The NOAO Distant Radio Galaxies Optically Non-detected in the SDSS (DRaGONS) Survey
There be DRaGONS.....
Massive radio galaxies provide a powerful tool to study the high redshift universe. With strong radio emission visible to z>7, they are known to form in the most dense regions of the early universe. In the standard CDM paradigm, they would be the first systems to form stars, possibly early enough to probe the epoch of reionization (Barkana & Loeb 2006). The presence of this early star formation, in conjunction with powerful AGN activity inferred from the high radio luminosity, makes HzRGs good candidates for examining feedback processes, and their role in the formation of the oldest and most massive galaxies. While radio surveys can address many fundamental questions in cosmology and galaxy formation the numbers of radio galaxies identified at high redshift remain small in comparison to low redshift surveys (Brand et al., 2005; Magliocchetti et al., 2004). Without large, statistically complete and homogeneously selected samples we cannot hope to constrain hierarchical galaxy formation models without the concern that sample variance might bias our analyses. For example, at redshifts z > 3 extensive optical and near-infrared (NIR) campaigns have yielded less than one hundred and fifty galaxies (van Breugel et al., 1998; De Breuck et al., 2001; Vardoulaki et al., 2006). The reason for the paucity of these samples comes from the necessity of surveying large volumes to identify the most massive systems. Given the broad redshift distribution of radio galaxies (e.g., Dunlop and Peacock, 1990), large numbers of radio targets must be observed in order to extract the high redshift component.
Selecting High Redshift Radio Galaxies
The main obstacle to identifying high redshift radio galaxies is screening out the low redshift foreground. Bright radio galaxies (>100 mJy) have a broad redshift peak at z ~2. Extracting only the high-z galaxies through blind spectroscopic follow up of radio surveys is inefficient. Our goal is to eliminate the low redshift contamination through the inclusion of multi-wavelength information. Figure 2 shows a simulation of the r-K colors as a function of redshifts for a range of galaxy spectral energy distributions (including dusty and evolved stellar populations). The thick line shows the expected colors of radio galaxies with an r-band magnitude r<24.1 Sources below the line on the diagram would be detected in by the SDSS (if we combine all passbands), while sources above would not. Note that there will be some scatter in the cutoff due to the scatter in the K-z Hubble diagram. The exact nature of the objects passing the magnitude cut will depend on the color dependence of the scatter in th K-z relation. We can, therefore, use the optical properties of the radio galaxies (or their lack of optical detection) to eliminate the low redshift component of the distribution.
We select all S_1.4 GHz,>100mJy FIRST sources that are not detected in SDSS optical images for follow up in the NIR K_S-band, with no consideration of the radio spectral slope.
An Infrared Survey of High Redshift Radio Galaxies
Using this selection technique we developed the DRaGONS survey; a K-band survey of 530 FIRST radio sources having with 1.4 Ghz fluxes > 100 mJy with no optical counterparts. DRaGONS covers over 5000 square degrees (correspnding to the union of the SDSS and FIRST footprints), which represents a volume at z = 2 of about 6x1010 Mpc3. Supported by the NOAO survey program we acquired 40 nights of K-band imaging data 3. The official survey began in 2005. Observing was scheduled in six night groups during bright time: 2005 September 16-21, 2006 March 14-19, 2006 May 11-16, 2006 October 5-10, 2007 March 29 to 2007 April 3, and 2007 May 26 to 2007 June 1. All observations were done on the KPNO 4-meter Mayall telescope using the Florida Multi-object Imaging Near-IR grism Observational Spectrometer (FLAMINGOS) instrument. The detector is a 2048x2048 HgCdTe wide-field IR imager and multi-slit spectrometer with a pixel size of 0.316", which gives a 10.8'x10.8' FOV on the 4 meter telescope
Figure 2 shows the estimated redshift distribution based on converting K-band magnitudes into redshifts using a linear fit to the K-z Hubble diagram. The shaded portion of the histogram represents the lower limits on the estimated redshifts for the non-detected objects. If we assign redshifts to the radio galaxies based on the linear fit to the K-z diagram, the mean redshift for this sample is z = 2.7 and the median redshift is z = 2.2. Simulations indicated we should exclude all sources with z < 1.8 and an average redshift for the sample of z~2.5; consistent with our results. We find no strong correlation between redshift and the slope of the radio spectrum for our sample (Ultra Steep Radio Sources are a common technique for identifying high redshift radio galaxies) in our sample., Our target selection is sensitive to a population of HzRGs missed by USS techniques.
Extremely Red Galaxies
The majority of the DRaGONS galaxy sample appears to be at z > 2 (as expected given our selection criteria). 10% of the sample are, however, anomalously bright in the K-band (K_Vega = 16.5 - 17.5) with extremely red colors, r - K > 6.5 (see Figure 3). These Red DRaGONS, as they are known, are redder than the red quasars of Glikman et al (2004), which have 4 < R - K < 6.5 (with the exception of one object at R-K=8) and are among the reddest AGN ever detected. Assuming an elliptical galaxy spectral energy distribution would require a reddening of A_R > 2 mag, an extreme global obscuration for AGNs. In comparison to local samples the most heavily obscured quasars in a recent study of type 2 quasars using SDSS data (Zakamska et al 2006) only have AV = 1.2 mag. The current work on DRaGONS is to address the nature of these extremely red galaxies and determine their abundance. If, for example, the Red DRaGONS reside at high redshift (z > 3) with extreme dust obscuration, this would place strong constraints on their star formation history (and the rate at which dust is generated) as well as the potential contribution of radio galaxies in the reionisation of the universe. If the Red Dragons are found contain both strong AGN and strong star formation processes this places limits on the extent of any quenching AGN feedback. To differentiate between a high redshift starburst, a heavily dust obscured low redshift elliptical, or perhaps an highly obscured type 2 AGN requires infra-red spectroscopy to measure the AGN contribution, redshift and the dust extinction.