A photometric redshift survey of SMGs

The LABOCA survey of the Extended Chandra Deep Field South: A photometric redshift survey of submillimetre galaxies


We derive photometric redshifts from 17-band optical to mid-infrared photometry of 74 robust radio, m and Spitzer IRAC counterparts to 68 of the 126 submillimetre galaxies (SMGs) selected at m by LABOCA observations in the Extended Chandra Deep Field South (ECDFS). We test the photometric redshifts of the SMGs against the extensive archival spectroscopy in the ECDFS. The median photometric redshift of identified SMGs is , the interquartile range is –2.7 and we identify 10 () high-redshift () SMGs. We derive a simple redshift estimator for SMGs based on the IRAC 3.6 and 8 m fluxes which is accurate to for SMGs at . A statistical analysis of sources around unidentified SMGs identifies a population of likely counterparts with a redshift distribution peaking at , which likely comprises of the unidentified SMGs. This confirms that the bulk of the undetected SMGs are co-eval with those detected in the radio/mid-infrared. We conclude that at most of all the SMGs are below the flux limits of our IRAC observations and lie at and hence around of all SMGs have . We estimate that the full  mJy SMG population has a median redshift of . In contrast to previous suggestions we find no significant correlation between submillimetre flux and redshift. The median stellar mass of the SMGs derived from SED fitting is and the interquartile range is (4.7–14), although we caution that the uncertainty in the star-formation histories results in a factor of uncertainty in these stellar masses. Using a single temperature modified blackbody fit with the median characteristic dust temperature of SMGs is  K and the interquartile range is –43.3 K. The infrared luminosity function shows that SMGs at –3 typically have higher far-infrared luminosities and luminosity density than those at –2. This is mirrored in the evolution of the star-formation rate density (SFRD) for SMGs which peaks at . The maximum contribution of bright SMGs to the global SFRD ( for SMGs with  mJy; for SMGs with  mJy) also occurs at .

submillimetre – galaxies: starburst – galaxies: evolution – galaxies: high-redshift

1 Introduction

Observations in the millimetre and submillimetre wavebands provide a uniquely powerful route to survey the distant Universe for intense dust-obscured starbursts (Blain & Longair, 1993). This is due to the negative K-correction arising from the shape of the spectral energy distribution (SED) of the dust emission in the rest-frame far-infrared, which results in an almost constant apparent flux for sources with a fixed luminosity at –8.

Over the past decade, a series of ever larger surveys in the submillimetre and millimetre wavebands have mapped out a population of sources at mJy-flux limits with a surprisingly high surface density (e.g. Smail et al., 1997; Barger et al., 1998; Hughes et al., 1998; Eales et al., 1999; Bertoldi et al., 2000, 2007; Coppin et al., 2006; Knudsen et al., 2008; Weiß et al., 2009; Austermann et al., 2010). The mJy fluxes of these sources imply far-infrared luminosities of  L, if the sources are at cosmological distances, , classing them as ultraluminous infrared galaxies (ULIRGs; Sanders & Mirabel, 1996). Their high surface density is far in excess of that expected from a “no evolution” model, suggesting very strong evolution of the population: (Smail et al., 1997; Blain et al., 1999). If this results from strong luminosity evolution of starburst galaxies (as opposed to obscured AGN; Alexander et al. 2005 then a significant fraction of the massive star formation (and metal production) at high redshift may be occurring in this population.

To confirm this evolution and understand the physical processes driving it requires redshifts for the submillimetre galaxies (SMGs). Due to the coarse spatial resolution of the submillimetre and millimetre maps from which the SMGs can be identified, combined with their optical faintness (in part due to their high dust obscuration), it has proved challenging to measure their spectroscopic redshift distribution (e.g. Barger et al., 1999; Chapman et al., 2003a, 2005).

In fact, spectroscopic redshifts are not necessary to map the broad evolution of the SMG population and cruder photometric redshifts can be sufficient, if they are shown to be reliable. Various photometric redshift techniques have therefore been applied in an attempt to trace the evolution of SMGs, using their optical/near-/mid-infrared or far-infrared/radio SEDs (e.g. Carilli & Yun, 1999; Smail et al., 2000; Ivison et al., 2004; Pope et al., 2005, 2006; Ivison et al., 2007; Aretxaga et al., 2007; Clements et al., 2008; Dye et al., 2008; Biggs et al., 2010).

