Large Area Survey for z=7 Galaxies in SDF and GOODS-N: Implications for Galaxy Formation and Cosmic Reionization
Masami Ouchi, Bahram Mobasher, Kazuhiro Shimasaku, Henry C. Ferguson, S. Michael Fall, Yoshiaki Ono, Nobunari Kashikawa, Tomoki Morokuma, Kimihiko Nakajima, Sadanori Okamura, Mark Dickinson, Mauro Giavalisco, Kouji Ohta
aa r X i v : . [ a s t r o - ph . C O ] O c t Accepted for Publication in The Astrophysical Journal
Preprint typeset using L A TEX style emulateapj v. 03/07/07
LARGE AREA SURVEY FOR z = 7 GALAXIES IN SDF AND GOODS-N:IMPLICATIONS FOR GALAXY FORMATION AND COSMIC REIONIZATION ‡ Masami Ouchi , Bahram Mobasher , Kazuhiro Shimasaku , Henry C. Ferguson ,S. Michael Fall , Yoshiaki Ono , Nobunari Kashikawa , Tomoki Morokuma ,Kimihiko Nakajima , Sadanori Okamura , Mark Dickinson ,Mauro Giavalisco , Kouji Ohta Accepted for Publication in The Astrophysical Journal
ABSTRACTWe present results of our large-area survey for z ′ -band dropout galaxies at z = 7 in a 1568 arcmin sky area covering the SDF and GOODS-N fields. Combining our ultra-deep Subaru/Suprime-Cam z ′ -and y -band ( λ eff = 1 µm ) images with legacy data of Subaru and HST, we have identified 22 bright z -dropout galaxies down to y = 26, one of which has a spectroscopic redshift of z = 6 .
96 determinedfrom Ly α emission. The z = 7 luminosity function (LF) yields the best-fit Schechter parameters of φ ∗ = 0 . +2 . − . × − Mpc − , M ∗ UV = − . ± .
76 mag, and α = − . ± .
65, and indicates adecrease from z = 6 at a >
95% confidence level. This decrease is beyond the cosmic variance in ourtwo fields, which is estimated to be a factor of .
2. We have found that the cosmic star formationrate density drops from the peak at z = 2 − z = 7 roughly by a factor of ∼
10 but not largerthan ∼ z = 7, or more likely that properties of galaxies at z = 7 aredifferent from those at low redshifts having, e.g., a larger escape fraction ( & . z -dropout galaxies appear to form 60-Mpc long filamentary structures,and the z = 6 .
96 galaxy with Ly α emission is located at the center of an overdense region consistingof four UV bright dropout candidates, which might suggest an existence of a well-developed ionizedbubble at z = 7. Subject headings: galaxies: formation — galaxies: high-redshift — cosmology: observations INTRODUCTION
Recent results from deep galaxy surveys have raisedexciting questions about cosmic reionization and theearly phases of galaxy formation. These surveys haveextended the redshift frontier to z ∼ −
10, usingvarious techniques (Iye et al. 2006; Stark et al. 2007;Ota et al. 2008; Richard et al. 2008; Bouwens et al. Observatories of the Carnegie Institution of Washington, 813Santa Barbara St., Pasadena, CA 91101 Carnegie Fellow; ouchi at obs.carnegiescience.edu Department of Physics and Astronomy, University of Califor-nia, Riverside, CA 92521, USA Department of Astronomy, School of Science, University ofTokyo, Tokyo 113-0033, Japan Research Center for the Early Universe, School of Science, Uni-versity of Tokyo, Tokyo 113-0033, Japan Space Telescope Science Institute, 3700 San Martin Drive, Bal-timore, MD 21218 Department of Physics and Astronomy, Johns Hopkins Univer-sity, 3400 N. Charles St., Baltimore, MD 21218 Optical and Infrared Astronomy Division, National Astronom-ical Observatory, Mitaka, Tokyo 181-8588, Japan Research Fellow of the Japan Society for the Promotion of Sci-ence NOAO, 950 N. Cherry Avenue, Tucson, AZ 85719 Department of Astronomy, University of Massachusetts,Amherst, MA 01003 Department of Astronomy, Kyoto University, Kyoto 606-8502,Japan ‡ Based on data obtained with the Subaru Telescope, theNASA/ESA Hubble Space Telescope (HST), and Spitzer SpaceTelescope. The Subaru Telescope is operated by the National As-tronomical Observatory of Japan. HST is operated by the Asso-ciation of Universities for Research in Astronomy (AURA), Inc.,under NASA contract NAS5-26555. The Spitzer Space Telescopeis operated by the Jet Propulsion Laboratory, California Instituteof Technology under a contract with NASA. z >
6, which is suggested by Fan et al.(2006) who find that the Gunn-Peterson (GP) opticaldepths of SDSS QSOs significantly increase at z ∼ z to z > z = 6. The ionizing sources at z ≃ i ′ -dropout technique(Stiavelli et al. 2004; cf. Bunker et al. 2004). However,the relation between reionization and ionizing sources,i.e. galaxies, are still not clear in our understandingof cosmic reionization. Since the WMAP5 polarizationdata indicate possible scenarios of an instantaneousreionization at z = 11 ± . z ∼ −
11 (Dunkley et al. 2009), the ionizing photonproduction rate of galaxies at this epoch would constrainthese models of reionization history. It is suggestedthat a substantial fraction of galaxies have completedtheir starburst phase before z ≃
6. Eyles et al. (2007)have shown that about 40% of the IRAC detected z ≃ z > It is Recently, Schaerer & de Barros (2009) argue that Eyles et al.(2007) would overestimate the Balmer breaks by the contributions
Ouchi et al.important to understand when and how the progenitorsof these post-starburst galaxies were formed. To addressthese questions, we need to study z & z ≃ z ≃ µ m(e.g. Bouwens et al. 2008; Oesch et al. 2009a). Al-though Hubble Space Telescope (HST) images can reachas deep as ∼
29 AB magnitude in near-infrared (NIR)bands with the state-of-the-art Wide Field Camera 3(WFC3), the number of z ≃ ≃ −
20, so far (Oesch et al.2009b; McLure et al. 2009b; Bunker et al. 2009; see alsoBouwens et al. 2009b). Moreover, the present studiescover only small areas ( ≃ for the HST/WFC3studies and ≃
100 arcmin even for recent VLT/HAWK-Iobservations; Castellano et al. 2009; Hickey et al. 2009)or small volumes by the gravitational lensing technique(e.g. ∼
100 Mpc ; Stark et al. 2007). Although grav-itational lensing technique can probe very faint high- z galaxies that cannot be found in blank fields with thecurrent facilities (Stark et al. 2007; Bradley et al. 2008;Bouwens et al. 2009a; Zheng et al. 2009), these pencilbeam surveys suffer from large uncertainties due to cos-mic variance, and miss a population of bright galaxiesat z >
7, which plays an important role in determin-ing UV luminosity density for cosmic reionization and inunderstanding forming massive galaxies in their majorstar-formation phase.Motivated by this, we have conducted a wide-field z -dropout galaxy survey by performing deep z ′ and y -bandimaging down to y = 26 . z -dropout galaxies with the rea-sonably deep magnitude limit, which strongly constrainsthe bright-end of UV LF at z = 7. In this paper, wepresent our bright z -dropout galaxy candidates foundby our Subaru observations, and constrain cosmic star-formation history and reionization in conjunction withfaint z -dropout galaxies identified by the recent deepHST/WFC3 studies. We describe our observations and z = 7 z -dropout galaxy samples in § §
3, respec-tively. We show the UV LF and UV luminosity densityof z = 7 galaxies in §
4. We discuss the cosmic star-formation history, cosmic reionization, and the distribu-tion of our z -dropout galaxies in §
5, and summarize ourresults in §
6. Throughout this paper, magnitudes are inthe AB system. We adopt ( h, Ω m , Ω Λ , Ω b h , n s , σ ) =(0 . , . , . , . , . , . DATA
Observations and Data Reduction
We carried out deep y - and z ′ - band imaging sur-vey with Suprime-Cam in the Subaru Deep Field (SDF;Kashikawa et al. 2004) and GOODS-N (Giavalisco et al.2004a) during the last three years from 2006 to 2009.The y band is a custom broad-band filter centered at of nebular emission to their broad band photometry. Fig. 1.—
Response curves of y and z ′ bands, together with modelspectra of a z ∼ .
0, while the black dotted lines denote those showing the total sys-tem throughput. Dark and light gray lines are the normalized to-tal y -band response curves for the flux-weighted SDF and GOODSdata. All of these response curves include the CCD quantum ef-ficiency, transmission+reflection of telescope+instrument optics,and atmospheric absorption. Red line is the model spectrummimicking the one of a typical dropout galaxy (Papovich et al.2001), but redshifted to z = 6 .
73. Green line is a template spec-trum of the typical local elliptical (old) galaxy placed at z = 1 . z = 1 . − z = 1 . z = 1 . z = 1 .
5, which is a reproduction of the typical dusty-starburst SED(Cimatti et al. 2002). Cyan line shows a Galactic T3 dwarf star,one of the reddest Galactic dwarf stars (Knapp et al. 2004). µ m. This filter was originally made for identifying z ≃ z R filter in Shimasaku et al.(2005), we refer to it as y filter for sake of simplicity.Figure 1 shows the total response of the y band fil-ter (central wavelength of 9860˚A and FWHM of 590˚A),which includes the CCD quantum efficiency, transmis-sion+reflection of telescope+instrument optics, and at-mospheric absorption. Each of SDF and GOODS-N fieldis covered by one pointing of Suprime-Cam whose FoV is918 arcmin . Table 1 summarizes our observations andthe y and z ′ data that we used. Before 2007, we took y data with MIT-Lincoln Laboratory (MIT-LL) CCDsinstalled in Suprime-Cam (Miyazaki et al. 2002), while2009 data were acquired with the new fully-depletedHamamatsu CCDs (Kamata et al. 2008). In our anal-ysis, we also use archival y and z ′ images taken in 2001-2006 to make the deepest stacking of y and z ′ data avail-able in SDF and GOODS-N. The total integration timeof our y band images are 26 and 33 hours in SDF andGOODS-N, respectively. The z ′ image of SDF is pro-duced via a number of variable object studies such as su-arge Area Survey for z = 7 Galaxies 3pernovae (Poznanski et al. 2007), active galactic nuclei(Morokuma T. et al. in preparation), and high propermotion stars (Richmond et al. 2009).Our data were reduced using Suprime-cam DeepField REDuction package (SDFRED; Yagi et al. 2002;Ouchi et al. 2004a). We have found that the totalthroughput in the y band is improved by nearly a fac-tor of 2 in 2009 with the new Hamamatsu CCDs, butthat the shapes of y -band total response curves are al-most identical between the data of MIT-LL (taken before2007) and Hamamatsu CCDs (taken in 2009; Figure 1).Thus, we combine these y -band data taken with MIT-LLand Hamamatsu CCDs. Figure 1 displays the normalizedtotal y -band response curves of the Hamamatsu CCDsand the flux-weighted MIT-LL+Hamamatsu CCDs forthe SDF and GOODS data. The differences between allthe response curves are negligibly small. Since we finda very small difference of . .
