Mid-Infrared Spectroscopy of Lensed Galaxies at 1<z<3: The Nature of Sources Near the MIPS Confusion Limit
J. R. Rigby, D. Marcillac, E. Egami, G. H. Rieke, J. Richard, J.-P. Kneib, D. Fadda, C. N. A. Willmer, C. Borys, P. P. van der Werf, P. G. Perez-Gonzalez, K. K. Knudsen, C. Papovich
aa r X i v : . [ a s t r o - ph ] N ov Resumbitted, 6 Sept 2007
Preprint typeset using L A TEX style emulateapj v. 03/07/07
MID-INFRARED SPECTROSCOPY OF LENSED GALAXIES AT 1 < Z <
3: THE NATURE OF SOURCESNEAR THE MIPS CONFUSION LIMIT
J. R. Rigby , D. Marcillac , E. Egami , G. H. Rieke , J. Richard , J.-P. Kneib , D. Fadda , C. N. A. Willmer ,C. Borys , P. P. van der Werf , P. G. P´erez-Gonz´alez , K. K. Knudsen , and C. Papovich Resumbitted, 6 Sept 2007
ABSTRACTWe present Spitzer/IRS mid-infrared spectra for 15 gravitationally lensed, 24 µ m–selected galaxies,and combine the results with 4 additional very faint galaxies with IRS spectra in the literature.The median intrinsic 24 µ m flux density of the sample is 130 µ Jy, enabling a systematic survey ofthe spectral properties of the very faint 24 µ m sources that dominate the number counts of Spitzercosmological surveys. Six of the 19 galaxy spectra (32%) show the strong mid-IR continuua expectedof AGN; X-ray detections confirm the presence of AGN in three of these cases, and reveal AGNs in twoother galaxies. These results suggest that nuclear accretion may contribute more flux to faint 24 µ m–selected samples than previously assumed. Almost all the spectra show some aromatic (PAH) emissionfeatures; the measured aromatic flux ratios do not show evolution from z = 0. In particular, the highS/N mid-IR spectrum of SMM J163554.2+661225 agrees remarkably well with low–redshift, lower–luminosity templates. We compare the rest-frame 8 µ m and total infrared luminosities of star–forminggalaxies, and find that the behavior of this ratio with total IR luminosity has evolved modestly fromz=2 to z=0. Since the high aromatic–to–continuum flux ratios in these galaxies rule out a dominantcontribution by AGN, this finding implies systematic evolution in the structure and/or metallicity ofinfrared sources with redshift. It also has implications for the estimates of star forming rates inferredfrom 24 µ m measurements, in the sense that at z ∼
2, a given observed frame 24 µ m luminositycorresponds to a lower bolometric luminosity than would be inferred from low-redshift templates ofsimilar luminosity at the corresponding rest wavelength. Subject headings: galaxies—infrared: galaxies: high-redshift—galaxies: evolution INTRODUCTIONThe Spitzer Space Telescope (Werner et al. 2004) hasbeen tremendously successful at detecting star–forminggalaxies and active galactic nuclei (AGN) at z > µ m band. Downto the confusion limit of ∼ µ Jy at 24 µ m (Dole etal. 2004), the MIPS instrument (Rieke et al. 2004) candetect luminous infrared galaxies (LIRGs) out to z ∼ out to z ∼
3, and hyper-luminous infrared galaxies (HLIRGs) outto even higher redshifts.Detectability at z > . µ m, that passthrough the 24 µ m band. Indeed, photometric redshiftsplace ∼ µ m sources at redshifts Electronic address: [email protected] Steward Observatory, University of Arizona, 933 N. CherryAve., Tucson, AZ 85721 Current address: Observatories, Carnegie Institution of Wash-ington, 813 Santa Barbara St., Pasadena, CA 91101 Spitzer Fellow Department of Astronomy, Caltech, 1200 E. California Blvd,Pasadena, CA 91125 Laboratoire d’Astrophysique de Marseille Leiden Observatory, Leiden University, P. O. Box 9513, NL-2300 RA Leiden, The Netherlands Departamento de Astrof´ısica y CC. de la Atm´osfera, Facul-tad de CC. F´ısicas, Universidad Complutense de Madrid, 28040Madrid, Spain Max-Planck-Institut f¨ur Astronomie, K¨onigstuhl 17, D-69117Heidelberg, Germany Defined as 11 < log L ( T IR ) L ⊙ <
12, where L(TIR) is the totalinfrared luminosity between 8 and 1000 µ m. Defined as 12 < log L ( T IR ) <
13 L ⊙ . Defined as log L ( T IR ) >
13 L ⊙ . above 1.4 (Le Floc’h et al. 2005; P´erez-Gonz´alez et al.2005; Caputi et al. 2006; Wang 2006).For extremely high–luminosity galaxies, spectra havebeen obtained with the Infrared Spectrograph (IRS,Houck et al. 2004), e.g. Houck et al. (2005); Yan etal. (2007); Men´endez-Delmestre et al. (2007). However,since spectral properties depend strongly on luminosity,results obtained for hyper-luminous galaxies may havelimited applicability for the bulk of the IR–detected pop-ulation. Spectra of lower–luminosity galaxies have beenobtained only in a few cases (e.g. Lutz et al. 2005; Teplitzet al. 2007).Numerous authors (e.g. Caputi et al. 2006; Choi et al.2006; Reddy et al. 2006) have used the observed 24 µ mband as an estimator of star formation rates. Since mostof the infrared power is radiated redward of the observedband, the 24 µ m diagnostic must be calibrated usinglow–redshift templates (e.g. Chary & Elbaz 2001; Dale& Helou 2002; Lagache et al. 2004; Brandl et al. 2006;Armus et al. 2007). The central assumption in theseworks is that the spectra and spectral energy distribu-tions of high-z galaxies are matched accurately by thelow–redshift templates. For 0 . < z <
3, the 24 µ m bandfluxes are strongly influenced by aromatic band emission.Evolution over time either in the behavior of these aro-matic bands, in metallicity, or in the geometry and ra-diative transfer within these infrared sources, could allundermine this central assumption.To probe these possibilities, we have been obtainingIRS spectra for intrinsically faint 24 µ m–selected galax-ies at 1 . z .
3. Our targets are strongly lensed by thegravitational potential of foreground clusters of galax- Rigby et al.ies, such that their fluxes are amplified by factors of 3–25. Though these galaxies have observed 24 µ m fluxdensities of ∼ ν (24 µ m) = 0 .
82 mJy, but the median intrinsicflux density is nearly an order of magnitude lower. Wecompare their aromatic feature flux ratios, and the ra-tios of their X-ray, aromatic, and total infrared luminosi-ties to those of low–redshift LIRGs and ULIRG samples.We examine in detail the high–quality spectrum of SMMJ163554.2+661225, a LIRG at z = 2 . µ m galaxy population.We assume Ω m = 0 . Λ = 0 . =72 km s − Mpc − (Spergel et al. 2003; Freedman et al.2001). SAMPLE SELECTIONOur goal was to select sources with intrinsic flux den-sities close to, or even below, the MIPS 24 µ m confusionlimit. We first obtained deep 24 µ m images of galaxyclusters, generally identified as being massive throughhigh X-ray luminosity. From these 24 µ m images, weselected sources within 1.5 ′ of the cluster center that: • had observed 24 µ m flux densities above 0.4 mJy(to make efficient use of IRS). • had faint, irregular, or arc–like HST counterparts.This excluded galaxies at the cluster redshift. • had known spectroscopic or probable photometricredshifts & . • had lensing amplification factors calculated to begreater than 3. This restricted selection to clus-ters with good lensing models. Since amplifica-tion depends on the source redshift, if the red-shift was unknown, we calculated the amplificationfor 1 < z < f ν (24 µ m)= 0.13–0.2 mJy; these areunlensed galaxies at 1 < z < A2219a and A2261a lack HST imagery; we instead usedground-based F606W images from the Steward Observatory 90 inchtelescope. Both galaxies are known sub-mm sources (see Table 1),and can thus be added to the sample despite the lack of high–resolution optical images. also use the lensed z = 2 .
81 sub-mm source in A370 fromLutz et al. (2005) with f ν (24 µ m)=1.36 mJy.Source coordinates, redshifts, amplifications, and fluxdensities or our sample and for the four literature sourcesare all listed in Tables 1 and Table 3. OBSERVATIONS AND DATA REDUCTIONThis paper is based on a combination of imaging withMIPS, spectroscopy with IRS, archival imaging fromChandra, and submillimeter photometry from the liter-ature. Exposure times are summarized in Table 1.3.1.
