Transcript Evolutionary properties of galaxies and mass assembly up

PI
Co-PI
O. Le Fèvre
G. Vettolani
PI : Carole Lonsdale
Evolutionary properties of galaxies and mass
assembly up to z ~ 2 from VVDS+SWIRE data
G. Zamorani
On behalf of the VVDS-SWIRE collaboration
Outline of the talk
The VVDS 02hr field (The “deep” part of the VVDS) :
A) Environmental effects up to z ~ 1.2
luminosity functions (Ilbert et al. 2006) and
colors
(Cucciati et al. 2006)
as a function of the local environment
B) K band luminosity function and stellar mass density
up to z ~ 2 from VVDS + SWIRE data (Arnouts et al.
and Pozzetti et al.)
The 02hr VVDS field
The 02hr VVDS field is an extragalactic field with extensive
multi-wavelength observations (radio (VLA), UV (GALEX),
far infrared (SPITZER), X-ray (XMM)) and deep optical
photometry (VVDS + CFHTLS) and spectroscopy (more
than 6000 galaxy redshifts from the VVDS over about 0.5
sq.deg. (IAB < 24.0 purely magnitude selected)
Very well adapted for investigations on galaxy
evolution and LSS (see Le Fevre talk on Monday)
A) Luminosity functions, colors and environment
At z ~ 0.1 and over a wide range of local galaxy densities, a
strong dependency of galaxy properties with environment is
observed.
In addition to the well defined spectral type-density relation,
Croton et al.
2005; 2dF
there are also significant differences in the LFs in overand under-dense regions
2dF Luminosity functions (Croton et al. 2005)
N ~ 50,000
Clusters
Voids
-17
-19
-21
MbJ
A larger fraction of bright (red) galaxies
in over-dense regions (see also Balogh et
al. 2004)
Similar slopes in different
environments, but brighter
M* in over-dense regions
Naïve question:
Why is the slope of the global LF essentially independent from
the environment, even if the fraction of different spectral types
is a strong function of the environment and the slopes of the
LFs of different types are so different from each other?
Possible explanation :
If the conditional luminosity function (CLF) of galaxies is
only a function of the halo masses ( (L,Mh); halo-occupation
models (Mo et al. 2004; Cooray 2005)) the -dependence of
the galaxy LF enters only through the conditional halo mass
function n(M | ) and therefore the slope of the LF is
approximately constant in different environments, while M* is
brighter in over-dense environments.
If this is the correct explanation, it has to be checked if these
environmental effects remain the same at high redshift
Estimate of local density in the VVDS 02hr field
02hr field: More than 6,500 highly reliable galaxies’ redshifts, in
about 0.5 sq.deg, with average spectroscopic sampling rate of ~
20% (and as high as ~ 35% in the central region), allow good 3D
density field reconstruction over scales of the order of 5 h-1 Mpc
Sampling rate :
For each ith galaxy the local density
contrast i with respect to the average
density <(zi)> is measured:
 i (RS) = (  i(RS) - < (zi)> ) / < (zi)> ,
where  i(RS) is computed with a gaussian
filter with smoothing length RS
Simulations show a good correspondence
between input and reconstructed density
fields for scales RS > 5 h-1 Mpc
3D galaxy density contrast:
(r,R)=
(r, R, <MC)=
F(R)
S(r,MC)
(m)
(z,m)
(,)
(r, R) - (r)
(r)
 S(r,M ) (m) (z,m) (,)
D(r-ri) F(|r-ri|/R)
C
 gaussian filter
 radial selection function
 target sampling rate
 spectroscopic success rate
 angular sampling rate
A critical factor for the reliability of the density contrast estimator is a
high spatial sampling rate. At z ~ 0.75 the VVDS mean inter-particle
separation is ~ 4.4h-1 Mpc, similar to that of the 2dFGRS at its median
depth.
