(see also the profile in the A0aper-ture in Fig.3).However,the shift in the centroid of the emission must still largely be due to a gradient in the high-velocity component,rather than any extension of the classical narrow-line region.From an aperture that includes essentially all of the high-velocity emission,our decomposition of the[O III] profile gives a value of491±40km s−1for the velocity offset between the two systems in the rest frame of the quasar.The[O III]fluxes in the two components are 2.9×10−13and4.1×10−14erg s−1cm−2,respectively,for the high-velocity and systemic velocity components;i.e., the high-velocity gas has∼7times the[O III]luminosity of that of the classical narrow-line region.We have attempted to estimate contributions to the Hβprofile from these two narrow-line components.We take the best-fitting Gaussian decomposition of the[O III] profile,shift these to the appropriate positions and widths for Hβat the same redshifts,and scale the two components independently to obtain the maximum sub-traction consistent with a smooth and fairly symmetric core for the Hβresidual.This procedure gives a narrow-line Hβ/[O III]ratio of∼0.1for the systemic-velocity component and∼0.08for the high-velocity component. We have also carried out decompositions of the[O II]λ3727doublet and the[Ne V]λ3425line.The[O II] case is complicated by the uncertainty in the density-dependent doublet line ratio for each of the components, but the formal bestfit gives a line ratio at or near the low-density limit and a totalflux∼16%of that of[O III]λ5007for the high-velocity component and a line ratio5Fig.3.—(left panel)Spectra of the region around the Hβand[O III]λλ4959,5007lines for the same apertures as are shown in Fig.2; but,in this case,a scaled quasar spectrum has been subtracted from each of the off-nuclear spectra to remove the scattered quasar light,as judged by the broad Hβemission.(right panel)Similar quasar spectra for apertures at a radius of0.′′4from the quasar.The B60spectrum at the bottom,shown at the same scale as the other spectra,indicates the relative scaling for the two panels.near unity(indicating an electron density of a few hun-dred per cm3)and a totalflux∼25%of that of[O III]λ5007for the component at the systemic velocity.Forthe[Ne V]line,the high-velocity and systemic compo-nents havefluxes around14%and20%of those of theirrespective[O III]profiles.Our ratios for the two veloc-ity systems for Hβ/[O III]and[O II]/[O III]are quiteconsistent with those reported by Chatzichristou et al.(1999).All in all,the ionization parameters of the twocomponents seem roughly similar.From our estimate of the Hβflux in the high-velocitycomponent,and assuming an electron density and tem-perature,the mass of the ionized gas in this componentis given by4πm p f Hβd2LM H=6Fig. 4.—The location of nuclear[O III]extended emission in3C48.The unresolved nuclear emission component has been removed from these images using plucy(Hook et al.1994),where the PSF has been determined from the average continuum profile on either side of the[O III]line.The black dot marks the position of the quasar.The[O III]line profile is shown to the left of each image,and the green bar indicates the wavelength region that has been summed to form the corresponding image.The number in the upper-left corner of the images gives the central velocity shift of the image,in km s−1in the quasar frame,from the systemic redshift of0.36933.The5 GHz Merlin radio contours from Feng et al.(2005)are shown superposed on the image near the peak of the high-velocity system in the lower-right panel.All images use a common intensity mapping and have North up and East to the left.panel),we overplot on the spectrum Bruzual&Charlot (2003)solar-metallicity instantaneous-burst models with ages ranging from100to200Myr,which appear to bracket the luminosity-weighted average age of the stars (we also include a10-Myr model for reference).The best fit,judged by the slope of the continuum shortward of the Balmer limit,is about140Myr.We have explored the possibility of younger stellar ages,combined with internal reddening,as a means of trying tofit the ob-served spectrum.The fundamental difficulty with this approach is the rapid decrease in the amplitude of the Balmer-limit break at ages below100Myr,as shown in the right panel of Fig.5.Because of the short wave-length interval involved(from∼3600˚A to∼4000˚A), the break cannot be significantly enhanced by redden-ing without doing unacceptable violence to the rest of the spectrum,and this conclusion holds for any plausi-ble reddening law.It has previously been noted(see, e.g.,Chatzichristou et al.1999)that our particular line of sight to3C48must have little or no significant red-dening,in spite of the evidence for a major starburst in the host galaxy,and this conclusion seems to hold for 3C48A as well,in spite of suggestions to the contrary by Zuther et al.(2004).The fact that there is some evidence for a Ca II K line stronger than in our models give encouraged us to try combinations of an old stellar population with a younger one.Such a composite population might be expected to allow a younger age for the young population.(Recall, however,that the host galaxy background near3C48A has been subtracted offtofirst order by our reduction procedure,so that any significant older population would likely have to be from3C48A itself).The general results of these experiments are that it is difficult to add any sub-stantial fraction(by light)of old stars without unaccept-able decreases in both the Hδstrength and the Balmer break amplitude,and,even with a fairly large fraction of old stars(20%at rest-frame4050˚A),the younger com-ponent still could not be much younger than108years. While the difficulties of attempting to obtain a clean spectrum of3C48A from ground-based observations are formidable because of scattered quasar light(and par-ticularly the difficulty of correcting for its wavelength variation)as well as contamination from other parts of7Fig. 5.—Both panels show the HST STIS spectrum of 3C 48A (black trace).The left-hand panel shows a 140-Myr-old Bruzual &Charlot (2003)instantaneous-burst spectral synthesis model (red trace),and the green and blue traces show 100-Myr and 200-Myr models,respectively.The right-hand panel shows a range of more extreme models withages from 10-Myr (which would be consistent with an estimate of the upper limit to the age of the stellar population in 3C 48A obtained by Chatzichristou et al.1999)to 2Gyr.Note that the Balmer lines and Balmer break constrain the possibility of trying to fit younger populations with substantial reddening to the data.Fig. 6.—Spectra and models of 3C 48A.The black trace is the spectrum of the 60◦aperture,after subtracting the mean of the other off-nuclear apertures and scaling to match the HST STIS spectrum,which is shown in red.The green trace is the long-wavelength portion of the Bruzual &Charlot (2003)140Myr model,using the same scaling as the same model in the left panel of Fig.5.the host galaxy,we were encouraged by the fact that our 60◦aperture,which covers 3C 48A,shows a substan-tially higher continuum level than the other apertures,even in the uncorrected spectra shown in Fig.2.We tried simply subtracting a straight mean of the other 5off-nuclear apertures,which are all at the same distance from the quasar.If the quasar profile is closely symmet-ric,this procedure should remove the scattered quasar light reasonably well.It will also,to first order,subtract the host-galaxy background,although there is certainly no guarantee that this will be very accurate,because of variations in the host galaxy surface brightness.The re-sult is shown in Fig.6.Naturally,one must be wary of putting too much trust in the details of this residual spectrum because of varia-tions among the 5spectra that have been subtracted from it.Nevertheless,it does agree with the general features of the STIS spectrum remarkably well,and it reinforces the conclusion that the dominant stellar population in 3C 48A must be old enough to show a substantial Balmer break.These results are difficult to reconcile with the sug-gestion that 3C 48A is in some way connected with the radio jet.The fact that 3C 48A is dominated by stel-lar radiation eliminates any possibility that it is due to any form of gaseous emission or emission connected with relativistic electrons,such as synchrotron radiation or Compton upscattering of photons.Dust scattering of starlight from elsewhere in the host galaxy is extremely unlikely,given that the most luminous source around by far is the quasar itself,and no hint of scattered quasar radiation is seen in the STIS spectrum.A remaining possibility is jet-induced star formation,as has been suggested by Chatzichristou et al.(1999).This now also seems unlikely because of the age of the stars.An approximate dynamical time scale for the jet can be estimated from its length and the velocity of theemission-line gas associated with it:τdyn ≈106D kpc v −11000yr.For D =5and v =0.8(see §3),τdyn ≈6×106yr.The apparent stellar age of 3C 48A of 108years is thus much older than the likely age of the radio jet.It is also much older than the likely dynamical time indi-cated by the distance of 3C 48A from the jet if it had been produced by it.If the arc-like structure were to indicate that we are seeing the edge of a “bubble”of star formation caused by the jet,the vector from the point of origin of the bubble to the arc must likely be close to the plane of the sky.Because there is every indica-tion that the host galaxy of 3C 48is quite massive,yet the observed radial velocity range of the stars is quite small (Canalizo &Stockton 2000),stellar motions must。

a r X i v :a s t r o -p h /0603437v 3 16 S e p 2006Mon.Not.R.Astron.Soc.000,1–??(2000)Printed 5February 2008(MN L A T E X style file v1.4)The properties of extragalactic radio sources selected at20GHzElaine M.Sadler 1,Roberto Ricci 2,Ronald D.Ekers 2,J.A.Ekers 2,Paul J.Hancock 1,Carole A.Jackson 2,Michael J.Kesteven 2,Tara Murphy 1,Chris Phillips 2Robert F.Reinfrank 3,Lister Staveley–Smith 2,Ravi Subrahmanyan 2,Mark A.Walker 1,2,4,Warwick E.Wilson 2,Gianfranco De Zotti 5,61Schoolof Physics,University of Sydney,NSW 2006,Australia2AustraliaTelescope National Facility,CSIRO,P.O.Box 76,Epping,NSW 1710,Australia3Department of Physics and Mathematical Physics,University of Adelaide,Adelaide,SA 5005,Australia 4MAW Technology Pty Led,3/22CliffSt.,Manly 2095,Australia 5SISSA/ISAS,Via Beirut 2–4,I-34014Trieste,Italy6INAF,Osservatorio Astronomico di Padova,Vicolo dell’Osservatorio 5,I-35122Padova,Italy5February 2008ABSTRACTWe present some first results on the variability,polarization and general properties of radio sources selected at 20GHz,the highest frequency at which a sensitive radio survey has been carried out over a large area of sky.Sources with flux densities above 100mJy in the ATCA 20GHz Pilot Survey at declination −60◦to −70◦were observed at up to three epochs during 2002–4,including near-simultaneous measurements at 5,8and 18GHz in 2003.Of the 173sources detected,65%are candidate QSOs or BL Lac objects,20%galaxies and 15%faint (b J >22mag)optical objects or blank fields.On a 1–2year timescale,the general level of variability at 20GHz appears to be low.For the 108sources with good–quality measurements in both 2003and 2004,the median variability index at 20GHz was 6.9%and only five sources varied by more than 30%in flux density.Most sources in our sample show low levels of linear polarization (typically 1–5%),with a median fractional polarization of 2.3%at 20GHz.There is a trend for fainter 20GHz sources to have higher fractional polarization.At least 40%of sources selected at 20GHz have strong spectral curvature over the frequency range 1–20GHz.We use a radio ‘two–colour diagram’to characterize the radio spectra of our sample,and confirm that the flux densities of radio sources at 20GHz (which are also the foreground point-source population for CMB anisotropy experiments like WMAP and Planck)cannot be reliably predicted by extrapolating from surveys at lower frequencies.As a result,direct selection at 20GHz appears to be a more efficient way of identifying 90GHz phase calibrators for ALMA than the currently–proposed technique of extrapolation from radio surveys at 1–5GHz.Key words:surveys —cosmic microwave background —galaxies:radio continuum —galaxies:active —radio continuum:general1INTRODUCTIONMost large–area radio imaging surveys have been carried out at frequencies of 1.4GHz or below,where the long–term variability of the radio–source population is generally low.The catalogued flux densities measured by such surveys can therefore continue to be used with a high level of confidence for many years after the survey was made.It is not clear to what extent this is true for radio sur-veys carried out at higher frequencies,where the source pop-ulation becomes increasingly dominated by compact,flat–spectrum sources which may be variable on timescales of a few years.We are currently carrying out a sensitive radio survey of the entire southern sky at 20GHz,using a wide-band ana-c2000RAS2Sadler et al.logue correlator on the Australia Telescope Compact Array (ATCA;see Ricci et al.2004a for an outline of the pilot study for this survey).We have therefore begun an investi-gation of the long–term variability of radio sources selected at20GHz,which will also help us estimate the likely long–term stability of our source catalogue.There is little information to guide us in what to expect. Only a few studies of radio–source variability have been car-ried out at frequencies above5GHz and these have generally targeted sources which were either known to be variable at lower frequencies,or were selected to haveflat or rising radio spectra at frequencies below about5GHz.Such objects may not be typical of the20GHz source population as a whole.The full AT20GHz(AT20G)survey,using the whole 8GHz bandwidth of the analogue correlator and coherently combining all three interferometer baselines,began in late 2004and has a detection limit of40–50mJy at20GHz,i.e. about a factor of two fainter than the sources discussed here. It will eventually cover the entire southern sky from decli-nation0◦to−90◦.Our reasons for carrying out a Pilot Survey in advance of the full AT20G survey were to characterize the high–frequency radio–source population,and to optimize the ob-servational techniques used in the two–step survey process (i.e.fast scans of large areas of sky with a wide-band ana-logue correlator,followed by snapshot imaging of candi-date detections)to maximize the completeness,reliability and uniformity of thefinal AT20G catalogue.Because of the slightly different observational techniques used in2002, 2003and2004,the Pilot Survey data are not as complete or uniform as the AT20G data are intended to be.The Pi-lot Survey nevertheless provides an importantfirst look at the faint radio–source population at20GHz.Since correc-tions for extragalactic foreground confusion will be critical for next–generation CMB surveys,a better knowledge of the properties of high–frequency radio sources(and especially their polarization and variability)is particularly desirable.This paper presents an analysis of the radio–source pop-ulation down to a limitingflux density of about100mJy at 20GHz,based on observations in the declination zone−60◦to−70◦scanned by the AT20GHz Pilot Survey in2002 and2003.Our aim is to provide somefirst answers to the following questions:•How does the radio–source population at20GHz re-late to the‘flat–spectrum’and‘steep–spectrum’populations identified at lower frequencies?•What fraction of radio sources selected at20GHz are variable on timescales of a few years,and how stable in time is a20GHz source catalogue?•What are the polarization properties of radio contin-uum sources selected at20GHz?2OBSER V ATIONS2.1The ATCA wide–band correlatorAn analogue correlator with8GHz bandwidth(Roberts et al.2006),originally developed for the Taiwanese CMB in-strument AMiBA(Lo et al.2001)is currently being used at the Australia Telescope Compact Array(ATCA)to carry out a radio continuum survey of the entire southern sky at2002Sep13–17218 3.430 2003Oct9–16317.6,20.46–730,30,60 Date ATCA Obs.Freq.N antconfig.(GHz)Table2.Log of follow–up ATCA imaging observations of sources detected in the scanning survey at20GHz.N ant shows the num-ber of antennas equipped with12mm receivers for each observing session.The angular resolution of the follow–up images is typi-cally8arcsec at4.8GHz,4arcsec at8.6GHz and15arcsec at 20GHz.20GHz.The wide bandwidth of this correlator,combined with the fast scanning speed of the ATCA,makes it pos-sible to scan large areas of sky at high sensitivity despite the small(2.3arcmin)field of view at20GHz.Since delay tracking cannot be performed with this wide-band analogue correlator,all scanning observations are carried out on the meridian(where the delay for an east–west interferometer is zero).The fast-scanning survey measures approximate posi-tions andflux densities for all candidate sources above the detection threshold of the survey.Follow-up20GHz imaging of these candidate detections is then carried out a few weeks later,using the ATCA in a hybrid configuration with its standard(delay–tracking)digital correlator.These follow–up images allow us to confirm detections,and to measure ac-curate positions andflux densities for the detected sources. Finally,the confirmed sources are also imaged at5and 8GHz to measure their radio spectra,polarisation and an-gular size.2.2Observations in the−60◦to−70◦declinationzoneTables1and2summarize the telescope and correlator con-figurations used for the observations discussed in this paper. There are three main data sets:(i)The ATCA Pilot Survey observations made in2002 and published by Ricci et al.(2004a).These are briefly de-scribed in§2.4below.(ii)Data from a resurvey of the same declination zone at 20GHz in2003,together with near–simultaneous observa-tions at4.8and8.6GHz of the confirmed sources(see§2.5). (iii)20GHz images made in2004of sources detected at 18GHz in2002and/or2003,as part of a program to monitorc 2000RAS,MNRAS000,1–??Extragalactic radio sources at20GHz3the long–term variability of the sources detected in the pilot survey(§2.6).Although our ATCA20GHz pilot survey covered the whole sky between declinations−60◦to−70◦,only sources with Galactic latitude|b|>10◦are discussed in this paper. While the source population at2<|b|<10◦is also dom-inated by extragalactic objects,it is very difficult to make optical identifications of radio sources close to the Galactic plane because of the high density of foreground stars.Since one aim of this study is to examine the optical properties of high–frequency radio sources,we therefore chose to exclude the small number of extragalactic sources which lay within ten degrees of the Galactic plane,or within5.5degrees of the centre of the Large Magellanic Cloud.2.3Theflux density scale of the ATCA at20GHz At centimetre wavelengths,the ATCA primaryflux cali-brator is the radio galaxy PKSB1934–638(Reynolds1994). Planets have traditionally been used to set theflux density scale in the12mm(18–25GHz)band,and the planets Mars and Jupiter were used as primaryflux calibrators during the first two years of operation of the ATCA12mm receivers in 2002–3.However,the use of planets to set theflux density scale has some significant disadvantages(Sault2003):•Their angular size(4–25arcsec for Mars and30–48arcsec for Jupiter)means that they can be resolved out at20GHz on baselines greater than a few hundred metres.•Their(northern)location on the ecliptic means that they are visible above the horizon for a much shorter time than a southern source like PKSB1934–38,and shadowing of northern sources can also be a problem in some compact ATCA configurations.PKSB1934–638was monitored regularly in the12mm band over a six–month period in2003,using Mars as primaryflux calibrator(Sault2003).These observations showed that theflux density of PKSB1934–638remained constant(varying by less than±1–2%at20GHz),making it suitable for use as aflux calibrator at these high frequen-cies.From2004,therefore,PKSB1934–638was used as the primaryflux ATCA calibrator at20GHz,whereas Mars was used in our2002and2003observations.2.42002observations2.4.1Scanning observationsThefirst observations of the declination strip−60◦to−70◦were made by Ricci et al.