Photometric estimates of stellar masses in high-redshift galaxies

We present a new tool for the photometric estimate of stellar masses in distant galaxies. The observed source spectral energy distributions are fitted by combining sets of various simple stellar populations, with different normalizations and different amounts of dust extinction, for a given (Salpeter) IMF. This treatment gives us the best flexibility and robustness when dealing with the widest variety of physical situation for the target galaxies, including inactive spheroidal and active starburst systems. We tested the code on three classes of sources: complete samples of dusty ISO-selected starbursts and of K-band selected ellipticals and S0s in the HDF South, and a representative sample of $z\sim 2$ to 3 Lyman-break galaxies in the HDF North. We pay particular attention in evaluating the uncertainties in the stellar mass estimate, due to degeneracies in the physical parameters, different star formation histories and different metallicities. Based on optical-NIR photometric data, the stellar masses are found to have overall uncertainties of a factor of ~2 for E/S0s, while for the starburst population these rise to factors 2-5 (even including ISO/15 μm photometric data), and up to ≥10 for Ly-break galaxies. Our analysis reveals in any case the latter to correspond to a galaxy population significantly less massive ($M<$ a few $10^{10}~M_\odot$) than those observed at lower redshifts (for which typically $M>$ several $10^{10}~M_\odot$), possibly indicating substantial stellar build-up happening at $z\sim 1$ to 2 in the field galaxy population. Using simulated deep SIRTF/IRAC observations of starbursts and Lyman-break galaxies, we investigate how an extension of the wavelength dynamic range will decrease the uncertainties in the stellar mass estimate, and find that they will reduce for both classes to factors of 2-3, comparable to what found for E/S0s and good enough for statistically reliable determinations of the galaxy evolutionary mass functions