The formation epoch of the galaxy red sequence Marc Balcells1 Instituto de Astrofı́sica de Canarias, 38200 La Laguna, Tenerife, Spain balcells@iac.es Summary. Galaxy number counts in the Ks, U and B bands have been used as inputs to galaxy formation models. We infer a typical formation redshift for earlytype galaxies of zf ∼ 1.5, corresponding to population ages of 9 Gyr, in excellent agreement with independent determinations of ages in field ellipticals. 1 Introduction Determining the formation epoch of elliptical galaxies has long been thought of as a key diagnostic on the validity of the two-stage galaxy formation theories based on CDM [22]. Indeed, if ellipticals formed at some point from mergers of smaller galaxies, then they should not be uniformly old. We live in fortunate times, since galaxies over the last 80% of the age of the Universe are now available for direct observation and study. Spectroscopic studies can be carried out with telescopes of the 10-m class, while direct imaging of galaxies at epochs interesting for elliptical galaxy formation can be carried out on 4-m class telescopes. We can therefore begin to address the detection of the formation phase of elliptical galaxies. The GTC, coupled to the EMIR NIR spectrograph, will undoubtedly make important contributions in this field. Observing in the NIR is essential for two reasons, first not to bias the sample selection to star-forming galaxies, and, second, to allow the mapping of the rest-frame optical spectrum, so that distant galaxies may be compared to nearby galaxies on the same set of observational parameters. An important breakthrough in our understanding of galaxy formation and evolution came from the realization that the color distribution of galaxies is bimodal. Analyzing thousands of galaxies from the Sloan Digital Sky Survey, Kauffmann et al. found that a number of age indicators, such as the 4000Å break, and the Hδ index showed a bi-modal distribution [16], defining two distinct sequences rather than a continuum. A similar bi-modal distribution was found for galaxy colors [15]. The red sequence, dominated by non-star 2 Marc Balcells forming galaxies, encompasses both ellipticals and lenticulars, while the blue cloud includes star-forming galaxies. The red sequence offers a means of addressing one aspect of the problem of dating of early-type galaxies. The emerging picture is one in which galaxies live on the blue cloud while they form stars, then eventually migrate to the red sequence when star formation ceases. Therefore, it suffices to trace the galaxy color distribution back in time. If the red sequence should disappear at a given redshift, we could naively conclude that the formation redshift for ellipticals has been found. In practice, we would have found the redshift at which galaxies became red, i.e., migrated from the blue cloud to the red sequence. Determining the redshift(s) at which galaxies migrated to the red sequence is an important astrophysical result. Work on optically-selected samples led to the conclusion that the red sequence is present at z = 1, and that the stellar mass in red galaxies may have roughly doubled between that epoch (∼8 Gyr ago) and the present [1]. Studying the red sequence beyond z = 1 faces the difficulty that imaging surveys utilizing CCDs, even in typical red bands such as the I-band, sample the UV continuum blue-ward of the 4000Å break, hence the resulting catalogs heavily penalize galaxies in which the light is dominated by old stellar populations. The problem does not come only from the uncertain K-corrections in UV bands, more important is that old objects fall altogether out of the optically-selected catalogs. The solution is simply, to work from NIR imaging surveys. NIR detectors have matured immensely in the past decade, in terms of higher array size, lower read-out noise, higher quantum efficiency and overall better uniformity [17]. Imaging in the 2.2 µ K-band samples the rest-frame optical continuum red-ward of 4000Å all the way up to z ∼ 3.9. At z = 2, the K-band samples the rest-frame I band, suitable for the detection and study of red populations. The GOYA project [13], in preparation for its spectroscopic mapping of the high-redshift galaxian populations, carried out a deep NIR and optical survey, the GOYA photometric survey [20]. Analysis of galaxy number counts from NIR and blue bands has allowed us to put constraints on the age when the red sequence appeared. 