Emergence of cosmic structures around distant radio galaxies and quasars

quasars

Overzier, Roderik Adriaan

Citation

Overzier, R. A. (2006, May 30). Emergence of cosmic structures around distant radio

galaxies and quasars. Retrieved from https://hdl.handle.net/1887/4415

Version:

Corrected Publisher’s Version

License:

Licence agreement concerning inclusion of doctoral thesis in the

_{Institutional Repository of the University of Leiden}

Downloaded from:

https://hdl.handle.net/1887/4415

Chapter 5

Star f

ormati

on,morphol

ogi

esand

c

l

us

teri

ngofgal

axi

esi

na radi

ogal

axy

protoc

l

us

ter at z

=

4

.

1

Abstract._{We presentdeepg}475r625i775z850KSobservationstowardsthe radiogalaxy TNJ1338–1942at

z=4.1. The radiogalaxy isa∼6L∗_z=4galaxy. The data allow ustostudy indetail12spectroscopi -cally confirmedcompanionspreviously foundthroughtheir excessLyαemissionby Venemansetal. (2002). We conclude thatthe Lyαemitters(LAEs) are young(a few×107yr), dust-free galaxiesbased onsmallsizes, steepUV slopes(β ≈ −2)andblue UV-opticalcolorswithstar formationrates(SFRs) of<14 M yr−1. Whenstackingthe KS-bandfluxes, the LAEsseem tobe lessmassive (massesof

a few×108M) thanUV-selectedLymanbreakgalaxies(LBGs) while havingcomparable UV SFRs.

We estimate the LAEAGNfractiontobe minimal.

The fieldfurther contains66 g475-dropoutstoz850₌27(5σ), 6 ofwhichare inthe LAEsample. Their

SFRs, sizes, morphologicalparameters, UV slope-magnitude and(i775–KS) vs. KS color-magnitude

relationsare allsimilar tothose foundfor LBGsinthe ‘field’. We quantify the number density andcosmicvariance ofz ∼4 g475-dropoutsextractedfrom a pixel-by-pixeltransformationofthe

B435V606i775z850GOODSsurvey tog475r625i775z850, andshow thatthe fieldofTN J1338–1942isricher

thanthe average fieldat∼3−5σsignificance. The angular distributionishighly filamentary, with abouthalfofthe objectsclusteredina 4.4 arcmin2

regionthatincludesthe radiogalaxy andthe

brightestLBGs. A second, butmuchlesspronouncedconcentrationofobjectsiss

eenaroundan-other∼6L∗ LBGlocatedwithinthe same field, for whichwe obtaineda spectroscopicredshiftof z=3.8. The generally fainter LAEsappear tofavour regionsthatare devoidofLBGs, while LBGs detectedinthe rest-frame optical(KS) tendtolie inthe richestregion, suggestinga formingage-or

mass-density relation. We compare the angular two-pointcorrelationfunction, w(θ) , tothe signal measuredinsimilarly sizedmocksampleswitha built-intwo-pointclusteringasmeasuredfor field LBGsatz∼4. We findanexcesssignal(2σ) atseparationsofθ. 2000, correspondingtothe typical halosize ofdarkmatter haloshostingbrightLBGs. The large galaxy overdensity, itscorresponding massoverdensity andthe sub-clusteringatthe approximate redshiftofTN J1338–1942suggestthe assemblage ofa>1014

Mstructure, possibly a ‘protocluster’.

R. A. Overzier, R. J. Bouwens, N. J. G. Cross, B.Venemans, G. K. Miley, A. W. Zirm, N. Ben´ıtez,J. P. Blakeslee, D. Coe, R. Demarco,H. C. Ford, N. Homeier, G. D. Illingworth,J.D. Kurk,

A. R. Martel, S.Mei, H.J.A. R¨ottgering,Z. Tsvetanov& W. Zheng SubmittedtoThe AstrophysicalJournal

5.1

I

n

t

r

o

d

u

c

t

i

o

n

The complexity of the present-day large-scale structure originates from small seed density fluctuations, as evidenced by the discovery of the minute cosmic microwave background (CMB) anisotropies. The power spectra of the CMB and galaxy redshift surveys have given important clues to the cosmic matter budget, the

density field at recombination (z≈1000) and the

development of galaxies and galaxy bias. The emerging picture is that of bound objects, form-ing on increasform-ingly larger spatial scales due to gravitational instabilities of the cold dark

mat-ter (CDM) in a flat, Λ-dominated universe. It is

currently believed that the universe became (re-)ionized by the first light of stars and/or quasars at z∼7−25 (Kogut et al. 2003).

Large samples of UV-selected Lyman break

galaxies (LBGs;masses of a few×1010 M

(Pa-povich et al. 2001;Barmby et al. 2004)) at 3<z< 5 have been used to determine the cosmic star

formation rate density well beyond z∼2 (e.g.

Madau et al. 1996;Steidel et al. 1996, 1999;Ouchi et al. 2004). LBGs, as well as the partially

over-lapping population of Lyαemitters (LAEs), are

strongly clustered at z=_{3−5, and are highly} bi-ased relative to predictions for the dark matter distribution (Giavalisco et al. 1998;Adelberger et al. 1998;Ouchi et al. 2004;Lee et al. 2005; Kashikawa et al. 2005). The biasing becomes stronger for galaxies with higher rest-frame UV luminosity (Giavalisco & Dickinson 2001). In an all-encompassing census of the clustering prop-erties of LBGs, Ouchi et al. (2004) found that the bias may also increase with redshift and dust extinction. They suggest that the reddest LBGs could be connected with the similarly strong clustered sub-mm sources (but see Webb et al. 2003) or the extremely red objects (EROs, Elston et al. 1988;McCarthy et al. 2001;Daddi et al. 2002). By comparing the number densities of LBGs to that of dark halos predicted by Sheth &

Tormen (1999), they concluded that z=_{4 LBGs}

are hosted by halos of 1×1011−5×1012 M,

and that the descendants of those halos at z=₀

have masses that are comparable to the masses of groups and clusters. The derived halo

occu-pation numbers of LBGs increase with luminos-ity from a few tenths to roughly unluminos-ity, implying that there is only one-to-one correspondence be-tween halos and LBGs at the highest masses. On the other hand, the halo occupation numbers of SCUBA sources and distant red galaxies (Franx et al. 2003) are significantly above unity, imply-ing that for a given massive halo only 10%of the galaxies would be identified as an LBG.

The evolution of the earliest objects into present-day galaxies can currently be probed

to z∼7, predominantly using the superb

res-olution and sensitivity of the AdvancedCamera

for Surveyson the HubbleSpaceTelescope(HST ACS, Ford et al. 1998). The deepest samples of high redshift galaxies currently available come from the Hubble Deep Field North and South (HDFs), the Great Observatories Origins Deep Fields (GOODS) CDF-S and HDF-N, the Hub-ble Ultra Deep Field (UDF) and the two UDF-Parallel Fields (UDF-Ps). Highlights from these deep studies include the discovery of signifi-cant samples of UV-bright i775-dropouts –

galax-ies with red enough i775–z850 to lie at z ∼ 6

(Bouwens et al. 2003b, 2004b, 2006;Stanway et al. 2003;Yan & Windhorst 2004;Dickinson et al. 2004), allowing the determination of the

cosmic star formation rate (SFR) from z=_{6 to}

z=_{0. There is a significant but modest decrease}

in the SFR from z∼2 out to z∼6 (Giavalisco

et al. 2004a;Bouwens et al. 2005b). Studies of galaxy sizes indicate that high-redshift galax-ies are compact in size (∼0.001–0.003), while large (&0.00

4) low surface brightness galaxies are rare (Bouwens et al. 2004a). Furthermore, there is a clear decrease in size with redshift for objects of fixed luminosity, with a preferred redshift size scaling (1+_z)−1.05±0.21(Bouwens et al. 2004a,see

also Ferguson et al. (2004)). Morphological anal-ysis of LBGs indicates that they often possess brighter nuclei and more disturbed profiles than local Hubble types degraded to the same image

quality (e.g. Lotz et al. 2004). NICMOS imaging

of the UDF in J110 and H160 suggest that

lumi-nous galaxies at z∼7−8 exist with a total UV

luminosity density that is still significant

sug-gests that cosmic star formation is a continuous process with only gradual changes in the UV

lu-minosity density since z∼6−7.

