SPITZER IMAGING OF i -DROP GALAXIES: OLD STARS AT z 6. 1 Introduction

Transcription

1 SPITZER IMAGING OF i -DROP GALAXIES: OLD STARS AT z 6 Laurence P. Eyles 1, Andrew J. Bunker 1, Elizabeth R. Stanway, Mark Lacy 3, Richard S. Ellis, Michelle Doherty 5 1 School of Physics, University of Exeter, Stocker Road, Exeter, EX QL, U.K. Astronomy Department, University of Wisconsin-Madison, 75 N. Charter Street, Madison, WI 5376, U.S.A. 3 Spitzer Science Center,California Institute of Technology, Mail Code -6, 1 E. California Blvd., Pasadena, CA 9115, U.S.A. California Institute of Technology, Mail Stop , Pasadena, CA 9119, U.S.A. 5 Institute of Astronomy, Madingley Road, Cambridge, CB3 HA, U.K. Abstract. We present new evidence for mature stellar populations with ages > 1 Myr in massive galaxies (M stellar > 1 1 M ) seen at a time when the Universe was less than 1 Gyr old. We analyse the prominent detections of two z 6 star-forming galaxies (SBM3#1 & #3) made at wavelengths corresponding to the rest-frame optical using the IRAC camera onboard the Spitzer Space Telescope. We had previously identified these galaxies in HST/ACS GOODS images of Chandra Deep Field South through the i -drop Lyman break technique, and subsequently confirmed spectroscopically with the Keck telescope. The new Spitzer photometry reveals significant Balmer/ Å discontinuities, indicative of dominant stellar populations with ages > 1 Myr. Fitting a range of population synthesis models (for normal initial mass functions) to the HST/Spitzer photometry yields ages of 5 65 Myr and implied formation redshifts z f in presently-accepted world models. Remarkably, our sources have best-fit stellar masses of M (95% confidence) assuming a Salpeter IMF. This indicates that at least some galaxies with stellar masses > % of those of a present-day L galaxy had already assembled within the first Gyr after the Big Bang. We also deduce that the past average star formation rate must be comparable to the current observed rate (SFR UV 5 3 M yr 1 ), suggesting that there may have been more vigorous episodes of star formation in such systems at higher redshifts. Although a small sample, limited primarily by Spitzer s detection efficiency, our result lends support to the hypothesis advocated in our earlier analyses of the Ultra Deep Field and GOODS HST/ACS data. The presence of established systems at z 6 suggests long-lived sources at earlier epochs (z >7) played a key role in reionizing the Universe. A full analysis is presented in Eyles et al. (5). 1 Introduction The redshift range z 6 is of great importance, as it heralds the end of the reionization of the Universe (Becker et al. 1; Kogut et al. 3) which might be achieved through star formation. Using the Hubble Space Telescope (HST) and the new Advanced Camera for Surveys (ACS; Ford et al. 3), the Lyman-break technique (Steidel, Pettini & Hamilton 1995; Steidel et al. 1996; 1999) has been pushed to this early epoch (Stanway, Bunker & McMahon 3; Bouwens et al. ; Yan & Windhorst ; Giavalisco et al. ) by using the i and z filters to isolate i -drop galaxies. This technique utilises the rest-uv continuum break seen shortwards of Lyman-α caused by H I absorption in the intergalactic medium. Ground-based follow-up spectroscopy (Bunker et al. 3; Stanway et al. a,b; Dickinson et al. ) has shown that a colour cut of (i z ) AB > 1.5 mag reliably finds star-forming galaxies at z 6 with modest foreground contamination (primarily from low-mass stars, and passively evolving galaxies at z 1 ). Yet an important part of the puzzle is missing: the i -band drops are selected in the rest-frame UV, and are therefore known to be actively forming stars. However, it is unclear whether these objects suffer from significant reddening due to dust (in which case the star formation rates will have been underestimated), or if there is an underlying older stellar population which has been recently rejuvenated. Publicly-available Spitzer imaging with the Infrared Array Camera (IRAC; Fazio et al. ) as part of the Great Observatories Origins Deep Survey 1 (GOODS; Dickinson & Giavalisco 3; Dickinson et al. in prep) allows us to address both questions. A key benefit of Spitzer photometry arises because the IRAC camera samples wavelengths longwards of the age-sensitive Balmer & Å breaksatz 6. Accordingly, we have analysed the IRAC Super-deep Legacy Programme images of the GOODS-South field (centred on the Chandra Deep Field South, Giacconi et al. ), using filters at approximately λ cent =3.6 µm,.5 µm, 5.6 µm and8. µm (channels 1 ), and widths of λ FWHM =.68,.87, 1.5,.53 µm respectively. The total exposure time in each channel is 86 ksec, depending on location. We have already selected i -drop galaxies from the GOODS-South ACS images (Stanway, Bunker & McMahon 3) and the Ultra-Deep Field (UDF; Bunker et al. ). 1 see