02 mag between the re-sponse curves of flux-weighted SDF and GOODS-N evenwith the model spectra of extremely red objects such asL/T-dwarf stars and z -dropout galaxies ( § y -band response curve of the flux-weighted SDF inour analysis. The 3 σ sky noise of the reduced imagesare ( y, z ′ ) = (26 . , .
7) and (26 . , .
9) magnitudes in a1 ′′ . inSDF and GOODS-N, respectively. The total area is 1568arcmin . The positions of y and z ′ are aligned based onhundreds of stellar objects commonly detected in bothimages. After the PSF sizes of these images are matched,FWHM sizes of PSFs are estimated to be ≃ ′′ .
91 and ≃ ′′ .
87 in SDF and GOODS-N, respectively.
Matched Images
Except for the y and z ′ bands, we use the legacyground-based deep optical imaging data for SDF(Kashikawa et al. 2004) and GOODS-N (Capak et al.2004), and the deep HST/ACS v2.0 and Spitzer/IRACv0.3 available for the central ≃
160 arcmin fieldof GOODS-N (Giavalisco et al. 2004a; Dickinson et al.2003). These data are registered with stellar objectsto match the coordinates of our Suprime-Cam y and z ′ images in SDF and GOODS-N. Table 2 summarizes allof imaging data used in our study. Again, the PSF sizesof our SDF y and z ′ images are matched to that of thepublic SDF images with a PSF FWHM of 0 ′′ .
99. Notethat the public ground-based GOODS-N data include U , V , and R -band images with a relatively poor seeing( F W HM ≃ ′′ . y and z ′ images. Because our pur-pose of the U BV R photometry is not to measure a colordefined by the same PSF+aperture but to confirm non-detections, we place the upper limits of detection with alarge, 3 ′′ -diameter, aperture size for those U , V , and R images.During the observations, we took images of spectropho- Note that HST/ACS v2.0 data are significantly deeper thanthe previously released images.
Fig. 2.—
Colors of z ′ − y for various objects as a functionof redshift. Red and blue lines represent the models of dropoutand dusty-starburst galaxies that are reproductions of the typicalSEDs shown in Papovich et al. (2001) and Cimatti et al. (2002),respectively. Green lines indicate elliptical, Sbc, Scd, and irregulargalaxies (Coleman, Wu, & Weedman 1980). Purple and cyan starmarks are Galactic stars (Gunn & Stryker 1983) and L1-L9/T0-T9 dwarf stars (Knapp et al. 2004). Black solid line indicates ourcolor selection criterion, z ′ − y > .
5. The black dotted line marksa redshift, z = 6 .
5, that is roughly a lower limit of our selection. tometric standard star of G191-B2B with y band filterin GOODS-N (Oke 1990; Bohlin et al. 1995). The stan-dard star was observed 4 times under photometric con-dition. We calculate photometric zero-points from thestandard star data. The photometric zero points of theother images, i.e. GOODS-N z ′ , SDF y , and SDF z ′ , aredetermined by matching the zero points with those ofimages taken by Capak et al. (2004), Shimasaku et al.(2005), and Kashikawa et al. (2004), respectively. Wecheck these photometric zero points based on colors ofstellar objects in our field and 175 Galactic stars calcu-lated from spectra given in Gunn & Stryker (1983). Wefind that the colors of stellar objects in our data areconsistent with those of Gunn & Stryker’s (1983) starswithin ≃ .
03 magnitude. CATALOGS AND SAMPLES
Photometric Catalogs
Source detection and photometry are performed us-ing SExtractor (Bertin & Arnouts 1996). The y im-ages are chosen for our source detection. We mea-sure 1 ′′ . z ′ − y color with these aperture magnitudes obtainedby the dual image mode of SExtractor. We cor-rect the magnitudes of objects for Galactic extinctionof E ( B − V ) = 0 .
018 (SDF) and 0 .
012 (GOODS-N;Schlegel, Finkbeiner, & Davis 1998). A total of 63,740and 55,559 objects are identified down to the 4 σ detec-tion limits in SDF ( y = 26 .
1) and GOODS-N ( y = 25 . Photometric Samples
We isolate z -dropout galaxy candidates at z ∼ z ∼ z galaxies and one of the red-dest T dwarf stars (Knapp et al. 2004) which are likelyto be prominent interlopers in our photometric sample.Figure 1 shows that the spectral feature of a significant1216˚A trough for z ∼ z ′ - and y -band wavelengths.In Figure 2, we present predicted z ′ − y colors as afunction of redshift for these model spectra. This fig-ure demonstrates that no objects except z ∼ z ′ − y & .
5. Since the wavelength coverages of z ′ - and y -bands are very close (Figure 1), the Ly α trough of z ∼ z & . z ′ − y ) & . − . z ′ − y color cutcan isolate z ∼ z ∼ α trough, e.g. z ′ − Y & z ′ − J &
1, with a broad band of Y / J whosecentral wavelength is redder than that of our y band(Bouwens et al. 2008; Oesch et al. 2009b; Bunker et al.2009; Castellano et al. 2009; Hickey et al. 2009). Thecolors of z ′ − Y ∼ z ′ − J ∼ z ′ − Y or z ′ − J color alone. Instead, thesestudies can distinguish foreground red galaxies with theiravailable deep NIR ( J , H , and/or K ) photometry (e.g.Figure 3 of Bunker et al. 2009). The idea of our candi-date selection is to discriminate foreground red galaxieswithout NIR photometry but with a color cut strongerthan the color criterion of the other studies. In addi-tion to a Ly α trough, a rest-frame far UV continuum be-low Lyman break (912˚A) is damped by IGM absorptionwith a large Lyman continuum opacity (Inoue & Iwata2008). Because no such continuum should be identifiedat a wavelength shorter than 7000( ≃ × [1 + 6 . z & .
7, non-detection criteria should begiven in U , B , V , and R bands whose bandpasses arebluer than 7000˚A. From the model colors and the lack ofa far UV continuum, we define the selection criteria of z ∼ z ′ − y > . U > U σ & B > B σ & V > V σ & R > R σ (1) where U σ , B σ , V σ , and R σ are the 2 σ limitingmagnitudes of U , B , V , and R images, respectively.The U -band criteria is only applied to the objects inGOODS-N, since there are no public U -band data inSDF. The 2 σ limiting magnitudes are ( B σ , V σ , R σ ) =(29 . , . , .
8) in SDF and ( U σ , B σ , V σ , R σ ) =(27 . , . , . , .
0) in GOODS-N. We select z ∼ σ limits of y = 26 . y = 25 . z ′ − y > . y -band detection limits, be- cause the 2 σ upper limits of z ′ images reach 28 . . ∼
160 arcmin field ofGOODS-N, we do not use these HST images at this stageto avoid making a heterogeneous sample given by the dif-ferent detection limits on the sky of GOODS-N. Instead,we take advantage of the deep HST images for confirm-ing non-detections of blue continuum for the candidatesfalling in the area with the HST images (see below).After we reject spurious sources near the spikes ofbright sources, the residuals of sky subtraction etc. byvisual inspection, we obtain z -dropout galaxy samplesconsisting of 15 and 7 candidates in SDF and GOODS-N, respectively. All of these candidates have magnitudesfainter than y = 25 . z ′ - as well as y -bands. Because the bandpassof z ′ band includes both red and blue sides of GP trough(1216˚A), the detections of faint z ′ counterparts are rea-sonable. The z ′ -band detections rather confirm that thecandidates are neither spurious sources nor transients ap-pearing in the y images. Although we do not apply acriterion of non-detection in i ′ / I band where a UV con-tinuum between Lyman break (912˚A) and GP trough(1216˚A) falls, none of our candidates have an i ′ / I -bandcounterpart with a flux beyond our detection limits.We check the spectroscopic catalogs of SDF andGOODS-N, which are obtained by Kashikawa et al.(2003), Shimasaku et al. (2003), Ouchi et al. (2004a),Shimasaku et al. (2006), Kashikawa et al. (2006),Yoshida et al. (2006), Iye et al. (2006), Nagao et al.(2007), Hayashi et al. (2009), and Ly et al. (2009) forSDF, and Wirth et al. (2004), Reddy et al. (2006),Barger et al. (2008), Cohen et al. (2000); Cohen (2001),Steidel et al. (1996, 1999, 2003), Phillips et al. (1997),Lowenthal et al. (1997), and Dawson et al. (2001)for GOODS-N. We find that one of our z -dropoutcandidates, SDF-63544, has a spectroscopic redshift of z = 6 .
96, which was originally identified by Iye et al.(2006) in their Ly α emitter (LAE) study. SDF-63544is the first dropout galaxy at z ≃ y -band magnitudeof y = 25 .
42, which is the brightest candidate in our z -dropout galaxy samples. It should be noted thatour photometric sample surely includes a real z ≃ z objects with a spectroscopicredshift. This confirms that our photometric criteria donot select obvious foreground objects.Two out of seven candidates in the GOODS-N field,GOODSN-152505 and GOODSN-108036, fall in the re-gion with the deep GOODS-N HST/ACS and Spitzer im-ages. We display snapshot images in Figure 4. We havefound that neither candidate is detected in HST/ACS B , V , and i bands . The Spitzer images ofGOODSN-152505 and GOODSN-108036 are confusedby the nearby bright objects due to large PSF sizes in GOODSN-152505 is located near the edge of HST GOODS-Nfield, and not covered by the B image. arge Area Survey for z = 7 Galaxies 5 Fig. 3.—
Snapshots of our z -dropout candidates identified inSDF (left) and GOODS-N (right). Each object has images of ( U ), B , V , R , i ′ , z ′ , and y bands with an ID number on the right. Thesize of images is 10 ′′ × ′′ . North is up and east is to the left. Fig. 4.—
Snapshots of two z -dropout candidates falling in thearea with the HST and Spitzer images in GOODS-N. From left toright, we display HST/ACS B , V , i , z , Suprime-Cam y , Spitzer/IRAC 3 . µ m, 4 . µ m, 5 . µ m, 8 . µ m, and Spitzer/MIPS24 µ m images. The Suprime-Cam y image is shown for comparison.GOODSN-152505 is not covered with the B image. The size ofall images is 10 ′′ × ′′ , but that of 24 µ m band is 11 ′′ × ′′ . Northis up and east is to the left. IRAC and MIPS data. However, there are some signa-tures of possible counterparts for GOODSN-152505 inthe 4 . µ m band and for GOODSN-108036 in the 3 . µ mand 4 . µ m bands, which would be different from the ef-fects of source confusion. Although photometry of thesepossible Spitzer counterparts are more or less contami-nated by fluxes of the nearby objects, we simply calculatetotal magnitudes from 3 ′′ -diameter aperture magnitudesand aperture corrections for IRAC and MIPS fluxes givenin Yan et al. (2005) and the Spitzer web page , respec-tively. Since these sources are confused by the nearbyobjects, we find that total magnitudes ( m . , m . ) ofGOODSN-152505 and GOODSN-108036 are fainter than(23 .
