The DEEP2 Galaxy Redshift Survey: Mean Ages and Metallicities of Red Field Galaxies at z ~ 0.9 from Stacked Keck DEIMOS Spectra

Abstract

As part of the DEEP2 galaxy redshift survey, we analyze absorption line strengths in stacked Keck DEIMOS spectra of red field galaxies with weak to no emission lines, at redshifts 0.7 ≤ z ≤ 1. Comparison with models of stellar population synthesis shows that red galaxies at z ~ 0.9 have mean luminosity-weighted ages of the order of only 1 Gyr and at least solar metallicities. These ages cannot be reconciled with a scenario in which all stars evolved passively after forming at very high z. Rather, a significant fraction of stars can be no more than 1 Gyr old, which means that some star formation in the stacked populations continued to at least z ~ 1.2. Furthermore, a comparison of these distant galaxies with a local SDSS sample, using stellar population synthesis models, shows that the drop in the equivalent width of Hδ from z ~ 0.9 to 0.1 is less than that predicted by passively evolving models. This admits two interpretations: either each individual galaxy experiences continuing low-level star formation, or the red-sequence galaxy population from z ~ 0.9 to 0.1 is continually being added to by new galaxies with younger stars.

Full text

Page 1

arXiv:astro-ph/0602248v2 13 Feb 2006
Version: Feb 9, 2006
Preprint typeset using LATEX style emulateapj v. 6/22/04
THE DEEP2 GALAXY REDSHIFT SURVEY: MEAN AGES AND METALLICITIES OF RED FIELD GALAXIES
AT Z ∼ 0.9 FROM STACKED KECK/DEIMOS SPECTRA1
Ricardo P. Schiavon2, S. M. Faber3, Nicholas Konidaris3, Genevieve Graves3, Christopher N.A. Willmer3,4,
Benjamin J. Weiner5, Alison L. Coil4,6,8, Michael C. Cooper6, Marc Davis6, Justin Harker3, David C. Koo3,
Jeffrey A. Newman6,7,8& Renbin Yan6
2Department of Astronomy, University of Virginia, P.O. Box 3818, Charlottesville, VA 22903-0818
3UCO/Lick Observatory/Department of Astronomy and Astrophysics, University of California, Santa Cruz, CA 95064
4Steward Observatory, University of Arizona, Tucson, AZ 85721
5Department of Astronomy, University of Maryland, College Park, MD 20742-2421
6Department of Astronomy, University of California, Berkeley, 601 Campbell Hall, Berkeley, CA 94720-3411
7Lawrence Berkeley National Laboratory, Berkeley, CA 94720 and
8Hubble Fellow
Version: Feb 9, 2006
ABSTRACT
As part of the DEEP2 galaxy redshift survey, we analyze absorption line strengths in stacked
Keck/DEIMOS spectra of red field galaxies with weak to no emission lines, at redshifts 0.7 ≤ z ≤
1. Comparison with models of stellar population synthesis shows that red galaxies at z ∼ 0.9 have
mean luminosity-weighted ages of the order of only 1 Gyr and at least solar metallicities. This result
cannot be reconciled with a scenario where all stars evolved passively after forming at very high z.
Rather, a significant fraction of stars can be no more than 1 Gyr old, which means that star formation
continued to at least z ∼ 1.2. Furthermore, a comparison of these distant galaxies with a local SDSS
sample, using stellar populations synthesis models, shows that the drop in the equivalent width of
Hδ from z ∼ 0.9 to 0.1 is less than predicted by passively evolving models. This admits of two
interpretations: either each individual galaxy experiences continuing low-level star formation, or the
