Discovery of Lyman Break Galaxies at z~7 from the ZFOURGE Survey
V. Tilvi, C. Papovich, K.-V. H. Tran, I. Labbe, L. R. Spitler, C. M. S. Straatman, S. E. Persson, A. Monson, K. Glazebrook, R. F. Quadri, P. van Dokkum, M. L. N. Ashby, S. M. Faber, G. G. Fazio, S. L. Finkelstein, H. C. Ferguson, N. A. Grogin, G. G. Kacprzak, D. D. Kelson, A. M. Koekemoer, D. Murphy, P. J. McCarthy, J. A. Newman, B. Salmon, S. P. Willner
aa r X i v : . [ a s t r o - ph . C O ] A p r Preprint typeset using L A TEX style emulateapj v. 5/2/11
DISCOVERY OF LYMAN BREAK GALAXIES AT Z ∼ V. Tilvi , C. Papovich , K.-V. H. Tran , I. Labb´e , L. R. Spitler , C. M. S. Straatman , S. E. Persson , A.Monson , K. Glazebrook , R. F. Quadri , P. van Dokkum , M. L. N. Ashby , S. M. Faber , G. G. Fazio S. L.Finkelstein
H. C. Ferguson , N. A. Grogin , G. G. Kacprzak , D. D. Kelson , A. M. Koekemoer , D.Murphy , P. J. McCarthy , J. A. Newman B. Salmon , S. P. Willner ABSTRACTStar-forming galaxies at redshifts z > z ∼ − break galaxies (LBGs) from a 155 arcmin area in the CANDELS/COSMOS field imaged bythe deep FourStar Galaxy Evolution (zFourGE ) survey. The FourStar medium-band filters providethe equivalent of R ∼
10 spectroscopy, which cleanly distinguishes between z ∼ z ∼ ∼ z ∼ M UV ∼ − . z ∼ z ∼
5. Fitting the galaxies’ spectralenergy distributions, we predict Lyman- α equivalent widths for the two brightest LBGs, and findthat the presence of a Lyman- α line affects the medium-band flux thereby changing the constraintson stellar masses and UV spectral slopes. This illustrates the limitations of deriving LBG propertiesusing only broad-band photometry. The derived specific star-formation rates for the bright LBGs are ∼
13 Gyr − , slightly higher than the lower-luminosity LBGs, implying that the star-formation rateincreases with stellar mass for these galaxies. Subject headings: galaxies: high-redshift — galaxies: Lyman break galaxies–galaxies:Luminosity Func-tion INTRODUCTION
Discovering high-redshift ( z >
6) galaxies, and un-derstanding their physical nature is of great importancein studying the early universe because these objectsare likely sources for reionization of the intergalacticmedium (IGM: e.g. Trenti et al. 2010; Salvaterra et al.2011; Finkelstein et al. 2012) at z & z & λ > µ m.There are two main techniques for finding high-redshift( z &
7) galaxies : (1) the Lyman-break or dropout George P. and Cynthia Woods Mitchell Institute for Funda-mental Physics and Astronomy, and Department of Physics andAstronomy, Texas A&M University, College Station, TX. Sterrewacht Leiden, Leiden University, NL-2300 RA Leiden,The Netherlands Centre for Astrophysics & Supercomputing, Swinburne Uni-versity, Hawthorn, VIC 3122, Australia Carnegie Observatories, Pasadena, CA 91101, USA Department of Astronomy, Yale University, New Haven, CT06520, USA Harvard-Smithsonian Center for Astrophysics, 60 GardenSt., Cambridge, MA 02138 USA UCO/Lick Observatory, Department of Astronomy and As-trophysics, University of California, Santa Cruz, CA, USA University of Texas, Austin, TX, USA Space Telescope Science Institute, Baltimore, MD, USA University of Pittsburgh, PA, USA Hubble Fellow Australian Research Council Super Science Fellow method (e.g. Steidel et al. 1995, 1999; Adelberger et al.2004; Dickinson et al. 2004; Giavalisco et al. 2004;Ouchi et al. 2004; Bouwens et al. 2007; McLure et al.2010), selected based on strong absorption in their spec-tral energy distribution (SED) at wavelengths shortwardof redshifted Lyman- α due to IGM HI absorption, and(2) the narrow-band imaging technique, which identifiesgalaxies with a strong, redshifted Lyman- α emission lineusing narrow-band filters (e.g. Malhotra & Rhoads 2002;Rhoads et al. 2005; Iye et al. 2006; Kashikawa et al.2006; Ouchi et al. 2009b; Hibon et al. 2010; Tilvi et al.2010; Krug et al. 2012). A unique advantage of narrow-band selected galaxies is that these galaxies are morelikely to be confirmed via spectroscopic followup be-cause they are pre-selected based on the strong Lyman- α emission line. On the other hand, LBGs selectedvia the broad-band dropout method are unbiased intheir selection (unlike Lyman- α selected galaxies), andLBG surveys probe larger cosmological volumes dueto broader filters. Currently there are many candi-date LBGs (e.g. Bouwens et al. 2007; Ouchi et al. 2009;McLure et al. 2010) at redshifts much greater than z ∼ z ∼ Fig. 1.—
Filter transmission curves of the FourStar medium-band compared to other broad-band filters. The top panel showsbroad-band Subaru z ′ , HST Y , J & H filters, while thebottom panel shows FourStar medium-band filters J , J , J , H s and H l , which provide R ∼
10 resolution spectroscopy. Black lineshows a template spectral energy distribution of a z = 7 . α emission line falling in the J filter. The topaxis marks the Lyman- α redshifts as a function of wavelength. lows us to study the galaxy evolution by comparing thenumber density of galaxies at different redshifts.Several studies have focused on understanding the evo-lution of the UV luminosity function from z ∼ z >
6. Current observations at z > ⋆ ∼ -19.8 mag (Bouwens et al. 2008;McLure et al. 2010; Ouchi et al. 2009; Yan et al. 2010;Castellano et al. 2010; Bouwens et al. 2010) comparedwith M ⋆ ∼ -20.9 mag at z ∼ z ∼
7, rely on
HST observations which have the advantage of prob-ing fainter galaxies, however with smaller survey areas.Recently, Ouchi et al. (2009), Castellano et al. (2010),Capak et al. (2011), and Bowler et al. (2012) found sev-eral candidates using ground-based observations, whichhave the advantage of surveying larger areas and thusprobing brighter, rarer galaxies.While it is relatively easy to construct the luminosityfunction, its accuracy depends on a number of systemat-ics, including the degree of contamination of the z ∼ z ∼ z ∼ −
2) dust obscured galaxies with very faint continuum in the visible bands, and transient ob-jects. The brown dwarf stars may be separated from(resolved) galaxies using high angular-resolution imag-ing. However, the ability to resolve galaxies can becomeunreliable at the faint magnitudes and compact sizes typ-ical of z ∼ HST observations.For example, the rejection of point sources from z ∼ z ∼ HST resolution (FWHM ≈ . ′′ z = 7), leading to sample incompleteness. These uncer-tainties need to be accounted for when constructing theluminosity function.In addition to studying the UV luminosity func-tion, understanding the physical properties of high-redshift galaxies is important to link their evolu-tion during the early stages of galaxy formationto galaxies at other epochs. As we probe red-shifts z &
2, galaxies on average have smallerstellar masses ∼ M ⊙ (e.g. Sawicki & Yee1998; Papovich et al. 2001; Shapley et al. 2001, 2005;Yan et al. 2005; Erb et al. 2006; Fontana et al. 2006;Reddy et al. 2006; Overzier et al. 2009), bluer UV col-ors (e.g. Dickinson et al. 2003; Papovich et al. 2004;Labb´e et al. 2007; Overzier et al. 2009; Bouwens et al.2009a), and younger stellar populations with smalleramounts of dust (e.g. Meurer et al. 1999; Papovich et al.2001; Shapley et al. 2001, 2005; Reddy et al. 2005,2006, 2008; Reddy & Steidel 2009; Bouwens et al. 2009a;Finkelstein et al. 2010, 2011). Moreover, the specificstar-formation rates (sSFR: the SFR per unit stellarmass) of galaxies at z & ∼ − ,albeit with large uncertainties (e.g. Stark et al. 2009;Gonz´alez et al. 2010; Bouwens et al. 2011; Reddy et al.2012). In all cases, the physical properties of themost distant galaxies ( z > z ∼ z ∼ ∼
10) of themedium-band filters. This led to a recent discovery ofa distant ( z = 2 .
