Astronomy & Astrophysics manuscript no. 4309dbur (cid:13)c ESO 2008 February 5, 2008 Ultraviolet-to-Far Infrared Properties of Lyman Break Galaxies ∼ 1 and Luminous Infrared Galaxies at z 1 2 2 2 1 D. Burgarella , Pablo G. P´erez-Gonz´alez , Krystal D. Tyler , George H. Rieke , V´eronique Buat , Tsutomu T. Takeuchi1,10, S´ebastien Lauger1, St´ephane Arnouts1, Olivier Ilbert3, Tom A. Barlow4, Luciana Bianchi5, Young-Wook Lee6, Barry F. Madore7,8, Roger F. Malina1, Alex S. Szalay9 and Sukyoung K. Yi6 6 1 ObservatoireAstronomiqueMarseilleProvence,Laboratoired’AstrophysiquedeMarseille,traversedusiphon,13376Marseille 0 cedex 12, France 0 2 e-mail: denis.burgarella@; veronique.buat@;tsutomu.takeuchi@; [email protected] 2 Steward Observatory,University of Arizona, 933 North Cherry Avenue,Tucson, AZ 85721, USA n e-mail: pgperez@; ktyler@; [email protected] a 3 Osservatorio Astronomico diBologna, via Ranzani, 1 - 47 Bologna - Italy e-mail: [email protected] J 4 California Institute of Technology, MC 405-47, 1200 E. California Boulevard, Pasadena, CA 91125, USA e-mail: 6 [email protected] 5 CenterforAstrophysicalSciences,TheJohnsHopkinsUniversity,3400NorthCharlesSt.,Baltimore,MD21218,USAe-mail: 1 v [email protected] 3 6 Center for SpaceAstrophysics, Yonsei University,Seoul 120-749, Korea e-mail: ywlee@;[email protected] 2 7 Observatories of the Carnegie Institution of Washington, 813 Santa Barbara Street, Pasadena 91101, USA e-mail: 1 [email protected] 1 8 NASA/IPACExtragalactic Database, California Institute of Technology, Mail Code 100-22, 770 S. Wilson Ave., Pasadena, 0 CA 91125, USA 6 9 Department of Physics and Astronomy, Johns Hopkins University, 3400 North Charles Street, Baltimore, MD 21218, USA 0 e-mail: [email protected] h/ 10 Present address: Astronomical Institute, Tohoku University Aoba, Aramaki, Aoba-ku, Sendai 980-8578, Japan e-mail: p [email protected] - o Received ; accepted r t s a ABSTRACT : v Aims. Wepresentthefirst large, unbiased sample of LymanBreak Galaxies (LBGs) at z∼1. Farultraviolet-dropout (1530 ˚A) i X galaxies intheChandraDeepField Southhavebeenselected usingGALEX data.Thisfirstlargesampleinthez∼1universe r providesus with a high quality reference sample of LBGs. a Methods. Weanalyzed the sample from theUV tothe IR usingGALEX, SPITZER,ESO and HST data. Results. The morphology (obtained from GOODS data) of 75 % of our LBGs is consistent with a disk. The vast majority of LBGs with an IR detection are also Luminous Infrared Galaxies (LIRGs). As a class, the galaxies not detected at 24µm are an order of magnitudefainter relative tothe UVcompared with those detected individually,suggesting that there may be two types of behavior within the sample. For the IR-bright galaxies, there is an apparent upper limit for the UV dust attenuation and this upper limit is anti-correlated with the observed UV luminosity. Previous estimates of dust attenuations based on the ultraviolet slopeare compared tonew onesbased on theFIR/UVratio(for LBGs detected at 24µm),which is usually amore reliable estimator. Depending on the calibration we use to estimate the total IR luminosity, β-based attenuations AFUV are larger by 0.2 to 0.6 mag. than the ones estimated from FIR/UV ratio. Finally, for IR-bright LBGs, median estimated β-based SFRsare2-3timeslargerthanthetotalSFRsestimatedasSFRTOT =SFRUV +SFRIR whileIR-basedSFRsprovidevalues below SFRTOT by 15 - 20 %. We use a stacking method to statistically constrain the 24µm flux of LBGs non individually detected. Theresults suggest that theseLBGs do not contain large amounts of dust. Key words.cosmology: observations – galaxies : starburst – ultraviolet : galaxies – infrared : galaxies – galaxies : extinction 1. Introduction LymanBreakGalaxies(LBGs)arethemostnumerousob- Send offprint requests to: D. Burgarella jectsobservedathighredshift(z >2−3)intherest-frame 2 Burgarella D.et al.: A Multi-λ Analysis of Lyman Break Galaxies at z ∼ 1 ultraviolet (UV). The discovery of a large population of forthedustattenuationtorecoverthefullStarFormation LBGs beginning with the work of Madau et al. (1996), Density (e.g. P´erez-Gonza´lezet al. 2005). followed by the spectral confirmation of their redshifts, The usual way of detecting and identifying LBGs is provided the astronomicalcommunity with the first large through the so-called dropout technique, i.e. the absence sample of confirmed high redshift galaxies (Steidel et al. of emission in the bluest of a series of bands due to the 1996;Lowenthalet al. 1997).The spectra of bright LBGs Lyman break feature moving redwards with the redshift (e.g. cB58 by Pettini et al. 2000; Teplitz et al. 2000) are (e.g.Steidel&Hamilton1993,Giavalisco2002).However, remarkably similar to those of local starbursts, indicat- it is well known that selection effects can have a very ing that these objects are forming stars at a high rate. strong influence on the deduced characterics of an ob- The observed colors of LBGs are redder than expected served galaxy sample (e.g. Buat et al. 2005; Burgarella fordust-freestar-formingobjects.Thisreddeningsuggests et al. 2005). Until now, there has been no way to detect thatsomedustispresentinthispopulation.However,the a