Документ взят из кэша поисковой машины. Адрес оригинального документа : http://www.stecf.org/goods/spectroscopy/CDFS_Mastercat/dickinson04.pdf Дата изменения: Thu Sep 16 11:25:44 2004 Дата индексирования: Tue Oct 2 01:16:00 2012 Кодировка: Поисковые слова: п п п п п п п п п п п п п

The Astrophysical Journal, 600:L99­L102, 2004 Januar y 10
2004. The American Astronomical Society. All rights reser ved. Printed in U.S.A.

COLOR-SELECTED GALAXIES AT z 6 IN THE GREAT OBSERVATORIES ORIGINS DEEP SURVEY

1

M. Dickinson,2 D. Stern,3 M. Giavalisco,2 H. C. Ferguson,2 Z. Tsvetanov,4 R. Chornock,5 S. Cristiani,6 S. Dawson,5 A. Dey,7 A. V. Filippenko,5 L. A. Moustakas,2 M. Nonino,6 C. Papovich,8 S. Ravindranath,2 A. Riess,2 P. Rosati,9 H. Spinrad,5 and E. Vanzella9,10
Received 2003 May 28; accepted 2003 October 31; published 2004 January 9

ABSTRACT We report early results on galaxies at z 6 selected from Hubble Space Telescope imaging for the Great Obser vatories Origins Deep Sur vey. Spectroscopy of one object with the Advanced Camera for Sur veys grism and from the Keck and Ver y Large Telescope obser vatories shows a strong continuum break and asymmetric line emission, identified as Lya at z p 5.83 . We find only five spatially extended candidates with signal-to-noise ratios greater than 10, two of which have spectroscopic confirmation. This is much fewer than would be expected if galaxies at z p 6 had the same luminosity function as those at z p 3 . There are many fainter candidates, but we expect substantial contamination from foreground interlopers and spurious detections. Our best estimates favor a z p 6 galaxy population with fainter luminosities, higher space density, and similar comoving ultraviolet emissivity to that at z p 3, but this depends critically on counts at fluxes fainter than those reliably probed by the current data. Subject headings: early universe -- galaxies: evolution -- galaxies: formation -- galaxies: high-redshift On-line material: machine-readable table
1. INTRODUCTION

Broadband color selection, based on ultraviolet (UV) spectral breaks caused by neutral hydrogen, is an efficient technique for identifying galaxies at z p 3 ­4 (Steidel et al. 1996; Madau et al. 1996). At higher redshifts and relatively bright magnitudes, i z colors from the Sloan Digital Sky Sur vey have been used to identify QSOs out to z p 6.4 (Fan et al. 2003). Some galaxies at z 1 5 have also been found in this way, but the required deep imaging and spectroscopy is extremely challenging. A Lyman break galaxy (LBG) with typical (L) UV luminosity at z p 3 (M1700 A p 21.0; Adelberger & Steidel ° 2000) would have m( z) p 26.0 if moved, without evolution, to z p 6 and would be undetected in the i band (hence, an "idropout"). At z 6.5, Lya shifts through the z band, and galaxies are lost to optical sight altogether. One goal of the Great Obser vatories Origins Deep Sur vey (GOODS) is to find and study large numbers of galaxies at 3.5 ! z ! 6.5. Here we report initial results on galaxy candi1 Based on obser vations taken with the NASA/ESA Hubble Space Telescope, which is operated by AURA, Inc., under NASA contract NAS5-26555, the W. M. Keck Obser vatories, and the Ver y Large Telescope (VLT) at Cerro Paranal, Chile, operated by the European Southern Obser vator y, under programs 170.A-0788 and 168.A-0485. 2 Space Telescope Science Institute (STScI), 3700 San Martin Drive, Baltimore, MD 21218. 3 Jet Propulsion Laborator y, California Institute of Technology, Mail Stop 169-506, Pasadena, CA 91109. 4 Department of Physics and Astronomy, Johns Hopkins University, 3400 North Charles Street, Baltimore, MD 21218-2686. 5 Department of Astronomy, University of California, Berkeley, Mail Code 3411, Berkeley, CA 94720. 6 Istituto Nazionale di Astrofisica, Osser vatorio Astronomico di Trieste, via G. B. Tiepolo 11, Trieste I-34131, Italy. 7 National Optical Astronomical Obser vator y, 950 North Cherr y Avenue, Tucson, AZ 85719. 8 Steward Obser vator y, University of Arizona, 933 Cherr y Avenue, Tucson, AZ 85721-0065. 9 European Southern Obser vator y, Karl-Schwarzschild-Strasse 2, D-85748 Garching bei Munchen, Germany. Ё 10 Dipartimento di Astronomia dell'Universita di Padova, Vicolo dell'Osser` vatorio 2, I-35122 Padua, Italy.

