A new estimate of the cosmic star formation density from a radio-selected sample, and the contribution of $H$-dark galaxies at $z\ge 3$

Abstract

The Star Formation Rate Density (SFRD) history of the Universe is well constrained up to redshift $z\sim 2$. At earlier cosmic epochs, the picture has been largely inferred from UV-selected galaxies (e.g. Lyman-break galaxies, LBGs). However, LBGs' inferred SFRs strongly depend on the assumed dust extinction correction, which is not well-constrained at high-$z$, while observations in the radio domain are not affected by this issue. In this work we measure the SFRD from a 1.4 GHz-selected sample of $\sim$600 galaxies in the GOODS-N field up to redshift $\sim 3.5$. We take into account the contribution of Active Galactic Nuclei from the Infrared-Radio correlation. We measure the radio luminosity function, fitted with a modified Schechter function, and derive the SFRD. The cosmic SFRD shows a rise up to $z\sim 2$ and then an almost flat plateau up to $z\sim 3.5$. Our SFRD is in agreement with the ones from other FIR/radio surveys and a factor 2 higher than those from LBG samples. We also estimate that galaxies lacking a counterpart in the HST/WFC3 H-band ($H$-dark) make up $\sim 25\%$ of the $\ddot{I}•$-integrated SFRD relative to the full sample at z $\sim 3.2$, and up to $58\%$ relative to LBG samples.

20 pages, 11 figures. Accepted for publication in ApJ