1) galaxy cluster in the COSMOS field(Spitler et al. 2012). Owing to very deep imaging andexcellent seeing, we are also able to place better con-straints on the redshifts and on the evolution of brightstar-forming galaxies out to z ∼ z ∼ f /f z ′ ( < r) U n c on v o l v ed Convolved 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4Radius(arcsec)0100200300400500600 S / N Fig. 2.—
Left panel: Curves-of-growth for point sources in the convolved and unconvolved images. Dashed and solid lines show fluxratio before and after convolution, respectively, where f is the flux in a given band and f z ′ is the z ′ (reference image) flux enclosedwithin radius r. Different colors indicate different filters. The flux in aperture radii larger than 0 . ′′
48 among convolved images contributeto <
2% uncertainty for point sources. Right panel: The signal-to-noise (S/N) for point sources as a function of radii, in the convolvedimages. Different lines show the S/N of same stars (here shown only the J filter for illustration purposes) that were used for creating theconvolution kernel (Section 2). For photometry, we use a circular aperture with 0 . ′′
55 radius as this yields maximum S/N.
TABLE 1Photometry for COSMOS field.
Filter λ ( µm ) Depth (mag) a Filter width(˚A) c g ′ r ′ i z ′ J J J J (F125W) H s H l H (F160W) K s . b d [4 . b da σ depth in 1 . ′′ b σ depth in 2 . ′′ c Filter width where filter transmission is > d Transmission is > This paper is organized as follows. Section 2 describesthe observations, data reduction, and catalog generation.In Section 3 we present our LBG color selection methodand sources of contamination and their identification.Section 4 describes the construction of the UV luminos-ity function and its comparison to previous studies. Wecompare the advantage of using medium-band over onlybroad-band photometry, and its influence on the derivedphysical properties of high-redshift galaxies in Section 5.We summarize our conclusions in Section 6. Throughoutthis paper we use AB magnitudes and standard cosmol-ogy with h = 0 .
7, Ω λ = 0 .
7, and Ω m = 0 . OBSERVATIONS AND DATA REDUCTION
We observed a 155 arcmin area in the CAN-DELS /COSMOS field (Scoville et al. 2007) centeredat RA: 10:00:31 and Dec:+02:17:21, using the FourStarinstrument, as part of the ongoing FourStar GalaxyEvolution survey (zFourGE ) during 2011-2012. TheFourStar instrument (Persson et al. 2013) is a near-IRcamera mounted on the Magellan/Baade telescope. Ithas 4096 × . ′′ ′ .8 × ′ .8 field of view. We observedwith five adjacent medium-band filters ( J , J , J , H s , H l : Table 1) with a resolution of R ∼
10, and onebroad-band ( K s ) filter (Labb´e et al in preparation). Fig-ure 1 shows filter transmission curves for the FourStarmedium-band filters compared to the broad-band filtersfrom Subaru/Suprime ( z ′ ), and HST ( Y , J , H )filters. The J filter has a similar central wavelengthto the Y filter (and also to many Y -filters on ground-based telescopes), but the J filter is much narrower com-pared with the HST Y filter.We processed all raw images using a custom de-signed IDL-based pipeline (described in more detail inLabb´e et al in preparation) adapted from the NEW-FIRM Medium Band Survey data reduction pipeline(NMBS: Whitaker et al. 2011). The data quality of theFourStar images is excellent and the point-spread func-tion (PSF) FWHM for the stacked FourStar images cor-respond to 0.55, 0.52, 0.51, 0.51, 0.57, & 0.48 arcsec inthe J through K s filters, respectively. In addition tothis data set, we used publicly available data in the op-tical (Subaru g ′ , r ′ , z ′ , & ACS i ( i )), deep near-IR HST
WFC3 J & H from the CANDELS survey(Koekemoer et al. 2011; Grogin et al. 2011), and infraredIRAC (Fazio et al. 2004) data. We also used deep 3.6and 4.5 µ m images (Ashby et al 2013 submitted) from theSpitzer Extended Deep Survey ( SEDS ) ; 5.8 and 8.0 µ m http://candels.ucolick.org/ http://z-fourge.strw.leidenuniv.nl/ Tilvi et al.data from the Spitzer Space Telescope (Werner et al.2004). Table 1 presents the central wavelengths and lim-iting magnitudes for the zFourGE data as well as infor-mation for the ancillary data.The PSF among different bands varies due to differingcentral wavelengths, different instruments used for imag-ing, and in the case of ground-based observations, due todifferent seeing conditions. We convolved all images toa reference image having the broadest PSF-FWHM (inour case the Subaru z ′ image with PSF FWHM=0 . ′′ . ′′ . ′′
55. Thereforewe use circular apertures with diameter of 1 . ′′ Spitzer/IRAC Photometry
As mentioned earlier, to extend the wavelength cover-age of our candidates and to derive the physical proper-ties as accurately as possible, we used recent deep SpitzerIRAC data (Ashby et al 2013 submitted) in two bands(3.6 µm , 4.5 µm ) obtained from the SEDS survey. How-ever, due to its low resolution, neighboring objects tendto blend resulting in incorrect flux measurements. To cir-cumvent this issue, we used GALFIT (Peng et al. 2002)code to measure the object fluxes, as described below.First, we fit the two dimensional surface brightness pro-files of all objects within 7 ′′ of the z ∼ z ∼ z ∼ ′′ diameter aperture and correcting to the total flux usingthe aperture corrections derived by taking the ratio be-tween 2 ′′ and 15 ′′ diameter aperture fluxes of isolatedsources. Finally, we derived the uncertainties by mea-suring the rms in apertures of the same size as above,randomly placed in regions devoid of objects in the IRACimages. Source Catalogs
We identified sources in each of the images using SourceExtractor software (SExtractor : Bertin & Arnouts1996) in dual image mode. In this mode, a detectionimage is used to identify the pixels associated with eachobject, while the fluxes are measured from a distinct pho-tometry image. For the detection image, we used a χ image, which optimally sums multiple images (account-ing for varying S/N) for source detection (Szalay et al. http://irsa.ipac.caltech.edu/data/COSMOS/ χ image ( J chq ) was constructed from the J , J , J images using SWarp (Bertin et al. 2002). Wemeasured object photometry from the PSF-matched im-ages in a fixed aperture size of 1.1 ′′ diameter.While SExtractor estimates the noise for each object,this estimate is incorrect in the convolved images as itdoes not account for correlated noise, which results fromthe image reduction and convolution processes. To esti-mate the flux uncertainty for each object in the convolvedimages we followed an empirical approach described inLabb´e et al. (2003). To estimate the pixel-to-pixel noisewe measured the rms background variation as a func-tion of linear size N = √ A where A is the area of anaperture. First, we measured the sky flux in increasinglylarger apertures that are randomly placed in the PSF-matched-convovled image, taking care to avoid detectedobjects. From the distribution of fluxes measured in agiven aperture size, we measured the standard deviation, σ . The 1 σ error on the flux was then calculated usingequation 5 from Whitaker et al. (2011). Our final cata-log contains aperture fluxes obtained from PSF-matchedimages in g ′ , r ′ , i , z ′ , J , J , J , J , H s , H l , H , and K s along with flux uncertainties. Photometric Redshifts
We derive photometric redshifts using the photomet-ric redshift code EAZY (Brammer et al. 2008) takingadvantage of extensive multi-wavelength data publiclyavailable in the COSMOS field for finding the best-fitsolution to the observed data. The EAZY code (v2.1;Brammer et al. 2011) contains seven different templatesspanning a large range of galaxy colors and incorporatesnebular emission lines. While our z ∼ z > LBG COLOR SELECTION AT Z ∼ Our initial color-selection criteria to identify candidategalaxies at z ∼ / N( J chq ) > . / N(optical) < . , J − J < . ,z ′ − J > . ,z ′ − J > . . J − J ) mag ,z ′ − J > − . J − J ) mag . Similar criteria have been used in other studies to identify z ∼ J filter as this filter is very effective in discriminat-ing T-dwarf stars (Tilvi et al in preparation), due to theirmethane and water absorption features (Tinney et al.2012), from high-redshift star-forming galaxies.Figure 3 shows a J - J vs z ′ - J color-color plot andillustrates the region (dashed line) that corresponds tothe selection criteria of Equation 1. Color-color tracksof star-forming galaxies with two different dust attenu-ation are shown in blue and red lines with redshift la-bels. These tracks are obtained from Bruzual & Charlotiscovery of Lyman Break Galaxies at z ∼ -0.5 0.0 0.5 1.0 1.5J -J (mag)01234 z ′ - J ( m ag ) z=6.0z=6.3z=6.5z=6.7z=6.9z=7.0 E ( B - V ) = E ( B - V ) = . brown dwarfs 2014 24834 26680 30058 4329 13965 25035 26343 26836 z~7 candidates Likely foreground galaxyT-dwarf candidates Fig. 3.—
Color-color plot for z ∼ z ∼ z ∼ z ∼ (2003) stellar population synthesis models by integratingthe model spectral energy distributions with the z ′ , J ,and J filter. The models assumed E(B-V)= 0.0 (blueline) and 0.1 (red line) with Z=0.2Z ⊙ metallicity, con-stant star-formation history, IGM opacity taken from(Meiksin 2006), and a constant age of t = 100 Myr. Basedon these color tracks it can be seen that the star-forminggalaxies occupy a specific region of the color-color space.We identified 9 objects in total satisfying the selec-tion criteria from Equation 1. Based on several tests de-scribed in the following sections, out of the 9 objects inthe initial sample (hereafter initial source list), four ob-jects are identified as z ∼ z <
3. To minimize the contamination fromstars and foreground galaxies, we further refine the color-color selection (Section 3.3) that include only the robust z ∼ z ∼ Sources of Contamination
The color-color selection of z ∼ Spurious detections
In the z ∼ > z ′ filter. This requirement makes spuriousdetections due to either background or electronic noiseextremely unlikely. Transient/High-proper motion objects
It is possible that transient objects including asteroids,and supernovae can be detected in more than one band.However, with the availability of two epoch data, sep-arated by about one year, we can eliminate any tran-sients by comparing their positional shift between thetwo epochs. Therefore, we conclude that the initialsource list does not contain any transient sources.
