The galaxy luminosity function at z similar or equal to 6 and evidence for rapid evolution in the bright end from z similar or equal to 7 to 5

International audienceWe present the results of a search for bright (-22.7 \textless= M-UV \textless= -20.5) Lyman-break galaxies at z similar or equal to 6 within a total of 1.65 deg(2) of imaging in the UltraVISTA/Cosmological Evolution Survey (COSMOS) and United Kingdom Infrared Telescope Deep Sky Survey (UKIDSS) Ultra Deep Survey (UDS) fields. The deep near-infrared imaging available in the two independent fields, in addition to deep optical (including z'-band) data, enables the sample of z similar or equal to 6 star-forming galaxies to be securely detected longward of the break (in contrast to several previous studies). We show that the expected contamination rate of our initial sample by cool Galactic brown dwarfs is less than or similar to 3 per cent and demonstrate that they can be effectively removed by fitting brown dwarf spectral templates to the photometry. At z similar or equal to 6, the galaxy surface density in the UltraVISTA field exceeds that in the UDS by a factor of similar or equal to 1.8, indicating strong cosmic variance even between degree-scale fields at z \textgreater 5. We calculate the bright end of the rest-frame Ultraviolet (UV) luminosity function (LF) at z similar or equal to 6. The galaxy number counts are a factor of similar to 1.7 lower than predicted by the recent LF determination by Bouwens et al. In comparison to other smaller area studies, we find an evolution in the characteristic magnitude between z similar or equal to 5 and z similar or equal to 7 of Delta M* similar to 0.4, and show that a double power law or a Schechter function can equally well describe the LF at z = 6. Furthermore, the bright end of the LF appears to steepen from z similar or equal to 7 to z similar or equal to 5, which could indicate the onset of mass quenching or the rise of dust obscuration, a conclusion supported by comparing the observed LFs to a range of theoretical model predictions