Abstract
Interpreting ultraviolet-to-infrared (UV-to-IR) observations of galaxies in terms of constraints on physical parameters-such as stellar mass ($M_{\ast}$) and star formation rate (SFR)-requires spectral synthesis modelling. We investigate how increasing the level of sophistication of the standard simplifying assumptions of such models can improve estimates of galaxy physical parameters. To achieve this, we compile a sample of 1048 galaxies at redshifts 0.7 < z < 2.8 with accurate photometry at rest-frame UV to near-IR wavelengths from the 3D-HST Survey. We compare the spectral energy distributions of these galaxies with those from different model spectral libraries to derive estimates of the physical parameters. We find that spectral libraries including sophisticated descriptions of galaxy star formation histories (SFHs) and prescriptions for attenuation by dust and nebular emission provide a much better representation of the observations than 'classical' spectral libraries, in which galaxy SFHs are assumed to be exponentially declining functions of time, associated with a simple prescription for dust attenuation free of nebular emission. As a result, for the galaxies in our sample, $M_{\ast}$ derived using classical spectral libraries tends to be systematically overestimated and SFRs systematically underestimated relative to the values derived adopting a more realistic spectral library. We conclude that the sophisticated approach considered here is required to reliably interpret fundamental diagnostics of galaxy evolution.