Download FIRST LIGHT IN THE UNIVERSE

FIRST LIGHT IN THE UNIVERSE
Saas-Fee, April 2006
Richard Ellis, Caltech
1. Role of Observations in Cosmology & Galaxy Formation
2. Galaxies & the Hubble Sequence
3. Cosmic Star Formation Histories
4. Stellar Mass Assembly
5. Witnessing the End of Cosmic Reionization
6. Into the Dark Ages: Lyman Drop Outs
7. Gravitational Lensing & Lyman Alpha Emitters
8. Cosmic Infrared Background
9. Future Observational Prospects
When Did Galaxies Form?
In the 1970-80’s, astronomers sought for a distinct era of galaxy formation
Synthesis models suggested E/S0s formed a high z via luminous `initial burst’
These searches for `primeval galaxies’ were unsuccessful
In 1990’s gradual assembly was deduced via studies of cosmic SF history
Shown to be consistent with CDM
Early Searches for Primeval Galaxies
What would they look like?
• Partridge & Peebles (1967, Ap J 147, 868): free-fall collapse - 700 L*
systems at z~10-30, large diffuse & red, possible L
• Meier (1976, Ap J 207, 343): compact stellar systems, possibly QSOs
• Tinsley (1980, Ap J 241, 41): MS brightening goes as dM/dt 1.3-0.3x
leading to observed excess in blue galaxy counts
• Bruzual (1983, Ap J 241, 41): updated synthesis models
Spectral evolution
log F
Hubble diagram
mB(z)
SF history
look-back time
log z
Pritchet 1994 PASP 106, 1052: Comprehensive review of a decade of searching!
Local Inventory of Stars
Relevant papers: Fukugita et al 1998 Ap J 503, 518
Fukugita & Peebles 2004 Ap J 616, 643
Survey data: Cole et al 2001 MNRAS 326, 255 (2dF+2MASS)
Kauffmann et al 2003 MNRAS 341, 33 (SDSS)
Stellar density: derives from local infrared LF,  (LK) scaled by a mean
mass/light ratio (M/LK) which depends on initial mass function
Useful stellar density is that corrected for fractional loss R of stellar
material due to winds & SNe: this should be integral of past SF
history (e.g. R~0.28 for Salpeter IMF)
Cole et al find stars h = 0.0014 ± 0.00013; M/LK = 0.73 (Miller-Scalo)
stars h = 0.0027 ± 0.00027; M/LK = 1.32 (Salpeter)
Fukugita & Peebles: stars h = 0.0027 ± 0.0005 (5% in brown dwarfs)
gas = 0.00078 ± 0.00016 (H I, He I, H2)
NB: only 6% of baryons are in stars!
Some Initial Mass Functions
Salpeter > 1 M
reproduces colors
and H properties
of spirals
Mass
fraction
per log
mass
bin
IMF < 1 M
makes minor
contribution
to light but is
very
important for
mass
inventory
Mass (M)
Stellar Mass Function at z=0
17,000 K-band selected galaxies with K<13.0
2dFGRS redshifts & 2MASS photometry
K
Stellar
mass
Cosmic Star Formation History
Various probes of the global SF rate:  (z) M yr-1 comoving Mpc-3
*
• UV continuum (GALEX, LBGs)
• H and [O II] emission in spectroscopic surveys
• mid-IR dust emission
• 1.4GHz radio emission
No simple `best method’: each has pros and cons (dust extinction, sample
depth, z range and physical calibration uncertainties)
Each has different time-sensitivity to main sequence activity so if SFR not
uniform do not expect same answers for the same sources
Would expect the integral of the past activity to agree with locallydetermined stellar density (Fukugita & Peebles 2004)
Can also determine the stellar growth rate for comparison with the stellar
mass assembly history (next lecture)
Recent review: Hopkins 2004 Ap 615, 209
Time-Dependence of Various SF Diagnostics
• Each SF diagnostic
arises from a
component of the
stellar population
whose lifetime is
different, e.g. in a
situation where the
SF is erratic. So there
is no single “best”
one
• Radio continuum is
thought to arise from
SN remnants and
offers the potential of
a dust-free diagnostic
Burst model
Comparison of UV and Hα for same local galaxies
• UV(2000Å) c.f. H
(corrected for extinction
via Balmer line ratio)
• Scatter cannot be
explained with a
dispersion in IMFs and
metallicities
• Suggests evidence for
non-uniform SF histories
&/or significant dust
complications
Sullivan et al (2000) MNRAS 312, 442
50-130 Myr;
10-30% mass
Cosmic SFH: Calibration
Kennicutt 1998 Ann Rev A&A 36, 189 (Salpeter IMF)
1.
UV continuum (1250-2500 Å) :
Pro: Extensive datasets over 0<z<6: easily calibrated via MS models
M>5M, timescales >108 yr, calibration largely independent of l
Con: dust! (A < 3 mag); IMF-dependent
2.
Line emission (H, [O II] :
Pro: Very sensitive probe, available to z~2: M>10Mtimescales <106 yr,
Con: uncertain fesc of ionizing photons; strong IMF-dependence (3),
excitation uncertainties [OII]
3.