Spitzer/IRS spectroscopy
Low-resolution IRS spectra were obtained as part ofSpitzer GTO programs 82 and 30775 (PI G. Rieke).Data reduction used a package developed by D. Fadda,which has already been successfully used to reduce low-resolution IRS spectra of faint high-z sources (Yan etal. 2007). Residual background, rogue pixels (pixelswith dark current values abnormally high and variablewith time) and cosmic ray hits have to be corrected toobtain an optimal reduction of the data. The packageconsiders all the bidimensional frames produced by theIRS/SSC pipeline. First, background and noise imagesare produced by coadding frames after masking targetand serendipitous spectra on each frame. (The back-ground is better estimated for the PID 30775 observa-tions where both LL1 and LL2 orders provide redundantbackground measurements, compared to the PID 82 ob-servations using only LL1, where the background canonly be determined by differencing nods and thus suffersfrom contamination by faint sources.)A robust statistical estimator (biweight) is used to min-imize the effects of deviant pixels on the coadded value.This part is iterative in order to allow manual identifi-cation of sources. Rogue pixels are then identified bycomputing the dispersion of the noise around every pixeland flagging pixels which are > σ deviant from the meanlocal value. Spectral extraction is done optimally (e.g.,by using the PSF profile to weight the spectrum), takinginto account the spectral distortion, and also rejectingpixels affected by cosmic rays. Finally, the spectra ob-tained at each position are co-added.We downloaded and reduced one archival IRS spec-trum, from program 3453 (PI P. van der Werf). Thisis SMM J163555.2+661238, image “A” in Kneib et al.(2004), one of three images of the same triply–imagedsub-mm galaxy in Abell 2218. We obtained an IRSspectrum of the brightest image of this galaxy, SMMJ163554.2+661225 (image “B” in Kneib et al. 2004);its 24 µ m flux is 1.8 times brighter than the SMMJ163555.2+661238. In the archival spectrum, the sourcewas mis-centered across the short axis of the slit, whichreduced throughput. Because its signal-to-noise ratio islow, we use the archival spectrum only to confirm thedetection of features, and use the spectrum from our pro-gram for all analysis.The IRS spectra are presented in Figure 2. Spectraobtained in programs 82 and 30775 for the same object The two other sources from those papers, “2-x” from Teplitzet al. (2007) and the Abell 2125 source from Lutz et al. (2005),have insufficient signal-to-noise for our analysis. http://ssc.spitzer.caltech.edu/irs/roguepixels/ RS spectra of lensed galaxies 3are plotted separately; observations from the latter pro-gram have superior background subtraction. Flux den-sities are as observed, without any correction for gravi-tational amplification. When redshifts were known fromthe literature, or are evident from aromatic emission fea-tures in the IRS spectrum, then the upper x-axis showsrest wavelength. Only one source (A2261a) lacks a red-shift. Redshifts are listed in Table 1, with a reference ifthe redshift is from the literature.The rest–frame spectra were run through PAHFIT(Smith et al. 2007) to fit the continuum and aro-matic (PAH) features simultaneously. Measured aro-matic fluxes are reported in Table 1.We allowed PAHFIT to fit the silicate absorption (e.g.,we did not specify “NO EXTINCTION”). Fixing the sil-icate optical depth at zero changes the best-fit 7.7 µ mfeature fluxes by 4% (median), the best-fit 8.6 µ m fea-ture fluxes by 8% (median), and the best-fit 11.3 µ mfeature fluxes by 11% (median).We report fluxes in table 1 for features in which themeasured flux is at least twice the uncertainty calculatedfor that line. We add back in a few features, mostly at11.3 µ m, which are obviously detected even though theerrorbars are large.3.2. Spitzer/MIPS 24 µ m and 70 µ m photometry MIPS photometry–mode images at 24 µ m and 70 µ mwere obtained through Spitzer GTO program 83 (PIG. Rieke). Exposure times are given in Table 1 and pho-tometry is given in Table 3. The data were reduced andmosaicked using the Data Analysis Tool (Gordon et al.2006) with a few additional processing steps (Egami et al.2006).Photometry at 24 µ m was obtained by PSF fitting,using the IRAF implementation of the DAOPHOT task allstar . The PSF was created empirically using all avail-able images from program 83. An aperture correction of1.131 was applied, calculated from a Tiny Tim model ofthe 24 µ m Spitzer PSF that extends to r = 220 ′′ (C. En-gelbracht, priv. comm.) Aperture photometry at 70 µ m was performed on the1.0 pixel scale images. The aperture radius was 16 ′′ (twice the half-width at half-max), the background an-nulus had radii of 18 ′′ and 39 ′′ (the position of the firstairy ring), and an aperture correction of 1.968 was ap-plied. If the source was undetected, we took the upperlimit as 2 × the 1 σ sky noise plus any positive flux in thesource aperture.3.3. Spitzer/IRAC photometry
IRAC images at 3.6, 4.5, 5.8, and 8 µ m were obtainedas part of GTO program 83 (PI G. Rieke) for five ofthe six clusters; IRAC data for Abell 1835 were obtainedthrough GTO program 64 (PI G. Fazio). Exposure timesare listed in Table 1. IRAC images were mosaicked asdescribed by Huang et al. (2004).Photometry was obtained by aperture photometrywith sky subtraction. The high density of cluster galax-ies in the cluster cores necessitated use of irregularly–shaped polygons for both target and sky apertures. The The aperture correction is specific to the DAOPHOT param-eters used, in this case sky annulus radii of 31 . ′′ and 40 ′′ , and aPSF defined out to r = 22 . ′′ . http://ssc.spitzer.caltech.edu/mips/apercorr/ flux density within these polygonal regions was deter-mined using CIAO Versions 3.2.1 and 3.4.1.1. IRACphotometry is listed in Table 3; the values do not includeaperture corrections.3.4.
HST photometry from WFPC2 and ACS
From the HST archive, we downloaded all publi-cally available ACS images for the clusters. We usedMultidrizzle (Koekemoer et al. 2002) to distortion–correct each flat-fielded pipeline image, and then cross–correlated the images to measure the small ( < ′′ ) coor-dinate registration offsets with high precision. The mea-sured offsets were then used by multidrizzle for the finalmosaicking. The data were photometered using the sametechnique as for the IRAC bands.We also used reduced WFPC2 images kindly madeavailable by D. Sand, which are described in Sand etal. (2005). ACS and WFPC2 cutouts of the galaxies inour sample are shown in Figure 1.3.5. Chandra ACIS imagery
From the Chandra archive, we downloaded all publi-cally available ACIS observations of clusters with IRStargets in this sample: observation ID numbers 529 and902 for MS0451; 1663, 5004, and 540 for A1689; 495 and496 for A1835; 1454, 1666, and 553 for A2218; 896 forA2219; 5007 and 550 for A2261; 4193, 500, and 501, forA2390; and 1562 for AC114.We used CIAO (3.3) to update the CTI correctionwhen necessary and remove data obtained during periodsof high solar flares. We mosaicked multiple overlappingobservations following the CIAO thread “ReprojectingImages: Making an Exposure–corrected Mosaic”.For X-ray detections, fluxes were determined from theindividual (not mosaicked) observations as described inMarcillac et al. (2007). X-ray flux upper limits (3 σ )were determined from the mosaicked images and mo-saicked exposure maps as in Donley et al. (2005), exceptthat the 90% encircled energy radius was used, and theMonte–Carlo sky apertures were taken close to the source(within 30 ′′ ) in order to sample a representative clusterbackground.Chandra X-ray fluxes and upper limits for the 2–8 keVand 0.5–8 keV bands are reported in Table 1.3.6. Long–wavelength photometry from the literature
Eight galaxies in our sample, and one galaxy from theextended (literature) sample, have submillimeter detec-tions in the literature; these are listed and referencedin Table 1. Four sources were detected at 15 µ m byISO: A851a and A2218b (Barvainis et al. 1999), A2390bLemonon et al. (1998), and A2390c (Metcalfe et al. 2003).We will use this photometry to help determine the totalinfrared luminosities of the sources. DISCUSSION4.1.
Aromatic Feature Flux Ratios in High–RedshiftGalaxies
We now test whether the mid-IR spectra of star–forming galaxies, with their prominent aromatic emission http://cxc.harvard.edu/ciao/ Rigby et al.features, show evidence for evolution from z = 0 to highredshift.Given its high quality, we first examine the IRS spec-trum for SMM J163554.2+661225 in detail. The imageis amplified by a factor of 22 ± µ m flux density wouldhave been 53 µ Jy, at the MIPS confusion limit. Thegalaxy redshift, from H α , is z = 2 . ± . µ m. (The lower-quality spec-trum from program 3453 confirms the detection of thesearomatic features, though the two longer–wavelength fea-tures are detected at low significance.) Overplotted witharbitrary normalization are two spectral templates: a)the starburst galaxy NGC 2798 from the SINGS survey(Dale et al. 2006); and b) the average spectrum of 13nearby starburst galaxies with little apparent AGN con-tribution (Brandl et al. 2006), whose mean IR luminosityis 4 . × L ⊙ . Coincidentally, this is also the L(TIR)of NGC 2798.The z = 2 .
516 spectrum in Figure 3 closely resem-bles the low–redshift, lower–luminosity (by a factor of 15)templates. Aromatic flux ratios are reported in Table 1,along with those we measure for the Brandl template,and those reported for the SINGS sample. The strengthof the 8.6 µ m feature in SMM J163554.2+661225 rela-tive to those at 6.2 µ m and 7.7 µ m, is within the 10%–90% variation within the SINGS sample. In addition, theflux ratios of SMM J163554.2+661225 are a close matchto those of the Brandl et al. (2006) average starbursttemplate; while 7.7 µ m is relatively stronger than thetemplate, the difference is comparable to the standarddeviation of line ratios for individual galaxies that wereaveraged to create the template. Thus, we concludethat this z = 2 . galaxy has aromatic feature fluxratios that are consistent with those observed forlower–luminosity, z = 0 starbursting galaxies. While the other spectra in the sample have lowersignal-to-noise ratios, they are still sufficient, in theaggregate, to test whether high–redshift galaxies havemarkedly different aromatic flux ratios. Figure 4 com-pares, for four aromatic feature flux ratios, the observedratios in our sample with those measured for the low–redshift Brandl et al. (2006) template. Though the er-rorbars in individual measurements are sizable, Figure 4confirms the result seen for SMM J163554.2+661225:star–forming galaxies at 1 < z < z = 0 star–forminggalaxies.Variations in aromatic feature ratios have been re-ported in certain HII regions—this may be attributableto variations in size, composition, and ionization state ofthe carriers (Draine & Li 2007 and references therein).In particular, large PAHs emit more strongly at 11.3 µ m,and neutral PAHs emit much more strongly at 3.3 and11.3 µ m than charged PAHs (Draine & Li 2007; Galliano2006; Allamandola et al. 1989). That our measured fea-ture ratios at z > z ∼ Evidence for Compact Source Accretion
Active galactic nuclei are detectable in multiple ways—by the dust they heat that radiates in the mid-IR; byX-rays emitted from the accretion disk coronae; and byhigh–excitation or broad emission lines. We now examinethe evidence for AGN activity in our extended sample.The first AGN diagnostic we examine is X-ray lumi-nosity, plotted in Figure 5 against aromatic luminos-ity. Most of the sample are X-ray non–detections, atlimiting luminosities that rule out X-ray–loud QSOs orbright Seyferts, although an X-ray–weak or highly ob-scured AGN could still be present (c.f. Donley et al.2005; Alonso-Herrero et al. 2006). Two of the fourgalaxies from the literature, and three of the galaxiesin our program are detected in X-rays: two have no ap-parent aromatic features (sources A2261a and A2390a);and the other, source A2390b, is interesting in that itcontains a luminous X-ray–emitting AGN, yet its spec-trum still shows aromatic features. That source emitsroughly equal power in the 7.7 µ m aromatic feature(scaled from the 11 µ m feature) and at 10–30 keV, remi-niscent of the more luminous z = 1 .