general, unbiased sample of LBGs at low redshift (i.e. amount of dust in LBGs (Baker et al. 2001; Chapman z ≤ 2), with the same dropout method very successfully et al. 2000), and therefore the reddening-corrrected star used at z ≥ 2 because we lacked an efficient observing formation rate (SFRc), is poorly known. Meurer et al. facility in the UV range. This was quite unfortunate be- (1999), Adelberger & Steidel (2000) and subsequent pa- cause the simple fact that the galaxies are closer to us perstriedtoestimatetheamountofdustattenuationfrom means that we can access much more information on the the β method (Calzetti et al. 1994).However,it has been morphology, detect fainter LBGs in the UV and in the shownrecentlythatthis approachonlyprovidesroughes- IR, and therefore harvest larger samples. GALEX and timates of the total UV attenuation in localgalaxies (e.g. SPITZERchangedthis situationandhaveallowedus to Buatetal.2005;Burgarellaetal.2005;Seibertetal.2005, define a large sample of LBGs at z ∼ 1 in the Chandra Goldader et al. 2002 and Bell 2002). Deep Field South (CDFS). In this study, we combine the detection in the UV of High redshift LBGs are mainly undetected at sub- true (i.e. with a detected Lyman break) LBGs at z ∼ 1 millimeter (sub-mm) wavelengths, where the emission of withGALEX andat24µmwithSPITZER/MIPS.These galaxies is dominated by the dust heated by young stars dataletusestimatethetotaldustemissionandtherefore, (Kennicutt 1998). Only the most extinguished LBGs are the dust-to-UV flux ratio, which provides a good tracer detected by SCUBA (Chapman et al. 2000, Ivison et al. of the dust attenuation in the UV. We also use high spa- 2005) and we have no idea of the dust attenuation for a tialresolutionimagestoanalysetheirmorphology.Weare representative sample. Very recently, Huang et al. (2005) therefore able to perform a complete analysis for the first observed a population of LBGs at 2 < z < 3 detected time on a large LBG sample. withSPITZER.Unfortunately,duetotheveryhighred- −1 −1 shift, the SPITZER/MIPS observations were not deep A cosmology with H0 =70 km.s .Mpc , ΩM =0.3 and ΩΛ =0.7 is assumed in this paper. enough to detect the thermally reradiated emission from very many LBGs at z ∼ 3 and the SPITZER/IRAC data, although deep enough to detect many LBGs, only 2. The Galaxy Sample probe the rest-frame NIR (i.e. no information about the GALEX (Martin et al. 2005) observed the CDFS field dustenshroudedstarformationcanbeinferred).Themor- for 44668 sec (Deep Imaging Survey = DIS) in both the phology of LBGs is also a matter of debate: early works far ultraviolet (FUV) and the near ultraviolet (NUV). (e.g. Giavalisco et al. 1996) suggested that LBGs could The GALEX field is centered at α = 03h32m30.7s, δ be ellipsoidal,andthereforeperhapsthe progenitorsofel- = -27deg52’16.9” (J2000.0). The GALEX IR1.1 pipeline lipticals or of the bulges of spiral galaxies. The problem identified 34610objects within the 1.25o-diameter field of is that we can hardly detect low surface brightness areas view. The GALEX resolution (image full width at half athighredshiftbecauseofthe cosmologicaldimming.For Maximum = FWHM) is about 4.5 arcsec in FUV and 6 instance, Burgarella et al. (2001) suggest that only com- arcsec in NUV. pactstar forming regionscould be easily detected in deep This field has also been observed by SPITZER using HST observations. MIPS (Rieke et al. 2004) in the guaranteed time observ- On the other hand, observations in the sub-mm range ing program. The MIPS observations provide about 7 - 8 have revealed a population of FIR-bright galaxies that sources arcmin−2 at 24 µm centered on a ∼ 1.45×0.4= might be similar to local (Ultra) Luminous IR galaxies 0.6 deg2 field of view. The SPITZER image FWHM is ((U)LIRGs) (Blain et al. 1999) with 1011L⊙ < LIR = about 6 arcsec and almost perfectly matches that of L(8−1000µm)<1012L⊙forLIRGsand1012L⊙ <LIR < GALEX. 1013L⊙ forULIRGs.Theseobjectsarelikelytodominate Redshifts from GOODS (Vanzella et al. 2005) and the Cosmic InfraredBackground(CIB; Elbaz& Cesarsky VVDS (Le F`evre et al. 2004) are available at the cen- 2003)athighredshift.ThelinkbetweenLBGsandLIRGs ter of the GALEX field. Part of the GALEX CDFS is still an open question: are there two classes of objects field was observed by COMBO 17 (Wolf et al. 2004) or are they related? If they are two facets of the same over 0.5 × 0.5 deg2. We made use of COMBO 17 red- population,thenwecould,forinstance,correctUVfluxes shifts for objects with r<24.5. In this range, the quality Burgarella D.et al.: A Multi-λ Analysis of Lyman Break Galaxies at z ∼ 1 3 δz/(1+z) remains within σ < 0.03 for 53 % of the ob- z jects.Finally,weobtainedphotometryfromtheEuropean Southern Observatory Imaging Survey (EIS) in U, B, V, R and I. WebuiltasampleofLBGsasfollows.Fromthesources with redshifts, we selected objects in the rest-frame FUV (i.e.intheobservedNUV)thatwecross-correlated(r =1 arcsec)firstwiththe ground-basedopticalEISdata,then (r =4arcsecs)withtheMIPS24µmdata.Intheresulting catalogue, we down-selected to objects that COMBO 17 puts in the ”GALAXY” class. Then, we extracted the sources with redshift 0.9 ≤ z ≤ 1.3. Since we wish to study Lyman break