Foreground galaxies
While the z ∼ < σ ) in the observed visible bands, it ispossible that the initial source list contains foreground Tilvi et al. zF25035 zF26836zF13965 zF26343 zF26680 g’ r’ i z’ J1 J2 J3 J125 Hs Hl H160 Ks [3.6] [4.5] [5.7] [8] z F z F z F P = . P = . P = . z F P = . T d w a r f z F P = . Fig. 4.—
Image cutouts of four z ∼ z ′ , J , H )image cutouts while lower two panels show image cutouts in the individual filters. The four z ∼ ′′ radius circle. These candidates are clearly detected in at least one of the medium-bandfilters while undetected ( < σ ) in any of the visible bands. The integrated probability distribution P (Equation 2) is shown on theright side. In addition to satisfying the color-color selection criteria, each of the candidates is required to have P > . < χ image. These two criteria minimizes the contamination from low-redshift galaxies. In the bottom panel, we show one ofthe candidate T-dwarf stars that was selected in the initial color-color selection (Oesch et al 2010). Such T-dwarf stars can be cleanlydistinguished from galaxies using medium-band photometry. objects, likely z ∼ − z ∼ z ∼ − . − [5 . . z ∼ z ∼ P given by P = R P ( z ) dz R P ( z ) dz , (2)where R P ( z ) dz is the integrated probability distribu-tion from z = 6 to z = 9. This probability distributionis normalized by R P ( z ) dz such that P ≤
1. We makeuse of the probability distribution P ( z ) derived using the photometric redshift code EAZY and require that eachof the z ∼ P > .
7. This means thateach of the z ∼ z > P > .
99 and in principle, we could restrict thesample to the stricter requirement of P > .
99, but thiswould have risked missing some z ∼ Visible χ Image
To further test the reliability of z ∼ g ′ , r ′ , and i images to construct a visible χ image; thecombined χ image is significantly deeper than individ-ual images. We then run SExtractor to measure the S/Nof all objects in the χ image. Based on S/N( >
2) inthe visible χ image, we reject one object (zF4329) outof 9 objects from the initial source list, as being a likelyforeground galaxy with very faint continuum. This re-jection is also supported by its strong detection in theMIPS (Rieke et al. 2004) 24 µm flux; there are no MIPSdetections for any of the remaining objects. Distinguishing Stars from Compact Galaxies
Galactic brown dwarfs ( M , L , and T ) are one of themain contaminants because their near-IR colors resem-iscovery of Lyman Break Galaxies at z ∼ z ∼
7) galaxy colors. While it is pos-sible to identify candidates for stars using their FWHMand stellarity index (produced by SExtractor) when us-ing
HST observations, these classifiers tend to becomeunreliable at fainter magnitudes. In addition, at high-redshifts we expect some galaxies to be very compactwith their FWHM ( < z ∼
7) comparable tothe PSF FWHM. In such cases, even with
HST obser-vations, it would be challenging to discriminate betweenstars and compact galaxies.
20 22 24 26 28J (mag)0.51.01.52.02.5 F W H M ( a r cs e c ) FourStar z~7 candidates Likely foreground galaxydwarf star
20 22 24 26 28J (mag)0.00.20.40.60.81.01.21.4 F W H M ( a r cs e c ) HST z~7 candidates Likely foreground galaxydwarf star
Fig. 5.—
FWHM as a function of magnitude. Stars (starsymbols) are identified based on their FWHM and stellarityindex > z ∼ J mag vs FWHM from the FourStarbands (ground-based observations). Bottom panel: J mag vsFWHM (space-based observations). The distinction between starsand compact high-redshift galaxies, based on their FWHM, isunreliable from ground-based observations, compared with HSTobservations. All fours dwarf stars (square symbols) have stellarityindices > J magnitude for the foreground object (triangle) is significantlyfainter compared with the J magnitude due to the fact thatin the raw (0 . ′′ J image, this object is identified as threedifferent objects. In the following subsections, we perform three tests todistinguish stars from compact galaxies: (i) tests com-paring the objects’ FWHM and stellarity indices, (ii)tests of the objects’ spectral features indicative of molec- ular absorption in stellar atmospheres of brown dwarfstars, and (iii) tests of the objects’ surface brightnessprofiles.
Identifying Stars using FWHM and Stellarity Index
Figure 5 shows the FWHM (obtained from SExtractor)for objects as a function of J magnitude as well as theFWHM measured in the (unconvolved) HST J bandto compare ground-based data with the space based ob-servations. The horizontal dashed line in each panel ofFigure 5 indicates the stellar PSF FWHM obtained bystacking ten isolated stars in the respective images.In addition to the FWHM, SExtractor produces a stel-larity index (CLASS STAR), which determines a likeli-hood of an object being either a point or an extendedsource using the self-training of a neural network. A stel-larity index closer to 1 is likely a point source while ex-tended objects like galaxies have stellarity indices closerto 0. Figure 5 (top panel), shows some objects classifiedas stars based on their high ( > .