Despite these advances in the study of the evolution of the highest redshift galaxies, galaxy

clusters have been studied out to only z=₁.4

(Mullis et al. 2005). High redshift clusters are X-ray luminous due to virialized gas and galax-ies moving within the cluster gravitational po-tential. These clusters contain populations of old and relatively massive galaxies, as well as younger star-forming galaxies (e.g. Dressler et al. 1999; van Dokkum et al. 2000; Goto et al. 2005). The scatter in the color-magnitude rela-tion for cluster ellipticals at z∼1 is virtually in-distinguishable from that at low redshift, sug-gesting that some of the galaxy populations in these clusters are already remarkably old (e.g. Stanford et al. 1998; Blakeslee et al. 2003a; Wuyts et al. 2004; Holden et al. 2005). Postman et al. (2005) measured the morphology-density

rela-tion (MDR) in seven z ∼1 clusters that have

been observed with the ACS. Evolution in the MDR appears to be primarily due to a deficit of S0 galaxies and an excess of Spiral/Irr galaxies relative to the local galaxy population, while the MDR for ellipticals exhibits no such significant

evolution between z∼1 and z=_0.

It has become clear that the rich clusters gan forming at earlier epochs than hitherto be-lieved, and their progenitors may be found at

much earlier epochs. Finding and studying

these progenitors may yield powerful tests for (semi-)analytical models and N-body simula-tions of structure formation. Although these models are relatively successful in reproduc-ing large-scale galaxy clusterreproduc-ing and galaxy lu-minosity functions, they still remain relatively untested on cluster-sized scales because of the absence of observed cluster progenitors beyond z∼1.

Several good candidates for galaxy overden-sities, possibly ‘protoclusters1

’, have been dis-1

The term protocluster iscommonly used todescribe galaxyoverdensitiesathigh redshift(z& 2)with masses -timatesthatare comparable tothose ofthe virializedgalaxy clusters,butwithoutanyevidence for a virializedint

ra-coveredatveryhighredshift(e.g. Pascarelle

etal.(1996);Keeletal.(1999);Francisetal.

(2001);M ¨oller & Fynbo (2001);Steideletal.

(1998,2005);Shimasakuetal.(2003);Ouchiet

al.(2005)).These structureshave beenfoundof

-tenasby-productsofwide fieldsurveysusing

broador narrow bandimaging.Overdensities

have beenfoundasby-productsofwide field

surveyswithbroador narrow bandimaging,

butalso throughanestablishedtechnique that isbasedonthe hypothesisthatluminousradio

sourcesare amongstthe mostmassive forming

galaxiesathighredshift(e.g.De Breucketal.

2002;Deyetal.1997;Pentericcietal.2001;Vill ar-Mart´ınetal.2005)thatmaypinpointthe l oca-tionofoverdense regions.The associationof distant,powerfulradio galaxieswithmassive

galaxyandcluster formationismainlybased

ontwo observationalclues.First,radio galaxies

form a brightenvelope inthe K-bandHubble

redshiftdiagram (De Breucketal.2002),s

ug-gestingthattheir hostgalaxiesare the prime candidatesfor later brightestcluster galaxies (BCGs)thatdominate the deeppotentialwells ofclusters.Second,highredshiftradio galaxies

have companiongalaxies,rangingfrom the red

galaxyoverdensitiesat1.5<z<2 (e.g.S´anchez & Gonz´alez-Serrano 1999,2002;Thompsonetal. 2000;Halletal.2001;Bestetal.2003;Woldetal.

2003)to the large excessesofLAEsdiscovered

throughdeepnarrow-bandimagingands

pec-troscopic follow-upwiththe VeryLarge Tel

e-scope (VLT)ofthe EuropeanSouthernObs

er-vatory(e.g.Pentericcietal.2000;Kurketal.

2003;Venemansetal.2002,2004;Venemansetal.

2005).

Buildingonthe excessesofLAEsdiscovered

inthe vicinityofdistantradio sources,we are

performinga surveyofcandidate LBGsinsuch

radio-selectedprotoclusterswithACS.InMi

-leyetal.(2004)andOverzier etal.(2006)we

reportedonthe detectionofa signi

ficantpop-ulationofLBGsaroundradio galaxiesatz=

4._{1 (TN J1338–1942)and z = 5}._{2 (TN J0924–}

2201).Here,we willpresenta detailedanal

y-sisofthe ACS observationsofprotocluster TN

J1338–1942 at z=4.1, augmented by ground-based observations with the VLT. This structure is amongst the handful of overdense regions so far discovered at z>4, as evidenced by 37 LAEs

that represent a surface overdensity of∼5

com-pared to other fields (Rhoads et al. 2000; Daw-son et al. 2004; Shimasaku et al. 2003). The

FWHM of the distribution is 625 km s−1_,∼4×

narrower than the narrowband filter used. The mass overdensity as well as the velocity struc-ture is consistent with the global properties of

z∼4 protoclusters derived from N-body

sim-ulations combined with semi-analytical model-ing (De Lucia et al. 2004; Venemans 2005, but see Monaco et al. (2005)). The protocluster may possibly harbor several sub-mm sources as well (De Breuck et al. 2004). The radio galaxy itself is extremely bright in the rest-frame UV/optical and the sub-mm, suggesting the formation of a massive galaxy. It has a complex morphol-ogy which we have interpreted as arising from AGN feedback on the forming ISM and a mas-sive starburst-driven wind (Zirm et al. 2005).

The main issues that we will attemt to ad-dress include the following. What can the ob-servational properties tell us about the star for-mation histories and physical sizes of LAEs and LBGs? In particular we wish to study these properties in relation to the overdense environ-ment that the TN1338 field is believed to be associated with, analogous to galaxy environ-mental dependencies that have been observed at lower redshifts and are predicted by models (e.g. Kauffmann et al. 2004; Postman et al. 2005; De Lucia et al. 2006). How do the clustering and mass overdensity of the TN1338 structure com-pare to the ‘field’, and what is the relation to lower redshift galaxy clusters? In Sect. 2 we will describe the observations, data reduction and methods. We present our sample of LBGs in Sect. 3, and describe the rest-frame UV and optical properties of LBGs and LAEs. In Sect. 4 we present the results of a nonparametric mor-phological analysis. In Sect. 5 we will present the evidence for a galaxy overdensity associated with TN J1338–1942 and investigate its cluster-ing properties. We conclude with a summary

of the main results and a discussion in Sect. 6.

We use a cosmology in which H0₌72 km s−1

Mpc−1_{, ΩM}

=0.27, and Ω_Λ=0.73 (Spergel et al. 2003). In this Universe, the luminosity distance is 37.1 Gpc and the angular scale size is 6.9 kpc arcsec−1 _{at z}

=4.1. The lookback time is 11.9

Gyr, corresponding to an epoch when the Uni-verse was approximately 11%of its current age. All colors and magnitudes quoted in this paper are expressed in the AB system (Oke 1971).

5.2

Ob

s

e

r

v

a

t

i

o

n

sa

n

dd

a

t

ar

e

d

u

c

t

i

o

n

5.2.1 ACSimaging

We observed one field with the ACS around the radio galaxy TN J1338–1942 (henceforward

‘TN1338’). These observations were part of the

ACS Guaranteed Time Observing high redshift cluster program. To search for candidate clus-ter members on the basis of a Lyman-break at

the approximate wavelength of Lyα_redshifted

to z=4.1, we used the Wide Field Channel to

obtain imaging through the broadband filters2

g475, r625, i775, and z850. The total observing time

of 18 orbits was split into 9400 s in each of g475

and r625, and 11700 s in each of i775and z850. The

filter transmission curves are indicated in Fig. 7.1. The r625-band may include Lyαif present.