galaxies, we omitted objects without observed NUV and U flux : we kept only objects down toGALEXNUV magnitudes=24.5correspondingtothe GALEX 80% completeness level and the U-band limiting Fig.2. The distribution in observedluminosity presented magnitude at U=25.1. We did not use a color-color se- herecorrespondstoUVluminositiesinblue(solid)andIR lection as performed by Steidel & Hamilton (1993). Since luminosities in red (dashed). Heavy lines are drawn from we are studying galaxies with known redshifts, the extra the populationofLBGswithdetectedcounterpartsinthe color is not needed to screen out interlopers. The selec- 24µm MIPS image and thin lines to upper limits only. tiononthex-axis(i.e.G−Rcolor)mightbiasthesample Note that the cut at low LIR is not sharp because the toward low-reddening LBGs that we wish to avoid so we 83µJy limit used here is not the detection limit but the can achieve a more general understanding of the sample. 80 % completeness limit. UV luminosities cover the same Infact,wefindthatthemembersofoursamplefallwithin rangeforthetwosamplesandreachuncorrectedUVlumi- the traditional color-color LBG range, or very close to it, nosities of LogLUV =11(inL⊙). This upper limit is con- asdiscussedinSection3.5.Forgalaxiesatz ∼1,GALEX sistentwithAdelberger&SteidelrangeforLBGsatz ∼3 FUV corresponds to a rest-frame wavelength of ∼ 765 but seems inconsistent with the Balmer-break sample at ˚A and GALEX NUV corresponds to ∼ 1155 ˚A. The ob- z ∼ 1. Two areas are shaded. Starting from the left, the servedFUVandNUVfiltersarethereforeinthesamerest- first corresponds to the range covered by UV Luminous framewavelengthrangesasU andGfiltersusedtoidentify Galaxies (open right ended) and the second one corre- Lyman breaks at high redshift (e.g. Giavalisco 2002 for a sponds to Luminous IR Galaxies. A few Ultra Luminous review). The observed FUV-NUV color thus gives a clear IR Galaxies are also detected. indicationoftheLymanbreak.Wepickedtheobjectswith the strongest indication of a break : FUV −NUV > 2; theirmorphologiesandtheirstarformationasrevealedby the final sample contains 297 LBGs. Of this list, 49 ob- the UV and IR data and finally discuss the implications jects (16.5 %) have a measured flux above the 80 % com- they have for studies centered on higher redshifts. pleteness limit ofthe SPITZER24 µm data (e.g. P´erez- Gonz´alez et al. 2005) and thus at a high enough ratio of signaltonoisetobetreatedindividually.Fig.1showstwo 3.1. Ultraviolet and Infrared Luminosities examples of these LBGs at several wavelengths including the two GALEX bands, the B and R EIS bands and the We find a large range of observed (i.e. un-corrected for 24µm SPITZER/MIPS band. Many additional galax- dust attenuation) FUV luminosities (λfλ in rest-frame ies were detected by SPITZER but at a weaker level; FUV) with 9.3 ≤ LogLUV[L⊙] ≤ 11.0. The lowest lu- below we will describe how we used a stacking technique minosity is set by the limiting magnitude in the U-band to probe their properties. Our sample of 297 UV-selected at U = 25.1. Below the break, the limiting magnitude LBGs in the redshift range 0.9 ≤ z ≤ 1.3 constitutes amounts to FUV = 26.0. Although fainter objects are the database that we will study in this work (except for detected in the NUV (down to NUV = 25.9), we use Sect. 3.5). The common GALEX - SPITZER/MIPS - a limiting magnitude of NUV = 24.5 to compute safer COMBO17 field ofview (mainly limited by COMBO17) FUV −NUV colorsandthereforeperformasaferLyman isabout0.25deg2,whichtranslatesto∼1180LBGsdeg−2 Break selection. The averagevalue is <LogLUV[L⊙]>= and ∼ 200 LBGs deg−2 for which we have individual IR 10.2±0.3forthesamplewithindividualIRdetections,and detections. <LogLUV[L⊙]>=10.1±0.3 for the rest of the sample. Total IR luminosities (L ) are estimated following IR the procedure described in P´erez-Gonza´lez et al. (2005). 3. Lyman Break Galaxies at z ∼ 1 Briefly, rest-frame 12 µm fluxes are calculated by com- Our LBGs provide the first opportunity to study an un- paring the observed mid-IR SEDs (including IRAC and biased sample of LBGs systematically at z ∼ 1. We ana- MIPS fluxes) of each individual galaxy with models of lyzetheUVandIRluminositiesofthesegalaxies,measure dust emission (e.g. Chary & Elbaz 2001 or Dale et al. 4 Burgarella D.et al.: A Multi-λ Analysis of Lyman Break Galaxies at z ∼ 1 Fig.1. Two galaxies from our LBG sample are shown here, from left to right in GALEX FUV and NUV, then EIS B, HST GOODS B, EIS I, HST GOODS I and finally SPITZER/MIPS 24µm band. For the two galaxies, the leftmost image (blue frame) is below the Lyman break at z ∼ 1 and the galaxy is not visible. The size of GOODS images is 4 × 4 arcsec2. The corresponding GOODS field (green frame) is plotted in the large (35 × 35 arcsec2) EIS images. a) a LBG classified as a disk-dominated galaxy; the more compact object is at z = 0.546 from VVDS, it appears reddish and should not contribute in ultraviolet and in far infrared; b) a LBG classified as a merger / interacting galaxy. 