8) stellarity index (starsymbol). All objects in the initial source list are shownin filled circle, box, or triangle symbols. As can be seenfrom the top panel, most of the objects appear to be re-solved in the FourStar image. However, this appears tobe due to the seeing conditions and the magnitude limitof the ground-based data. The lower panel in Figure 5shows that five of the objects in our sample (zF2014,zF24834, zF30058, zF26680, and zF25035) have FWHMin the HST image much closer to that of a point source.Four of these objects (square symbols in Figure 5) havehigh stellarity indices ( > z ∼ Spectral Fitting to Medium-band Photometry
Due to the availability of medium-band photometry, inaddition to using stellarity indices and FWHM, we canuse a spectral fitting technique whereby observed spec-tral templates are fitted to the medium-band photome-try. This method works very well especially in identify-ing T-dwarf stars because the central wavelengths of theFourStar medium-band filters trace the strong absorp-tion features of T-dwarf stars; this would not be pos-sible using broad-band photometry. van Dokkum et al.(2009) have shown one of the ways to identify T-dwarfstars using medium-band photometry from NEWFIRM(see their Figure 5).To identify dwarf stars, we synthesized medium-bandphotometry from the M, L, and T-dwarf observed spec-tra (Burgasser et al. 2006) and fit these to the FourStarphotometry of all candidates to obtain the minimum χ between the candidate photometry and the dwarf tem-plate spectra. Figure 6 demonstrates the effectivenessof medium-band filters in identifying T-dwarf stars andbest-fit spectral energy distribution (SED) of z ∼ λ ( µ m)-0.20.00.20.40.60.81.0 F l u x ( F ν ) zF25035z~7 galaxy p ( z ) λ ( µ m)0.00.51.01.52.0 F l u x ( f ν ) zF26680Tdwarf T7 templatespectrum λ ( µ m)0.00.20.40.60.8 F l u x ( F ν ) zF13965P =0.99 p ( z ) λ ( µ m)0.00.10.20.30.4 F l u x ( F ν ) zF26836P =0.99 p ( z ) Fig. 6.—
Spectral energy distributions (SEDs) of the three z ∼ z ∼ χ between the FourStar photometry and the expected flux density synthesized from spectral templates of M, L, and T-dwarfstars from Burgasser et al. (2006). The medium-band photometry clearly traces the T-dwarf observed template spectrum (lower leftpanel) and can distinguish between different T-dwarf types (Tilvi et al., in preparation). the medium-band photometry, while the remaining pan-els show z ∼ J FWHM, as dwarf stars.For the compact object (zF25035) however, we did notfind a good fit with any of the M, L, or T-dwarf observedspectra. Thus, this test favors the conclusion that thisobject is a compact z ∼ Surface Brightness Profile
Figure 7 shows the surface brightness ( J ) profilesof five objects: two T-dwarf candidates (with medium-band colors indicative of brown dwarfs and high stellar-ity); two z ∼ > .
9. The twoT-dwarf stars both have stellarity indices > . ′′ z ∼ z ∼ z ∼ z ∼
7, there is a population of very compact (FWHM . > . H &
25 mag), even with
HST -like image quality.In such cases, the near-IR medium-band photometry willallow us to cleanly distinguish between the spectral fea-tures of brown dwarf stars and z ∼ z ∼ Final Sample and New Selection Criteria usingMedium-bands
Using the initial color selection criteria defined inEquation 1, out of the 9 objects in the initial source list,four objects are identified as nearby dwarf stars based ontheir FWHM and spectral template fitting test, one ob-ject is likely a foreground galaxy based on the S/N( > χ image, and the remaining four objectsare identified as z ∼ z ∼ z ∼ / N( J chq ) > . / N( J ) > . / N( χ ) < . , J − J < . ,z ′ − J > . . This color-color selection is indicated in Figure 3 as theshaded region within the dotted polygon. As can beseen, this color-color selection isolates z ∼ z ∼ z ∼ z ∼ LUMINOSITY FUNCTION AT Z ∼ The UV luminosity function provides a direct obser-vational measurement on the number density of star-forming galaxies at a given redshift. We derive thecompleteness-corrected luminosity function φ ( m ) as φ ( m ) = 1 dm N ( m ) V eff , (4)where N ( m ) is the number of detected objects with mag-nitude between m and m + dm , and V eff is the effectivesurvey volume. The Effective Survey Volume
In order to accurately estimate the observed luminos-ity function we must calculate the effective survey vol-ume since this volume depends on both the apparentmagnitude of the selected objects and their redshift dis-tributions. We estimate the effective volume ( V eff ) as R e l a t i v e s u r f a c e b r i gh t ne ss T - d w a r f s t a r sz ~ L B G s Random stellar-like object
Fig. 7.—
Comparison of surface-brightness profiles ( J ) of thecompact z ∼ > z ∼ > . HST image quality, discrimination between galaxies and stars based ontheir stellarity index is unreliable especially at fainter magnitudes. described in Steidel et al. (1999): V eff ( m ) = Z dV ( z ) dz p ( m, z ) dz, (5)where p ( m, z ) is a probability of detecting a galaxy ofapparent magnitude m at redshift z specific to the imagequality, depth, and passbands of our dataset, and dV /dz is the comoving volume in a redshift interval dz .In addition, due to photometric errors, objects canscatter in and out of the color-color selection region(shaded region; Figure 3). In the following sections wedescribe how we estimate the effective volume of our sur-vey. In short, we first compute p ( m, z ) by inserting andthen recovering artificial galaxies in the real images as afunction of apparent magnitude and redshift. We thencalculate dV /dz for each redshift bin. Modeling the Colors of High- z Galaxies
To generate artificial galaxies that are representativeof high-z galaxies, we created a grid of spectral energydistributions (SEDs) of star-forming galaxies using theBruzual & Charlot (2003) models. In these models weused the Chabrier initial mass function, solar metallic-ity, and exponentially rising star-formation history. Fora given redshift, we simulated the effect of dust andIGM using the Calzetti (1997) starburst model and theMeiksin (2006) IGM prescription, respectively. We thenconvolved the resultant SEDs to calculate the bandpassmagnitudes and colors through the z ′ and J filters atvarious redshifts (see Figure 3).To calculate the J - J colors, we followed a differentprocedure instead of using galaxy tracks because themodel colors are nearly constant (see Figure 3) over theredshift range we are currently probing. Here we as-sumed that the flux at a given wavelength can be de-0 Tilvi et al. TABLE 2Photometry of candidate galaxies at z ∼ . ID RA Dec z ′ J J J J H s H l H K s [3 .
5] [4 . > ± ± ± ± ± ± ± ± ± ± > ± ± ± ± ± ± ± ± ± ± > > > ± ± ± ± ± ± ± ± a > ± ± ± ± ± ± ± ± ± ± σ limits in 1 . ′′ g ′ =28.9; r ′ =28.9; i =27.6. a This candidate falls outside the refined color selection. scribed as f λ ∝ λ β where β is the UV continuum slope.For a given J magnitude (calculated from the modelSEDs as described in the previous paragraph), we calcu-late the J magnitude assuming β = − .
0. This valueof β is consistent with z ∼ The Probability Function p(m, z)
We computed the probability function (completenessfunction) by inserting artificial sources based on theirredshifts and colors, and recovering these artificial ob-jects using the same procedure as for the real galaxies.Now that we have magnitudes and colors of high-z galax-ies as a function of redshift, we first have to create anartificial object with some shape, and assign it a magni-tude. To create an artificial galaxy with its shape as closeas a real observed galaxy, we choose a random object ina given magnitude bin from the science image itself (inthis case J ).To insert this object in the z ′ , J , and J images wechoose 500 random positions avoiding any already exist-ing sources ( > σ flux) in the J image. Each object wasassigned a z ′ , J , and J magnitude based on the slope β , and the color obtained from SED models, respectively.We simulate artificial sources in a redshift bin of 0.2 andmagnitude bin dm = 0 . J =24.8 magin Figure 8 (shaded area). To compare this complete-ness function to the broad-band filters, we also overplotthe completeness function for the z ∼ F , Y , and J filters from Oesch et al. (2012)and Bouwens et al. (2012). We clearly see a broadercompleteness function for broad-band filters comparedto our medium-band selection. In contrast, at J =24.8mag, p ( m, z ) is much lower than that derived from muchdeeper HST broad-band surveys. Now that we have p ( m, z ), we calculate the effective volume by integratingthe product of p ( m, z ) and dV /dz (equation 5). Evolution of UV Luminosity Function at z ∼ p ( m , z ) J =24.8 mag Medium-band filters Broadband filters
Fig. 8.—
Probability of recovering artificial galaxies in oursurvey, shown for a source with J =24.8 mag (filled curve).Dashed and dotted lines show HST broad-band completenessfunction from Oesch et al (2009) and Bouwens et al (2012),respectively. The broad-band completeness distribution achieves ahigher peak p ( m, z ) due to the HST survey depths, but is slightlybroader compared to the FourStar completeness distribution. Thenarrower distribution of medium-bands allows for much tighterredshift probability distribution.
