Observational Properties of z~6 Galaxies

Observational Properties of z~6 Galaxies

Rychard J. BouwensUCSC

Special thanks to Roderik Overzier, Mauro Giavalisco, Haojing Yan for helping me prepare this talk

The End of the Dark Ages / STScI / March 14, 2005

Collaborators:

Garth Illingworth, Ivo Labbe, Marijn Franx, Roderik Overzier, John Blakeslee, Dan Magee

z~6 -- An Exciting Epoch!

Mass of ~L* galaxiesSpringel et al. (2005)

Rapid Buildup of L* galaxies

z~6 represents a key transition

point of change between z~10

and z~3

High Redshift Frontier

1990: z = 3.80 Radio Galaxy (Chambers et al.)1997: z = 4.92 Lensed Dropout around CL1358+62 (Franx et al.)1998: z = 5.34 Lyman-alpha emitting Object (Dey et al.) 1998: z = 5.60 LBG in HDF North (Weymann et al.)1999: z = 5.74 Lyman-alpha emitter (Hu et al.)2001: z = 6.28 SDSS quasar (Fan et al.)2002: z = 6.56 Lyman-alpha emitter (Hu et al.)2003: z = 6.58 Lyman-alpha emitter (Kodaira et al.)2004: z ~ 6.6 Lensed Dropout (Kneib et al.) 2005: z = 6.7 Malhotra et al.

2

Highest Redshift Spectroscopically Confirmed Object / Some Mileposts

Wasn’t until the 2000s that we crossed the z~6 barrier…

Interesting how so many different techniques have been useful in finding the highest redshift objects: * Lyman-alpha emitters, QSOs, Lyman Break Galaxies * gravitational lensing, wide-area surveys, deep HST surveys

Finding Sources at z~6

z~6 Sloan QSOs

z = 6.56 Ly emitter (Hu et al. 2002)

(leverages i+z band imaging over very large area)

Ly

(leverages narrowband preselection + gravitational lensing)

9120 N R

HST WFPC2

Space has big advantages in searching for high-z objects due to much lower background.

However, until 2002, WFPC2 was the only camera in space to use for exploring the z>5 universe.

U B V I

HST Advanced Camera for Surveys

Redder, more efficient filters for exploring z > 5.5 universe

U B V i z

i

HST ACS

U B V I

HST WFPC2

Can select dropouts in much redder filter

with ACS!

Redder, more efficient filters for exploring z > 5.5 universe

U B V i z

i

HST ACS

U B V I

HST WFPC2

Can select dropouts in much redder filter

with ACS!

From Stanway et al. (2003) z~6 galaxy cuts off at the boundary

beween the i and z filters

Ideally we would do the z~6 i-dropout selection using the familiar two color diagram, i.e.,

Lyman Break Color

Strong Break

No Break

Continuum ColorBlue Red

z~6 objects

U B V i z

Ideally we would do the z~6 i-dropout selection using the familiar two color diagram, i.e.,

Lyman Break Color

Strong Break

No Break

Continuum ColorBlue Red

Unfortunately, you get the continuum

color you need deep infrared

imaging which is very expensive

z~6 objects

U B V i z IR

Single color i-dropout selection

Lyman Break Color

Strong Break i - z > 1.3

selection

No Break

RedshiftBunker et al. (2004)

Initial Round of Papers on i-dropouts

Stanway et al. (2003)

Yan et al. (2003) Bouwens et al. (2003)

Dickinson et al. (2004)~ 6 candidates

~ 30 candidates ~ 23 candidates

~ 251 candidates

Finding Real z~6 Galaxies Amongst Possible Contaminants

Evolved z~2-3 Sources did not appear to be an important concern

Bouwens et al. (2003)

Lyman Break Color

Strong Break

No Break

Continuum Color

Blue Red

All resolved sources here Stellar Locus

Size

z850 bandmag

Stanway et al. (2003)

i-dropouts are sufficiently resolved to exclude stellar contaminants

Stars

Galaxies

Surface Brightness Selection Biases(Incompleteness)

U-dropout from HDF-N artificially redshifted to z~6.0

Cosmic Surface Brightness Dimming Substantial

Factor of 10 from z~3

to z~6

Did surface brightness selection effects represent an important bias for the SFH?

SFRdensity

Redshift

Lanzetta et al. (2002)

After correction for SB selection effects?

Or selection effects not so significant?

High Redshift Size Evolution

Sizes

Redshift

Ferguson et al. (2004) did not plot a point at z~6 since surface brightness selection biases were still very important in the data used to construct this plot.