2002). This procedure is meant to cope with the strong to Log(LIR) ≥ 11L⊙, that is, nearly all the IR-detected K-corrections observed in the mid-IR due to the emis- LBGs are LIRGs (95.9 %) and there are 2 ULIRGs (4.1 sion from aromatic molecules. We use the formulations of %). An association between UVLGs and LIRGs was pro- Takeuchiet al. (2005)and Chary & Elbaz (2001)for con- posed by Burgarella et al. (2005). We confirm here this versionfrom12µmflux density to L .Inadditionto the association and extend it to LBGs. IR intrinsic differences due to the two calibrations (the for- mer provides L lower by 0.2 dex), the conversion from IR 3.2. The Morphology L12µm to LIR can introduce errors up to a factor of 2 for individual normal galaxies and 4 for galaxies with SED The Great Observatories Origins Deep Survey (GOODS) variations over the full IRAS sample range (Takeuchi et provides high resolution and high signal-to-noise images al. 2005; Dale et al. 2005). The effects of these errors are of some of our LBGs, which can be used to study their greatly reduced in this work because we discuss average morphology in the rest-frame B band. An advantage of properties, not those of individual galaxies; in this case, our low redshift sample of LBGs is that the images ex- theuncertaintiesarelikelytobecommensuratewiththose tend to low surface brightness and hence morphologies in the overall conversions to LIR (i.e., still ∼ 0.2 dex). can be determined well. We compute the asymmetry and There are also uncertainties in the luminosity values due concentration (Fig. 3) as in Lauger et al. (2005a) from to the distance (since we use photometric redshifts with the objects within the GOODS field which have a signal- δz/(z+1) ≤ 0.03), but they are less than 10% for LUV to-noise ratio larger than S/N ≈ 1 per pixel and whose and LIR. coordinates are within 2 arcsecs from the GALEX detec- Itisdifficulttocompareoursampletopreviouslypub- tion. We were able to obtain the morphology for only 36 lished ones since none was available in this redshift range LBGS out of our 300 LBGs (about 1/4 of our GALEX before GALEX. However, Adelberger & Steidel (2000) + SPITZER + COMBO 17 field is covered by GOODS). have discussed a Balmer-break sample at z ∼ 1. The dis- Fig. 3 leads us to two conclusions: i) all but one LBG in tributions in L and L for our sample are shown in our sample are located on the disk side of the line sepa- UV IR Fig. 2. The lower limit is set by the flux limits but the ratingdisk-dominatedandbulge-dominatedgalaxies,and upper limit of our LBG sample is about the same as the ii) part of them (22 %) are in the top part of the dia- z >3 one in Adelberger & Steidel’s (2000). However, the gram, i.e. with an Asymmetry larger than 0.25 and could upper limit of our sample is higher by a factor of ∼ 4 be interpretedas mergers.This kind of quantitative anal- (assuming H0 = 70) than Adelberger & Steidel’s (2000) ysis is also applied to higher redshift LBGs, however, we sample. mustbecarefulintheinterpretationbecause,evenifdisks Heckman et al. (2005)defined UV Luminous Galaxies are present, it would be very difficult to detect them due (UVLGs) as galaxies with UV luminosities above to the cosmologicaldimming (e.g. Burgarellaet al. 2001). Log(LUV)=10.3L⊙.TheyfoundthattheseUVLGsbear Indeed,indeepspectroscopicobservations(e.g.Moorwood similarities to LBGs, especially a sub-sample of compact et al. 2000,Pettini et al. 2001), the profiles of the optical ones.Inoursample,22.2%ofthe LBGsareUVLGs,30.6 nebular lines suggestthe presence of disks in some LBGs. % of the LBGs with an IR counterpart are UVLGs. The Anumberofstudieshavebeendevotedtogalaxymor- 83µJydetectionlimitat24µmapproximatelycorresponds phology in the redshift range 0.6 ≤ z ≤ 1.2. At z ∼ 0.7, Burgarella D.et al.: A Multi-λ Analysis of Lyman Break Galaxies at z ∼ 1 5 most of the works seem to agree that about 60 - 70% ofthe objectscanbe classifiedasdisk-dominatedgalaxies (mainlyspiralsandMagellanicirregulars)and10-20%as mergers/interactinggalaxies.Herespiralsareasub-group of disks which exhibit a more symmetric (spiral) struc- ture than irregular-like objects similar to the Magellanic clouds. Their asymmetry is therefore lower. Lauger et al. (2005b) found about 70 - 80% of disk-dominated galaxies atz ∼1.Zhengetal.(2004)studiedtheHST morphology of a sample of LIRGs and also found that a large major- ity (∼ 85 %) of them are associated with disk-dominated galaxies.This conclusion is reached whether the selection is in the rest-frame ultraviolet(Wolf etal. 2005)orin the infrared (Bell et al. 2005). However, some dispersion due tothe cosmicvariancemightexist(Conselice,Blackburne & Papovich 2005). Fig.3. The morphology of the LBG sample is quantita- tively estimated from the asymmetry and the concentra- Therefore, overall it appears that the majority of the tion (sub-sample drawn from a larger CDFS analysis by star formation at 0.6 < z < 1 resides in disks and about Lauger et al. 2005b). The smaller number of objects pre- half of it in spirals. The numbers that we draw for our sented here is due to the smaller field of view of GOODS LBG sample at z ∼ 1 are globally consistent: we find as compared to ours. Black circless are LBGs without that ∼22% of the LBGs are likely mergers (e.g. Fig. 1b), IR counterpartswhile reddiamonds representLBGs with ∼ 75% are disks (e.g. Fig. 