Luminosity functions in under- and over-dense regions
N = 924
N = 623
N = 1440
N = 641
N = 820
N = 471
N = 194
N = 193
Red : galaxies in over-dense
regions
Blue: galaxies in under-dense
regions
Up to z ~ 1.2, we find:
Brigther galaxies preferentially seen
in denser regions (same as locally)
Significant differences in the LFs :
flatter slopes in over-dense
regions ( ~ 0.2 – 0.3), while the
M* values are consistent with
each other in the two
environments (differently from local
samples)
Is the difference in slope, mainly due to different fractions of
early and late type galaxies in different environments?
This can be tested in the redshift range
0.6 – 0.9, where we have the highest
statistics, deriving separately the LFs of
red (left; MU – MV > 1.5) and blue
(right; MU – MV < 1.5) galaxies
Blue curve : LF in under-dense regions
Red curve : LF in over-dense regions
Not only the global LFs, but
also the type-specific LFs
significantly depend on
environment
The LFs are steeper in underdense regions both for red and
blue galaxies
Comparison with local results
While the higher fraction of bright galaxies in over-dense regions seen
locally is already present at z ~ 1.2, the environmental dependency of
the LF shape at high redshift appears to be different from what is seen
at low redshift.
Is this difference due to an increase with cosmic time of the number of
faint red galaxies, developing a steep slope of LF of early type
galaxies in over-dense regions as in 2dF ?
But :
a) We do not see any change on the environmental effect from z ~ 0.3
up to z ~ 1.2…
b) is it reasonable that the environmental effect on the LF changes so
quickly from our first redshift bin (z ~ 0.3) to z ~ 0?
To be better understood which physical mechanism can produce
this change with redshift of environmental effects on the LF
Evolution with z and luminosity of the color-density
Select Blue (U-V < 0.65) and red (U-V
> 1.40) galaxies and compute their
fraction as a function of  for different
absolute magnitudes and redshifts
A well defined color–density relation is
seen up to different z for different MB :
for MB <-19.0 up to z~0.6
for MB <-20.0 up to z~0.9
for MB <-21.0 up to z~1.2
The color - density relation is building
up, differentially with luminosity, as
cosmic time goes by, earlier for brighter
galaxies and later for fainter galaxies …
What does it mean?
Assuming that the adopted colors correspond to different
star formation histories, these data suggest that:
a) Star formation is differentially suppressed in high and
low density regions : the drop in star formation occurred
earlier in higher density environment
b) The drop in star formation is also a function of
luminosity (and mass?)
Both these findings are consistent with a downsizing
scenario in which star-formation activity is progressively
shifted with cosmic time toward lower luminosity
systems and out of high density peaks
B) The K band LF’s and the mass density up to z ~2
SWIRE+VVDS in the 02hr field
We use the photometric
data collected by SWIRE
(red), VVDS in BVRI
and JK (magenta) and by
CFHTLS survey in ugriz
(white). The VVDSspectroscopic survey has
collected >6500 spectra
up to I=24 (blue). In the
following, we restrict the
analysis to the area in
common between SWIRE
and CFHTLS ( ~ 0.85 sq.
deg.; ~ half of which with
spectroscopic redshifts)
IRAC Number counts
4.5m counts are shifted
by 2.5 mag for clarity
The IRAC 3.6 (red) and 4.5m
(green) (shifted by 2.5
magnitudes) differential galaxy
(circles) and star (triangles)
number counts. Adopted flux
limits are 9Jy and 15Jy or
21.5 and 21.0 (in AB mag),
respectively (70% completeness
at these limits). A small
completeness correction at the
faint end is applied for the
luminosity function analysis to
match the Fazio et al. (2004)
number counts (red and green
lines).
The combined dataset
Matching :
The SWIRE sources (~ 25,500 objects) have been
matched with the optical catalogs. Less than 1% of
the sources are not detected in the optical. The
matched SWIRE catalog in the CFHTLS area
consists of 22,300 galaxies with 3.6<21.5 and
19000 with 4.5<21.0 (~2300 are galaxies with a
reliable spectroscopic redshift from VVDS)
Color-magnitude diagram :
Spectroscopic sources are shown with red
(galaxies), yellow (stars) and green (QSOs)
symbols. The stars classified using profile
information and SED fitting from photo-z are
shown in magenta.