15 source CXO GWSJ141741.9+522823 discussed by Le Floc’h et al. (2007).We now consider a second AGN diagnostic, the rela-tive contribution of aromatic versus continuum emissionto the mid-IR flux. At low redshift, low aromatic featureequivalent widths have been demonstrated as an effectiveAGN diagnostic (Armus et al. 2007; Brandl et al. 2006,and Devost et al. (in prep.)). However, it is extremelydifficult to measure equivalent widths accurately for thehigh–redshift IRS sources, due to the limited wavelengthbaseline, the fact that the aromatic features have broadwings, and most importantly, imperfect sky subtractionwhich adds a (positive or negative) pedestal to each spec-trum. For example, for SMM J163554.2+661225 (ourhighest–quality spectrum), we cannot measure an accu-rate equivalent width due to the difficulty in determiningthe true continuum level. We are able to set a lower limitof > . µ m (rest-frame, conservatively assuming a highcontinuum level), which is a typical value for low–redshiftstar–forming galaxies. For the other spectra in the sam-ple, equivalent widths are even harder to measure.Therefore, we instead create an aromatic-to-mid-IRflux ratio, which we define as the ratio of the flux in the7.7 µ m PAH feature (in erg s − cm − , as fit by PAHFIT)to the MIPS photometry at 24 µ m (in Jy, corrected forbandwidth compression to z = 0.) This metric does notsuffer as strongly from the difficulty in determining thecontinuum level. In the low–redshift, star–forming com-parison sample we construct in § × − . We thus take 7 × − as a dividingline between spectra dominated by star formation (abovethe value), and spectra with substantial AGN contribu-tion (below the value). This discriminant selects the fol-lowing mid-IR spectra as having substantial AGN con-tribution: A370a, MS0451a, A1689a, A2218c, A2261a,A2390a. Thus, of 19 galaxies with adequate spectra,6 show strong indications of AGN contribution to theirmid-IR outputs.A third indication of AGN activity is a mid-IR spec-trum that rises steeply with increasing wavelength. Four For A2390b, we scale from the flux in the 11 µ m aromaticfeature, using the average 7.7/11 µ m flux ratio of 3.6 from Smithet al. (2007). RS spectra of lensed galaxies 5spectra (A2261a, A2218c, A1689a, and A370a) show thisbehavior; none has detected aromatic emission.A fourth indication of AGN activity is the presenceof broad or highly excited lines in an optical spectrum.Of the spectroscopy that has been published, or is inpreparation, for our sample (see references in Table 1),three show evidence for AGN: A1689a has a Keck spec-trum showing highly ionized neon (J. Richard et al. inprep.); A2390a has a Keck spectrum (see Appendix)which shows Lyman α with FWHM ∼
945 km s − , typ-ical of an AGN narrow line region; and the spectrumof A370a is reported to contain AGN lines (Lutz et al.2005).A fifth indication of AGN activity is a high ratio of[Ne III] 15.5 µ mto [Ne II] 12.8 µ m. Solar–metallicitystarbursts are not observed to excite this line ratioabove unity (Thornley et al. 2000; Rigby & Rieke 2004),whereas the harder continuum of AGN do. Eight spec-tra cover rest-frame 12.8 µ m; [Ne II] is detected in 6 ofthese. Six spectra cover rest-frame 15.5 µ m; [Ne III]is detected in 4. Of the four galaxies where both linesare detected (A1689a, Ab2218b, A2390b, and A2667a),only in A1689a does the [NeIII]/[NeII] ratio exceed unity:2 . ± .
76. This high line ratio offers additional evidencethat A1689a hosts a luminous AGN. The other, lower lineratios are more difficult to interpret.The results for these AGN diagnostics are compiled inTable 1. Combining these diagnostics, 8 galaxies out ofthe sample of 19 have at least one indication of AGNactivity (X-ray detection, low aromatic contribution tothe mid-IR flux, rising mid-IR spectrum, or optical AGNlines). Five galaxies have two different diagnostics thatindicate AGN. Thus, we find that 42% of the faint 24 µ msources in this sample have some nuclear activity; the95% confidence interval is 23%–63%. Further, in 32% ofthe sample, the AGN strongly affects or dominates themid-IR spectrum; the 95% confidence interval is 15%–53%. (Confidence intervals were calculated using eqn. 26of Gehrels (1986).)The high proportion of AGN in our sample of high–redshift LIRGS and ULIRGs mirrors the general behav-ior of local IR-luminous galaxies, where AGN becomesubstantially more common with increasing luminosity(e.g., Lutz et al. 1998). Though the errorbars are largegiven the small sample size, the high proportion (32%) ofAGN sufficiently powerful to dominate the mid–IR spec-tra may be inconsistent with the frequent assumptionthat ∼ µ Jy, z > µ m–derived star formation rates in deep surveys,where it is often assumed that the mid-IR outputs ofhigh–redshift, faint galaxies are dominated by strong aro-matic features from star formation. It also implies thata higher fraction of the integrated 24 µ m flux on the skymay arise from accretion than previously thought.4.3. Total infrared luminosity
We now determine the total infrared (TIR) luminosi-ties of the lensed sources, so that we may examine thedependence of aromatic feature emission on total infraredluminosity, and how this behavior may evolve with red-shift. 4.3.1.
SMM J163554.2+661225
The sub-mm galaxy behind Abell 2218 has exception-ally good photometric coverage from 0.4 to 850 µ m.Since it is triply–lensed, we increase the signal-to-noiseratio of its multiband photometry by using fluxes fromall three images, weighed by the amplifications given inKneib et al. (2004). (Bright neighboring galaxies forcedus to use only image A and B for the optical and IRACphotometry.) Figure 6 plots this photometry.We fit redshifted Chary & Elbaz (2001) and Dale &Helou (2002) templates to the 24, 70, 450, and 850 µ mphotometry, corrected for lensing amplification. Foreach template family, we determine the best–fit templateand the range of acceptable templates, then determineL(TIR) for those templates in a method consistent withthe definition given in Table 1 of Sanders & Mirabel(1996) using 12, 25, 60, and 100 µ m photometry. Weestimate the intrinsic total infrared luminosity of SMMJ163554.2+661225 as: • . ± . × L ⊙ for Dale & Helou (2002) tem-plates. • . +1 . − . × L ⊙ for Chary & Elbaz (2001) tem-plates.As Figure 6 shows, the best–fit Chary & Elbaz (2001)template over-predicts all > µ m points, especially at70 µ m. The best–fit Dale & Helou (2002) template isa better match to the shape of the overall infrared SEDand also has high aromatic feature equivalent widths,consistent with the spectrum of the source. Thus, thisgalaxy’s intrinsic luminosity is that of a LIRG.4.3.2. Other Galaxies
We adopt a similar approach for the other galaxieswith sufficient far–infrared and submillimeter data toprovide reasonable constraints. We tried both Dale &Helou (2002) and Chary & Elbaz (2001) templates, fit-ted to fall just below any upper limits so long as theywere compatible with the other constraints, i.e., to givea maximal luminosity. Two galaxies are fit poorly by astar–forming template, and much better by an AGN tem-plate, Mrk 231; MS0451a may be, as well. The ranges ofacceptable luminosities and the preferred templates arelisted in Table 3. Dale & Helou (2002) templates pro-vided generally better fits to our sample than Chary &Elbaz (2001) templates, as expected given the results ofMarcillac et al. (2006) for 0 < z < µ m (rest)-to-Total Luminosity Ratio We now consider the aromatic and total infrared (TIR)luminosities of the star–forming galaxies in the lensedsample, and how this ratio behaves as a function L(TIR).We compare to higher–luminosity star–forming galaxiesat z ∼
2, as well as comparable–luminosity star–forminggalaxies at z ∼ . z ∼
0. We interpret the prop-erties of the high redshift galaxies strictly in terms ofproperties that are well–determined for local analogs, sowe can test whether inconsistencies emerge.4.4.1.