Both spectroscopic and photometric analyses suggest that the bulk of the SMG population lies at , with an apparent peak at for the subset of SMGs which can be located through their Jy radio emission (Chapman et al., 2005). Nevertheless, there are significant disagreements between the different studies (see e.g. Chapman et al., 2005; Clements et al., 2008; Dye et al., 2008), which may arise in part due to differing levels and types of incompleteness in the identifications and biases in the redshift measurements. The most serious of these is the incompleteness due to challenges in reliably locating the correct SMG counterpart. They are typically identified through statistical arguments and physical correlations based on radio, mid- or near-infrared emission (e.g. Ivison et al., 1998, 2000, 2005; Smail et al., 1999; Pope et al., 2005; Bertoldi et al., 2007; Hainline et al., 2009; Biggs et al., 2010), but these locate only –80% of SMGs. The expectation is that the SMGs whose counterparts are missed could potentially include the highest redshift (and thus the faintest in the radio and mid-infrared) examples, biasing the derived evolution (Ivison et al., 2005). Attempts to address this incompleteness through time-intensive submillimetre interferometry have located a small fraction of previously unidentified SMGs (e.g. Dannerbauer et al., 2002, 2008; Younger et al., 2007, 2009; Wang et al., 2007) but the nature and redshifts of this unidentified subset of SMGs remains a critical issue for studies of the population as a whole.

In this paper we use optical, near- and mid-infrared photometry to study SMGs detected in the Extended Chandra Deep Field South (ECDFS) by the Large APEX BOlometer CAmera (LABOCA; Siringo et al., 2009) on the Atacama Pathfinder EXperiment (APEX; Güsten et al., 2006) 12-m telescope in the LABOCA ECDFS Submillimetre Survey (LESS; Weiß et al., 2009). LESS mapped the full ECDFS at -m to a noise level of mJy beam, for a beam with angular resolution of . 126 SMGs were detected at significance (equivalent to a false-detection rate of , Weiß et al. 2009) and robust or tentative radio, m or IRAC mid-infrared counterparts are identified to 93 (71 robust and 22 tentative) SMGs (Biggs et al., 2010). Here we determine photometric redshifts for the 91 (68 robust and 23 tentative) of these SMGs with detectable optical and near-infrared counterparts in new and archival multiband photometry of the ECDFS (described in §2). LESS is an ideal survey for this purpose because of its panoramic, deep and uniform submillimetre coverage and extensive auxiliary data, including spectroscopy of sufficient SMG counterparts to adequately test our photometric redshifts. In addition, the large size of the survey allows us to statistically measure the redshift distribution of the SMGs that we are unable to locate directly, in order to test if their redshift distribution differs significantly from the identified population.

The plan of the paper is as follows: in §2 we derive multiband photometry from new and archival observations; while in §3 we describe our the photometric redshift estimates and tests of their reliability. The photometric redshifts, SED fits, absolute -band magnitudes, infrared luminosities, dust temperatures and star-formation rates of SMGs are presented and discussed in §4 and we present our conclusions in §5. Throughout this paper we use deboosted submillimetre fluxes from Weiß et al. (2009), J2000 coordinates and CDM cosmology with , and . All photometry is on the AB magnitude system, in which , unless otherwise stated.

2 Observations and Data Reduction

Filter Detection limit Reference
(m) (; mag)
MUSYC WFI 0.35 26.9 Taylor et al. (2009b)
MUSYC WFI 0.37 25.4 Taylor et al. (2009b)
VIMOS 0.38  28.43 Nonino et al. (2009)
MUSYC WFI 0.46 26.8 Taylor et al. (2009b)
MUSYC WFI 0.54 26.7 Taylor et al. (2009b)
MUSYC WFI 0.66 25.8 Taylor et al. (2009b)
MUSYC WFI 0.87 24.9 Taylor et al. (2009b)
MUSYC Mosaic II 0.91 24.5 Taylor et al. (2009b)
MUSYC ISPI 1.25 23.6 Taylor et al. (2009b)
HAWK-I 1.26 25.7 Zibetti et al. (in prep.)
MUSYC SofI 1.66 23.0 Taylor et al. (2009b)
MUSYC ISPI 2.13 22.7 Taylor et al. (2009b)
HAWK-I 2.15 25.3 Zibetti et al. (in prep.)
SIMPLE IRAC m 3.58 24.6 Damen et al. (2010)
SIMPLE IRAC m 4.53 24.4 Damen et al. (2010)
SIMPLE IRAC m 5.79 22.8 Damen et al. (2010)
SIMPLE IRAC m 8.05 23.5 Damen et al. (2010)
Table 1: Summary of photometry employed in this paper.