The lines show the bright (I<17.5) and faint (I<24)
limits of the spectroscopic survey.
Note the substantial number of red galaxies
(I – 3.6 > 3.5) fainter than the VVDS
spectroscopic limit
Photometric redshifts and their accuracy
Photo-z are derived by using the code «Le Phare» (Arnouts et al.) on the optical bands
BVRI,ugriz,(+JK) and the IRAC bands 3.6 and 4.5 m and exploiting the large
spectroscopic redshift sample to train the photo-z estimates (see Ilbert et al. 2006).
The main steps are: Iterative zero-points adjustments using spectroscopic sources +
optimization of the original templates by reconstructing the SEDs from the observed flux
of the spectroscopic sample
Comparison between spectroscopic and
photometric redshifts for the 3.6 m
sample with secure redshifts (1431
galaxies). ( a correction of 5% to the IRAC
fluxes improves the comparison)
The filled/open circles show photo-z with
estimated errors smaller/larger than 0.3.
Results:
No systematic shift is observed :
< z > = 0.00 for 0 < z < 1.5
Very small statistical error
 (z/(1+z)) = 0.03
Very small number of catastrophic errors
(~ 1.7% of the galaxies have z>0.15(1+z))
Redshift distributions :
The red curve shows the N(zspec) distribution,
while the dashed black line shows the N(zphot)
of the same objects. The solid black line shows
the normalized N(z) for the total 3.6 m zphot
sample.
Excellent agreement between the zphot
and zspec distributions (clear peaks are
detected in both samples at z ~ 0.3 and z ~
0.9, while the peak at z=0.6, although
visible, is somewhat diluted in the zphot
sample)
Prominent high redshift tail in the total
zphot 3.6 m sample (objects with red
colors seen as black dots in the green ellipse
in the previous slide)
Redshift segregation in a color – color plot
This color-color plot (r-i vs. i-3.6 m),
with different color contours (red for
z <1 and blue for z >1) for different
zphot intervals, shows how galaxies
move in this plane as a function of
redshift.
stars (magenta) are easily separated
from the galaxy population.
The distribution of galaxies at z<1 is
well distinguished from the highest z
population (see solid lines
Many of the galaxies with zphot > 1.5
have such a red i-3.6 color to be
excluded in an I-selected sample, thus
depleting the high z tail in the
spectroscopic sample.
K rest-frame Luminosity Functions from 3.6 and 4.5 m samples
The IRAC 3.6 and 4.5 m data probe the
2.2m stellar light at z ≈ 0.6 and z ≈1
respectively and are therefore the best data to
estimate, with minimal K-corrections, the K
rest-frame LFs at these redshifts.
Our LFs are very well determined up to
z=2.5 and are in excellent agreement
with each other at all z-bins (also with
the LF derived for our spectroscopic
sample)
With respect to local LF (Kochanek et
al.) we detect a clear evolution with z.
Samples:
3.6 with zphot (filled)
or zspec (open)
4.5 with zphot
----- Kochanek et al. (2001)
----- Drory et al. (2003)
----- Caputi et al. 2005)
Good agreement also with Pozzetti et
al. (2003; K20), Drory et al. (2003;
MUNICS) for z < 1.2, and Caputi et al.
(2005,2006; GOODS/ CDFS) up to z =
2.5 from an area 20 times smaller.
No evidence of the bright end to decrease
up to z = 2.5
Color bimodality : separation between early and late type populations
See also poster by D.Vergani et al.
A clear bimodality is seen in the
NUV-r intrinsic color, with a red
peak at NUV-r ~ 4.8.
This color is a good proxy of the b
parameter (SFR/<SFR>) as found by
GALEX-SDSS (Salim et al. 2005).
Excellent correspondence
between NUV-r color and our
SED 2 fitting, which we
therefore used to select a
population of early type galaxies
Red contours : elliptical templates in SED fitting
Blue contours : all other templates
With such a separation we
obtain a sample of ~ 4400
Early-type and ~16800 Latetype galaxies.