The apparent magnitude can be converted to the rest-frame UV magnitude M UV (at λ = 1500˚ A ) using M = J − DM ( z ) − k, (6)where DM is the distance modulus and k is the k -correction between the rest-frame wavelength 1500˚A and J filter. Assuming a negligible k correction, the aboveequation can be rewritten as M = J − D L / pc ) + 2 . z ) + 5 , (7)where D L is the luminosity distance at redshift z .We divided our z ∼ z ∼ M UV = -21.5 mag (Ouchi et al. 2009; Capak et al.2011; Bowler et al. 2012; Hathi et al. 2012). Our resultis consistent with previous conclusions about the evolu-tion of the UV luminosity function: the UV luminosityfunction evolves from z ∼ z ∼ z ∼ -23 -22 -21 -20 -19M -7-6-5-4-3-2 Log N ( M p c - M ag - ) zFourGE Bouwens+07Bouwens+11Ouchi+ 09Bouwens+ 10aCastellano+ 10Bowler+ 12McLure+ 10 zFourGE (+contaminants) z ∼ z ∼ z ∼ ∼ Fig. 9.—
Binned UV luminosity function at z ∼
7. Blue filled circles represent the number densities of our z ∼ dM =1 mag. Error bars on the abscissa show the width of the magnitude bin. Error bars in the ordinateare a posteriori errors, calculated by marginalizing over all Poisson distributions given the observed number of two objects in the brighterbin and one object in the fainter bin. The red colored filled diamonds, triangles, stars, squares, and downward pointing triangles showluminosity functions from Ouchi et al. (2009), Bouwens et al. (2010), Castellano et al. (2010), Bowler et al. (2012), and McLure et al.(2010) respectively. The respective survey area for these studies are 1568 arcmin (Ouchi et al. 2009), 58 arcmin (Bouwens et al. 2010),161 arcmin (Castellano et al. 2010), 3240 arcmin (Bowler et al. 2012), and 4.5 arcmin (McLure et al. 2010). The dashed line anddot-dashed line show the best-fit Schechter functions at z ∼ z ∼ M UV =-21.6 mag), consistent with previous observations.The open circles show our estimates if we include foreground contaminants namely brown-dwarfs. Thus, confusing brown-dwarfs forcompact galaxies, which is possible for ground-based broad-band observations, will overestimate the UV LF. ( M UV ∼ − . ∼
400 Myr the numberdensity of M UV = −
21 to −
22 mag galaxies has increasedby a factor of ∼ z ∼ z ∼
5, which impliesa rapid build-up of the star-formation rates in galaxiesduring this short period.
Comparison to other z ∼ Samples
While there are a few other studies focused on z ∼ HST observations, currently there are onlyfour other ground-based surveys that have identified z ∼ z ∼ J =26.1mag (this filter is similar to the Y filter), but covers asmaller survey area. Castellano et al. (2010) found eight z -dropout galaxies in the GOODS-S field using HAWK-I and FORS2 broad-band (Z, Y, J) observations in asurvey area comparable to ours, however with shallowersurvey depth. Recently, Bowler et al. (2012) found aboutten z ∼ J =25 mag), their survey area is much larger allowing them to probe the brightestand rarest galaxies at this redshift. Finally, Hathi et al.(2012) found two possible LBGs brighter than J =24.5mag from a slightly wider ( ∼
169 arcmin ) but muchshallower ( J ∼
25; 3 σ ) survey in the GOODS-N field.On the other hand, using HST observations, sev-eral studies have found z ∼ M UV & − . HST observations are appropriate in prob-ing the faint end of the UV luminosity function, ground-based surveys have the advantage of probing much largervolume and thus finding brighter and rarer galaxies, al-lowing us to constrain the bright end of the luminosityfunction as can be seen in Figure 9. All the observationsbrighter than M UV = -21.5 mag are currently obtainedfrom ground-based observations.From Figure 9 we can see that at brighter mag-nitudes ( M UV brighter than -22 mag), the best-fitSchechter functions obtained for the HST observa-tions (Bouwens et al. 2007, 2011) prefer a significantevolution from z ∼ z ∼
7. On the other2 Tilvi et al.hand, the ground-based observations (Ouchi et al. 2009;Bowler et al. 2012) suggest a less dramatic evolution atthe bright end, however due to large error bars, strongevolution can not be ruled out. It is possible that inthe absence of a clean method to identify contaminants,namely T-dwarfs (square symbols in Fig. 3), the ob-served z ∼ z ∼ z ∼ PHYSICAL PROPERTIES OF Z ∼ Because near-IR medium-band filters provide a higherspectral resolution compared to the broad-band data,the constraints on physical parameters of these galaxiesis improved (photometric redshifts, stellar masses, star-formation rates (SFRs), ages, extinction, etc.). Here,we constrain the physical properties of the two brightestgalaxies, which have high S/N in the zFourGE bands.It is standard practice to compare galaxy photome-try with stellar population models that vary over somerange of parameter space. Uncertainties on the stellarpopulation parameters (e.g., mass, age, extinction, star-formation history, metallicity) are either derived throughMonte Carlo methods or by marginalizing over someprobability distribution function. Both of these meth-ods assume a prior that the models represent the data.In the sections that follow we compare the physical prop-erties derived using broad-band photometry alone, andthen including the medium-band photometry. For illus-trative purpose, we plot the relevant figures either forthe brightest (zF13965) or the 2nd brightest (zF25035)LBG (whichever yields extreme values) but tabulate thederived properties for the two brightest galaxies in Table2-3.
Stellar Population Synthesis Modelling
To derive the physical properties of our z ∼ ⊙ and assumed a Chabrier initial mass function (IMF) –for a Salpeter IMF the masses would be roughly 0.2 dexlarger, but this would have no effect on the rest-frameUV-optical colors of the galaxies as the slopes of the IMFsat the high-mass end are identical between the two IMFs.In addition, we included nebular emission lines in oursynthetic model spectra. These lines have been shown to zF13965 Fig. 10.—
Comparison of photometric redshift probability dis-tribution of the brightest object (zF13965) using only broad-bandphotometry (
HST +Subaru+IRAC: hashed region), and addingmedium-band filters (FourStar +
HST +Subaru +IRAC bands :filled red region). The photometric redshifts using the FourStarmedium-band filters prefers a lower as well as narrower redshiftrange. affect the derived physical properties of galaxies in the lo-cal universe (e.g., Papaderos et al. 2002; Pustilnik et al.2004) and at high-redshifts (e.g., Schaerer & de Barros2009; Schaerer et al. 2011; Finkelstein et al. 2012, B.Salmon et al 2013 in preparation). We constrained τ to be negative (rising SFH), constant, or τ >
300 Myras the SFHs of these galaxies are expected (on average)to be increasing with time (Papovich et al. 2011). Next,we applied the dust attenuation to our model spectrausing the Calzetti dust law (Calzetti et al. 2000), whichis appropriate to starburst galaxies. The spectra werethen redshifted and attenuated using the IGM attenu-ation from Meiksin (2006). Finally, we computed thebandpass averaged fluxes in each of the filters. The best-fit model was obtained by minimizing the χ between thebandpass averaged fluxes and the model spectra. To un-derstand the effect of different dust attenuation laws onthe SED-derived parameters, we also used the SMC-likedust law (Pei 1992) described in Section 5.3. Photometric Redshift Distribution
Figure 10 shows the photometric redshift prob-ability distribution for the brightest object. Forthe brightest object (zF13965), the photometric red-shifts derived for the
HST +Subaru bands tend to behigher with z phot =8.11 +0 . − . , while inclusion of medium-band photometry lowers the photometric redshift to z phot =7.24 +0 . − . (where the error bars are all 99% con-fidence; Figure 10). This is also true for the 2ndbrightest object (zF25035), with z phot =7.66 +0 . − . and z phot =7.16 +0 . − . when derived using broad-band photom-etry alone, and then adding medium-band filters, respec-tively. The addition of medium-band photometry lowersthe median redshift because of the J filter, which bet-ter constrains the Lyman-break. Surveys using a Y-band We model the star-formation rate as a function of e − t/τ . iscovery of Lyman Break Galaxies at z ∼ Fig. 11.—
Comparison of SED of our brightest candidate(zF13965) using all bands (top panel:
HST +Subaru + FourStarmedium-band +IRAC) and excluding medium-band photometry(middle panel:
HST +Subaru +IRAC bands). The SED modelassumes 0.2Z ⊙ metallicity, a Chabrier IMF, the Bruzual &Charlot 2003 model, and include nebular emission lines. The lowerpanel shows best-fit SED without IRAC bands. Adding FourStarmedium-band filters improves the constraints on the photometricredshift and other physical properties, while exclusion of IRACbands yields poorest constraints. near-IR filter may expect similar photometric redshifts.Including the medium-band filters, the photometric red-shift probability distribution functions favor lower red-shift solutions, with a narrower range of favored redsh-fits. We fit the stellar-population models to the measuredphotometry over a range of redshift spanned by theseprobability distribution functions, and we marginalizeover these properties when determining constraints onstellar population parameters. Best-fit SED
Figure 11 shows the best-fit model spectrum with andwithout the FourStar medium-band filters for our bright-est z ∼ HST andvisible bands while the middle panel shows the best-fitSED without the FourStar medium-band data. The bot-tom panel shows the effects when the IRAC bands areexcluded. In all three cases, we use the photometric red-shifts obtained from EAZY, using respective bands only.The behavior of the best-fit SEDs for the second bright-est z ∼ UV Spectral Slope
The UV continuum light is sensitive to dust absorptionand its slope provides a good tracer of the dust attenu-ation for star-forming galaxies as the two are well corre-lated in UV-luminous, star-forming galaxies from z ∼ ∼ z ∼ β by fitting a power-law to the best-fit SEDin the the rest-frame wavelength range 1250 ≤ λ ≤ β =-1.88 ± . HST bands while adding the medium-band photometry givesa steeper value of β =-2.08 ± . β with a shallower slope, β = − . ± .