Far IR emission (10-300 m) :
Pro: Independent method, available for obscured sources to high z:
Con: uncertain source of dust heating (AGN/SF?); age of stellar popn,
primarily applicable in starbursts, bolometric FIR flux required
Some Popular Dust Extinction Laws
How It Works: Early Estimate of Cosmic SFH
Ellis 1997 Ann Rev A&A 35, 389
Madau et al (astro-ph/9901237)
Luminosity density L(z)
SFR(z)
Rapid rise in blue light to z~1 has its origins
in galaxy count excesses back to 1980
Field redshift surveys to z~1 (Lilly et al 1996, Ellis et al 1996)
Counts of LBGs z>2 (Madau et al 1996)
Cosmic SFH: Recent Compilation
Hopkins 2004 Ap J 615, 209 (see also Hogg astro-ph/0105280):
- standardized all measures to same IMF, cosmology, extinction law
- integrated LF over standardized range for each diagnostic (except at v high z)
 (1+z)3.1 (z<1)
*
Star
formation
rate per
unit
comoving
volume
Fossil
record
decline?
Implications of Cosmic SFH
Hopkins & Beacom (astro-ph/0601463)
Fitting parametric SFH can predict  (z) in absolute units
*
Star formation history
GALEX,
SDSS UV
ACS
dropouts
Mass assembly history
Cole et al 2dF
Spitzer FIR
• Satisfactory agreement with local 2dF/2MASS mass density
• Data suggests half the local mass in stars is in place at z~2  0.2
• Major uncertainties are IMF and luminosity-dependent extinction
Theoretical Estimates of Star Formation History: - I
Baugh et al 1998, 2000
Baugh et al 2005
quiescent
burst
• Extended SF histories was an early prediction of CDM models
• But considerable flexibility in matching changing datasets!!
• Energetic sub-mm sources posed a major challenge
Invoke quiescent & burst modes of star formation
Can only fit optical & thermal IR LFs if bursts have top-heavy IMF
Theoretical Estimates of Star Formation History: - II
Nagamine et al 2004
In same cosmology, hydrodynamic simulations predict more SF at high z
- more than observed (even including the sub-mm sources!)
- a “missing high redshift galaxies” problem!
- zstar(50%)  2.0-2.5 c.f. zstar(50%)  1.3 for semi-analytic models
Towards a Unified View of the Various High z Populations
Integrating to produce a comoving cosmic SFH dodges the important question of
the physical relevance of the seemingly diverse categories of high z galaxies
(e.g. LBGs, sub-mm, DRGs).
Given they co-exist at 1<z<3 what is the relationship between these objects?
Key variables:
- basic physical properties (masses, SFRs, ages etc)
- relative contributions to SF rate at a given redshift
- degree of overlap (e.g. how many sub-mm sources are LBGs etc)
- spatial clustering (relevant to bias)
Some recent articles:
Papovich et al (astro-ph/0511289)
Reddy et al (astro-ph/0602596)

Lyman Break Galaxies - Clustering
 (r)  A(r r0 )
UV bright galaxies at z~3 are
clustered nearly as strongly as
bright galaxies in the present
Universe. Of what population are
they the progenitors?
What are the masses of these
galaxies (both dark and stellar)?
Adelberger et al (1998) demonstrated strong clustering of
LBGs consistent with their hosting massive DM halos perhaps
as progenitors of massive ellipticals (Baugh et al 1998)
V-band Luminosity Function at z~3
Local LF
Key to physical nature of LBGs is origin of intense SF. Is it:
- prolonged due to formation at z~3 (Baugh et al 1998)
- temporary due to merger-induced star burst (Somerville et al 2001)
Shapley et al 2001 Ap J 562, 95
LBG Properties (z~3)
<age> = 320 Myr @ z = 3
<E(B-V)> =0.15
AUV~1.7 ~5
<M*> = ~2 x 1010 M
<SFR> ~ 45 M yr-1
Extinction correlates with age– young galaxies are much dustier
SFR for youngest galaxies average 275 M yr-1 ; oldest average 30 M yr-1
Objects with the highest SFRs are the dustiest objects
Shapley et al 2001 Ap J 562, 95
Composite Spectra: Young vs. Old
• Young LBGs also
have much weaker Ly
emission, stronger
interstellar absorption
lines and redder
spectral continua
• Galaxy-scale outflows
(“superwinds”), with
velocities ~500 kms s-1,
are present in
essentially every case
examined in sufficient
detail
LBG Summary
• Period of elevated star formation (~100’s M yr-1) for ~50
Myr with large dust opacity (sub-mm galaxy overlap)
• Superwinds drive out both gas and dust, resulting in more
quiescent star formation (10’s M yr-1) and smaller UV
extinction
• Quiescent star formation phase lasts for at least a few
hundred Myr; by end at least a few 1010 M of stars have
formed
• All phases are observable because of near-constant far-UV
luminosity
So how is this LBG-submm connection viewed from
the sub-mm point of view?