red-sequence galaxy population from z ∼ 0.9 to 0.1 is continually being added to by new galaxies with
younger stars.
Subject headings: Galaxies: evolution — Galaxies: stellar content — Galaxies: distances and redshifts
1. INTRODUCTION
The formation of early-type galaxies is one of the on-
going riddles of modern extragalactic astrophysics. Ac-
cording to the leading models, massive early-type galax-
ies have been assembled hierarchically, from the merg-
ing of less massive structures.
are seen locally to be accompanied by star formation
(e.g., Schweizer & Seitzer 1992), one of the best ways
to test the hierarchical formation paradigm is by deter-
mining the star formation history of early-type galax-
ies. This can be achieved by estimating the ages of stars
in galaxies from their integrated light, through compar-
ison with stellar population synthesis models. Several
groups have attempted to achieve this goal from obser-
vations of massive galaxies in a range of redshifts (e.g.,
Le Borgne et al. 2006, Treu et al. 2005, Daddi et al. 2005,
Longhetti et al. 2005, and references therein). However,
spectroscopic dating of stellar populations older than ∼
1 Gyr is best achieved by simultaneously matching the
strengths of Balmer and metal lines in their integrated
spectra, in order to avoid spurious effects due to the age-
metallicity degeneracy. So far, observational difficulties
have prevented such detailed studies for all but local sam-
ples (e.g., Gonz´ alez 1993, Trager et al. 2000, Kuntschner
2000, Caldwell et al. 2003, Eisenstein et al. 2003, Thomas
et al. 2005, Schiavon 2006, and references therein).
In this Letter we present the analysis of absorption line
strengths measured in stacked integrated Keck/DEIMOS
spectra of red galaxies with redshifts between 0.7 and 1,
Because such mergers
1Based on observations taken at the W. M. Keck Observatory
as part of the DEEP2 survey (Davis et al. 2003). We
find that the stars in these galaxies have mean light-
weighted single stellar population (SSP) ages of order
only 1 Gyr, and their metallicities are at least solar. Since
these objects are observed several billion years after the
big bang, this result suggests that stars inhabiting red
galaxies were formed during an extended period of time.
2. SAMPLE AND DATA
The data used in this Letter consist of k-corrected ab-
solute magnitudes in the Vega system and 1-hour ex-
posure Keck/DEIMOS (Faber et al. 2003) spectra from
DEEP2 (Davis et al. 2003). Redshift determinations are
described in Davis et al. (2003), and restframe MBmag-
nitudes and U–B colors were derived from CFHT BRI
photometry and redshifts by Willmer et al. (2006). The
S/N of each 1-hour exposure spectrum is not high enough
for accurate measurement of absorption line indices, so
we stack spectra of hundreds of galaxies, selected in bins
of color, luminosity, and redshift. Further details can be
found in Schiavon et al. (in preparation).
2.1. Sample Selection
Our goal is to study the evolution of red-sequence
early-types, so we first select galaxies by color.
selection criterion is illustrated in the left panel of Fig-
ure 1, where data for 17,745 DEEP2 galaxies with 0.7 ≤
z ≤ 1.05 were used to produce a contour plot on the
restframe color-magnitude space. Red-sequence galaxies
(RSGs) are chosen to be those with U–B ≥ 0.25, making
up a total of 1941 objects. Ideally, we would also like to
This

