Star formation at high redshift: The confrontation between theory and observations

We investigate properties of Lyman-break galaxies by statistically comparing photometric observations with predictions derived from semi-analytic models based on the ΛCDM theory of hierarchical structure formation. We construct samples of U, B435, and V775-dropouts produced by GOODS, and complement the ACS optical B435, V606, i775, and z850 data with the VLT ISAAC J, H, and Ks and IRAC 3.6, 4.5, 5.8, and 8.0 observations. We produce model dropout galaxies derived from semi-analytic model runs, where parameters controlling star formation and dust content are varied. We then construct model density functions and convolve them with observational scatter derived from Monte-Carlo simulations. We find the best-fit models by computing likelihoods using the data and model dropouts and the UV-continuum and Balmer-break color-magnitude diagnostics. We find that we cannot discern among models with varying starburst efficiencies due to data limitations. However, we do favor models with enhanced quiescent star formation. Our best-fit models rule out any strong dependence of quiescent star formation on circular velocities. We also favor dusty models. Using the best-fit models we present predictions for the stellar masses, SFRs, and ages of the z ∼ 3, z ∼ 4, and z ∼ 5 Lyman-break samples. We find that even though the current optical surveys are effective at selecting UV-bright, massive galaxies, they fail to select most of the stellar mass, which remains hidden in UV-faint and moderately massive galaxies. Our best-fit models predict a ∼70% mass build-up between the z ∼ 4 and z ∼ 3 epochs for UV rest-frame L* galaxies, and a smaller, ∼50%, build-up between the z ∼ 5 and z ∼ 4 epochs. This implies an on-going process of quite active stellar-mass assembly between the z ∼ 5 epoch and the z ∼ 3 epoch. Furthermore, for the z ∼ 3 sample, the stellar masses range from 108 to 1010 M⊙ , roughly 1.5 orders of magnitude less than the stellar masses of the present day L* spirals and ellipticals---this indicates that the z ∼ 3 Lyman-break galaxies are not the fully assembled progenitors of the present-day L > L* galaxies. Finally, we find that quite a few of the z ∼ 5 galaxies have stellar masses of >1010 M⊙ , and that the median age of the z ∼ 5 population is 240 Myrs. This points to an already active star formation well before the z ∼ 5 epoch.