In this paper we consider the optical and infrared counterparts to 126 SMGs in the ECDFS detected at (Weiß et al., 2009) and identified by VLA radio, MIPS (Rieke et al., 2004) 24 m and IRAC (Fazio et al., 2004) emission (Biggs et al., 2010). Following convention and Biggs et al. (2010) we consider robust counterparts as those with a corrected Poissonian probability of being unassociated with the submillimetre source (p; Downes et al., 1986) of in one or more of the radio, 24m, or IRAC datasets, or – 0.10 in two or more; tentative counterparts are those with – 0.10 in only one of the three bands.

Six of the SMGs have multiple robust counterparts; of these four SMGs (LESS 2, LESS 27, LESS 49 and LESS 74) have two counterparts with photometric redshifts (§3.1) consistent with them being at the same distance and possibly physically associated. Two SMGs (LESS 10 and LESS 49) each have two robust counterparts with photometric redshifts and SEDs that suggest they are not physically associated. In these cases, from the information currently available, it is not possible to determine which of the two counterparts is the source of the submillimetre flux, or whether the LABOCA detection is a blend of emission from two galaxies. To avoid bias we have included all of the multiple counterparts in our analysis, but we note that their small number means that their inclusion does not significantly affect our results.

2.1 Optical and infrared photometry

SMGs typically have faint optical and near-infrared counterparts (e.g. Ivison et al., 2002) so we require deep photometry for accurate photometric redshift estimates. The ECDFS was chosen for this survey because it is an exceptionally well-studied field, and as such we are able to utilise data from extensive archival imaging and spectroscopic surveys. For completeness and uniformity we only consider surveys that cover a large fraction of the ECDFS rather than the smaller and deeper central CDFS region. Therefore, we utilise the MUltiwavelength Survey by Yale-Chile (MUSYC; Gawiser et al., 2006) near-infrared survey for to -band imaging (Taylor et al., 2009b), and the Spitzer IRAC/MUSYC Public Legacy in ECDFS (SIMPLE) imaging for Spitzer IRAC data (Damen et al., 2010). We also include -band data from the deep GOODS/VIMOS imaging survey of the CDFS (Nonino et al., 2009); although this covers only of LESS SMGs it is valuable for galaxies that are undetected at short wavelengths in the shallower MUSYC survey.

In addition, we have carried out deep near-infrared observations in the and bands with HAWK-I (Pirard et al., 2004; Casali et al., 2006; Kissler-Patig et al., 2008) at the ESO-VLT (ID: 082.A-0890, P.I. N. Padilla). The ECDFS was covered with a mosaic of 16 pointings in each band, with a total exposure time of 0.75 and 1.1 hours per pointing, in the and bands respectively. The median seeing is 0.7 in and 0.5 in . Data reduction has been performed using an upgraded version of the official ESO pipeline for HAWK-I, customized calibration has been obtained from observations of photometric standard stars. More details and catalogues will be published in Zibetti et al. (in preparation).

For accurate photometric redshifts we require consistent photometry in apertures which sample the same emitting area in each of the 17 filters. For consistency between surveys and to ensure that all detected SMG counterparts are included in this study we extract photometry from the available survey imaging rather than relying on the catalogued sources. SMGs are typically brighter at mid-infrared than optical wavelengths due to their high redshifts and extreme dust obscuration. Therefore, we use SExtractor (Bertin & Arnouts, 1996) to create a source list from a combined image of the four IRAC channels, which is weighted such that a given magnitude receives equal contributions from all of the input images. Real sources are required to have at least 4 contiguous pixels with fluxes at least 1.5 times the background noise. In addition, we visually check the area within of each LABOCA source to ensure that no potential SMG counterparts are missed.

