Deep Observations of Lyman Break Galaxies Max Pettini1, Charles C. Steidel2, Alice E. Shapley2, Kurt L. Adelberger2, 3 3 4 Alan F.M. Moorwood , Jean-Gabriel Cuby , Mark Dickinson , and Mauro Giavalisco4 1 Instituteof Astronomy,Madingley Road, Cambridge CB3 0HA, England 1 2 Palomar Observatory,California Instituteof Technology, MS 105–24, Pasadena, 0 CA 91125, USA 0 3 European Southern Observatory,Karl-Schwarzschild-Str.2, D-85748 Garching, 2 Germany; and Alonso de Cordova 3107, Santiago, Chile 4 Space Telescope Science Institute,3700 San Martin Drive,Baltimore, MD n 21218, USA a J 6 1 Abstract. We summarise the main results of recent work on the Lyman break galaxy population which takes advantage of newly commissioned instrumentation 1 on the VLT and Keck telescopes to push the detection of these objects to new v wavelengths and more sensitive limits. 1 4 5 2 1 Introduction 1 0 1 Given the specialist nature of this meeting, I shall concentrate on the most 0 / recentresultsinourstudyofLymanbreakgalaxies(LBGs),ratherthangive h abroadperspectiveofthehighredshiftgalaxypopulation;uptodatereviews p - of this subject can be found in the articles by Dickinson (2000), Ferguson et o al. (2000), and Pettini (2000). As we shall see, the work I am about to de- r t scribeuses‘deep’observationsofLBGswhichpushthecapabilitiesofpresent s instrumentation to its limits. a : Since the commissioning of the high resolution, near infrared spectro- v graphs on the VLT (ISAAC) and Keck (NIRSPEC) telescopes in mid-1999, i X we have been engaged in an extensive programme to record the rest-frame r optical spectra of the brightest Lyman break galaxies with the aim of com- a plementing and extending the information provided by their rest-frame UV spectraonwhichmostofourknowledgeoftheseobjectsrestsatthemoment. Apart from the surprises often associated with opening a new wavelength window, there are several obvious scientific motivations. TheluminosityoftheBalmerlines,primarilyHαandHβ,givesameasure ofthestarformationratewhichisdirectlycomparabletothevaluesdeduced inlocalsurveys.Furthermore,sincetheopticalemissionlinesandthefar-UV continuumdonotrespondtodustextinctiontothesamedegree,therelative 1 Proceedings of the ESO Deep Fields Symposium. To be published in ESO Astrophysics Symposia, ed. S. Cristiani (Berlin: Springer) 2 Max Pettini luminosity of a galaxy in these two tracers of star formation could in princi- ple be used as a reddening indicator. When integratedoveran entire galaxy, the widths of the nebular lines should reflect the velocity dispersion of the H II regions within the overall gravitational potential so that a kinematical massmaybededuced.ThisisnotpossibleintheUV,becausetheinterstellar absorptionlines are sensitive to smallamounts ofgas acceleratedto high ve- locities by energetic events such as supernova explosions and bulk outflows, while the stellar absorption lines from OB stars are intrinsically broad. Fi- nally,therearewellestablishedchemicalabundancediagnosticsbasedonthe relative strengths of nebular emission lines, primarily [O II], [O III] and Hβ. Abundancemeasurementsaremuchmoredifficultinthe UVwherethe more easily observedinterstellar lines are generallysaturated so that their equiva- lent widths depend mostly on the velocity dispersion of the gas and only to a lesser extent on the column density of the absorbing ions. With the large sample of Lyman break galaxies now available (nearly 1000 spectroscopically confirmed objects), it is possible to isolate redshifts which place the transitions of interest in gaps between the strong OH lines which dominate the near-IR sky; here the background is sufficiently dark for faint extragalactic work to become possible. Figure 1 shows an example of the quality of spectra which can be secured with a 2-3 hour integration with NIRSPEC or ISAAC. Together with observations from the literature, Fig.1. Exampleof a NIRSPECK-band spectrum of a Lyman