Detections of Lyman Continuum from Star-forming Galaxies at z~3 Through SubaruSuprime-Cam N

a r X i v :0805.4012v 1 [a s t r o -p h ] 26 M a y 2008Submitted to ApJLPreprint typeset using L A T E X style emulateapj v.03/07/07DETECTIONS OF LYMAN CONTINUUM FROM STAR-FORMING GALAXIES AT Z ∼3THROUGHSUBARU/SUPRIME-CAM NARROW-BAND IMAGING 1I.Iwata 2,A.K.Inoue 3,Y.Matsuda 4,5,H.Furusawa 6,T.Hayashino 7,K.Kousai 7,M.Akiyama 6,8,T.Yamada 8,D.Burgarella 9,and J.-M.Deharveng 9Submitted to ApJLABSTRACTKnowing the amount of ionizing photons from young star-forming galaxies is of particular impor-tance to understanding the reionization process.Here we report initial results of Subaru/Suprime-Cam deep imaging observation with a special narrow-band filter to optimally trace ionizing radiation from galaxies at z >3.The unique wide field-of-view of Suprime-Cam enabled us to search for ionizing photons from 198galaxies with spectroscopically measured redshifts z ≃3.1.We detected ionizing radiation from 7Lyman break galaxies (LBGs),as well as from 10Ly-αemitter (LAE)candidates.Some of the detected galaxies show significant offsets of ionizing radiation from non-ionizing UV emission.As an average of the 7detected LBGs,the observed flux density ratio of non-ionizing UV to ionizing radiation is estimated to be 4.9,which is smaller than values expected from population synthesis models with a standard Salpeter initial mass function (IMF)and dust attenuation.This implies an intrinsically bluer spectral energy distribution,e.g,that produced by a top-heavy IMF,for these LBGs.The observed flux density ratios of the detected LAEs are even smaller than those expected from a top-heavy IMF and QSOs if they are truly at z ≃3.1.We find that the average escape fraction of ionizing photons for the detected LBGs should be higher than 15%.Subject headings:galaxies:evolution —galaxies:high-redshift —cosmology:observations —inter-galactic medium —diffuse radiation1.INTRODUCTIONIonizing radiation from star-forming galaxies is a likely primary source of cosmic reionization.Although the ra-tio of the flux density of Lyman continuum escaping from a galaxy to that produced in the galaxy,the so-called es-cape fraction (f esc ),is a key parameter for evaluating the contribution of galaxies to cosmic reionization,it has been poorly constrained due to the fact that Lyman con-tinuum photons are easily absorbed by the intergalactic medium (IGM).Direct observations of Lyman continuum from z >4are virtually impossible because of a rapid increase in the number density of Lyman limit systems toward high redshifts (Inoue &Iwata 2008).Therefore,we must focus on z ∼3where the IGM optical depth is still about unity on average.Steidel et al.(2001)de-tected Lyman continuum in a composite spectrum of 29Lyman break galaxies (LBGs)at z =3.4.The ob-served flux density ratio was f 1500/f 900=17.7±3.8.1Based on data collected with the Subaru Telescope,which is operated by the National Astronomical Observatory of Japan.2Okayama Astrophysical Observatory,National Astronomical Observatory of Japan,Honjo,Kamogata,Asakuchi,Okayama 719-0232;iwata@oao.nao.ac.jp3College of General Education,Osaka Sangyo University,3-1-1Nakagaito,Daito,Okaka 574-8530,Japan4Department of Astronomy,Graduate School of Science,Kyoto University,Kyoto 606-8502,Japan5Optical and Infrared Astronomy Division,National Astronom-ical Observatory of Japan,2-21-1Osawa,Mitaka,Tokyo 181-85886Subaru Telescope,National Astronomical Observatory of Japan,650North A’ohoku Place,Hilo,HI 96720.7Research Center for Neutrino Science,Graduate School of Sci-ence,Tohoku University,Sendai 980-8578,Japan8Astronomical Institute,Graduate School of Science,Tohoku University,Aramaki,Aoba-ku,Sendai 980-8578,Japan9Laboratoire d’Astrophysique de Marseille,Traverse du Siphon,BP 8,13376Marseille Cedex 12,FranceShapley et al.