Ferguson et al. (2004)

Standard ruler

H(z)-1 ~ (1+z)-1.5

H(z)-2/3 ~ (1+z)-1

Data appear to be in good agreement with the scalings expected from this simple theory

High redshift galaxies are expected to be smaller because their halos collapse earlier and therefore more concentrated

Extending Size Measurements to z~6

Size vs. redshift

The sizes of i-dropouts are in good agreement with size-redshift trends found in Ferguson et al. (2004)

Sizes


This suggests z>7 galaxies are likely to have half-light radii of ~0.1”

Size vs. magnitude

Sizes

~0.14”-0.15”

UDF

i-dropouts are small (~0.15”)

Bouwens et al. (2006); see also Bunker et al. (2004)

z~6 observations versus z~3

Volume Density

Rest-frame UV Continuum Luminosity Function

z~3 (Steidel et al. 1999)

Rest frame UV 1350 Å


Volume Density



Yan (2003)



Volume Density



Yan (2003)

Stanway (2003)

6x



Volume Density



Yan (2003)

Stanway (2003)

Bouwens (2003)

6x



Volume Density



Yan (2003)

Stanway (2003)

Bouwens (2003)

6x

Dickinson (2004)



Volume Density



Yan (2003)

Stanway (2003)

Bouwens (2003)

6x

Dickinson (2004)



Volume Density



Stanway (2003)

Bouwens (2003)

6x

Dickinson (2004)



Volume Density



Stanway (2003)

Bouwens (2003)

6x

Dickinson (2004)

Disagree?


The two GOODS fields (~150 arcmin2 each) were key search areas in the earlier work.

Overall approach:

UDF UDF-IR

Hubble Ultra-Deep Field

UDF UDF-IR

ACS and NICMOS

5 limit: in UDF is ~30-31AB mag in BViz; in UDF-IR ~27.5AB mag in JH

Images of z~6 Galaxies

>100 i-dropouts in the UDF

Credit: Image by Zolt Levay

Yan & Windhorst (2005); Bouwens et al. (2006) see also Bunker et al. (2004)

(vs. much smaller numbers in the other fields)

Message from HUDF was that there are many faint galaxies

Surface Density of i-dropouts

z-band magnitude

No-evolution (NE) predictions from z~3

At bright mags: z~6 observations are much

lower than NE z~3 predictions

At faint mags:z~6 observations nearly equal to

NE z~3 predictions


Bright Faint

Observedsurface density of z~6 galaxies

(uses UDF + shallower datasets)

Early work by Dickinson et al. (2004) before HUDF suggested there were more

faint galaxies than bright ones.


No-evolution (NE) Predictions From z~3

Corrected i-dropoutcounts

Surface Density

z-band magnitude

BrightFaint

Many fewer bright z~6 objects found predicted from

z~3 assuming NE


Yan & Windhorst (2004):Used UDF to argue faint-end slope of z~6 LF was very steep, = 1.8

z~6 LF

Faint-end slope at z~6

Bright Faint

Malhotra et al. (2005):

Best fit to z~6 galaxies (HUDF) had a fainter characteristic luminosity than at z~3

(compare to = -1.6 at z=3)

Bright Faint

Galaxies at z~6 (i-dropouts):

Bouwens et al 2006

Wide Deep

z850,AB~ 27.1 (10) (vers: 1.0)

UDF-Parallels UDF

z850,AB~ 28.4 (10)z850,AB~ 29.2 (10)

506 z~6 i-dropouts!

17 arcmin2

11 arcmin2

316 arcmin2

GOODS

CDF-SHDF-N

1.927.5 Since original GOODS program, a significant amount of SNe search data has been taken

over the GOODS fields.

z~6 UV Luminosity Function

Applied a well-tested i-z > 1.3 criterion to select i-dropouts in all fields.

Used detailed degradation experiments on our deeper fields to perform completeness and flux corrections.

Carefully matched up surface densities of all fields to remove field-to-field variations (35% effect)

Accounted for blending with foreground objects (5-10% effect)

Determined contamination level (5-10% effect): Intrinsically-red objects Photometric scatter Stars Spurious sources

Selection function determined by using best estimates of UV colors and sizes of z~6 objects.