(2004a).Using a single analogue correlator with3GHz bandwidth and two ATCA antennas on a single30m baseline,they detected123extragalactic (|b|>5◦)sources at18GHz above a limitingflux density of100mJy.The2002observations did not completely cover the whole−60◦to−70◦declination strip because of tech-nical problems which interrupted some of the fast scanning runs.Figure4of Ricci et al.(2004a)shows the2002sky coverage and the missing regions,which are mainly in the RA range5–8h.The declination−60◦to−70◦strip was therefore reobserved at22GHz in2003,and full coverage was then achieved.The region overlapped by the2002and 2003observations gives a useful test of the completeness of the scanning survey technique,as discussed in§4.2.4.2Follow–up imaging andflux–density errorsFollow-up synthesis imaging of the candidate sources de-tected in the2002scans was carried out at18GHz with the ATCA as described by Ricci et al.(2004a).It is important to note that,because the candidate source positions obtained from the wide-band scans in2002were typically accurate to ∼1arcmin,and the primary beam of the ATCA antennas at20GHz is only∼2.3arcmin,about30%of the sources detected in the follow–up images were offset by80arcsec or more from the pointing centre,and so required large (more than a factor of two)corrections to their observed flux densities to correct for the attenuation of the primary beam.These corrections were made by Ricci et al.(2004a), but were not explicitly discussed in their paper.It has sub-sequently become clear that uncertainties in the primary beam correction at very large offsets from thefield centre can sometimes introduce large systematic errors into the ob-servedfluxes.For this reason,we now regard the18GHzflux density measurements listed by Ricci et al.(2004a)as unre-liable for sources observed at more than80arcsec from the imagingfield centre.For follow–up imaging in2003and sub-sequent years,sources more than80arcsec from the imaging field centre were re-observed at the correct position when-ever possible.2.52003observations2.5.1Scanning observationsIn2003,we used three analogue correlators and three ATCA antennas,giving us three independent baselines(of30,30 and60m).The correlators also had a new design with the potential for8GHz operation(Roberts et al.2006).The 2003fast scans were carried out using three ATCA antennas separated by30m on an east–west baseline,and scanned in a trellis pattern at15deg min−1with11–degree scans from declination−59.5◦to−70.5◦,interleaved with2.3arcmin separation and sampled at54ms.The system temperature was continually monitored at 17.6and20.4GHz and periods with high sky noise(i.e.due to clouds or rain)wereflagged out and repeated later.Cal-ibration sources were observed approximately once per day by tracking them through transit(±5min).Due to an unforeseen problem matching the wide-band receiver output to thefibre modulator,there was a15db slope across the bandpass.When we transformed the16lag channels observed into8complex frequency channels,the resulting bandpass was uncalibratable and unphysical.This occurred because we had an analogue correlator and there is no exact Fourier Transform relation between delay and frequency(Harris&Zmuidzinas2001).The actual bandpass was measured by taking the Fourier Transform of the time sequence obtained while tracking a calibrator source through transit.In this case we have a physical delay which changes as the earth rotates and we can get a sensible bandpass.In the end only two channels were usable,giving a total band width of3GHz. It was also impractical to make a phase calibration of thec 2000RAS,MNRAS000,1–??4Sadler etal.Figure parison of the 18GHz flux densities measured in 2002and 2003for sources detected independently in the scanning process.Sources which were detected in 2002but not recovered in 2003are shown as open triangles with a flux density limit of 100mJy for 2003.As discussed in the text,the error bars on the 2002flux density measurements are significantly larger than for 2003.Open squares show sources with offsets of more than 80arcsec from the imaging field centre in the 2002data.three interferometers with this data.As a result the sensi-tivity in 2003was only marginally better than that in 2002,and overlapping scans could not be combined coherently.To extract a candidate source list from the 2003raster scans,the correlator delays were cross-matched with the template delay pattern of a strong calibrator.The correlator coefficient for each time stamp along the scans was recorded,and values from overlapping scans were incoherently com-bined to form images in 12equal-area zenithal projection maps (each two hours wide in right ascension).The source finding algorithm imsad implemented in Miriad was used to extract candidate sources above a 5σthreshold.2.5.2Follow–up imagingA list of 1350candidate sources detected in the scanning survey was observed at 17,19,21and 23GHz as noted in Table 2.As in 2002,the planet Mars was used as the primary flux calibrator.In the 2003follow–up imaging,the data were reduced as the observations progressed,and sources which were more than 80arcsec from the imaging centre were reob-served if possible.This significantly improved the accuracy of the flux density measurements for the 2003images com-pared to those made in 2002,as can be seen in Figure 1.Images of each follow–up field were made at 18and 22GHz using the multi–frequency synthesis (MFS)tech-nique (Conway et al.1990;Sault &Wieringa 1994).Since the signal-to-noise ratio in the 18GHz band was signifi-cantly higher than at 22GHz,we used only the 18GHz data in our subsequent analysis.The median rms noise in thefollow–up images was 1.5mJy/beam at 18GHz,and sources stronger than five times the rms noise level (estimated from the Stokes-V images)were considered to be genuine detec-tions.The 364sources with confirmed detections at 18GHz (including some Galactic plane sources)were imaged at 5and 8.6GHz in November 2003.The total integration time for these follow–up images was 80s (2cuts)at 17–19and 21–23GHz,and 180s (6cuts)at 5and 8GHz.2.62004observationsA sample of 200sources detected at 18GHz in 2002and/or 2003was re-imaged on 22October 2004in a series of tar-geted observations at 19and 21GHz,using the ATCA hy-brid configuration H214.All these imaging observations were centred at the source position measured in 2002/3,so that positional offsets from the imaging field centre were negligi-ble.The 19and 21GHz data were combined to produce a single 20GHz image of each target source.The total inte-gration time at 20GHz was 240s (2cuts),and the median rms noise in the final images was 0.7mJy rms.3DATA REDUCTION AND SOURCE–FITTING3.1Reduction of the follow–up imagesFor the 2003data,deconvolved images of the confirmed sources were made at 5,8and 18GHz and positions and peak flux densities were measured using the Miriad task maxfit ,which is optimum for a point source.We also used the Miriad task imfit to measure the integrated flux density and angular extent of extended sources.Where necessary,the fit-ted flux densities were then corrected for the primary beam attenuation at frequencies between 17and 23GHz based on a polynomial model of the Compact Array antenna pattern.Positional errors were estimated by quadratically adding a systematic term and a noise term:the systematic term was assessed by cross-matching the 18GHz source po-sitions with the Ma et al.(1998)International Coordinate Reference Frame (ICRF)source positions;the noise term is calculated from the synthesized beam size divided by the flux S/N.The median position erors are 1.3arcsec in right ascension and 0.6arcsec in declination.To estimate the flux density errors,we quadratically added the rms noise from V-Stokes images to a multiplica-tive gain error estimated from the scatter between snapshot observations of the strongest sources.The median percent-age gain errors were 2%at 5and 8GHz,and 5%at 18GHz.For the 2004data,the 19and 21GHz visibilities were amplitude and phase calibrated in Miriad.As noted in §2.3,PKSB 1934-638was used as the primary flux calibrator.The calibrated visibilities were combined to form 20GHz images using the MFS technique and peak fluxes were worked out using the Miriad task maxfit .Position and flux errors were determined in the same way as for the 2003data.3.2Polarization measurementsAs all four Stokes parameters were available,linear polar-ization measurements were carried out on the 2003andc2000RAS,MNRAS 000,1–??Extragalactic radio sources at20GHz5parison of18GHzflux densities measured in2002and2003with20GHzflux densities measured in2004.The horizontal dotted line shows the sensitivity limit of the2002and2003surveys.2004data.Q-Stokes,U-Stokes,and polarisedflux P=6Sadler et al.The columns in Table3are as follows:(1)The AT source name,followed by#if the source is resolved or double at20GHz(see§3.4).(2)The radio position(J2000.0)measured from the20GHz images.For resolved doubles,the listed position is the radio centroid.(3)For sources where we were able to make an optical iden-tification on the Digitized Sky Survey,this column lists theb J magnitude from the Supercosmos database.(4)The object type of the optical ID,as classified in Su-percosmos:T=1for a galaxy,T=2for a stellar object(QSO candidate).T=0indicates either a blankfield at the source position or a faint(>22mag)object for which the Supercos-mos star/galaxy separation is unreliable.(5)The18GHzflux density measured in2002,followed by its error.For resolved doubles,we list the integratedflux density over the source.Flux densities in square brackets[ ]are measurements made at offsets of more than80arcsec from the imagingfield centre at18–20GHz,and should beregarded as unreliable because of the large primary–beam correction.Flux densities followed by a colon are measured at offsets of60–80arcsec from thefield centre,but should be reliable.(7)The18GHzflux density measured in2003,and its error.(9)The20GHzflux density measured in2004,and its error.(11)The8.6GHzflux density measured in2003,and its er-ror.(13)The4.8GHzflux density measured in2003,and its er-ror.(15)The integratedflux density at843MHz and its error, from the Sydney University Molonglo Sky Survey(SUMSS) catalogue(Mauch et al.2003).(17)The fractional linear polarization at20GHz measured in2004,and its error.(19)The debiased variability index at20GHz,calculated as described in§5.1.(20)Alternative source name,from the NASA Extragalactic Database.(21)Notes on individual sources,coded as follows:C=listed in the online ATCA calibrator catalogue,E=possible EGRET gamma–ray source(Tornikoski et al. 2002),I=listed as an IRAS galaxy in the online NASA Extra-galactic Database(NED),M=galaxy detected in the near-infrared Two-Micron All-Sky Survey(2MASS),P=in the Parkes quarter-Jy sample(Jackson et al.2002), Q=listed as a QSO in NED,V=VLBI observation with the VSOP satellite(Hirabayashi et al.2000)W=source detected in thefirst–year WMAP data(Bennett et al.2003),X=listed as an X-ray source in NED,=polarization observation by Ricci et al.(2004b).3.4Extended sources at20GHzThe great majority of the sources detected in the20GHz Pilot Survey are unresolved in our follow–up images at5, 8and20GHz.The source–detection algorithm used in the Pilot Survey was optimized for point sources,and therewillGalaxyCandidate QSOMag.limitFigure4.Optical identifications for the20GHz radio sources in Table3.Galaxies and stellar objects(QSO candidates)are shown separately.Only27sources(13%of the sample)are unidentified down to b J<22mag.be some bias against extended sources with angular sizes larger than about30arcsec.For sources larger than1arcmin in size,the totalflux densities listed in Table3may also be underestimated.Only eleven of the173sources in Table3were resolved in our(15arcsec resolution)20GHz images.The overall properties of extended sources in the current sample are as follows:•Three objects,J0103–6439,J2157–6941and J2358–6052/J2359–6057,are very extended double–lobed radio sources which are too large to be imaged with these ATCA snapshots.As a result,the totalflux densities listed in Table 3are lower limits to the correct value.•Another seven sources are resolved in our ATCA im-ages,but still lie within the2.2arcmin primary beam of the ATCA at20GHz.Details of these objects are given in Ap-pendix A.•Five of the extended sources(J0103–6439,J0121-6309,J0257-6112,J0743–6726and J2157–6941)have aflat–spectrum core which dominates theflux density at20GHz. Since the number of extended sources is small,and they ap-pear to be somewhat diverse in nature,we defer any detailed discussion of the extended radio–source population to a later paper.3.5Optical identification of the20GHz sources We examined all the sources in Table3in the SuperCOS-MOS online catalogue and images(Hambly et al.2001).An optical object was accepted as the correct ID for a20GHz radio source if it was brighter than b J=22mag.and lay within2.5arcsec of the radio position.For one source in Ta-ble3(J0715–6829)the optical image was saturated by lightc 2000RAS,MNRAS000,1–??Extragalactic radio sources at20GHz7 Figure5.Relation between SuperCOSMOS b J magnitude andredshift for those objects in our sample which have a publishedredshift.Open circles show galaxies from the20GHz andfilledcircles QSOs.The small crosses show a representative subsampleof2dFGRS radio galaxies selected at1.4GHz(Sadler et al.2002).The highest redshift so far measured for an object in this sampleis for J1940–6907,a QSO at z=3.154.from a nearby11th magnitude star and so no identifica-tion could be attempted.Of the remaining172sources,146(85%)had an optical ID which met the criteria listed above.Monte Carlo tests(based on matching the SuperCOSMOScatalogue with radio positions randomly offset from thosein Table3)imply that at least97%of these IDs are likelyto be genuine associations,rather than a chance alignmentwith a foreground or background object.As can be seen from Figure4,the majority(65%)of ra-dio sources selected at20GHz have stellar IDs on the DSSB images,and are candidate QSOs or BL Lac objects.20%of the radio sample are identified with galaxies and15%are faint objects or blankfields.The overall optical identi-fication rate of85%for radio sources selected at20GHz issignificantly higher than the identification rate for brightradio sources selected at1.4GHz(typically∼30%aboveB∼22mag),but is closer to that found by Bolton et al.(2004)for aflux–limited sample of radio sources selectedat15GHz,as discussed in§8.1.1.Figure5shows the relation between b J magnitude andredshift for the22sources(13%of the objects in Figure4)which currently have a published redshift.A represen-tative sample of nearby radio galaxies(Sadler et al.2002)selected from the2dF Galaxy Redshift Survey(2dFGRS;Colless et al.2001)is shown for comparison.Galaxies de-tected in our20GHz survey appear to span a narrow rangein optical luminosity similar to that seen in nearby radiogalaxies selected at lower frequencies,though we cautionthat the sub-sample of sources with published redshifts isinhomogeneous in nature and may be biased in luminosity<100600%101–12513431%126–1508563%151–2001212100%>200504794%8Sadler etal.Figure6.Examples of radio spectra for each of the four spectral classes identified in the text(Upturn,Rising,Steep and Peak),together with a spectrum classified as Flat(|α|<0.1for both0.84–5GHz and8–20GHz).Where available,a408MHzflux density from the MRC (Large et al.1981)is plotted in addition to the data from Table3.5RADIO SPECTRA OF THE20GHZSOURCES5.1Representative radio spectra at0.8to20GHzFigure6shows some representative radio spectra for sourcesin our sample.It is clear we see a wide variety of spectralshapes,most of which cannot befitted by a single power–law over the frequency range1–20GHz.We can distinguishfour main kinds of spectra:(a)Sources with steep(falling)spectra over the whole range843MHz to20GHz(e.g.J0408–6545in Fig.6).(b)Sources with peaked(GPS)spectra,in which thefluxdensity rises at low frequency and falls at high frequency(e.g.J0201–6638).(c)Sources with inverted(rising)radio spectra over thewhole frequency range(e.g.J0113–6753).(d)Sources with an upturn in their spectrum,where thefluxdensity is falling at lower frequencies,but then turns up andbegins to rise above5–8GHz(e.g.J2213–6330).In addition,a small number of sources haveflat radiospectra in which theflux density is essentially constant overthe entire frequency range observed(e.g.J0220–6330in Fig.6).The radio spectral indexα=(log S1−log S2)。

a r X i v :a s t r o -p h /9910422v 1 22 O c t 1999Mon.Not.R.Astron.Soc.000,000–000(0000)Printed 5February 2008(MN L A T E X style file v1.4)The hyperluminous infrared quasar 3C 318and itsimplications for interpreting sub-mm detections of high–redshift radio galaxiesChris J.Willott 1,2⋆,Steve Rawlings 2,Matt J.Jarvis 21Instituto de Astrof´ısica de Canarias,C/Via Lactea s/n,38200La Laguna,Tenerife,Spain 2Astrophysics,Department of Physics,Keble Road,Oxford,OX13RH,U.K.5February 2008ABSTRACTWe present near-infrared spectroscopy and imaging of the compact steep-spectrum radio source 3C 318which shows it to be a quasar at redshift z =1.574(the z =0.752value previously reported is incorrect).3C 318is an IRAS,ISO and SCUBA source so its new redshift makes it the most intrinsically luminous far-infrared (FIR)source in the 3C catalogue (there is no evidence of strong gravitational lensing effects).Its bolometric luminosity greatly exceeds the 1013L ⊙level above which an object is said to be hyperluminous.Its spectral energy distribution (SED)requires that the quasar heats the dust responsible for the FIR flux,as is believed to be the case in other hyperluminous galaxies,and contributes (at the >10%level)to the heating of the dust responsible for the sub-mm emission.We cannot determine whether a starburst makes an important contribution to the heating of the coolest dust,so evidence for a high star-formation rate is circumstantial being based on the high dust,and hence gas,mass required by its sub-mm detection.We show that the current sub-mm and FIR data available for the highest-redshift radio galaxies are consistent with SEDs similar to that of 3C 318.This indicates that at least some of this population may be detected in the sub-mm because of dust heated by the quasar nucleus,and that interpreting sub-mm detection as evidence for very high (>∼1000M ⊙yr −1)star–formation rates may not always be valid.We show that the 3C318quasar is slightly reddened (A V ≈0.5),the most likely cause of which is SMC-type dust in the host galaxy.If very distant radio galaxies are reddened in a similar way then we show that only slightly greater amounts of dust could obscure the quasars in these sources.We speculate that the low fraction of quasars amongst the very high redshift (z >∼3)objects in low–frequency radio–selected samples is the result of such obscuration.The highest-z objects might be preferentially obscured because like 3C318they are inevitably observed very shortly after the jet-triggering event,or because their host galaxies are richer in dust and gas at earlier cosmic epochs,or because of some combination of these two effects.Key words:galaxies:active –galaxies:nuclei –quasars:general –galaxies:evolution1INTRODUCTIONThe compact radio source 3C 318was identified with a faint “galaxy”on Palomar sky survey plates by V´e ron (1966)and Wyndham (1966).A spectrum of this object was obtained at the Lick Observatory by Spinrad &Smith (1976;here-after SS76),which showed a faint,red continuum and two weak emission features which they identified as MgII λ2799⋆Email:cjw@ll.iac.es and [OII]λ3727at a redshift of z =0.752.Their MgII lineappeared to be broad and they classified the object as an N galaxy.At this time it was the galaxy with the highest known redshift in the 3CR sample (Smith,Smith &Spinrad 1976).Willott et al.