(2006,hereafter S06)detected Lyman con-tinuum individually from two out of 14observed LBGs at z ∼3through a deep long-slit spectroscopy,and derived f 1500/f 900=12.7±1.8and 7.5±1.0for the two LBGs.So far the number of galaxies with direct observation of Lyman continuum is too small to find common proper-ties shared by galaxies with large f esc ,or to estimate a typical value of f esc .In this letter we report the initial results of the obser-vations of Lyman continuum from galaxies at z ∼3with Subaru/Suprime-cam (S-Cam;Miyazaki et al.2002).A narrow-band filter imaging was adopted to search for Lyman continuum,as pioneered by Inoue et al.(2005)for galaxies at z ∼3.The use of narrow-band imag-ing instead of slit spectroscopy enables us to examine ionizing radiation from a large number of galaxies simul-taneously,as well as to examine emission offsets from rest-frame non-ionizing UV radiation.The target field of the present study is the SSA22field where the promi-nent proto-cluster of galaxies at z =3.09has been dis-covered (Steidel et al.1998,2000;Hayashino et al.2004,hereafter H04).We adopt a cosmology with the param-eters ΩM =0.3,ΩΛ=0.7and H 0=70km/s/Mpc.All magnitudes are measured in the AB system.2.OBSERVATIONS AND DATA REDUCTIONWe produced a special narrow-band filter NB359which has a central wavelength of 359nm with a FWHM of 15nm (and >10%transmittance between 350nm and 371nm).This filter is designed to trace Lyman contin-uum of galaxies at z 3.06with contamination from non-ionizing radiation at less than 1%for a typical star-forming galaxy.In the laboratory measurement it was verified that the transmission of the filter is less than 0.01%in the wavelength range of 400nm–1200nm.We2Iwata et al.also confirmed that the central wavelength is stable with rms=0.47nm at every position on thefilter.The imaging observations using Subaru/S-Cam with the NB359filter were carried out in2007September 10–14(UT).Three nights were photometric while in the other two nights cloud coverage was relatively high. The seeing condition was good throughout the observ-ing dates,with a FWHM of0.′′5–0.′′9in NB359images. The pointing center(α=22h17m26s.1,δ=+00◦16′11′′[J2000])is aligned to the previous Subaru/S-Cam obser-vations of the SSA22field(Matsuda et al.2004,hereafter M04;H04).For half of the shots the position angle was switched from90◦to270◦,in order to reduce the effect of chip-to-chip sensitivity variation.The data reduction was made using SDFRED version 1.2.5(Yagi et al.2002;Ouchi et al.2004).The rejection of frames under poor sky condition led a selection of45 frames with1,800sec exposures to produce the mosaic image.Theflux calibration was made using measure-ments of spectroscopic standard stars.The3σlimiting magnitude of thefinal mosaic image is estimated to be 27.33AB mag for1.′′2diameter aperture.Source detection and photometry were done using SEx-tractor(Bertin&Arnouts1996)version 2.5.0in the double-image mode;source detection was performed with the R-band image,and photometry in other band images was done at each position of a source detected in the R-band.Galactic extinction has been corrected using the dust map by Schlegel,Finkbeiner&Davis(1998).As described below,there are objects detected in NB359but their positions in NB359are slightly offset from those in the R-band.In such cases we performed source detection and photometry with the NB359image.3.RESULTSThere are198objects with spectroscopically measured redshifts larger than3.0in the S-Camfield-of-view(32′×26′).Ten objects classified as QSOs/AGNs from optical spectra and/or X-ray observation have been excluded. Among the198objects there are44LBGs reported in the literature(Steidel et al.2003,S06),29LBGs selected as U-dropout galaxies and identified spectroscopically with VLT/VIMOS(Kousai et al.,in preparation),and125 Ly-αemitters(LAEs)and Ly-α’blobs’selected through a narrow-band(NB497)imaging with Subaru/S-Cam (H04,M04)and identified spectroscopically with Sub-aru/FOCAS and Keck/DEIMOS(Matsuda et al.2006, Matsuda et al.,in preparation).We detected16objects with secure detection(>3σ) within1.