Page 2

2 Schiavon et al.
Fig. 1.— Left panel: Restframe color-magnitude diagram of
17,745 DEEP2 galaxies. Our sample of RSGs is defined by U–B
> 0.25 (objects above the dotted line). Right panel: A histogram
of [OII] λ 3727 EWs for RSGs. No [OII] emission happens when
EW[OII]=3.7˚ A (dashed line) and lower values indicate the presence
of line emission. All galaxies with EW[OII] ≤ –5˚ A are excluded
from the sample.
select galaxies on the basis of morphology, but we un-
fortunately lack that information for the sample under
analysis. Therefore, in order to minimize contamination
by strongly reddened late-type galaxies we impose an-
other cut, based on the equivalent width (EW) of [OII]
3727 (see definition by Fisher et al. 1998). This is illus-
trated in the right panel of Figure 1, where a histogram
of [OII] EWs is shown for all RSGs in our sample. Strong
emission-line RSGs have very negative values of EW[OII]
while the zero of [OII] emission is at ∼ 3.7˚ A (Konidaris
et al. in preparation). The distribution is strongly peaked
at very low [OII] emission values with a long tail towards
galaxies with strong [OII] emission. Contamination by
reddened late-type galaxies is likely to be more impor-
tant in the strong-emission line regime, so that we re-
move from our sample all galaxies with [OII] EW ≤ –5˚ A.
This emission-line cut admittedly leaves in our sample a
large number of galaxies in the low-emission line regime.
These are mostly AGN on the basis of the ratios between
[OII] and residual Balmer line emission (Schiavon et al.
in preparation, Konidaris et al. in preparation), as has
been also found for low redshift RSGs (Phillips et al.
1986, Rampazzo et al. 2005, and Yan et al. 2006).
The color and emission-line cuts leave us with a sam-
ple of 1160 galaxies. We create six subsamples out of
this set of galaxies, three with varying colors but the
same redshift range, and three with varying redshifts,
but consistent colors and luminosities. The color and
redshift limits of each bin are listed in Table 1. In an
attempt to compare objects with similar masses, galax-
ies in the color and redshift sub-samples were further
selected within 1 mag-wide MBintervals where the cen-
tral magnitude was chosen to be consistent with passive
evolution from the age and metallicity of the high-z sam-
ple. However, adopting the exact same MB interval for
each z bin does not change the results. The numbers of
galaxies in each bin are listed in Table 1.
2.2. Stacked Spectra and Lick Indices
The 1160 galaxy spectra were visually inspected in or-
der to clean the sample from a few misclassified stars,
galaxies with wrong redshifts, and zero-S/N spectra. A
rough relative fluxing was achieved by dividing each spec-
trum by the normalized throughput of the DEIMOS
spectrograph with the 1200 l/mm grating. Before coad-
dition, the spectra were brought to restframe and then
normalized through division by the average (σ-clipped)
counts within the λλ 3900–4100˚ A interval. Coaddition
was performed adopting a σ-clipping procedure to elim-
inate sky-subtraction residuals, zero-count pixels due to
CCD gaps, and other spectral blemishes. After several
tests the best results were obtained when a single 3-σ
clipping iteration was adopted. On average more than
90% of all galaxies in a given bin contribute to the stacked
spectrum at any given wavelength. No clipping was per-
formed in the region of the [OII] λ 3727˚ A line. In Fig-
ure 2a we compare one of our stacked spectra with a
SSP model from Schiavon (2006).
the overall flux distribution of the theoretical spectrum,
the observed spectrum was dereddened by E(B–V)=0.2.
Since the observations were not properly flux-calibrated,
this E(B-V) value does not reflect the average redden-
ing in the sample galaxies and this correction has the
sole purpose of bringing observations and theory to a
common relative scale so as to highlight the outstanding
agreement between line strengths in the observed and
synthetic spectra.
All Lick indices in the λλ 4000–4500˚ A region were
measured in the stacked spectra, but we focus here on
the HδF and Fe4383 indices, which are chiefly sensitive
to age and [Fe/H], respectively. The spectra were first
broadened to the Lick resolution as given by Worthey &
Ottaviani (1997), and the indices were measured follow-
ing definitions by those authors and by Worthey et al.
(1994). Velocity dispersions (σ) were measured in the
stacked spectra through Fourier cross correlation using
the IRAF rv.fxcor routine. The template adopted was
a model spectrum from Schiavon (2006) for a SSP with
solar metallicity and an age of 2 Gyr. The same model
spectrum was used to infer corrections to the line indices
for the effect of σ-broadening. The indices were all cor-
rected to σ = 0 km/s using the σ determined for each
stacked spectrum. The latter are listed in Table 1. We
do not attempt to convert the line indices to the Lick
system, aside from smoothing them to the Lick/IDS res-
olution. However, zero-point differences should be very
small, given that the Schiavon (2006) models are based
on fluxed spectra and the DEEP2 spectra are corrected
from instrumental throughput.
were corrected for emission-line in-fill, which was esti-
mated from EW[OII], adopting EW[OII]/EW(Hα) = 6
(Yan et al. 2006) and EW(Hδ) = 0.13 EW(Hα). The
correction to HδF is smaller than 0.2˚ A, corresponding
to less than 1 Gyr in age.
In order to match
Finally, Balmer lines
2.3. Local Sample
Galaxy evolution is better assessed when distant and
local samples of similar objects are contrasted using evo-
lutionary models. Moreover, it is vital that the nearby
and distant samples are defined as consistently as pos-
sible, to ensure that the two samples represent objects
of the same class. For a local counterpart to the dis-
tant DEEP2 sample we use the SDSS data from Eisen-
stein et al. (2003), who provide stacked flux-calibrated
spectra of RSGs, binned by absolute magnitude, envi-
ronment, and redshift. Because the DEEP2 stacked spec-