Rigorous i-dropout luminosity function determination


Bouwens et al 2006


Log # mag-1 Mpc-3

z~6

Bright Faint


Bouwens et al 2006


Log # mag-1 Mpc-3

z~6

z~3LF at z~6: goes ~3 mag below L*

Bright Faint


Bouwens et al 2006


Luminosity evolutionprovides the best fit - not density evolution

Log # mag-1 Mpc-3

z~6

z~3

Luminosity Evolution Provides a good fit

Bright Faint


Bouwens et al 2006 Rest frame UV 1350 Å

z~6Faint-endSlope

The characteristic luminosity at z~6

(L*UV,z~6) is ~50% of (L*UV,z~3) at z~3.Faint Bright


Bouwens et al 2006 Rest frame UV 1350 Å

z~6Faint-endSlope

The characteristic luminosity at z~6

(L*UV,z~6) is ~50% of (L*UV,z~3) at z~3.Faint Bright

Weak constraints on faint-end slope

Star Formation History

Luminosity Density (Star Formation Rate

Density - no extinction)

Log10

M yr-1Mpc-3

Star Formation History -- z ~ 0 - 6

z~6 result

z~6 result

Brighter FluxLimit

Fainter FluxLimit

Evolution in SFR density is much more

dramatic to brighter flux limits

Bouwens et al. 2006




Star Formation History -- z ~ 0 - 6

Shimasaku et al. 2005

Bright (zR<25.4) wide-area i-dropout search with Subaru

Fainter i-dropout search (Bouwens et al. 2004)

SFR density to bright limit

SFR density to fainter limit

Evolution of the UV LF

Hierarchical Buildup

AGN Feedback?Gas Exhaustion?

Transition between Hot/Cold Cooling

Flows?

Bright

Faint


Volume Density



Stanway (2003)

Bouwens (2003)

6x

Dickinson (2004)

Disagree?



Volume Density



z~3 (Steidel et al. 1999)Dickinson (2004)

Bouwens (2003)

Stanway (2003)

6x

?


Volume Density



Stanway (2003)

Bouwens (2003)

6x

Dickinson (2004)

Evolution Factor is Luminosity Dependent

Don’t Disagree!


Implications for Reionization

Using the standard Madau description, we find that the number of I-dropouts at z~6 appears to be approximately consistent with the numbers necessary to reionize the universe,assuming an escape fraction of 0.5 and clumping factor of 30.

6x

Dickinson (2004)


Field-to-field variations can be significant

# of i-dropouts / field at same depth

UDF

First ACS parallel to UDF NICMOS field

Second ACS parallel to UDF NICMOS field

18

50

44

~35% RMS variations for single ACS fields (Bouwens et al. 2006; see also

Bunker et al. 2004)

Large Scale Structure significantly limits our

ability to determine M*,

Surface Density of i-dropouts from GOODS + UDF-Ps + UDF

Significant Poisson Noise

UDF +UDF-Ps

GOODS

Relative normalization of bright + faint probes uncertain due to large-

scale structure

~ L*z=6

Unfortunately, L* is just at the edge of what can be probed with

the wide-area GOODS fields

Including LSS uncertainties

Ignoring LSS uncertainties

Large Scale Structure significantly limits our

ability to determine M*,

Surface Density of i-dropouts from GOODS + UDF-Ps + UDF

Significant Poisson Noise

UDF +UDF-Ps

GOODS

Relative normalization of bright + faint probes uncertain due to large-

scale structure

~ L*z=6

Unfortunately, L* is just at the edge of what can be probed with

the wide-area GOODS fields

Including LSS uncertainties

Ignoring LSS uncertainties

==> Need more deep fields

Deep i-dropout Search Fields

ACS Parallels to the UDF NICMOS data

UDF

CDF South

Deep i-dropout Search Fields

ACS Parallels to the UDF NICMOS data

UDF

UDF05 (PI: Stiavelli)

CDF South

Key New Data

Ground Based Spectroscopy

Keck

GeminiVLT Subaru

z~6 spectroscopic samples

Malhotra et al. 2005

GRAPES

23 objects

Dow-Hygelund et al. 2006

8 objects

UCSC/Keck

GOODS Team

GLARE/Exeter~20 objects

+ 80(?) more from the PEARS program

=> ~100 z~6 objects spectroscopically confirmed!

47 objects

Vanzella et al. 2005, 2006; Stern et al., in prep; Dawson et al. 2002

(also includes a few redshifts from GRAPES here)

Spectroscopy on z~6 galaxies

Bunker et al. 2003

z=5.78

Vanzella et al. 2006, in prep

z=5.52

Dow-Hygelund et al. 2005

Some noteworthy examples of z~6 spectra

Stack of 25 emission line galaxies

Composition of z~6 spectroscopic samples

~30% of i-dropouts show Ly emission (EW: >20 A)vs. 25% of U-dropouts at z~3 (Dow-Hygelund et al. 2006)

Contamination rates for current I-dropout selections appears to very low.