Each orbit of observation time was split into two 1200 s exposures to facilitate the removal of cosmic rays. The data were reduced using the ACS pipeline science investigation software (Apsis; Blakeslee et al. 2003). After initial pro-cessing of the raw data through CALACS at STScI (bias/dark subtraction and flat-fielding), the following processing was performed by

Ap-sis: empirical determination of image offsets

and rotations using a triangle matching

algo-rithm, background subtraction, the rejection of

cosmic rays and the geometric correction and combining of exposures through drizzling us-ing the STSDAS Dither package. The final sci-ence images have a scale of 0.00_{05 pixel}−1_{. The}

total field of view is 11.7 arcmin2

. The radio 2

We use g475, r625, i775and z850to denote magnitudes in

Figure 5.1—Total effective throughput of the HST ACS g475r625i775z850and VLT/ISAAC KSfilters. The SED template shown

is the SB2 template from Ben´ıtez (2000) redshifted to z=4.1 following the attenuation prescription of Madau et al. (1996). The GOODS B435V606filters are indicated by dashed lines.

galaxy (α_J2000=13h₃₈m₃₀s_,δ

J2000= −19◦4203000)

is located about 10 _{away from the image}

cen-tre. The field further includes 12 of the 37 spec-troscopically confirmed LAEs (Venemans et al. 2002; Venemans et al. 2006). The resultant color image of the field is shown in Fig. 7.2, with g475

in blue, r625in green, and z850in red. The radio

galaxy clearly stands out as the sole ‘green’ ob-ject in the entire field, due to its prominent halo of Lyα_{emission observed in r}625(see Zirm et al.

2005).

We used the latest ACS zeropoints from Siri-anni et al. (2005), and an extinction value of

E(B−V)= 0.096 from Schlegel et al. (1998).

We measured the limiting magnitudes from the RMS of noise fluctuations in 10000 square aper-tures of varying size that were distributed over the images in regions free of objects. A sum-mary of the observations and the limiting mag-nitudes are listed in Table 5.1.

5.2.2 VLT optical spectroscopy and NIR

imaging

We obtained 10 hours of VLT/FORS2

spec-troscopy in service mode3

. The instrumental setup, the seeing conditions, and the method of processing of the data were similar as described in Venemans et al. (2002); Venemans et al. (2005).

Near-infrared data in the KS-band were

ob-tained with VLT/ISAAC4

. We observed a 2.0_4×

2.0_{4 field for 2.1 hours in March 2002, and for}

5.4 hours in a partly overlapping field in 2004. After dark subtraction, flat fielding and rejec-tion of science frames of poor quality, the data for each night was individually processed into a combined image using the XDIMSUM package in IRAF. Since only the data taken on the night of March 26 2002 was considered photometric, the combined images of the other nights were scaled to match that particular night using sev-eral unsaturated stars for reference. We derived the zeropoint based on observations of the

near-3

Program ID:071.A-0495(A)

4

Figure 5.2—ACS color image showing g475in blue, r625 in green and z850 in red. The field measures 11.7 arcmin 2

. The observer is observing the radio galaxy TN J1338–1942.

IR photometric standard FS 142. However, we had to adjust the zeropoint by 0.2 magnitudes to match the magnitudes of several 2MASS stars in the field. The seeing was∼0.005 (FWHM), and the galactic extinction in KSwas 0.036 mag.

Next, each of the combined images with the

native ISAAC scale of 0.00_{148 pixel}−1 _was

pro-jected onto the 0.00₀₅−1_{pixel ACS i775image}

us-ing the tasks GEOMAP/GEOTRAN_{in IRAF. Using}

andthe ISAAC images, the projection hada

typ-icalaccuracyof1.5 ACS pixels (RMS). Where

available, reference stars close to the corners

andedges ofthe images were selectedso as to

take out the effect ofthe geometric distortion on

the ISAAC frames. The registeredimages were

combinedusing a weighting basedon the

vari-ance measuredin a source-free region ofeach

image. The limiting 2σdepth in the AB5

system

was 25.2 magnitudes for a circular aperture of

1.00_{4 diameter. Areas that are onl}

ycoveredbyei-ther the 2002 or the 2004 data are shallower by

0.5 and0.3 magnitudes, respectively. The KS

-banddata cover 81%ofthe ACS field,

andcon-tain the radio galaxyand11 LAEs.

5.2.3 Objectdetectionandphotometry

Object detection andphotometrywas done

us-ing the SExtractor software package ofBertin

& Arnouts (1996). We used SExtractor in

double-image mode, where object detection and

aperture determination are carriedout on the

so-called“detection image”, andthe

photom-etryis carriedout on the individualfilter

im-ages. For the detection image we usedan in-verse variance weightedaverage ofthe r625, i775

and z850images, and a map ofthe

totalex-posure time per pixelwas usedas the

detec-tion weight map. Photometric errors were cal

-culatedusing the root mean square (RMS)

im-ages from Apsis. These images contain the

abso-lute error per pixelfor each output science

im-age. We detectedobjects byrequiring a

mini-mum of5 connectedpixels at a thresholdof1.5

times the localbackground(S/N of>3.35). The

values for SExtractor’s deblending parameters

(DEBLEND MINCONT=0.1, DEBLEND NTHRESH=8)

were chosen to limit the extent to which our of

-ten clumpyz∼4 g475-dropouts were split into

multiple objects. Our ‘raw’detection catalog

contained3994 objects. We rejectedallobjects

which hadS/N less than 5 in z850, where we

define S/N as the ratio ofcounts in the

isopho-talaperture to the errors on the counts. The

re-maining 2022 objects were

consideredrealob-5

K_s ,AB=

K_s

,Vega+1.86

jects, although theystillcontain a smallfraction ofartefacts.

We use SExtractor’s MAG AUTO to estimate

to-talobject magnitudes within an aperture radius

of2.5×rKron(Kron 1980). However, when

ac-curate color estimation is more important than

estimating a galaxy’s totalflux, for example

in the case ofcolor-selection or when

deter-mining photometric redshifts,

isophotalmagni-tudes are preferredbecause ofthe higher S/N

and the smaller contribution ofneighboring

sources. Therefore we calculate galaxycolors

from isophotalmagnitudes. These procedures

are optimalfor (faint) object detection

andaper-ture photometrywith ACS (Ben´ıtezet al. 2004).

Optical-NIR(observed-frame) colors were

de-rivedfrom combining the ACS data with l

ower-resolution groundbasedKS data in the foll

ow-ing way. We usedPSFMATCH in IRAFto

de-termine the 2Dkernelthat willmatch the point

spreadfunction in the ACS images to that

ob-tainedin the KS-band, andconvolvedthe ACS

images with this kernel. The photometrywas

done using SExtractor in double image mode,

using the KS-bandimage for object detection.

Colors involving the NIRdata were determined

in circular apertures with a diameter of1.00_{4. For}

the radio galaxyTN J1338–1942 a circular

aper-ture of3.00_{0 diameter was used, due to its signif}

-icantlylarger size.

5.2.4 Apertureandcompletenesscorrections

The photometric properties ofgalaxies are

usu-ally measured using source extraction al

go-rithms such as SExtractor. We can conveniently

use this software to determine aperture

correc-tions andcompleteness limits as a function of

e.g., the ‘intrinsic’or realapparent magnitude,

half-light radius (rhl) or the shape ofthe galaxy

surface brightness profile (see also Ben´ıtezet al.