05 using
HST broad-band photometry and a steeper β = − . ± . bluer ) UV spectralslopes when the medium-band photometry are included.The UV slope measured for both objects are between β =-1.2 to β =-2.1, within the range seen in other studiesat z ∼
7. Finkelstein et al (2011) found a median valueof β =-2.1 for galaxies with M UV = − .
6. Similarly,Dunlop et al (2012) found an average value of β =-2 forbright z ∼ UV = − . Constraints on Dust Attenuation from SED Fitting
Finkelstein et al. (2010) found the best-fit A(V)=0 − z ∼ HST
WFC3 in the HUDF field, while Labb´e et al. (2010)found A(V)=0 for a stacked sample of z ∼ HST . Figure 12 shows the cumulativeprobability distribution of the dust content E(B-V) forour 2nd brightest object (zF25035) from the SED fit-ting. The shaded regions show 68% and 95% confidenceregions. The top panel shows E(B-V) with broad-band+FourStar photometry, while the bottom panel showsthe probability when including broad-band photometryalone (Table 3).While the best-fit model, with and without themedium-bands, yield similar range of E(B-V), the proba-bility distribution with the medium-bands spans a widerrange with a median E(B-V) = 0.22, slightly lowercompared with the median E(B-V) that excludes themedium-band filters. For the brightest LBG (zF13965),median E(B-V)=0.01 and E(B-V)=0.09 with and with-out FourStar bands respectively. For the reasons de-scribed above, the FourStar bands provide a more robustconstraint on the physical properties of galaxies owing totheir higher spectral resolution.
Stellar Mass and Age
Bowler et al. (2012) found that LBGs at z ∼ J ∼
25 mag) have stellar mass ∼ × M ⊙ .For fainter magnitudes, depending on the range of ex-tinction, the estimated stellar masses range from 3 × TABLE 3Physical properties derived from spectral energy distribution fitting for the two brightest LBG candidates (zF13965 &zF25035).
ID Filters z phot β E(B-V) a Log Stellar Stellar age b Best-fitmass M ⊙ Myr τ (Gyr)zF13965 HST +Subaru+IRAC+FourStar 7.24 +0 . − . -2.08 +0 . − . +0 . − . +0 . − .
251 100
HST +Subaru+IRAC 8.11 +0 . − . -1.88 +0 . − . +0 . − . +0 . − .
159 100
HST +Subaru++no IRAC+FourStar 7.24 +0 . − . -1.17 +0 . − . - 11.3 +0 . − .
630 100zF25035
HST +Subaru+FourStar 7.16 +0 . − . -1.44 +0 . − . +0 . − . +0 . − .
321 100
HST +Subaru 7.66 +0 . − . -1.29 +0 . − . +0 . − . +0 . − .
640 100
HST +Subaru++no IRAC+FourStar 7.16 +0 . − . -1.52 +0 . − . - 10.4 +0 . − .
640 100 a Median value. Errors indicate upper and lower 68% values. b We report the best-fit ages from the spectral-energy distribution modeling only, as the age constraints depend strongly on the assumedstar-formation history. zF25035
Fig. 12.—
Cumulative probability distribution of E(B-V),for the 2 nd brightest candidate (zF25035), using broad-bandphotometry alone (top panel), while the bottom panel includesmedium-band photometry. While the median E(B-V)=0.2 arenearly same for both, the E(B-V) derived using medium-bandfilters allows a wider distribution compared to the the distributionobtained using broad-band photometry alone. - 1 × M ⊙ (Finkelstein et al. 2010; Labb´e et al. 2010).For our brightest object (zF13965), the favored rangeof stellar mass shifts to lower values after including themedium-band filters, with the most likely median massdecreasing by a factor of 1.5. As can be seen in Fig-ure 13, the probability distribution is narrower with me-dian mass 6 . × M ⊙ when the medium-band filtersare included; the absence of the FourStar medium-bandphotometry shifts the masses to higher values (medianmass = 1 × M ⊙ ). On the other hand, for the second zF13965 Fig. 13.—
Cumulative distribution of stellar mass for themost luminous z ∼ × M ⊙ to 6 . × M ⊙ ) because itbetter measures the spectrum of the galaxy. Removing the IRACdata allows for a large range of stellar-population parameters,including older age models, which greatly extends the range ofstellar masses to higher values. brightest object (zF25035), the median mass is higher(2 . × M ⊙ ) when medium-band filters are included.Figure 13 (bottom panel) shows the constraints if weremove the IRAC bands. In this case, the allowed rangeof stellar population parameters that fits the rest-frameiscovery of Lyman Break Galaxies at z ∼ µ m band over the photometric redshift range. Thisboosts the flux in the model, and requires lower stellarmasses.The stellar ages for the two brightest objects rangefrom 250 to 640 Myr, albeit with large uncertaintieswhich depend on the assumed star-formation history. Wecan however estimate the timescale over which the stellarmass doubles. Because the best-fit τ (Table 3) for bothobjects is large (100 Gyr), the timescale is simply stellarmass/SFR. This yields a timescale of about 150 Myr todouble the stellar mass.In summary, using broad-band photometry alone,broadens the photometric redshift range compared to theresult when the medium-band data are included. This isexpected as the medium-band filters more accurately iso-late the location of the Lyman-break. In addition, cur-rently there are no broad-band filters corresponding tothe central wavelengths of some of the FourStar medium-band filters (see Figure 1).While our comparison is based on only two z ∼ z ∼ z ∼ Effect of Lyman- α Line and J Filter on SEDFitting
We further investigated the presence of a Lyman- α line and the effect of the J filter on the derived phys-ical properties of our bright LBGs. To do this, we ob-tained the best-fit SED (section 5.1) by first including noLyman- α emission and then by excluding the J filter.Excluding the J filter or absence of a Lyman- α fluxresults in essentially the same values as that for broad-band only derived parameters. This reinforces the ideathat we need Lyman- α equivalent width measurementsto understand (1) the dust extinction and (2) the stellarmasses, especially for galaxies with Lyman- α emissionline. Lyman- α Equivalent Width
In principle, the Lyman- α equivalent width W pro-vides a proxy to the star-formation rate in the galaxy,and defined as W ≡ F line /F cλ , where F line is the totalline flux (erg s − cm − ) and F cλ is the continuum fluxdensity (erg s − cm − ˚A − ). If our z ∼ α line, it will contribute a flux excessin the J filter, thereby decreasing (i.e., making brighter)the J magnitude (Papovich et al. 2001);∆ m ≃ − . (cid:20) W (1 + z )∆ λ (cid:21) , (8) TABLE 4Estimated rest-frame Lyman- α equivalent widths andLyman- α line fluxes for the two brightest z ∼ LBGcandidates.