′′2diameter apertures in NB359;6are LBGs and10are LAEs.The redshift range of the LBGs is 3.04<z<3.31,while for the LAEs it is3.07<z<3.10. In order to eliminate a possibility of spurious detections, we split the frames used to create thefinal NB359image into two sets and generated two mosaic images.All16 objects were detected in both images at 2σlevel.We also executed a detection test using a negative image in NB359and found that a probability of spurious detection at>3σlevel is∼0.4%.Therefore,we are confident in the reality of source detection in NB359.In Figure1we show four postage stamp images as examples.In addition,SSA22a-C49,from which S06detected Lyman continuum,has 2.95σsignificance in our im-age.The measuredflux density within the1.′′2apertureFig.1.—Examples of objects detected in the NB359image.In each panel NB359(left)and R-band(right)images are shown.For the R-band images a contour map of the NB359image(2and3σ)is over-plotted.Field of view is20′′×20′′.In the bottom line SSA22a-C49and D3,which are reported to be detected in S06are also shown.D3is not detected in NB359.(5.5±1.0×10−2µJy)is consistent with that measured with a long-slit spectroscopy by S06(6.9±1.0×10−2µJy).We add this object to the detection list in NB359, bringing the total number of detected objects to17(7 LBGs and10LAEs).On the other hand,we could not find any significant signal in NB359at the position of SSA22a-D3which is another object detected in Lyman continuum by S06Since the3σdetection limit within 1.′′2aperture at the position of this object(5.8×10−2µJy)is well below theflux density reported by S06 (11.8±1.1×10−2µJy),it is not clear why the object is not visible in our NB359image.In Figure1images of SSA22a-C49and D3are also shown.All of the10LAEs detected in NB359show a promi-nent emission line around497nm in their spectra taken with FOCAS and DEIMOS,but continua are not well detected due to their faintness.All but one spectrum of these10objects have a resolution high enough to distin-guish the doublet if the emission line detected with the NB497filter is[O ii]λ3727at z=0.33,and wefind no such feature.However,there is still a possibility that they are AGNs at lower redshifts if the emission line is C ivλ1549at z=2.21or Mg iiλ2798at z=0.78, although some of them show a spatially extended emis-sion in the NB497image.Unfortunately,the wavelength coverage of the spectra obtained so far is too narrow to definitely rule out the low redshift possibility.Follow-up spectroscopy would be required to clarify whether these objects are really LAEs at z≃3.09.If a faint foreground object lies very close to an ob-ject at z≃3.1,they might mimic Lyman continuum in our NB359,and it would be difficult to distinguish it at longer wavelengths if the object at z≃3.1is brighter than the foreground object.Siana et al.(2007)discussed such a possibility of contamination by faint foreground objects in the z∼3Lyman continuum survey.Follow-ing Siana et al.(2007),we roughly estimate the proba-bility of such a case by assuming a surface density of faint galaxies and their spatial distribution to be purely random.The apparent magnitudes of galaxies detected in NB359range from26.5mag to27.5mag.The sur-face number density of galaxies in this magnitude range from U-band number count by Williams et al.(1996)is ∼105mag−1deg−2.If we consider that a foreground ob-Lyman Continuum from z=3Galaxies3ject within a0.′′5radius cannot be distinguished in our NB359image,each object has∼0.6%chance of such fore-ground contamination.Since we have198spectroscopic sample galaxies,one or two objects among the galaxies detected in NB359may be explained by contamination by such a foreground object.However,it would be dif-ficult to imagine that all17detections could be due to foreground contamination.4.DISCUSSION4.1.Properties of Detected ObjectsAs seen inFigure1,shapes and positions of the emit-ting regions in the17detected galaxies in NB359are quite different from those in the R-band in many cases.The rms of positional offsets between the NB359and the R-band images for foreground objects is∼0.