2004;Giavalisco et al. 2004). To this endwe

populated the ACS z850image with artificial

galaxies consisting ofa 50/50

mixofexponen-tialandde Vaucouleurs profiles. We simulated

∼10,000 galaxies with∼200 per simul

atedim-age to avoidover-crowding. We tookuniformly

Table5.1—_{Summaryofobservations.}

Filter Date Texp A Depth

g475(F475W) July 11–12 2002 9400 0.359 28.46a 27.47a

r625(F625W) July 8–9 2002 9400 0.256 28.23a 27.23a

i775(F775W) July 8–9 2002 11700 0.193 28.07a 27.08a

z850(F850LP) July 11–12 2003 11800 0.141 27.73a 26.73a

KS 2002,2004 27000 0.036 25.15b 24.16b

a_{Measured in 0.}00_{45 diameter square apertures.} b_{Measured in 1.}00_{4 diameter circular apertures.}

and uniformly distributed axial ratios in the range 0.1–1.0. Galaxies were placed on the im-ages with random position angles on the sky. Using the zeropoint we scale the counts of each

galaxy to uniformly populate the range z850=20–

28 magnitudes. We added Poisson noise to the

simulated profiles, and convolved with the z850

point spread function (PSF). Next, SExtractor was used to recover the model galaxies as de-scribed in Sect. 5.2.3.

Approximately 75% of the artificial galaxies were detected. In Fig. 5.3 (top left) we show the measured rhl versus the input ‘intrinsic’ rhl.

Radii are increasingly underestimated as the in-put radii become larger, because the surface brightness gets fainter as r2

hl. On average, the

radius is underestimated by about 50% for a z850∼26 magnitude object with an intrinsic

half-light radius of 0.00_{4. The discrepancy between}

in-put and outin-put radius is generally smaller for an exponential than for a de Vaucouleurs profile.

In Fig. 5.3 (top right) we show the aperture corrections defined by the difference between the Sextractor MAG AUTO magnitude and the total magnitude of the simulated profile. The amount of flux missed rises significantly to-wards fainter magnitudes, with a 0.5–1.0 nitude correction for objects with output

mag-nitudes of z850₌25₋27. For future reference,

we also show the modeled versus the recovered ellipticities, b/a, in Fig. 5.3 (bottom left). The correspondence becomes poorer towards fainter and smaller objects, as well as more elongated ones. This is because the PSF causes small

ob-jects to appear generally rounder, and because the structural parameters are measured by SEx-tractor after convolving the image with a (Gaus-sian) detection kernel. Finally, in Fig. 5.3

(bot-tom right) we show the z850completeness limits

as a function of z850 and r_hl. About 50%

com-pleteness is reached at z850₌26−26.5 for

unre-solved or slightly reunre-solved sources. Note that the 50% completeness limit will lie at measured MAG AUTO magnitudes that are fainter by 0.5– 1.0 magnitude, given the significant aperture corrections presented in Fig. 5.3 (top right).

In the analysis that follows we will apply ap-proximate corrections to the physical quantities

derived from measured rhland magnitudes (e.g.

physical sizes, luminosities, and SFRs) based on the above results for exponential profiles. The quoted angular sizes and magnitudes are al-ways as measured.

5.2.5 Photometric redshift technique

We will use the Bayesian Photometric Redshift code (BPZ) of Ben´ıtez (2000) to estimate galaxy

redshifts, zB. For a complete description of

BPZand the robustness of its results, we refer the reader to Ben´ıtez (2000) and Ben´ıtez et al. (2004). Our library of galaxy spectra is based on the elliptical, intermediate (Sbc) and late type spiral (Scd), and irregular templates of Cole-man et al. (1980), augmented by two starburst

galaxy templates with E(B−V)∼0.3 (SB2) and

23 24 25 26 27 28 mtrue 0.2 0.4 0.6 0.8 rhl, tr u e ( a rc se c ) 1₀ 1₀ 2 0 2₀ 3 0 3₀ 5₀ 5₀ 7 0 7₀ 8₀ 8₀ 90 90 9₅ 95

Figure 5.3— Topleft:Intrinsicrhl,zversus rhl,zmeasuredby SExtractor for de Vaucouleurs profiles (squares) andexponentials (circles). The difference between intrinsicandmeasuredradius is smaller for exponentials. The intrinsicsizes are increasingly underestimatedwhen goingto fainter magnitudes,e.g. from z850∼24(upper lines) to z850∼26 (lower lines). – Topright:The difference between z850MAG AUTO andtotal‘intrinsic’magnitudes for de Vaucouleurs profiles (squares) andexponentials

(circles). – Bottomleft:Input vs. recoveredellipticity (b/a) for sources with0.00₁<_rhl<_0.00_{3 (blue), 0.}00₃<_rhl<_0.00_{5 (green),} 0.00₅<rhl<_0.00_{7 (yellow), and0.}00₇<rhl<_0.00_{9 (red).– Bo}ttomright:Compl_{etenesslimitsinz}850asa functionoftotal‘intrinsic’

magnitudesandrhl, where completenessisdefinedasthe ratioofthe numberofobjectsdetectedtothe numberofartificial

Figure 5.4—g475–r625 versus r625–z850 for model SEDs

(points) simulated using the Bruzual & Charlot (2003) li-braries. The model parameter grid is given in Table 5.2. Galaxies at z=₄._{1 are shown as (yellow) large solid} cir-cles. The shaded area is defined by g475–r625≥1.5, g475–

r625≥r625–z850+1.1, r625–z850≤1.0. The spectral tracks are

an elliptical (red solid line), an Sbc (red dashed line), an Scd (red dotted line), and a 100 Myr constant star formation model with E(B−V)=₀._{0 (blue solid line) and E(B}−V)= 0._{2 (blue dotted line). Redshifts are indicated along the} tracks. The redshift of the overdensity of Venemans et al. (2002) is marked by blue squares (z≈4._{1). Green stars mark} the stellar locus based on the stellar SED library of Pickles (1998).

Charlot(2003). The latter twotemplateshave

beenfoundtoimprove the accuracyofBPZfor

veryblue, younghighredshiftgalaxiesinthe

UDF(Coe etal. 2005). BPZmakesuse ofa

pa-rameter ‘ODDS’definedasP(|z−zB| <∆z) that

givesthe totalprobabilitythatthe true redshift iswithinanuncertainty∆z. For the uncertainty we cantake the empiricalaccuracyofBPZfor

the HDF-N whichhasσ =₀.06(1+_zB). For

a Gaussianprobabilitydistributiona 2σ c

on-fidence intervalcenteredonzB wouldgetan

ODDS of>0.95. The empiricalaccuracyofBPZ isσ≈0.1(1+_zB) for objectswithI814. 24 and

z . 4 observedinthe B435V606I814-bandswith

0 1 2 3 4 5 6 z 0.0 0.2 0.4 0.6 0.8 1.0 S el ec ti o n e ff ic ie n cy 0 1 2 3 4 5 6 z 0.0 0.2 0.4 0.6 0.8 1.0 S el ec ti o n e ff ic ie n cy 0 1 2 3 4 5 6 z 0.0 0.2 0.4 0.6 0.8 1.0 S el ec ti o n e ff ic ie n cy

Figure 5.5—Selection efficiency for z∼4 LBGs. The solid histogram shows the fraction of model galaxies that meet the selection criteria in each redshift bin. The dotted his-togram shows the selection efficiency for model galaxies with ages less than 100 Myr and 0<_E(B−V)<₀._{3. The} dashed histogram shows the fraction of models with ages greater than 0.5 Gyr selected, illustrating possible contam-ination of our z∼4._{1 sample by relatively old galaxies at} z∼2._{5. Another source of contamination is the inclusion of} Balmer-breakobjects at z∼0._{5. The shaded region indicates} the redshift interval (z=4.₀₇−4._{13) of the protocluster of} LAEs.