4, 24 .
2) and (24 . . y − m . and y − m . are . − .
1. BecauseEyles et al. (2007) report that their z ∼ − z ′ − m . and z ′ − m . (as well as J − m . J − m . ),the colors of our two candidates are comparable to thoseof z ∼ z ∼ −
3, suchreported by Yan et al. (2004), whose optical to IRAC col-ors ( z − m . ) exceed 3.3. Although the possible IRACcounterparts of our z -dropout galaxy candidates wouldindicate that there exist post-starburst galaxies even at z ∼
7, there remains the possibility that these two candi-dates are foreground interlopers. We will discuss stellarpopulation of these two candidates via detailed spectralenergy distribution model fitting after we confirm theredshifts of these candidates by spectroscopy.There are no obvious counterparts of these two candi-dates at longer wavelengths; IRAC 5 . µ m, 8 . µ m, andMIPS 24 µ m bands. The MIPS snapshot of GOODSN-152505 shows a source at the left side, but this MIPSsource is a counterpart of a bright source located nearthe left corner. Because the detection limits of the threebands are too shallow (21 −
23 mag; Table 2) to iden-tify galaxies at large distances, the non detections, again,confirm that these two candidates are neither extremelyred galaxies nor AGN at low redshifts. LUMINOSITY FUNCTION
Surface Number Densities and DetectionCompleteness
We obtain the number counts of all y -band detectedobjects, N all ( m ), and our z -dropout galaxy candidates, N cand ( m ), from our photometric catalogs. We calculate http://ssc.spitzer.caltech.edu/mips/apercorr/ Ouchi et al.
Fig. 5.—
Top panel: Detection completeness of our y -bandimages in percentage. Black and gray solid lines represent thecompleteness for 1 ′′ . y data. Lower and upper sequences ofpoints show surface densities of our z -dropout galaxy candidatesand all objects detected in the y images, respectively. The surfacedensities are shown with the squares (SDF) and circle (GOODS-N). In the upper sequence, filled black symbols are those with thecompleteness correction, while open symbols are not applied forthe correction. In the lower sequence, filled gray symbols are sur-face densities that are subtracted with the numbers of contami-nation, and filled black symbols denote the best-estimate of our z -dropout galaxy surface densities with both the contaminationand completeness corrections (see § z -dropout galaxies is carried out with a 1 ′′ . z -dropout galax-ies in the luminosity function plot of Figure 7. For the presen-tation purpose, we slightly shift the open and gray points alongthe abscissa. The exact magnitudes are the same as magnitudesof black filled points. The vertical axis on the right side indicatesthe number counts of objects, i.e. N/ (0 . / (0 . ), whichapproximately correspond to numbers of z -dropout galaxies iden-tified in each target field. Dotted line represents the best-fit powerlaw to the completeness-corrected surface densities of all objects. the surface number densities, Σ obs , by dividing N all ( m )and N cand ( m ) by our respective survey areas. The re-sults are presented in the bottom panel of Figure 5. Sincethe surface number densities of faint objects are affectedby detection incompleteness, we estimate detection com-pleteness as a function of y magnitude by Monte Carlosimulations. We distribute 7240 artificial objects with apoint spread function on our y -band images after addingphoton Poisson noise, and detect them in the same man-ner as for the detection for our photometric catalogs withSExtractor. We repeat this process 20 times, and com-pute the ratio of recovered objects to the input objects.The top panel of Figure 5 shows the detection complete-ness of our y -band images. We find that the detection completeness is typically &
70% for relatively luminoussources with y . .
5. The detection completeness is >
50% even in the faintest magnitude bins centered at y = 25 .
85 (SDF) and y = 25 .
65 (GOODS-N). We correctthe surface number densities for the detection complete-ness,and present them in the bottom panel of Figure 5.
Contamination
There are four sources of contamination in our z ∼ y -band sources made ofnoise fluctuations, 2) transients, such as faint variablestars+AGN and supernovae detected in our y -band im-ages, 3) foreground red objects entering our samplesdue to photometric errors, 4) L/T dwarf stars satisfy-ing our color selection criteria. We define the numbersof contamination for 1), 2), 3), and 4) in our samplesas N ( m ), N ( m ), N ( m ), and N ( m ), respec-tively. Below, we check the effects of contamination, andestimate their impacts on our z -dropout galaxy samples.1) Spurious sources:Because we push our y -band detection limits, our z ∼ y -band spurious sourcescould pass our selection criteria of eq. (1). To estimatehow much spurious sources are included in our z -dropoutgalaxy samples, we carry out source detection and colorselection same as those in §
3, but with images whoseADU counts are multiplied by −
1. We run SExtractorwith these negative-count images, and make negative- y band detection catalogs. We apply the color criteriaof eq. (1), and reject sources apparently made by theresiduals of sky subtraction in the same manner as forthe real z -dropout galaxy selection. We find 0 and 1spurious z ∼ N ( m ) = 0 (SDF) and 1 . ± . y images are the stacked data that wereacquired in a 6-year (3-year) long period from 2003(2006) to 2009 for SDF (and GOODS-N; see Table 1), weinvestigate the possibility of transients for our z -dropoutcandidates. We stack y data taken before and after 2008,and obtain y -band images for the two epochs. The y -band detection limits of the former ( y epoch1lim ) and the lat-ter ( y epoch2lim ) images are ( y epoch1lim , y epoch2lim ) = (25 . , . . , .
6) in GOODS-N. We have carriedout photometry at the positions of our z -dropout galaxycandidates, and found no candidates detected at the 3 σ levels that show a significant magnitude change betweenthese two epochs by & . − N ( m ) = 0 inboth SDF and GOODS-N. Checking this result, we calcu-late an expected number of transients based on the deepand wide-field transient study results of Morokuma et al.(2008). Morokuma et al. (2008) present that the numberdensity of transients (with the timescale greater than200 days) in the magnitude range of i ′ ≃ . − . ≃ . Assuming the difference of magnitudes be-tween y and i ′ is negligible for transients which are mostlynearby objects, the expected number of transients in ourSDF and GOODS-N samples is only ∼ . z = 7 Galaxies 7tection completeness correction. This estimate is consis-tent with our conclusion of no transients in our z -dropoutgalaxy samples.3) Foreground red objects entering our samples due tophotometric errors:It is possible that some foreground objects, such as redgalaxies at intermediate redshifts, enter our color crite-ria by photometric errors, although their intrinsic colorscannot satisfy the criterion of z − y > .
5. We make aninput mock catalog mimicking foreground objects, andcarry out Monte Carlo simulations with the mock catalogto estimate the numbers of foreground interlopers. Themock catalog has the same number-density distributionas that of all y -detected objects corrected for the detec-tion completeness (Filled circles at the upper sequencein the bottom panel of Figure 5). In the mock catalog,colors of the bright ( y < .
5) objects are the same asthose of the observed y -detected objects. Because faintobjects have moderately large photometric errors, we donot use colors of objects with a magnitude fainter than y = 24 .
5. Instead, we assign color distribution of ob-served y -detected objects with y = 23 . − . y > . § N ( m ) = 0 .
85 in SDF down to y = 26 .
1, and 0 .
80 inGOODS-N down to y = 25 . z − y > . z -dropout galaxy samples. Estimating the num-bers of late-type stars which contaminate our z -dropoutsamples, we carry out Monte-Carlo simulations same as3), but with an input mock catalog of late-type stars.We use the number density of L/T dwarfs as a func-tion of Galactic latitude presented in Ryan et al. (2005)who derive the number densities in 15 deep HST/ACSfields down to z = 26 . N ( m ) = 5 . y = 26 . .
80 down to y = 25 . . − . Redshift Distribution
We have estimated redshift distribution of our z -dropout galaxies, C ( m, z ), by Monte-Carlo simulationswith an input mock catalog of high- z galaxies. Themock catalog consists of high- z galaxies whose propertiesare given with the probability distributions of i) numbercount, ii) continuum color, and iii) Ly α emissivity. First,for the probability distribution of i), we use the surfacenumber densities of our z -dropout galaxies corrected forcontamination and completeness (black filled points inFigure 5). Second, we assume that high- z galaxies havethe average UV continuum slope of β = − Fig. 6.—
Redshift distribution of our z -dropout galaxies inSDF (top) and GOODS-N (bottom). Solid lines plot the redshiftdistributions averaged over magnitudes weighted with the numberdensity distributions of our z -dropout galaxies. Gray shades rep-resent errors of the redshift distribution estimates obtained by ourMonte-Carlo simulations. Arrow indicates the redshift of our z -dropout galaxy with the spectroscopic confirmation (SDF-63544; z spec = 6 . found in z ∼ β is a Gaussian func-tion with a standard deviation of σ β = 0 .
5. We gener-ate spectra with the stellar population synthesis modelof Bruzual & Charlot (2003), and obtain galaxy spectrawith β = ( − . − ( − . z = 3 (Papovich et al.2001), but with a young age of 4 Myr and Calzetti dustextinction ranging from E ( B − V ) = 0 .
008 to 0 . β = − .