break galaxy. The objectstargetedinoursurveytypicallyhaveK =21(ontheVegascale)andremain undetected in the continuum, although the emission lines are easily seen with 2-3 hourexposures. The dotted line shows the 1σ error spectrum. our survey consists of 19 LBGs drawn from the bright end of the luminosity Lyman Break Galaxies 3 function, from ∼ L∗ to ∼ 4L∗. Overall we find that galaxies observed are very uniform in their near-IR properties. The spectra are dominated by the emission lines, and the continuum is detected in only two objects, one of which—West MMD11—has an unusually red optical-to-infrared color with (R−KAB)= 2.72. In all cases [O III] is stronger than Hβ and [O II]. The line widths span a relatively small range, with values of the one dimensional velocity dispersion σ between 50 and 115km s−1. 2 Star Formation Rates and Dust Extinction The star formation rates deduced from the luminosity of the Hβ emission line agree within a factor of ∼ 2 with the values implied by the continuum luminosity at 1500˚A before any corrections for dust extinction are applied (see Figure 2). There is no trend in the present sample for the former to Fig.2. Comparison between the values of star formation rate deduced from the luminositiesintheHβ emissionlineandintheUVcontinuumat1500˚A.Thecolor (G−R)measurestheintrinsicUVspectralslopeafterstatisticalcorrectionforthe Lyα forest opacity. Note that in the SCUBA source West MMD11, which is also the reddest object in the present sample, the strength of Hβ relative to the UV continuum is typical of therest of thesample and SFRHβ ≃SFRUV. be larger than the latter, as may have been expected from the shape of all reddening curves which rise from the optical to the UV. Evidently, any such differentialextinctionmustbe smallcomparedwiththe uncertaintiesincali- brating these two different measuresof the SFR (Kennicutt 1998;Charlot& Longhetti 2001). This conclusion is in agreement with the results of similar recent studies of UV-selected star-forming galaxies at z ∼< 1 (e.g. Flores et al. 1999; Sullivan et al. 2000; Bell & Kennicutt 2001) and contradicts the 4 Max Pettini commonly held view that the Balmer lines are more reliable star formation indicators than the UV continuum—from our sample one would obtain es- sentially the same star formation rate density using either method. 3 Oxygen Abundance In five cases (four new ones and one previously published) we attempted to deduce values of the abundance of oxygen by applying the familiar R23 ([O II]+[O III]/Hβ) method which has been extensively used in local H II regions. We found that generally there remains a significant uncertainty, by upto1dex,inthe valueof(O/H)becauseofthe double-valuednatureofthe R23 calibrator. Thus, in the galaxies observed oxygen could be as abundant as in the interstellar medium near the Sun, or as low as ∼1/10 solar.While this degeneracy can in principle be resolved by measuring the [N II]/Hα ratio (and in the one case where this has proved possible—Teplitz et al. 2000—values of (O/H) near the upper end of the range are indicated), this option is not normally available for galaxies at z ≃ 3 because the relevant lines are redshifted beyond the K-band. Even so, it is still possible to draw Fig.3. Metallicity-luminosity relation for local galaxies, from the compilation by Kobulnicky & Koo (2001) adjusted to the H0 = 70km s−1 Mpc−1, ΩM = 0.3, ΩΛ =0.7cosmologyadoptedinthiswork.Theverticalbarinthebottomright-hand cornergivesanindicationofthetypicalerrorinlog(O/H).IntheSun12+log(O/H) = 8.83 (Grevesse & Sauval 1998). The box shows the approximate location of the Lyman break galaxies in our sample at a median z = 3.1. Like many local H II galaxies,LBGsareoverluminousfortheirmetallicity.Theheightoftheboxresults largely from the double-valued nature of the calibration of (O/H) in terms of the R23 index; the one case where the ambiguity can be resolved (MS 1512-cB58) lies in theupperhalf