2010MNRAS.409..855B

We have searched for star-forming galaxies at z ~ 7-10 by applying the Lyman-break technique to newly released Y-, J- and H-band images (1.1, 1.25 and 1.6µm) from Wide Field Camera 3 (WFC3) on the Hubble Space Telescope. By comparing these images of the Hubble Ultra Deep Field with the Advanced Camera for Surveys (ACS) z'-band (0.85µm) images, we identify objects with red colours, (z' - Y)_AB> 1.3, consistent with the Lyman α forest absorption at z ~ 6.7-8.8. We identify 12 of these z'-drops down to a limiting magnitude Y_AB < 28.5 (equivalent to a star formation rate of 1.3M_☉/yr at z = 7.1), all of which are undetected in the other ACS filters. We use the WFC3 J-band image to eliminate contaminant low-mass Galactic stars, which typically have redder colours than z ~ 7 galaxies. One of our z'-drops is probably a T-dwarf star. The z ~ 7z'-drops appear to have much bluer spectral slopes than Lyman-break galaxies at lower redshift. Our brightest z'-drop is not present in the NICMOS J-band image of the same field taken 5 years before, and is a possible transient object. From the 10 remaining z ~ 7 candidates we determine a lower limit on the star formation rate density of 0.0017M_☉/yr/Mpc3 for a Salpeter initial mass function, which rises to 0.0025-0.004M_☉/yr/Mpc3 after correction for luminosity bias. The star formation rate density is a factor of ~10 less than that of Lyman-break galaxies at z = 3-4, and is about half the value at z ~ 6. We also present the discovery of seven Y-drop objects with (Y - J)_AB> 1.0 and J_AB < 28.5 which are candidate star-forming galaxies at higher redshifts (z ~ 8-9). We find no robust J-drop candidates at z ~ 10. While based on a single deep field, our results suggest that this star formation rate density would produce insufficient Lyman continuum photons to reionize the Universe unless the escape fraction of these photons is extremely high (f_esc> 0.5), and the clumping factor of the Universe is low. Even then, we need to invoke a large contribution from galaxies below our detection limit (a steep faint-end slope). The apparent shortfall in ionizing photons might be alleviated if stellar populations at high redshift are of low metallicity or have a top-heavy initial mass function.