(1998)classified 3C 318as a broad-line radio galaxy on the basis of its calculated absolute blue magnitude falling just fainter than the quasar threshold of M B =−23.3C 318is a member of the class of radio sources known as compact steep-spectrum (CSS)sources.These sources arec0000RAS2Willott et al.usually defined as having projected linear sizes≤30kpc and high-frequency radio spectral indicesαrad≥0.5†.There is still some debate over whether these sources are small because they are young,or whether they are confined by anomalously dense environments(e.g.Fanti et al.1995).Re-cent evidence strongly supports the former hypothesis(e.g. Murgia et al.1999),although selection effects(e.g.Blundell &Rawlings1999)may also favour unusually dense envi-ronments for the CSS population.Taylor,Inoue&Tabara (1992)found a high Faraday rotation measure in3C318. This,together with a radio spectrumflattening at low fre-quency(perhaps due to thermal absorption),is suggestive of a dense environment.Another peculiarity of3C318is its high IRAS far-infrared(FIR)flux.Hes,Barthel&Hoekstra(1995)report aflux of F(60µm)=148±24mJy.Very few other3CR sources at z>0.5were detected by IRAS and those that were are all quasars,mostly with strong radio cores,where the FIRflux is likely to be dominated by a non-thermal beamed component(Hoekstra,Barthel&Hes1997).As-suming that the FIR radiation from3C318is not beamed (the probable radio core is very weak,L¨u dke et al.1998),it implies an enormous infrared luminosity,due to dust heated by either the active nucleus or a starburst.Due to the fact that,despite the clearly broadened emis-sion line in the spectrum of SS76,3C318is still considered by many authors as a narrow-line radio galaxy,we obtained near-infrared spectra to search for evidence of other broad emission lines,such as Hα.In Section2we present our ob-servations,clearly showing that3C318actually lies at a much higher redshift than previously believed and is lightly reddened in the observed-frame optical.In Section3we con-sider the infrared and sub-mm properties of3C318given its new redshift.In Section4we consider the implications of our findings for the population of high–redshift radio sources de-tected at sub-mm wavelengths.2OBSER V ATIONS2.1Near-infrared spectroscopy3C318was observed under photometric conditions with the CGS4spectrometer at UKIRT on1999January12.We used the40lines mm−1grating and the long(300mm focal-length)camera.Exposures were made in bothfirst and sec-ond orders to give wavelength coverage of most of the region from1–2µm.Thefirst order spectrum was centred at1.71µm and has a resolution of0.008µm.The second order spectrum was centred at1.20µm and gives a resolution of 0.003µm.A2-pixel(1.2arcsec)slit was centred on the posi-tion of the radio core[RA15h17m50.64s,DEC+20◦26′53.3′′(B1950.0)]oriented at a position angle of90◦.Total exposure times were800s infirst order and1600s in second order.The standard stars HD18881and HD44612were observed to en-ableflux-calibration.In addition an F star was observed at the same airmass as the second order observation to enable †We assume throughout that H◦=50km s−1Mpc−1and q0=0.5,unless stated otherwise.The convention for all spec-tral indices,α,is that Sν∝ν−α,where Sνis theflux density at frequencyν.removal of atmospheric absorption features.Unfortunately, there was insufficient time left to observe this star infirst order,so we use one from earlier in the night at a similar airmass of1.1.The observations were reduced in a standard way. Briefly,the steps involved were;combineflat-fielded ob-servations taken at different positions along the slit,wave-length calibrate using sky OH emission lines,subtract resid-ual background,extract2pixel(1.2arcsec)and6pixel(3.6 arcsec)wide aperture spectra from the2-D images,combine ‘positive’and‘negative’spectra,flux-calibrate using stan-dard stars andfinally removal of atmospheric absorption by dividing by normalised F stars.The resulting2pixel wide aperture spectra are shown in Figure1.These spectra have been boxcar-smoothed by3 pixels(36˚A)infirst order and5pixels(30˚A)in second order. 6emission lines are clearly detected with good signal-to-noise.The emission line data are presented in Table1.Note that these lines clearly indicate a much higher redshift than that of z=0.752deduced by SS76.The redshift adopted here is z=1.574±0.001,determined from a gaussianfit to the narrow[OIII]λ5007line.This redshift is consistent with the broad feature at4906˚A identified by SS76as MgII λ2799,actually being CIII]λ1909.The narrow emission feature at6528˚A in the spectrum of SS76,which they identified as[OII],is either spurious or from another object along the line of sight.It corresponds to a rest-frame wavelength of2536±2˚A with our new redshift determination,which does not correspond to any known,bright emission line.Gelderman&Whittle(1994) looked for[OIII]λ5007emission in3C318at the redshift of SS76.They did not detect any line emission down to a limit of2.10−19Wm−2,but detected the continuum at a comparable strength to SS76.This would imply a ratio of [OIII]/[OII]flux less than one(which although highly un-usual for a powerful radio source does not rule out emission from a foreground object).Furthermore,no Hαat z=0.752 (λobs=1.150µm)is detected in our NIR spectrum to a limit of10−19Wm−2.Therefore we conclude that this feature is probably spurious.Fig.1shows all the characteristics of a quasar;a blue continuum slope(αopt≈0)and broad permitted lines.The Hβline suffers from noisy features in its wings,so no esti-mate of its width can be made.However the Hαline clearly has a broad base,although there is some evidence for a strong narrow central component.Unfortunately the reso-lution of this spectrum is insufficient to fully deblend the various components.A single Lorentzian line profile with F W HM=220˚A(4000km s−1)provides a goodfit to the line(reducedχ2=1.5),but the best-fit single Gaussian profile is significantly worse(reducedχ2=4.0).Note that the broad Hαis blended on the red side with narrow[NII] emission which is just visible as a shoulder on the line profile in Fig.1,but this makes a negligible difference to the line fitting given above.Based on these broad emission lines(Hαand CIII]from SS76)we clearly identify3C318as a quasar. Due to the increase in its redshift,its absolute magnitude is now much more luminous than the division between quasars and BLRGs of M B=−ing the magnitudes reported in Section2.2we determine an absolute magnitude in the rest-frame of M B=−24.9.c 0000RAS,MNRAS000,000–000The hyperluminous infrared quasar3C3183Figure1.The near-infrared spectrum of3C318.Note its typical blue quasar colour and broad Hαemission line.CIII]19094906 1.570— 1.130SS76Hγ434011186 1.577110 1.0±0.320±7Hβ486112524 1.576— 4.2±1.4110±60low snr-uncertain centreHα656316899 1.57522016.0±3.0700±200blended with[NII][OIII]495912755 1.57260 3.5±0.580±20[OIII]500712886 1.574659.5±1.0200±50[SII]6716/673117304 1.57495 1.1±0.460±20blend4Willott etal.Figure2.Central30×30arcsec2of thefield of3C318at K-band.This image has been convolved with a gaussian(σ=1 pixel),since the pixel scale oversamples the seeing by a factor of 10.North is to the top and east to the left.F702Wfilter(Figure3).3C318has a very strong point-source component at0.1arcsec resolution,with possible very faint extended emission(≈1arcsec).Unfortunately the lack of a bright star in thefield to give a reliable point-spread function prevents us from doing a PSF-subtraction analysis.However,we note that theflux,size and appar-ent symmetry of the faint,extended emission are consistent with the expected properties of a radio source host galaxy at z=1.5.The close companion seen in the K-band im-age has a peculiar structure in the HST image,with an apparent bright nucleus and a possible spiral arm to the south(but it is impossible to be certain that the bright unresolved nucleus is not actually due to a cosmic ray im-pact).The optical-NIR photometry of the nearby galaxy (R=21.93±0.10,J=20.14±0.20,K=18.23±0.20)is fitted by a power-law withαopt=1.7.The R magnitude was measured from the HST image in a1arcsec aperture and the J and K magnitudes in1.5arcsec apertures to avoid significantflux from the quasar.Since the power-law spec-trum of this object cannot constrain its redshift,we instead consider its luminosity.If this galaxy is at the same redshift as3C318,then it would have a luminosity of5L∗.If it is an L∗galaxy then it would be at z=0.85.Given the strong nuclear component in this object,we conclude it is possible that it is a very luminous galaxy at the same redshift as the quasar,possibly interacting with the quasar on the evidence of the HST morphology.However,it is also possible that it is at a significantly lower redshift.Future optical spec-troscopy in good seeing is required to determine its redshift and whether it too hosts an AGN.Photometry of the HST image is consistent with that of our ground-based R image.Wefind that the nearby galaxy contributes0.16magnitudes to the IAC-80magnitude at R-band.Making the assumption that this fractional contri-Figure3.HST F702W snapshot image of the samefield as in Fig.2.Note that the image is littered with cosmic rays,since with just one exposure it is impossible to remove them.Inset is the central4×4arcsec2showing3C318(left)and the nearby galaxy(right).In this image,obvious cosmic rays have been edited out and it been smoothed with a kernel of0.02arcsec to bring out the morphology of the nearby galaxy.bution is the same in the B,V,R,I bands(an assumption consistent with the observed optical-NIR photometry),all the IAC-80magnitudes of3C318have been corrected for this contamination.In Fig.4we show the rest-frame UV to optical photom-etry of3C318.The shape of the spectrum is rather unusual with a steep spectral index in the UV,whichflattens at λrest≈5000˚A and then increases again at≈8000˚A.This sharp change between H and K may be due to a photomet-ric error at H-band since this has been calculated assuming a J−H colour from the NIR spectroscopy.Alternatively,it may be a real effect due to strong emission from hot dust dominating at K-band.Note that we have corrected for the strong emission lines in the J and H bands and line contri-butions in other bands are expected to be negligible.Making the assumption that the quasar continuum dominates at all wavelengths,we have attempted tofit these data with a red-dened quasar.Wefind that there is no way that reddening by galactic-type dust canfit the photometry.SMC-type dust (which has a much higher ratio of absorption in the UV to in the optical than that of our galaxy)provides a reasonable fit as shown in Fig.4with A V=0.5.Note that reddening by intervening material at a lower redshift than the quasar (i.e.in our galaxy or in an object along the line-of-sight) would give a much worsefit and is extremely unlikely.Other estimates of quasar reddening may be derived from relative emission line strengths.Wefind that the Balmer decrement(the ratio of thefluxes of Hαand Hβ)is ≈4,a typical value for unreddened quasars.This is consis-tent with the lack of severe reddening seen in the observed-frame NIR.For small values of reddening it is necessary toc 0000RAS,MNRAS000,000–000The hyperluminous infrared quasar3C3185 Figure4.Rest-frame UV to optical photometry of3C318.Thesolid line shows a composite quasar spectrum(Francis et al.1991)reddened by A V=0.5of SMC-type dust at the redshift of thequasar.The dotted line is the same quasar spectrum reddenedby A V=0.8of galactic–type dust which does notfit as well asthe SMC-type dust.Beyond5400˚A the unreddened quasar isapproximated by a power-law withαopt=0.2move to the rest-frame UV lines to get a clearer indication ofBLR reddening(although ionisation and metallicity uncer-tainties introduce scatter here).The ratio of observed HγtoCIII]fluxes is1.This compares with a ratio of2for optically-selected quasars(Francis et al.1991).If the broad lines in3C318undergo the same reddening as the continuum shownin Fig.4,then one would expect the ratio of Hγto CIII]toincrease by a factor of2from its intrinsic value.Thus theobserved ratios are perfectly consistent with this amount ofreddening.Therefore we conclude that there is A V≈0.5ex-tinction towards the nucleus of3C318.Whether this occursin the host galaxy or due to the edge of a dusty torus(e.g.Baker1997)is not clear from these observations.33C318:A HYPERLUMINOUS INFRAREDQUASARGiven the new redshift of3C318presented in this paper,thehigh infraredflux of this source is even more interesting.Itis the most luminous60µm source in the3CR sample withlog10νLν(60µm)=40.1W(assumingα60=1as in Hes etal.1995).Recent ISOPHOT observations detect the quasarat60and90µm,however it was undetected at the longerwavelengths of174and200µm(C.Fanti et al.in prep.).Wenote that the ISO90µmflux is inconsistent with the IRAS100µmflux given by Heckman et al.(1992)–see Figure5.Given the much smaller error of the ISO data,we assumehere that the IRAS100µm measurement(which is3timesgreater than the ISO90µmflux)is erroneous.The60µmfluxes from IRAS and ISO are consistent within the errors.3C318has also recently been reliably detected in the sub-mm at850µm with SCUBA at the JCMT with a marginaldetection(≈2σ)at450µm(E.Archibald;m.).Auseful upper limit at1300µm has been obtained with IRAM(Murgia et al.1999).In Fig.5we plot the rest-frame radio to X-ray spec-tral energy distribution(SED)of3C318using data fromthis paper,the literature and unpublished data reported tous by E.Archibald and C.Fanti.The high-frequency radiospectrum is steep(αrad=1.5)up to the highest frequenciesmeasured(νrest=58GHz).The sub-mm and IR data showa clear excess above the extrapolation of the radio SED.Thus we can be confident that the high FIR luminosity isnot due to beamed synchrotron radiation as in the othermost extreme FIR-luminous3CR quasars(Hoekstra et al.1997).Indeed,the shape of the spectrum from the sub-mmto the mid-infrared(MIR)is characteristic of a thermal dustspectrum,peaking at∼1013Hz.There is now a consensus view concerning the interpre-tation of the thermal dust spectra of quasars at(rest-frame)mid-IR(3-30µm)wavelengths.This emission is thoughtto be heated primarily by the quasar nucleus,and this isstrongly supported by the remarkably tight correlation be-tween the optical and mid-IR luminosities of quasars(e.g.Rowan-Robinson1995;Andreani,Franceschini&Granato1999).The shape of the thermal-IR SEDs of quasars,specif-ically itsflatness followed by a gradual roll-over atλ>∼30µm,favours dust in a torus which extends from r min∼1pc(inner edge controlled by distance at which the temperaturefalls below the dust sublimation value)to r max of at least100pc.The more compact tori postulated by Pier and Kro-lik(1992)are ruled out because these would peak at∼10µm and cut-offsharply at longer wavelengths.It is therefore natural to interpret the ISO detections of3C318in this ing the dereddened optical-UVfluxof3C318as shown in Fig.5and assuming aα=1power-law for the far-UV up to a cut-offat1016Hz,we calculatean integrated optical-UV luminosity of log10L UV=39.9W(q0=0.5).The mid-IR luminosity is a factor of3-6greaterthan this depending upon the form of the SED at rest wave-lengthsλ<20µm,where there are only upper limits fromIRAS.This high ratio of MIR to UV luminosity is reminis-cent of the Cloverleaf quasar(Granato,Danese&Frances-chini1996)and suggests a very high covering factor for thedust.There is,however,controversy concerning the interpre-tation of the far-IR(30-1000µm)dust spectra of quasars.On the one hand it is possible that it is just an extrapola-tion of the quasar-heated mid-IR spectrum(e.g.Granato etal.1996;Andreani et al.1999).These models are attractivebecause of the similar ratios of far-IR to mid-IRfluxes inquasars of widely different luminosities(Haas et al.1998;Wilkes et al.1999).However,they require a large amount ofrelatively cool dust at large radii(up to∼1kpc)from thequasar.On the other hand,Rowan-Robinson(1995)arguesthat dust tori cannot provide sufficient far-IR luminosityand the FIR radiation is due to a massive starburst.Sup-port for this idea comes from the observation that in NGC1068the bulk of the FIR emission is resolved into a star-burst ring approximately3kpc from the nucleus(Telesco etal.1984).Another example is IRAS F10214where gravita-tional lensing magnifies the inner regions and the CO emis-sion is observed to come from a region of intrinsic size<∼400pc(Downes,Solomon&Radford1995).This would appear c 0000RAS,MNRAS000,000–0006Willott etal.Figure5.Rest-frame radio to X-ray SED of3C318.Radio data are from Taylor et al.(1992),van Breugel et al.(1992)and Waldram et al.(1996);sub-mm detections at850and450µm are from E.Archibald(m.)and the1300µm upper limit from Murgia et al.(1999);ISO photometry at200,174,90and60µm come from Fanti et al.(in prep.);100,25and12µm IRAS data are from Heckman et al.(1992)and the60µm point from Hes et al.(1995)(IRAS data shown as triangles);near-infrared and optical data are in this paper –open symbols show thesefluxes after correction for reddening(A V=0.5of SMC-type dust);the X-ray detection with the Einstein satellite is from the LEDAS database.The dashed curve shows a dust spectrum at a temperature of50Kfitted to the sub-mm and ISO 90µm data.The solid curves show an AGN dust model from Granato&Danese(1994)as described in the text with r max=800pc (upper curve is a pole-on view and the lower curve an edge-on view towards the nucleus).A model with a smaller maximum torus radius of r max=160pc(with the same other parameters)is shown as a dotted curve(pole-on view only).to be at odds with the∼2kpc disk required by the mod-els of Granato et al.(1996),subject to uncertainties dueto the lensing.Detailed studies of IRAS F10214(cy,Rawlings&Serjeant1998)conclude that there is substantialstar-formation going on on100pc scales that,provided it isshielded from the nuclear radiationfield,could heat up cooldust on the requisite scales.To investigate whether the observed sub-mm emissionfrom3C318could be due to dust in an extended torusheated by the quasar,we use the models of Granato&Danese(1994)and compare them to the observed SED inFig.5.The models used here have a covering factor of0.8and a constant dust density in the disk with an optical depthat0.3µm of50.A covering factor as large as this is suggestedby the ratio of MIR to UV luminosities discussed previously.There is a strong viewing angle dependence forλ<∼20µm,but this is not important for comparing the ISO and sub-mmfluxes.The extent of the disk is determined by the ratioof the maximum to minimum radii r max/r min.For3C318,r min≈0.8pc from the optical-UV luminosity calculatedpreviously.The ratio of FIR to MIR luminosity increases asr max increases.To simultaneouslyfit the mid-IR and far-IRdata,a large maximum radius of≈1kpc is required.If onereduces r max to≈150pc,then the model accounts for only10%of the sub-mm emission observed at850µm.In thiscase,the bulk of the sub-mm emission would have to be dueto star-formation.Note that the opticalflux of3C318fallsbetween the pole-on and edge-on models plotted in Fig.5,possibly indicating an intermediate orientation(as well asthe high dust covering factor discussed earlier).Deriving the temperature of the coolest dust from ob-servations such as these is a difficult problem,due to uncer-tainty in several parameters such as the dust optical depthand emissivity(e.g.Hughes et al.1993).Here we attempt tomodel the dust spectrum as an isothermal greybody and de-rive limits on the dust temperature allowed by the data.Wefind that isothermal models cannot provide an acceptablefit(reducedχ2<1).This is because the ISO measurements at90and60µm are almost certainly due to emission from hotdust in the torus as explained above,whereas the sub-mmemission is from a cooler dust component.Excluding the60µm data point from thefit leads to models providing accept-ablefits with temperatures in the range45K≤T eff≤100K.However even this determination is probably not validdue to the inclusion of the90µm point and we concludethat the temperature of the cool dust component is unde-termined from these observations.In Fig.5we plot a dustmodel with T eff=50K,which is consistent with the sub-mmand90µm data.Assuming a temperature of50K for thecool dust,we calculate the mass of dust responsible for thec 0000RAS,MNRAS000,000–000The hyperluminous infrared quasar3C3187optically thin sub-mm/FIR luminosity as in Hughes,Dun-lop&Rawlings(1997).Wefind that there is approximately 5×108M⊙of dust,a value similar to those of the z≈4ra-dio galaxies4C41.17and8C1435+635(Hughes&Dunlop 1999).From the SED as shown in Fig.5we have attempted to calculate the total sub-mm/IR luminosity of3C318.Fit-ting a T=100K isothermal dust model through the sub-mm and ISO data gives a reasonable estimate of the FIR luminosity of log10L FIR=13.9L⊙(log10L FIR=14.2L⊙for q0=0).In fact,the total IR luminosity is likely to be greater than twice this value,due to significant emission at wavelengths below25µm which is not well-determined from current observations.Thus3C318meets the criteria of log10L FIR≥13L⊙to be classified as a hyperluminous in-frared galaxy(HyLIG).As noted by Rowan-Robinson(1996) and Hines et al.