0) to the moderately red ( β = − . α emitting galaxies, we assume that 30% of z = 7dropout galaxies have a Ly α emission line with a rest-frame equivalent width ( EW ) of > z = 6 dropout galaxies downto ∼ L ∗ (Stanway et al. 2004a,b; Vanzella et al. 2006a;Dow-Hygelund et al. 2007; Stanway et al. 2007). We adda Ly α luminosity to these 30% of dropout galaxies withthe EW probability distribution of LAEs at z = 5 . z = 3 LBGs havea Ly α EW of ≤ z = 7 dropout galaxies have no Ly α emission line. Forthe rest of 20% of high- z galaxies, we add a very weak EW = 0 −
20 Ly α line to their spectra.Finally, we produce an input mock catalog of high- z galaxies that are distributed in the radshift space ho-mogeneously, and apply statistical weights following theprobability distributions of i), ii), and iii) shown above. Ouchi et al.This mock catalog is used to carry out Monte-Carlo sim-ulations in the same manner as those in § y = 27, andproduce the mock catalog including faint sources belowour detection limit. We obtain y -band detection catalogsby the Monte-Carlo simulations, and select artificial z -dropout galaxies with the color criteria (eq. 1) to drawmock samples of z -dropout galaxies. We calculate theratio of the selected objects to the input objects as afunction of redshift, which corresponds to the redshiftdistribution of our z -dropout galaxies. Figure 6 plotsthe redshift distribution of our z -dropout galaxies, C ( z ),averaged over magnitudes with the probability distribu-tions of i). For both data of SDF and GOODS-N, thepeak redshift of C ( z ) is z = 6 .
9, and 90 percent of the z -dropout galaxies fall in z = 6 . − .
1. Thus, the redshiftwindow of our z -dropout selection is z = 6 . +0 . − . . UV Luminosity Function
We derive the UV luminosity function of z -dropoutgalaxies based on the numbers of our candidate galax-ies ( § § § n ( m ),of z -dropout galaxies at each field in a given magnitudebin by n ( m ) = (cid:2) N cand ( m ) − P ni =1 N i cont ( m ) (cid:3)R ∞ dVdz C ( m, z ) dz , (2)where n represents the four kinds of contaminants sat-isfying our selection criteria in a given magnitude bin( n = 4; see § dVdz is the differential cosmic vol-ume with an area of SDF or GOODS-N. Because we ap-ply a 1 ′′ . z -dropout galaxy pho-tometry to maximize the signal-to-noise ratio, we needto brighten the y magnitudes with an aperture correc-tion to estimate total fluxes. On the other hand, our y -band magnitudes can be contaminated with Ly α emis-sion lines. We should subtract the contributions of Ly α fluxes to obtain UV-continuum magnitudes, and dim the y -band brightness accordingly. Moreover, we should ap-ply k-correction to get UV-continuum luminosities at therest-frame ∼ − α forest below GP trough,which enter the bandpass of our y filter. We estimatea correction factor to derive total UV-continuum mag-nitudes at ∼ y -band magnitudes basedon the results of our Monte-Carlo simulations in § y -band magnitudes averaged with the statisti-cal weights from the probability distribution functions( § .
01 and+0 .
05 for SDF and GOODS-N samples, respectively. Weapply these small corrections to our UV-continuum mag-nitude estimatesFigure 7 presents UV LF of z -dropout galaxies fromour samples with the red filled squares (SDF) and cir-cle (GOODS) as well as our upper limits with the red Fig. 7.—
UV luminosity function (LF) of z -dropout galax-ies, together with those at lower redshifts. Two red filled squaresand one red filled circle present UV LF of our z -dropout galaxiesin SDF and GOODS-N, respectively. Two red arrows with opencircles indicate upper limits of z -dropout galaxies estimated fromour GOODS-N data (right) and the combination of our SDF andGOODS-N data (left). Magenta arrow displays the lower limitestimated from our spectroscopically-identified z -dropout galaxy.Magenta inverse-triangles represent the maximal LF, i.e., the se-cure upper limits of our LF estimates that include no correc-tion for contamination. Recent measurements and upper limitsincluding those from HST/WFC3 studies are also plotted withred star marks (Bouwens et al. 2008), hexagons (McLure et al.2009b), black filled pentagons (Oesch et al. 2009b), open pentagons(Oesch et al. 2009a), open diamonds (Mannucci et al. 2007), andcrosses (Richard et al. 2006). Although previous studies define M UV at the rest-frame ≃ − ≃ ∼ .
07 (Oesch et al. 2009a).Red solid line plots our best-fit Schechter function of z -dropoutgalaxies at z = 7. Dotted line is UV LF at z ∼ z ∼ z = 7.At around the top of this plot, we also tick the corresponding star-formation rates estimated from eq. (3) with no dust extinctioncorrection. open circles. We also plot z -dropout galaxy UV LFs de-rived by the deep HST NICMOS+WFC3 (Bouwens et al.2008; Oesch et al. 2009a,b; McLure et al. 2009b) andthe ground-based (Richard et al. 2006; Mannucci et al.2007) studies, together with UV LFs at low redshifts. Because we have a spectroscopically identified galaxyat z = 6 .
96, we can also place a lower limit on the UVLF. We estimate a 1500˚A-continuum magnitude of thisgalaxy to be M UV = − . ± .
31 from the y-band pho-tometry ( y = 25 . α flux (2 × − erg s − cm − ;Iye et al. 2006; Ota et al. 2008), and the mean modelspectrum at z = 6 .
96 with β = − . § z = 7 UV LF at the bright magnitude of We cannot include the recent results of Bunker et al. (2009),Castellano et al. (2009), and Hickey et al. (2009), because their UVLF measurements are not apparently presented. arge Area Survey for z = 7 Galaxies 9 − < M UV < −
21 as well as the lower limit basedon the spectroscopically-identified z -dropout galaxy. Wefind that our UV LF at z = 7 falls significantly belowthat at z = 6. Because our z -dropout galaxy sam-ples are largely corrected for contamination estimatedby the simulations ( § z -dropout galaxies could be due to over es-timates of the contaminants. We derive UV LF of our z -dropout galaxies in the same manner as above, butwith no contamination subtraction, and refer to theseestimates as the maximal LF which provides conserva-tive upper limits. We show the maximal LF with themagenta inverse-triangles in Figure 7. These magentainverse-triangles also fall below the z = 6 UV LF mea-surements. Note that the z = 6 UV LF measurementswould also have the similar problems in contaminationestimates, but our magenta inverse-triangles come belowthe gray shade area in Figure 7 that represents a vari-ance of z = 6 LFs derived by different studies with var-ious contamination estimates. Thus, we conclude withour LF measurements and maximal LF estimates that z = 7 UV LF definitely decreases from z = 6 at thebright end. Our conclusion is consistent with the claim ofMannucci et al. (2007); Castellano et al. (2009). More-over, this decreasing tendency is similar to that found atthe faint magnitudes by the HST studies ( M UV > − z = 6 to 7. Wefit a Schechter function to the LF measurements from ourand the HST studies, and obtain the best-fit parametersof φ ∗ = 0 . +2 . − . × − Mpc − , M ∗ UV = − . ± . α = − . ± .
65 that maximize the likelihood, L = Π i p [ N obs ( m i ) , N exp ( m i ; φ ∗ , M ∗ UV , α )], where p [ x, µ ]is the Gaussian distribution with a mean µ evaluated at x , and N obs and N exp are, respectively, the numbers ofgalaxies within a magnitude bin of m i from observationsand expectations for a given set of Schechter parame-ters. We summarize the best-fit Schechter parametersin Table 4, and plot the best-fit function in Figure 7 withthe red line. Although the constraints on α are weak, asteep slope of α is suggestive, which is similar to that at z . z -dropout galaxies. First, the num-bers (+Poisson errors) of z -dropout galaxies in samplesof (SDF, GOODS-N) are (2 ± .
4, 1 ± .
0) at y < . ± .
0, 3 ± .
7) at y < .
7, and (8 ± .
8, 7 ± . y < . . Since the survey areas of SDF andGOODS-N are comparable ( ≃ . ; § To avoid using the dependent measurements from the sameHST data, we exclude the results of Oesch et al. (2009a,b) in thefitting. The inclusion of Oesch et al. (2009a,b) data does notchange our conclusions but with artificially small errors. Seven (= 15 −
8) candidates of SDF fall in a narrow magnitudewindow of y = 25 . − .
1. Because the test in § − z -dropout galaxiesat this magnitude regime. Fig. 8.—
Error ellipses of Schechter parameters, M ∗ and φ ∗ ,at the 1 and 2 σ confidence levels. Red lines represent our resultsof z -dropout galaxies at z = 7. Blue, cyan, and green contoursdenote error ellipses for galaxies at z = 4, 5, and 6 obtained byBouwens et al. (2008) (large contours) and McLure et al. (2009a)(small contours only for z = 5 and 6). The dotted contours areerror ellipses of z = 7 galaxies estimated by Bouwens et al. (2008).All of these Schechter fits are based on α ≃ − .
7. Our measure-ments of the red contours indicate that the Schechter parametersof z = 7 LF differ from those of z ≤ > σ (i.e. > SDF and GOODS-N in Figure 7, and find that the cos-mic variance is smaller than a factor of ∼ . . We estimatethe expected cosmic variance with the analytic CDMmodel of Sheth & Tormen (1999) from the number den-sity of our z -dropout galaxies down to our magnitudelimit (0 . +2 . − . × − Mpc − at M UV < b = 7 . − . at z = 6 . − .
1. We check our calculations with theCosmic Variance Calculator (Trenti & Stiavelli 2008),and find that this calculator returns a very comparablenumber of ≃
30% for the cosmic variance after subtrac-tion of Poisson error term. We obtain even a smallercosmic variance of ≃ −
12% with the CDM model ofSheth & Tormen (1999), if we do not assume the one-to-one correspondence, but adopt clustering bias of galaxiesmeasured at a slightly lower redshift of z ∼ b ≃ − .
2, in the twoareas of 0.2 deg .Figure 8 shows the error ellipses of our Schechter pa-rameters, M ∗ and φ ∗ , for α = − .
72 at the 1 and 2 σ con-fidence levels. We also present those of LFs at z = 6 and5 (Bouwens et al. 2008; McLure et al. 2009a) and z = 4(Bouwens et al. 2008). Note that all of these Schechterfits are based on α ≃ − . α = [ − . − [ − . http://solo.colorado.edu/ ∼ trenti/CosmicVariance.html α = [ − . − [ − .