The z ∼ comparison sample Rigby et al.To probe rest-frame 8 µ m and total infrared (TIR)luminosities for local galaxies, we use the SINGS sam-ple (photometry from Dale et al. 2005) supplementedin the ULIRG range by IRAS 09111-1007, 10565+2448,12112+0305, 14348-1447, 17208-0014, 22491-1808, andArp220 (Farrah et al. 2003; Armus et al. 2007), allselected because they appear to be dominated byyoung stars rather than AGN. We also include IRAS00262+4251, 01388-4618, 02364-4751, 16474+3430,23128-5919, and 23365+3604 from Rigopoulou et al.(1999). The strength of the aromatic features relativeto the continuum indicates that all these galaxies arealso dominated by star formation. For the galaxies fromFarrah et al. (2003) and Armus et al. (2007), we deter-mined 8 µ m photometry from archival IRAC images. Inour reductions, we included a correction for the extendedsource calibration. For those galaxies in Rigopoulou etal. (1999), we converted the tabulated continuum andaromatic 7.7 µ m peak fluxes into 8 µ m photometric val-ues by comparing with templates for galaxies in com-mon between Armus et al. (2007): IRAS 12112+0305,14348-1447, 1525+3609, and 22491-1808, plus Arp 220.The peak–to–peak scatter was a factor of two (in agree-ment with the finding of Chary & Elbaz (2001)). Weadd a variety of LIRGS (Alonso-Herrero et al. 2006b),again avoiding any galaxies with indications of signifi-cant levels of AGN power. We took IRAS measurementsfrom Sanders et al. (2003), and computed L(TIR) to beconsistent with the approach taken in that paper. Wedefine L(8 µ m) as being proportional to νf ν , and usesimilar definitions for all luminosities at specific wave-lengths. Figure 7 shows the relation between rest–frameL(8 µ m) and L(TIR) for these z ∼ Results from the literature at z ∼ We add a sample of infrared galaxies at z ∼ .
85 fromMarcillac et al. (2006). This paper utilizes 15 µ m datafrom ISO as well as 24 µ m data from Spitzer and, for asubset, radio data from the VLA; the authors found thatthe galaxies had rest–frame L(8 µ m)/L(12 µ m) ratios atthe high end of the local galaxy dispersion. We makethe assumption that L(12 µ m) correlates with L(TIR) inthat sample, as it is observed to do well at z = 0. Wecomputed rest 8 µ m luminosities for the members of thesample with 0.55 < z < µ m pho-tometry, and corrected them to the IRAC 8 µ m band bycomparison with the IRS spectrum of IRAS 2249, a lo-cal star-formation–dominated ULIRG. We took L(TIR)from the Dale & Helou (2002) template fits (Marcillac etal. 2006). The results are shown in green in Figure 8.There is no mid-IR spectroscopy for these galaxies, butif they had strong mid-IR contributions from AGN wewould expect substantial dispersion in the L(8 µ m) toL(TIR) ratio. Such a large dispersion is not seen.4.4.3. Results from the Literature at z ∼ We now compare the behavior of the 8 µ m luminosityvs. L(TIR) at z ∼ IRAS 12112+0305 and 14348-447 are double nuclei sources.
First, we took the stacked z ≈ . galaxy tem-plate of Daddi et al. (2005) and compared the ratioof observed 850 µ m to 24 µ m flux densities, referred tothe appropriate rest wavelengths for each of the localstar–forming galaxies where there is sufficient data toconstruct an accurate SED template (IRAS 1211, 1434,17208, 2249, and Arp 220). Excluding Arp 220, forwhich the ratio is 54, the average value for the other fourgalaxies is 14, whereas the Daddi et al. (2005) templateindicates a value of 6.5. Thus, the template indicatesstronger emission at rest 8 µ m relative to the submmthan is typical of local star–forming ULIRGs.We also compare with total infrared luminosity de-rived from submm measurements. Since the rest wave-length for the observed 850 µ m point in the Daddi et al.template is 293 µ m, we construct the average ratio ofL(TIR)/L(293 µ m) for local LIRGs and ULIRGs with4 × < L ( T IR ) < . × L ⊙ and high qualitymeasurements at 350–450 µ m (IRAS 09111, 1211, 1434,and 17208, Arp 220, and NGC 1614). We correct theflux densities for the differing effective filter bandwidthsusing the IRAS 2259 template. We correct to 293 µ m as-suming a spectral slope of -3.3, determined from a largersample of local ULIRG templates. We findL(TIR) = (98 ±
40) L(293 µ m)and then plot the Daddi et al. point as a red square inFigure 7. We did not fit the observations of lower lumi-nosity galaxies, because the scatter at lower luminositiesis very large, probably due to varying amounts of coldand warm dust in the infrared-emitting objects. Second, we make use of the stacking analysiscarried out by Papovich et al. (2007).
They re-port observed–frame 24 µ m and 70 µ m stacked photom-etry for galaxies that are both undetected in X-rays andhave non–power–law IRAC SEDS, and are therefore as-sumed to be powered predominantly by star formation.We computed the rest 8 µ m and 24 µ m flux densitiesfor this category (since they quote values for z = 2, therest wavelengths are represented directly by the mea-surements). We corrected the 8 µ m flux density (derivedfrom that observed at 24 µ m) to the equivalent for theIRAC band, again using the SED of IRAS 2249. Finally,we compute L(TIR) from the relation derived from ourlocal galaxy sample:L(TIR) = 86 . × L(24 µ m, rest) α , where α = 0 . .
14 dex.The resulting values are shown in Figure 7 as red dia-monds (we omitted the lowest flux density bin from Pa-povich et al. because the signal to noise ratio at 70 µ mis too low). Third, we use the results of Yan et al. (2007).
We select those galaxies that have z > .
5, and rest–frame 6.2 µ m equivalent widths (measured by Sajina etal. 2007) greater than 0.3 µ m. From Armus et al. (2007),this equivalent width threshold should select galaxiesdominated by star formation. The galaxies selected(MIPS 289, 8493, 15928, 16144, 22530) are all catego-rized as “type 1” by Yan et al. (2007), meaning theirspectra contain strong aromatic features. The medianrest-frame EW(7.7 µ m) is 2.5 µ m, corroborating thatstar formation dominates the mid-IR spectra. We con-verted the tabulated L(5.8 µ m) for these galaxies intoL(8 µ m) from an average template for our local ULIRGsample (IRAS 2249).RS spectra of lensed galaxies 7We estimate L(TIR) for these sources using the radio-infrared relation. We describe our adjustment of thisrelation for use at high redshift in detail in a forthcom-ing publication (N. Seymour et al., in preparation) butgive a short summary here. Yun et al. (2001) have quan-tified the relation for a large number of local galaxies,finding that the average ratio of flux density at 60 µ mto 1.4GHz is 110 independent of luminosity. The resultsof Yun et al. (2001) have not been K-corrected; sincethe infrared SEDs drop rapidly toward the blue, whilethe radio spectrum drops slowly, these corrections are afairly steep function of redshift. They enter at a signifi-cant level only for the high luminosity galaxies, becausethey tend to be at higher redshift than the lower luminos-ity ones. Including them increases the indicated ratio to ∼
170 at a L(TIR) = 10 L ⊙ . A second issue is that theradio outputs of LIRGs and ULIRGs often show signsof free-free absorption, which can decrease the outputsat 1.4GHz. To estimate this effect, we took the 19 starforming galaxies with adequate data from Condon et al.(1991) and computed an average slope for them of − . − . − .
58 and − .
7. The second concern is that the radio-infrared ratio may be a function of redshift (Kov´acs et al.2006; Vlahakis et al. 2007). We tested this suggestion bydetermining a K-correction between 1.4 GHz and 850 µ mfor an average of local LIRG and ULIRG templates (tobe published by Alonso-Herrero et al.) and applying itto a large sample of high redshift sources. We find thatthe K-correction accurately reproduces the observed be-havior with no offset, indicating there is no significantevolution of the relation. We have therefore used a valueof 140 for the intrinsic flux density ratio at rest wave-lengths to convert the radio measurements to TIR ones.We extrapolated the radio fluxes observed to a rest fre-quency of 1.4 GHz with a slope of − .
7. The redshiftusefully makes the observed rest frequency significantlyhigher than 1.4 GHz, reducing the effects of free-free ab-sorption for the high-z galaxies. To convert to L(TIR),we use the relationship from our local sample,L(TIR)= 740 × L(60) α , where α = 0 . reducing the change wefind in L(8 µ m)/L(TIR). That is, using the relation de-rived by Yun et al. (2001) and applying the proposedchange at high redshift (Kov´acs et al. 2006; Vlahakiset al. 2007) would have both resulted in a much largerchange than reported. Therefore, we are taking a con-servative approach.We take radio flux densities for the galaxies observedby Yan et al. (2007) from Condon et al. (2003) as tabu-lated on the SSC website . Two of the remaining galax-ies are detected at low significance as indicated by in-spection of the grey scale images for them (MIPS 429and 506), so we enter rough values of 0.1 mJy for them.For MIPS 289, which is undetected, we assigned an up- http://ssc.spitzer.caltech.edu/fls/extragal/vla.html per limit of 0.1 mJy. We reject the source MIPS 8493because of fringing in the VLA image at its position.The other four (MIPS 289, 15928, 16144, and 22530) arein regions where the radio images are clean. We correcttheir observed radio measurements to the rest frame as-suming a radio k-correction of (1 + z ) . , since a typicalslope for a starburst radio spectrum is − . µ m) val-ues in Figure 8 as circles (filled for the detections, openfor the upper limit.)4.4.4. µ m-to-TIR ratio for the z ∼ lensed galaxies The lensed galaxies described in this paper bridge theluminosity divide between the z ∼ z ∼ z = 0 . µ m, the SEDs are well enough constrained fora direct determination of L(TIR). Therefore, the lensedgalaxies can cleanly test for evolution with redshift in thebehavior of the L(8 µ m)/L(TIR) ratio.We use the aromatic-to-mid-IR flux ratio introducedin § × − ,with an average value of 16 × − . By comparison,our sample of local star–forming LIRGs and ULIRGshas a range in ratio of 7–32 × − , with an averageof 21 × − . That is, the relative aromatic featurestrength is virtually the same for the two samples, withno evidence that AGN have significantly augmented the8 µ m fluxes of our star–forming high–z sub-sample. TheEW(6.2 µ m) criterion we used for the Yan et al. (2007)sample would have returned the same sample of star–forming lensed galaxies, except that A2219a lacks 6.2 µ mcoverage and thus would not have been selected.We plot the four lensed galaxies as red triangles inFigure 8.4.5. Comparison of 8 µ m-to-TIR ratio at high and lowredshift The L(8 µ m)/L(TIR) ratio is important for two rea-sons: 1) it is one of the most accessible measures of thephysical behavior of infrared emission as a function ofredshift; and 2) many studies rely on fluxes at observed–frame 15 or 24 µ m to deduce L(TIR) and related proper-ties of galaxies at z ∼ µ m)/L(TIR). Figures 7 and 8 illustratethe important result that this ratio is well–behaved forstar–forming galaxies out to z ∼ .