2 redshifts, based on Lyman-α emission; two sources from Stanway, Bunker & McMahon (3): the brightest confirmed i -drop in the GOODS-South field (with z =.7mag) SBM3#3 (with a spectroscopic redshift of z =5.78 from Keck/DEIMOS, Bunker et al. 3); and the brightest i -dropintheudf,sbm3#1with z =5.3mag (spectroscopically confirmed with Keck/DEIMOS by Stanway et al. a as having a redshift of z =5.83). Both of these spectroscopic redshifts were independently confirmed by Dickinson et al. (). The other two sources come from the Gemini Lyman-α at Reionization Era (GLARE, Stanway et al. b) spectroscopy with Gemini/GMOS: GLARE#31 (z =5.79, z =6.mag) and GLARE#311 (z =5.9, z =7.mag). Throughout we adopt the standard concordance cosmology of Ω M =.3, Ω Λ =.7, and use H = 7kms 1 Mpc 1 in this cosmology, the Universe today is Gyr old, and at z =5.8 the age was 99 Myr. All magnitudes are on the AB system (Oke & Gunn 1983). Observations, Data Reduction and Stellar Population Fitting To construct spectral energy distributions of our spectroscopically-confirmed i -band drop galaxies in the GOODS-South field we undertook aperture photometry in the various images. At Spitzer resolution, these galaxies are clearly unresolved and so we treat them as point sources. Images of SBM3#1 in the various wavebands are shown in Figure 1. Checks were made to ensure the objects were not contaminated by the presence of any neighbouring bright foreground sources. SBM3#3 has a brighter neighbour that lies away; this source was subtracted using the GALFIT software (Peng et al. ). For the four IRAC channels, we derived 3σ limiting AB magnitudes of 6.5 and 6.1 in. apertures in channels 1 and, respectively, and 3.8 and 3.5 in 3. apertures in channels 3 and, respectively. In the short-wavelength channels, SBM3#1 & #3 are well-detected (> 1 σ, AB ), and we have a more marginal 3 σ detection of GLARE#31 (z AB =6.) at AB 6. Only SBM3#1 is detected at.5 µm SBM3#3 falls outside the region surveyed so far in this filter, and GLARE#31 is undetected. The fainter GLARE#311 (z AB =7.) was undetected at all IRAC wavelengths. For the HST photometry, we use the ACS i -band and z -band magnitudes from Stanway, Bunker & McMahon (3) from GOODS-South and, in the case of SBM3#1, from the deeper UDF (Bunker et al. ). These magnitudes are already corrected to a total flux, also through an aperture correction. We use NICMOS magnitudes (Thompson et al. 5) in F11W and F16W ( J and H band) from Stanway, McMahon & Bunker (5) for SBM3#1, and for the others we used v1. of the ESO VLT/ISAAC GOODS/EIS images in the J and K s bands (Vandame et al. in prep). The final step in the reduction process is the construction of spectral energy distributions (SEDs) for the chosen sources. In order to compare our photometry with stellar spectral synthesis models, we utilise the latest Bruzual & Charlot (3, hereafter B&C) isochrone synthesis code, with the Padova evolutionary tracks (preferred by B&C 3). Models with Salpeter (1955) initial mass functions (IMF) were selected, although we also consider the effect of adopting a Chabrier (3) IMF, and models of solar and 1/5th solar metallicities were explored. From several star formation histories available, a single stellar population (SSP an instantaneous burst), a constant star formation rate (SFR), and exponentially decaying (τ) SFR models were used. For each of the four i -drops with spectroscopic redshifts, the filters were corrected to their rest-frame wavelengths by the appropriate redshift factor. The measured flux was folded through the filter transmission profiles, and the best-fit age model was computed by minimising the reduced χ, using the measured errors on the magnitudes. The number of degrees of freedom is the number of independent data points (magnitudes in different wavebands). The normalisation of each best-fit B&C model was returned by the fitting routine and then used to calculate the corresponding best-fit stellar mass. Although some of our data points (particularly from the HST/ACS imaging) have S/N > 1, we set the minimum magnitude error to be (mag) =.1 to account for flux calibration uncertainties. During the SEDfitting process, the i -band flux was ignored, as this band is prone to contamination due to Lyman-α forest absorption shortwards of Lyman-α (λ rest = 116 Å) and also emission line contamination due to Lyman-α itself. 