We next use apphot in iraf to measure the fluxes in diameter apertures for each of the four IRAC bands. We then cut the catalogues to based on the background noise, and finally apply aperture corrections as derived by the SWIRE team (Surace et al., 2005) to obtain total source magnitudes. The resolution in the - to -band imaging is better than IRAC (FWHM compared to for IRAC) and so we convolve each - to -band image to match the seeing of the worst band. We next use apphot to measure photometry in diameter apertures at the positions of the IRAC-selected sources. In all cases, we only allow apphot to re-centroid the aperture if centroiding does not cause the extraction region to be moved to a nearby source, as flagged by iraf’s CIER parameter when the centroid shift is . We have not performed any deblending of the photometry but examination of the images suggests fewer than of the SMG counterparts are affected. We note here that the photometric extraction process is not restricted to SMGs and yields photometry (which allows us to calculate consistent photometric redshifts) for IRAC-selected sources throughout the ECDFS.

Finally, to ensure equivalent photometry between the IRAC and optical-to-near-infrared filters we create simulated IRAC images of point sources. Using these images we calculate that the correction between the measured IRAC total magnitudes and the photometry extracted from diameter apertures on seeing images is magnitudes, and as such we do not apply any systematic corrections to the IRAC magnitudes at this stage. In §3.1 we calibrate the photometry prior to photometric redshift calculation in a process which corrects for small residual offsets. A summary of our photometry is presented in Table 1.

The median number of photometric filters per SMG counterpart is 15 and we require detections in at least three photometric filters in order to calculate photometric redshifts. Our final sample therefore contains 74 optical counterparts to the 68 robustly identified SMGs with sufficient detectable optical-to-infrared emission. In §3.3 we show that the exclusion of the additional 23, tentatively identified, counterparts does not bias our results.

2.2 Spectroscopy

We employ spectroscopy of the ECDFS to calibrate our photometry with the SED templates (§3.1) and test our photometric redshifts (§3.2). We have examined the spectroscopic redshift catalogues from many archival surveys (Cristiani et al., 2000; Croom et al., 2001; Bunker et al., 2003; Dickinson et al., 2004; Le Fèvre et al., 2004; Stanway et al., 2004; Strolger et al., 2004; Szokoly et al., 2004; van der Wel et al., 2004; Zheng et al., 2004; Daddi et al., 2005; Doherty et al., 2005; Mignoli et al., 2005; Grazian et al., 2006; Ravikumar et al., 2007; Kriek et al., 2008; Vanzella et al., 2008; Popesso et al., 2009; Treister et al., 2009; Balestra et al., 2010, Koposov et al. in prep.) and also our own on-going spectroscopic survey of LESS sources with the VLT (PID: 183.A-0666, P.I. I. Smail), which will be published in full in Danielson et al. (in prep.).

3 Analysis

3.1 Photometric redshift calculation

We use Hyperz4 (Bolzonella et al., 2000) to calculate the photometric redshifts of counterparts to LESS SMGs (Biggs et al., 2010). Hyperz compares a model SED to observed magnitudes and computes for each combination of spectral type, age, reddening and redshift and thus statistically determines the most likely redshift of the galaxy. We use the elliptical (E), Sb, single burst (Burst) and constant star-formation (Im) spectral templates from Bruzual & Charlot (1993) which are provided with Hyperz, and allow reddening (Calzetti et al., 2000) of –5 in steps of 0.2. This combination of templates and was shown by Wardlow et al. (2010) to be sufficient for calculating photometric redshifts of SMGs. Redshifts between 0 and 7 are considered and galaxy ages are required to be less than the age of the Universe at the appropriate redshift. In §3.5 we show that the Hyperz-derived galaxy ages cannot be reliably determined, but we note here that the requirement for SMGs to be younger than the Universe does not significantly affect the derived redshifts. Galaxies are assigned zero flux in any filter in which they are not detected, with an error equal to the detection limit of that filter. To ensure that galaxies at –3 do not have their redshifts systematically underestimated we have modified the handling of the Lyman- forest in Hyperz, such that intragalactic absorption in the models is increased and three different levels of absorption are considered in the fitting process. The reliability of the calculated redshifts and the validity of these settings is tested in full in §3.2.