1a) and only ∼ 3% (i.e. 1 a 24µm MIPS detection. The line corresponding to the galaxy) is possibly a spheroid. In the cases where our limitbetweendisk-dominatedandbulge-dominatedgalax- LBGs canalso be classifiedasLIRGS fromtheir IR lumi- ies (Lauger et al. 2005a).The location in the diagram re- nosity,ourmeasuredmorphologiesareconsistentwiththe flects the morphological type of the galaxies: more asym- LIRG morphology measurements of Zheng et al. (2004). metrical LBGs (e.g. mergers) are in the top part of the diagram (A > 0.25) while early-type spirals would have 3.3. Ultraviolet Dust Attenuations A<0.1.TheLBGsampleismainlydominated(75%)by disk-dominated galaxies and the contribution from merg- Until now, it has been difficult to estimate the validity ers amounts to ∼21 %. of dust attenuation estimates for distant LBGs, because we had no clear idea of their L . Adelberger & Steidel IR (2000) tried to estimate the 800µm fluxes of their LBG There are now studies of high-redshift LBGs (z > 2) sample from the β method and compared the results to with SPITZER (e.g. Labb´e et al. 2005; Huang et al. observations. However, only the most extreme LBGs can 2005). However,the results are so far inconsistent: Labb´e be detected either directly in the sub-millimeter range or et al. (2005) found that LBGs are consistent with low- intheradiorangeat1.4GHz andthereforecouldbeused reddening models while Huang et al. (2005) found more in this comparison. reddened LBGs. Further observations will resolve these In this paper, we use total IR luminosities, L , and differences and provide a firm basis for comparison with IR L to compute the FIR/UV ratio, which is calibrated our sample. UV into FUV dust attenuation A (e.g. Burgarella et al. With the 24µm SPITZER flux for the individually FUV 2005). This method has been shown to provide more ac- detectedz ∼1objects,wecangoastepfurtherandcheck curatedustattenuationsthanthose fromthe UV slopeβ. how UV dust attenuation estimations carried out from Fig. 4 shows an apparent anti-correlation of A with the β method compare with the better IR/UV-based es- FUV the UV luminosity. It is not clear, however, whether this timates. This comparisonhas already been performed for relationship is real or only observational. Indeed, in ad- nearby galaxies (Buat et al. 2005, Burgarella et al. 2005, dition to the observational cut at low L , the 24µm Seibertetal.2005andreferencestherein).Giventhewide FUV lower limiting flux means that we cannot detect individ- use ofthe β method on highredshift LBGs, it is useful to ually low-luminosity galaxies with low dust attenuations. compare with our lower-redshift sample. It is very interesting to note that we do not detect LBGs Using the equations in Adelberger & Steidel (2000) withbothahighUVluminosityandahighUVdustatten- (deduced from Meurer et al. (1999)), we estimate the IR uation,andthis cannotbe causedbyobservationallimits. luminosity that is used to compute A and the total FUV In other words, we seem to observe a population of high luminosity for each LBG. We observe a small overesti- L LBGs (which qualify as UVLGs) with dust attenua- mation of the dust attenuation as compared to the ones UV tions similar to UV-selected galaxies in the local universe estimated from SPITZER/MIPS data and the dust-to- (e.g. Buat et al. 2005). UVLG galaxies are LIRGs with UV flux ratio. The β-based mean dust attenuation esti- the lowest A . mated for our z ∼ 1 LBG sample is A = 2.76±0.13 FUV FUV 6 Burgarella D.et al.: A Multi-λ Analysis of Lyman Break Galaxies at z ∼ 1 3.4. Star Formation Rates and Implications for the Cosmic Star Formation Density We estimate SFRs for our LBG sample from the IR lu- minosities and after applying dust corrections estimated fromβ andwe comparethem, inFig.5,to the totalSFR: SFR =SFR +SFR where SFR is not cor- TOT UV IR UV rected for dust attenuation. SFR is assumed to be TOT the best SFR estimate and we use it as a reference. The first conclusion is that the dispersion is much larger for UV SFRcs computed with β dust correctionsthanfor IR SFRs.Butthemedianvaluesarealsodifferent:SFRc = UV 112.3 ± 33.8 (σ = 260.9) M⊙.yr−1 while SFRIR = 49.9 ± 13.1 (σ = 100.9) M⊙.yr−1 if we use Chary & Elbaz (2001) and SFRIR = 30.5 ± 6.0 (σ = 46.6) M⊙.yr−1 Fig.4. Blue and red symbols are the same objects but if we use Takeuchi, Buat & Burgarella (2005). Median luminosities are LUV for the former and LIR for the lat- SFRTOT for the two above calibrations are, respectiv- ter. The dust attenuation strongly decreases while LUV elly, SFRTOT = 62.8 ± 13.2 (σ = 102.1) M⊙.yr−1 and increases. Part of this apparent correlation might be due SFRTOT =41.1±6.3(σ =48.6)M⊙.yr−1.Asexpected, to the fact that observational limits prevent us from de- for LBGs detected in IR, SFRIR is therefore a better es- tecting low-luminosity LBGs with low dust attenuation. timate. About 22 % of our LBGs with an IR detection The clear cut on the upper parts of the box cloud cannot have SFRTOT > 100 M⊙.yr−1 (using Chary & Elbaz’s be due to observational biases. We do not seem to ob- calibration)ascomparedtolessthan1%inFloresetal.’s serve UVLGs with high dust attenuations. On the other (1999) galaxy sample