3 Analysis For the two i -drops with the best Spitzer detections (SBM3#1 & #3) we find evidence of a significant Balmer/ Å break (Figures & 3): the brightening by a factor of in f ν from the near-infrared (.9. µm) to 3.6 µm implies a break amplitude of 1.7 (forf λ flux densities). The presence of a Balmer/Å break is suggestive of a system viewed a significant time after a major epoch of star formation. The z,j& K s colours are relatively flat in f ν, which favours the spectral break interpretation rather than dust reddening. The break amplitude of 1.7 is comparable to that seen by Le Borgne et al. (5) in massive post-starburst galaxies at much lower redshifts (z 1) from the Gemini Deep Deep Survey project. Indeed, the break amplitude in our two significant z 6 cases is only slightly less than the D (Bruzual 1983) observed at z in the Sloan Digital Sky Survey (e.g., Kauffmann et al. 3), with D In order to produce the Balmer/Å break amplitude at λ rest Å, most of the stellar mass probably formed well before the current starburst, most likely > 1 Myr previously. Hence the galaxies SBM3#1 & #3

3 than 1 Gyr old. The constraints for the stellar ages of GLARE#31 & 311 are weak, but they are consistent with younger stellar populations than SBM3#1 & #3, which are brighter and presumably more massive. We cover a range of star formation histories designed to bracket most plausible evolution scenarios. Only a subset of these star formation histories provided acceptable fits to our photometry. Our SED fitting shows that the broad-band colours can be fit with a variety of stellar ages/star formation histories. The lower limit on the age is > 1 Myr (from an SSP model) with the oldest allowed models comparable to the age of the Universe at z 6. Our best-fits to the broad-band photometry come from exponentially-decaying star formation histories, or a two-component model where only.5 5% of the stellar mass is forming in an ongoing starburst (Figure 3). These best-fit models have mean stellar ages of 6 6 Myr, which would require formation redshifts of z f In all our models, the stellar masses of SBM3#1 & 3 were > 1 1 M at 95% confidence ( σ). Our preferred models (τ =7 5Myr and the two-component stellar population) have stellar masses of 1 1 M for a Salpeter IMF. These masses are surprisingly large, supporting our contention that at least these two objects are well-established galaxies. The stellar mass is equivalent to % of that for a L galaxy today, using L r = 1.1 from the SDSS analysis of Blanton et al. (3) and taking M/L V 5 M /L (appropriate for a 1 Gyr old population from B&C models using a Salpeter IMF) to obtain M = M, comparable to the estimate of M = M from Cole et al. (1) for our adopted Salpeter IMF. When conducting the SED fitting, two metallicities were explored: solar (Z ) and a sub-solar model (.Z ). The ages and masses of the best-fit models were similar for both metallicities, although the sub-solar model returned slightly better fits to the data, with smaller reduced χ min values. We also considered whether the red optical-infrared colours (spanning the rest-frame UV to optical) could be attributable to dust reddening, by applying the reddening model of Calzetti (1997), appropriate for starburst galaxies. We find little evidence for substantial dust reddening in the detected starlight out to 5 µm (Figure ). Finally, we tested the effects on the derived stellar masses of the assumed initial mass function, by re-running the two-component stellar population model using a Chabrier, rather than Salpeter IMF. Both IMFs produced comparably good fits (near-identical χ values), the same ages, and near-identical burst fractions (for SBM3#1,.6% and.7% by mass for the Chabrier and Salpeter IMFs). The main difference came in the best-fitting total stellar mass: the Chabrier model produced a mass 3% less than the Salpeter IMF. This effect is mainly a mass re-scaling independent of star formation history, as the