We test for small systematic discrepancies between the photometry and model SEDs prior to using Hyperz to calculate photometric redshifts of SMGs. This is done by running Hyperz on 1796 galaxies and AGN with spectroscopic redshifts in the ECDFS and requiring a fit at the observed redshift. We then compare the model and measured magnitudes for each galaxy, and iteratively adjust the zeropoints of the filters with the largest systematic offsets. This yields significant offsets for the following filters: VIMOS ( mag), MUSYC ( mag), ( mag), ( mag), ( mag), ( mag), HAWK-I ( mag), IRAC 3.6 ( mag) and IRAC 8.0 m ( mag). The typical uncertainties in these corrections are and the remaining eight filters have no significant corrections.

The calibrated photometry of the robust LESS SMG counterparts is listed in Table 5 and in Table 2 we provide the coordinates, photometric redshifts, absolute rest-frame -band magnitudes, far-infrared luminosities and characteristic dust temperatures of the SMGs (§4.6). We also provide the reduced of the best fit SED at the derived photometric redshift and the number of filters in which the SMG was detected and undetected (but observed). We caution that the reduced for galaxies with only a few photometric detections is typically low () but the error on the photometric redshift is typically large, since there are only weak limits on the SED from the photometry. Therefore, the values of the reduced should be considered in conjunction with the number of photometric detections when considering the reliability of the photometric redshifts.

The median reduced of the SMG counterparts is 2.3 (2.1 if only the galaxies with reduced are considered). This suggests that our photometric errors are slightly overestimated and lead to apparently overly-precise photometric redshift limits. Indeed, we find that the Hyperz 99% confidence intervals more reliably represent the errors, yielding of SMGs with photometric redshifts consistent with the spectroscopic redshifts. Therefore, throughout this paper we use the Hyperz 99% confidence intervals on the photometric redshift estimates to represent the 1- uncertainty. Of the 74 SMG counterparts examined there are eight with poor fits of the SED to the photometry (indicated with reduced ). Of these, one (LESS 39) is blended in the optical imaging and two (LESS 66 and LESS 81) lie in stellar halos. LESS 66 is also likely to be a quasar, as is LESS 96, and another four SMGs with reduced (LESS 19, LESS 57, LESS 75 and LESS 111) have excess 8 m flux compared to the best-fit SED, which is indicative of an AGN component (see §3.4 for a full discussion). Since we did not include any quasar or AGN templates in the fitting procedure it is unsurprising that these sources are not well represented by the employed SEDs. We note here that, as we show in §3.4, the exclusion of AGN templates does not bias our photometric redshift estimates.