5. In fact, this goodbehavior is a reasonably demanding test, since the var-ious samples at z ∼ µ m, and we estimated L(TIR) from the radio- Rigby et al.infrared relation. The stacked galaxies from Papovich etal. (2007) were selected based on faint K-band and 24 µ mdetections; we estimated L(TIR) from rest 24 µ m mea-surements. The stacked galaxies from Daddi et al. (2005)were originally selected from BzK colors indicating theyare at z ∼ µ m)/L(TIR) to decrease withincreasing luminosity for the local galaxies for L > L ⊙ , but this tendency seems reduced for the galaxies at z ∼
2. In fact, the high redshift galaxies appear to haveSED behavior similar to normal local galaxies of lowerluminosity in this regard.We have made two tests of the significance of this dif-ference. The first is based on the stacking results ofDaddi et al. (2005) and Papovich et al. (2007). Bothstudies select the galaxies to stack on the basis of K-banddetections, and therefore the mid– and far–infrared char-acteristics of the galaxies should be representative of theentire population of IR-active galaxies (i.e., there is nobias toward a specific L(8 µ m)/L(TIR). To have an anal-ogous set of data for the local galaxies, we have aver-aged their L(8 µ m)/L(TIR) ratios in luminosity intervalsof 10 . ; these averages are shown as the large blue trian-gles in Figure 7. We have then fitted the averaged datapoints, shown as a blue line. To test whether the highredshift stacked results were compatible with the samefit, we carried out a second fit in which we included thethree stacked points. For the galaxies above 10 L ⊙ , wethen compared χ for the low redshift galaxies only in thetwo cases. We found that this χ is 4.5 times larger if thethree stacked points at z ∼ z ∼ To determine whether a few anomalous local galax-ies could affect this conclusion, we repeated the proce-dure but using a more robust average for the luminosity-binned ratios. For each bin with ten or more items, wediscarded the two highest and two lowest ratios and thenaveraged the remainder. Where there were less than tenitems, we retained the straight average. The highest lu-minosity bin, which is responsible for most of the changein slope of the relation between the two luminosities, hasten objects and hence its average was recomputed. Theresults were virtually identical to those with the straightaverages, as shown in Figure 7.A second test was applied to the high redshift galax-ies measured individually, either by Yan et al. (2007),or by us. In the case of the Yan et al. sample, we ex-pect a strong bias toward objects with a large ratio ofL(8 µ m)/L(TIR), because they selected relatively brightobjects for spectroscopy, given the sensitivity limitationsof the IRS. Our sample was selected on a combination of24 µ m flux density and lensing amplification, so a bias islikely, although perhaps less severe. This analysis makes the assumption that low–luminositygalaxies have similar behavior at low and high redshift.
To compare with these two studies, we selected a lo-cal sample with relatively high 8 µ m output by removingall the galaxies that are below the fit made to the aver-aged values (above). This sample is shown in Figure 8.It is noteworthy that the scatter in L(8 µ m)/L(TIR) issmall, with no outliers toward high L(8 µ m), so this is anappropriate comparison sample for cases where a bias issuspected. As before, we carried out two fits, one justto the local galaxies, and another to them plus the eightindividually measured high redshift ones. We found that χ for the low redshift sample with L > L ⊙ in-creased by a factor of 2.5 if we included the high redshiftgalaxies in the fit. Comparing Figure 8 with Figure 7,the discrepancy would be substantially larger if we hadincluded all the local galaxies in our fits. This result con-firms that found for the stacked samples, namely that asingle functional fit is not a good fit to the data over allredshifts.Another comparison can be made with the sample ofinfrared galaxies at z ∼ .
85 from Marcillac et al. (2006).The results fall in a region occupied by both low andhigh redshift fits in Figure 8, which confirms the rela-tively modest change in L(8 µ m)/L(TIR) with redshift,but (not surprisingly given the small difference between z ∼ µ m observed flux densities. If insteadwe treat them as an unbiased sample and average themin luminosity intervals, they fall along (or slightly above)the fit for the z ∼ µ memission could arise either through a contribution of anextra, AGN-powered component at 8 µ m (e.g. the ex-tra 5.8 µ m continuum claimed by Sajina et al. 2007 andthe mid-IR excess galaxies with hard X-ray emission dis-cussed by Daddi et al. 2007b), or to a change in theproperties of the stellar–powered infrared–emitting re-gions. The AGN explanation runs counter to our selec-tion of individual galaxies with strong aromatic featuresthat correspond to equivalent widths typical of purelystar forming local galaxies. It is also counter to the pre-cautions taken in the stacking analyses to exclude AGN(e.g., Papovich et al. 2007). In addition, AGN contribu-tions would come in a broad range, so it would be hardto explain the consistent behavior of the five high red-shift samples in Figures 7 and 8 if AGN were causingthe departure from the behavior from local star forminggalaxies. Finally, the behavior we see is similar to thatfound by Zheng et al. (2007) in a stacking analysis at z ∼ .
7; specifically, they report that the far-IR SEDsappear to resemble those of local galaxies of lower lumi-nosity. Thus, while AGN may boost 8 µ m emission inother sources, it is unlikely that AGN contamination isresponsible for the behavior seen in Figures 7 and 8.We conclude that star-forming galaxies have a mildtrend toward increasing L(8 µ m) / L(TIR) ratio with red-shift that reaches up to a factor of ∼ z = 2. Inother words, L(8 µ m) does not turn over with increasingL(TIR) for z ∼ . µ m aromatic equivalent widths than areobserved for local ULIRGs (Desai et al. 2007; Men´endez-Delmestre et al. 2007; Sajina et al. 2007), though directRS spectra of lensed galaxies 9comparisons cannot be made because there are not localgalaxies with such extremely high luminosities. Becausethe lensed galaxies and stacked samples directly probethe 10 –10 . L ⊙ range at z ∼
2, we can directly com-pare to z ∼ z = 0 trend line.Use of local templates to estimate L(TIR) for high-zgalaxies detected at 24 µ m has indicated the existenceof very high luminosities at z ∼ µ m)/L(TIR) sug-gested here would reduce L(TIR) for a given L(8 µ m)by a factor of ∼
4. As shown in the figure, in our sam-ple there is only one galaxy selected to be dominated bystar formation that has a L(TIR) significantly exceeding10 L ⊙ . A luminosity of 10 L ⊙ corresponds to astar formation rate of ∼ ⊙ yr − if we use thelocal calibration (Kennicutt 1998), and is suggestivelyclose to the theoretical maximum luminosity that a star-burst is allowed by causality arguments—that star for-mation throughout a galaxy should not transpire fasterthan the dynamical timescale (Lehnert & Heckman 1996;Elmegreen 1999). (The actual maximum value dependson efficiency and dynamical mass.) Thus, our lower cal-ibration for L(8 µ m)/L(TIR) may remove the discrep-ancy between L(TIR) inferred for high redshift galaxiesand plausible upper limits on what a starburst can pro-duce. Spatially–resolved imaging of high–redshift star-bursts (e.g. Nesvadba et al. 2007) can test whether theyare in fact forming stars at the dynamical limit. SUMMARY AND IMPLICATIONSThe 24 µ m photometric band of Spitzer efficientlydetects galaxies out to z ∼
3, and is being used asan extinction–robust measure of star formation rates inthe distant universe. We present new IRS spectra andMIPS photometry of gravitationally–amplified galaxies,and use them in concert with archival Chandra imagesand published submillimeter photometry to investigatethe nature of the very faint 24 µ m sources.The aromatic feature flux ratios observed in our spec-tra agree well with those of star–forming template spec- tra at z ∼
0. This result implies that dust size and ioniza-tion distribution are not strongly evolving with redshift.Thus, for z & z ∼ z ∼
2, the 24 µ m data probe the 8 µ m rest wave-length region. We find that the ratio of rest-frame 8 µ mluminosity to total infrared luminosity is approximatelythe same as for local galaxies of similar L(TIR). How-ever, the z ∼ z ≈ µ m) luminosities for a given L(TIR)than low–redshift analogs. Functionally, the break inthe L(8 µ m) vs. L(TIR) relationship occurs at a 10-times-higher L(TIR) at z ∼ z ∼ z ∼
2, and may indicate a maximumstar–forming luminosity of a few × L ⊙ .The results just summarized were found for sourceswhere AGN do not contribute substantially to the mid-IR spectra, as determined by the aromatic-to-mid-IR fluxratio. But 6 of the 19 mid-IR spectra in our extendedsample do show a strong AGN. Including other AGN di-agnostics (X-ray detections and optical emission lines),8 of the 19 galaxies show evidence for nuclear activity.The prevalence of AGN in this sample may contradictthe frequent assumption that the 24 µ m flux density in100 µ Jy sources at 1 < z < This galaxy was selected as star–forming based on itsEW(6.2 µ m) (Sajina et al. 2007), but its EW(7.7 µ m) is unusu- ally low, which may indicate a substantial AGN contribution. APPENDIX
NOTES ON INDIVIDUAL SOURCES
Source A2390a
Johan Richard obtained a Keck spectrum for this source, shown in Figure 9. The strong Lyman α emission isredshifted by z = 2 . ∼
945 km s − ) indicates that the source is an AGN. Source A2261a
This sub-mm galaxy (Chapman et al. 2002) lacks a spectroscopic redshift; submillimeter and radio-based photometricestimates place it at z > . . < z < µ m complex for 0 . < z < .