Page 3

Red Galaxies at z ∼ 0.93
TABLE 1
Data for galaxies used in stacked DEEP2 spectra
Bin
zMB/Mr
U − BHδF
Fe4383
σ (km/s)N
Low z
Intermediate z
High z
Red
Intermediate
Blue
SDSS - Lum 215
SDSS - Lum 210
[0.7,0.8]
[0.8,0.9]
[0.9,1.0]
[0.75,0.95]
[0.75,0.95]
[0.75,0.95]
0.171
0.143
[–21.57,–20.57]
[–21.70,–20.70]
[–21.86,–20.86]
[–21.76,–20.76]
[–21.76,–20.76]
[–21.76,–20.76]
[–22.0,–21.5]
[–21.5,–21.0]
[0.25,0.60]
[0.25,0.60]
[0.25,0.60]
[0.45,0.60]
[0.35,0.45]
[0.25,0.35]
–
–
1.8±0.1
1.8±0.1
1.8±0.2
1.6±0.2
1.8±0.1
2.3±0.2
0.66± 0.01
0.82± 0.01
3.7±0.5
3.0±0.2
—
3.7±0.3
3.2±0.2
2.4±0.3
4.32±0.02
4.22±0.02
190
170
180
190
170
170
235
210
113
288
167
129
228
119
5412
6477
Note.
measurements taken in stacked spectra or, in the case of U–B for DEEP2 and z for SDSS, average values within a given bin. N is the
number of galaxies in a bin. Absolute magnitudes are MBfor DEEP2 and Mr for SDSS.
— Numbers in brackets correspond to intervals adopted to select galaxies in different bins.Single numbers correspond to
tra include galaxies from all environments, we chose to
use stacked spectra from a similarly defined sample from
Eisenstein et al. (their “All” sample). Furthermore, in
order to match the relative position of our absolute mag-
nitude bins along the red sequence, we chose to exclude
both the lowest and highest magnitude bins of the Eisen-
stein et al. sample from our analysis. The spectra were
downloaded from D. Eisenstein’s website and submitted
to the same treatment as described above for the DEEP2
spectra. Key data for the Eisenstein et al. (2003) sample
are listed in Table 1
3. MEAN AGES AND METALLICITIES
Lick indices measured in the stacked spectra are com-
pared with SSP models in Figure 2b. Shown are the
indices measured in Eisenstein et al. (2003) spectra and
those from the color-selected DEEP2 sub-samples. Sym-
bol size for the Eisenstein et al. galaxies is propor-
tional to luminosity. The data are compared to models
computed adopting [α/Fe] = +0.4, [C/Fe]=+0.15, and
[N/Fe]=+0.3 (See Schiavon et al. in preparation for de-
tails).
The main result of this Letter is immediately apparent
in Figure 2b. The stellar populations of field RSGs at
z ∼ 0.9 are young, with a mean luminosity-weighted age
of only ∼ 1.2 Gyr. Adoption of solar-scaled models would
imply older ages by roughly 0.5 Gyr. Iron abundances
range from [Fe/H] ∼ 0 to ∼ +0.3. As expected, the stellar
populations in the local SDSS galaxies are older than
those at z ∼ 0.9, with mean ages ranging between 3 and
6 Gyr, with above solar [Fe/H]. The time difference from
the characteristic redshifts of DEEP2 to SDSS samples is
over 5 Gyr, so the expected mean ages of SDSS galaxies
under passive evolution should be over 6 Gyr. We note
that the Schiavon (2006) models match the same data
for Galactic globular clusters and—most importantly for
this study—those for the open cluster M 67 (3.5 Gyr,
[Fe/H]=0), to within 1 Gyr in age and 0.1 dex in [Fe/H].
Another interesting way of viewing this result is illus-