discrepancy primarily arises from the different mass fraction in long-lived low-mass stars. We can usefully compare the ongoing star formation rate, the stellar mass and implied age, to deduce whether our two selected z 6 galaxies are being seen during a fairly quiet or active period in their history. In our first analyses of i -drop galaxies in the GOODS ACS images (Stanway, Bunker & McMahon 3; Stanway et al. a) and the Ultra Deep Field (Bunker et al. ), based on the z -band magnitudes we inferred unobscured SFRs of 19.5&33.8M yr 1 for SBM3#1 & #3, after accounting for Lyman forest blanketting of D A.95 shortwards of Lyman-α. The fits of the B&C stellar synthesis models to the broad-band photometry provide estimates of the current SFR for a range of histories. However, a firm lower limit arises from our Keck/DEIMOS spectroscopy (Bunker et al. 3, Stanway et al. a), which revealed Lyman-α emission. Our spectroscopic lower limit of an SFR > 6 M yr 1 rules out the simple SSP model and those declining star formation rate models with the shortest decay times (τ = 1 3 Myr). Our favoured models are an exponentially decaying SFR with decay time τ 3 5 Myr, and two-component models with 1 % of the stellar mass created in an ongoing starburst which began 1 1 Myr prior to the epoch of observation. These all indicate similar SFRs to our original estimates of 3 M yr 1 from the z -band flux. However, the past-average SFR must actually be greater than the current SFR in order to build our 1 1 M galaxies, given the short time available prior to z 6(1Gyr). Indeed,toformthe 1 1 M of stars requires an average star formation rate of 5 1 M yr 1 over the redshift range z f = favoured for the formation of the old stellar component. Hence the past-average star formation rates of SBM3#1 & 3 are factors of 5greater than their current rates at the epoch of observation (z 5.8). Although our chosen galaxies are the brightest confirmed i -drops in the GOODS-South field and possibly unrepresentative, the present work does imply that in these galaxies at least there was a yet earlier vigorous phase of activity, possibly at z>1, which may have played a key role in reionizing the Universe. This may be consistent with the measurement of temperature-polarization correlation of the cosmic microwave background from the Wilkinson MAP satellite by Kogut et al. (3). Conclusions Our group previously identified and spectroscopically-confirmed z 6 galaxies through HST/ACS i -drop imaging and Keck/DEIMOS & Gemini/GMOS spectroscopy. Here, and in Eyles et al. (5), we have presented the first infrared detections of this population using Spitzer/IRAC. We have significant ( 1 σ, AB mag) detections at 3.6 µm of the Stanway, Bunker & McMahon (3) galaxies #1 and #3 (at z spec =5.83, 5.78), and a more marginal detection of the z spec =5.79 galaxy GLARE#31 (Stanway et al. b). We also detect SBM3#1 at.5 µm (SBM3#3 is outside the field of view of this.5 µm filter). OurSpitzer detection of SBM3#1 in the UDF has subsequently been independently confirmed by Yan et al. (5).

4 1 8.µm 3.6µm 5.6µm Ks band.5µm 1.6µm f ν / 1 3 erg/cm /s/hz µm z band i band λ obs / µm Figure 1: (Left) Images of SBM3#1 (z =5.83), taken with HST/ACS (i -andz -band); HST/NICMOS (1.1 µm F11W J-band, and 1.6 µm F16W H-band ), VLT/ISAAC K s -band (smoothed with a 3-pixel. 5 boxcar) and the Spitzer channels ( µm). Each panel is 8 across (a projected distance of 5 kpc). Figure : (Right) One of the best-fit B&C models for SBM3#3: an exponentially decaying star formation rate with τ = 5 Myr, viewed 6 Myr after the onset of star formation, and with a stellar mass of M. Flux density is in f ν units. The K s band magnitude (from VLT/ISAAC imaging) is anomalously faint..3 f ν / 1 3 erg/cm /s/hz x1 1 M sun 5Myr 1 Myr burst (.7%) reddening E(B V) burst const SFR λ obs / µm age / yr Figure 3: (Left) The best-fitting two-component stellar population model (Salpeter IMF) for SBM3#1: a dominant 5 Myr population of mass M, with some ongoing star formation activity (a burst for the last 1 Myr involving.7% of the stellar mass). Figure : (Right) SBM3#1: A plot showing the reddening E(B-V) values for the two limiting scenarios of an instantaneous burst model and a constant SFR model, for the case metallicity = Z. Contours are 68% confidence (solid line), 95% confidence (dashed line) and 99% confidence (dotted line) for reduced χ of 1,, 3.

