Determining the stellar masses of submillimetre galaxies: the critical importance of star formation histories

Submillimetre (submm) galaxies are among the most rapidly star-forming and most massive high-redshift galaxies; thus, their properties provide important constraints on galaxy evolution models. However, there is still a debate about their stellar masses and their nature in the context of the general galaxy population. To test the reliability of their stellar mass determinations, we used a sample of simulated submm galaxies for which we created synthetic photometry. The photometry were used to derived their stellar masses via spectral energy distribution (SED) modelling, as is generally done with real observations. We used various SED codes (GRASIL, MAGPHYS, HYPERZ, and LEPHARE) and various alternative assumed star formation histories (SFHs). We found that the assumption of SFHs with two independent components enables the SED modelling codes to most accurately recover the true stellar masses of the simulated submm galaxies. Exponentially declining SFHs (tau models) lead to lower masses (albeit still formally consistent with the true stellar masses), while the assumption of single-burst SFHs results in a significant underestimation of the stellar masses. Thus, we conclude that studies based on the higher masses inferred from fitting the SEDs of real submm galaxies with double SFHs are most likely to be correct, implying that submm galaxies lie on the high-mass end of the main sequence of star-forming galaxies. This conclusion appears robust to assumptions of whether submm galaxies are driven by major mergers, since the suite of simulated galaxies modelled here contains examples of both merging and isolated galaxies. We identified discrepancies between the true and inferred stellar ages (rather than the dust attenuation) as the primary determinant of the success or failure of the mass recovery. Regardless of the choice of SFH, the SED-derived stellar masses exhibit a factor of similar to 2 scatter around the true value, and this scatter is an inherent limitation of the SED modelling due to simplified assumptions (regarding, e.g., the SFH, detailed galaxy geometry and wavelength dependence of the dust attenuation). Finally, we found that the contribution of active galactic nuclei (<60% at the K-band in these simulations) does not have any significant impact on the derived stellar masses.