trated in Figure 2c, where DEEP2 and SDSS galaxies are
compared with passively evolving SSP models in a red-
shift vs. HδF plot. The DEEP2 data plotted in this case
come from the redshift-selected sub-sample (see Table 1).
The lines correspond to model predictions for SSP evo-
lution assuming various redshifts of formation (zform)
and adopting a concordance WMAP cosmology (Spergel
et al. 2003). Dashed (solid) lines represent models with
[α/Fe] = 0.0 (+0.4) and super-solar [Fe/H]. Figure 2c
shows that the data for field RSGs require zform∼ 1.1–
1.3, when they are modeled using SSPs with super-solar
metallicity. Moreover, the distant and local samples are
not connected by lines of passive evolution, which indi-
cates the occurrence of star formation between z ∼ 0.9
to ∼ 0.1 (see also Gebhardt et al. 2003). This could be
due either to in situ star formation or to the incorpora-
tion in the red sequence of new galaxies coming from the
blue cloud after cessation of star formation. Figure 2c
also makes clear that these results are valid regardless of
the [α/Fe] ratio of the models adopted. The results also
remain qualitatively unchanged when models with lower
metallicity are used. Adoption of solar metallicity mod-
els (not shown) would result in slightly higher redshifts
of star formation, zform∼ 1.3–1.5.
4. CONCLUSIONS AND CAVEATS
The results presented in this Letter need to be inter-
preted with caution. Ages and metallicities inferred from
comparison of integrated galaxy spectra with SSP models
are luminosity-weighted averages, whereas the real stel-
lar populations doubtless consist of stars with a range of
ages. Therefore, we are not claiming that the stars in
the DEEP2 galaxies plotted in Figure 2b are uniformly
∼ 1.2 Gyr old. Likewise, we are not proposing that the
mean galaxies plotted in Figure 2c sprang into existence
at z ∼ 1.3. But the strength of Hδ in the stacked inte-
grated spectra of field RSGs indicates that they harbour
young and/or intermediate-age stars both locally and at
z ∼ 0.9. Since the universe was roughly 6 Gyr old at
z ∼ 0.9, this result cannot be reconciled with models in
which all the stars in these galaxies were formed at very
high redshifts (z > 3) and evolved passively ever since.
In fact, it appears that star formation in these galaxies
was prolonged, and that it continued, in small amounts,
between z ∼ 0.9 and 0.1. The fact that we see no evo-
lution in [Fe/H] seems to indicate that the bulk of star
formation has occurred before z ∼ 0.9.
The presence of young/intermediate-age stars in early-
type galaxies can be accounted for by at least two sce-
narios. The so-called frosting models (e.g., Trager et al.
2000) propose small amounts of recent in situ star forma-
tion originated these stars. On the other hand, quench-
ing models (e.g., Bell et al. 2004, Faber et al. 2005) sug-
gest that blue galaxies migrate to the red sequence after
cessation of star formation, possibly associated with a