which confirms that our LBGs are hand, the more dispersed but well-known increase of the forming stars very actively. However, none of the LBGs dust attenuation with LIR is observed here. Larger sym- undetected at 24µm is above SFRTOT >100 M⊙.yr−1. bols correspond to UVLGs (LogLUV >10.3L⊙). Most of The higher SFRs reached when dust attenuations are them havelowdust attenuationsbut one is a ULIRGand computed with the UV slope β (depending on the L IR has A ∼4. calibration,+79to+173%)leadtoanoverestimatedcon- FUV tributionofLBGstotheCosmicStarFormationDensityif the same quantitative difference exists at higher redshift. However,Takeuchi,Buat&Burgarella(2005)showedthat the current assumption of a constant dust attenuation (σ = 1.02) while the dust-to-UV dust attenuation gives does not seem to be verified. The increase of the mean AFUV = 2.16±0.11 (σ = 0.84) with LIR from Takeuchi AFUV from 1.3 to 2.3 from z = 0 to z = 1 means etal.(2005)andAFUV =2.53±0.12(σ =0.94)withLIR that, for a given observed FUV luminosity density, the fromChary& Elbaz(2001).The averagevalue ofthe two dust-corrected star formation density would vary. Note dust-to-UVestimatesisconsistentwithTakeuchi,Buat& that those mean dust attenuations cannot be compared Burgarella(2005).Theneteffectisthattotalluminosities with the numbers given for the Kaplan-Meier estimates basedon the β method are slightly larger than the actual which are biased toward large M LBGs while fainter FUV values and the deduced SFRs are therefore overestimated LBGs seem to have larger dust attenuations in our sam- (see next section). The mean of the ratios of the β vs. ple.AlthoughFIRdataarenotalwaysavailable,itisvery FIR/UV AFUV = 1.31±0.07 (σ = 0.52) for Takeuchi et important that one is aware of these uncertainties when al. (2005) and AFUV = 1.09±0.05 (σ = 0.40) for Chary using Star Formation Densities derived from UV values & Elbaz’s calibration. corrected from the β method for LBGs with high dust Applying the Kaplan-Meier estimator (and using attenuations, especially at high redshift where we have a Chary & Elbaz’s calibrations), we can take upper lim- very poor knowledge of actual attenuations. These ambi- its into account to estimate mean dust attenuations. guities may be reduced by further study of Spitzer data, We find moderate values < A >= 1.36 ± 0.07 for and with Herschel. FUV M ≤ −22 (5 data points of which 1 is an upper FUV limit) and < A >= 1.08± 0.11 for M ≤ −21 FUV FUV 3.5. Extension to Galaxies Faint at 24µm i.e. L at z=3 (35 data points of which 65 % are upper ⋆ limits). Using only detections, we reached, respectivelly, So far, most of our arguments have been based on the < A >= 1.54±0.09 and < A >= 1.59±0.18. 49 galaxies individually detected at 24µm well above the FUV FUV Since we only use a small number of bright LBGs, this is completeness limit. To put the behavior of these galaxies hardlycomparableto the numbers quotedin the previous inabroadercontext,wehavedeterminedaverageinfrared paragraph. However, it suggests that lower L LBGs fluxdensitiesforgroupsofgalaxiesbystacking.Imagesat FUV have higher dust attenuations (see Fig. 4). 24µm are shifted to a common center on the basis of the Burgarella D.et al.: A Multi-λ Analysis of Lyman Break Galaxies at z ∼ 1 7 and 1.1×1044 ergs/s, respectively. However, the average infrared flux densities differ by a factor of ten: 13µJy for the undetected group and 143µJy for the detected one. There is no significant difference in average infrared flux density among the subgroups in the detected group. These results indicate that the LBGs divide into two classes. About 40% of them are infrared bright. The av- erage NUV flux density for this group is 1.63µJy, so as measured in νF , the NUV and 24µm luminosities are ν similar. Since there is a substantial correctionto total far infraredluminosity,theinfraredcomponenttotheoutput fromyoungstarsissignificant,probablyaccountingforthe majority of the luminosity for these objects. The remain- ing 60%are infraredfaint: νF is about ten times greater ν in the NUV than at 24µm, indicating that their outputs Fig.5. Blue open circles compare SFRUcV to SFRTOT aredominatedby the UV.The resultsofthis paperapply whileredfilledboxescompareSFRFIRtoSFRTOT.Both to galaxies like those in the detected group only. SFRs are computed from Kennicutt (1998). The median To explore other possible differences between these SFRFIR is underestimated by about 80 % and the un- classes,wecomputed the averageB (i.e.rest-frameNUV) derestimationincreasesatlowerSFRs,whichisconsistent flux densities for the same two groups and three sub- withthefactthatthe UVcontribution(notaccountedfor groups.Althoughitisinfluencedbyotherfactors,wetake from LIR) is usually higher at low SFRs than at high the ratio of B to NUV flux densities (or equivalently the SFRs. For this sample of LBG detected at 24 µm, the NUV - B color) to be an indicator of the level of redden- median value of SFRUcV is overestimated by a factor of ing, and the ratio of 24µm to B flux density to measure 2 - 3 in average with a possible trend for the difference the relative portion of the luminosity from young stars to increase at high SFRs. The dispersion