5 (as would be expected in these rest-uv-selected objects). However, the preceding star formation history has not been explored until now. In the two best-detected galaxies, we have evidence of a significant Balmer/Å break, indicative of a prominent older stellar population which probably dominates the stellar mass. Exploring a range of population synthesis models indicates that the average stellar age is > 1 Myr; our best-fit models suggest preferred ages of 5 65 Myr for an exponentially-declining star formation rate (of decay time τ 7 5Myr) or a two-component model (with an ongoing starburst responsible for.5 5% of the total stellar mass). This implies formation epochs of z f for the galaxies SBM3#1 & #3. In all our models, the best-fit stellar masses are > 1 1 M, with 95% confidence masses of M. This indicates that at least some galaxies with stellar masses > % the mass of L galaxies today were already assembled within the first Gyr of the Universe. For these objects, the past average star formation rate is comparable to, or greater than the current SFR, implying that there may have been even more vigorous episodes of star formation at higher redshifts. These may have played a key role in reionizing the Universe, consistent with the earlier studies of Bunker et al. () and Egami et al. (5). References Becker, R.H., et al. 1, AJ, 1, 85 Beckwith, S., Somerville, R., Stiavelli, M. 3, STScI Newsletter vol issue Blanton, M.R., et al. 3, ApJ, 59, 819 Bouwens, R.J., et al., ApJL, 66, 5 Bruzual, G.A. 1983, ApJ, 73, 15 Bruzual, G.A., Charlot, S. 3, MNRAS, 3, 1 Bunker, A.J., et al. 3, MNRAS, 3L, 7 Bunker, A.J., Stanway, E.R., Ellis, R.S., McMahon, R.G., MNRAS, 355, 37 Calzetti, D. 1997, AJ, 113, 16 Chabrier, G. 3, PASP, 115, 763 Cole, S., et al. 1, MNRAS, 36, 55 Dickinson, M.E., et al., ApJ, 6L, 99 Dickinson, M., Giavalisco, M. 3, Massive Galaxies at Low and High Redshift, p3, ESO, Springer Egami, E., et al. 5, ApJL, 618, 5 Eyles, L.P., et al. 5, MNRAS, 36, 3 Fazio, G.G., et al., ApJS, 15, 1 Ford, H.C., et al. 3, SPIE, 85, 81 Giavalisco, M. 3, AAS,, 173 Giavalisco, M., et al., ApJL, 6, 13 Giacconi, R., et al., ApJS, 139, 369 Kauffmann, G., et al. 3, MNRAS, 31, 33 Kogut, A., et al. 3, ApJS, 18, 161 Le Borgne, D., et al. 5, submitted to ApJ, astro-ph/531 Oke, J.B., Gunn, J.E. 1983, ApJ, 66, 713 Peng, Ho, Impey, Rix., AJ, 1, 66 Salpeter E. E., 1955, ApJ, 11, 161 Stanway, E.R., Bunker, A.J., McMahon, R.G. 3, MNRAS, 3, 39 Stanway, E.R., McMahon, R.G., Bunker, A.J. 5, MNRAS (accepted, online early), astro-ph/3585 Stanway, E.R., et al. a, ApJ, 67, 7 Stanway, E.R., et al. b, ApJL, 6, 13 Steidel, C.C., Pettini, M., Hamilton, D. 1995, AJ, 11, 519 Steidel, C.C., Giavalisco, M., Pettini, M., Dickinson, M.E., Adelberger, K.L. 1996, ApJL, 6, 17 Steidel, C.C., Adelberger, K.L., Giavalisco, M., Dickinson, M.E., Pettini, M. 1999, ApJ, 519, 1 Thompson, R.I., et al. 5, AJ, 13, 1 Yan, H., Windhorst, R.A., ApJL, 61, 93 Yan, H., et al. 5, ApJ (accepted), astro-ph/57673