of thebox. Lyman Break Galaxies 5 someinterestingconclusions.First,LBGsaredefinitelymoremetal-richthan damped Lyα systems at the same epoch, which typically have metallicities Z ≈ 1/30Z⊙. This conclusion is consistent with the view that DLAs are drawnpreferentiallyfromthefaintendofthegalaxyluminosityfunctionand are not the most actively star forming galaxies, as indicated by essentially all attempts up to now to detect them via direct imaging. Second, LBGs do not conform to today’s metallicity-luminosity relation and are overluminous for their oxygen abundance (see Figure 3). This is probably an indication that they have relatively low mass-to-light ratios, as also suggested by their kinematicalmasses;anadditionalpossibilityisthatthewhole(O/H)vs.MB correlationshiftstolowermetallicitiesathighz,whengalaxieswereyounger. 4 Kinematical Masses If the emission line widths reflect the relative motions of H II regions within thegravitationalpotentialofthegalaxies,theimpliedmassesareoftheorder 10 of 10 M⊙ within half-light radii of ∼ 2.5kpc. This is likely to be a lower limit to the total masses of the galaxies as would be obtained, for example, if we could trace their rotation curves. A more serious uncertainty, however, is the real origin of the velocity dispersions we measure. We do not see any Fig.4. One dimensional velocity dispersion of nebular emission lines in Lyman break galaxies as a function of absolute magnitude in the rest-frame far-UV (left) and B-band(right). Thevertical barshows thetypicalerror on themeasurements of σ.Curiously, theSCUBA source West MMD11, which has thereddest (R−K) color in thepresent sample, exhibitsone of thesmallest velocity dispersions. correlationbetweenσandgalaxyluminosityineitherourlimitedsample(see Figure 4) nor in an on-going study by some of us of a much larger sample of galaxies at z ≃ 1 which span five magnitudes in luminosity and yet show 6 Max Pettini very similar line widths to those found here. In two cases we have found hints of orderedmotions in spatially resolvedprofiles ofthe [O III] lines, but attempts to use highresolutionimagesto clarify whether they areindicative of rotating disks proved to be inconclusive. 5 Galactic Superwinds Inallthe galaxiesobservedwe findevidence forbulk motionsofseveralhun- dredkms−1fromthevelocitiesoftheinterstellarabsorptionlines—whichare systematically blueshifted—and Lyα emission—which is always redshifted— relativetothenebularemissionlines(seeFigure5).Weinterpretthiseffectas Fig.5. Velocity offsets of theinterstellar absorption lines (blue ordark grey) and of the Lyα emission line (red or light grey) relative to [O III] and Hβ. Large scale motions of the order of several hundred km s−1 are indicated by the systematic tendency for the former to be blueshifted and the latter redshifted relative to the nebular emission lines. indicativeofgalaxy-wideoutflowswhichappeartobe acommoncharacteris- ticofgalaxieswithlargeratesofstarformationperunitareaathigh,aswell as low,redshifts (e.g. Heckman2000).Such‘superwinds’ involvecomparable amounts of matter as is being turned into stars (the mass outflow rate is of thesameorderasthestarformationrate)andabout10%ofthetotalkinetic energy delivered by the starburst (Pettini et al. 2000). Furthermore, they have a number of important astrophysical consequences. They provide self- regulationtothestarformationprocess(Efstathiou2000);candistributethe products of stellar nucleosynthesis over large volumes (the outflow speeds Lyman Break Galaxies 7 often exceed the escape velocities—Ferrara et al. 2000); may account for some of the ‘missing’ metals at high redshift (Pettini 1999; Pagel2000);and may also allow Lyman continuum photons to leak from the galaxies into the intergalactic medium, easing the problem of how the universe came to be reionized. 