ACS toa depthcomparable toour observations

(Ben´ıtezetal. 2004). Note thatwe willbe appl y-ingBPZtogenerallyfainter objectsatz∼4 ob-serveding475r625i775z850. The true accuracyfor

sucha sample hasyettobe determinedempi

r-ically. The accuracyofBPZmaybe improved

byusingcertainpriors. We applythe commonly

usedmagnitude prior thatisbasedonthe

mag-nitude distributionofgalaxiesinrealobs erva-tions(e.g. the HDF).

5.2.6 Template-basedcolor-colorselectionof protoclusterLBGcandidates

We extracted LBGsfrom our catalogsusing

color criteria thatare optimizedfor detecting

1999;Ouchi et al. 2004a;Giavalisco et al. 2004a). To define the optimal selection for our filters we followed the approach employed by Madau et al. (1996). We used the evolutionary stel-lar population synthesis model code GALAXEV (Bruzual & Charlot 2003) to simulate a large variety of galaxy spectral energy distributions (SEDs) using:(i) the Padova 1994 simple stel-lar population model with a Salpeter (1955)

IMF with lower and upper mass cutoffs mL =

0.1 M and mU=100 M of three metallicities

(0.2Z,0.4Z,Z), and (ii) the predefined star

formation histories for instantaneous burst,

ex-ponentially declining (τ =0.01 Gyr) and

con-stant (t=₀.1,1.0 Gyr) star formation. We ex-tracted spectra with ages between 1 Myr and 13 Gyr, applied the reddening law of Calzetti

et al. (2000) with E(B−V) of 0.0–0.5, and

redshifted each spectrum to redshifts between 0.001 and 6.0, including the effects of

attenua-tion by the IGM using the Madau et al. (1996)

recipe. Galaxies were required to be younger than the age of the universe at their redshift, but other parameters were not tied to redshift. The full parameter grid is summarized in Ta-ble 5.2. While this approach is rather simplis-tic due to the fact that the model spectra are not directly tied to real observed spectra and lumi-nosity functions, it is reasonable to expect that they at least span the range of allowed physi-cal spectra. The resulting library can then be used to define a robust set of color criteria for selecting star-forming galaxies at the appropri-ate redshift, and estimating color-completeness

and contamination (Madau et al. 1996)6

. We extracted the model colors by folding each spectrum through the corresponding ACS fil-ter transmission curves. No photometric scat-ter was applied to the models. The g475–r625and

r625–z850 color-color diagram is shown in Fig.

6

Alternatively, these quantities can be estimated with more accuracy by carrying out extensive Monte Carlo simu-lations of model galaxies that follow the observed size and color distributions of dropout galaxies when observed with the typical photometric quality of the data (e.g. Steidel et al. 1999; Giavalisco et al. 2004a; Bouwens et al. 2005b). How-ever, this requires samples that are significantly larger than our current sample for TN1338.

5.4. LBGs at z∼4 can be isolated from lower

redshift objects by a selection that is based on the g475–r625 and r625–z850colors. For

compari-son, we have overplotted the color-color tracks of the standard spectral types from Ben´ıtez

(2000). We elected to use the r625–z850 color

in defining our selection region (instead of the r625–i775color used in Miley et al. (2004)) due to

the greater leverage in wavelength.

The color-color region that we use to select

z∼4.1 LBGs is defined as:

g475−r625≥1.5,

g475−r625≥r625−z850+1.1,

r625−z850≤1.0. (5.1)

Fig. 5.5 shows the color selection efficiency as a function of redshift, defined as the number of galaxies selected in a redshift bin, divided by the total number of model galaxies in that red-shift bin. The solid histogram indicates the frac-tion of model galaxies meeting the selecfrac-tion cri-teria. The resulting redshift distribution has an approximately constant maximum efficiency of ∼45% for 3.5<_z<4.5. If we limitthe model galaxiesto ageslessthan100 Myr and0<E(B₋ V)<0.3 (consistentwiththe average LBGpop-ulationatz∼3−4 (Papovichetal. 2001;Stei -deletal. 1999)), the color completeness(dot

-tedhistogram) becomes∼ 90% for modelsat

z∼4.1. The dashedhistogram showsthe frac -tionof modelswithagesgreater than0.5 Gyr s e-lected, illustratingthe mainsourcesof cont am-inationinour z∼4.1 sample, namelyfrom rel -ativelyoldgalaxiesatz∼2.5 andthe possible inclusionof Balmer-breakobjectsatz∼0.5.

5.2.7 GOODSsimulatedimages

To determine whether TN1338isalso hostto an

overdensityof LBGsatz ∼4.1, we willwant

to compare the number of g475-dropoutsfound

inour ACSfieldwiththatfoundina random

fieldonthe sky. Unfortunately, atpresent, there

are notmanyACSfieldsavailable, withc

om-parable depthsing475, r625, andz850to carry

outsuchcomparison. We therefore avai

Table5.2—_{Parametergridofsyntheticspectra.}

SED Parameter Values

Instantaneous burst –

Exponential star formation timescales 0.01 Gyr

Constant star formation durations 0.1 1.0 Gyr

Ages 0.001 0.005 0.01 0.03 0.07 0.1 0.2 0.3 0.4 0.5 0.60.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 2.0 3.0 5.0 7.0 13.0 Gyr Metallicities 0._2Z0.4Z1.0Z E(B−V) 0.0 0.1 0.2 0.3 0.4 0.5 Redshifts 0.001 0.01 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.60.65 0.7 0.75 0.8 0.85 0.9 0.95 1.0 1.1 1.2 1.3 1.4 1.5 1.61.7 1.8 1.9 2.0 2.1 2.2 2.3 2.4 2.5 2.62.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 3.5 3.63.7 3.8 3.9 4.0 4.1 4.2 4.3 4.4 4.5 4.64.7 4.8 4.9 5.0 5.1 5.2 5.3 5.4 5.5 5.65.7 5.8 5.9 6.0

control. The 3 orbit B435, 2.5 orbit V606, 2.5

orbit i775, and 5 orbit z850 coverage is

strik-ingly similar in depth and much larger in cov-erage, to the g475r625i775z850 imaging we have

on TN1338, suggesting that with simple wave-length interpolation, we should be able to mir-ror our TN1338 selection.

Though there are many ways to have per-formed this interpolation, we chose to perform the interpolation directly on the ACS data itself, changing it from the observed B435V606i775z850

filter set to the g475r625i775z850 filter set. This

transformation was performed on a pixel-by-pixel basis, using the formula:

fY i,j = _Ii ,jg(SED,Y,i775,z) + ΣXH X=XL  |λ(Y)− λ(X)| λ(XH)− λ(XL)  × g(SED,Y,X,_{z)∆ f}X i,j, (5.2) (5.3) where fY

i,jis the fluxatpixel(i,j) insome band

Y, Ii,j is the best-fitfluxes ineac

hpixel(ex-pressedas ani775-bandflux), g(SED,Y,X,z) is

a generalizedk-correctionfrom some bandX

toanother bandY for some SED andredshift

z,λ(X) is the meanwavelengthfor some band

X, the summationΣXH

X=XL runs over those bands

whichimmediatelystraddle the Y band,andthe

∆ fX

i,jterms whichaccountfor the error inthe fits

toindividualpixels.The best-fitfluxes Ii,jwere

determinedbyminimizing χ2=ΣX " Ii,jg(SED,X,i775,z)−f X i,j σ_iX ,j # , (5.4) where fX

i,j andσi,Xj are the fluxandits unc

er-tainty, respectively, inthe X bandatpixelpo-sition(i,j).The error terms ∆ fX

i,j are equalto

fi,j−g(SED,X,i775,z).The firstterm inEq.5.3

is a generalizedk-correctionappliedtothe best

-fitmodelSEDs, while the secondis anint

erpo-lationappliedtothe fluxresiduals from the fit.

This is nearlyidenticaltoexpressions from

Ap-pendixB1of Bouwens etal.(2003)

andrepre-sents a slightupdate tothatprocedure.