ID EW Lyman- α flux(˚A) (10 − erg s − cm − ) Method 1 zF13965 16.5 +15 . − . +1 . − . zF25035 26.2 +16 . − . +0 . − . Method 2 zF13965 -1.0 +12 . − . +1 . − . zF25035 13.5 +14 . − . +0 . − . where W is the rest-frame Lyman- α equivalent widthand ∆ λ is the J filter width.To measure ∆ m , we followed two different methods totake advantage of the presence of J filter. In the 1 st method, we first obtain the best-fit SED using all pho-tometry and including the emission line (see Section 5.1).We then use this best-fit SED, but remove the emissionline, and measure the bandpass J magnitude. This mag-nitude compared with the observed J magnitude yields∆ m . In the 2 nd method, instead of removing the emis-sion line, we obtain the best-fit excluding the J filter andcompare the observed J magnitude with the synthesized J magnitude to get ∆ m . We calculate the rest-frameequivalent widths using Equation 8 for our two brightest z ∼ W , wecalculate the Lyman- α line flux: F line = F cλ W (1 + z ) . (9)Table 4 shows the estimated Lyman- α EWs andline fluxes, obtained from both methods, for the the twobrightest z ∼ α EWsare smaller when the J filter is excluded ( 2 nd method).On the other hand, the Lyman- α fluxes obtained fromthe 1 st method are similar to the recent observationsof spectroscopically confirmed LBGs at z =7.008 and z =7.109, with line fluxes 1.62 × − and 1.21 × − erg s − cm − , respectively (Vanzella et al. 2011). Theseline fluxes are accessible to deep spectroscopy with > Specific Star Formation Rate
The specific star formation rate, sSFR (SFR/ M ⋆ ) pro-vides a way to quantify the evolution of mass build-upgiven a certain SFR. Many observations suggest that thesSFR is nearly constant from z ∼ z ∼ z ∼ Fig. 14.—
The redshift evolution of the specific SFR. Bluefilled circles represent the two brightest galaxies from FourStarobservations. Open diamond, open circles, and open trianglerepresent data from Salim et al. (2007), Stark et al. (2009), andGonz´alez et al. (2010) respectively while filled symbols- greysquares, cyan square, stars, and red squares show observationsfrom Noeske et al. (2007), Daddi et al. (2007), Reddy et al.(2012), and Bouwens et al. (2011) respectively. These galaxieshave stellar masses M ⋆ = (1 − × M ⊙ . The error bars onthe FourStar observations are the uncertainties in the estimatedstellar mass obtained from the best-fit SED. The sSFR remainsnearly constant over 1 Gyr period from z = 8 to z = 4, whichfavors the idea of rising star-formation history with increasingstellar mass (Papovich et al. 2011). parameters derived from the fits to the FourStar medium-band + HST + IRAC data. We derived the SFRs us-ing the best-fit SEDs: for the brightest LBG (zF13965),the median SFR=77.6 +62 − M ⊙ yr − , much smaller com-pared with the 2nd brightest LBG (zF25035) with me-dian SFR=285.7 +346 − M ⊙ yr − (Table 5).Our derived sSFR ∼
13 Gyr − for the two brightest z ∼ z ∼
7, which implies a nearly constant sSFRfrom z ∼ z ∼ × M ⊙ to 2 × M ⊙ ). These sSFRs favors the idea that galax-ies at these epochs have rising star-formation historiessuch that the sSFR is nearly independent of mass at afixed redshift (e.g. Papovich et al. 2011). Models withconstant or declining SFHs would imply decreasing sSFRwith increasing mass at fixed redshift, which is disfa-vored. Comparison Using Calzetti & SMC DustExtinction Laws
To compare the effect of different extinction laws on thederived physical properties of z ∼ z ∼ χ ) using SMC dust ex-tinction law decrease except for the dust extinction (seeTable 5). For the 2nd brightest LBG (zF25035), stel- zF25035 Fig. 15.—
Comparison of dust extinction derived from the best-fit SED using Calzetti and SMC dust extinction law. Here we haveshown the dust extinction of the 2nd brightest LBG (zF25035) asthis object yields significantly different extinctions for SMC andCalzetti dust extinction law, compared with the brightest LBG(zF13965). The SMC-like law yields a median E(B-V)=0.1, a valuemuch smaller compared with median E(B-V)=0.22 obtained usingthe Calzetti dust extinction law.
TABLE 5Comparison of derived physical parameters from thebest-fit SEDs using Calzetti and SMC dust extinctionlaws for the two brightest LBGs. zF13965 zF25035 χ /ν ⊙ ) 9.8 +0 . − . [9.6 +0 . − . ] 10.3 +0 . − . [9.8 +0 . − . ]E(B-V) 0.09 +0 . − . [0.05 +0 . − . ] 0.22 +0 . − . [0.1 +0 . − . ]SFR(M ⊙ yr − ) 77.6 +62 . − . [57.8 +32 . − . ] 285.7 +346 . − . [62.7 +75 . − . ]sSFR(Gyr − ) 13.5 +1 . − . [13.5 +2 . − . ] 13.5 +2 . − . [10.7 +2 . − . ]Values in square brackets are derived using SMC extinction law. lar mass, extinction and χ either decrease or remainunchanged except for the stellar age. A major differ-ence among these two LBGs is that the dust extinctionfor zF25035 is significantly reduced (A UV =0.9) when us-ing the SMC extinction law (see Fig. 15). This smallerextinction value also gives rise to the lower SFR com-pared with the Calzetti dust extinction-derived SFR.This demonstrates that the assumption of the extinc-tion law affects the interpretation of the dust content inhigh-redshift galaxies. In particular, the SMC-type law- which may be more applicable in these distant galaxies(Oesch et al. 2012b) - reduces the implied dust contentcompared to a Calzetti-type law for low-redshift lumi-nous starburst galaxies.iscovery of Lyman Break Galaxies at z ∼ SUMMARY & CONCLUSIONS
We have obtained a sample of z ∼ z ∼ z ∼ in the COSMOS field,which has very deep publically available data. The avail-ability of near-IR medium-band and broad-band photom-etry allowed us to compare the derived physical proper-ties of high-redshift galaxies and how these properties areinfluenced by the choice of filter widths. Our principalfindings are summarized below. Stars vs compact galaxies − While the contamina-tion from nearby T-dwarf stars is a serious concernfor dropout-selected galaxies, especially when usingthe ground-based broad-band photometry, the FourStarmedium-band filters allow us to cleanly distinguish be-tween nearby dwarf stars and high-redshift galaxies thatare relatively compact.One of the three z ∼ z ∼ Bright end of the z ∼ UV luminosty function − Us-ing the number density of z ∼ z ∼ z ∼
5. The numberdensity of bright LBGs ( M UV ∼ -21.5) increases by a fac-tor of 4 from z ∼ z ∼ HST -based studies.
Physical properties of z ∼ galaxies − In general, thepresence of medium-band photometry yield tighter con-straints on the photometric redshifts and improved phys-ical constraints on the stellar population parameters ofthese galaxies.(i) For our two brightest LBGs, the best-fit SEDmodel with broad-band + medium-band photom-etry yields lower photometric redshifts, and muchtighter redshift probability distribution comparedto the z phot ∼ β tends tobe shallower when derived using broad-band pho-tometry alone. For the brightest object, β changesfrom -1.88 to -2.08 when FourStar medium-bandphotometry is included.(ii) The SED modeling prefers a lower dust attenuationwith median E(B-V)=0 .