(1995),thefirst HyLIGs to be discovered all contained obscured AGN and this trend appears to be con-tinuing as more are discovered(e.g.van der Werf et al.1999). Genzel et al.(1998)and Lutz et al.(1998)have found that in low-redshift ultraluminous infrared galaxies(ULIRGS, log10L FIR≥12L⊙),it is starbursts and not AGN which are the predominant source of heating for the mid-IR,but that the fraction of AGN increases with luminosity.Thus it would appear that for the most luminous infrared galaxies, an active nucleus is a requirement and that there is an upper limit to the infrared luminosity due to star-formation.This idea will be developed further in the following section.Note that3C318would not have been selected as a HyLIG by the selection criteria of van der Werf et al.(1999)because, despite the reddening of the UV continuum,its ratio of60µm to B-band luminosity,R≈60is below their adopted limit of100.Thus HyLIGs with a strong quasar component which is not heavily obscured,or heavily scattered,will be missed using such criteria.We should however sound a note of caution here.The three HyLIGs discovered with the highest FIR luminosities of log10L FIR≈15L⊙(APM08279+5255,H1413+117and IRAS F10214+4724)are gravitationally lensed and their FIR luminosities are magnified by an order of magnitude or more(see Ibata et al.1999and references therein).Is it possible that3C318is also being lensed?The HST image shows no evidence of possible lensing galaxies,or distortion of the quasar continuum.The nearby galaxy south-west of the quasar is2arcsec away,which is significantly more than the≈1arcsec Einstein radius of an isolated massive galaxy (e.g.Peacock1999).For an Einstein radius of2arcsec,a group or cluster would be necessary,of which there is no sign in our images.In addition,there are no obvious distor-tions of the compact radio structure over a variety of scales from30milliarcsec to1arcsec(Spencer et al.1991;Taylor et al.1992),suggesting it is not strongly lensed.4THE RELATIONSHIP OF3C318TO THE DISTANT RADIO GALAXY POPULATION 4.1Implications for star-formation ratesBlankfield surveys at850µm with SCUBA are now reveal-ing a population of objects with very high FIR luminosities which are likely to be at high redshifts(Hughes et al.1998;Smail et al.1998;Eales et al.1999).Optical follow-up has shown that at least20%of these sources are associated with AGN(e.g.Barger et al.1999),although the fraction of the integrated observed sub-mm luminosity heated by AGN is extremely uncertain.It could be higher than this value if the other80%of sources contain obscured AGN or it could be lower if even in sources with identified AGN,the far-IR ra-diation is due to starbursts.Recently,models of the sub-mm contribution from the AGN responsible for the X-ray back-ground have shown that it is indeed expected to be at ap-proximately the20%level but there are still large uncertain-ties with respect to the fraction of obscured AGN and their evolution(Almaini,Lawrence&Boyle1999;Gunn&Shanks 1999).The other population of known high-z sources tar-geted(and often detected)with SCUBA are powerful AGN –radio galaxies and radio-quiet quasars.In these cases it is normally assumed that the huge FIR luminosities are due to a massive starburst(SFR>∼1000M⊙yr−1),however it is possible that it is actually the AGN that heats the dust.In this section we investigate the relationship of3C318to the population of high-redshift(z>∼3)radio galaxies detected in the sub-mm and the cause of dust-heating in these objects.In Section3,we showed that models of AGN dust tori can account for both the ISO and SCUBAfluxes of3C318if the dust is distributed in an extended torus/disk out to≈1 kpc from the quasar.If a large amount of dust on these scales is not present then the sub-mmflux must be accounted for by a starburst.In Fig.6we plot the sub-mm–IR SEDs of3C 318,the Cloverleaf quasar H1413+117,IRAS F10214+4724 and two z∼4radio galaxies8C1435+635and4C41.17.The SED of3C318is strikingly similar to that of H1413and F10214,with a similar ratio of FIR to MIRfluxes.A simi-lar correlation between the luminosities of hot and cool dust components for PG quasars was found by Rowan-Robinson (1995).This correlation can be interpreted in two ways.The correlation is a natural consequence if the cool dust is in the outer regions of a dusty torus/disk and the hot dust in the inner regions and they are both heated by the quasar.Al-ternatively,if the cool dust is heated by a massive starburst, this must be connected to the AGN in terms of triggering and fueling.Rowan-Robinson(1995)argued that the larger scatter in the correlation between FIR and opticalfluxes than that of MIR and opticalfluxes showed evidence for the latter view.In Fig.6we plot the same model SED from Granato& Danese(1994)which we used for3C318in Section3.The same model(with r max/r min=1000)is used in all cases with only the normalisation differing.The upper curves show a direct view of the quasar and the lower curves an edge-on view to the nucleus.This modelfits the observations of3C 318,H1413and F10214well.Similarly these SEDs could be fit by two-component models involving an AGN+starburst (e.g.Rowan-Robinson&Crawford1989)The z∼4radio galaxies8C1435+635and4C41.17 have both been detected at850and1300µm.The sub-mm –FIR SEDs of these galaxies have previously beenfitted by T eff≈50K isothermal dust and hence very high star–formation rates of>∼1000M⊙yr−1inferred(e.g.Hughes et al.1997).Fig.6shows that the same model as applied to the quasars above is also consistent with the SEDs of these two radio galaxies.But since the only detections are at mm and sub-mm wavelengths,the data are equally consistentc 0000RAS,MNRAS000,000–000。

a r X i v :a s t r o -p h /0002386v 1 21 F eb 2000Dust Emission from High Redshift QSOsC.L.Carilli 1,2,F.Bertoldi 1,K.M.Menten 1,M.P.Rupen 2,E.Kreysa 1,Xiaohui Fan 3,Michael A.Strauss 3,Donald P.Schneider 4,A.Bertarini 1,M.S.Yun 2,R.Zylka 11Max-Planck-Institut f¨u r Radioastronomie,Auf dem H¨u gel 69,D-53121Bonn,Germany 2National Radio Astronomy Observatory,P.O.Box O,Socorro,NM 87801,USA 3Princeton University Observatory,Peyton Hall,Princeton,NJ 08544,USA 4Dept.of Astronomy,Pennsylvania State University,University Park,PA 16802,USA ccarilli@ Received to appear in the Astrophysical Journal (Letters)ABSTRACTWe present detections of emission at250GHz(1.2mm)from two high redshift QSOs from the Sloan Digital Sky Survey sample using the bolometer array at the IRAM30m telescope.The sources are SDSSp015048.83+004126.2 at z=3.7,and SDSSp J033829.31+002156.3at z=5.0,which is the third highest redshift QSO known,and the highest redshift mm emitting source yet identified.We also present deep radio continuum imaging of these two sources at1.4GHz using the Very Large Array.The combination of cm and mm observations indicate that the250GHz emission is most likely thermal dust emission,with implied dust masses≈108M⊙.We consider possible dust heating mechanisms,including UV emission from the active nucleus(AGN), and a massive starburst concurrent with the AGN,with implied star formation rates>103M⊙year−1.Subject headings:dust:galaxies—radio continuum:galaxies—infrared: galaxies—galaxies:starburst,evolution,active1.IntroductionModern telescopes operating from radio through optical wavelengths are detecting star forming galaxies out to redshifts z>4(Steidel et al.1999,Bunker&van Breugel2000, Adelberger&Steidel2000).These observations are pushing into the‘dark ages,’the epoch when thefirst stars and/or black holes may have formed(Rees1999).Millimeter(mm) and submm observations provide a powerful probe into this era due to the sharp rise of observedflux density with increasing frequency in the modified Rayleigh-Jeans portion of the grey-body spectrum for thermal dust emission from limeter and submm surveys thereby provide a uniquely distance independent sample of objects in the universe for z>0.5(Blain&Longair1993).These surveys have revealed a population of dusty, luminous star forming galaxies at high redshift which may correspond to forming spheroidal galaxies(Smail,Ivison,&Blain1998,Hughes et al.1998,Barger et al.1998,Eales et al. 1998,Bertoldi et al.2000).An interesting sub-sample of dust emitting sources at high redshift are active galaxies,including powerful radio galaxies(Chini&Kr¨u gel1994,Hughes &Dunlop1999,Cimatti et al.1999,Best et al.1999,Papadopoulos et al.1999,Carilli et al.2000),and optically selected QSOs(Omont et al.1996a).In an extensive survey at240GHz,Chini,Kreysa,&Biermann(1989)found that the majority of z<1QSOs show dust emission with dust masses≈few×107M⊙,comparable to normal spiral galaxies.They argue that the dominant dust heating mechanism is radiation from the active nucleus(AGN),based primarily on spectral indices between mm and X-ray wavelengths.On the other hand,Sanders et al.(1989)showed that the majority of radio quiet QSOs in the PG sample show spectral energy distributions from cm to submm wavelengths consistent with star forming galaxies.However,they suggest that this may be coincidental,since concurrent starbursts would require large star formation rates to power the dust emission,while it would require the absorption of only a fraction of theAGN UV luminosity by dust.Omont et al.(1996a)extended the240GHz study of QSOs to high redshift by observing a sample of z>4QSOs from the Automatic Plate Measuring(APM)survey. They found that6of16sources show dust emission at3mJy or greater,with implied FIR luminosities≥1013L⊙,and dust masses≥108M⊙.Follow-up observations of three of these dust-emitting QSOs revealed CO emission as well,with implied molecular gas masses ≈few1010M⊙(Guilloteau et al.1997,1999,Ohta et al.1996,Omont et al.1996b,Carilli, Menten,&Yun1999).Given the large dust and gas masses,Omont et al.(1996a)made the circumstantial argument that the dominant dust heating mechanism may be star formation. Supporting evidence came from deep radio observations at1.4GHz,which showed that the ratio of the radio continuum to submm continuum emission from these sources is consistent with the well established radio-to-far IR correlation for low redshift star forming galaxies (Yun et al.1999).All these data(dust,CO,radio continuum)suggest that the host galaxies of these QSOs are gas-rich,and may be forming stars at a rate≥103M⊙year−1, although it remains unclear to what extent the AGN plays a role in heating the dust and powering the radio emission.We have begun an extensive observational program at cm and mm wavelengths ona sample of high redshift QSOs from the Sloan Digital Sky Survey(SDSS;York et al. 2000).The sample is the result of optical spectroscopy of objects of unusual color from the northern Galactic Cap and the Southern Equatorial Stripe,which has yielded40QSOs with z≥3.6,including the four highest redshift QSOs known(Fan et al.1999,Fan et al. 2000,Schneider et al.2000).This sample presents an ideal opportunity to investigate the properties of the most distant QSOs and their host galaxies.The SDSS sample spans a range of M B=−26.1to−28.8,and a redshift range parative numbers for the APM sample observed by Omont et al.(1996a)are M B=−26.8to−28.5,and4.0≤z≤4.7.Our observations of the SDSS QSO sample include sensitive radio continuum imaging at1.4GHz with the Very Large Array(VLA),and photometry at250GHz using the Max-Planck Millimeter Bolometer array(MAMBO)at the IRAM30m telescope.These observations are a factor three more sensitive than previous studies of high redshift QSOs at these wavelengths(Schmidt et al.1995,Omont et al.1996a),and are adequate to detect emission powered by star formation in the host galaxies of the QSOs at the level seen in low redshift ultra-luminous infrared galaxies(L F IR≥1012L⊙;Sanders and Mirabel1999). We will use these data to measure the correlations between optical,mm,and cm continuum properties,and optical emission line properties,and look for trends as a function of redshift.In this letter we present thefirst two mm detections from this study.The sources are SDSSp J033829.31+002156.3(hereafter SDSS0338+0021),and SDSSpJ015048.83+004126.2(hereafter SDSS0150+0041);these objects’names are their J2000 coordinates.They have z=5.00±0.04with i∗=19.96,and z=3.67±0.02with i∗=18.20 (Fan et al.1999).The absolute blue magnitudes of these quasars are–26.56and–27.75, respectively,assuming H o=50km s−1Mpc−1and q o=0.5.These two sources were taken from the Fall survey sample(Fan et al.1999),which covered a total area of140deg2.Note that J0150+0041is the second most lumininous QSO in the combined Fall and Spring samples of Fan et al.(1999,2000).2.Observations and ResultsObservations were made using MAMBO(Kreysa et al.1999)at the IRAM30m telescope in December1999and February2000.MAMBO is a37element bolometer array sensitive between190and315GHz.The half-power sensitivity range is210to290GHz,and the effective central frequency for a typical dust emitting source at high redshift is250 GHz.The beam for the feed horn of each bolometer is matched to the telescope beam of 10.6′′,and the bolometers are arranged in an hexagonal pattern with a beam separation of 22′′.Observations were made in standard on-offmode,with2Hz chopping of the secondary by50′′in azimuth.The data were reduced using IRAM’s NIC and MOPSI software packages(Zylka1998).Pointing was monitored every hour,and was found to be repeatable to within2′′.The sky opacity was monitored every hour,with zenith opacities between 0.23and0.36.Gain calibration was performed using observations of Mars,Neptune,and Uranus.We estimate a20%uncertainty in absoluteflux density calibration based on these observations.The target sources were centered on the central bolometer in the array (channel1),and the temporally correlated variations of the sky signal(sky-noise)detected in the remaining bolometers was subtracted from all the bolometer signals.The totalon-target plus off-target observing time for SDSS0338+0021was108min,while that for SDSS0150+0041was77min.The source SDSS0338+0021was detected with aflux density of3.7±0.5mJy(Figure 1).At z=5.0,250GHz corresponds to an emitted frequency of1500GHz,or a wavelength of200µm.The source SDSS0150+0041was detected at250GHz with aflux density of 2.0±0.4mJy(Figure1).At z=3.7,250GHz corresponds to an emitted frequency of1180 GHz,or a wavelength of254µm.The quoted errors influx density do not include the20% uncertainty in gain calibration.The sources were not seen to vary dramatically(≤30%) between December1999and February2000.Assuming a dust spectrum of the type seen in the low redshift starburst galaxy Arp220 (corresponding roughly to a modified black body spectrum of index1.5and temperature of 50K),the implied60µm luminosity(Rowan-Robinson et al.1997)for SDSS0338+0021is L60=8.2±1.0×1012L⊙,while that for0150+021is L60=5.0±1.0×1012L⊙.Increasingthe temperature to100K would increase the values of L60by a factor of about3.5,while using the spectrum of M82as a template would increase the values by a factor of1.5.The1.4GHz VLA observations were made in the A(30km)and BnA(mixed30 km and10km)configurations on July8,August14,September30,and October8,1999, using a total bandwidth of100MHz with two orthogonal polarizations.Each source was observed for a total of2hours,with short scans made over a large range in hour angle to improve Fourier spacing coverage.Standard phase and amplitude calibration were applied, as well as self-calibration using background sources in the telescope beam.The absolute flux density scale was set using observations of3C48.Thefinal images were generated using the widefield imaging and deconvolution capabilities of the AIPS task IMAGR.The observed rms noise values on the images were within30%of the expected theoretical noise. The Gaussian restoring CLEAN beam was between1.5′′and2′′FWHM.Neither source was detected at1.4GHz.For SDSS0338+0021the observed valueat the source position was37±24µJy beam−1,and the peak in an8′′box centered on the source position was60µJy beam−1located3.4′′southwest of the source position.At z=5.0,1.4GHz corresponds to an emitted frequency of8.4GHz.For SDSS0150+0041 the observed value at the source position was−3.0±19µJy beam−1,and the peak in an 8′′box centered on the source position was55µJy beam−1located2.8′′north of the source position.At z=3.7,1.4GHz corresponds to an emitted frequency of6.6GHz.We quote maximum and minimum values in an8′′box in the eventuality that the dust emission is not centered on the QSO position(see discussion below).In the analysis below we adopt an upper limit of60µJy beam−1for both sources.Assuming a radio spectral index of–0.8,the3σupper limit to the rest frame spectral luminosity at5GHz of SDSS0338+0021is3.3×1031erg s−1Hz−1,while that for0150+0041 is1.7×1031erg s−1Hz−1.These upper limits place the sources in the radio quiet regime,using the division at1033erg s−1Hz−1at5GHz suggested by Miller et al.(1990).Note that90%of optically selected high redshift QSOs are radio quiet according to this definition (Schmidt et al.1995).3.DiscussionThe fact that the cmflux densities for these sources are at least two orders of magnitude below the mmflux densities,and that the sources do not appear to be highly variable at 250GHz,argues that the mm signal is thermal emission from warm dust.Adopting the relation between L60and dust mass for hyper-luminous infrared galaxies(i.e.galaxies with L60≥1013L⊙),derived by Rowan-Robinson(1999),the dust mass in SDSS0338+0021 is3.5×108M⊙,while that in SDSS0150+0021is2×108M⊙.This gives gas masses of 10×1010M⊙,and6×1010M⊙,respectively,using the dust-to-molecular gas mass ratio of 300suggested by Rowan-Robinson(1999).Of course,such estimates using data at a single frequency are quite uncertain due to the lack of knowledge of the dust temperature and emissivity law,and uncertainties in the gas-to-dust ratio.We are currently searching for CO emission from these two sources to obtain a better understanding of the gas content of these QSOs.Omont et al.(1996a)found that dust emission seems to be correlated with broad absorption line systems(BALs)in the APM QSO sample.The source SDSS0150+0041 has been classified as a‘mini-BAL’by Fan et al.(1999).The source SDSS0338+0021also shows strong and broad associated C IV absorption in high signal-to-noise spectra(Songaila et al.1999).The presence of strong associated absorption is yet another indication of a rich gaseous environment in these systems.The critical question when interpreting thermal dust emission from high redshift QSOsis whether the emission is powered by the AGN or star formation.This question has been considered in great detail for ultra-and hyper-luminous infrared galaxies(Sanders& Mirabel1996,Sanders et al.1989,Genzel et al.1998),and a review of this question can be found in Rowan-Robinson(1999).Dust emitting luminous QSOs,such as SDSS0338+0021 and SDSS0150+0041,comprise a subset of the ultra-and hyper-luminous infrared galaxies. In QSOs,typically less than30%of the bolometric luminosity is emitted in the infrared. Using the blue luminosity bolometric correction factor of16.5derived for the PG QSO sample by Sanders et al.(1989),the FIR luminosity of SDSS0338+0021accounts for only 15%of its bolometric luminosity,and the FIR emission of SDSS0150+0041comprises only 3%of its bolometric luminosity.It is important to keep in mind that the FIR luminosity estimates are based on measurements at a single frequency,and hence have at least a factor two uncertainty(Adelberger&Steidel2000),while the bolometric luminosity estimates require an extrapolation from the rest frame UV measurements into the blue,and hence have comparable uncertainties.Still,it is likely that only a minor fraction of the total AGN luminosity needs to be absorbed and re-emitted by dust to explain the FIR luminosity in both SDSS0338+0021and SDSS0150+0041.An important related point is that the IR emitting region has a minimum size of about0.5kpc for sources such as SDSS0338+0021and SDSS0150+0041,as set by the far IR luminosity and assuming optically thick dust emission with a dust temperature of 50K(Benford et al.1999,Carilli et al.1999).In the Sanders et al.