dates at z 6, including spectroscopy for one galaxy, CDFS J033240.0 274815 (hereafter SiD2). We use AB magnitudes (AB { 31.4 2.5 log A fn /n JyS) and assume a cosmology with Q tot , QM , QL p 1.0, 0.3, 0.7 and H0 p 70 km s 1 Mpc 1.
2. IMAGING, PHOTOMETRY, AND COLOR SELECTION

The GOODS Treasur y Program covers areas around the Chandra Deep Field­South (CDF-S) and Hubble Deep Field­ North (HDF-N) with mosaics of Advanced Camera for Sur veys (ACS) images. The obser vations, data reduction, and catalogs are described in Giavalisco et al. (2004b). Our present analysis uses three-epoch co-added images for both fields, with 3, 1.5, 1.5, and 3 orbit depth in the F435W, F606W, F775W, and F850LP filters (hereafter B 435 , V606 , i 775, and z 850, respectively). After discarding regions near the image borders or without fourband coverage, the sur vey solid angle is 316 arcmin2. We detect objects in the z 850 images using SExtractor (Bertin & Arnouts 1996), and we measure photometr y through matched apertures in all bands. Here we use z 850 "total" magnitudes (SExtractor MAG_AUTO), and colors measured through isophotal apertures defined in the z 850 image. We estimate the reddest colors expected for ordinar y galaxies with spectral templates (Coleman, Wu, & Weedman 1980) integrated through the ACS bandpasses. The redshifted colors of an elliptical galaxy peak11 at i 775 z 850 1.2 for z 1.1. There is only one "bright" galaxy in the GOODS fields with i 775 z 850 1 1.3 (z 850 p 23.9; i 775 z 850 p 1.32). It is well detected at V606, bright in the near-infrared (IR), and certainly has z K 6. Redder colors may be explained by dust obscuration, high metallicity, strong line emission in the z 850 band, or intergalactic medium (IGM) absorption at high redshift. For the range of UV colors for LBGs at z 3, i 775 z 850 1 1.3 is crossed at z p 5.5­5.7. Cool stars may also be this red, but only a tiny minority of high-latitude stars have i z 1 1.3 (Fan et al.
11 The redshifted elliptical template has redder colors at 1.7 ! z ! 2.3, but the UV spectrum of any galaxy at that redshift is unlikely to resemble that of an old elliptical galaxy at z 0.

L99


L100

DICKINSON ET AL.

Vol. 600

Fig. 1.--Color-magnitude diagram for the combined GOODS fields. The dashed line shows our i-dropout color selection limit. Light blue points are extended objects that do not meet the color and S/N(B435, V606) ! 2 criteria. Triangles mark 1 j lower color limits for objects undetected in i775. Red circles are i-dropout candidates. The larger, filled circles with error bars (1 j) are candidates with S/N(z850) 1 10; stars mark point sources. Two galaxies with spectroscopic redshifts are highlighted, as is a possible T dwarf. The mauve cur ve shows the color-magnitude track for an unevolving, L elliptical galaxy; vertical crosses mark z p 0.5, 1, 1.5 and 2.0. The blue cur ve shows the track for an unevolving Lp3 LBG with average UV colors; vertical crosses mark z z p 2, 3, 4, 5, and 6, while mult crosses mark z p 5.5 and 6.5.

Fig. 2.--Top: ACS grism spectrum of SiD2. Pixels in the spectrum are correlated by the data reduction process and thus have smaller scatter than suggested by the 1 j error bars. The cur ves show unnormalized bandpass functions for the i775 and z850 filters. Bottom: Keck VLT spectrum of SiD2, slightly smoothed. The inset panel shows a magnified (unsmoothed) view of the Lya emission line.

