1. Figure 5: Example of stacked images. Figure 6: Number count plot where the diamonds are the simulated data assuming no evolution from z=3-4 to z=5 and the crosses are our new data. Search For z~5 Galaxies Laura Douglas¹, Malcolm Bremer¹, Matt Lehnert² 1.HH Wills Physics Laboratory, University of Bristol, Tyndall Avenue, Bristol BS8 1TL, U.K. 2. Max-Planck-Institut für extraterrestrische Physik, Glessenbachstraβe, 85748 Garching bei München Introduction Availability of deeper images and more sensitive equipment has lead to identification of ever increasing numbers of high-redshift galaxies. Until recently the size of candidate samples has been small, limiting the reliability of the statistics of these sources. To remedy this we have begun a survey of 800 arcmin 2 of sky over 20 fields using extremely deep multicolour imaging. The data used were obtained by the EDisCS project (P.I. S. White). Deep optical photometry was taken using FORS2 at the VLT in the V-, R- and I-bands (2 hours in each, with typical detection limits of AB>27), with complementary deep J and Ks imaging from the NTT and I-band ACS images from the HST. The data were originally obtained to identify clusters at 0.4<z<0.8. In addition to standard reduction techniques the images were corrected for galactic extinction and the lensing effects of the observed clusters, using the lensing maps provided by Clowe et. al. (2004). An I-band cluster image and its corresponding lensing map can be seen in Fig 1. The high-redshift galaxies were identified using the dropout technique pioneered by Steidel and collaborators. The technique uses the fact that emission from these high redshift sources at wavelengths shortward of Lyman alpha is absorbed by the intervening neutral hydrogen in the IGM. This break in the spectrum can be observed using filters shortward and longward of the Lyman break. In practice this means we select objects with I AB 1.3. Objects with a m AB =26 have a M AB (1700 Å)>-21. Object Selection A object catalogue was prepared, using Sextractor, from the reduced EDisCS images. Apertures were defined by the I-band image and were then applied to the V-, R-, J- and Ks-band images. High-redshift candidates were selected using the colour cuts above, determined from modelling of expected colours of high-redshift galaxies, and their half-light radii from the high resolution HST images. Once this sample had been identified, photometric and morphological cuts were used to remove any contaminants. High-redshift galaxies are not the only sources to be identified by the colour cuts used, so the IR and HST data provided possible J- or Ks-band detections and morphological information that could identify EROs (Extremely Red Objects, typically galaxies at z~1) and cool stars such as M-type dwarfs, the main contaminants in a sample like this. Despite the fact that the IR data is not deep enough to identify high-redshift galaxies it is deep enough to identify EROs and cool sub stellar objects. Also, the HST data can identify stars as they are unresolved and EROs which have half-light radii >0.5” whereas true high-redshift galaxies are resolved but with small half-light radii, 0.1”-0.2” (Bremer et. al., 2004, MNRAS, 347,L7). Results Number Counts and the Bright-end Slope of the Luminosity Function Lehnert & Bremer (2003, ApJ, 593,630) claimed that the number of z>5 sources selected in a similar manner to this work is less than that expected from the z=3-4 luminosity function of Lyman break galaxies, at least at I AB <26.3. Although their data was taken with the same instrument as ours their conclusion was drawn from only 10% of the area of sky of this most recent survey. Nevertheless, our average number counts over 10 fields, also implies a lack of bright sources compared to that expected if there was no evolution in the luminosity function between z=3 and z=5 (Fig 6). A caveat is that we removed contaminants from the brighter bins using the IR data but not from the fainter bins. Although Lehnert and Bremer (2003) found no contaminants in their fainter bins we clearly need to confirm this assumption with spectroscopy. Figure 3: Images of objects in the V-, R-, I-, J- and Ks-bands with composite optical image. Top object is an example of a good high-redshift candidate, the middle object is an example of a likely M type dwarf and the bottom object is an example of a likely ERO. HSTVRIJKs Ks-Band Stacked Images - Average Colours of z=5 Galaxies 13 areas of the same field the image was over 1 magnitude deeper than an individual frame. Fig 5 shows a non- detection in the stacked image K AB >25, implying that the average colour of these objects is I-K 10, therefore the universe could not have been ionised at these redshifts. If the high-redshift candidates were just below the detection limit of the Ks-band with a range of I-K colours the stacked image would give us a detection, so the lack of detection tells us that all the candidates are well below the detection limit. Figure 4: Images of objects in the V-, R-, I-, J- and Ks-bands with composite optical image. The top object is a possible double object and the bottom object shows a tail-like feature. HSTVRIJKs I AB Number Counts Number (0.5 Mag bin) -1 Figure 1: I-band image of cluster CL1054-1245 and its corresponding lensing map. The lensing map shows the number of magnitudes correction that should be applied as a function of position for z=5 objects. 0.5mag 0mag Our work, and that of others, on the Hubble Ultra Deep Field (see presentation by M. Bremer) have shown that in addition to compact objects a significant fraction of high-redshift candidates are part of a double system or exhibit tail-like features. Two examples of this morphology have been found in our sample (Fig 4). The greater resolution of the HST images reveals a double object, in which each member shows similar colours, and an object with a tail of similar colour to the main detection. Examples of good candidates and contaminants can be seen in Figs 3 and 4. After removing the majority of contaminants we have identified 150 excellent candidates in the first 430 arcmin 2. The number of candidates in each field ranges from 5 to 47. This significant difference does not appear to be correlated with the lensing properties of the different clusters, but rather a consequence of cosmic variance which demonstrates the importance of having a large survey area. A break in colours between candidate high-redshift galaxies and the redder EROs is shown in Fig 2. The colours of stars are not such a useful tool but with the added half-light radii information from the HST these can still be easily identified. Even without the removal of contaminants there would still be a lack of sources of ionising photons. This paucity implies that the unobscured star formation density and the UV density provided by the brightest galaxies decreases by a factor of ~3, which in turn is insufficient to ionise the universe implying that the majority of ionising photons originated from fainter sources. (Lehnert & Bremer, 2003). The I-K colour of a galaxy is an important value as it shows how long the galaxy has had active star formation: the redder the colour, the longer the star formation episode. To obtain an average I-K colour for the candidate galaxies we stacked the Ks-band data to see if there was a combined detection. By combining up to Figure 2: Colour-colour plot where the diamonds are EROs and the triangles are identified stars. The arrows depict the colour limits of the candidates as they are all detected in the I-band but not in the R-band, to a limit of 27.8, and the Ks-band, to a limit of 22.3. Two of the unresolved objects, previously identified as stars, have colours similar to the high-redshift candidates, suggesting they could be quasars. I AB -K AB R AB -I AB Colour-Colour Plot