The redshifts z andSEDs SED we use for

individualpixels are baseduponaninit

ialob-jectcatalogwe made of eachfieldbefore

do-ingthe transformation. Objects are detected off a χ2

image (Szalayetal.1999) constructed from the V606i775z850-bandusinga fairl

(SEx-tractor DEBLEND MINCONT=0.005). Best-fit redshifts and SEDs are then estimated for each object from the photometry. These model pa-rameters, in turn, are assigned to all the pix-els which make up these objects (according to the SExtractor deblending maps), and thus used in the transformation given by Eq. 5.3. Only

pixels belonging to objects with colors (B435₋

V606)>0.8,(B435₋V606)>0.6(V606₋z850)+0.5,

(B435₋V606) >3.375(V606₋z850)₋4.575 were

transformed.

Since our ACS reduction of the TN1338field had a different pixel scale (i.e., 0.00_{05) thanthat}

ofthe GOODSv1.0 reduction(0.00

03:Giavalisco etal. 2004), we didnotuse thatreductionas the

basis for our simulationofthe CDF-SGOODS

field. Instead, we made use ofanindependent

reductionwe hadmade ofthe GOODSfield

withApsis. Thatreductionwas performedon

a 0.00_{05 grid, using a procedure nearlyidentical}

tothatdescribedinBouwens etal. (2005b), but

using a ‘Lanzcos3’kernel(whichmatches the

TN1338 ACSreduction).

5.3

Pr

ope

r

t

i

e

sofLBGsa

ndLAEsi

n

TN1

338

We willapplythe color-color sectiondefinedin the previous sectiontothe TN1338 fieldt ose-lecta sample ofcandidate z∼4.1 LBGs (g475

-dropouts) and studytheir properties inSect. 5.3.1. InSect. 5.3.2 we willstudythe same prop-erties for the sample ofz=4.1 LAEs withinthe

ACSfield.

5.3.1 Theg475-dropoutsample

Using the selectioncriteria definedinEq. 5.1

we extracteda sample ofLBGs fromthe TN1338

field. Intotalthere are 66 suchobjects inTN1338 withz850<27.0, 51 ofwhichhave z850<26.5, and

32 ofwhichhave z850<26.0. The color-color di

-agram is showninFig. 7.7. Althoughthe stel -lar locus (basedonPickles (1998)) lies outside the regiondefinedbyour selectioncriteria, we requiredobjects tohave a SExtractor stellarity indexof< 0.85 (non-stellar objects withhigh confidence). This shouldexclude essentiallyall

Figure 5.6 —Color-color diagram ofg475-dropoutsin

TN1338(circles)andthe detectioncatalog(points). The shadedregionshowsour selectionwindow (Eq.1) .Con-firmedLAEsfrom Venemansetal.(2002)are markedbyred stars,andthe radio galaxybythe square.See the captionof Fig.5.4for further details.

star-like objects from our sample. 5.3.1.1 Starformationrates The characteristicluminosity, L∗

z=4, ofthe LBG

luminosity function atz ∼ 4 corresponds to z850∼25.0 (Steideletal. 1999). The sample c

on-tains twoobjects, one ofwhich is the radio galaxy, witha luminosityof∼ 6L∗ (i775≈23).

The remainder ofthe sample spans luminosi

-ties inthe range ∼0.4−2L∗, where we have appliedaperture corrections ofupto∼1 mag-nitude basedonthe exponentialprofiles inFig. 5.3.

We calculated SFRs from the emission-line free UVfluxat1500 ˚A (i775) using the c

onver-sionbetweenluminosityandSFRfor a Salpeter initialmass function(IMF) giveninMadauet

al. (1998): SFR (M yr

−1

) = L1500˚A (erg s −1

Hz−1

23 24 25 26 27 z850 -4 -3 -2 -1 0 1 β iz E(B-V)= 0.1 0.005< age< 0.15Gyr E(B-V)= 0.0 0.05< age< 0.3 Gyr E(B-V)= 0.3 0.001 < age< 0.3 Gyr 23 24 25 26 27 z850 -4 -3 -2 -1 0 1 β iz

Figure 5.7 — z850 versus βiz for g475-dropouts

de-tected/undetected in KS(filled/open circles), LAEs (stars),

and the radio galaxy (square). The best-fit linear relations are indicated (thick lines, see text for details). The thin solid line is the relation for B435-dropouts (Bouwens et al. 2005b).

The best-fit SED from Papovich et al. (2001) redshifted to z=4 hasβiz≈ −1.4 (thindashedline). Shadedregionsare for E(B−V)=0.0 withagesbetween0.05and0.3 Gyr (bot -tom lightshadedregion),E(B−V)=0.1 withagesbetween 0.005and0.15Gyr (darkshadedregion),andE(B−V)=0.3 withagesbetween0.001 and0.3 Gyr (toplights hadedre-gion),assuminganexponentialstar formationhistory(τ = 10 Myr) with0.2Zmetallicityanda Salpeter IMF.

the mainsequence, the UV luminosityi

spro-portionaltothe SFR, relativelyindependent of

the prior star formationhistory. The SFRsare listed inTable 7.3. The radiogalaxyand object

#367each have a SFRof∼95 Myr−1. The

me-dianSFRofthe entire sample is∼8 M yr−1.

Although we assumed here that the LBGsare

dust-free, one could multiplythe SFRsbya fac

-tor of2.5 tocorrect for anaverage LBGextinc

-tionofE(B−V)≈0.1 (see next section) giving

a medianSFRof∼20 Myr

−1

.

5.3.1.2 UVContinuum colors

We calculate the UVcontinuum slopesfrom the i775–z850color. Thiscolor spansthe rest-frame

wavelength range from∼ 1400 ˚A to∼ 2000

˚

A. We assume a standard power-law spectrum

with slopeβ( fλ∝ λβ, sothat a spectrum that is

flat infν hasβ = −2). We calculate

βiz= log10Q_Q850 775−0.4(i775−z850) log10λ 775 λ850 −2, (5.5)

where λ775and λ850are the effective bandpass

wavelengths, and Q775and Q850are the fractions

ofthe continuum fluxesremainingafter appl

y-ingthe recipe for foreground neutralhydrogen

absorptionofMadau(1995). The breakat res

t-frame 1216˚A onlystartstoenter the i775-band

for galaxiesat z & 4.7. ThusQ775and Q850are

unityand β willbe relativelyindependent of

redshift for 3.5 . z . 4.5. The uncertaintieson βizwere obtained bypropagatingthe individual

errorsonthe measured magnitudes. The

mea-sured slopesare plotted inFig. 5.7. Excluding the twobrightest sources, we findhβizi = −1.95.

Thisissignificantlybluer thanthat found byPa-povich et al. (2001), although it isconsistent at

the bright magnitude end where the compari

-sonwith L∗

galaxiesisappropriate (thindashed line).

