2001ApJ...559..620P

Using deep near-infrared and optical observations of the Hubble Deep Field-North from the Hubble Space Telescope NICMOS and WFPC2 instruments and from the ground, we examine the spectral energy distributions (SEDs) of Lyman break galaxies (LBGs) at 2.0≲z≲3.5 in order to investigate their stellar population properties. The ultraviolet-optical rest-frame SEDs of the galaxies are much bluer than those of present-day spiral and elliptical galaxies and are generally similar to those of local starburst galaxies with modest amounts of reddening. We use stellar population synthesis models to study the properties of the stars that dominate the light from LBGs. Under the assumption that the star formation rate is continuous or decreasing with time, the best-fitting models provide a lower bound on the LBG mass estimates. LBGs with ``L*'' UV luminosities are estimated to have minimum stellar masses ∼10<sup>10</sup> M<sub>☉</sub>, or roughly 1/10 that of a present-day L* galaxy, similar to the mass of the Milky Way bulge. By considering the photometric effects of a second stellar population component of maximally old stars, we set an upper bound on the stellar masses that is ∼3-8 times the minimum mass estimate. The stellar masses derived for bright LBGs are similar to published estimates of their dynamical masses based on nebular emission line widths, suggesting that such kinematic measurements may substantially underestimate the total masses of the dark matter halos. We find only loose constraints on the individual galaxy ages, extinction, metallicities, initial mass functions, and prior star formation histories. Most LBGs are well fitted by models with population ages that range from 30 Myr to ∼1 Gyr, although for models with subsolar metallicities a significant minority of galaxies are well fitted by very young (≲10 Myr), very dusty stellar populations, A(1700 Å)>2.5 mag. We find no galaxies whose SEDs are consistent with young (≲10<sup>8</sup> yr), dust-free objects, which suggests that LBGs are not dominated by ``first-generation'' stars and that such objects are rare at these redshifts. We also find that the typical ages for the observed star formation events are significantly younger than the time interval covered by this redshift range (∼1.5 Gyr). From this, and from the relative absence of candidates for quiescent, non-star-forming galaxies at these redshifts in the NICMOS data that might correspond to the fading remnants of galaxies formed at higher redshift, we suggest that star formation in LBGs may be recurrent, with short duty cycles and a timescale between star formation events of ≲1 Gyr.