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