2003), and our ACS imaging provides a robust measure of stellarity for z 850 ! 26.2. The three-epoch ACS mosaics have ver y small misalignments between images, and these misalignments can trigger overrejection in the cores of point sources during cosmic-ray removal in the V606 and i 775 bands (only the B 435 and z 850 images are reduced differently). There is virtually no photometric impact for extended sources, but the i 775 z 850 colors of brighter stars can be biased redward, and we treat them with caution here. We are interested in objects near our detection limits. The signalto-noise ratio (S/N) of a measurement depends on the source flux and size and on the exposure time, which varies with position in our mosaics. The significance of a source is best estimated not from its magnitude but from S/ N(z 850 ) in the detection aperture. Our photometric errors are computed from noise maps that account for interpixel correlations. To verify their reliability, we added artificial objects to the z 850 images (only) and detected them with SExtractor. Background-subtracted counts (Si) were measured through matched apertures for the other bands and compared with the predicted uncertainties (ji) from the noise maps. The distribution of Si /ji is nearly Gaussian with a mode of 0 and an rms of 1, showing that our error estimates are reliable, except for a positive tail due to blending with other objects. Because of this tail, 14% of z 6 galaxies would have S/ N 1 2 in the B 435 or V606 bands.12 We consider this an acceptable loss rate and adopt i 775 z 850 1 1.3 and S/ N(B 435 , V606 ) ! 2 as our i-dropout criteria (Fig. 1).
3. SPECTROSCOPY

G800L obser vation of SN 2002FW (Riess et al. 2004) was obtained on UT 2002 October 1 and included SiD2. It was reduced with the CALACS pipeline and the aXe extraction software. The spectrum (Fig. 2, top) shows a flat continuum ° with a sharp break at l 8300 A. We obser ved SiD2 with the Low Resolution Imaging Spectrometer (LRIS; Oke et al. 1995) on the Keck I telescope on

Supernovae found by GOODS are studied in a Target-ofOpportunity program that, in some cases, obtains low-resolution, slitless spectra with the ACS G800L grism. An 18,840 s
12

On average, this limit corresponds to B435 1 29.1 or V606 1 29.1.

Fig. 3.--Optical-infrared two-color diagram for a portion of the CDF-S (using ISAAC near-IR data). The circles show objects with S/N(z) 1 7, with sizes proportional to their J magnitudes. The filled circles and stars with error bars (1 j) are objects (extended and unresolved) that meet our i-dropout criteria. The filled red circle for SiD2 is circled in black. Open circles show extended objects that do not meet these criteria. The mauve cur ves show modeled colors of ordinar y, low-redshift galaxies, redshifted over 0 ! z ! 2. The blue cur ves show the expected colors of LBGs at 2 ! z ! 7 (z p 6 at the "bend"), spanning the range of UV spectral slopes seen in LBGs at z 3. The crosses on the cur ves mark the same redshifts as in Fig. 1.


No. 2, 2004

GALAXIES AT z 6 FROM GOODS
TABLE 1 The z 6 Galaxy Candidates, Ordered by S/N(z850)a

L101

ID SiD001 SiD002 SiD003 NiD001 SiD004 ...... ...... ...... ...... ......

R.A. (J2000) 03 03 03 12 03 32 32 32 36 32 25.60 40.02 19.07 19.90 33.20

Decl. (J2000) 27 27 27 62 27 5548.6 48 15.0 54 21.9 09 34.2 39 49.2

S/N(z850) 20.45 12.72 11.72 10.55 10.24

m(z850) 24.65 25.20 25.91 25.63 25.38 0.06 0.12 0.13 0.13 0.13

i775 z8 1.63 1.51 1.39

b 50

FWHM(z850) (arcsec) 0.18 0.19 0.27 0.19 0.70

Notes SBM object 3, z p 5.78 (Bunker et al. 2003) SBM object 1, z p 5.83 (this Letter); ISAAC SBM object 7; faint IR (SOFI) Faint IR (SOFI)

0.15 0.23 0.28 12.29 12.21

Note.--Units of right ascension are hours, minutes, and seconds, and units of declination are degrees, arcminutes, and arcseconds. SOFI p Son of ISAAC. Table 1 is published in its entirety in the electronic edition of the Astrophysical Journal. A portion is shown here for guidance regarding its form and content. a The robust sample is that with S/N(z850) 1 10. Objects with S/N ! 10 should be regarded with caution and may include spurious contaminants (see text). b Color limits are reported at 2 j.