(1989)model for AGN-powered IR emission from QSOs,dust heating on kpc scales is facilitated by assuming that the dust is distributed in a kpc-scale warped disk,thereby allowing UV radiation from the AGN to illuminate the outer regions of the disk.Detailed models by Andreani, Franceschini,and Granato(1999)and Willott et al.(1999)show that the dust emission spectra from3µm to30µm can be explained by such a model.The alternative to dust heating by the AGN is to assume that there is active star formation co-eval with the AGN in these systems.Omont et al.(1996)and Rowan-Robinson (1999)argue that star formation would be a natural,although not required,consequence of the large gas masses in these systems.Imaging of a few high redshift dust emitting AGN shows that in some cases the dust and CO emission come from regions that are separated from the location of the optical AGN by tens of kpc(Omont et al.1999b,Papadopoulos et al.2000,Carilli et al.2000).Such a morphology argues strongly for a dust heating mechanism unrelated to UV emission from the AGN,at least in these stly, Rowan-Robinson(1999)performed a detailed analysis of the spectral energy distributions (SEDs)of hyper-luminous infrared galaxies and concluded that,although≥50%of these systems show evidence for an AGN in the optical spectra,the SEDs between50µm and 1mm are best explained by dust heated by star formation.Yun et al.(2000)have extended this argument to cm wavelengths and reach a similar conclusion.Carilli&Yun(2000)present a model for the expected behavior with redshift of the observed spectral index between cm and submm wavelengths for star forming galaxies, relying on the tight radio-to-far IR correlation found for nearby star forming galaxies (Condon1992).Using the observed mmflux densities of SDSS0150+0041and SDSS 0338+0021,these models predictflux densities at1.4GHz between15and60µJy for both sources.Radio images with a factor three better sensitivity are required to determine if these two sources have cm-to-mm SEDs consistent with low redshift star forming galaxies.If the dust and radio continuum emission is powered by star formation in SDSS 0338+0021and SDSS0150+0041,then the star formation rates in these galaxies are high. Using the relation between L60and total star formation rate given in Rowan-Robinson (1999)for a108year starburst assuming a standard Salpeter IMF,the implied rate for SDSS 0338+0021is2700M⊙year−1,while that for SDSS0150+0041is1800M⊙year−1.Notethat the high redshift QSOs from the APM sample were detected atflux levels between 3mJy and12mJy at240GHz(Omont et al.1996a),hence the required star formation rates may be even larger in some systems.The implication is that we may be witnessing the formation of a large fraction of the stars of the AGN host galaxy on a timescale≤108years.Magnification by gravitational lensing would lower the required luminosities,bringing the sources more in-line with known ultra-luminous infrared galaxies.Two of theAPM QSOs detected at240GHz by Omont et al.(1996a)show possible evidence for gravitational lensing,while two other high redshift dust emitting QSOs,H1413+117and APM08279+5255,are known to be gravitationally lensed(Barvainis et al.1994,Downes et al.1999).However,thus far there is no evidence for multiple imaging on arc-second scales for either SDSS0338+0021or SDSS0150+0041in optical and near IR images(Fan et al.2000).Imaging at sub-arcsecond resolution is required to determine if these sources are gravitationally lensed.An important question concerning the formation of objects in the universe is:which camefirst,black holes or stars(Rocca-Volmerange et al.1993)?If the dust emissionis powered by a starburst in SDSS0338+0021and SDSS0150+0041,then the answerto the above question in some systems may be:both.Co-eval starbursts and AGN at high redshift may not be surprising,since both may occur in violent galaxy mergers as predicted in models of structure formation via hierarchical clustering(Franceschini et al. 1999,Taniguchi,Ikeuchi,&Shioya1999,Blain et al.1999,Kauffmann&Haehnelt2000, Granato et al.2000).To properly address the interesting question of AGN versus starburst dust heating in these high redshift QSOs requires well sampled SEDs from cm to optical wavelengths,and perhaps most importantly,imaging of the mm and cm continuum,and CO emission,with sub-arcsecond resolution.The VLA is a facility of the National Radio Astronomy Observatory(NRAO),whichis operated by Associated Universities,Inc.under a cooperative agreement with the National Science Foundation.This work was based on observations carried out with the IRAM30m telescope.IRAM is supported by INSU/CNRS(France),MPG(Germany) and IGN(Spain).This research made use of the NASA/IPAC Extragalactic Data Base (NED)which is operated by the Jet propulsion Lab,Caltech,under contract with NASA. CC acknowledges support from the Alexander von Humboldt Society.DPS acknowledges support from National Science Foundation Grant AST99-00703.XF and MAS acknowledge support from the Research Corporation,NSF grant AST96-16901,and an Advisory Council Scholarship.REFERENCESAdelberger,K.L&Steidel,C.C.2000,ApJ,in press(astroph0001126)Andreani,P.,Franceschini,A.,and Granato,G.1999,MNRAS,306,161Barger,A.,Cowie,L.L.,Sanders,D.B.,Fulton,E.,Taniguchi,Y.,Sato,Y.,Kawara,K.,& Okuda,H.1998,Nature,394,248Barvainis,R.,Tacconi,L.,Antonucci,R.,Alloin,D.and Coleman,P.1994,Nature,371, 586Benford,D.J.,Cox,P.,Omont,A.,Phillips,T.G.,&McMahon,R.G.1999,ApJ(letters), 518,65Bertoldi,F.et al.2000,in preparationBest,P.et al.1999,MNRAS(letters),301,15Blain,A.W.,&Longair,M.S.1993,MNRAS,264,509Blain,A.W.,Jameson,A.,Smail,I.,Longair,M.S.,Kneib,J.-P.,and Ivison,R.J.1999, MNRAS,309,715Bunker,A.J.and van Breugel,W.J.M.(eds)1999,The Hy-Redshift Universe(PASP,San Francisco)Carilli,C.L.&Yun,M.S.1999,ApJ,513,L13Carilli,C.L.,Menten,K.M.&Yun,M.S.1999,ApJ(lettes),521,L25Carilli,C.L.&Yun,M.S.2000,ApJ,in pressCarilli,C.L.et al.2000,in preparationCarilli,C.L.,Menten,K.M.,Yun,M.S.,Bertoldi,F.,Owen,F.,&Dey,A.1999,in Scientific imperatives at centimeter&meter wavelengths,eds.M.P.van Haarlem&J.M.van der Hulst(NFRA,Dwingeloo)Chini,R.,Kreysa,E.,&Biermann,P.L.1989,A&A,219,87Chini,R.&Kr¨u gel,E.1994,A&A(letters),288,33Cimatti,A.,Freudling,W.,R¨o ttgering,H.,Ivison,R.J.,&Mazzei,P.1998,A&A,329,399 Condon,J.J.1992,ARAA,30,575Downes,D.,Neri,R.,Wiklind,T.,Wilner,D.J.,and Shaver,P.A.1999,ApJ(letters),513, 1Eales,S.et al.1999,ApJ,515,518Fan,X.et al.1999,AJ,118,1Fan,X.et al.2000,AJ,119,1Franceschini,A.,Hasinger,G.,Mayaji,T.,&Malquori,D.1999,MNRAS(letters),310,5 Granato,G.L.et al.2000,MNRAS,in press(astroph9911304)Guilloteau,S.,Omont,A.,McMahon,R.G.,Cox,P.,&Petitjean,P.1997,A&A,328,L1 Guilloteau,S.,Omont,A.,McMahon,R.G.,Cox,P.,&Petitjean,P.1999,A&A,349,363 Hughes,D.H.et al.1998,Nature,394,241Hughes,D.H.&Dunlop,J.S.1999,in Highly Redshifted Radio Lines,eds.C.Carilli,S.Radford,K.Menten,&ngston(PASP,San Francisco),p.103Kreysa,E.et al.1999,SPIE,3357,319Kauffmann,G.&Haehnelt,M.2000,MNRAS,311,576Ohta,K.,Yamada,T.,Nakanishi,K.,Kohno,K.,Akiyama,M.,&Kawabe,R.1996, Nature,382,426Omont,A.,McMahon,R.G.,Cox,P.,Kreysa,E.,Bergeron,J.,Pajot,F.,&Storrie-Lombardi,L.J.1996a,A&A,315,1Omont,A.,Petitjean,P.,Guilloteau,S.,McMahon,R.G.,Solomon,P.M.,&Pecontal,E.1996b,Nature,382,428Papadopoulos,P.P.et al.2000,ApJ,528,626(astroph9908286)Rees,M.J.1999,in After the Dark Ages,eds.S.Holt&E.Smith(AIP,Washington),p.13 Rowan-Robinson,M.1999,MNRAS,in press(astroph9912286)Rowan-Robinson,M.et al.1997,MNRAS,298,490Rocca-Volmerange,B.,Guiderdoni,B.,Dennefeld,M.,&Van,Tran-Thanh1993,First Light in the Universe,(Editions Frontieres,Cedex,France)Sanders,D.B.&Mirabel,I.F.1996,ARAA,34,749Sanders,D.B.,Phinney,E.S.,Neugebauer,G.,Soifer,B.T.,&Matthews,K.1989,ApJ, 347,29Schmidt,M.,van Gorkom,J.H.,Schneider,D.P.,&Gunn,J.E.1995,AJ,109,473 Schneider,D.P.et al.2000,PASP,112,6Songaila,A.,Hu,E.M.,Cowie,L.L.,&McMahon,R.G.1999,ApJ(letters),525,5 Steidel,C.C.,Adelberger,K.L.,Giavalisco,M.,Dickinson,M.,&Pettini,M.1999,ApJ, 519,1Taniguchi,Y.,Ikeuchi,S.,&Shioya,Y.1999,ApJ(letters),514,9Willott,C.J.,Rawlings,S.,and Jarvis,M.J.2000,MNRAS,in press(astroph9910422) York,D.G.et al.2000,AJ,submittedYun,M.S.,Carilli,C.L.,Kawabe,R.,Tutui,Y.,Kohno,K.&Ohta,K.2000,ApJ,in press Zylka,R.1998,MOPSI Users Manual,(IRAM:Grenoble)Fig.1.—The time averaged,opacity corrected,sky-noise subtractedflux densities derived for35of the37bolometers(two channels were disfunctional).The thick line represents the time-averaged signal from the central,on-source bolometer(channel1).The thin solid line is the mean of the off-source bolometers,and the dashed lines show the1σnoise envelope for the off-source bolometers.The upperfigure shows observations of SDSS0338+0021and the lowerfigure shows observations of SDSS0150+0041.。

a r X i v :a s t r o -p h /0401354v 1 18 J a n 2004Draft version February 2,2008Preprint typeset using L A T E X style emulateapj v.04/03/99A LARGE SAMPLE OF SPECTROSCOPIC REDSHIFTS IN THE ACS-GOODS REGIONOF THE HDF-NL.L.Cowie,1,2A.J.Barger,1,3,4,2E.M.Hu,1,2P.Capak 1,2A.Songaila 1,2Submitted to the Astronomical JournalABSTRACTWe report the results of an extensive spectroscopic survey of galaxies in the roughly 160arcmin 2ACS-GOODS region surrounding the HDF-N.We have identified 787galaxies or stars with z ′<24,R <24.5,or B <25lying in the region.The spectra were obtained with either the DEIMOS or LRIS spectrographs on the Keck 10m telescopes.The results are compared with photometric redshift estimates and with redshifts from the literature,as well as with the redshifts of a parallel effort led by a group at Keck.Our sample,when combined with the literature data,provides identifications for 1180sources.We use our results to determine the redshift distributions with magnitude,to analyze the rest-frame color distributions with redshift and spectral type,and to investigate the dependence of the X-ray galaxy properties on the local galaxy density in the redshift interval z =0−1.5.We find the rather surprising result that the galaxy X-ray properties are not strongly dependent on the local galaxy density for galaxies in the same luminosity range.Subject headings:cosmology:observations —early universe —galaxies:distances andredshifts —galaxies:evolution —galaxies:formation1.INTRODUCTIONThe Hubble Deep Field-North (HDF-N;Williams et al.1996;Ferguson,Dickinson,&Williams 2000)remains the single most intensively studied region on the sky.In addition to the defining HST WFPC2ob-servations of the central 5.3arcmin 2area (the HDF-N proper),the field was also the target of the deepest X-ray exposure to date,the 2Ms Chandra Deep Field-North (CDF-N)observation (Alexander et al.2003;Barger et al.2003),which covers about 460arcmin 2.Even wider areas have been covered with ultradeep ground-based optical and near-infrared imaging (Ca-pak et al.2003and references therein)and with very deep 20cm radio observations with the VLA (Richards 2000).Most recently,a roughly 160arcmin 2region,cov-ering the HDF-N proper and lying mostly within the CDF-N,was observed with the ACS camera on HST in four passbands (F435W,F606W,F775W,F850LP)as part of the GOODS HST Treasury program (here-after,ACS-GOODS;Giavalisco et al.2003).This re-gion will also be the target of the deepest SIRTF observations,as part of the GOODS SIRTF Legacyprogram (M.Dickinson,PI).While not quite as deep as the HDF-N proper,the ACS-GOODS survey pro-vides a very deep,high spatial resolution image of a moderately large area of sky.While photometric redshift surveys can cover large areas—avoiding the problems of cosmic variance,which can be significant in areas the size of the ACS-GOODS region—and are useful for determin-ing the evolution of the luminosity function and the universal star formation history (see Capak et al.2004for a photometric redshift analysis of the 600arcmin 2Suprime-Cam field around the HDF-N),spectroscopic redshift surveys are needed to deter-mine the evolution of clustering and the environmen-tal dependences of galaxy properties,such as their X-ray,submillimeter,and radio emission.Spectroscopic redshifts are also required for determining the distri-bution of spectral types,which may be combined with galaxy morphologies from the ACS-GOODS observa-tions,and for determining line luminosities,which provide an alternative way of approaching the star formation history.Finally,spectroscopic redshifts may be used to refine photometric redshift estimates and to test what fraction of photometric redshifts are1Visiting Astronomer,W.M.Keck Observatory,which is jointly operated by the California Institute of Technology,the University of California,and the National Aeronautics and Space Administration.2Institute for Astronomy,University of Hawaii,2680Woodlawn Drive,Honolulu,HI 96822;cowie@;barger@;hu@;capak@ 3Department of Astronomy,University of Wisconsin-Madison,475North Charter Street,Madison,WI 53706.4Department of Physics and Astronomy,University of Hawaii,2505Correa Road,Honolulu,HI 96822.12in error in a givenfield.Some classes of sources,such as broad-line AGNs or sources that have very strong emission lines,frequently have bad photometric red-shifts(e.g.,see Barger et al.2003),and spectroscopic measurements can be used to determine how common such sources are.The present paper,and a parallel paper by Wirth et al.(2004),describe two efforts to spectroscopically identify large,well-defined samples in the ACS-GOODS region.The region around the HDF-N was intensively ob-served spectroscopically by a number of groups us-ing the Low-Resolution Imaging Spectrograph(LRIS; Oke et al.1995)on the Keck10m telescopes.This was a labor-intensive effort,since typically only20−30sources could be observed in each mask,and∼1hr observations were required.However,spectra were obtained for671sources(see the compilation of Co-hen et al.2000for full referencing).Analyses of vari-ous aspects of this sample are given in Cohen(2001) and Cohen(2002).The recent implementation of the Deep Extragalac-tic Imaging Multi-Object Spectrograph(DEIMOS; Faber et al.2003)on the Keck II10m telescope has greatly increased the speed with which such obser-vations may be carried out.DEIMOS can observe around100or more sources per mask,and its higher resolution and quantum efficiency make identifica-tions more straightforward.The long axis of thefield-of-view of the instrument is well matched to the18 arcminute long axis of the ACS-GOODSfield,and most of the ACS-GOODSfield can be covered in two DEIMOS pointings.With DEIMOS it should be pos-sible to identify nearly all of the sources with optical AB magnitudes brighter than24in the red through the entire ACS-GOODS area.Recognizing this,two parallel efforts were begun in Spring2003to map spectroscopically the galaxies in the ACS-GOODS area with DEIMOS.A team led by Keck astronomers(hereafter,the Keck Team Redshift Survey or KTRS)focused purely on an R magnitude selected sample in the ACS-GOODS region,using ob-serving and reduction techniques developed for the DEEP-2project(Davis et al.2003).This sample is described in the paper by Wirth et al.(2004).Our own sample was embedded within observations tar-geted at high-redshift galaxies and X-ray and radio selected galaxies in the wider regions,as described in Hu et al.(2004),Barger et al.(2003),and Cowie et al.(2004).Free regions of the masks that lay within the ACS-GOODS area werefilled with magnitude se-lected samples chosen to have z′<23.5,R<24,or B<24.5.Some fainter sources were also included where no suitable brighter targets could be obtained. Combining our DEIMOS data with our LRIS data from earlier runs,we now have spectroscopic iden-tifications for787sources in thefield with z′<24, R<24.5,or B<bining this sample with data from the literature(Cohen et al.2000,hereafter C00;Dawson et al.2001,hereafter D01;Steidel et al. 2003,hereafter S03)increases the identified sample to1180.In the present paper we describe our spectroscopic sample and compare it with other spectroscopic and photometric redshift samples,focusing on our sam-ple’s completeness.We then analyze our z′<23.5 selected sample in detail to look at how the colors and the star formation and AGN signatures of the galaxies depend on their density environment over the z=0−1.5redshift range.We take H0= 65h65km s−1Mpc−1and use anΩM=13 cosmology.2.OBSERVATIONSSince the ACS-GOODS observations were in progress at the time of our mask designs,the astro-metric and photometric properties of the sample were taken from the ground-based catalog of Capak et al. (2003).In the present paper we use the astrometric positions from the ACS-GOODS catalog,which were kindly provided by Mauro Giavalisco,but we con-tinue to use the photometric magnitudes from Capak et al.We chose the DEIMOS sample from galaxies in the area with z′<23.5,supplemented by galaxies with R<24or with B<24.5.Where no primary target could befitted into a slit mask,we included sources up to half a magnitude fainter.As noted above,many of the sources already had redshifts in the literature or from our own previous LRIS spec-troscopy.In planning our DEIMOS observations,we prioritized unobserved sources over those with exist-ing redshifts.However,where no unobserved target was available,we included sources with known red-shifts,since one of our primary goals was to obtain as uniform a set of spectra for the sources as possible. The new spectroscopic observations were obtained with DEIMOS on Keck II the nights of UT2003Jan-uary29–30,March27,April25–27,and May27. The observations were made with the600lines per mm grating,giving a resolution of3.5˚A and a wave-length coverage of5300˚A.The spectra were centered at an average wavelength of7200˚A,though the exact wavelength range for each spectrum depends on the slit position.Each∼1hr exposure was broken into three subsets,with the sources stepped along the slit by1.5′′in each direction.The spectra were reduced in the same way as previous LRIS spectra(see Cowie et al.1996)by forming a primary sky subtraction from the median of the dithered images,registering and combining the images with a cosmic ray rejec-tionfilter,removing geometric distortion,and thenCowie et al.3applying a profile weighted extraction to obtain the spectrum.The wavelength calibration was made us-ing a polynomialfit to the night sky spectrum. Each spectrum was individually inspected,and a redshift was measured only where a robust identi-fication was possible.The high resolution of the DEIMOS observations greatly simplifies the identi-fication of sources dominated by single strong emis-sion lines since it resolves the doublet structure of the[OII]3727˚A line.As discussed below,most of the magnitude-limited sample sources could be iden-tified.The incompleteness is largest at the faint end, but in a small number of cases(about5−10%,de-pending on the mask)the spectra were problematic in some way(e.g.,the sources were at the edge of a CCD chip or in the heavily distorted and vignetted regions at the edge of the mask);these are present at all mag-nitudes.Sources that were observed but could not be identified(independent of the cause)are retained in our main table(Table1).In Table1we list the905sources that we observed with either DEIMOS or LRIS and have spectra for, from any part of the ACS-GOODS area,and satisfy-ing the magnitude constraints z′<24,R<24.5,or B<25.Table1is ordered by R magnitude and gives the right ascension and declination in decimal degrees (from the ACS-GOODS catalog),the z′,R,and B magnitudes(from the ground-based data of Capak et al.2003),and the measured redshift.The magni-tudes are3′′diameter aperture magnitudes corrected to total magnitudes.Where a source is saturated or otherwise problematic,the magnitudes are omitted. In the redshift column,sources which were observed but not identified are shown as“obs”.Of the sources in the table,787are identified,comprising704galax-ies and83stars.We refer to the sample in Table1 as the“Hawaii”sample.In Table2we combined the identified sources in the “Hawaii”sample with redshifts from the literature to form a more complete table.Including the data from C00,D01,and S03,we have redshifts or stellar iden-tifications for1180sources.Table2gives the right ascension and declination in decimal degrees from the ACS-GOODS catalog,the HK′,z′,I,R,V,B,and U magnitudes from Capak et al.