1.1 Data As part of the GOYA photometric survey, we imaged 180 arcmin2 of sky from the Groth field [12] and the Coppi [7] fields in the Ks-band. Exposures of 1.5-2 hr per pointing with the WHT/INGRID camera (Hawaii-1 1024×1024 Rockwell array), and seeing ranging from 0.6” to 1.0”, led to a depth (50% detection efficiency) of 20.6 to 21.0 Vega magnitudes. Galaxy detection and photometry were carried out with SExtractor [2]. Inferred counts were duly corrected from detection efficiency, spurious detections, foreground extinction and star counts. The counts are presented and analyzed in reference [8]. The final counts [8] are shown in Figure 1b. Galaxy Red Sequence 3 Fig. 1. Galaxy number counts in the Groth field. (a) B-band. (b) Ks-band. The latter include counts from the Coppi field as well. Ann abrupt change of slope in the Ks number counts is apparent. Lines depict the contribution of each galaxy class to the total counts (see text). We imaged a large area of sky around the Groth strip in U - and B-band with the INT/WFC camera. An area of 0.29 square degree was covered with exposures of 4 hr and 3 hr in U and B, respectively, reaching 50% detection efficiencies of 24.8 mag and 25.5 mag (Vega system). Corrections were applied to the counts as described in the previous paragraph. See [9] for details. The final B counts [9] are shown in Figure 1a. 2 Number count logarithmic slopes Inspection of Figures 1ab shows clear differences in the behavior of the blue and NIR counts. Blue counts closely follow a power-law extending over ∼14 mag up to B ∼ 25. In contrast, NIR counts show an abrupt slope change around Ks = 17.5. This feature had been reported previousy [11], and was confirmed by our group [8] thanks to the high product depth×field of our survey. On the bright side, measured slopes are γ ≡ d log N/d mag = 0.59. This is very close to γ = 0.60, the expected value for a uniform distribution of sources in an euclidean Universe. Faint-ward of the break, the slope is γ = 0.25. In the B band, slope is γB = 0.50 ± 0.03. Interpreting the slope change in the NIR number counts is the main driver of this paper. At first, one may easily interpret the result as saying that, beyond a given redshift galaxies were bluer [11]. If the slope change appeared in the blue bands, then a change in the star formation activity would be called for. Given that the change occurs in the NIR counts, something more 4 Marc Balcells fundamental must have occurred to the galaxian population: a decrease in numbers, or in stellar masses, back from a given redshift. 3 Galaxy evolution models To put our interpretation of the count slopes in a more quantitative footing, we ran number count models using the ncmod code [10]. ncmod simply evolves the local luminosity function (LF) back in time according to the prescriptions of population synthesis models. Details of the models are given in ref [9]. Briefly, the populations are anchored at z = 0 to the morphologicallydependent LF’s from SDSS [18], and evolved back in time using the 2003 Bruzual & Charlot models [5]. Four galaxy classes are considered, namely ellipticals and S0’s; early spirals; late spirals; and irregulars. A merger-driven, luminosity conserving number evolution is assumed, that scales as (1 + z)2.0 . Assuming extinction (τB = 0.6) for all galaxy classes is essential in order to fit the blue bands. Other modeling parameter have a minor contribution to the final results. Early-type galaxies dominate the Ks LF at z = 0 and therefore dominate the counts at low to intermediate z. They are likely to be responsible for the change in the slope of the NIR counts. Extensive testing led to the conclusion that the only way to reproduce the downward change in the Ks count slope is to delay the formation of a dominant population to moderate redshifts. While in U or B it is comparatively easy to break a power-law behavior by playing with extinction, or star formation activity, both of which strongly modify the UV continuum emission of galaxies, for the Ks counts only a drastic z-evolution of the galaxy numbers can break the power-law behavior of the counts, given the weak effects of dust extinction in Ks, the small Kcorrections and the slow sensitivity of M /LK to star formation episodes. Our model is able to reproduce the Ks = 17.5 slope change and the lack of such feature in the blue passbands by assuming a formation redshift of zf = 1.5 for the precursors of ellipticals and S0’s. The model predictions, shown in Figs. 1ab, successfully reproduce ranges of 16 mag in U (not shown), 18 mag in B, and 10 mag in Ks. The zf = 1.5 for early types gives the Ks = 17.5 knee; extinction prevents the knee from appearing in the blue passbands; and, number evolution gives the overall right slope; thanks to the assumed number evolution, we find no need to introduce a disappearing population of dwarfs at intermediate redshifts. 