Red

UV continuum slope

Blue

Dust Properties(important for calculating unobscured star formation rate)

()

UV

UV

Low Dust

extinction High Dust

extinction

Red

UV continuum slope

Blue

Most LightAbsorbed By

Dust First

Infrared Light

UV Light

Most LightEscapesWithout

Absorption

Dust Properties(important for calculating unobscured star formation rate)

Correction Factor (Meurer et al. 1999)

()

UV

UV

Low Dust

extinction High Dust

extinction

Evolution in UV Continuum Slope

UV continuum slope vs. z

Bouwens et al. 2004, 2006b,c; See also Stanway et al. 2005; Lehnert et al. 2003; Yan et al. 2005

Red

UV continuum slope

Blue

Dusty

Dust Free

Galaxies appear to become less dusty at high redshift

Significant Dust at z~6.5?

Chary et al. (2005)

HCM6AAbell 370Hu et al. (2002)

z = 6.56

Anomalous jump in the flux density at ~6000 A rest-

frame

Is this due to Hemission?

If so, suggestssignificant dust

extinction ?

The Evolution of the SFR density

Bouwens et al. 2006

SFRdensity

TrueSFR density

SFR density (not counting

for dust)

Correcting for dust extinction accentuates the size of SFR density

peak at z~1-3

X-ray Properties

Chandra

Independent Method:SFR density from X-ray emission

Lehnert et al 2005A Recent stack of a larger ~400 object i-dropout sample from GOODS is undetected in x-ray

Lehnert et al 2006, private communication

- There is a known incidence of high mass x-ray binaries in SF regions

- X-ray light is much less affected by dust than UV light

Spitzer Space Telescope

(Measuring Stellar Masses)

UV Optical Rest-frame

Size of break tells us how

many old stars there

are

Age

Age

Age

Age

NIR Observed IRAC

Rest-optical & -IR at z~6

Kneib et al. (2004) lensed object at z~6.6

J

Hz~6.6 source

3.6 4.5

Stellar Mass = ~109 Msol

Best Fit (e-decay) = 100 Myr

z~6.6 source

IRAC Imaging

Egami et al. (2005)

Stellar Masses in select GOODS/UDF i-dropouts

ch2, 4.5mrest=6600A

Yan et al. (2005); see also Eyles et al. (2005)

Major results:• Very massive galaxies (M>1010 Msun)

existed at z ~ 6• A few hundred million years old (must

form well before z ~ 6)• Modest reddening (best-fits all have zero

reddening)

IRAC 3.6J110z850

z850

(zoomed)

SED fitting using Bruzual & Charlot SED fitting using Bruzual & Charlot (2003) models with exponentially decay (2003) models with exponentially decay star formation historiesstar formation histories

Significantly Larger z~6 Samples

Yan et al. 2006 (to be submitted)12’

12’

53 i-dropouts from GOODS with firm IRAC detections 79 i-dropouts are invisible

in their individual IRAC exposures

Move to complete samples of i-dropouts over the GOODS fields (~200 objects)

To make statistically significant statements:

z850 - IRAC 3.6 colors for ~170 i-dropouts

Yan et al. 2006 (to be submitted)

Balmer Break

z850-band magnitude

i-dropouts detected in 3.6 IRAC imaging

i-dropouts undetected in 3.6 IRAC imaging

Old

Young

Bright FaintControl

Stack of I-dropouts which are undetected individually

Implications


<z850 - 3.6>AB = 1.33

Detected with IRAC (~40% of the sample)

Individually undetected with IRAC (~60% of the sample)

<z850 - 3.6>AB = 0.4

Constant SFR

Simple Stellar Population

>100 Myr old=> 1 Gyr of constant SF is not enough

Stellar mass density lower limits based on full-epoch GOODS results


Stellar Mass Density

Integral ofSFR History

Diagram

1 + Redshift

New Point

None, or at least very few, of these objects appear to have solar masses as large as the

Mobasher et al. (2005) JD2 object or the Wiklind et al. (2006) objects.