Sub-mm Sources
Sub-mm source counts
Extragalactic background
COBE
Source counts already provide bulk of the measured FIR background, so
provided N(z) is unbiased, z~2-3 is where most of the sub-mm sources lie
because of the negative k-term
Blain et al 2002 Phys. Rep. 369, 111
What about clustering of sub-mm sources?
Clustering allows us to determine the typical halo mass
in which different galaxy types live
Correlation length r0
Angular correlation function
Tentative evidence for stronger
clustering than LBGs (but N=73 cf.
N>1000 LBGs!) suggesting more
massive subset in dense structures
Blain et al 2004 Ap J 611, 725
Passive Galaxies: The Classical Picture
Homogeneity of Cluster E/S0 U-V Colors
z  0.0
z  0.5 (HST)
Virgo & Coma: (U-V)o< 0.05 (Bower, Lucey & Ellis 1990, Bower et al 1998)
Morphs: <z> = 0.5 sample: (U-V)o< 0.07 (Ellis et al 1997)
Tight color-luminosity relations: stars are old
z  0.5
z0
U-V
(U-V)0
Universal relation for Es and S0s (Sandage & Visvanathan 1978)
Scatter dominated by observational errors (Bower et al 1990, Bower et al 1998)
(U-V) is sensitive probe of decline rate of MS component (Buzzoni 1989)
 uniform star formation history:
synchronisation of recent activity or old stellar population zF > 3
Stellar Mass Assembly History (CDM)
Evolution of stellar mass function
Merger trees
z=5,3,2,1,0.5,0
time
z=5,3,2,1,0.5,0
time
z=1
CDM predicts recent growth in assembly of spheroids, slower growth in disks
Declining Red Sequence to z=1: Agreement with CDM?
Color-photometric z’s in COMBO-17
Red luminosity density
COMBO-17 data suggests 3 decline in `red sequence’ luminosity density to
z=1: consistent with hierarchical predictions (Bell et al ApJ 608, 752 2004)
Gemini Deep Deep Survey
By contrast, the Gemini
DD Survey find an
abundance of high
mass old objects at
redshifts z>1 - in
seeming contradiction
with the COMBO-17
results to z~1?
R-K
(Can reconcile these
contradictory observations
if mass assembly is itself
mass-dependent)
Redshift
Glazebrook et al Nature 430, 181 (2004)
Gemini Deep Deep Survey: Spectroscopic Age-dating
• 20 red galaxies z~1.5, age 1.2 - 2.3 Gyr, zF=2.4 - 3.3
• Progenitors have SFRs ~ 300-500 M yr-1 (sub-mm gals?)
McCarthy et al Ap J 614, L9 (2004)
`BzK’ selection of passive and SF z>1.4 galaxies
New apparently
less-biased
technique for
finding all
galaxies
1.4<z<2.5
(z-K)
sBzK: star
forming
galaxies
pBzK:
quiescent
galaxies
WHERE DO THESE FIT IN?
(B-z)
Daddi et al 2004 Ap J 617, 746
LBG & `BzK/SF’ z~2 populations are the same
LBG
Fraction of BzK/SF galaxies
selected as LBGs and v.v.
Contribution to SF density
(including X-ray AGN)
(excluding X-ray AGN)
Reddy et al 2005 Ap J 633, 248
Subaru/VLT Survey of K<20 `BzK’ Galaxies
Source Counts
Clustering
Large panoramic survey - but no spectroscopy
Clustering of sBzK and pBzK galaxies is identical, suggesting
they are the same massive population and SF is simply transient
Space density of pBzKs with M>1011 M is ~ 20% local value
Kong et al 2006 Ap J 638, 72
Are SF and Passive z~2 populations distinct?
log
stellar
mass
Passive 1.6<z<2.9
LBG 1.5<z<2.9
K
Stellar mass distributions overlap indicating primary difference is current SF
Reddy et al 2005 Ap J 633, 248
Spitzer Studies of Massive Red Galaxies (J-K>2.3)
Stellar mass
Specific SFR (/mass)
K-selected sample of 153 DRGs z<3; many with M>1011 M ; 25% with AGN
Specific SFR (including IR dust emission) ~2.4 Gyr-1; >> than for z<1 galaxies
Witnessing bulk of SF in massive galaxies over 1.5<z<3
Papovich et al astro-ph/0511289
Summary of Lecture #3
• Multi-wavelength (optical/UV, near-IR, Spitzer and sub-mm)
observations have led to a revolution in tracking the history of star
formation in the Universe
• We have a good understanding of the evolution of the co-moving
density of SF since z~3 which accounts for the observed stellar mass
density at z=0. Half the stars we see today were formed by z~2.
• Galaxy populations identified by various means (sub-mm, LBGs,
BzK, DRG..) can be connected by their clustering, intermittent SF and
dust content.
• The emerging picture has the bulk of the star formation in massive
galaxies complete by z~1.5; subsequent evolution is largely occurring
in the demise of activity in lower mass systems
• This `downsizing’ phenomenon (SF completed in massive galaxies
earlier than in lower mass systems) is counter to simple hierarchical
assembly and arises from feedback (Lecture #4).