(2003),the redshift and its source(H=Hawaii,C00,D01,or S03),the photometric redshift from Capak et al.(2004),and the odds assigned to it(see§3.2).Sources in the sam-ple that are compact in the ACS-GOODS images and are classified as stars from the empirical star-galaxy separation criterion(V−I)−2.0(I−HK′)=−1.1 given in Barger et al.(1999)are assigned as stars in the photometric redshift list.We refer to the sample in Table2as the“total”sample.There are a very small number of discrepant sources,where different groups have assigned differ-ent redshifts.In only one instance does a Hawaii DEIMOS redshift differ from a Hawaii LRIS redshift. In this case,the DEIMOS redshift also turns out to be inconsistent with other groups’measurements (see Table3),so we adopt the LRIS redshift in our “Hawaii”sample.3.TESTSparison with other catalogsIn order to compare the“Hawaii”sample with the KTRS sample of Wirth et al.(2003)and the litera-ture samples(C00,D01,and S03),wefirst combined the KTRS and literature samples to form the“other”sample.In the small number of cases where there were discrepancies,we somewhat arbitrarily resolved these in favor of KTRS,then C00,then D01,and then S03.There are552overlapping identifications be-tween the“Hawaii”sample and the“other”sample. Of these,512are galaxies and40are stars.All of the star identifications are consistent in the two samples. In Figure1we compare the“Hawaii”galaxy redshifts with the“other”galaxy redshifts andfind that nearly all of these are also consistent.Considering a redshift to be discrepant if it deviates by more than0.01,we find that only seven of the overlap galaxies are dis-crepant,in addition to the one internally inconsistent galaxy discussed in§2;the latter source is not shown in Fig.1because we adopted our LRIS redshift in-stead of our DEIMOS redshift.These eight galaxies give an upper bound of1.4%of sources where the identifications might be problematic in the“Hawaii”sample;they are summarized and briefly described in Table3.HAWAII REDSHIFTOTHERREDSHIFTFig. 1.—“Hawaii”spectroscopic redshift versus“other”spectroscopic redshift(from the KTRS and literature sam-ples).There are512overlapping galaxy identifications be-tween the“Hawaii”sample and the“other”sample,only seven of which have discrepant redshifts.3.2.Photometric redshift comparison4We next compared the “total”sample of redshifts (from the Hawaii observations and the literature)with the photometric redshift catalog of Capak et al.(2004).The photometric redshift estimates are based on ultradeep imaging in seven colors from U to HK ′(Capak et al.2003)and were computed using the Bayesian code of Ben´ıtez (2000).This code as-signs odds that the redshift lies close to the estimated value.We only use those cases where the odds lie above 90%.The imaging data also saturate at bright magnitudes,so we restrict the comparison to sources with R >20.This gives a sample of 1078sources with both spectroscopic and photometric redshifts,which we compare in Figure 2.Fourteen of the sources are broad-line AGNs,where the photometric estimates may be poorer since the templates do not properly model such sources.We have distinguished these sources in Figure 2with larger solid squares.97%of the photometric redshifts agree to within a multi-plicative factor of 1.3with the spectrosopic redshifts.This is consistent with the expectation from the odds assigned to the photometric redshifts.The agreement is independent of magnitude over the R =20−24range.SPECTROSCOPIC REDSHIFTP H O T O M E T R I CFig.2.—Photometric redshift versus spectroscopic redshift for the 1078sources from the “total”sample with 20<R <24having both redshift measurements.The 14broad-line AGNs are denoted by large solid squares.Only 38of the photomet-ric redshifts differ by more than a multiplicative factor of 1.3from the spectroscopic redshifts;eight of these are broad-line AGNs.We may also use the photometric redshifts to test whether there are redshift intervals where the spec-troscopic identifications are substantially incomplete.In Figure 3we compare for three magnitude bins the histogram (dotted line )of the spectroscopic redshifts (from the “total”sample)with the histogram (dashed line )of the photometric redshifts.In the R =20−22and R =22−23bins,the shapes of the two his-tograms are similar,and an appropriate renormal-ization (solid line )produces close agreement between the two distributions.However,in the R =23−24interval,a higher fraction of sources are expected to lie at z >1.5on the basis of the photometric redshifts than are spectroscopically observed at such redshifts.This is the well-known spectroscopic “desert”,where the [OII]3727˚A line has moved out of the optical window,while the ultraviolet signatures have not yet entered.123REDSHIFT1101001000N U M B E R20-22123REDSHIFT1101001000N U M B E R22-23123REDSHIFT1101001000N U M B E R23-24Fig. 3.—Observed spectroscopic (from the “total”sam-ple;dotted )and photometric (dashed )redshift distributions for three magnitude bins:(a)R =20−22,(b)R =22−23,and (c)R =23−24.Solid line shows the spectroscopic dis-tribution renormalized to match the total number of sources.Only region of significant incompleteness is the spectroscopic desert at z =1.5−2in the faintest magnitude (R =23−24)interval.Cowie et al.5R MAGNITUDEI D E N T I F I E D F R A C T I ONZ MAGNITUDEI D E N T I F I E D F R A C T I O NFig. 4.—Fractional completeness in 0.5mag intervals of the “Hawaii”DEIMOS spectroscopic identifications using (a)the R selected sample and (b)the z ′selected sample.pletenessWe next tested the fractional completeness of the “Hawaii”DEIMOS spectroscopic identifications ver-sus magnitude.The results,which should be typical of what can be achieved with 1hour exposures on this instrument,are given in Figure 4.Below R =23,nearly all of the observed sources are identified,ex-cept in a very small fraction of cases where there were problems with the spectra.The completenessbegins to drop offbeyond R =23,and by R =24,it has fallen to about 60%.Some part of this may be a consequence of there being a higher fraction of z >1.5sources at these fainter magnitudes,since such sources in the spectroscopic desert are hard to identify.The photometric redshift estimates shown in Figure 3predict that 13%of the sources in the R =23.75−24.25interval should lie between z =1.5and z =2,while only 5sources are spectroscopically identified in this redshift interval.This suggests that about a third of the incompleteness is caused by red-shift evolution moving the sources into difficult red-shift ranges and that the remaining incompleteness is primarily caused by the lower S/N in the spectra ofthe faintersources.Fig. 5.—Magnitude selected sources in the ACS-GOODS region.To simplify the complex geometry and eliminate sources that are poorly covered in the ACS tiling,we restrict to sources lying within the rectangular area shown.Rectangle is centered at 189.227018and 62.238091,with a long axis of 15.63arcmin and a short axis of 9.28arcmin.The long axis is oriented at 45degrees to the N-S direction and runs from NE to SW.3.4.Fraction ObservedThe ACS-GOODS region has a complex morphol-ogy due to the tiling process (see Fig.5).To eliminate the more poorly covered regions,from now on we re-strict our analysis to the rectangular 145arcmin 2area shown in Figure 5;hereafter we refer to this as the “restricted”field.The coordinates and size of this area are given in the Figure 5caption.The area con-tains 3460sources that satisfy the primary magnitude selection criteria (z ′<23.5,R <24,or B <24.5).Figure 6shows the fractional identifications versus R and z ′magnitude for the “total”sample in the restricted field.Figure 7shows the same versus posi-tion along the long axis of the rectangle.More than a third of the sources have been identified to R =24and z ′=23.5,with the fractions being higher at the brighter magnitudes.The “total”sample is rather strongly concentrated around the HDF-N proper (see Fig.7).In order to show the current completeness of all spectrscopic data on the field we have combined the present data with the KTRS sample of Wirth et al.(2003).In the restricted field this raises the number of identified sources to slightly more than 1800,or more than half the sources to R =24(see Fig.6).The inclusion of the KTRS data also substantially smooths the spatial distribution (see Fig.7).6R MAGNITUDEO B S E R V E D F R A C T I O NFig. 6.—Fractional completeness (open squares )of the R <24“total”sample in the restricted field (as compared to all of the sources in the restricted field,whether observed or not)vs.R magnitude.Total number of sources in each magnitude bin is shown along the top.Solid squares show the fractional completeness when combined with the KTRS sample in the restricted field.POSITION (ARCMINUTES)O B S E R V E D F R A C T I O NFig.7.—Fractional completeness (open squares )of the R <24“total”sample in the restricted field (as compared to all of the sources in the restricted field,whether observed or not)vs.position along the long axis of the ACS-GOODS field.Total number of sources at each position is shown along the top.Solid squares show the fractional completeness when combined with the KTRS sample in the restricted field.4.DISCUSSION4.1.Redshift distributionFigure 8shows the redshift-magnitude diagrams for the R <24and z ′<23.5selected sources in the “to-tal”sample in the restricted field.At magnitudes R <23or z ′<22.5,nearly all of the sources at red-shifts higher than z =1.5are broad-line AGNs.At fainter magnitudes,the more normal galaxy popula-tion stretches beyond z =1.5into the redshift desert (see §3.3),and the truncation of the redshift distri-bution at z =1.5is clearly seen.A small number of galaxies have measured redshifts above z =2.A substantial fraction of the z >2sources have X-ray counterparts (large open squares )and may have sub-stantial AGN contributions to their light,but many do not (see Barger et al.2003).R MAGNITUDER E D S H I F TZ MAGNITUDER E D S H I F TFig.8.—Spectroscopic redshift versus magnitude for the (a)R <24and (b)z ′<23.5selected sources in the “total”sam-ple in the restricted field.Broad-line sources are denoted by large solid squares,and X-ray sources by large open squares.Stars are denoted by asterisks.Cowie et al.7REDSHIFTD I S T A N CE (M P c )REDSHIFTD I S T A N CE (M P c )Fig.9.—Pie diagrams for the spectroscopically identified sources in the “total”sample in the restricted field.(a)Slice parallel to the long axis of the ACS-GOODS region.(b)Slice parallel to the short axis.Angular distances are plotted versus redshift.Broad-line sources are denoted by large solid squares,and X-ray sources by large open squares.All of the broad-line AGNs are also X-ray sources.The well-known “velocity sheets”(C00)that are a consequence of the large scale clustering can be seen in the redshift-magnitude diagrams.These are more clearly visible in the pie diagrams of Figure 9,which show the distributions (a)parallel and (b)perpendic-ular to the long axis of the ACS-GOODS rectangle.The most prominent of the sheets lie at z =0.85and z =1.01.The sheets extend most of the way along the long axis (the drop-offat the ends is a consequence of the higher incompleteness of the ob-servations at these positions—see Figure 7),but the z =1.01sheet is primarily found on the eastern side of the rectangle.The X-ray source positions (large open squares )and broad-line AGN positions (large solid squares )are shown on the pie diagrams.As has been previously noted,these also show velocity sheet structures (Barger et al.2002,2003;Gilli et al.2003).We shall return to the question of whether they are more concentrated into velocity sheets than the nor-mal galaxies in §4.3.4.2.Galaxy colors versus redshiftFigure 10shows R −z ′color versus redshift for the z ′<23.5“total”sample in the restricted field.The upper bound of the distribution matches extremely well to the expected color of an elliptical galaxy.Ap-parently,very few galaxies are sufficiently reddened to lie above this envelope.There are only six such sources,three of which are galaxy pairs and radio sources (solid squares ),suggesting they may be dusty mergers.It is unclear why the other three sources lie at these colors.REDSHIFTR -z ’Fig.10.—R −z ′color versus redshift for the z ′=23.5“total”sample in the restricted field.Only sources with spec-troscopic redshifts are shown.Upper curve shows the color of the Coleman,Wu,&Weedman (1980)elliptical galaxy versus redshift,which almost perfectly maps the upper envelope of the distribution over the z =0−1redshift range.Only 6sources lie above the envelope,three of which are galaxy pairs and radio sources (solid squares ).Because of the extensive color information avail-able (Table 2),we may analyze the galaxy colors in a more model-independent fashion by interpolating to form the rest-frame colors at each redshift.In Fig-ure 11we show the rest-frame 2800−8000˚A color versus redshift.The upper bound matches the color of an elliptical galaxy and is nearly invariant from z =0−1.Beyond this redshift,the z ′selection be-comes biased against elliptical galaxies as the 4000˚Abreak enters the z ′band.The bottom of the dis-tribution corresponds to irregulars,star formers,andbroad-line AGNs (the latter are denoted by large solid squares).The type selection versus absolute magnitude is il-lustrated in Figure 12,where we show the rest-frame absolute (a)8000˚A and (b)2800˚A magnitudes ver-sus redshift.For Figure 12a we use z ′<23.5,but we implicitly have a second selection since we need to interpolate between the HK ′magnitudes and the z ′magnitudes to determine the 8000˚A magnitudes at high redshifts.To formalize this,we also limit the sample to HK ′<22.5.This HK ′selection is illustrated with a dashed curve computed for a flat-spectrum source.Above this curve,all sources should8be included,if we assume that all sources are redder than flat f ν.However,the z ′selection means that we will omit redder sources at the higher redshifts and fainter magnitudes.We illustrate this for the case of an elliptical galaxy (solid curve )and a spi-ral galaxy (dashed curve ).At z =1.2we will omit ellipticals fainter than about −22in the rest-frame 8000˚A band.For Figure 12b we use R <24.Since this is a red-selected sample,we again illustrate the selection threshold for a flat f νspectrum.All sources above this line should be included.REDSHIFTR E S T 2800-8000Fig.11.—Rest-frame 2800−8000˚Acolor versus redshift for the z ′<23.5sample.Only sources with spectroscopic red-shifts are rge solid squares denote broad-line AGNs,and large open squares denote Chandra sources.In Figures 12a and 12b,we also show the po-sitions of the sources with broad emission lines in their spectra (large solid squares )and the positions of the sources with high X-ray luminosities (L X >1042ergs s −1;large open squares ).Most of the high ultraviolet luminosity sources fall in these categories and may have substantial AGN contributions.Pre-vious analyses have not been able to distinguish be-tween these sources and galaxies where the ultravi-olet light is dominated by star formation.Further discussion of the evolution of the light density and luminosity functions with redshift may be found in Capak et al.(2004).REDSHIFTR E S T 8000A A B S M A GREDSHIFTR E S T 2800A A B S M A GFig.12.—Absolute rest-frame magnitude versus red-shift at (a)8000˚A and (b)2800˚A .In each case,sources with spectroscopic redshifts are denoted by small solid sym-bols,and sources with photometric redshifts are denoted by plus signs.Sources with broad emission lines are denoted by large solid squares,and sources with X-ray luminosities L X >1042ergs s −1by large open squares.Dotted curve in (a)shows the HK ′<22.5selection for sources with a flat f νspectrum,while solid and dashed curves show the z ′<23.5selection for an elliptical galaxy and a spiral galaxy,respec-tively.Solid curve in (b)shows the R <24selection for a flat f νspectrum galaxy.The horizontal dashed lines show a reference absolute magnitude of -23for 8000˚A and -21.5for2800˚A which roughly corresponds to the maximum observed values at high redshift for objects without AGN signatures.4.3.X-ray properties versus environmentA large fraction of the most luminous galaxies har-bor AGNs.For all galaxies with M 8000˚A<−21in the z =0−1range,based on either spectro-scopic or photometric redshifts,the fraction with L X >1042ergs s −1is 6%,while for M 8000˚A<−22,it is 15%,in agreement with previous estimates (e.g.,Barger et al.2001).Conversely,nearly all of the lumi-nous X-ray sources lie in optically luminous galaxies.Of the sources with L X >1042ergs s −1,82%lie in galaxies with M 8000˚A<−21in the z =0−1.2redshift range,where lower optical luminosity sources would still be seen.Neither result shows any strong redshift dependence.。

a r X i v :a s t r o -p h /0105560v 1 31 M a y 2001The ELAIS Deep X-ray Survey C.J.Willott Astrophysics,University of Oxford,Keble Road,Oxford OX13RH,U.K.O.Almaini,J.Manners,O.Johnson,wrence,J.S.Dunlop,R.G.Mann Institute for Astronomy,University of Edinburgh,Blackford Hill,Edinburgh EH93HJ,U.K.I.Perez-Fournon,E.Gonzalez-Solares,F.Cabrera-Guerra Instituto de Astrof´ısica de Canarias,C/Via Lactea s/n,38200La Laguna,Tenerife,Spain S.Serjeant Unit for Space Sciences and Astrophysics,The University of Kent,Canterbury,Kent CT27NR,UK S.J.Oliver Astronomy Centre,CPES,University of Sussex,Falmer,Brighton,Sussex BN19QJ,UK M.Rowan-Robinson Astrophysics Group,Blackett Laboratory,Imperial College of Science Technology &Medicine (ICSTM),Prince Consort Rd,London SW72BZ,UK We present initial follow-up results of the ELAIS Deep X-ray Survey which is being undertaken with the Chandra and XMM-Newton Observatories.235X-ray sources are detected in our two 75ks ACIS-I observations in the well-studied ELAIS N1and N2areas.90%of the X-ray sources are identified optically to R =26with a median magnitude of R =24.We show that objects which are unresolved optically (i.e.quasars)follow a correlation between their opticaland X-ray fluxes,whereas galaxies do not.We also find that the quasars with fainter optical counterparts have harder X-ray spectra,consistent with absorption at both wavebands.Initial spectroscopic follow-up has revealed a large fraction of high-luminosity Type 2quasars.The prospects for studying the evolution of the host galaxies of X-ray selected Type 2AGN are considered.1IntroductionX-ray astronomy is moving into a new era with the successful launch of the Chandra and XMM-Newton observatories.The combination of large collecting area,high spatial resolution and wide spectral range enable new surveys to probe deeper than ever before and resolve most of the hard X-ray (2-10keV)background (XRB).The peak intensity of the XRB lies at 30keV and models of the XRB spectrum (astri et al.1995;Gilli et al.1999)predict that obscured AGN with very hard spectra must be much more common in the Universe than naked quasars.Ordinary,optically selected AGN perhaps account for only 10-20%of the total hard XRB.The direct implication is that most of the accretion activity in the Universe is absorbed.We are conducting a deep X-ray survey with Chandra and XMM-Newton in fields which have been well-studied at other wavebands.In each of the two fields,designated N1and N2,Chandra ACIS-I observations of duration75ks have been made.A total of235X-ray sources are detected in the full band images above aflux limit of S0.5−10keV>1.3×10−15erg cm−2s−1. We have also been awarded time for a150ks observation of N2with XMM-Newton.Multi-wavelength coverage is very important for understanding the nature of the X-ray sources.Our fields lie within the European Large-Area ISO Survey(ELAIS)and have been observed with ISO at7,15,90and175µm(Oliver et al.2000)and with the VLA at1.4GHz(Ciliegi et al. 1999).One of the two X-rayfields is co-incident with the widest survey yet made with SCUBA on the JCMT–the UK