LF per types up to z = 2.5
Very well defined LFs for both
types of galaxies (early and late)
up to z ~ 2.5
Significant difference in the
slopes at all redshifts:
-0.5 <  < 0 for early
-1.4 <  < -1.1 for late
The brightest objects (KAB < -24)
are about 50% early and 50% late
up to z ~ 1.2
At z > 1.5 there is a significant
decline of the early type
population
Evolution of the NIR Luminosity versus z
The total luminosity density (black
circles) is dominated by emission
from the late type galaxies (blue),
which stays approximately constant
up to z~2, while the NIR luminosity
density of early type galaxies (red)
shows a quick drop at z>1.2.
Good agreement, but over a much
larger area, with the results from the
K20 results (Pozzetti et al. 2003) and
GOODS/CDFS results for K-selected
galaxies (~130 sq.arcmin) (blue
dashed line; Caputi et al. 2005)
Anything about downsizing?
Differential evolution of number density as a function of mass
(see poster by
Pozzetti et al.)
Number density of most
massive galaxies consistent
with local density up to z = 2

Decrease with redshift of
the density of less massive
galaxies

Evolution of the stellar mass density versus z
Slow decline of the global mass density up
to z ~1.2 and faster decline for 1.2<z<2.0 :
~60% (~25%) of the stellar mass is already
formed at z~1.2 (z~2.0)
Mass density ~ constant up to z ~ 1.4 for
the late type galaxies
Slow decline up to z ~ 1.2 also for the
early type population (~50% of the mass
of evolved galaxies already assembled at
z ~1.2), with a significant drop for 1.2<z<2.
The red curve is the mass assembly in ellipticals
from the Protracted Assembly model by F06 :
stars in today ellipticals born at high redshift,
most of which (~80%) dynamically assembled in
spheroids at 0.9 < z < 1.6.
The mass assembly in our “color defined”
early-type galaxies is consistent with what
found by Franceschini et al. (2006) based
on a morphological definition in an area ~
20 times smaller (GOODS) (red curve)
The growth of the stellar mass is in
excellent agreement with the integrated
dust-corrected SFR from GALEX data in the
same field (Schiminovich et al. 2005)
Conclusions: Environmental effects up to z ~ 1.2
Luminosity function:
While the higher fraction of bright galaxies in over-dense regions seen locally
is already present at z ~ 1.2, the environmental dependency of the LF shape at
high redshift appears to be different from what is seen at low redshift 
implications on models?
Color – density relation:
The color – density relation changes with redshift and luminosity:
Star formation is differentially suppressed in high and low density regions
and for galaxies with different luminosities (masses?): the drop in star
formation occurred earlier in higher density environment and in
higher luminosity galaxies  consistent with a downsizing scenario in
which star-formation activity is progressively shifted with cosmic time in
smaller systems and out of high density peaks
Conclusions : K-band LF and mass density up to z~2
A significant evolution of the bright end of the LF is detected, with no indication
of a decrease up to z ~ 2.5
At all redshifts there is a significant difference in the slopes of the early
(-0.5 <  < 0) and late types (-1.4 <  < -1.1)
The brightest objects (KAB < -24) are about 50% early and 50% late up to z ~ 1.2
Slow decline of the global mass density up to z ~1.2 and faster decline for
1.2<z<2.0 : ~60% (~25%) of the stellar mass is already formed at z~1.2 (z~2.0)
(but ~ 100% for galaxies more massive than 1011 solar masses)
Slow decline up to z ~ 1.2 also for the late type population (~50% of the mass of
evolved galaxies already assembled at z ~1.2), with a significant drop for
1.2<z<2.
Given the same assumption on the IMF, there is an excellent agreement between the
stellar mass density from LF +SED fitting ( M/LK) and that obtained with the
integrated dust-corrected SFR from GALEX data in the same field (Schiminovich
et al. 2005)