2002ApJ...565..174R

It appears that the majority of rapidly and well-localized gamma-ray bursts with undetected or dark optical afterglows (``dark bursts'') occur in clouds of size R≳10L^1/2₄₉ pc and mass M≳3x10⁵L₄₉ M_☉, where L is the isotropic-equivalent peak luminosity of the optical flash. We show that clouds of this size and mass cannot be modeled as a gas that is bound by pressure equilibrium with a warm or hot phase of the interstellar medium (i.e., a diffuse cloud): such a cloud would be unstable to gravitational collapse, resulting in the collapse and fragmentation of the cloud until a burst of star formation reestablishes pressure equilibrium within the fragments and the fragments are bound by self-gravity (i.e., a molecular cloud). Consequently, dark bursts probably occur in molecular clouds, in which case dark bursts are probably a by-product of this burst of star formation, if the molecular cloud formed recently, and/or the result of lingering or latter generation star formation if the molecular cloud formed some time ago. We then show that if bursts occur in Galactic-like molecular clouds, the column densities of which might be universal, the number of dark bursts can be comparable to the number of bursts with detected optical afterglows: this is what is observed, which suggests that the bursts with detected optical afterglows might also occur in molecular clouds. We confirm this by modeling and constraining the distribution of column densities, measured from absorption of the X-ray afterglow, of the bursts with detected optical afterglows: we find that this distribution is consistent with the expectation for bursts that occur in molecular clouds and is not consistent with the expectation for bursts that occur in diffuse clouds. Consequently, we find that all but perhaps a few bursts, dark or otherwise, probably occur in molecular clouds. Finally, we show that the limited information that is available on the column densities of the dark bursts is not consistent with the idea that the dark bursts occur in the nuclear regions of ultraluminous infrared/submillimeter-bright galaxies, from which we draw conclusions about the nature of star formation in the universe.