Using very deep Spitzer MIPS 24 \mu m observations , we present an analysis of the bolometric luminosities and UV extinction properties of more than 200 spectroscopically identified , optically selected ( U _ { n } G { \cal R } ) z \sim 2 galaxies in the GOODS-N field .
The large spectroscopic sample of rest-UV selected galaxies is supplemented with photometrically identified near-IR-selected ( “ BzK ” and “ DRG ” ) galaxies and sub-mm sources at similar redshifts in the same field , providing a representative collection of relatively massive ( M ^ { \ast } > 10 ^ { 10 } M _ { \odot } ) galaxies at high redshifts .
We focus on the redshift range 1.5 < z < 2.6 , for which the 24 \mu m observations provide a direct measurement of the strength of the mid-IR PAH features in the galaxy spectra ; the rest-frame 5 - 8.5 \mu m luminosities ( L _ { 5 - 8.5 \mu m } ) are particularly tightly constrained for the objects in our sample with precise spectroscopic redshifts .
We demonstrate , using stacked X-ray observations and a subset of galaxies with H \alpha measurements , that L _ { 5 - 8.5 \mu m } provides a reliable estimate of L _ { IR } for most star forming galaxies at z \sim 2 .
We show that the range of L _ { IR } in the optical/near IR-selected samples considered extends from \simeq 10 ^ { 10 } L _ { \odot } to > 10 ^ { 12 } L _ { \odot } , with a mean \langle L _ { IR } \rangle \simeq 2 \times 10 ^ { 11 } L _ { \odot } .
The LIRG population at z \sim 2 is essentially the same population of galaxies that are selected by their optical/near-IR colors .
Objects with LIRG to ULIRG luminosities are present over the full range of stellar masses in the samples , from 2 \times 10 ^ { 9 } M _ { \odot } to 5 \times 10 ^ { 11 } M _ { \odot } .
We use the MIPS 24 \mu m observations for an independent examination of dust extinction in high redshift galaxies , and demonstrate that , as in the local universe , the obscuration ( L _ { IR } \over L _ { 1600 } ) is strongly dependent on bolometric luminosity , and ranges in value from < 1 to \sim 1000 within the sample considered .
However , the obscuration is \sim 10 times smaller at a given L _ { bol } ( or , equivalently , a similar level of obscuration occurs at luminosities \sim 10 times larger ) at z \sim 2 than at z \sim 0 .
We show that the values of L _ { IR } and obscuration inferred from the UV spectral slope \beta generally agree well with the values inferred from L _ { 5 - 8.5 \mu m } for L _ { bol } < 10 ^ { 12 } L _ { \odot } .
As found previously by several investigators , for “ ultraluminous ” objects with L _ { bol } > 10 ^ { 12 } L _ { \odot } it is common for UV-based estimates to underpredict L _ { IR } by a factor of \sim 10 - 100 .
Using the specific star formation rate of galaxies ( SFR per unit stellar mass ) as a proxy for cold gas fraction , we find a wide range in the evolutionary state of galaxies at z \sim 2 , from galaxies that have just begun to form stars to those which have already accumulated most of their stellar mass and are about to become , or already are , passively-evolving .