′′25.The offsets of4LBGs among the7LBGs detected in Ly-man continuum exceed3σ,and the average offset is0.′′97 (3.8σ,corresponding to7.4kpc at z=3.09).Such differ-ences in the shape and the position between the emitting regions of ionizing radiation and those of non-ionizing UV may give us a clue to understand how Lyman continuum escapes from galaxies.For instance,Lyman continuum may escape through a chimney-like structure in the in-terstellar medium(Razoumov&Sommer-Larsen2007), and we may see the emission only from some limited regions.Another possibility is that the spatial distribu-tion of the Lyman continuum sources is different from that of the non-ionizing UV emitting stars.For some detected galaxies,high resolution images taken with the HST/ACS using the F814Wfilter are available.Wefind that many of them show multiple knotty structures and the Lyman continuum emission appears to be associated with one of such knots.We will discuss these geomet-rical features further in a future publication.Note that at the moment with the F814W image we cannot distin-guish between superposition of a foreground object and multiple knotty substructures frequently seen in LBGs (e.g.,Lowenthal et al.1997).The position offsets of Ly-man continuum from non-ionizing UV peak also imply a possibility that long-slit spectroscopy may miss Lyman continuum.In Figure2the NB359−R colors of the198objects with spectroscopic redshifts larger than3.0are plotted against their R-band magnitudes.The NB359−R color corresponds to the apparentflux density ratio of ioniz-ing to non-ionizing UV photons for objects at z∼3. Wefind that luminous(R 25)objects,which roughly correspond to those with L>L∗of LBGs at z∼3 (Sawicki&Thompson2006),show relatively red colors of NB359−R 2.On the other hand,some of the less luminous objects show bluer colors,0 NB359−R<2.If the IGM opacity does not depend on the source luminosity,this suggests that the ionizing-to-non-ionizing UV escapeflux density ratio(a proxy of f esc;see Inoue,Iwata&Deharveng2006)is lower in relatively lu-minous objects.Interestingly,this appears contrary to the argument by Gnedin,Kravtsov,&Chen(2008)that f esc is smaller for galaxies with smaller star formation rates(or mass).We use deep multi-band optical imaging data(H04) to investigate spectral energy distribution(SED)of the detected sample.The UV spectral slopes of the detected LBGs areflat(0<V−i′<0.5).There are two clearlyFig. 2.—NB359−R colors of objects in the SSA22field with spectroscopic redshifts z>3.0.The higher and lower dashed lines indicate approximately1σand3σlimits in the NB359image.The objects without significant detection( 3σ)in the NB359image are plotted with open circles and upper arrows.The detected LBGs are shown with solid Es are classified with two types according to their rest-frame UV slopes:red UV colors(filled cir-cles)and blue UV colors(triangles).See text for details.distinctive sub-groups of the detected LAEs from their V−i′colors:one is red(0.5<V−i′<1.0),and the other is blue(−0.5<V−i′<0).In Figure2different symbols are used for different types of SEDs.Blue LAEs show extremely blue NB359−R colors(<0.3),while LBGs have NB359−R larger than0.7.The difference between these three types becomes clear if we place them into the NB359−R and V−i′two-color plane,as shown in Figure3.To compare with these observed colors we calculated predictions of colors for young star-forming galaxies with a population synthesis code(Starburst99 version5.1;Leitherer et al.1999).A model at zero age, with the Salpeter IMF(0.1–120M⊙;Salpeter1955)and Padova evolutionary tracks with metallicity Z=4×10−4 is used as afiducial one.Since models with higher metal-licity or older age have redder colors,this model has the bluest SED under an assumption of the Salpeter IMF. The colors of the model without IGM and dust attenua-tion and assuming f esc=1at z=3.0and3.3are shown asfilled and open circles connected with a solid line in Figure3.An arrow indicates the direction of dust atten-uation following a prescription by Calzetti et al.