8, and to the7.7 µ m aromatic feature for 1 . < z < .
1. Thus, for the aromatic features to fall outside the IRS spectral coverage,the source would have to have very low ( z <
1) or very high ( z > .
1) redshift. If the source were at z > .
1, it mustbe ∼ µ m flux density. It seems more probablethat this source lies at 1 < z < REFERENCESAkiyama, M., Ueda, Y., Ohta, K., Takahashi, T., & Yamada, T.2003, ApJS, 148, 275Allamandola, L. J., Tielens, G. G. M., & Barker, J. R. 1989, ApJS,71, 733Alonso-Herrero, A., et al. 2006, ApJ, 640, 167Alonso-Herrero, A., Rieke, G. H., Rieke, M. J., Colina, L., P´erez-Gonz´alez, P. G., and Ryder, S. D. 2006, ApJ, 640, 167Appleton, P. N., et al. 2004, ApJS, 154, 147Aretxaga, I., Hughes, D. H., Chapin, E. L., Gazta˜naga, E., Dunlop,J. S., & Ivison, R. J. 2003, MNRAS, 342, 759Armus, L., et al. 2007, ApJ, 656, 148Barvainis, R., Antonucci, R., & Helou, G. 1999, AJ, 118, 645Bicay, M. D., & Helou, G. 1990, ApJ, 362, 59Brandl, B. R., et al. 2006, ApJ, 653, 1129Caputi, K. I., et al. 2006, ApJ, 637, 727Chapman, S. C., Scott, D., Borys, C., & Fahlman, G. G. 2002,MNRAS, 330, 92Charmandaris, V., et al. 2004, ApJS, 154, 142Chary, R., & Elbaz, D. 2001, ApJ, 556, 562Choi, P. I., et al. 2006, ApJ, 637, 227Condon, J. J., Huang, A.-P., Yin, Q. F., & Thuan, T. X. 1991,ApJ, 378, 65Condon, J. J., Cotton, W. D., Yin, Q. F., Shupe, D. L., Storrie-Lombardi, L. J., Helou, G., Soifer, B. T., & Werner, M. W. 2003,AJ, 125, 2411Cowie, L. L., Barger, A. J., & Kneib, J.-P. 2002, AJ, 123, 2197Daddi, E., et al. 2007a, submitted to ApJ, astro-ph/0705.2831v1Daddi, E., et al. 2007b, submitted to ApJ, astro-ph/0705.2832v1Daddi, E., et al. 2005, ApJ, 631, L13Dale, D. A., et al. 2005, ApJ, 633, 857Dale, D. A., Smith, J.D., Armus, L., et al. 2006, ApJ, 646, 161Dale, D. A., & Helou, G. 2002, ApJ, 576, 159Desai, V., et al. 2007, ApJin press, (arXiv:0707.4190)Dole, H., et al. 2004, ApJS, 154, 93Donley, J. L., Rieke, G. H., Rigby, J. R., & P´erez-Gonz´alez, P. G.2005, ApJ, 634, 169Downes, D., & Solomon, P. M. 2003, ApJ, 582, 372007ApJ...663..866D Draine, B. T., et al. 2007, ApJ, 663, 866Draine, B. T., & Li, A. 2007, ApJ, 657, 810Ebbels, T., Ellis, R., Kneib, J.-P., Leborgne, J.-F., Pello, R., Smail,I., & Sanahuja, B. 1998, MNRAS, 295, 75Egami, E., Rieke, G. H., Fadda, D., & Hines, D. C. 2006, ApJ, 652,L21Egami, E. et al. 2006, ApJ, 647, 922Elmegreen, B. G. 1999, ApJ, 517, 103Farrah, D., Afonso, J., Efstathiou, A., Rowan-Robinson, M., Fox,M., & Clements, D. 2003, MNRAS, 343, 585Frayer, D. T., Ivison, R. J., Scoville, N. Z., Yun, M., Evans, A. S.,Smail, I., Blain, A. W., & Kneib, J.-P. 1998, ApJ, 506, L7Frayer, D. T., et al. 1999, ApJ, 514, L13Freedman, W. L., et al. 2001, ApJ, 553, 47Frye, B., & Broadhurst, T. 1998, ApJ, 499, L115Garrett, M. A., Knudsen, K. K., & van der Werf, P. P. 2005, A&A,431, L21Galliano, F. 2006, ArXiv Astrophysics e-prints, arXiv:astro-ph/0610852Gehrels, N. 1986, ApJ, 303, 336Gordon, K. D., Rieke, G. H., Engelbracht, C. W. et al. 2005, PASP,117, 503 Helou, G., Soifer, B. T., & Rowan-Robinson, M. 1985, ApJ, 298,L7Houck, J. R., et al. 2004, ApJS, 154, 18Houck, J. R., et al. 2005, ApJ, 622, L105Huang, J.-S., et al. 2004, ApJS, 154, 44Ivison, R. J., Smail, I., Barger, A. J., Kneib, J.-P., Blain, A. W.,Owen, F. N., Kerr, T. H., & Cowie, L. L. 2000, MNRAS, 315,209Kennicutt, R. C., Jr. 1998, ARA&A, 36, 189Kneib, J.-P., van der Werf, P. P., Kraiberg Knudsen, K., Smail, I.,Blain, A., Frayer, D., Barnard, V., & Ivison, R. 2004, MNRAS,349, 1211Knudsen, K. K., van der Werf, P. P., Kneib, J.-P. 2007, MNRAS,submittedKnudsen, K. K. 2004, Ph.D. Thesis, Leiden UniversityKnudsen, K. K., et al. 2006, MNRAS, 368, 487Koekemoer, A., Fruchter, A., Hook, R., Hack, W. 2002, HSTCalibration Workshop, p. 337Kov´acs, A., Chapman, S. C., Dowell, C. D., Blain, A. W., Ivison,R. J., Smail, I., & Phillips, T. G. 2006, ApJ, 650, 592Lagache, G., et al. 2004, ApJS, 154, 112Le Floc’h, E., et al. 2007, ApJ, 660, L65Le Floc’h, E., et al. 2005, ApJ, 632, 169Lehnert, M. D., & Heckman, T. M. 1996, ApJ, 472, 546Lemonon, L., Pierre, M., Cesarsky, C. J., Elbaz, D., Pello, R.,Soucail, G., & Vigroux, L. 1998, A&A, 334, L21Lutz, D., Spoon, H. W. W., Rigopoulou, D., Moorwood, A. F. M.,& Genzel, R. 1998, ApJ, 505, L103Lutz, D., Valiante, E., Sturm, E., Genzel, R., Tacconi, L. J.,Lehnert, M. D., Sternberg, A., & Baker, A. J. 2005, ApJ, 625,L83Marcillac, D., Elbaz, D., Chary, R. R., Dickinson, M. E., Galliano,F., & Morrison, G. 2006, A&A, 451, 57Marcillac, D., Rigby, J. R., Rieke, G. H., & Kelly, D. M. 2007, ApJ,654, 825Men´endez-Delmestre, K., et al. 2007, ApJ, 655, L65Metcalfe, L., et al. 2003, A&A, 407, 791Motohara, K., et al. 2005, AJ, 129, 53Nesvadba, N. P. H., et al. 2007, ApJ, 657, 725Papovich, C. et al. 2007, ApJ, submittedPello, R., Sanahuja, B., Le Borgne, J.-F., Soucail, G., & Mellier,Y. 1991, ApJ, 366, 405P´erez-Gonz´alez, P. G., et al. 2005, ApJ, 630, 82Reddy, N. A., Steidel, C. C., Fadda, D., Yan, L., Pettini, M.,Shapley, A. E., Erb, D. K., & Adelberger, K. L. 2006, ApJ, 644,792Rieke, G. H., et al. 2004, ApJS, 154, 25Rigby, J. R., & Rieke, G. H. 2004, ApJ, 606, 237Rigopoulou, D., Spoon, H. W. W., Genzel, R., Lutz, D., Moorwood,A. F. M., & Tran, Q. D. 1999, AJ, 118, 2625Sand, D. J., Treu, T., Ellis, R. S., & Smith, G. P. 2005, ApJ, 627,32Sanders, D. B., Mazzarella, J. M., Kim, D.-C., Surace, J. A., &Soifer, B. T. 2003, AJ, 126, 1607Sanders, D. B., & Mirabel, I. F. 1996, ARA&A, 34, 749Sajina, A. Yan, L., Armus, L., Choi P., Fadda, D., Helou, G., &Spoon, H. 2007, astro-ph/0704.1765Siebenmorgen, R., & Kr¨ugel, E. 2007, A&A, 461, 445Smail, I., Ivison, R. J., Owen, F. N., Blain, A. W., & Kneib, J.-P.2000, ApJ, 528, 612
RS spectra of lensed galaxies 11
Smail, I., Ivison, R. J., Blain, A. W., & Kneib, J.-P. 2002, MNRAS,331, 495Smail, I., Smith, G. P., & Ivison, R. J. 2005, ApJ, 631, 12Smith, G. P., Kneib, J.-P., Smail, I., Mazzotta, P., Ebeling, H., &Czoske, O. 2005, MNRAS, 359, 417Smith, J. D. T., Draine, B. T., Dale., D. A., et al. 2007, ApJ, 656,770Spergel, D. N., et al. 2003, ApJS, 148, 175Teplitz, H. I., et al. 2007, ApJ, 659, 941Thornley, M. D., Schreiber, N. M. F., Lutz, D., Genzel, R., Spoon,H. W. W., Kunze, D., & Sternberg, A. 2000, ApJ, 539, 641 Vlahakis, C., Eales, S., & Dunne, L. 2007, MNRAS, 379, 1042Wang, Y. P. 2006, ApJ, 638, 138Werner, M. W., et al. 2004, ApJS, 154, 1Yan, L., et al. 2007, ApJ, 658, 778Yun, M. S., Reddy, N. S., & Condon, J. J. 2001, ApJ, 554, 803Zheng, X. Z., Dole, H., Bell, E. F., Le Floc’h, E., Rieke, G. H.,Rix, H.-W., & Schiminovich, D. 2007, ApJ in press, astro-ph/0706.0003
TABLE 1The IRS lensed sample.