Submillimetre Consortium’s8mJy Survey(Dunlop et al.2000).The goals of our survey are to obtain a better understanding of the nature of the AGN responsible for the X-ray background(or equivalently,the accretion history of the Universe) and the host galaxies they reside in.Some of the questions to be answered are:What is the connection between AGN activity and star-formation?Do X-ray AGN live in old,evolved galaxies or young galaxies with high levels of star-formation?What are the clustering properties of X-ray sources,both with other X-ray sources and with the general galaxy population?How much of the accretion energy of the Universe is obscured and reprocessed by dust?Is the dust distribution well-represented by torus models with typical galactic gas-to-dust ratios?What X-ray luminosity dependence is there on host galaxy/torus properties?The answers to these questions will come from detailed follow-up of deep X-ray surveys.Note that the small sky areas covered in each pointing with Chandra(equivalent to∼7×7 Mpc at z>∼1)mean that the effects of small-scale clustering and cosmic variance are important. For example,in the N1field the density of X-ray sources brighter than a givenflux-limit is approximately30%higher than that in N2.Thus,no single ultradeep survey(such as the Chandra Deep Field South or the Hubble Deep Field North)will be able to accurately determine the contribution to the XRB,the redshift distribution or luminosity function of X-ray AGN. Only by combining the results of several deep surveys will such results be obtainable.Aflat cosmology with parameters H0=70km s−1Mpc−1,ΩM=0.3andΩΛ=0.7is assumed throughout.2Optical Identification of the ELAIS Deep X-ray SurveyOptical imaging data covering the Chandrafields has been obtained with the William Herschel Telescope(WHT)Prime-Focus Camera and the Isaac Newton Telescope(INT)Wide-Field Cam-era.The WHT data are only in N2and reach to limiting Vega magnitudes of R≈26and I≈26. INT data in bothfields reach limiting AB magnitudes of g′≈25.5,r′≈25,i′≈24.5.In the near-infrared there is full coverage of bothfields at H-band with CIRSI on the INT and partial K-band coverage from INGRID on the WHT and UFTI on the UKIRT.Full details of all these observations will be presented in forthcoming papers.For initial identification of the X-ray sources detected in the full(0.5-10keV)band,we consider only the WHT R-band imaging in N2and the INT r′-band imaging in N1.The X-ray and optical astrometric frames are registered using≈20X-ray sources which are optically bright.The adopted identification procedure takes into account the positional uncertainty of the Chandra source,which is dependent upon the number of photons detected and the off-axis angle (since the PSF is degraded as one moves further from the centre of thefield).The positional uncertainties range from0.5to2.5arcsec.A modified version of the likelihood ratio procedure (Sutherland&Saunders1992)was used to determine the likelihood of possible identifications. Full details of the adopted method will be given in Gonzalez-Solares et al.(in prep.).The median magnitude of the X-ray source counterparts is R≈24with90%identified to a limit of R=26.Our results show a small,but non-negligible,percentage of chance associations (5%)of optical counterparts to Chandra sources,most of which would occur near thefield edges where the Chandra PSF is quite large.Note that this will be a serious problem for deep XMM-Figure1:Magnitude histogram of Chandra source optical identifications in the ELAIS N2region.The histogram is split into those objects with a quasar-like unresolved optical appearance and those which appear to be galaxies. Newton surveys which have lower spatial resolution than bining XMM-Newton and Chandra data will help to alleviate this problem.The good seeing(0.8arcsec)of the optical imaging allows us to make a morphological clas-sification of the optical counterparts of X-ray sources.We have used the Sextractor program to generate image source catalogues and the parameter CLASSFigure2:0.5-10keVflux against R magnitude for ELAIS Chandra sources.Sources also detected in the hard band images(2-10keV)are shown as symbols with a black border.The correlation between optical and X-rayflux exhibited by the quasars,but not the galaxies, explains why the fraction of galaxy counterparts increases fainter than R=24,since it is at this magnitude where L X≈L opt at our X-rayflux limit.3Comparison of optical properties with X-ray hardnessAn analysis of the hardness ratios,HR=[(H−S)/(H+S)],of the Chandra sources shows that HR increases at fainterfluxes(as seen in the CDFS–Tozzi et al.2001).The simplest explanation of this effect is that sources get harder at lower luminosities,since the redshift distribution is unlikely to change significantly from10−14to10−15erg cm−2s−1.It is worth pointing out that HR alone is not a good measure of the absorption of an absorbed power-law X-ray source.Due to the negative k-correction of a highly absorbed source,sources at high redshift will have much lower values of HR than sources with the same rest-frame spectra at lower redshifts.In Figure3we plot the hardness ratio HR against the R magnitude.Note that the hardest sources are galaxies,indicating a lack of quasars with hard spectra.Many of the hard(HR>0)galaxies are very faint,which suggests they are at high redshifts(z>1)and have large absorbing columns.There is no overall correlation of HR with magnitude for the galaxies, although this non-correlation is not trivial to interpret given the positive correlation between R and z for galaxies and the negative correlation between HR and z for a given absorbing column. This could be indicating a more rapid evolution of hard sources than soft sources,as requiredFigure3:Hardness ratio against optical magnitude for ELAIS Chandra sources.by some XRB models(e.g.Gilli et al.2001).The hardness ratio of an unabsorbed,broad line quasar is typically HR=−0.5.Most of the optically bright quasars(R<23)in Fig.3show values similar to this.However,the fainter quasars tend to have harder X-ray spectra such that at R=24,HR≈0.This cannot be a redshift effect because HR decreases at high redshift and if these sources were at low redshift then their galaxies would outshine the nucleus in the optical.A simple explanation for this change is that some of the optically fainter quasars are undergoing absorption in their environments,both in the optical by dust and in the X-rays by cool gas.The nuclear region in such sources may suffer a few magnitudes of extinction or may be totally obscured,in which case the light we observe is either scattered into our line-of-sight or emitted from a region beyond the obscuring medium.For a galactic gas-to-dust ratio,the large absorbing columns necessary for such hard X-ray spectra would equate to very high optical reddening,suggesting that either the light we observe is scattered or there is a very high gas-to-dust ratio as found in some low-redshift Seyferts(Maiolino et al.2001).4The fraction of high-luminosity Type2quasarsModels of the X-ray background have long postulated the existence of high-luminosity(L2−10keV> 1044erg cm−2s−1),obscured Type2quasars,showing only narrow emission lines.Within the unified scheme for AGN,it is expected that the central regions of Type2quasars are obscured along our line-of-sight by a dusty torus(e.g.Antonucci1993).However,previous hard X-rayFigure4:Optical spectra of hard X-ray selected ELAIS Chandra sources.The sources are radio-quiet Type2quasars.surveys with ASCA and BeppoSax have revealed few such objects leading some authors to doubt their existence at all(Halpern et al.1999).Of course,radio-loud Type2quasars(i.e.radio galaxies)have been known for a long time and it is worth considering the known properties of high-redshift radio galaxies when searching for similar objects which are radio-quiet.The ratio of narrow-line to broad-line high-luminosity radio-loud AGN is well determined by virtue of the isotropy of extended low-frequency radio emission and is1.5(Barthel1989).At lower radio luminosities(151MHz luminosities L151<3×1026W Hz−1sr−1)this ratio increases to≈7(Willott et al.2000).There is no evidence for evolution of this ratio as a function of redshift.The perceived picture of the optical appearance of high redshift radio galaxies is biased towards the most luminous objects(e.g.3C radio galaxies)which have been most widely studied. High-redshift3C radio galaxies have very strong high-ionization emission lines and tend to have blue optical-to-near-infrared colours due to extended rest-frame UV emission which is either non-stellar in origin or due to jet-triggered star-formation(e.g Best et al.1998).Due to the correlation between emission line strengths and radio luminosity(Rawlings&Saunders1991), lower luminosity radio galaxies have much weaker lines.They also have less rest-frame UV emission and a large fraction of them are classified as extremely red galaxies(R−K>5)with colours typical of old galaxies at z>1(Willott et al.2001).The identification of hard Chandra sources with galaxies which are faint in the optical and very red(Giacconi et al.2001;Crawford et al.2000;Barger et al.2001)suggests that many of these may also be evolved galaxies at z>1.Cowie et al.(2001)present high quality data on a couple of such objects showing that they do indeed appear to be red galaxies with little contribution to the broad band magnitudes by AGN emission.Although such objects do not always show strong narrow emission lines,their high redshifts imply high X-ray luminositiesfor samples withflux-limits similar to the ELAIS Chandra observations.Accounting for the hard X-ray absorption,(which can be a factor of10for Type2quasars detected at2-10keV; Norman et al.2001)then the intrinsic luminosities would be very high and well above the L2−10keV>1044erg cm−2s−1criterion.The large number of faint galaxy counterparts in Fig. 2suggest that a sizeable fraction of ELAIS Chandra sources will be Type2quasars at z>1.We have obtained optical spectra from the WHT for a small sample offive hard X-ray selected ELAIS Chandra sources.Redshifts were determined for three of thefive sources(the two undetected sources have R∼25and very red optical to near-IR colours).Spectra of the three sources with emission lines are shown in Figure4.Two sources have spectra which are very similar to those of high-redshift radio galaxies,with narrow(FWHM<1500kms−1) lines.CXON2Figure5:The near-infrared Hubble diagram for radio galaxies from the3C,6C and7C samples and Type2X-ray sources from Chandra surveys.Note that the X-ray source host galaxies have a much larger range of K-band luminosities than the radio galaxies.The curves are passive evolution models for galaxies which formed all their stars in a short burst at high redshift(z=10).The upper curve corresponds to L⋆and the lower curve to5L⋆4.Best P.N.,Longair M.S.,R¨o ttgering H.J.A.,1998,MNRAS,295,5495.Ciliegi P.,et al.,1999,MNRAS,302,222astri A.,Setti G.,Zamorani G.,Hasinger G.,1995,A&A,296,17.Cowie L.L.,et al.,2001,ApJ,551L,98.Crawford C.S.,Fabian A.C.,Gandhi P.,Wilman R.J.,Johnstone R.M.,2000,MNRAS,submitted,astro-ph/00052429.Dunlop J.S.,2000,UMass/INAOE conference proceedings on‘Deep millimeter surveys’,eds.J.Lowenthal and D.Hughes,World Scientific,astro-ph/001107710.Eales S.A.,Rawlings S.,Law-Green J.D.B.,Cotter G.,Lacy M.,1997,MNRAS,291,59311.Giacconi R.,et al.,2001,ApJ,551,62412.Gebhardt K.,et al.,2000,ApJ,539,1313.Gilli R.,Salvati M.,Hasinger G.,2001,A&A,366,40714.Gilli R.,Risaliti G.,Salvati,M.,1999,AN,320,28215.Halpern J.P.,Turner T.J.,George I.M.,1999,MNRAS,307L,4716.Hasinger G.,et al.,in”Highlights in X-ray Astronomy”,MPE report272,p.19917.Hornschemeier A.E.,et al.,2001,in press,astro-ph/010149418.Jarvis M.J.,et al.,2001,MNRAS,in press19.Magorrian J.et al.,1998,AJ,115,228520.Maiolino R.,et al.,2001,A&A,365,2821.Norman C.,et al.,ApJ,submitted,astro-ph/010319822.Oliver S.,et al.,2000,MNRAS,316,74923.Rawlings S.,Saunders R.,1991,Nature,349,13824.Schreier E.J.,et al.,2001,ApJ,in press,astro-ph/010527125.Sutherland W.,Saunders W.,1992,MNRAS,259,41326.Tozzi P.,et al.,2001,ApJ,submitted,astro-ph/010301427.Willott C.J.,Rawlings S.,Blundell K.M.,2001,MNRAS,324,128.Willott C.J.,Rawlings S.,Blundell K.M.,Lacy M.,2000,MNRAS,316,449。

a r X i v :a s t r o -p h /9605009v 1 2 M a y 1996A&A manuscript no.(will be inserted by hand later)Key words:atomic processes –ISM:dust –Galaxies:cooling flows,radio galaxies,galaxy formation 1.IntroductionThe strong,spatially extended,rest-frame ultraviolet emission lines observed in high redshift radio galaxies pro-vide one of the principal diagnostics in establishing the state of the interstellar medium in galaxies at early epochs.2Resonance scattering,dust and the UV spectrum of RGscattering,more susceptible to absorption.We explore the fact that any resonance line will be extremely sensitive to geometrical factors,an aspect of the problem which has so far been overlooked in modeling the UV lines.If in radio-galaxies the distant gas clouds are photoionized from the outside by partially collimated UV radiation emitted by the nucleus,the line formation process—particularly for the resonance lines—is very different from internally ion-ized HII regions.The escape of resonance line photons is strongly influenced by the presence of spaces between the line emitting clouds.We have collected from the literature the observed line ratios for a number of high z radio-galaxies in which no contribution from any nuclear BLR is apparent.We have built a diagnostic diagram consisting of the lines CIVλ1549/Lyαvs.CIVλ1549/CIII]λ1909,in which we compare the position of the objects with photoionization models which not only consider the effects of internal dust but also those of the viewing perspective—the angle be-tween the incoming ionizing radiation and the observer’s line of sight.Our concentration on the particular class of radio galaxies is purely for pragmatic reasons.It is these objects,which we presume to harbour a powerful quasar which is hidden at optical/ultraviolet wavelengths to our line of sight,which are most readily found and studied at the high redshifts where we have access to the ultraviolet spectrum from groundbased observations.Our conclusions should be equally applicable to other classes of AGN.For some objects,Lyαis observed to be fainter with respect to CIV than predicted by dust-free photoioniza-tion models.The explanation previously proposed to ex-plain the weakness of Lyαwith respect Hαor Hβhas been dust destruction of resonant Lyαphotons.This is not borne out by our calculations in which we have used arbitrary amounts of dust and found that this cannot si-multaneously weaken Lyαwhile leaving the CIV/CIII]ra-tio relatively unchanged since resonant CIV suffers also from dust absorption.Alternatively,by varying the pro-portions of the illuminated and the shadowed cloud faces which contribute to the observed spectrum,we are bet-ter able to match the data.geometric explanation could naturally explain that some of the brightness asymmetries noted by McCarthy et al.(1991a)on sides of the nucleus.As wefind that geometry alone(with or without inter-nal dust)can in principle explain most of the specific line ratios observed(fainter Lyαcompared with either CIV or HeII),we also discuss the possibility of a patchy outer halo of neutral gas to account for the diffuse Lyαseen in some cases to extend much beyond the CIV emitting region and even the outermost radio lobes.Reflection by cold gas of the brighter Lyαemitting side of the ionized clouds would lead to a narrower profile for such a dif-fuse component.Another possibility is that part of the beamed nuclear continuum and BLR radiation might be reflected at the wavelength of Lyαby thin matter-bounded photoionized gas at very large distances from the nucleus leading to a diffuse Lyαcomponent aligned with the radio axis.It appears to us that geometrical perspective effects are an essential component of the interpretation of the UV spectrum of radio-galaxies whether or not dust is present. Furthermore,a spectrum in which only Lyαappears does not necessarily imply starburst activity,other lines must be observed before the existence of an HII region can be inferred.2.Data sample and modeling procedure2.1.The dataWe have constructed a data sample containing galaxies at high z for which the CIV,Lyαand,in most cases CIII], emission lines have been measured.Since very high den-sities such as those encountered in the BLR alter signifi-cantly the line formation and transfer processes,we have excluded those objects which show evidence of a BLR.In Table1,we list the object names,the line ratios of inter-est to us here,the redshift and the reference to the ob-servations.The larger fraction of the data are taken from the recent thesis by van Ojik(1995)which includes ob-jects selected on the basis of a very steep radio spectrum. Probably by virtue of the radio selection,these sources populate the region of the line ratio diagram(Fig.4)with lower CIV/CIII]ratios(lower ionization parameter)than the previously published objects.The line measurements refer to the integrated emission from the object collected with a long slit aligned with the major axis.2.2.The model and its parametersThe data are compared to photoionization models com-puted using the multipurpose photoionization–shock code MAPPINGS I.The version described in Binette et al. (1993a,b)is particularly suited to the problem since it considers both the effects of the observer’s position with respect to the emitting slab and the ionizing source(cf, Fig.3).We distinguish between the spectrum seen from the back and from the front of the slab.The code also considers the effects of dust mixed with the ionized gas: extinction of the ionizing continuum and of the emission lines,scattering by the dust and heating by dust pho-toionization.The treatment of the escape of resonant CIV and Lyαphotons in a dusty medium is described in Ap-pendix B of Binette et al.(1993a)and is based on the results of Hummer&Kunasz(1980).The dust content of the photoionized plasma is de-scribed by the quantityµwhich is the dust-to-gas ratio of the plasma expressed in units of the solar neighbourhood dust-to-gas ratio.To specify the gas metallicity,we scale with a factor Z the solar abundance set of trace elements from Anders&Grevesse(1989).He/H is kept constant at 0.1.We generally consider the solar case with Z=1.Since the presence of dust implies depletion of refractory trace elements,we use the prescription given in Appendix A ofResonance scattering,dust and the UV spectrum of RG 3Binette et al.(1993a,hereafter BWVM3)to derive the gaseous phase abundances for any metallicity Z .The de-pletion algorithm is function of the ratio µ/Z and makes use of the depletion indices listed in Whittet (1992)for the different metals.The calculations consider the gas pressure to be con-stant (isobaric models)and so the density behaviour with depth in the cloud is determined by the behaviour of the temperature and the ionization fraction of the gas.The ionization parameter,a measure of the excitation level of the ionized gas,is defined as the quotient of the density of ionizing photons incident on the cloud and the gas density:U =∞νof νdν/hν4Resonance scattering,dust and the UV spectrum of RGTable1.Observed UV line ratios for several high z RG with not apparent broad componentAverage RG 2.0540.118McCarthy(1993)Resonance scattering,dust and the UV spectrum of RG5Fig.2.Observed and predicted UV line ratios.Filled trian-gles are the observed ratios of objects observed by different au-thors.Open circles correspond to data taken from van Ojik’s thesis,selected on the basis of a very steep radio spectrum.The open diamond is the average radiogalaxy spectrum of Mc-Carthy (1993).Dotted lines represent models in which the front and back spectra have been summed.It is clear that high val-ues of the ionization parameter U are required to reproduce the high CIV/CIII]values observed.photons by resonance scattering in the presence of dust is the explanation for its faintness,but why does not the same process reduce CIV which is also a resonance line?Is there an alternative explanation for this selective dimming of Ly α?To answer this question,we examine the geometrical aspects of the line formation process.Each emitting cloud is approximated as a plane parallel slab which contains a fully ionized region and a partially ionized zone where low ionization species co-exist (e.g.,O 0,S +,etc)with a mix-ture of H 0and H +.In principle there can be an additional neutral zone which does not contribute to the emission line intensities (see Fig.3).In this section we consider the dust-free case.For most lines (like CIII],HeII,H β,[OIII],etc)line opacity is neg-ligible and the line is emitted isotropically with photons escaping freely in all directions.However,when the line opacity is important as it is for CIV and Ly α,line scat-tering occurs which increases the path length.Another im-portant effect of large optical depths is that the line pho-ton will not escape isotropically.A resonance line photon following many scatterings must statistically escape in the direction of highest escape probability which can be shown to be the front for the photoionized slab depicted in Fig.3.In the case of Ly α,the reason is that —while Ly αpho-tons are generated more or less uniformly within theslabFig.3.Adopted slab geometry for the constant pressure pho-toionization calculations.The slab comprises a fully ionized zone and a partially ionized zone (PIZ)where H +and H 0co-exist.Beyond the PIZ we may have a region of neutral gas.