Clustering of i-dropouts

181 i-dropouts

CDF South GOODS

Overzier et al. (2006)

HDF North GOODS

151 i-dropouts

Bouwens et al. (2006) sample of i-dropoutsBased upon original GOODS v1.0 data + SNe search data

(twice as deep)

Useful for learning about the halo masses

Clustering of i-dropouts

Clusteringw()

Angular Separation (“)

Clustering significant at 99.9% confidence

Overzier et al. (2006)It is true that better statistics would be ideal, but larger

samples are unlikely to be available soon

Invert using Limber’s Equation


Redshift Distribution

€

w(θ) = Awθ−β

Angular Correlation Function

Limber’s Equation

€

Aw =C0γ

F(z)Dθ1−γ (z)N(z)2g(z)dz

0

∞

∫N(z)2dz

0

∞

∫

€

γ= +1

Real Space Correlation Function

€

ξ(r) = (r /r0)−γ

Results in Real Space


Correlation LengthsMore luminous i-dropouts appear to be more clustered than the faint ones.

Bright

Correlation Length

Strongly clustered

Weakly clustered

FaintSimilar to findings at z~3-5 (Giavalisco & Dickinson 2001; Ouchi et al. 2004;

Lee et al. 2006) Suggests that the most luminous starbursts live in the most massive halos.

i-dropoutsz~6

z~4

z~5

Bias / Halo Mass

Lee et al. (2006); Overzier et al. (2006)

Galaxies at z~3-4 live

in ~1012 Msol halos

z~6

z~4

z~5

Bias

However, galaxies at z~5-6 appear to live in ~1011 Msol halos

This suggests star formation is much more efficient at z>5 than it is at z~3-4 in producing UV photons

May be partially due to an evolution in dust content:

i.e., more dust at z~3-4=> less UV photons escaping

M1700 < -20.0

Bias / Halo Mass

Lee et al. (2006); Overzier et al. (2006)

Galaxies at z~3-4 live

in ~1012 Msol halos

z~6

z~4

z~5

Bias

However, galaxies at z~5-6 appear to live in ~1011 Msol halos

This suggests star formation is much more efficient at z>5 than it is at z~3-4 in producing UV photons

May be partially due to an evolution in dust content:

i.e., more dust at z~3-4=> less UV photons escaping

M1700 < -20.0

Change in Efficiency can Explain Slow Evolution in LF from z~6 to z~3

Mass of ~L* galaxies

Springel et al. (2005)

Evolution of Mass Function

~10x increase in number

of 1012 Msol halos from z~6 to z~3

z=6

z=3


Mass of ~L* galaxies



But the efficiency of star formation

changes from z~6 to z~3

Compare 1011 Msol halos at

z~6

z=6

z=3 with1012 Msol halos at

z~3




Compare 1011 Msol halos at

z~6

z=6

z=3 with1012 Msol halos at

z~3

Now measure increase differently

Change between z~6 and z~3 much less




2006 End of Dark Ages 03/14/06 RJB

Log10

M yr-1Mpc-3

z~6 result

Star Formation History -- z > 6




Log10

M yr-1Mpc-3

UDF z~7-8 sample

z~6 result

Star Formation History -- Previous Results



Previous J-dropout search



Log10

M yr-1Mpc-3

UDF z~7-8 sample

z~6 result

Star Formation History -- Previous Results




Log10

M yr-1Mpc-3

z~6 result



Samples of >500 galaxies are now available at z~6 from HST data.

z~6 UV LF rigorously determined to ~3 magnitudes below L*.

Substantial evolution occurs at the bright end of the UV LF from z~6 to z~3. The characteristic luminosity at z~6 (L*UV) is 2 smaller than what it is at z~3.

z~6 galaxies appear to be less dusty on average than galaxies at lower redshift. This accentuates the rise in SFR density from z~6 to z~3.

~80-100 z~6 objects have now been spectroscopically confirmed.


Conclusions


Conclusions A small fraction z~6 objects has solar masses in excess of 1010 Msolar.

40% appear to be at least 100 Myr old.

L* objects at z~6 appear to predominantly live in 1011 solar mass halos. This is smaller than at z~3, suggesting there is an increase in the SF efficiency from z~6 to z~3.

Our improved knowledge of the z~6 universe puts us in an ideal

position to interpret the z>6 universe.

Determining the z~6 UV LF

But, to determine the LF, we need to divide the numbers

by the volume, i.e.,

Surface Density of i-dropouts with magnitude z850,AB

€

ϕ (m) =Number(m)

Volume (m)GOODS

UDF-Ps

UDF

After many corrections

Estimating the Selection Volume

Probability of Selecting a galaxy with magnitude z850

and redshift z as an i-dropout

Too faint to be

detected

How to calculate?

1. Create artificial galaxies

2. Add these galaxies to real images

3. Reapply Selection Procedure


Bouwens et al 2006


Log # mag-1 Mpc-3

z~6

Documents

Observational Properties of z~6 Galaxies