RA(J2000) DEC(J2000) sub-mm name ref PID z ref ampOur sampleMS0451a 04:54:07.12 − .
95 1 2 . .
38 1 1 . , . − . . ± . − .
63 2 25–47A1835a 14:01:04.96 +02:52:24.8 SMM J140105+025223.5 C 12 2 .
565 3 3 . ± . a A2218a 16:35:54.18 +66:12:24.8 SMM J163554.2+661225 D 1 2 .
516 4 22 ± .
034 5 6 . .
97 1 6.7A2219a 16:40:19.50 +46:44:00.5 SMM J16403+46440 F 1 2 .
03 1 3 . < z < . .
858 6 10 . .
913 7 3 . .
91 8 10 . − .
47 1 9 . − .
034 9 17Additional sources from the literatureTeplitz-1 3:32:44.00 -27:46:35.0 8 2.69 10 1.0Teplitz-1-BzK 3:32:38.52 -27:46:33.5 8 2.55 10 1.0Teplitz-2 3:32:34.85 -27:46:40.4 8 1.09 10 1.0A370a 2:39:51.87 -1:35:58.78 SMM J02399-0136 A 9 2.81 11 2.45
Note . — Columns:1) source;2–3) coordinates;4) sub-mm name, if source has a sub-mm detection in the literature;5) reference for detected sub-mm counterpart: “A” is Smail et al. (2002); “B” is Knudsen (2004) and Knudsen (2007); “C”is Smail et al. (2000); Ivison et al. (2000); “D” is Kneib et al. (2004); “E” is Knudsen et al. (2006); “F” is Chapman et al.(2002); “G” is Cowie et al. (2002).6) Spitzer program in which IRS spectra were obtained: “1” signifies PID 82; “2” signifies PID 30775; “8” signifies PID252; “9” signifies PID 3241;7) redshift;8) redshift reference: “1” means redshift was determined from our IRS spectra; “2” is based on Keck spectra (J. Richardet al. in prep.); “3” is from Frayer et al. (1999); “4” is from Kneib et al. (2004); “5” is from Ebbels et al. (1998); “6” isfrom Keck spectra presented in the Appendix; “7” is from Pello et al. (1991); “8” is from Frye & Broadhurst (1998); “9”is from Sand et al. (2005); “10” is from Teplitz et al. (2007); “11” is from Frayer et al. (1998) and also Lutz et al. (2005).9) amplification estimated from the lensing models. Values from the literature are footnoted; un-footnoted values are new; a The amplification for this source has been the subject of several papers; Downes & Solomon (2003) and Motohara et al.(2005) argue that an intervening cluster galaxy boosts the amplification to ∼
25. Smail et al. (2005) and Smith et al. (2005)find that the intervening galaxy is a dwarf with small velocity dispersion and thus contributes only modestly to the totalamplification, which they estimate as 3 . ± .
5. We note the controversy and adopt the Smail et al. (2005) amplification.
RS spectra of lensed galaxies 13
TABLE 2Exposure times.
Cluster t(24 µ m) t(3.6 µ m) t(IRS, LL) t(Chandra)MS0451 3 .
68 2 .
40 3.66 b .
78 2 .
40 6.34 a .
78 2 .
40 40a 7.31 a b 14.63 b A1835 2 .
77 6 .
00 3.66 b ; 3.66 a .
78 2 .
40 58a 3.66 b b 3.66 a c 7.31 a A2219 2 .
78 2 .
40 3.66 b .
77 2 .
40 5.85 b ; 7.31 a .
77 2 .
40 109a 3.66 b ; 5.61 a b 3.66 b ; 5.61 a c 7.31 a AC114 2 .
77 2 .
40 7.31 a .
77 2 .
40 1.95 a Note . — Exposure times, in kiloseconds, for the following bands:Spitzer/MIPS 24 µ m; Spitzer/IRAC 3.6 µ m; and Spitzer/IRS long-low grating. The 24 µ m exposure time is the median (per-pixel)exposure time within the central 4 ′ by 4 ′ box. a Both LL1 and LL2 were given this exposure time. b Only order LL1.
TABLE 3Long-wavelength photometry and template fits. f(3.5 µ m) f(4.5 µ m) f(5.7 µ m) f(8.0 µ m) f(24 µ m) (PAH/tot) f(70 µ m) L(TIR) templateMS0451a 32 ± ± ± ± .
32 6.3E-11 5 . ± . < . × CE a A851a 29 ± ± ± ± .
63 8.4E-11 · · · × DHA1689a 36 ± ± ± ± . · · · · · · A1689b 0 .
32 2.4E-10 · · · × DHA1835a 88 ± ± ± ± .
99 1.4E-10 <
13 6.8 (4.2–11) × DHA2218a 67 ± ± ± ±
10 1 .
16 2.0E-10 < × DHA2218b 182 ± ± ± ± .
67 7 . ± . × DHA2218c 99 ± ± ± ±
11 0 .
50 10 ± × Mrk 231A2219a 0 .
82 7.1E-11 · · · × DHA2261a 142 ± ± ± ± . < . c A2390a 39 ± ± ± ± .
83 5.8E-11 < . · · · A2390b 462 ± ± ± ± .
88 5 . ± . × Mrk 231A2390c 0 . < × DHAC114a 91 ± ± ± ± .
41 1.4E-10 · · · · · ·
A2667a 466 ± ± ± ± . · · · · · · Literature sampleTeplitz-1 0 . · · · Teplitz-1-BzK 0 . · · · Teplitz-2 0 . · · · A370a 1 . · · · Note . — Columns: 1) Source name. 2–5) IRAC photometry, in µ Jy. Formal errors from the photometry are quoted,though field crowding probably and aperture corrections probably introduce additional 5–10% errors. The photometry forA2218a is superior since it is multiply imaged (with different crowding). Crowding is too severe to quote photometry forA1689b, A2219a, and A2390c. 6) Observed 24 µ m flux density, in mJy, from DAOPHOT PSF-fitting. Errorbars fromDAOPHOT are overly optimistic and not quoted; we use errorbars of 0.1 mJy for template fitting. 7) Aromatic–to–mid-IRflux ratio, as defined in § z > . µ m flux density, in mJy, from aperture photometry.Errorbars are dominated by the sky. 9) Total infrared luminosity, our best–fit and the range of acceptable templates, in L ⊙ .10) Best-fitting template: Chary & Elbaz (2001) or Dale & Helou (2002). a Mrk 231 template may also fit. b Using amplification factor of 1.3. c L(TIR) cannot be usefully constrained because the redshift is unknown.