(except within the PIZ)by recombination —the neutral fraction and therefore the incremental line opacity dτL /dx increases monotonically as a function of depth as discussed in more detail by BWVM3.This means that for an open geometry like that shown in Fig.3,the zone of equal es-cape probability of front vs back occurs far beyond the point where half the luminosity of Ly αis produced.For the collisionally excited CIV line,the tendency to escape from the front also exists although it is less pronounced.It arises mainly because the emissivity of CIV is larger towards the front due to the temperature gradient across the C +3zone.Note that while CIV is emitted and scat-ters within the rather limited zone containing C +3,Ly αremains subject to scattering outside the region where it is produced.While most Ly αemission is produced in the fully ionized zone,most of the line opacity occurs within the PIZ.The presence of a layer of neutral gas beyond the PIZ will increase the anisotropy of Ly αescape.The effects described qualitatively above are shown in Fig.4using detailed photoionization calculations.We present the same sequence of dust-free models as in Fig.2but distinguish between the spectrum seen from the back —equivalent to observing the clouds through the PIZ —from that seen from the front —equivalent to seeing the UV irradiated side.The differences are striking:both Ly αand to a lesser extent CIV are fainter when seen from the back.The CIII]line is isotropic in the dust-free case.The fact that Ly αis more affected by perspective is due to the significant amount of neutral hydrogen (i.e.,large line opacity)in the PIZ which acts as a mirror.Although we have considered a very simplified geometry in our calcu-lations,the method nevertheless treats properly the es-sential physical effects and indicates how important the viewing direction is in this open geometry.6Resonance scattering,dust and the UV spectrum ofRGFig.4.Influence of viewing geometry on the UV line ratios.The dotted line corresponds to the line spectrum seen from the back of the slab (cf,Fig 3)while the dashed line corresponds to the spectrum seen from the front (UV irradiated face).It is apparent that perspective plays a very important role on the CIV/Ly αratio.The solid line represents models obtained by summing the back and front spectra which would represent symmetric case where equal numbers of clouds are observed with shadowed and illuminated faces.Fig.4suggests that perspective effects alone (without any dust )are sufficient to explain the weak Ly αseen in some objects.Ionization bounded calculations with U =0.1imply total hydrogen column densities (H +re-gion +PIZ)N H ∼1022cm −2(of which about 60%is ionized).Adding a modest neutral zone beyond the PIZ of ≃41021cm −2would double the CIV/Ly αratio with-out affecting in any way the CIV/CIII]ratio.It seems,therefore,that a geometry where we see preferentially the ionized gas from the side of the PIZ gives us an explana-tion of the weakness of Ly α.How would this apply to the EELR of the powerful radio galaxies?In a very simplified scheme,we can imag-ine (see Fig.5)that the clouds seen from the nearside cone are seen from a direction which we approximate as the back perspective in our slab calculations while clouds on the far side would be seen from the front.The stud-ies of McCarthy et al.(1991a)which emphasized the one-sideness of the line brightness suggest that the observer with limited spatial resolution at very high z will be bi-ased towards either a back or front dominated perspective depending on whether it is the near or the farside illumi-nated cone which is intrinsically brighter.Our proposed interpretation of the high CIV/Ly αratio of some objects in Fig.4is that they correspond to the case where the brightest clouds are seen from behind.Note that,becauseof the intensity weighting,there is rather little difference between the pure front perspective and the sum of equal numbers of front-and back-clouds,a small effect which is not easily distinguished from that of slightly reducing the value of U.Fig.5.Our conclusion from this section is that perspective effects in an open —externally illuminated cloud —ge-ometry go a long way towards explaining the behaviour of the CIV,Ly α,CIII]line ratio diagram.It implies that,when spatially resolved spectra become available,marked asymmetries in the CIV/Ly αratio could arise when the axis of the illuminated cones forms a large angle with the sky plane.3.3.The effects of internal dustThere is evidence for the existence of dust in some high z radio galaxies.The detection of 4C41.17(z =3.8)and B20902+34(z =3.5)(Chini &Kr¨u gel 1994;Dunlop et al.1994)and 8C1435+635(z =4.26)(Ivison 1995)in the mm spectral range is attributed to warm dust.Also,the IRAS galaxy F10214+4724(z =2.29)has been shown,from the far infrared flux,to contain ∼108M sol of dust,although this figure may be reduced by the gravitational lensing am-plification factor (Serjeant et al.1995).In addition,the mid-to far-infrared measurements at intermediate z are explained as emission from warm dust (Heckman,Cham-bers &Postman 1992).We should emphasize,however,that we have no direct measurement of how this dust is spatially distributed.In the event that this infrared emis-sion arises from the reprocessing of higher energy photons from the AGN by a dusty torus,it has no direct bearing on our modeling of ionized gas at tens of kpc.However,there is considerable evidence that aligned blue polarized continuum is the result of scattering of the anisotropic nu-clear radiation field by dust (e.g.,Cimatti et al.1993).In this case it is very likely that at least some of the dust is internal to the extended line emission regions.In thisResonance scattering,dust and the UV spectrum of RG 7section we consider the effects of including dust within the gas clouds which emit the UVlines.Fig.6.Effects of varying the amount of internal dust as seen from the front perspective.Short dashed line corresponds to models with µ=1.0,the long-dashed to models with µ=0.3and solid line to dust free models.We illustrate the effects of dust mixed with the emit-ting gas in Fig.6where we plot sequence of models which correspond to the front perspective for three different dust-to-gas ratios:µ=0,0.3and 1.The effects on the line ratios are evident.The resonance scattering suffered by CIV and Ly αincreases their pathlengths many times and,therefore,the probability of their being absorbed by dust grains is much higher than for CIII].In the case of Ly α,however,the geometrical thickness of the H +region exceeds greatly that of the C +3since we observe more than one stage of ionization of metals (e.g.C +2).Any reasonable parameters for the ionization structure of a photoionized slab with Z ∼1indicates that the opac-ity in Ly αgreatly exceeds that of CIV,which implies a larger pathlength increase for Ly αthan for CIV.This re-sults in relatively more Ly αabsorption by dust.This effect explains how the ratio CIV/Ly αincreases somewhat with increasing µ.Dust absorption of resonant CIV on the other hand causes a comparable decrease in CIV/CIII].What is important in these results is that,even with a concentra-tion of internal dust as high as µ=1(equivalent to that in solar neighbourhood cold clouds),it is not possible to reproduce the high ratio CIV/Ly α≥1which is observed in some objects and has been attributed to dust.We em-phasize that higher amounts of dust do not change these results.For instance models with Z =µ=2do not lead to any higher Ly α/CIV ratio than the models shown in Fig.6.Fig.7.Effects of perspective combined with small quantities of internal dust.The dust-free U sequences of Fig.5are re-peated in this figure.The nearly vertical dotted line with open circles corresponds to the last model with U =0.1(but with metallicity Z =0.3)in which the dust content is increased in proportion up to µ=rger quantities of dust do not produce any further increase in Ly α/CIV as seen from the back since it severely extinguishes both the CIII]and CIV lines which are produced in the front layers.When the clouds are viewed from behind,even with µ=0.3they are sufficiently opaque that all the UV lines are severely absorbed.With τV =510−22µN H ∼5and µ=1,all of the high excitation UV lines become absorbed within the PIZ.To produce an acceptable spectrum,with-out reddened CIV/CIII]and CIV/HeII ratios,as seen from the back of an ionization bounded slab with U =0.1,we have to use much smaller amounts of dust like µ≃0.017(2%of local ISM)in models.In Fig.7,we plot the back and front sequences for such models.These can reproduce the weak Ly αobjects although this result is obtained only for the back spectrum,emphasizing that perspective is the dominant factor.It should be noted that the amount of extinction within the ionization bounded slab implied by µ=0.017(i.e.,A V ≃0.1–0.2)is consistent with the quantity of small dust grains needed to explain the ex-tremely blue continuum of the “detached”ionized cloud in PKS 2152-69(di Serego Alighieri et al.1988;Magris &Binette private communication),supposing that the con-tinuum energy distribution is the result of dust scattering of the nuclear radiation.Note that because the scattering/absorbing dust is lo-cally kinematically linked to the emission line gas,we require much smaller column densities of HI than the absorbing screen proposed by van Ojik et al.(1994)(∼1023cm −2)to explain CIV/Ly α∼1.The line widths and8Resonance scattering,dust and the UV spectrum of RGcentroids of the PIZ and of the fully ionized gas are ex-pected to be quite similar within each of the emitting clouds,the ensemble of which could have a greater‘tur-bulent’velocity dispersion.To reproduce the extreme case of the IRAS galaxy F10214+4724with CIV/Lyα∼10,we need additional neutral gas beyond the PIZ.For instance,it requires only a column density of N H0∼1.51021cm−2assuming µ=0.017.Alternatively we might consider thatµwithin the PIZ increases with depth as the degree of ionization decreases.In this case,no additional HI gas would be re-quired.If this IRAS galaxy is indeed an extreme Seyfert2 as argued by Elston et al.(1994),then a closed,dust en-shrouded geometry as proposed by BWVM3to explain the Lyman and Balmer decrements in Seyfert2would be more appropriate than the open geometry adopted here which may be applicable only to the truly extended large scale gas.It is likely that objects with unusually strong NVλ1240emission(such as in F10214+4724and TX0211-122)are cases where the NV originates predominantly from the inner NLR.A high NV/CIV ratio indicates very enriched gas which is not unexpected within the inner parts of an AGN(Hamman&Ferland1993)and is con-sistent with the lack of convincing evidence that NV is spatially resolved in any of these objects.3.4.Neutral gas mirrorsSo far we have considered the effects of scattering by gas which forms part of the line emitting clouds:we refer to this as an‘intrinsic’process.Similar effects could be produced by a large-scale distribution of predominantly neutral material surrounding the emitting regions:we will refer to this as the‘extrinsic’case.Any extrinsic neutral gas component with a non-negligible covering factor could affect the observed spectra in a way which mimics that of the back perspective described earlier.Let us suppose that this outer material is broken up into cold gas clumps which are randomly distributed.In such a case,some Lyαphotons which leave the ionized cones will escape the re-gion through the holes between the external clumps while others will strike the neutral clumps and be immediately scattered away to escape eventually through another hole in a different direction(see Fig.8).If observed with suf-ficient spatial resolution,such a geometry would result in holes in the Lyαbrightness due to reflection by intervening clumps as well as diffuse sources corresponding to reflec-tion from clumps on the far side of the source.The bulk of the Lyαluminosity would be preserved but redistributed on an apparently larger scale than the true line emitting clouds.Only a more closed geometry would result in a sig-nificant destruction of Lyα,assuming that the interclump space does not contain pure dust segregated from the gas phase.The reflection efficiency of clouds will,of course,de-pend strongly on the relative velocityfields of the emitting and the cold regions.If the extended gas has a large scale ordered motion but small‘microturbulence’,the reflection effects would be localized.Broad Lyαemission and con-tinuum from the AGN could,however,be scattered by any of the extranuclear clouds(see Sect.3.4.2).In gen-eral,we would expect the diffuse,scattered Lyαto show a narrower line than the integrated emission profile.We will review the observational evidence for the ex-istence of such large scale mirrors.3.4.1.The radio galaxy PKS2104-242PKS2104-242(McCarthy et al.1990b)shows very ex-tended emission lines of Lyα,CIV and HeII associated with continuum knots which lie between the radio lobes and are aligned with the radio axis.The Lyαimage shows emission resolved into three distinct clumps,two of them corresponding roughly with the two continuum knots.In addition,there appears to be a low surface brightness halo of emission surrounding the entire object.Is this halo a consequence of reflection by neutral material of either Lyαemission or of the intense nuclear continuum,or are these photons emitted locally by H+recombination?A way of discriminating between these two possibilities is to look for the detection of any other emission line in the halo.If Lyαis the result of reflection by cold HI,there will be no other lines(except possibly resonant MgII).If it is instead produced by recombination,other lines should be detected in the rest-frame optical band like Hα,[OIII]λ5007,etc. Fig.8.The extrinsic case.Neutral material external to the ion-ized regions can strongly influence the appearance of the object in Lyα.In thisfigure,the neutral clumps(black clouds)cover a significant fraction of the ionized regions(grey+black clouds). The solid arrows represent Lyαphotons emitted directly in the ionized regions.Dashed arrows represent Lyαphotons that af-ter striking neutral clumps are reflected in other directions. The reflection by the neutral H atoms is so effective that the dust has little chance to interact with the photons before they find a hole to escape.Resonance scattering,dust and the UV spectrum of RG93.4.2.The radiogalaxy3C294From images of this object,McCarthy et al.(1990a)re-port the existence of Lyαemission extending over a region covering170kpc.The Lyαemission is elongated and well aligned with the inner radio-source.An intriguing aspect of the observations is the one-sidedness of the CIV emis-sion(obtained with an long slit aligned with the axis of elongation).The decreasing linewidth of Lyαon the side where CIV is absent(South)suggests to us the possibility that Lyαin the south corresponds to scattered continuum and BLR Lyαphotons by HI clumps lying along this di-rection provided the ionizing radiation has already been reprocessed(filtered out)within the nuclear regions into NLR or BLR emission.Failing that(scattering by HI clumps),an alternative possibility resides in very thin matter-bounded photoion-ized sheets of gas which can be very efficient at reflecting the intense(beamed)nuclear continuum as well as BLR Lyαphotons.Due to its large scattering cross section, very small column depths of H0(a trace specie within the ionized phase)are sufficient to scatter effectively the im-pingingflux within the core of the Lyαthermal profile. If we suppose the existence within the cone of ionization of the radio-galaxy of a population of clouds of similar physical conditions to that thought to apply to Lyαforest clouds,the energy reflected due to resonant scattering by HI typically exceeds that generated within the clouds by reprocessing of the ionizing radiation(i.e.by recombina-tion).In effect,adopting N H0=1013.8cm−2as a typi-cal column density of such cloud,we derive an equivalent width in absorption(i.e.the scattered intensity)of EW absLyα≈0.151With these parameters,the total column density is given by N H=N H+≈2.3104N H0.expected to increase up to8times the Lyαemitted by re-combination.To account for the observed FWHM<∼1000 km s−1in3C294(South),one most rely on a turbulent velocityfield for the the Lyαclouds.We ought to consider seriously the possibility that Lyαon the southern knot of3C294corresponds to large scale scattered light by either HI or even ionized gas unless other bonafide emission lines were detected.So far,no other lines than Lyαare observed which is consistent with our suggestion.The only certain way of discriminating be-tween reflection and in situ emission would be the detec-tion at this radius of any non-resonance line in either the optical([OII],[OIII],Hβetc.)or the UV(CII],CIII]).To conclude,we believe that HII regions or starbursts should not be considered the default emission mechanism in cases where only Lyαis present,especially along the axis where the AGN supposedly collimates its intense nuclear contin-uum+Lyα(BLR)light.3.4.3.Lyαabsorption in0943-242The radio galaxy0943-242shows a Lyαprofile with sev-eral absorption features which cover the whole of spatial extent of the Lyαemission(R¨o ttgering et al.1995).The strongest of the absorbers is at least as extended spatially as the emitting region(∼13kpc)and its HI column den-sity is1019cm−2.The absorption line is blue-shifted by 250km s−1with respect to the emission peak.The‘screen’is kinematically distinct from the emitting gas and there-fore clearly external to the emitting clouds).This is a clear demonstration that HI screens—or mirrors depending on the perspective—exist on galaxy scales at these early epochs.From Fig.5of R¨o ttgering et al.,we estimate that this cloud would reflect∼30%of the Lyα.Thicker clouds could exist around other objects but would be hard to detect when only the faint wings at the extremities of the emission profile were transmitted.In the case of0943-242,even if the screen contained dust,its effects would be negligible.The reason,as stated before,is that Lyαphotons(seen by the cloud)are incident from the exterior and will be very effectively scattered away before being absorbed by the dust.Even the photons getting through (i.e.,without any scattering)to us because they are suffi-ciently far in the wings of the profile would not see much dust in this particular cloud.With such low column densi-ties,τV∼510−22µN H=7.510−5forµ=0.015and0.005 forµ=1.Much thicker clouds may result in some non-negligible extinction but again this does not arise because of the multiple scattering of Lyα.4.ConclusionsIf the large-scale extended emission line regions(EELR) in radio galaxies are photoionized predominantly by the collimated UV radiation emitted by a hidden AGN,the line emission and transfer processes are characterizedby。