We have modeled the dependenciesofthe

slope on age and dust usinganexponential

star formationhistory(τ =10 Myr) with 0.2Z

metallicityand a Salpeter IMF. For a constant

E(B−V)≈0.0 the range ofslopesfavoursages

inthe range 50–300 Myr. A high dust content

(E(B−V)≈0.3) isincompatible with the ma-jorityofthe slopesobserved. A linear fit tothe data gave a slope-magnitude relationofβiz= (−0.16±0.05)(z850−25)−1.84 (thicksolid line),

which remainsvirtuallyunchanged whenwe

excluded the twobrightest objects(thickdashed line). There could be a possible higher inc om-pletenesstowardsfaint, relativelyred objects

(e.g. Ouchiet al. 2004a). However, the effect

islikelytobe much smaller thanthe observed correlationasshownbysimulationsinc orporat-ingB435-dropout selectioninGOODS(Bouwens

et al. 2005b). Our relationisingood

agree-ment with that ofB435-dropoutsinGOODSof

−0.21±0.03 mag−1 _{found byBouwenset al.}

(2005b). The best-fit relationspansagesinthe range 5–150 Myr for a constant E(B−V)≈0.1. A similar slope-magnitude relationisals

Figure 5.8—Rest-frame UV-optical colors of the g475

-dropouts(circles),the LAEsL9andL25 (stars),andthe radiogalaxy(square).Arrowsindicate 2σ_limitsfor non-detectionsinKS(errorsomittedfor clarity).Linesindicate

the colorsofaτ =10 Myr SED (0._{2Z) withagesinGyr} alongthe trackfor E(B−_V)=0._{0 (dotted) andE(B}−_V)= 0._{15 (solid).The large circle wasobtainedfrom a KS-band} stackof12 g475-dropoutshaving25.3<i775<26.4.The large

star wasobtainedfrom a KS-bandstackof5 LAEswithin

a similar magnitude range.Their i775–K_S colorsdiffer by

∼₀._{7 magnitude.}

et al. 2004a) andmayimplya mass-extinction

or a mass-metallicityrelationrather thana

re-lationwithage (Bouwenset al. 2005b). I

nter-pretingthe slope-magnitude relationasa mass -extinctionrelationimpliesE(B−V)≈ 0.13at z850_≈23andE(B₋V)_≈0.0 at z850_≈27for a

fixedage of70 Myr.

5.3.1.3 Rest-frameUVtoopticalcolors At z∼4._{1,the filtersi}775, z850andKSprobe the

rest-frame at ∼1500 ˚A, 1800 ˚A and4300 ˚A, re-spectively. We detected13ofthe g475-dropouts

inthe KS-bandat >2σ. InFig. 5.8we show

the i775–KSversusi775–z850color diagram. i775–

KS color ismore sensitive tothe effectsofage

anddust thani775–z850, due toitslonger lever

Figure 5.9 —Color magnitude diagram ofthe g475

-dropouts(circles).The dashedline indicatesthe approxi -mate 2σ_{detectionlimits.The tracksare forτ =10 Myr SEDs} withdifferentstellar masses(atts f= ∞) of0.03,0.1,0.5 and 2×1010Mfor E(B−V)₌0._{0 (dotted) andE(B}−V)₌0.₁₅ (solid).The thicksolidline indicatesthe ‘blue envelope’of Papovichetal.(2004),andsuggestsa color-magnitude rel a-tioninwhichluminositycorrelateswitheither age or dust. See the captionofFig.5.8for further details.

arm inwavelength. Comparingthe colorsto

the best-fit LBGSEDfrom Papovichet al. (2001) redshiftedtoz∼4 showsthat the observedcol -orsare consistent withagesinthe range 10-100

Myr, althoughthere willbe degeneracywith

dust. Non-detectionsinthe KS-bandsuggests

that more than50% ofthe g475-dropoutshave

ageslessthan70 Myr, witha significant fraction

lessthan30 Myr. The radiogalaxyisamong

the reddest objects, althoughit haslarge gra-dientsini775–KS amongitsvariousstellar and

AGN components(see Zirm et al. 2005). The

age of∼100 Myr wasderivedbasedoni

tsaver-age i775–KScolor, it mayactuallyconsist ofstel

-lar componentsthat are bothsignificantlyolder

andyounger than100 Myr, anddoesnot rule

out a significantlyhigher mass-weightedage for

In Fig. 5.9we plot the i775–KS versus KS

color-magnitude diagram. Papovich et al. (2004) found evidence for a trend of generally red-der colors for galaxies that are brighter in KSin

GOODS. The effect is not likely to be a selection effect because the objects are selected in the UV. Papovich et al. (2004) suggest that age and/or

dust of LBGs at z∼3−4 may increase with

in-creasing rest-frame optical luminosity. Our data are consistent with that conclusion.

5.3.1.4 Sizes

We measured rhl in z850 using SExtractor by

analysing the growth curve for each object out to 2.5_×rKron. Excluding the exceptionally large

radio galaxy, the measured radii range from unresolved (∼0.00_{07) to 0.}00_{42, corresponding to}

physical diameters of . 7 kpc at z∼4. The

aver-age radius is 0.00_{17 or∼1}._{4 kpc. If we divide our} sample into two magnitude bins each contain-ing an approximately equal number of objects

(achieved by placing a cut at z850₌26.1

magni-tude), the mean rhl are 0.0021±0.0001 (error

repre-sents the standard deviation of the mean) and 0.00_14±0.00_{01 in the bright and faint bins,}

respec-tively. The difference is expected to be largely due to a larger flux loss in the fainter sample (see Fig. 5.3), although fainter galaxies are also likely to be smaller because of the r200 ∼ Vc ∼ L1/3

luminosity-size relationship, where r200 is the

virial radius and Vc is the circular darkmatter

halo velocity (see Mo et al. 1998).

The i775-band morphologies of the g475

-dropouts are shown in Fig. 5.10. A separate sec-tion will be devoted to a nonparametric analysis of these morphologies and a comparison to field samples (see Sect. 5.4).

5.3.2 _Lyαgalaxies

Venemans etal. (2002) found anoverdensity

ofLAEs (EW0,Lyα>15 ˚A), allspectroscopically

confirmed tolie within625±150 km s−1

ofz=

4.11. Allofthe 12 LAEs inthe ACSfield have

beendetected inr625, i775and z850(see Table 7.2).

5.3.2.1 Starformationrates

The z850 magnitudes are inthe range 25.3–27.4,

corresponding toa luminosity range of∼0.2−

1.0L∗. The SFRs are ∼1−14 M yr −1 with a medianof5.1 M yr −1 (notincluding the effectofdust). Venemans etal. (2005) calc

u-lated the SFRs from Lyα using SFR(M yr −1 ) =8.7LLyα(erg s−1)/1.12×10 41 , from Kennicutt

(1998) withthe standard assumptionofcase B

recombination(Brocklehurst1971,LHα/LLyα = 8.7 for gas thatis optically thicktoHI

reso-nance scattering and nodust). Ingeneral, we

find good agreementbetweenthe SFRs calc

u-lated from the UV compared toLyαwitha me-dianUV-to-LyαSFRratioof1.3.

5.3.2.2 UVcontinuum colors

The UV slopes, βiz, ofthe Lyαemitters are i

n-dicated inFig. 5.7 (stars). The slope canbe constrained relatively wellfor the four bright -estemitters, whichhave−2.1±0.4,−2.0±0.6, −1.9±0.7, and−2.5±0.5. The LAEslopes scat -ter around the βiz-magnitude relationfor the

g475-dropouts found inSect. 5.3.1.2, witha

sam-ple average of−1.7±1.2. These slopes are c on-sistentwitha flat(in fν) continuum, thereby

favouring relatively lowages and little dust. 5.3.2.3 Rest-frameUVtoopticalcolors

None ofthe 11 LAEs covered were detected in

KSatthe>2σ level. We created a stackofthe

KS-band fluxes for the 5 LAEs thatfellinthe

deepestpartofour NIRimage. The

subsam-ple had 25.3<z850<26.4 andh(i775–z850)i ≈0.0.