of β-based UV emerging in the infrared compared with the UV. The re- SFRcs (σSFR ∼ 200M⊙.yr−1) is much larger than IR- sults are in Table 1. First, they demonstrate that all the based SFR (σSFR ∼ 30−40M⊙.yr−1 depending on the galaxiesinour selectionfallin, or close,to the color-color calibration into LIR). LBGzoneasadjustedfromhighztoz∼1(seeGiavalisco 2002). There is a trend for LBGs with a high 24/B ratio to present a high B/NUV, which is consistent with the 24µm coordinates if the object was well enough detected relation for the UV slope β and the FIR/UV ratio found in that band, or the coordinates of the optical identifi- by Meurer et al. (2000) on a sample of local starburst cation otherwise. We use sigma-clipping to eliminate sur- galaxies. An analysis based on detections is required to roundingsources;the levelatwhichclipping occursis ad- check whether Meurer et al.’s law can be applied safely justed empirically to provide the smoothest possible sky to those LBGs while we showed in the previous section image.Thequotedresultsareforalevelof5-σ forsources that it provides dispersed SFRs for the detected sam- detected at 24µm and 4-σ for undetected ones (see be- ple. Finally, the amount of dust attenuation for the un- low), but they are not sensitive to modest adjustments in detected group is very low (A ∼0.5−0.6 for a mean FUV this level.Ingeneral,the resultingbackgroundshaveonly LogL ≈ 10.5) which corresponds to LBGs with the TOT weak structure and, where the sources are fairly bright lowestreddening found by Adelberger andSteidel (2000). in the infrared, the 24µm stacked image is similar to the This very low reddening is consistent with the very blue point spread function of the instrument. We confirmed UV slope β ≈ −2.4. If confirmed, this would mean that that stacking sources of known flux density gave consis- about half of the LBGs do not contain large amounts of tent results. These behaviors validate the procedure. dust. For this study, we useda total of336 sources,selected similarly to those discussed above (but without screening 4. Conclusions for COMBO-17 type classification). We divided the sam- ple into two groups. The first, hereafter the undetected We use multi-wavelength data in the CDFS to define the group, includes 201 UV objects (60% of the total) for first large sample of Lyman Break Galaxies at z ∼ 1; which visual inspection indicated no reliable 24µm detec- GALEX is used to observe the Lyman break. Redshifts tion. The second (40%), hereafter the detected group, is aretakenfromspectraandfromCOMBO17.Quantitative the objects withevidence foraninfrareddetection;it was morphologies (Lauger et al. 2005a) are estimated from in turn divided into three equal subgroups according to high spatial resolution images. Finally, dust attenuations NUV luminosity. As shown in Table 1, the two groups and total luminosities are computed from SPITZER have very similar average NUV flux density, 1.89µJy for measurements at 24µm extrapolated to get the total IR the undetected and 1.63µJy for the detected group. The luminosity. averageNUVluminositiesarealsosimilar,1.3×1044erg/s The main results of this analysis are: 8 Burgarella D.et al.: A Multi-λ Analysis of Lyman Break Galaxies at z ∼ 1 Table 1. Stacking Analysis Results number LUV F(24µm) F(B) F(NUV) F(B)/F(NUV) F(24)/F(B) 1044ergs/sec mJy mJy mJy All detected 135 1.11 0.14 0.0018 0.0016 1.11 80 low UV 45 0.67 0.16 0.0014 0.0012 1.21 113 middle UV 45 0.99 0.16 0.0017 0.0014 1.19 95 high UV 45 1.67 0.14 0.0023 0.0024 0.98 59 Undetected 201 1.33 0.013 0.0015 0.0019 0.79 8.7 1. We detect LBGs in the range 9.3 ≤ LogLFUV[L⊙] ≤ support for construction, operation and science analysis for 11.0, i.e. well into the UVLG class as defined theGALEXmissiondevelopedincooperation withtheCentre by Heckman et al. (2005). For the same objects, Nationald’EtudesSpatialesofFranceandtheKoreanMinistry LogLIR ≥ 11L⊙, which means that almost all the ofScienceandTechnology.Finally,wethankEmericLeFloc’h for his help and discussions during thiswork. LBGs with a SPITZER confirmed detection are LIRGs (and 1 ULIRG) at z ∼1. 2. LBGsatz ∼1aremainlydisk-dominatedgalaxies(75 References %) with a small contribution of interacting/merging Adelberger K.L., Steidel C.C. 2000, ApJ, 544, 218 galaxies (22 %) and a negligible (3 %) fraction are BakerA.J.,LutzD.,GenzelR.,TacconiL.J.,LehnertM.D., spheroids. The morphologies of our LBG sample are 2001 A&A,372, 37 consistent with star-forming galaxies. Bell E. F. 2002 ApJ, 577, 150 3. The FUV dust attenuation appears to be anti- Bell E. F., Papovich C., Wolf C., LeFloc’h E., Caldwell J. A. correlated with the observed FUV luminosity. Part of R., Barden M., Egami E., McIntosh D. H., Meisenheimer this correlation might be due to observational limits K., Prez-Gonzlez P.G. et al. 2005 ApJ, 625, 23 at 24 µm. However, non-detection of LBG with high BlainA.W.,SmailI.,IvisonR.J.,KneibJ.