UT 2002 October 9 in poor weather conditions and detected ° line emission at 8303 A. Deeper obser vations (7.8 ks) with the 400 line mm 1 grating (R 1000) were obtained on UT 2002 November 8. Obser vations (12 ks) with the FOcal Reducer/ low dispersion Spectrograph 2 (FORS2) on the Yepun telescope (VLT 4) were obtained on UT 2002 December 8 with the 300I grism (R 860). We reduced the data with IRAF following standard procedures and combined the LRIS and FORS2 data with appropriate weighting. The final spectrum (Fig. 2, bottom) shows Lya emission at z p 5.829 with flux 1.6 # 10 17 ergs cm 2 s 1, The blue cutoff, characteristic of high-redshift Lya, and the Lya forest continuum break are clearly evident. The emission line is not seen in the ACS grism spectrum. The ACS exposure time calculator predicts a detection with S/ N 15 for a point source, but extended emission with a continuum will be harder to detect. We obtained an LRIS spectrum of the only extended HDF-N candidate with S/ N(z 850 ) 1 10 (J123619.9 620934) on UT 2003 May 1 but did not successfully measure a redshift.
4. OTHER CANDIDATE z 6 OBJECTS

There are 16 objects with S/ N( z 850 ) 1 10 that meet our selection criteria. Eleven are point sources (3.4% of the stellar objects with 24 ! z 850 ! 26.2; 0 .12 ! F WHM ! 0 .16 vs. 0 .18­

Fig. 4.--Number counts of i-dropout candidates vs. z850 magnitude (a) and vs. the S/N of the z850 detection image (b). The small and large open circles show "raw" counts with and without point sources. The filled red circles are counts after statistical correction for spurious objects. The vertical error bars show N counting statistics. The arrows indicate locations for the five most secure candidates. The lines show predicted counts from simulations with various assumptions about the galaxy LF. The dot-dashed line uses the z p 3 LBG LF, while the dashed line uses the same L but reduces f by a factor of 5. The solid line shows the best-fit Schechter function to the corrected N(z850) points, with L p 0.4Lp3 and f p 3.8fp3. z z

0 . 7 for extended S/N 1 10 candidates), whose i 775 z 850 colors are suspect (§ 2). Deep near-IR imaging from the VLT Infrared Spectrometer And Array Camera (ISAAC) covers 30% of the GOODS/CDF-S (Giavalisco et al. 2004b), including all three CDF-S stellar i-dropout candidates, whose z 850 J colors are redder than expectations for high-redshift objects (Fig. 3). Although some of the eight HDF-N point sources might be at high redshift, we believe that they are probably stars and will not consider them further here. This leaves five extended candidates with S/ N( z 850 ) 1 10, or 0.016 arcmin 2 (Table 1). Stanway, Bunker, & McMahon (2003, hereafter SBM) used the public release version 0.5 GOODS ACS CDF-S data to identify nine i-dropout candidates. Three are in our sample, and two have been confirmed spectroscopically (Bunker et al. 2003; this Letter). SBM object 5 is unresolved, with the reddest i 775 z 850 color (12.7 at 2 j) of any GOODS object. Its exceptionally blue near-IR colors (J H p 0.3, H K p 0.5, AB) suggest that it may be a T dwarf (see also SBM). Another point source, SBM object 4, was obser ved in the GOODS spectroscopic program and is a cool star (approximately L0 V). SBM objects 2, 4, 8, and 9 have S/ N 1 2 in V606 and/or B 435, are fairly bright in near-IR images, and are thus unlikely to be at z 6. SBM object 6 falls outside the area analyzed here, with shallow Viz (and no B 435) data. In summar y, three (perhaps four) of the nine SBM objects are good z 6 candidates. Data artifacts (space junk trails, reflection ghosts, diffraction spikes, residual cosmic rays) produce spurious z 850 detections without shorter wavelength counterparts that mimic i-dropouts. We have removed most of these by visual inspection. This is generally easy at S/ N 1 10, but this corresponds to Az 850 S 25.3, which is fairly bright for galaxies at z 6. Our catalogs push deeper; ex post facto, we truncate them at S/ N 5 and reject sources that are too small or sharp to be real. However, some spurious sources probably remain. As one check, we masked areas with objects and detected "negative sources." We find 57 that qualify as i-dropouts. All have S/N ! 8, and 75% have 5 ! S/ N ! 6. The vast majority of real, faint galaxies have i 775 z 850 ! 1.3 and z K 6, but measurement errors may scatter a small fraction to redder colors. We estimate this contamination using brighter objects [S/ N( z 850 ) 1 20]. We randomly assign their colors to fainter objects and then perturb the fluxes using the error distributions quantified in § 2. Only approximately two foreground interlopers with S/ N 1 10 would (barely) meet the i-dropout criteria, while 49 objects with 5 ! S/ N ! 10 could do so. Altogether, possible contaminants represent less than 0.7% of the 16,000 GOODS sources with 5 ! S/ N( z 850 ) ! 10 but may contribute 45% of the faint z 6 candidates. After subtracting the expected contamination, we estimate that there are 145 candidates with S/ N 1 5 (0.46 arcmin 2 ), more than