71] forMcLure et al. 2009a), and that Figure 8 exclusively com-pares two parameters of M ∗ and φ ∗ . Our measurements(red contours in Figure 8) indicate that the Schechter pa-rameters of z = 7 LF differ from those of z ≤ z = 6 to 7 at morethan the 2 σ (i.e. 95%) level. Moreover, our constraintsof z = 7 Schechter parameters are consistent with thoseof Bouwens et al. (2008) (dotted lines in Figure 8), butare stronger than those, which allow us to rule out noevolution at the >
95% level. Although the errors of ourmeasurements are too large to distinguish between lumi-nosity ( L ∗ ) and number ( φ ∗ ) evolutions, Figure 8 impliesthat a decrease in L ∗ would be the dominant factor ofthe LF evolution from z = 5 − UV Luminosity Density
We calculate UV-luminosity densities at z = 7 fromour UV LF with the best-fit Schechter parameters givenin § z -dropout galaxies of theHST/WFC3 studies, which was used for our Schechterparameter fitting (i.e., down to M UV = −
18 or ≃ . L ∗ ),and obtain the observed UV-luminosity density, ρ obsUV .Because a total UV-luminosity density has to be largerthan ρ obsUV by the amount of the contribution from galax-ies fainter than the limiting magnitude, ρ obsUV correspondsto the lower limit of the UV-luminosity density. We ex-trapolate our LF down to L = 0 to estimate the con-tribution from such faint galaxies, and obtain the UV-luminosity density, ρ upperUV . Since there exist no galaxieswith L ∼
0, this UV-luminosity density with the LF ex-trapolation, ρ upperUV , corresponds to the upper limit of UVluminosity density. We estimate ρ obsUV = 4 . +6 . − . × erg s − Hz − Mpc − and ρ upperUV = 1 . +1 . − . × erg s − Hz − Mpc − . Both ρ obsUV and ρ upperUV are summarized inTable 4. We discuss evolution of cosmic star-formationrate density ( § § α = − .
72, which we have weakly constrained.If we assume the best-fit Schechter parameters but witha steep slope of α = − . ρ upperUV would increase onlyby a factor of 2, which just corresponds to the 1 σ -uppererror value (2 . × erg s − Hz − Mpc − ) of our ρ upperUV .On the other hand, a very steep slope of α = − .
90 couldpush up ρ upperUV by a factor of 3, and an extreme value of α = − .
97 may boost ρ upperUV by a factor of 10. However,Bouwens et al. (2008) and McLure et al. (2009a) have re-ported α measurements similar to ours but with a smalluncertainty, α = − . ± .
16 and α = − . ± . z = 6. Since it is unlikely that the faint-end slope evolveslargely between z = 6 to 7, ρ upperUV would not be well be-yond its 1 σ -upper error value in the reasonable range of α . DISCUSSION
Cosmic Star-Formation History
We calculate cosmic star-formation rate densities(SFRDs) from the UV-luminosity densities, ρ obsUV and ρ upperUV . We use the relation between UV lu-minosity and star-formation rate (SFR) given by Fig. 9.—
Cosmic star-formation rate density (SFRD) as afunction of redshift. Red square and inverse-triangle representthe extinction corrected SFRDs integrated down to L ≃ . L ∗ (SFRD obscorr ), and to L = 0 (SFRD uppercorr ), respectively. Open squareand inverse-triangle are the same, but with no extinction correc-tion, i.e., SFRD obs and SFRD upper . Magenta line with trianglesgive the allowed SFRDs at z = 7 which are defined by SFRD obs and SFRD uppercorr with associated errors. We shift the magenta linewith the triangles along the abscissa for the presentation purpose.Filled circles indicate total SFRDs at z . z = 0 − . L ∗ ( z = 3) with extinction cor-rection, which are recently reported by Bouwens et al. (2009b).Because their z = 7 measurement is very close to our results, theirtriangle symbol at z = 7 is hidden behind the red filled square. Madau, Pozzetti, & Dickinson (1998):SFR( M ⊙ yr − ) = L UV (erg s − Hz − ) / (8 × ) , (3)where L UV is UV luminosity measured at 1500˚A. This re-lation assumes that galaxies have the Salpeter IMF withsolar metallicity. We obtain SFRD obs = 4 . +7 . − . × − M ⊙ yr − Mpc − and SFRD upper = 1 . +1 . − . × − M ⊙ yr − Mpc − from ρ obsUV and ρ upperUV , respectively.We apply extinction correction to the SFRDs, assum-ing the empirical relation between the UV slope, β , andextinction, A , for starburst galaxies, A = 4 .
43 + 1 . β (4)(Meurer, Heckman, & Calzetti 1999). Following theobservational results of z ∼ β = − z -dropoutgalaxies. We estimate the extinction-corrected SFRDsto be SFRD obscorr = 7 . +11 . − . × − M ⊙ yr − Mpc − andSFRD uppercorr = 2 . +2 . − . × − M ⊙ yr − Mpc − . SinceSFRD uppercorr is the SFRD with dust correction and extrap-olation of the LF down to L = 0, SFRD uppercorr is an upperlimit of our SFRD measurements. On the other hand,SFRD obs is the SFRD estimated with neither dust ex-tinction correction nor LF extrapolation. Thus, SFRD obs is regarded as a conservative lower limit.Figure 9 plots the cosmic SFRDs from our measure-ments (squares and inverse-triangles) as well as our up-per and lower limits including the 1 σ errors (magentaarge Area Survey for z = 7 Galaxies 11line). Figure 9 also displays the cosmic SFRDs ob-tained from previous studies with the assumption ofSalpeter IMF. At z ≃ −
6, we show the cosmicSFRD measurements compiled by Hopkins & Beacom(2006). The compilation of Hopkins & Beacom (2006)covers most of SFRD measurements made, to date,in various wavelength including H α (Glazebrook et al.1999; Tresse et al. 2002; Hanish et al. 2006), mid-infrared (Flores et al. 1999; P´erez-Gonz´alez et al. 2005),submm (Barger et al. 2000; Hughes et al. 1998), ra-dio (Condon et al. 2002; Serjeant et al. 2002), and X-ray (Georgakakis et al. 2003). It also includes resultsof Giavalisco et al. (2004b), Bunker et al. (2004), andOuchi et al. (2004a) for z > z . . L ∗ ( z = 3). Their measure-ments are interpreted as lower limits of total SFRDs thatare counterparts of our SFRD obscorr measurement. Ourmeasurement is consistent with that of Bouwens et al.(2008). In Figure 9, comparing the Hopkins & Beacom’s(2006) best-fit model function (dotted line) with our con-straints of upper and lower limits (the magenta line withthe error bar), we find that the cosmic SFRD drops fromthe peak at z = 2 − z = 7 roughly by a factor of ∼
10 (at least by a factor of & ∼ z = 6 to 7, whichis originated from the decrease of UV LF from z = 6 to7. Note that this decline of SFRD could be weaker, ifthere exist a large population of very faint galaxies, suchsuggested by Stark et al. (2007) for z = 9 −
10 galax-ies, that the present blank field surveys cannot iden-tify. However, the decreasing tendency of SFRD from z = 2 − α & − .
97, that changes ρ upperUV , i.e.,SFRD uppercorr by a factor of .
10 (see § z = 7. Ionization Photon Budget Near the ReionizationEpoch
We evaluate emission rate of hydrogen ionizing pho-ton per comoving Mpc , ˙ N ion , and discuss ionizing pho-ton budget, i.e., whether the photon production rate ofgalaxies is larger than the recombination rate of hydro-gen IGM. We calculate ˙ N ion for galaxies with˙ N ion (s − Mpc − ) = 10 . (cid:18) ǫ g (cid:19) (cid:16) α s (cid:17) − (cid:18) f esc . (cid:19) , (5)where ǫ g is the ionizing emission density at the Ly-man limit in units of erg s − Hz − Mpc − , α s is thespectral index of ionizing emission, and f esc is the es-cape fraction of ionizing photons (Bolton & Haehnelt2007). We adopt ǫ g = ρ UV / Fig. 10.—
Emission rate of ionizing photon per comoving Mpc ,˙ N ion , as a function of redshift. We assume f esc = 0 .
2, if not oth-erwise specified. Square and triangle present the lower and upperlimits of ˙ N ion at z ≃ ρ obsUV and ρ upperUV , respectively.Three thick magenta lines with triangle/inverse-triangle representthe allowed ˙ N ion ranges at z = 7 for f esc = 0 .
2, 0 .
05, and 1 . N ion . Thin magenta lines denote the associated 1 σ errorscorresponding to 1 σ errors of the upper and lower limit estimates.For the presentation purpose, we shift the red square and the ma-genta lines along the abscissa. The exact redshift is the same asthe one of red triangle. Solid lines plot the model predictions of˙ N ion that is required for maintaining hydrogen ionization in IGM(Madau et al. 1999) with clumping factors of C HII = 1, 3, 10,and 30, from bottom to top. Dark gray area indicates that eventhe homogeneous Universe ( C HII = 1) lacks ionizing photons tomaintain hydrogen ionization of IGM in the model of Madau et al.(1999). The photoionization rates inferred from the Ly α forest areshown with purple circles (Bolton & Haehnelt 2007). Light grayshade displays constraints from Bolton & Haehnelt (2007) who findthat the clumping factor is C HII . z ∼ z & N ion at z = 4 − N ion calculated from their LFs down to theobservation limiting luminosities and L = 0, respectively, whichare counterparts of our ρ obsUV and ρ upperUV measurements. Althoughour ˙ N ion at z = 7 includes no AGN contribution due to no AGNLF measurements at z = 7, the contribution of AGN is probablysmaller than . .
08 dex at log ˙ N ion ≃
50 (see text). at the Lyman limit, where ρ UV is the UV luminos-ity density at ∼ α s = 3 that corresponds to a model spectrumof Leitherer et al. (1999) with continuous star-formationhistory, Salpeter IMF, and a metallicity of Z = 0 . Z ⊙ (Bolton & Haehnelt 2007). We apply three f esc values; f esc ≃ . f esc ≃ .
05 found in LBGs at z ∼ f esc = 1 . f esc ≃ .
05 is also sug-gested for z > N ion val-ues at z = 7 estimated from our ρ obsUV and ρ upperUV with f esc = 0 .
2. We regard these two ˙ N ion as the lower andupper limits (see § N ion with the threedifferent escape fractions, f esc = 0 .
2, 0 .
05, and 1 .
0. Wecalculate ˙ N ion at z = 4 − N ion at z = 4 −
6, weintegrate UV LFs down to L = 0 for the upper limitsand down to the observed magnitudes of M UV ≃ − z = 6), M UV ≃ −
17 ( z = 5), and M UV ≃ − z = 4) for the lower limits of Bouwens et al. (2008) andMcLure et al. (2009a) . Similarly, we integrate UV LFsdown to M UV = − . − . z = 5 and 4, respectively. TheseUV luminosity densities are used to estimate ˙ N ion witheq. (5) and f esc = 0 .
