STELLAR MASS FUNCTIONS OF GALAXIES AT 4 < z < 7 FROM AN IRAC-SELECTED SAMPLE IN COSMOS/ULTRAVISTA: LIMITS ON THE ABUNDANCE OF VERY MASSIVE GALAXIES

We build a Spitzer IRAC-complete catalog of objects complementing the K-s-band selected UltraVISTA catalog with objects detected in IRAC only. To identify massive (log (M-*/M-circle dot) > 11) galaxies at 4 <z <7, we consider the systematic effects on photometric redshift measurements from the introduction of an old and dusty template and of a bayesian prior on luminosity, as well as the systematic effects from different star formation histories (SFHs) and from nebular emission lines in estimated stellar population properties. Our results are most affected by the luminosity prior, while nebular lines and SFHs marginally increase the measurement dispersion; the samples include 52 to 382 galaxies, depending on the adopted configuration. Using these results we investigate, for the first time, the evolution of the massive end of the stellar mass functions (SMFs) at 4 <z <7. Given the rarity of massive galaxies at these redshifts, cosmic variance and Poisson noise dominate the total error budget. The SMFs obtained excluding the luminosity prior show no evolution from z similar to 6.5 to z similar to 3.5, indicating that massive galaxies could already be present at early epochs. The luminosity prior reduces the number of z > 4 massive galaxies by 83%, implying a rapid growth of massive galaxies in the first 1.5 Gyr of cosmic history. The stellar-mass complete sample includes one candidate of a very massive (log (M-*/M-circle dot) similar to 11.5), quiescent galaxy at z similar to 5.4 with MIPS 24 mu m detection, suggesting the presence of an obscured active galactic nucleus. Finally, we show that the observed number of 4 <z <7 massive galaxies matches the number of massive galaxies at 3 <z <6 predicted by current galaxy formation models.