L102

DICKINSON ET AL.

Vol. 600

50% of which have 5 ! S/ N ! 6. We stress that many of the candidates with S/ N( z 850 ) ! 10 in Table 1 may be spurious or interlopers; this list should be used with caution. There are four extended candidates with S/ N 1 7 in the portion of the CDF-S with deep ISAAC imaging (Fig. 3). One has red z 850 J and bright IR magnitudes and is unlikely to be at z 6. Two or perhaps three candidates, including SiD2, are ver y faint in the near-IR (24.7 ! JAB ! 24.9), with colors expected for galaxies at 5.5 ! z ! 6.
5. DISCUSSION

The i 775 z 850 1 1.3 limit sets a lower redshift bound for idropouts, while IGM suppression in z 850 makes the upper bound, and hence the sampling volume, a strong function of luminosity. We use simulations (Giavalisco et al. 2004a) to predict the number counts of candidates, including photometric biases. We generated artificial galaxies with a mixture of disk and bulge surface brightness profiles, ellipticities, and orientations. Their sizes were drawn from a lognormal distribution tuned to reproduce measurements at z 3­5 (Ferguson et al. 2004). Their spectra have a distribution of UV spectral slopes that matches the obser ved colors of LBGs at z 3 (Adelberger & Steidel 2000). We distributed the galaxies in redshift, modulated their spectra by IGM opacity (Madau 1995), convolved them with ACS point-spread functions, added them to the GOODS images at various magnitudes, and recovered them with SExtractor. Figure 4 compares the number of candidates with simulations for various assumptions about the UV luminosity function (LF), which we model as a Schechter (1976) distribution with a faintend slope a p 1.6, as measured for LBGs at z p 3 (Adelberger & Steidel 2000). The number of bright galaxies is smaller at z 6 than at z p 3 (see also SBM and Lehnert & Bremer 2003 for z 5.3 LBGs from ground-based data). The z p 3 LF predicts 30 galaxies with S/ N(z 850 ) 1 10 versus S/ N(z 850 ) 1 5 obser ved, and 26 with z 850 ! 25 versus z 850 7 obser ved,13 and is excluded with a high degree of confidence (P ! 10 8). A change in the number of bright objects does not require comparable evolution in the total luminosity density of the population since the number of bright sources is exponentially sensitive to the value of L. Schechter functions fitted to the counts in Figure 4a favor fainter L and higher f compared with their values at z p 3. Integrating acceptable fits for M1700 A ! 19.4 ° (0.2 Lzp3), the UV emissivity is similar to that at z p 3 0.2 9 [r( L zp6 )/r( L zp3 ) p 0.77 0.23, 95.4% confidence]. However, the fitted L -values, and hence most of the inferred rL, are at z 850 1 26.4, where the current data are uncertain. The fit is strongly driven by objects with 5 ! S/ N ! 6.3 (Fig. 4b). A model with Lzp6 p Lzp3 and 5 times smaller f (and rL) is consistent
13 Out of seven candidates with z850 ! 25, only one has S/N(z850) 1 10. The others may be real, but contamination may be substantial.