2. Then we add ionizing photonsfrom AGN given by Bolton & Haehnelt (2007), and plotthem in Figure 10. We find that ˙ N ion decreases mono-tonically from z = 4 to 7.Figure 10 also shows ˙ N ion that is required to balancerecombination of hydrogen IGM based on the model ofMadau et al. (1999),˙ N ion (s − Mpc − ) = 10 . C HII (1 + z ) (6)with clumping factors of C HII = 1, 3, 10, and 30. Notethat C HII = 1 corresponds to the homogeneous Universe,and that the Universe at z ∼ C HII > C HII =1, hydrogen IGM cannot maintain the ionized state forany clumping factors taken at the redshift. On the otherhand, the photoionization rates inferred from the Ly α forest indicate that clumping factor is as small as C HII . z ∼ C HII should monotonically decreasetowards high redshifts in the hierarchical Universe, themodels with C HII . z &
6. If ˙ N ion of objects falls in or beyond the model of1 < C HII . z &
6, the ionizing photon productionrate is high enough to maintain the ionized IGM.Figure 10 presents that, in the cases of f esc = 0 . f esc = 1 .
0, ˙ N ion values of z = 7 galaxies are comparablewith those predicted by the models of 1 < C HII . N ion to ourestimate at z = 7, because no AGN UV LF data areavailable at this redshift. However, AGN contributionof ionizing photon production is only log ˙ N ion ≃ . z = 6 (Bolton & Haehnelt 2007). Because the comovingdensity of luminous QSOs at z ∼ z ∼ z ∼
7. Even if weassume no evolution of AGN LF from z = 6 to 7, theAGN contribution is negligible; only pushing 0.08 dex atlog ˙ N ion ≃
50 in Figure 10.Our results indicate that the ionizing photon budgetjust balances at z = 7 for f esc & .
2. On the otherhand, ˙ N ion of z = 7 galaxies with f esc = 0 .
05 is a fac-tor of three below the model of C HII = 1, which hasthe ≃
95% (2 σ ) confidence level. Note that f esc ≃ . The measurements of McLure et al. (2009a) include onlythose at z = 6 and 5. is the measured escape fraction at z ∼ N ion estimates, i.e. the spectral index and break, are plausi-ble ones for low- z star-forming galaxies. Thus, there aretwo scenarios. a) If no properties of star-forming galax-ies at z = 7 are different from those at low redshifts,the universe could not be totally ionized by only galax-ies (and AGN) at z = 7 at the ≃
95% confidence level.b) If the properties of star-forming galaxies evolve fromlow redshifts, e.g. larger f esc ( f esc & . z = 7 is ionized and close to being in balancebetween the rates of ionizing photon production and re-combination of hydrogen IGM. If the scenario a) is true,the hydrogen IGM would experience a deficit of ionizingphoton at z = 7. This implies that the Universe maynot complete the reionization by z = 7. In this case,the Universe would start reionization right after z = 7and almost complete it by z ∼ z = 8 . .
7) is rejected at the 2 σ (3 σ ) level (Dunkley et al.2009). It is unlikely that reionization of the Universeis completed at the late epoch of z ∼ − f esc increases towards high redshifts at0 < z < f esc , which reaches 0.8at z = 10, and that the angular averaged escape fractionof f esc = 0 . − . z ∼ ≃ − M ⊙ ). Metallicity of galaxies also impacton the production of ionizing photons. Stiavelli et al.(2004) claim that the ionizing efficiency of a stellar pop-ulation increases by a factor of 3 for Salpeter IMF anda factor of 10 for a top-heavy IMF as the metallicity de-creases from Z = Z ⊙ to Z = 0 (see also Schaerer 2003). Moreover, dust absorption may be important in de-termination of escape fraction, as demonstrated by thesimulations of Laursen et al. (2009). Because Ly α emis-sivity would be higher at z ∼ z as suggestedby LAE studies (e.g. Ouchi et al. 2008), a flatter IMF aswell as lower metallicity and/or less dust extinction to-wards high- z may be plausible. It should be noted that,even in the scenario b), our observational constraints areclose to being in balance between ionizing photon pro-duction and recombination rates at z ∼
7. In otherwords, we might be witnessing the final stage of reioniza- Since the temperature of IGM increases from 10 ,
000 K(for the solar metallicity) to ∼ ,
000 K (for low metallicity;Osterbrock 1989), the low metallicity in IGM would reduce therecombination rate of IGM with the solar metallicity by a factor of ∼ arge Area Survey for z = 7 Galaxies 13tion with the closely balanced photon budget. It wouldprovide signatures of the neutral fraction evolution thatare claimed by Iye et al. (2006) and Ota et al. (2008)who find significantly less number of LAEs at z ∼ z ∼
6. Moreover, the scenario of b) is very consis-tent with the extended ( z ∼ −
11) reionization picturesuggested by Dunkley et al. (2009). Note that these ar-guments assume that there is no emergence of a largepopulation of very faint galaxies at z = 7 beyond theSchechter function, such claimed by Stark et al. (2007)for the earlier epoch of z = 9 −
10. On the other hand,Santos et al. (2004) have found that star-formation ac-tivities of low-mass galaxies are suppressed at z ≃ z = 7 galaxies would be needed to cor-rectly understand the contribution from these very faintgalaxies. Although there remain the arguments of thevery faint galaxy population, the ˙ N ion of z = 7 galaxieswith f esc = 0 .
05 still falls below the model of C HII = 1with a very steep faint-end slope down to α ≃ − . ρ upperUV , i.e., the upper limit of ˙ N ion bya factor of ≃ § Distribution of Dropouts: Indication of IonizedBubble?
Figure 11 presents the sky distribution of our z -dropout galaxies in SDF and GOODS-N. Although thenumbers of galaxies in each field are small, they appearto be clustered on the sky. The distribution in SDFshows possible three filamentary structures crossing ataround the center of the field; from top to the center,the center to bottom right, and the center to bottomleft. The possible filaments would extend up to ∼ z = 7 galaxies with a bright UVluminosity ( M UV ∼ − SF R nodust = 10 −
30) wouldbe strongly clustered, which are similar to those at z ∼ y < . z = 7.We find that these brightest galaxies are located at thehigh density regions of z -dropout galaxies both in SDFand GOODS-N. In the SDF panel of Figure 11, we alsomark the Ly α emitting dropout galaxy, SDF-63544, at z spec = 6 .
96 confirmed by spectroscopy (Iye et al. 2006).Interestingly, this Ly α emitting dropout galaxy sits atthe center of the 4 UV brightest dropout galaxies whosedistribution extends by ∼
30 Mpc in projection (Figure11). Because Ota et al. (2008) did not confirm the otherLy α emitter candidate found by their narrow-band sur-vey of z = 7 even with their deep spectroscopic data,this SDF-63544 may be only one with an observable Ly α emission line at z = 7 in the SDF. It would be possiblethat an overdense region of the 4 UV brightest dropoutswould make a well-established ionized bubble of IGM inthe cosmic volume with a size of &
30 Mpc, and that theionized bubble may allow SDF-63544 to transmit Ly α toobservers with no strong Ly α damping absorption givenby neutral hydrogen of IGM. To evaluate how much Ly α flux is absorbed by IGM, we estimate a Ly α equiva-lent width of the Ly α emitting dropout galaxy from theUV continuum magnitude ( − . ± .
31) and Ly α flux Fig. 11.—
Sky distribution of our z = 7 dropout galaxies down to y = 26 . y = 25 . y < . α emission at z spec = 6 .
96. Gray shades are masked areas where wedid not use the data for our analysis. The scales on the maps aremarked in both degrees and comoving megaparsecs in projectionat z = 7. North is up and east is to the left in these images. (2 × − erg s − cm − ; see § EW = 37 ± EW is about a half of the one for the case B recombinationwith no absorption of Ly α (68˚A; Nagamine et al. 2008;Ono Y. et al. 2009 submitted to MNRAS). We definethe escape fraction of Ly α emission, f Ly α esc , by f Ly α esc = L Ly α obs L Ly α int , (7)where L Ly α obs and L Ly α int are observed and intrinsic Ly α luminosities, respectively. If we assume the case B re-4 Ouchi et al.combination, L Ly α int [ergs − ] = 1 . × SFR[ M ⊙ yr − ],and SFR from M UV with no dust extinction correction(eq. 3) , we estimate the Ly α escape fraction of theLy α emitting dropout to be f Ly α esc = 0 . ± .
12. This f Ly α esc would be comparable to the one at z ∼ α , because the aver-age Ly α escape fraction is f Ly α esc = 0 . − .
54 at z ∼ α opacities of Fan et al. (2006),Madau (1995), and Meiksin (2006) for the case that IGMabsorbs a blue half of symmetric Ly α emission line (see § f Ly α esc wouldsupport the idea that the Ly α emitting dropout galaxysits inside a well-established ionized bubble with a neu-tral fraction as low as that at z ∼ z = 7.Assuming that this is a lower limit of the character-istic bubble size, we find that the analytic models ofFurlanetto et al. (2006) would suggest an upper limit ofneutral fraction of x HI .
20% at z = 7. This small up-per limit of x HI may indicate that the Universe is notfully neutral at z = 7. The combination of reionizationmodels and LAE LF (+clustering) gives constraints onneutral fraction of x HI .
50% at z = 6 . x HI . −
60% at z = 7 (Kobayashi et al. 2007; Ota et al.2008). Our possible upper limit of x HI .
20% at z = 7is consistent with those of previous results from the in-dependent observational probes. This implies that thereis no strong evidence rejecting the presence of the & z = 7. CONCLUSIONS
We have identified 22 z -dropout galaxy candidates inthe 0.4 deg area of SDF and GOODS-N down to y = 26with deep ≃ y -band images and Subaruand HST legacy imaging data. One out of 22 z -dropoutgalaxies in the SDF has a spectroscopic redshift of z =6 .
96 determined from Ly α emission. We have derivedthe bright-end UV LF of galaxies at z = 7. Based onour bright z -dropout galaxies as well as faint z -dropoutgalaxies obtained by the recent HST/WFC3 studies, wehave constrained the early stage of galaxy formation andphoton budget of cosmic reionization at z = 7. We havealso discussed the distributions of our z -dropout galaxies.The major results of our study are summarized below.1. We find that our bright-end UV LF shows a decreasefrom z ∼ . area.The best-fit Schechter parameters of z = 7 galaxies are φ ∗ = 0 . +2 . − . × − Mpc − , M ∗ UV = − . ± .