4 Formation of early types The knee in the Ks counts, if indeed related to a major epoch of formation of early types, constraints the time of such formation to 1 < z < 2. E.g., setting zf = 2.5 for ellipticals and S0’s leads to a clear mismatch with the data. Galaxy Red Sequence 5 Our result applies to the dominant elliptical population. Some elliptical formation at earlier times is compatible with the data, as is residual elliptical formation at z < 1, but the present modeling cannot put strong limits to these fractions, or on the presence of downsizing. Our zf = 1.5 result (age 9 Gyr) is in excellent agreement with recent inferences on the formation epoch of ellipticals. After correcting for ’progenitor bias’, van Dokkum & Franx infer zf = 2.0+0.3 −0.2 , which is consistent with our result [21]. ’Archaelogical’ dating of elliptical populations [19] leads to 1 < zf < 2 outside cluster environments (and zf = 3 to 5 in dense environments). Morphology studies also support our inferred formation epoch, given that morphologically-selected ellipticals are not found at z = 2 in the K20GOODSS sample [6]. In contrast, the formation redshift for early-type spirals is much less constrained by the data: zf in the range 1 to 4 yields moderately good matches to the U BKs counts. We obtain further clues on the zf for early-type spirals by looking at an observable that is independent from the counts – the color distributions. Simple evolution models such as ncmod are known to yield too narrow color distributions as compared to data [10]. Our modeling relates this problem to the assumption of a single formation epoch for each galaxy class. By assuming a range of zf from 1 to 4, models tend to reproduce the width of the observed color distributions. 5 Physical processes Because the slope in the bright K counts is so close to Euclidean, the lower slope at K > 17.4 may be ascribed to a period of growth of the stellar mass of galaxies, and the slope change to a near-cessation of star formation. After that and until the present, aging of the stellar population (with small effects in the NIR bands), is roughly compensated by residual amounts of star formation. Several processes may be responsible for the appearance of a red population at z ∼ 1.5. In a conservative interpretation, we are seeing the formation of red, early-type galaxies by mergers of spirals. In this view, most stars formed well before; only a fraction formed in a burst-like event triggered by the merger, which caused the enhanced α/Fe ratios typical of ellipticals. Alternatively, the slope change identifies the end of a period of very active star formation in galaxies, perhaps linked to mergers as well, which built up the stellar masses of all types of galaxies, including future ellipticals. What stopped star formation in galaxies set to become ellipticals? The role of mergers in quenching star formation is well known, although the exact mechanisms are not clear. They may include gas removal through starburstdriven galaxy-scale winds, although, energetically this may only work in lowmass systems. Alternatively, gas may be used up in star formation, if efficiencies are high, or heated to galaxian virial temperatures by gravitational energy of the merger or by AGN energy injection. Gas heating to virial temperatures 6 Marc Balcells is predicted for accretion onto halos when halo masses reach over a critical mass of about log(M ) ∼ 11.5 [3], a process that should operate irrespective of the growth mode of the DM halos. Acknowledgements M.-Carmen Eliche-Moral, David Cristóbal-Hornillos, Mercedes Prieto, C.Enrique Garcı́a-Dabo contributed to this research. My thanks to them and to other members of the GOYA team, including Lilian Domı́nguez, David Abreu, Carlos López, Marc Vallbé, Rafael Guzmán, Jesús Gallego, and Roser Pelló, for our ongoing collaboration on galaxy formation. References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. Bell, E. F., et al. 2004, ApJ, 608, 752 Bertin, E. & Arnouts, S. 1996, A&AS, 117, 393 Dekel, A., & Birnboim, Y. 2006, MNRAS, 368, 2 Blumenthal, G. R., Faber, S. M., Primack, J. R., & Rees, M. J. 1984, Nature, 311, 517 Bruzual, G. A., & Charlot, S. 1993, ApJ, 405, 538 Cassata, P., et al. 2005, MNRAS, 357, 903 Coppi, P. S., & Guzman, R. 2001, Two Years of Science with Chandra, Abstracts from the Symposium held in Washington, DC, 5-7 September, 2001 Cristóbal-Hornillos, D., Balcells, M., Prieto, M., Guzmán, R., Gallego, J., Cardiel, N., Serrano, A., & Pelló, R. 2003, ApJ, 595, 71 (CH03) Eliche-Moral, M. C., Balcells, M., Prieto, M., Garcı́a-Dabó, C. E., Erwin, P., & Cristóbal-Hornillos, D. 2006, ApJ, 639, 644 Gardner, J. P. 1998, PASP, 110, 291 Gardner, J. P., Cowie, L. L., & Wainscoat, R. J. 1993, ApJ, 415, L9 Groth, E. J., Kristian, J. A., Lynds, R., O’Neil, E. J., Jr., Balsano, R., Rhodes, J., & WFPC-1 IDT 1994, Bulletin of the American Astronomical Society, 26, 1403 Guzman, R. 2003, Revista Mexicana de Astronomia y Astrofisica Conference Series, 16, 209 Heavens, A., Panter, B., Jimenez, R., & Dunlop, J. 2004, Nature, 428, 625 Hogg, D. W., et al. 2004, ApJ, 601, L29 Kauffmann, G., et al. 2003, MNRAS, 341, 54 McLean, I. S. 1997, Electronic imaging in astronomy. Detectors and instrumentation, Publisher: Chichester, UK Wiley, 1997 Physical description xxx, 472 p. Series Wiley-PRAXIS series in astronomy and astrophysics Published in association with Praxis Publishing, Chichester ISBN0471969710, Nakamura, O., Fukugita, M., Yasuda, N., Loveday, J., Brinkmann, J., Schneider, D.P., Shimasaku, K., & SubbaRao, M. 2003, AJ, 125, 1682N Thomas, D., Maraston, C., Bender, R., & Mendes de Oliveira, C. 2005, ApJ, 621, 673 Vallbé, M. 2006, these proceedings. van Dokkum, P. G., & Franx, M. 2001, ApJ, 553, 90 White, S. D. M., & Rees, M. J. 1978, MNRAS, 183, 341