-P.,1999MNRAS, L and high A cannot be explained by obser- 302, 632 FUV FUV vational biases. BuatV.,Iglesias-PramoJ.,SeibertM.,BurgarellaD.,Charlot 4. About40%ofoursampleofLBGsisdetectedat24µm. S., Martin D. C., Xu C. K., Heckman T. M., Boissier S., Boselli A.et al., 2005 ApJ,619, L51 Theseobjectsalsoshowevidenceforreddeningintheir Burgarella D., Buat V., Donas J., Milliard B., Chapelon S. UVcontinua.Theremaining60%ofthesampleareon 2001 AJ, 369, 421 averageanorderofmagnitude lessluminous inthe in- Burgarella D., Buat V., Small T., Barlow T, A., Boissier S., frared compared with the rest frame near UV. Their Gil de Paz A., Heckman T. M., Madore B. F., Martin D. UV continua also appear to be significantly less red- C., Rich R.M. et al., 2005 MNRAS,360, 1413 dened. Calzetti,D.,Kinney,A.L.,Storchi-BergmannT.1994,ApJ429, 5. Dust corrections of our IR-bright LBG sample com- 582 puted via the β method are over-estimated by ∼ 0.2 ChapmanS.C.,ScottD.,SteidelC.C.,BorysC.,HalpernM., mag. if we use Chary & Elbaz (2001) to compute Morris S. L., Adelberger K. L., Dickinson M., Giavalisco L and by ∼ 0.6 mag. if we use Takeuchi, Buat & M., Pettini M. 2000 MNRAS,319, 318 IR Burgarella (2005). Chary R.& Elbaz D., 2001 ApJ, 556, 562 Conselice C. J., Blackburne, J. A., Papovich C, 2005 ApJ620, 6. Dustattenuationsestimatedbytheβ methodforsuch 564 galaxies lead to overestimation of the SFRs at z ∼ 1 Dale D.A. et al., 2005 ApJin press (astro-ph/0507645) by a factor of 2 to 3, depending on the calibration of Dale D.A., Helou G. 2002 ApJ576, 159 the 24µm flux to L while IR-based SFRs are of the IR Elbaz D.,Cesarsky C.J. 2003 Science 300, 270 same order as SFR . TOT FloresH.,HammerF.,ThuanT.X.,CsarskyC.,DesertF.X., 7. By using the stackingmethod, wefind thatLBGs non OmontA.,Lilly S.J.,Eales S.,Crampton D.,LeF`evreO. detected at 24µm seem to have very low dust attenu- 1999 ApJ517, 148 ations. Giavalisco M., Steidel C. C., Macchetto F. D 1996 ApJ, 470, 189 Acknowledgements. TTT has been supported by the Japan Giavalisco M. 2002 ARA&A,40, 579 Society for the Promotion of Sciences (April2004 - December GoldaderJ.D.,MeurerG.,HeckmanT.M.,SeibertM.,Sanders 2005). Support for this work was provided by NASA through D.B., Calzetti D., Steidel C.C. 2002, ApJ568, 651 ContractNumber960785issuedbyJPL/Caltech.Thisworkis HeckmanT.M.,HoopesC.G.,SeibertM.,MartinD.C.,Salim basedonobservationsmadewiththeSpitzer,SpaceTelescope, S., Rich R. M., Kauffmann G., Charlot S., Barlow T. A., whichisoperatedbytheJetPropulsionLaboratory,California Bianchi L. et al. 2005 ApJ619, L35 Institute of Technology under NASA contract 1407. Analysis HuangJ.-S. et al., 2005 ApJ in press (astro-ph/0507685) ofthedataforthispaperwassupportedbythecontracttothe Ivison R.J., Smail I., Bentz M., Stevens J. A., Menendez- MIPS team, 1255094. We also thank the French Programme Delmestre K., Chapman S. C., Blain A. W. 2005 National Galaxies and the Programme National Cosmologie MNRAS362, 535 fortheirfinancialsupport.WegratefullyacknowledgeNASA’s Kennicutt R.C., 1988 ARA&A36,189 Burgarella D.et al.: A Multi-λ Analysis of Lyman Break Galaxies at z ∼ 1 9 Labb´eI., HuangJ., Franx M., RudnickG., Barmby P., Daddi E., van Dokkum P. G., Fazio G. G., Forster Schreiber N. M., Moorwood A. F. M. et al. 2005 ApJ624, L81 Lauger S., Burgarella D., Buat V.2005a A&A434, 77 Lauger S. et al., 2005b A&A submitted LeF`evreO.,VettolaniG., PaltaniS.,Tresse L.,Zamorani G., Le Brun V., Moreau C., Bottini D.,Maccagni D.,Picat J. P., et al. 2004 A&A428, 1043 LowenthalJ.D.,KooDavidC.,GuzmanR.,GallegoJ.Phillips A.C.,FaberS.M.,VogtN.P.,IllingworthG.D.,Gronwall C. 1997 ApJ481, 673 Madau P., Ferguson H.C., Dickinson M.E., Giavalisco M., Steidel C.C., FruchterA. 1996 MNRAS,283, 1388 Martin D. C., Fanson J., Schiminovich D., Morrissey P., Friedman P. G., Barlow T. A., Conrow T., Grange R., Jelinsky P. N. et al., 2005. ApJ,619, L1 Meurer, G.R., Heckman, T.M., & Calzetti, D. 1999, ApJ 521, 64 Moorwood A.F. M., vanderWerfP.P., CubyJ. G., OlivaE. 2000. A&A,362, 9 P´erez-Gonz´alez, P.G. et al., 2005 ApJ in press, (astro-ph/0505101) Pettini M., Steidel C. C., Adelberger K. L., Dickinson M., Giavalisco M. 2000 ApJ528, 96 PettiniM.,ShapleyA.E.,SteidelC.C.,CubyJ.-G.,Dickinson M., Moorwood A. F. M., Adelberger K. L., Giavalisco M. 2001 ApJ554, 981 Rieke G.H., Young E. T., Engelbracht C. W., Kelly D. M., Low F. J., Haller E. E., Beeman J. W., Gordon K. D., Stansberry J. A., Misselt K. A.et al. 2004 ApJS154, 25 SeibertM.,MartinD.C.,HeckmanT.M.,BuatV.HoopesC., Barlow T., Bianchi L., ByunY.-I.,DonasJ., Forster K.et al. 2005 ApJ, 619, L59 Steidel, C.C. & Hamilton D., 1993 AJ, 105, 201 Steidel, C.C., Giavalisco, M., Pettini, M., Dickinson, M., & Adelberger, K.L. 1996, ApJ 462, L17 TeplitzH.,McLeanI.S.,BecklinE.E.,FigerD.F.,GilbertA. M., Graham J. R., Larkin J. E., Levenson N. A., Wilcox, M. K.ApJ, 533, 65 Takeuchi T.T., Buat V., Iglesias-Paramo J., Boselli A., Burgarella D. 2005 A&A,432, 423 TakeuchiT.T.,BuatV.,&BurgarellaD.2005,A&A,440,L17 Vanzella E., Cristiani S., Dickinson M., Kuntschner H., MoustakasL.A.,NoninoM.,RosatiP.,SternD.,Cesarsky C., Ettori S.,et al. 2005 A&A434, 53 WolfC.,MeisenheimerK.,KleinheinrichM.,BorchA.,DyeS., Gray M., Wisotzki L., Bell E. F., Rix H.-W., Cimatti A., et al. 2004 A&A421, 913 Wolf C., Bell E. F., McIntosh D. H., Rix H.-W., Barden M., Beckwith S. V. W., Borch A., Caldwell J. A. R., Haussler B., HeymansC., al., 2005 ApJ630, 771 ZhengX.Z.,HammerF., FloresH.,Assmat F.,Pelat D.2004 A&A421, 847