76 mag, and α = − . ± .
65. Our Schechter parameter fitresults reject no evolution of UV LF from z = 6 to 7 atthe >
95% confidence level. A more dominant decreaseof L ∗ than φ ∗ is preferable from z = 5 − z = 2 − z = 7 bya factor of at least &
6. It is likely that the cosmic SFRDdecreases roughly by a factor of ∼
10, but not larger than ∼ z star-forminggalaxies including the spectral shape and escape fraction( f esc ≃ . z = 7 galaxies falls below the hydrogen IGMrecombination rate predicted by the analytic models ofMadau et al. (1999) even in the homogeneous Universe( C HII = 1) at the ≃
95% (2 σ ) confidence level. Althoughit implies that the Universe cannot be totally ionized byonly galaxies at z = 7, but we think that properties ofgalaxies at z = 7 are just different from those at low red-shifts with, e.g., a larger escape fraction ( f esc & . z ∼ z -dropout galaxies may be strongly clustered bothin SDF and GOODS-N. We find that the distribution of z -dropout galaxies in SDF appears to be a filamentaryshape which extends up to 60 Mpc in projection, and thatthe z = 6 .
96 dropout galaxy with a Ly α line is locatedat the center of the overdense region consisting of the 4UV brightest dropout galaxy candidates. This impliesthat there may exist a well-established ionized bubblemade by the 4 UV brightest dropout galaxies, and thatthe ionized bubble might help to transmit the Ly α linein IGM at z = 7.We thank Ross Mclure, Rychard Bouwens, PascalOesch, and Andrew Hopkins for providing their data.We are grateful to Daniel Schaerer, Wei Zheng, Min-Su Shin, and David Sobral for their useful comments.We acknowledge the current and former Subaru Obser-vatory staff, especially Hisanori Furusawa, Akito Tajitsu,Miki Ishii, Michihiro Takami, and Fumiaki Nakata, fortheir invaluable help that made this challenging and long-standing project possible. M.O. has been supported viaCarnegie Fellowship. Facilities:
Subaru (Suprime-Cam) SFR thus obtained is 19.7 M ⊙ yr − for the Ly α emittingdropout. REFERENCESBarger, A. J., Cowie, L. L., & Richards, E. A. 2000, AJ, 119, 2092Barger, A. J., Cowie, L. L., & Wang, W.-H. 2008, ApJ, 689, 687Becker, G. D., Rauch, M., & Sargent, W. L. W. 2007, ApJ, 662, 72Bertin, E. & Arnouts, S. 1996, A&AS, 117, 393Bohlin, R. C., Colina, L., & Finley, D. S. 1995, AJ, 110, 1316Bolton, J. S., & Haehnelt, M. G. 2007, MNRAS, 382, 325Bouwens, R. J., Illingworth, G. D., Blakeslee, J. P., & Franx, M.2006, ApJ, 653, 53 Bouwens, R. J., Illingworth, G. D., Franx, M., & Ford, H. 2007,ApJ, 670, 928Bouwens, R. J., Illingworth, G. D., Franx, M., & Ford, H. 2008,ApJ, 686, 230Bouwens, R. J., et al. 2009a, ApJ, 690, 1764Bouwens, R. J., et al. 2009b, arXiv:0909.1803Bouwens, R. J., et al. 2009c, arXiv:0909.4074Bradley, L. D., et al. 2008, ApJ, 678, 647 arge Area Survey for z = 7 Galaxies 15 Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000Bunker, A. J., Stanway, E. R., Ellis, R. S., & McMahon, R. G.2004, MNRAS, 355, 374Bunker, A., et al. 2009, arXiv:0909.2255Capak, P., et al. 2004, AJ, 127, 180Castellano, M., et al. 2009, arXiv:0909.2853Cimatti, A., et al. 2002, A&A, 381, L68Cohen, J. G., Hogg, D. W., Blandford, R., Cowie, L. L., Hu, E.,Songaila, A., Shopbell, P., & Richberg, K. 2000, ApJ, 538, 29Cohen, J. G. 2001, AJ, 121, 2895Cole, S. et al. 2001, MNRAS, 326, 255Coleman, G. D., Wu,C.-C., & Weedman, D. W. 1980, ApJS, 43,393Condon, J. J., Cotton, W. D., & Broderick, J. J. 2002, AJ, 124,675Dawson, S., Stern, D., Bunker, A. J., Spinrad, H., & Dey, A. 2001,AJ, 122, 598Dickinson, M., Giavalisco, M., & The Goods Team 2003, The Massof Galaxies at Low and High Redshift, 324Dijkstra, M., Wyithe, J. S. B., & Haiman, Z. 2007, MNRAS, 379,253Dow-Hygelund, C. C., et al. 2007, ApJ, 660, 47Dunkley, J., et al. 2009, ApJS, 180, 306Eddington, A. S., Sir 1940, MNRAS, 100, 354Eyles, L. P., Bunker, A. J., Ellis, R. S., Lacy, M., Stanway, E. R.,Stark, D. P., & Chiu, K. 2007, MNRAS, 374, 910Fan, X., et al. 2004, AJ, 128, 515Fan, X., et al. 2006, AJ, 132, 117Flores, H., et al. 1999, ApJ, 517, 148Furlanetto, S. R., Zaldarriaga, M., & Hernquist, L. 2006, MNRAS,365, 1012Georgakakis, A., Hopkins, A. M., Sullivan, M., Afonso, J.,Georgantopoulos, I., Mobasher, B., & Cram, L. E. 2003,MNRAS, 345, 939Giavalisco, M., et al. 2004a, ApJ, 600, L93Giavalisco, M., et al. 2004, ApJ, 600, L103Glazebrook, K., Blake, C., Economou, F., Lilly, S., & Colless, M.1999, MNRAS, 306, 843Gunn, J. E. & Stryker, L. L. 1983, ApJS, 52, 121Hughes, D. H., et al. 1998, Nature, 394, 241Inoue, A. K., Iwata, I., & Deharveng, J.-M. 2006, MNRAS, 371,L1Inoue, A. K., & Iwata, I. 2008, MNRAS, 387, 1681Iwata, I., et al. 2009, ApJ, 692, 1287Iye, M., et al. 2006, Nature, 443, 186Hanish, D. J., et al. 2006, ApJ, 649, 150Hayashi, M., et al. 2009, ApJ, 691, 140Henry, A. L., Malkan, M. A., Colbert, J. W., Siana, B., Teplitz,H. I., McCarthy, P., & Yan, L. 2007, ApJ, 656, L1Henry, A. L., Malkan, M. A., Colbert, J. W., Siana, B., Teplitz,H. I., & McCarthy, P. 2008, ApJ, 680, L97Henry, A. L., et al. 2009, ApJ, 697, 1128Hickey, S., Bunker, A., Jarvis, M. J., Chiu, K., & Bonfield, D. 2009,arXiv:0909.4205Hopkins, A. M., & Beacom, J. F. 2006, ApJ, 651, 142Iliev, I. T., Shapiro, P. R., McDonald, P., Mellema, G., & Pen,U.-L. 2008, MNRAS, 391, 63Kamata, Y., et al. 2008, Proc. SPIE, 7021,Kashikawa, N., et al. 2003, AJ, 125, 53Kashikawa, N., et al. 2004, PASJ, 56, 1011Kashikawa, N., et al. 2006, ApJ, 648, 7Knapp, G. R., et al. 2004, AJ, 127, 3553Kneib, J.-P., Ellis, R. S., Santos, M. R., & Richard, J. 2004, ApJ,607, 697Kobayashi, M. A. R., Totani, T., & Nagashima, M. 2007, ApJ, 670,919Laursen, P., Sommer-Larsen, J., & Andersen, A. C. 2009,arXiv:0907.2698Lee, K.-S., Giavalisco, M., Conroy, C., Wechsler, R. H., Ferguson,H. C., Somerville, R. S., Dickinson, M. E., & Urry, C. M. 2009,ApJ, 695, 368Leitherer, C., et al. 1999, ApJS, 123, 3Lowenthal, J. D., et al. 1997, ApJ, 481, 673Ly, C., et al. 2009, ApJ, 697, 1410Madau, P. 1995, ApJ, 441, 18Madau, P., Pozzetti, L., & Dickinson, M. 1998, ApJ, 498, 106Madau, P., Haardt, F., & Rees, M. J. 1999, ApJ, 514, 648 Mannucci, F., Buttery, H., Maiolino, R., Marconi, A., & Pozzetti,L. 2007, A&A, 461, 423McLure, R. J., Cirasuolo, M., Dunlop, J. S., Foucaud, S., &Almaini, O. 2009a, MNRAS, 395, 2196McLure, R. J., Dunlop, J. S., Cirasuolo, M., Koekemoer, A. M.,Sabbi, E., Stark, D. P., Targett, T. A., & Ellis, R. S. 2009b,arXiv:0909.2437McQuinn, M., Hernquist, L., Zaldarriaga, M., & Dutta, S. 2007,MNRAS, 381, 75Meiksin, A. 2006, MNRAS, 365, 807Meurer,G. R., Heckman, T. M., & Calzetti, D. 1999, ApJ, 521, 64Miyazaki, S. et al. 2002, PASJ, 54, 833Morokuma, T., et al. 2008, ApJ, 676, 163Nagamine, K., Ouchi, M., Springel, V., & Hernquist, L. 2008,arXiv:0802.0228Nagao, T., et al. 2007, A&A, 468, 877Oesch, P. A., et al. 2009a, ApJ, 690, 1350Oesch, P. A., et al. 2009b, arXiv:0909.1806Oke, J. B. 1990, AJ, 99, 1621Osterbrock, D. E. 1989, Astrophysics of gaseous nebulae and activegalactic nuclei (Mill Valley: University Science Books)Ota, K., et al. 2008, ApJ, 677, 12Ouchi, M., et al. 2004a, ApJ, 611, 660Ouchi, M., et al. 2005, ApJ, 620, L1Ouchi, M., et al. 2008, ApJS, 176, 301Ouchi, M., et al. 2009, ApJ, 696, 1164Overzier, R. A., Bouwens, R. J., Illingworth, G. D., & Franx, M.2006, ApJ, 648, L5Papovich, C., Dickinson, M., & Ferguson, H. C. 2001, ApJ, 559,620P´erez-Gonz´alez, P. G., et al. 2005, ApJ, 630, 82Phillips, A. C., Guzman, R., Gallego, J., Koo, D. C., Lowenthal,J. D., Vogt, N. P., Faber, S. M., & Illingworth, G. D. 1997, ApJ,489, 543Razoumov, A. O., & Sommer-Larsen, J. 2009, arXiv:0903.2045Reddy, N. A., Steidel, C. C., Erb, D. K., Shapley, A. E., & Pettini,M. 2006, ApJ, 653, 1004Richard, J., Pell´o, R., Schaerer, D., Le Borgne, J.