We obtained a 3σdetectionfor the stackfinding KS=25.8+0.4−0.362and hence i775–K_S≈0.0. We c

om-pared this toa stackof12 g475-dropouts witha

similar range inz850 magnitudes and i775–z850

colors, whichgave a 7σ detectionwithKS= 25.14+0.16

−0.13and i775–KS≈0.7. The results from

the stacks have beenindicated inFigs. 5.8 and 5.9. The difference inthe i775–K_Scolor is signi

fi-cantat∼2σ. A difference betweenthe KSlumi

Table5.3—_{Propertiesofthez∼4 Lymanbreaksample.} ID α_J2000 δ_{J2000 (g}475–r625)a (r625–z850)a (i775–z850)a z850b rhl,z SFR c UV 2707/RG 13:38:26.05 –19:42:30.47 3._42±0._{17 −0}._61±0._{03 0}._09±0._{03 23}._05±0._{05 0.}00_{62 93}.₇+2 .71 −2 .63 367 13:38:32.75 –19:44:37.27 1._70±0._{06 0}._52±0._{02 0}._07±0._{02 23}._10±0._{02 0.}00_{20 94}.₇+1.10 −1 .09 1991 13:38:27.84 –19:43:15.19 1._88±0._{25 0}._43±0._{08 0}._05±0._{07 24}._43±0._{12 0.}00_{42 28}.₈+2.24 −2 .08 3018 13:38:24.31 –19:42:58.06 1._73±0._{14 0}._29±0._{06 −0}._01±0._{05 24}._49±0._{07 0.}00_{20 34}.₂+1.19 −1 .15 3216 13:38:22.37 –19:43:32.41 1._86±0._{15 0}._33±0._{05 0}._10±0._{05 24}._54±0._{06 0.}00_{21 26}.₃+0.92 −0 .89 3116 13:38:24.21 –19:42:41.85 1._55±0._{14 0}._36±0._{06 −0}._00±0._{06 24}._67±0._{06 0.}00_{20 25}.₄+0.93 −0 .90 959 13:38:32.67 –19:43:3.673 1._88±0._{25 0}._55±0._{07 0}._28±0._{07 24}._73±0._{09 0.}00_{23 18}.₂+1.19 −1 .12 2913 13:38:23.68 –19:43:36.59 1._77±0._{22 0}._47±0._{07 −0}._01±0._{06 24}._94±0._{10 0.}00_{26 24}.₂+1.16 −1 .11 2152 13:38:26.92 –19:43:27.60 1._56±0._{17 0}._15±0._{08 −0}._01±0._{08 24}._97±0._{15 0.}00_{32 22}.₄+1.74 −1 .62 2799 13:38:24.88 –19:43:7.415 1._72±0._{17 0}._19±0._{07 −0}._03±0._{07 25}._03±0._{09 0.}00_{18 18}.₆+0.97 −0 .92 2439 13:38:25.35 –19:43:43.65 1._69±0._{24 0}._46±0._{09 0}._14±0._{08 25}._08±0._{10 0.}00_{25 17}.₇+1.03 −0 .97 3430 13:38:21.21 –19:43:41.99 1._74±0._{22 0}._49±0._{08 −0}._03±0._{07 25}._10±0._{09 0.}00_{16 17}.₀+0.88 −0 .83 2407 13:38:24.35 –19:44:29.15 1._54±0._{21 0}._43±0._{09 0}._04±0._{08 25}._11±0._{11 0.}00_{26 15}.₆+1.04 −0 .97 2839 13:38:25.90 –19:42:18.39 2._23±0._{45 0}._75±0._{09 0}._12±0._{07 25}._14±0._{10 0.}00_{15 15}.₆+0.98 −0 .92 1252 13:38:31.98 –19:42:37.47 1._57±0._{16 0}._39±0._{07 −0}._03±0._{06 25}._25±0._{08 0.}00_{11 16}.₃+0.70 −0 .67 227 13:38:33.02 –19:44:47.57 1._57±0._{17 −0}._06±0._{09 −0}._14±0._{09 25}._29±0._{19 0.}00_{23 20}.₄+1.22 −1 .15 2710/L9 13:38:25.10 –19:43:10.77 1._78±0._{25 0}._49±0._{09 −0}._02±0._{07 25}._34±0._{08 0.}00_{14 14}.₄+0.60 −0 .58 1815 13:38:29.01 –19:43:3.275 1._67±0._{24 0}._27±0._{11 0}._01±0._{10 25}._50±0._{12 0.}00_{17 11}.₈+0.80 −0 .75 1152 13:38:32.62 –19:42:25.15 >₂._{13 0}._48±0._{14 −0}._19±0._{11 25}._57±0._{20 0.}00_{28 15}.₀+1.28 −1 .18 2755 13:38:24.95 –19:43:16.89 1._78±0._{53 0}._36±0._{19 0}._12±0._{18 25}._59±0._{21 0.}00_{33 10}.₃+1.40 −1 .23 3304 13:38:23.67 –19:42:27.37 1._66±0._{33 0}._56±0._{12 0}._08±0._{10 25}._60±0._{12 0.}00_{15 10}.₂+0.77 −0 .71 1819 13:38:29.61 –19:42:38.19 1._98±0._{33 0}._39±0._{10 0}._15±0._{09 25}._60±0._{15 0.}00_{14 11}.₀+0.97 −0 .89 3159 13:38:22.21 –19:43:50.13 >₁._{59 0}._49±0._{23 0}._11±0._{20 25}._63±0._{16 0.}00_{41 7}.₈₃+1.04 −0 .92 309 13:38:34.77 –19:43:27.59 1._55±0._{27 0}._20±0._{13 0}._06±0._{12 25}._69±0._{15 0.}00_{21 10}.₁+0.92 −0 .85 1808 13:38:30.04 –19:42:22.51 1._61±0._{27 0}._36±0._{11 −0}._10±0._{10 25}._71±0._{12 0.}00_{15 10}.₁+0.73 −0 .68 3670 13:38:20.73 –19:43:16.32 2._09±0._{47 0}._50±0._{12 0}._07±0._{10 25}._81±0._{11 0.}00_{14 8}.₇₇+0.61 −0 .57 633/L25 13:38:34.96 –19:42:24.95 1._68±0._{33 0}._32±0._{13 0}._00±0._{12 25}._81±0._{15 0.}00_{24 9}.₉₄+0.78 −0 .72 2524 13:38:24.47 –19:44:7.263 >₂._{01 0}._60±0._{14 −0}._02±0._{12 25}._86±0._{21 0.}00_{17 13}.₃+1.03 −0 .96 1461 13:38:31.37 –19:42:30.95 >₁._{97 0}._56±0._{15 0}._07±0._{13 25}._89±0._{15 0.}00_{20 8}.₂₂+0.77 −0 .70 3177 13:38:22.97 –19:43:16.07 1._61±0._{35 0}._27±0._{16 0}._12±0._{15 25}._99±0._{23 0.}00_{18 8}.₇₀+1.09 −0 .97 1668 13:38:26.93 –19:44:53.25 >₁._{82 0}._72±0._{17 0}._16±0._{13 25}._99±0._{16 0.}00_{17 5}.₅₈+0.73 −0 .64 358 13:38:32.12 –19:45:4.687 2._00±0._{54 0}._48±0._{15 0}._03±0._{13 25}._99±0._{19 0.}00_{20 7}.₈₇+0.87 −0 .79 2569 13:38:26.38 –19:42:43.55 1._94±0._{44 0}._36±0._{14 0}._03±0._{13 26}._04±0._{15 0.}00_{13 7}.₀₈+0.64 −0 .59 3131 13:38:25.27 –19:41:55.49 >₂._{15 0}._28±0._{15 −0}._09±0._{14 26}._04±0._{19 0.}00_{16 8}.₂₉+0.83 −0 .75 a

Isophotalcolors. The limitsare 2σ_.

b

Totalmagnitudes.

c

SFRestimatedfrom the UVcontinuum flux(i775).

discussed inSect. 5.6.

5.3.2.4 Sizesandmorphologies

We calculated rhl,r from the r625-band, the filter

thatincludes Lyα, and compared ittothe rhl

ofthe continuum calculated from the z850-band

(Table 7.2). The meanrhl are 0.0013 inr625 and

0.00_{12inz850. Atz=4}.1, the measured angular sizes correspond tophysicalradii of<3 kpc,

witha meanvalue of∼1 kpc. We donotfind

evidence forthe sources tobe more extended in r625 thanthey are inz850, suggestingthatLyα

emissionis distributed ina very similarway

tothe continuum. One exceptionis source L7

whichhas rhl,r=0. 00_{18compared tor} hl,i=0. 00₁₃ and rhl,z=0. 00_11.

We have measured the rhl from a sample of