2009ApJ...701.1839M

We present spectral energy distribution (SED) modeling of a sample of 34 K-selected galaxies at z∼ 2.3. These galaxies have near-infrared (NIR) spectroscopy that samples the rest-frame Balmer/4000 Å break as well as deep photometry in 13 broadband filters. New to our analysis is Infrared Array Camera (IRAC) data that extend the SEDs into the rest-frame NIR. Comparing parameters determined from SED fits with and without the IRAC data we find that the IRAC photometry significantly improves the confidence intervals of τ, A_v, M_star, and SFR for individual galaxies, but does not systematically alter the mean parameters of the sample. We use the IRAC data to assess how well current stellar population synthesis codes describe the rest-frame NIR SEDs of young galaxies where discrepancies between treatments of the thermally pulsating asymptotic giant branch phase of stellar evolution are most pronounced. The models of Bruzual and Charlot, Maraston, and Charlot and Bruzual all successfully reproduce the SEDs of our galaxies with ≤5% differences in the quality of fit; however, the best-fit masses from each code differ systematically by as much as a factor of 1.5, and other parameters vary more, up to factors of 2-3. A comparison of best-fit stellar population parameters from different stellar population synthesis (SPS) codes, dust laws, and metallicities shows that the choice of SPS code is the largest systematic uncertainty in most parameters, and that systematic uncertainties are typically larger than the formal random uncertainties. The SED fitting confirms our previous result that galaxies with strongly suppressed SF account for ∼50% of the K-bright population at z∼ 2.3; however, the uncertainty in this fraction is large due to systematic differences in the specific star formation rates derived from the three SPS models.