-F., & Kneib,J.-P. 2006, A&A, 456, 861Richard, J., Stark, D. P., Ellis, R. S., George, M. R., Egami, E.,Kneib, J.-P., & Smith, G. P. 2008, ApJ, 685, 705Ricotti, M., & Ostriker, J. P. 2004, MNRAS, 352, 547Ricotti, M., Ostriker, J. P., & Mack, K. J. 2008, ApJ, 680, 829Poznanski, D., et al. 2007, MNRAS, 382, 1169Richmond, M. W., Morokuma, T., Doi, M., Komiyama, Y., Yasuda,N., Okamura, S., & Gal-Yam, A. 2009, PASJ, 61, 97Ryan, R. E., Jr., Hathi, N. P., Cohen, S. H., & Windhorst, R. A.2005, ApJ, 631, L159Santos, M. R., Ellis, R. S., Kneib, J.-P., Richard, J., & Kuijken, K.2004, ApJ, 606, 683Schaerer, D. 2003, A&A, 397, 527Serjeant, S., Gruppioni, C., & Oliver, S. 2002, MNRAS, 330, 621Shapley, A. E., Steidel, C. C., Pettini, M., & Adelberger, K. L.2003, ApJ, 588, 65Shapley, A. E., Steidel, C. C., Pettini, M., Adelberger, K. L., &Erb, D. K. 2006, ApJ, 651, 688Sheth, R. K. & Tormen, G. 1999, MNRAS, 308, 119Schaerer, D., & de Barros, S. 2009, A&A, 502, 423Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500, 525Shimasaku, K., et al. 2003, ApJ, 586, L111Shimasaku, K., Ouchi, M., Furusawa, H., Yoshida, M., Kashikawa,N., & Okamura, S. 2005, PASJ, 57, 447Shimasaku, K., et al. 2006, PASJ, 58, 313Sobral, D., et al. 2009, MNRAS, 398, L68Stanway, E. R., et al. 2004a, ApJ, 604, L13Stanway, E. R., Bunker, A. J., McMahon, R. G., Ellis, R. S., Treu,T., & McCarthy, P. J. 2004b, ApJ, 607, 704Stanway, E. R., McMahon, R. G., & Bunker, A. J. 2005, MNRAS,359, 1184Stanway, E. R., et al. 2007, MNRAS, 376, 727Stark, D. P., Ellis, R. S., Richard, J., Kneib, J.-P., Smith, G. P., &Santos, M. R. 2007, ApJ, 663, 10Steidel, C. C., Giavalisco, M., Dickinson, M., & Adelberger, K. L.1996, AJ, 112, 352Steidel, C. C., Adelberger, K. L., Giavalisco, M., Dickinson, M., &Pettini, M. 1999, ApJ, 519, 1Steidel, C. C., Adelberger, K. L., Shapley, A. E., Pettini, M.,Dickinson, M., & Giavalisco, M. 2003, ApJ, 592, 728
Stiavelli, M., Fall, S. M., & Panagia, N. 2004, ApJ, 610, L1Trenti, M., & Stiavelli, M. 2008, ApJ, 676, 767Tresse, L., Maddox, S. J., Le F`evre, O., & Cuby, J.-G. 2002,MNRAS, 337, 369Vanzella, E., et al. 2006a, A&A, 454, 423Wirth, G. D., et al. 2004, AJ, 127, 3121Wyithe, S., Hopkins, A. M., Kistler, M. D., Yuksel, H., & Beacom,J. F. 2009, arXiv:0908.0193Yagi, M., Kashikawa, N., Sekiguchi, M., Doi, M., Yasuda, N.,Shimasaku, K., & Okamura, S. 2002, AJ, 123, 66 Yajima, H., Umemura, M., Mori, M., & Nakamoto, T. 2009,arXiv:0906.1658Yan, H., et al. 2004, ApJ, 616, 63Yan, H., et al. 2005, ApJ, 634, 109Yoshida, M., et al. 2006, ApJ, 653, 988Zheng, W., et al. 2009, ApJ, 697, 1907 arge Area Survey for z = 7 Galaxies 17 TABLE 1Summary of Imaging Observations and Data
Field Band CCD t exp
PSF size a Area m limb Date of Observations and Note(sec) (arcsec) (arcmin ) (3 σ AB mag)SDF y MIT-LL 50614 · · · · · · · · · c y Hamamatsu 43901 · · · · · · · · · y (total) 94515 0.79 (0.99) 810 26.4 · · · z ′ MIT-LL 104069 0.91 (0.99) 810 27.7 Data from Morokuma et al. in preparation d GOODS-N y MIT-LL 89308 · · · · · · · · · y Hamamatsu 29101 · · · · · · · · · y (total) 118409 0.87 (0.87) 758 26.2 · · · z ′ MIT-LL 39150 0.87 (0.87) 758 26.9 2007 Apr. 19, (2001 Apr. 20, 22,2003 Apr. 6, 2004 Mar. 16, 2006 Feb. 23-24) ca The FWHM of PSFs in the reduced image. The values in parenthesis indicate the FWHM of PSFs that are matched with broad-band images ineach field. b The limiting magnitude defined by a 3 σ sky noise level in a 1 ′′ . c The observation dates in parenthesis are those of the Subaru archival data taken by the other teams. d See § TABLE 2Limiting Magnitudes
Band GOODS-N SDF(1) (2) U . · · · B . . B . · · · V . . V . · · · R . . I or i ′ . . i . · · · z . · · · z ′ . . y . . m . . · · · m . . · · · m . . · · · m . . · · · m . · · · Note . — Three sigma limit-ing magnitudes in GOODS-N (1)and SDF (2). The magnitudesare defined with a 1 ′′ . UBV Rizy ), ex-cept for the GOODS-N U , V ,and R , data. We apply a 0 ′′ . B , V , i , z ),and a 3 ′′ . . µ m, 4 . µ m, 5 . µ m,8 . µ m, and 24 µ m images ( m . , m . , m . , m . , m ). Thelimiting magnitudes of Spitzerdata include the offsets of aper-ture corrections (see text). The detection limits ofground-based GOODS-N U , V ,and R images are defined witha 3 ′′ . F W HM ≃ ′′ .
5; see the text). The 4 σ detection limits of y magnitudes are 25 . . σ up-per limits of z ′ magnitudes are27 . . arge Area Survey for z = 7 Galaxies 19 TABLE 3 z = 7 Galaxy Candidates ID U B V R I/i z y m . m . z ′ − y z spec Note
GOODS-N
GOODSN152505 a > . > . > . > . > . > . . > . b > . b > . · · · Source conf. in Spitzer bands( > .
4) ( > .
6) ( > .
9) ( > . > . > . > . > . > . > . . · · · · · · > . · · · · · · GOODSN201340 > . > . > . > . > . b > . . · · · · · · > . · · · Source confusionGOODSN104059 > . > . > . > . > . > . . · · · · · · > . · · · · · · GOODSN108036 a > . > . > . > . > . > . . > . b > . b > . · · · Source conf. in Spitzer bands( > .
4) ( > .
6) ( > .
9) ( > . > . > . > . > . > . > . . · · · · · · > . · · · Spurious source?GOODSN134896 > . > . > . > . > . > . . · · · · · · > . · · · · · · SDF
SDF63544 · · · > . > . > . > . .
02 25 . · · · · · · · · · > . > . > . > . .
04 25 . · · · · · · · · · · · · SDF46975 · · · > . > . > . > . .
48 25 . · · · · · · · · · · · · SDF76507 · · · > . > . > . > . .
11 25 . · · · · · · · · · · · · SDF123919 · · · > . > . > . > . .
51 25 . · · · · · · · · · · · · SDF77202 · · · > . > . > . > . .
50 25 . · · · · · · · · · · · · SDF75298 · · · > . > . > . > . .
50 25 . · · · · · · · · · · · · SDF20911 · · · > . > . > . > . .
41 25 . · · · · · · · · · · · · SDF121488 · · · > . > . > . > . .
49 25 . · · · · · · · · · · · · SDF84539 · · · > . > . > . > . .
16 25 . · · · · · · · · · · · · SDF16416 · · · > . > . > . > . > . . · · · · · · > . · · · · · · SDF64206 · · · > . > . > . > . > . . · · · · · · > . · · · · · · SDF107344 · · · > . > . > . > . > . . · · · · · · > . · · · · · · SDF136726 · · · > . > . > . > . .
59 26 . · · · · · · · · · · · · SDF41484 · · · > . > . > . > . b .
05 26 . · · · · · · · · · Source confusion
Note . — The upper limits in the ground-based images,
UBV RIizy , are defined by the 2 σ level. The numbers in parenthesis are 3 σ upper limits ofHST/ACS B , V , i , and z bands. a Our candidates falling in the central ∼
160 arcmin field of GOODS-N with HST and Spitzer images. There are no obvious counterparts in IRAC 5 . µ m,8 . µ m, and MIPS 24 µ m bands (see Figure 4). The 3 σ upper limits in these bands are ( m . , m . , m ) = (23 . , . , . b Fluxes of these objects are contaminated by close bright objects on the sky.
TABLE 4UV Luminosity Function at z = 7 φ ∗ M ∗ UV α Mag. Range n obs ρ obsUV ρ upperUV (10 − Mpc − ) (mag) (mag) (10 − Mpc − ) (10 erg s − Hz − Mpc − ) (10 erg s − Hz − Mpc − )(1) (2) (3) (4) (5) (6) (7)0 . +2 . − . − . ± . − . ± . − . < M < − . . +2 . − . . +6 . − . . +10 . − . Note . — (1)-(3): Best-fit Schechter parameters. The values of φ ∗ and M ∗ UV are given in units of 10 − Mpc − and AB magnitude, respectively. Thereduced χ of the fitting is 0.13. (4): Magnitude range of UV LFs that are used for the fitting. (5)-(6): Number density (in 10 − Mpc − ) and UV luminositydensity (in 10 erg s − Hz − Mpc − ) calculated with the best-fit Schechter parameters down to the limit of UV magnitude, M UV ≤ −
18, defined by theHST/WFC3 studies. (7): The upper limit UV luminosity density which is the integral of the best-fit Schechter function down to M UV = ∞∞