Katsianis, A.Tescari, E.Wyithe, J. S.B.2026-01-022026-01-020035-8711ORCID:/0000-0001-7956-9758/work/196588364https://hdl.handle.net/1885/733802443We investigate the evolution of the star formation rate-stellar mass relation (SFR-M*) and galaxy stellar mass function (GSMF) of z ~ 4-7 galaxies, using cosmological simulations run with the smoothed particle hydrodynamics code P-GADGET3(XXL). We explore the effects of different feedback prescriptions (supernova-driven galactic winds and AGN feedback), initial stellar mass functions and metal cooling. We show that our fiducial model, with strong energy-driven winds and early active galactic nuclei (AGN) feedback, is able to reproduce the observed stellar mass function obtained from Lyman-break selected samples of star-forming galaxies at redshift 6 ≤ z ≤ 7. At z ~ 4, observed estimates of the GSMF vary according to how the sample was selected. Our simulations are more consistent with recent results from K-selected samples, which provide a better proxy of stellar masses and are more complete at the high-mass end of the distribution. We find that in some cases simulated and observed SFR-M* relations are in tension, and this can lead to numerical predictions for the GSMF in excess of the GSMF observed. By combining the simulated SFR(M)* relationship with the observed star formation rate function at a given redshift, we argue that this disagreement may be the result of the uncertainty in the SFR-M* (LUV-M*) conversion. Our simulations predict a population of faint galaxies not seen by current observations.21enPublisher Copyright: © 2015 The Author Published by Oxford University Press on behalf of the Royal Astronomical Society.Cosmology: TheoryEvolutionGalaxiesGalaxies: FormationMethodsNumerical2015-04-2110.1093/mnras/stv16084942163878