SMG Short name RA Dec Filters ID type
(mag) () (K)
LESSJ033302.5-275643 LESS 2a 03330255 56447 16 [1] M
LESSJ033302.5-275643 LESS 2b 03330268 56426 8 [9] R
LESSJ033321.5-275520 LESS 3 03332150 55201 5 [10] M
LESSJ033257.1-280102 LESS 6 03325715 01015 16 [1] RM
LESSJ033315.6-274523 LESS 7 03331541 45240 16 [1] RM
LESSJ033211.3-275210 LESS 9 03321135 52129 6 [9] RM
LESSJ033219.0-275219 LESS 10a 03321904 52143 12 [5] R
LESSJ033219.0-275219 LESS 10b 03321930 52191 15 [2] R
LESSJ033213.6-275602 LESS 11 03321384 55598 7 [9] R
LESSJ033248.1-275414 LESS 12 03324796 54161 6 [11] RM
LESSJ033152.6-280320 LESS 14 03315247 03186 7 [9] RM
LESSJ033333.4-275930 LESS 15 03333335 59294 4 [8] M
LESSJ033218.9-273738 LESS 16 03321870 37435 17 [0] R
LESSJ033207.6-275123 LESS 17 03320726 51201 17 [0] RM
LESSJ033205.1-274652 LESS 18 03320487 46474 16 [1] RM
LESSJ033208.1-275818 LESS 19 03320823 58137 10 [6] RI
LESSJ033316.6-280018 LESS 20 03331677 00158 9 [7] RM
LESSJ033147.0-273243 LESS 22 03314690 32388 6 [4] RM
LESSJ033336.8-274401 LESS 24 03333697 43581 11 [2] RM
LESSJ033157.1-275940 LESS 25 03315685 59389 13 [2] RM
LESSJ033149.7-273432 LESS 27a 03314988 34304 4 [11] I
LESSJ033149.7-273432 LESS 27b 03314992 34367 7 [6] MI
LESSJ033336.9-275813 LESS 29 03333688 58088 4 [8] R
LESSJ033150.0-275743 LESS 31 03314977 57404 6 [9] RI
LESSJ033217.6-275230 LESS 34 03321760 52281 17 [0] M
LESSJ033149.2-280208 LESS 36 03314894 02136 7 [7] RM
LESSJ033336.0-275347 LESS 37 03333601 53494 11 [1] M
LESSJ033144.9-273435 LESS 39 03314500 34363 13 [1] RM
LESSJ033246.7-275120 LESS 40 03324677 51207 17 [0] RM
LESSJ033110.5-275233 LESS 41 03311009 52363 4 [0] I
LESSJ033307.0-274801 LESS 43 03330663 48019 8 [9] MI
LESSJ033131.0-273238 LESS 44 03313119 32386 11 [0] RM
LESSJ033256.0-273317 LESS 47 03325599 33189 8 [6] MI
LESSJ033237.8-273202 LESS 48 03323800 31594 4 [1] RM
LESSJ033124.5-275040 LESS 49a 03312445 50375 12 [1] RM
LESSJ033124.5-275040 LESS 49b 03312469 50464 11 [2] R
LESSJ033141.2-274441 LESS 50a 03314111 44424 17 [0] M
LESSJ033141.2-274441 LESS 50b 03314097 44348 11 [5] RM
LESSJ033243.6-273353 LESS 54 03324362 33566 7 [6] M
LESSJ033153.2-273936 LESS 56 03315311 39373 9 [8] RM
LESSJ033152.0-275329 LESS 57 03315193 53268 11 [6] RM
LESSJ033303.9-274412 LESS 59 03330362 44126 13 [4] RM
LESSJ033317.5-275121 LESS 60 03331753 51275 17 [0] RM
LESSJ033236.4-273452 LESS 62 03323652 34530 16 [1] RM
LESSJ033308.5-280044 LESS 63 03330849 00428 15 [1] RM
LESSJ033201.0-280025 LESS 64 03320098 00253 11 [4] RM
LESSJ033331.7-275406 LESS 66 03333192 54103 14 [0] RM
LESSJ033243.3-275517 LESS 67 03324318 55142 16 [1] RM
LESSJ033144.0-273832 LESS 70 03314392 38352 17 [0] RM
LESSJ033229.3-275619 LESS 73 03322928 56189 8 [9] R
LESSJ033309.3-274809 LESS 74a 03330934 48159 10 [6] RI
LESSJ033309.3-274809 LESS 74b 03330914 48166 10 [6] RI
LESSJ033126.8-275554 LESS 75 03312717 55509 15 [0] RM
LESSJ033221.3-275623 LESS 79 03322161 56231 16 [1] RM
LESSJ033127.5-274440 LESS 81 03312754 44395 14 [1] RM
Table 2: The catalogue of 74 robust counterparts to LESS SMGs, their photometric redshift estimates, reduced of the best-fit SED and the number of photometric filters in which the galaxy is observed. We also present the absolute rest-frame -band magnitudes, the derived far-infrared luminosities and characteristic dust temperatures of the SMGs.
SMG Short name RA Dec Filters ID type
(mag) () (K)
LESSJ033154.2-275109 LESS 84 03315449 51053 14 [3] I
LESSJ033251.1-273143 LESS 87 03325083 31412 5 [0] RM
LESSJ033155.2-275345 LESS 88 03315481 53409 16 [1] R
LESSJ033313.0-275556 LESS 96 03331262 55516 17 [0] RM
LESSJ033130.2-275726 LESS 98 03312989 57224 10 [4] RM
LESSJ033151.5-274552 LESS 101 03315153 45531 10 [7] R
LESSJ033335.6-274020 LESS 102 03333556 40232 11 [2] M
LESSJ033325.4-273400 LESS 103 03332537 33585 5 [7] M
LESSJ033140.1-275631 LESS 106 03314017 56224 11 [5] RI
LESSJ033316.4-275033 LESS 108 03331651 50393 15 [0] RM
LESSJ033122.6-275417 LESS 110 03312263