(2000) for a galaxy at z=3.Forλ<1200˚A we simply extrap-olated their attenuation law.Such smooth extrapolation toward extreme-UV may be reasonable up toλ≃800˚A (Draine2003).IGM attenuation has a large dispersion for different lines of sight.We show colors with a median value of IGM attenuation for z=3.0and z=3.2cases, with the optical depth range containing68%of all sight-lines of Monte-Carlo simulations(Inoue&Iwata2008). The shaded area shows colors of model galaxies which can be explained with dust and/or IGM attenuation. The dashed line in the area indicates colors with IGM opacity with1%probability of occurrence(i.e.,very rare transparent sight-lines);99%of objects with the bluest SED with the Salpeter IMF at z∼3should show redder NB359−R colors than this dashed line.However,all but one object of the detected sample show bluer NB359−R colors.The V−i′colors of the detected LBGs suggest moder-ate attenuation by dust,E(B−V)∼0.1–0.3,consistent with previous studies(e.g.,Iwata,Inoue&Burgarella4Iwata et al.Fig. 3.—NB359−R and V−i′colors of objects detected inNB359.LBGs detected in NB359are shown with squares,and twoLAE sub-groups divided by their UV slopes are shown with opencircles and triangles.Solid line is the color track of a model galaxywith the bluest SED with the Salpeter IMF from z=3.0to3.3.The shaded area indicates a color range which can be explainedwith attenuation by dust and IGM.See text for details.The crossindicates the expected color of a model QSO with a folded SED(α=−0.6and−0.2forλ<1100˚A andλ≥1100˚A,respectively.fν∝να),without IGM attenuaion.2005).However,the NB359−R colors of the detectedLBGs are difficult to explain with these models.Even ifwe assume no IGM attenuation for these objects,colorsof four LBGs bluest in NB359−R cannot be explained.A QSO-like spectrum may be able to explain these col-ors(see a cross shown in thefigure),although these ob-jects are not classified as AGNs,according to the catalog (Steidel et al.2003).To explain these strong emissionin Lyman continuum,models which can produce muchbluer intrinsic colors–such as those with a top-heavyIMF–would be required.Indeed,model SEDs witha top-heavy IMF with a slope1.0(much steeper thanthe slope of the Salpeter IMF,2.35)have∼0.3magbluer color in NB359−R.Alternatively,dust attenuationlaw might be quite different from that by Calzetti et al. (2000).For example,if the amount of dust attenuationat900˚A is much smaller than that at1500˚A,the dust attenuation arrow approaches vertical in Figure3,andthe NB359−R and V−i′colors of the LBGs may be explained simultaneously.Colors of the detected LAEs are much harder to ex-plain with the models considered here.Deep spec-troscopy with wide wavelength coverage is required toconfirm their redshifts and explore their nature.In the following discussion for constraints on f esc,we only usethe detected LBG sample.4.2.Constraints on the Escape FractionThe relative escape fraction is defined as(Steidel et al. 2001):(f1500/f900)intf esc,rel=Lyman Continuum from z=3Galaxies5Matsuda,Y.,et al.2004,AJ,128,569(M04)Matsuda,Y.,Yamada,T.,Hayashino,T.,Yamauchi,R.,& Nakamura,Y.2006,ApJ,640,L123Miyazaki,S.,et al.2002,PASJ,54,833Ouchi,M.,et al.2004,ApJ,611,660Razoumov,A.O.,&Sommer-Larsen,J.2007,ApJ668,674 Salpeter E.E.1955,ApJ,121,161Sawicki,M.,&Thompson,D.2006,ApJ642,653Schlegel,D.J.,Finkbeiner,D.P.,&Davis,M.1998,ApJ500,525 Shapley,A.E.,Steidel,C.C.,Pettini,M.,Adelberger,K.L.,& Erb,D.K.2006,ApJ,651,688(S06)Siana,B.,et al.2007,ApJ,668,62Steidel,C.C.,Adelberger,K.L.,Dickinson,M.,Giavalisco,M., Pettini,M.,&Kellogg,M.1998,ApJ,492,428Steidel, C. C.,Adelberger,K.L.,Shapley, A. E.,Pettini,M., Dickinson,M.,&Giavalisco,M.2000,ApJ,532,170Steidel,C.C.,Pettini,M.,&Adelberger,K.L.2001,ApJ,546, 665Steidel, C. C.,Adelberger,K.L.,Shapley, A. E.,Pettini,M., Dickinson,M.,&Giavalisco,M.2003,ApJ,592,728 Williams,R.E.,et al.1996,AJ,112,1335Yagi,M.,et al.2002,AJ,123,66。