TABLE 4Measured Aromatic fluxes. src 6.2 σ σ σ σ σ σ σ σ σ σ σ MS0451a 28 6 4 . . . . . A1689b 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Note . — Measured aromatic dust emission feature fluxes and uncertainties, measured with PAHFIT. Units are 10 − erg s − cm − . Columnnheaders give the central wavelength, in µ m, of each feature. For features comprised of multiple components (like 7.7 µ m), total flux and uncertaintywere calculated using PAHFIT’s “main feature” function. TABLE 5X-ray fluxes and σ upper limits. src f(2–8 keV) f(0.5–8 keV)erg s − cm − erg s − cm − MS0451a < . × − < . × − A851a · · · · · ·
A1689a < . × − < . × − A1689b < . × − < . × − A1835a < . × − < . × − A2218a < . × − < . × − A2218b < . × − < . × − A2218c < . × − < . × − A2219a < . × − < . × − A2261a 3 . × − . × − A2390a 5 . × − . × − A2390b 4 . × − . × − A2390c < . × − < . × − AC114a < . × − < . × − A2667a < . × − < . × − Note . — X-ray limits are the 3 σ sky back-ground plus any additional counts at the sourceposition, as described in § TABLE 6Aromatic ratios for SMM J163554.2+661225 lines ratio SINGS Starburst7.7 µ m/6.2 µ m 3 . ± . µ m/8.6 µ m 4 . ± . µ m/8.6 µ m 1 . ± . Note . — Columns: (1) line wavelengths ( µ m); (2) mea-sured flux ratios; (3) median flux ratios and 10%–90%range of variation from the SINGS sample (Smith et al.2007); (4) flux ratios of the average starburst template ofBrandl et al. (2006). RS spectra of lensed galaxies 15
TABLE 7[Ne III] 15.5 µ m and [Ne II] 12.8 µ m. src [Ne II] σ [Ne III] σ [Ne III]/[Ne II] σ MS0451 – · · ·
A851 · · · · · ·
A1689a 0.93 0.26 2.0 0.43 2.15 0.76A1689b · · · · · ·
A2218a · · · · · ·
A2218b 2.9 0.6 0.91 0.31 0.31 0.13A2218c – –A2219 · · · · · ·
A2390a · · · · · ·
A2390b 2.2 0.57 0.90 0.69 0.41 0.33A2390c 0.83 0.47 –AC114a 1.6 0.29 · · ·
A1835a · · · · · ·
A2261a · · · · · ·
A2667a 2.5 0.49 0.92 0.51 0.37 0.22Tep1 · · · · · ·
T1-bzk · · · · · ·
Tep 2 · · · · · ·
A370a · · · · · ·
Note . — Measured fine structure line fluxes, from PAHFIT, in unitsof 10 − erg s − cm − . “ · · · ” means the wavelength regime was notcovered; “–” means the line was covered but not detected. TABLE 8AGN Diagnostics. src X weak aromatic rising optMS0451a X · · · · · ·
A851a · · · · · · · · ·
A1689a X X XA1689b · · ·
A1835a · · ·
A2218a · · ·
A2218bA2218c X X · · ·
A2219a · · · · · ·
A2261a X X X · · ·
A2390a X X · · ·
XA2390b XA2390cAC114a · · ·
A2667aTep-1 X · · ·
Tep-1-BzK · · · · · ·
Tep-2A370a X X X · · · X Note . — Columns: 1) source name 2) X-ray detection;3) low aromatic flux contribution, as defined in § Fig. 1.—
Postage stamps of the lensed galaxies. Images are from the Hubble Space Telescope, Advanced Camera for Surveys, filterF850LP, with the following exceptions: the MS0451a image is HST ACS F814W; the A2667a and A851a images are HST WFPC2 F814W;and for A2261a and A2219a, the images are from the Steward Observatory 90 inch, F606W. Each postage stamp is 12 ′′ by 12 ′′ . Circles of R = 6 ′′ are drawn to illustrate the FWHM beam of Spitzer at 24 µ m. RS spectra of lensed galaxies 17
15 20 25 30 35 40-10123456 6 8 10 12
MS0451a, z=1.95
15 20 25 30 35 40-10123456 6 8 10
A851a, z=2.38
15 20 25 30 35 40-10123456 8 10 12 14 16 18
A1689a, z=1.15
15 20 25 30 35 40-10123456 4 5 6 7 8 9 10
A1689b, z=2.68
15 20 25 30 35 40-101234564 6 8 10
A2218a, z=2.52
15 20 25 30 35 40-10123456 8 10 12 14 16 18
A2218b, z=1.03
15 20 25 30 35 40-10123456 8 10 12 14 16 18 20
A2218c, z=0.97
15 20 25 30 35 40-10123456 6 8 10 12
A2219a, z=2.03
15 20 25 30 35 40-10123456 4 5 6 7 8 9 10
A2390a, z=2.86
15 20 25 30 35 40-10123456 8 10 12 14 16 18 20
A2390b, z=0.91
Fig. 2.—
IRS spectra. Lower x-axes show observed wavelength in µ m; upper x-axes show rest wavelength ( µ m) if redshift is known.Y-axes plot observed flux density (mJy). Spectra obtained for the same object in different programs are plotted separately, with program82 plotted first. Vertical dotted lines show the positions of expected aromatic and fine structure emission features. Note that artifacts arepresent at λ > µ m.
15 20 25 30 35 40-10123456 8 10 12 14 16 18 20
A2390c, z=0.92
15 20 25 30 35 40-10123456 6 8 10 12 14 16
AC114a, z=1.47
15 20 25 30 35 40-10123456 4 6 8 10
A1835a, z=2.56
15 20 25 30 35 40-10123456 4 6 8 10
A1835a, z=2.56
15 20 25 30 35 40-10123456
A2261a, z=?
15 20 25 30 35 40-10123456
A2261a, z=?
15 20 25 30 35 40-10123456 8 10 12 14 16 18
A2667a, z=1.04
Fig. 2 — continued.RS spectra of lensed galaxies 19 f ν ( m Jy ) observed wavelength ( µ m)rest wavelength ( µ m) Fig. 3.—
IRS spectrum of z = 2 .
516 lensed source SMM J163554.2+661225 behind Abell 2218 (solid thick line) . Overplotted are two z ∼ (thin solid line) ; and the average of 13 starburst galaxiesfrom Brandl et al. (2006) (thin dashed line) . Crosses show the wavelengths of known aromatic components from Smith et al. (2007). R i g b y e t a l. . m / . m f l ux r a ti o observed f(7.7um) (10 -15 erg s -1 cm -2 ) 0 5 10 15 20 0 20 40 60 80 100 . m / . m f l ux r a ti o observed f(7.7um) (10 -15 erg s -1 cm -2 ) 0 2 4 6 8 10 12 14 0 20 40 60 80 100 . m / . m f l ux r a ti o observed f(7.7um) (10 -15 erg s -1 cm -2 ) 0 5 10 15 20 0 20 40 60 80 100 . m / . m f l ux r a ti o observed f(7.7um) (10 -15 erg s -1 cm -2 ) F i g . . — A r o m a t i c flu x r a t i o s . T h e a r o m a t i c flu x r a t i o s a r e p l o tt e d aga i n s t o b s e r v e dflu x i n t h e . µ m f e a t u r e . F l u x e s w e r e fi t u s i n g P AH F I T . A m p li fi c a t i o n s h a v e n o t b ee nd i v i d e d o u t . A l s o p l o tt e d a r e t h e flu x r a t i o s o f t h e l o w – r e d s h i f t B r a nd l e t a l. ( ) s t a r bu r s tt e m p l a t e (s o l i dho r i z o n t a ll i n e s) , w i t h t h e ± s t a nd a r dd e v i a t i o n s f r o m t h a t s a m p l e ( do tt e dho r i z o n t a ll i n e s) . RS spectra of lensed galaxies 21 -15-14-13 -15 -14 -13 l og ob s . µ m P AH f l ux ( e r g s - c m - ) log obs 0.5 - 8 keV flux (erg s -1 cm -2 ) l og L ( . µ m P AH ) log rest-frame L(10-30 keV) Fig. 5.—
Comparison of aromatic and X-ray fluxes and luminosities. SMM J163554.2+661225 is represented by filled symbols. f ν ( m Jy ) observed wavelength ( µ m)A2218a CEDH
Fig. 6.—
Photometry and L(TIR) fits. Photometry and IRS spectra (thin lines) are plotted, with flux densities and wavelengths in theobserved frame. Overplotted are best-fit templates: Dale & Helou (2002) ( thick solid line ), Chary & Elbaz (2001) ( thick dashed line ), andMrk 231 (dotted line) . IRAC and HST photometry are plotted for SMM J163554.2+661225to demonstrate that its SED is dominated bystar formation.
RS spectra of lensed galaxies 23
MS0451a A851aA1689b A1835aA2218b A2218cA2219a A2261a f ν ( m Jy ) observed wavelength ( µ m) A2390b 0.1 1 10 100 10 100 1000 f ν ( m Jy ) observed wavelength ( µ m) A2390c
Fig. 6.—
Photometry and L(TIR) fits, continued.
Fig. 7.—
Comparison of L(8 µ m) and L(TIR) locally and for stacked measurements at z ∼
2. The individual local galaxies are indicatedas small blue squares, while the values averaged over L(TIR) intervals of 0.5 dex are shown as triangles. The sizes of triangles indicate theuncertanties in the average values for the bins, estimated from the scatter in the values. A fit to the averages is shown as a blue line. Theblack diamonds and line show a similar fit after rejecting high and low outliers in the luminosity bins (see text). The stacked results at z ∼ z ∼ L ⊙ . Fig. 8.—
Comparison of L(8 µ m) and L(TIR) locally and for galaxies at z ∼ µ m)/L(TIR) are indicated as blue squares. The blue line is a fit to them. The high-z galaxies from Yan et al. (2007) areshown as red circles, filled for those detected at 1.4GHz and open for the undetected example. The filled triangles are the lensed galaxiesfrom this work. The errors have been consolidated into L(TIR). They are shown as a factor of 1.4 for the Yan et al. (2007) galaxies,equivalent to the ratio of flux densities at 60 µ m to 1.4GHz varying from 100 to 200. For the lensed galaxies, we show errors by a factor of1.5, based on the range of luminosities indicated by our template fits. The red line is a fit to the z ∼ L ⊙ . The green points are individual galaxies at z ∼ .
85, from Marcillac et al. (2006) and the green line is a fit tothose points constrained to agree with the local fit at 10 L ⊙ . RS spectra of lensed galaxies 25 -1e-18 4.01e-18 4500 4550 4600 4650 4700 4750 4800 4850 4900 f l a m bda ( e r g / c m ^ / s / A ) observed wavelength (angstrom) Fig. 9.—
Keck/LRIS spectrum of A2390a. The strong feature at λ = 4690 ˚A is presumably Lyman α at z = 2 ..