with the data at bright magnitudes and high S/Ns but drastically underpredicts the counts at low S/N and z 850 1 25.5. Fits excluding the lowest S/N bin leave rL essentially unconstrained. Two other studies have analyzed i-dropouts from somewhat deeper ACS images. Yan, Windhorst, & Cohen (2003) found 2.3 candidates per arcmin2 with S/ N( z 850 ) 1 7.2 in an ACS image with a 1.5 times longer exposure in z 850 than the threeepoch GOODS data but a 4.9 times longer exposure in i 775, and thus more robust color discrimination against interlopers. Their density is 10 times larger than ours at the same S/N threshold. They estimate that their catalogs are 100% complete for z 850 27.5, whereas ours are only 50% complete for point sources at z 850 p 26.7 (Giavalisco et al. 2004b). Yan et al. (2003) may have underestimated their source fluxes or spurious detection rate, but it is notable that they also find very few bright candidates (none with z 850 ! 26.8). Bouwens et al. (2003) identified 0.5 candidates per arcmin2 with z 850 ! 27.3 from imaging (5­20 orbits in z 850) covering 46 arcmin2. They also find few bright candidates (only one with z 850 ! 25.5) and estimate r(L zp6 )/r(L zp3 ) p 0.6 0.2. In summar y, we have identified five spatially extended, S/ N( z 850 ) 1 10 candidates for galaxies at z 6 in early GOODS ACS imaging. Two have confirmed redshifts z 5.8. There are many fainter candidates, but we estimate that 45% may be spurious detections or foreground interlopers. The number of robust candidates is smaller than is predicted if the LF were the same as that at z 3. Our best estimates find fainter L, larger f, and moderately smaller rL compared with z p 3, but this strongly depends on the number of objects at z 850 1 26, which is as yet poorly measured. Constant L with smaller f and rL are consistent with the bright counts but greatly underpredict the number of faint sources. The measurements do not require (or robustly exclude) a dramatic change in rL from z 6 to 3, especially if L is evolving with redshift. Giavalisco et al. (2004a) find only a modest change ( 30% 10%) in the luminosity density from z p 3 to z 5 where the GOODS LBG sample is much better characterized. Our best estimates are consistent with an extrapolation of those results to z p 6, but deeper data are needed for a robust measurement. The final GOODS images are deeper, with fewer contaminating artifacts, and other forthcoming data (e.g., the ACS Ultra Deep Field) will provide tighter constraints on the galaxy population at these highest optically accessible redshifts. We thank other members of the GOODS team and the staffs at STScI, ESO, and the Keck Obser vator y, who made this project possible. Support was provided by NASA through grant GO09583.01-96A from STScI, which is operated by AURA, Inc., under NASA contract NAS5-26555. Work by L. A. M. and D. S. was supported by NASA through the SIRTF Legacy Science Program, through contract 1224666, issued by JPL, California Institute of Technology, under NASA contract 1407.

REFERENCES Adelberger, K. L., & Steidel, C. C. 2000, ApJ, 544, 218 Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393 Bouwens, R. J., et al. 2003, ApJ, 595, 589 Bunker, A. J., Stanway, E. R., Ellis, R. S., McMahon, R. G., & McCarthy, P. J. 2003, MNRAS, 342, L47 Coleman, G. D., Wu, C.-C., & Weedman, D. W. 1980, ApJS, 43, 393 Ferguson, H. C., et al. 2004, ApJ, 600, L107 Fan, X., et al. 2003, AJ, 125, 1649 Giavalisco, M., et al. 2004a, ApJ, 600, L103 ------. 2004b, ApJ, 600, L93 Lehnert, M. D., & Bremer, M. 2003, ApJ, 593, 630 Madau, P. 1995, ApJ, 441, 18 Madau, P., Ferguson, H. C., Dickinson, M., Giavalisco, M., Steidel, C. C., & Fruchter, A. 1996, MNRAS, 283, 1388 Oke, J. B., et al. 1995, PASP, 107, 375 Riess, A., et al. 2004, ApJ, 600, L163 Schechter, P. 1976, ApJ, 203, 297 Stanway, E. R., Bunker, A. J., & McMahon, R. G. 2003, MNRAS, 342, 439 (SBM) Steidel, C. C., Giavalisco, M., Pettini, M., Dickinson, M., & Adelberger, K. L. 1996, ApJ, 462, L17 Yan, H., Windhorst, R. A., & Cohen, S. H. 2003, ApJ, 585, L93