Page 4

4 Schiavon et al.
Fig. 2.— a) Comparison between the stacked spectrum for the intermediate color bin (see Table 1) and that for a model SSP from
Schiavon (2006). The two absorption lines studied here are indicated. This plot illustrates the high S/N of the stacked spectra and also the
high quality of the models, which reproduce all the main absorption lines in the observed spectrum very well. b) Measurements from RSG
stacked spectra versus predictions of SSP models. Squares with error bars indicate DEEP2 galaxies and triangles indicate SDSS galaxies
from Eisenstein et al. (2003). For SDSS data, symbol size correlates with luminosity. DEEP2 galaxies are binned by color and have absolute
magnitude and redshifts within the intervals indicated in the label and in Table 1. The mean colors of DEEP2 bins are indicated to the
right of each data point. The models are those of Schiavon (2006) for [α/Fe] = +0.4. Same-age (same-metallicity) models are connected
by solid (dashed) lines. Model age ([Fe/H]) ranges from 1.2 to 14 Gyr (-0.8 to +0.3 dex), as indicated by the labels. The mean SSP ages of
DEEP2 galaxies are ∼ 1.2 Gyr, regardless of color. They are younger than the SDSS galaxies, as expected. Their [Fe/H]s range between
solar and +0.3. c) DEEP2 data from the redshift-selected sample compared with SSP models in a redshift vs. HδF plot. Predictions from
SSP models with super-solar metallicity and a range of formation redshifs, zform, are shown as solid ([Fe/H]=+0.2, [α/Fe]=0) and dashed
([Fe/H]=+0.3, [α/Fe]=+0.4) lines. DEEP2 galaxies are consistent with zformranging from 1.1 to 1.3 (1.3 to 1.5 if [Fe/H]=0 SSP models
are adopted). This indicates that star formation was probably prolonged in those galaxies. Galaxies at high and low z are not connected
by lines of passive evolution, which probably indicates that star formation did not cease from z ∼ 0.9 to the present day.
merger event and/or enhanced AGN activity. In a sep-
arate study (Harker et al. in preparation) we make an
exploration in that direction, by showing that quenching
models provide a good match to the data analyzed in
this Letter. Quenching models were also shown by Faber
et al. (2005) to provide a good match to the evolution of
the luminosity function of red galaxies from z ∼ 1.4 to
the present day. An estimate of the fraction of the total
stellar mass allocated in these young/intermediate-age
stars (e.g., Leonardi & Rose 1996) might help discrimi-
nate between the different scenarios.
A few important caveats must be kept in mind when
considering these results. Of most importance, our sam-
ple is not yet selected on the basis of morphology. Al-
though contamination by late-type, star-forming galaxies
is low (Konidaris et al. 2005 in preparation), a stronger
statement on the history of star formation of early-type
galaxies awaits an analysis of a morphologically selected
sample of distant objects (e.g., Treu et al. 2005). It is also
important to keep in mind that we are dealing with a field
sample. Previous studies of cluster samples at compara-
ble redshifts tell a different story, with cluster RSGs be-
ing compatible with high zform(e.g., Kelson et al. 2001).
Lastly, the local SDSS sample used here does not match
perfectly the selection criteria adopted for our DEEP2
sample. Work aimed at producing more adequate local
counterparts to our distant DEEP2 sample is currently
under way (Graves et al. 2005 in preparation).
This project was supported in part by NSF grants AST
00-71198 and AST 00-71408, and AST-0507483. R.P.S.
acknowledges financial support from HST Treasury Pro-
gram grant GO-09455.05-A to the University of Virginia.
R.P.S. thanks Bob O’Connell and Jim Rose for helpful
comments on an early version of the manuscript. J.A.N.
acknowledges support from NASA through Hubble Fel-
lowship grant HST-HF-01165.01-Aawarded by the Space
Telescope Science Institute, which is operated by the As-
sociation of Universities for Research in Astronomy, Inc.,
for NASA, under contract NAS 5-26555. We thank the
Hawaiian people for allowing us to conduct observations
from their sacred mountain.
REFERENCES
Bell, E. et al. 2004, ApJ, 608, 752
Caldwell, N., Rose, J.A. & Concannon, K.D. 2003, AJ, 125, 2891
Daddi, E. et al. 2005, ApJ, 626, 680
Davis, M. et al. 2003, SPIE, 4834, 161
Eisenstein, D. et al. 2003, ApJ, 585, 694
Faber, S.M. et al. 2003, SPIE, 4841, 1657
Faber, S.M. et al. 2005, ApJ, in press, astro-ph/0506044
Fisher, D., Fabricant, D., Franx, M. & van Dokkum, P. 1998, ApJ,
498, 195
Gebhardt, K. et al. 2003, ApJ, 597, 239
Gonz´ alez, J.J. 1993, PhD Thesis, University of California, Santa
Cruz
Kelson, D.D., Illingworth, G.D., Franx, M. & van Dokkum, P.G.
2001, ApJ, 552, 17
Kuntschner, H. 2000, MNRAS, 315, 184
Le Borgne, D. et al. 2006, ApJ, in press, astro-ph/0503401
Leonardi, A.J. & Rose, J.A. 1996, AJ, 111, 182
Longhetti, M. et al. 2005, MNRAS, 361, 897
Phillips, M.M., Jenkins, C.R., Dopita, M.A., Sadler, A.M. &
Binette, L. 1986, AJ, 91, 1062
Rampazzo, R., Annibali, F., Bressan, A., Longhetti, M., Padoan,
F. & Rodighiero, G. 2005, A&A, 433, 497
Schiavon, R.P. 2006, ApJ, submitted
Schweizer, F. & Seitzer, P. 1992, AJ, 104, 1039

Page 5

Red Galaxies at z ∼ 0.95
Spergel, D.N. et al. 2003, ApJS, 148, 175
Thomas, D., Maraston, C., Bender, R. & Mendes de Oliveira, C.
2005, ApJ, 621, 673
Trager, S.C., Faber, S.M., Worthey, G. & Gonz´ alez, J.J. 2000, AJ,
120, 188
Treu, T. et al. 2005, ApJ, 633, 174
Willmer, C.N.A. et al. 2006, ApJ, in press, astro-ph/0506041
Worthey, G., Faber, S.M., Gonz´ alez, J.J. & Burstein, D. 1994,
ApJS, 94, 687
Worthey, G. & Ottaviani, D.L. 1997, ApJS, 111, 377
Yan, R., Newman, J.A., Faber, S.M., Konidaris, N., Koo, D. &
Davis, M. 2006, ApJ, submitted, astro-ph/0512446