01 for broad-band photom-etry, while it increases to median E(B-V)=0 . α equivalent widths andline fluxes of the two brightest candidates and findthat the presence of Lyman- α line influences thederived physical properties of z ∼ lower mass galaxies at z ∼
7. This favors the ideal that theSFR increases with increasing stellar mass at thisredshift.While it is early to reach robust conclusions about theeffect of medium-band photometry on the derived prop-erties of z ∼ z ∼ REFERENCESAdelberger, K. L., Steidel, C. C., Shapley, A. E., et al. 2004, ApJ,607, 226Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393Bertin, E., Mellier, Y., Radovich, M., et al. 2002, AstronomicalData Analysis Software and Systems XI, 281, 228Bouwens, R. J., Illingworth, G. D., Blakeslee, J. P., & Franx, M.2006, ApJ, 653, 53Bouwens, 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., Illingworth, G. D., Franx, M., et al. 2009, ApJ,705, 936Bouwens, R. J., Illingworth, G. D., Oesch, P. A., et al. 2010, ApJ,708, L69Bouwens, R. J., Illingworth, G. D., Labbe, I., et al. 2011, Nature,469, 504Bouwens, R. J., Illingworth, G. D., Oesch, P. A., et al. 2012, ApJ,752, L5Bowler, R. A. A., Dunlop, J. S., McLure, R. J., et al. 2012,arXiv:1205.4270Bradley, L. D., Bouwens, R. J., Zitrin, A., et al. 2012, ApJ, 747, 3Brammer, G. B., van Dokkum, P. G., & Coppi, P. 2008, ApJ,686, 1503Brammer, G. B., Whitaker, K. E., van Dokkum, P. G., et al.2011, ApJ, 739, 24Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000Burgarella, D., Buat, V., & Iglesias-P´aramo, J. 2005, MNRAS,360, 1413Burgasser, A. J., Geballe, T. R., Leggett, S. K., Kirkpatrick,J. D., & Golimowski, D. A. 2006, ApJ, 637, 1067Calzetti, D., Kinney, A. L., & Storchi-Bergmann, T. 1994, ApJ,429, 582Calzetti, D. 1997, AJ, 113, 162Calzetti, D., Armus, L., Bohlin, R. C., et al. 2000, ApJ, 533, 682Capak, P., Mobasher, B., Scoville, N. Z., et al. 2011, ApJ, 730, 68Castellano, M., Fontana, A., Boutsia, K., et al. 2010, A&A, 511,A20Daddi, E., Cimatti, A., Renzini, A., et al. 2004, ApJ, 617, 746Daddi, E., Dickinson, M., Morrison, G., et al. 2007, ApJ, 670, 156Dale, D. A., Gil de Paz, A., Gordon, K. D., et al. 2007, ApJ, 655,863Dickinson, M., Stern, D., Giavalisco, M., et al. 2004, ApJ, 600,L99Dickinson, M., Papovich, C., Ferguson, H. C., & Budav´ari, T.2003, ApJ, 587, 25Erb, D. K., Shapley, A. E., Pettini, M., et al. 2006, ApJ, 644, 813Fan, X., Narayanan, V. K., Strauss, M. A., et al. 2002, AJ, 123,1247Fazio, G. G., Hora, J. L., Allen, L. E., et al. 2004, ApJS, 154, 10Finkelstein, S. L., Papovich, C., Giavalisco, M., et al. 2010, ApJ,719, 1250Finkelstein, S. L., Cohen, S. H., Moustakas, J., et al. 2011, ApJ,733, 117Finkelstein, S. L., Papovich, C., Ryan, R. E., Jr., et al. 2012,arXiv:1206.0735Fontana, A., Salimbeni, S., Grazian, A., et al. 2006, A&A, 459,745Giavalisco, M., Ferguson, H. C., Koekemoer, A. M., et al. 2004,ApJ, 600, L93Gonz´alez, V., Labb´e, I., Bouwens, R. J., et al. 2010, ApJ, 713, 115Grazian, A., Castellano, M., Koekemoer, A. M., et al. 2011,A&A, 532, A33Grogin, N. A., Kocevski, D. D., Faber, S. M., et al. 2011, ApJS,197, 35Hathi, N. P., Mobasher, B., Capak, P., Wang, W.-H., & Ferguson,H. C. 2012, ApJ, 757, 43Hibon, P., Cuby, J.-G., Willis, J., et al. 2010, A&A, 515, A97Iye, M., Ota, K., Kashikawa, N., et al. 2006, Nature, 443, 186Kennicutt, R. C., Jr. 1998, ARA&A, 36, 189Koekemoer, A. M., Faber, S. M., Ferguson, H. C., et al. 2011,ApJS, 197, 36Komatsu, E., Smith, K. M., Dunkley, J., et al. 2011, ApJS, 192,18 Kashikawa, N., Shimasaku, K., Malkan, M. A., et al. 2006, ApJ,648, 7Krug, H. B., Veilleux, S., Tilvi, V., et al. 2012, ApJ, 745, 122Labb´e, I., Franx, M., Rudnick, G., et al. 2007, ApJ, 665, 944Labb´e, I., Franx, M., Rudnick, G., et al. 2003, AJ, 125, 1107Labb´e, I., Gonz´alez, V., Bouwens, R. J., et al. 2010, ApJ, 716,L103Laird, E. S., Nandra, K., Adelberger, K. L., Steidel, C. C., &Reddy, N. A. 2005, MNRAS, 359, 47Madau, P. 1995, ApJ, 441, 18Malhotra, S., & Rhoads, J. E. 2002, ApJ, 565, L71McLure, R. J., Dunlop, J. S., Cirasuolo, M., et al. 2010, MNRAS,403, 960Meiksin, A. 2006, MNRAS, 365, 807Meurer, G. R., Heckman, T. M., Leitherer, C., et al. 1995, AJ,110, 2665Meurer, G. R., Heckman, T. M., Lehnert, M. D., Leitherer, C., &Lowenthal, J. 1997, AJ, 114, 54Meurer, G. R., Heckman, T. M., & Calzetti, D. 1999, ApJ, 521, 64Noeske, K. G., Faber, S. M., Weiner, B. J., et al. 2007, ApJ, 660,L47Oesch, P. A., Bouwens, R. J., Illingworth, G. D., et al. 2010, ApJ,709, L16Ono, Y., Ouchi, M., Mobasher, B., et al. 2012, ApJ, 744, 83Ouchi, M., Mobasher, B., Shimasaku, K., et al. 2009, ApJ, 706,1136Oesch, P. A., Bouwens, R. J., Illingworth, G. D., et al. 2012, ApJ,745, 110Oesch, P. A., Labbe, I., Bouwens, R. J., et al. 2012,arXiv:1211.1010Ouchi, M., Shimasaku, K., Okamura, S., et al. 2004, ApJ, 611, 660Overzier, R. A., Shu, X., Zheng, W., et al. 2009, ApJ, 704, 548Ouchi, M., Ono, Y., Egami, E., et al. 2009, ApJ, 696, 1164Papaderos, P., Izotov, Y. I., Thuan, T. X., et al. 2002, A&A, 393,461Papovich, C., Dickinson, M., & Ferguson, H. C. 2001, ApJ, 559,620Papovich, C., Dickinson, M., Ferguson, H. C., et al. 2004, ApJ,600, L111Papovich, C., Finkelstein, S. L., Ferguson, H. C., Lotz, J. M., &Giavalisco, M. 2011, MNRAS, 412, 1123Pei, Y. C. 1992, ApJ, 395, 130Peng, C. Y., Ho, L. C., Impey, C. D., & Rix, H.-W. 2002, AJ,124, 266Pustilnik, S. A., Pramskij, A. G., & Kniazev, A. Y. 2004, A&A,425, 51Reddy, N. A., & Steidel, C. C. 2004, ApJ, 603, L13Reddy, N. A., Erb, D. K., Steidel, C. C., et al. 2005, ApJ, 633, 748Reddy, N. A., Steidel, C. C., Erb, D. K., Shapley, A. E., &Pettini, M. 2006, ApJ, 653, 1004Reddy, N. A., Steidel, C. C., Pettini, M., et al. 2008, ApJS, 175,48Reddy, N. A., & Steidel, C. C. 2009, ApJ, 692, 778Reddy, N. A., Pettini, M., Steidel, C. C., et al. 2012, ApJ, 754, 25Rhoads, J. E., Panagia, N., Windhorst, R. A., et al. 2005, ApJ,621, 582Rieke, G. H., Young, E. T., Engelbracht, C. W., et al. 2004,ApJS, 154, 25Salim, S., Rich, R. M., Charlot, S., et al. 2007, ApJS, 173, 267Shapley, A. E., Steidel, C. C., Adelberger, K. L., et al. 2001, ApJ,562, 95Shapley, A. E., Steidel, C. C., Erb, D. K., et al. 2005, ApJ, 626,698Stanway, E. R., Bunker, A. J., & McMahon, R. G. 2003,MNRAS, 342, 439Shimasaku, K., Ouchi, M., Furusawa, H., et al. 2005, PASJ, 57,447Sawicki, M., & Yee, H. K. C. 1998, AJ, 115, 1329Salvaterra, R., Ferrara, A., & Dayal, P. 2011, MNRAS, 414, 847Schenker, M. A., Stark, D. P., Ellis, R. S., et al. 2012, ApJ, 744,179Schaerer, D., & de Barros, S. 2009, A&A, 502, 423Schaerer, D., de Barros, S., & Stark, D. P. 2011, A&A, 536, A72Schaerer, D., & de Barros, S. 2012, IAU Symposium, 284, 20 iscovery of Lyman Break Galaxies at z ∼7 from the zFourGE Survey 19