6 The Contribution of Galaxies to the Ionising Background Threeofus(Steideletal.2001)haverecentlyaddressedthis lastpointquan- titatively by adding together the spectra of 29 Lyman break galaxies at a mean redshift hzi = 3.40±0.09 which places the Lyman continuum region at wavelengths accessible to LRIS. The composite spectrum, reproduced in Figure 6, shows a positive signal at the 4.8σ level between rest-frame wave- lengths 880 and 910˚A. After allowance for the opacity of the intergalactic Fig.6. Composite spectrum of 29 Lyman break galaxies at hzi = 3.40. Note the average residual flux in the Lyman continuum between 880 and 910˚A in the rest frame. The most prominent interstellar (blue and above the zero level) and stellar (green and below thezero level) absorption lines are identified. medium, even this weak signal implies a surprisingly large escape fraction— the emergent flux density below the Lyman limit is 20% of that at 1500˚A, 8 Max Pettini which has been well quantified by extensive surveys. This may well be an upper limit to the value appropriate to the whole population because the galaxieswhich contributed to the composite spectrum in Figure 6 are drawn from the bluest quartile of our full sample and may thus the most ‘leaky’ objects among a spread of properties. On the other hand, the ubiquitous presenceofsuperwindsinLymanbreakgalaxiessuggeststhattheirinterstel- larmediamaycommonlydeveloplargecavitieswhichcanprovidearoutefor the Lyman continuum photons to escape from the regions of star formation into the IGM. If the escape fraction (as defined above) of 20% applies generally, the contribution of galaxies to the metagalactic radiation field at z ≃ 3.5 is Jν ≈1.2±0.3×10−21ergs s−1 cm−2 Hz−1 sr−1. This value is close to most estimates of Jν based on the proximity effect in the Lyα forest, and exceeds the contributionfrom QSOs by about a factor of 5.Presumably,the balance between galaxies and QSOs as the providers of Lyman continuum photons shifts further in favour of the former as we move to even higher redshifts (Steidel et al. 1999; Fan et al. 2001), making it quite plausible that stars, rather than AGNs, were responsible for the end of the ‘dark ages’. This conclusion, if supported by data soon to be secured with the new UV-sensitivearmofLRIS,bodeswellforthe‘VeryDeepFields’ whicharethe goaloftheAdvancedCameraforSurveysonHST andeventuallytheNGST. Max Pettini would like to express his gratitude to the organisers of this enjoyable and timely meeting. References 1. Bell, E.F. & Kennicutt Jr., R.C. 2001, ApJ,in press (astro-ph/0010340) 2. Charlot, S. & Longhetti, M. 2001, MNRAS,in press (astro-ph/0101097) 3. Dickinson, M. 2000, Philos. Trans. R. Soc. Lond. A,358, 2001 4. Efstathiou, G. 2000, MNRAS,317, 697 5. Fan, X., et al. 2001, AJ, in press (astro-ph/0008123) 6. Ferguson, H.C., Dickinson, M., & Williams, R.2000, ARA&A,38, 667 7. Ferrara, A., Pettini, M., & Shchekinov,Y. 2000, MNRAS,319, 539 8. Flores, H.,et al. 1999, ApJ, 517, 148 9. Grevesse, N., & Sauval, A.J. 1998, SpaceSci Rev,85, 161 10. Heckman,T.M.2000,inASPConf.Ser.,GasandGalaxyEvolution,ed.J.E.Hi- bbard,Mp.P.Rupen,&J.H.vanGorkom,(SanFrancisco:ASP),inpress(astro- ph/0009075) 11. KennicuttJr., R.C. 1998, ARA&A,36, 189 12. Kobulnicky,H.A., & Koo, D.2001, ApJ,in press (astro-ph/0008242) 13. Pagel, B.E.J. 2000, in Galaxies in the Young Universe, ed. H. Hippelein (Berlin:Springer-Verlag), in press (astro-ph/9911204) 14. Pettini, M. 1999, in Chemical Evolution from Zero to High Redshift, ed. J.R. Walsh, & M.R. Rosa (Berlin:Springer-Verlag), 233 15. Pettini, M. 2000, Philos. Trans. R. Soc. Lond. A,358, 2035 Lyman Break Galaxies 9 16. Pettini, M., Steidel, C.C., Adelberger, K.L., Dickinson, M., & Giavalisco, M. 2000, ApJ,528, 96 17. Steidel, C.C., Adelberger, K.L., Giavalisco, M., Dickinson, M., & Pettini, M. 1999, ApJ,519, 1 18. Steidel, C.C., Pettini, M., & Adelberger, K.L. 2000, ApJ, 546 in press 19. Sullivan,M., Treyer,M.A.,Ellis, R.S.,Bridges, T.J., Milliard, B.,& Donas,J. 2000, MNRAS,312, 442 20. Teplitz, H.I., et al. 2000, ApJ,533, L65