The metals-to-dust ratio to very low metallicities using GRB and QSO absorbers; extremely rapid dust formation

Among the key parameters defining the interstellar media (ISM) of galaxies is the fraction of the metals that are locked up in dust: the metals-to-dust ratio. This ratio bears not only on the ISM and its evolution, but also particularly on the origin of cosmic dust. We combine extinction and abundance data from gamma-ray burst (GRB) afterglows with similar data from quasar (QSO) foreground absorbers, as well as from multiply-imaged galaxy-lensed QSOs, to determine the metals-to-dust ratios for lines of sight through a wide diversity of galaxies from blue, dwarf starbursts to massive ellipticals, across a vast range of redshifts z = 0.1-6.3, and nearly three orders of magnitude of column density and metal abundance. The GRB and lensed QSO extinction methods are the most reliable that are available outside the Local Group (LG), allowing absolute extinction measurements. We thus determine the metals-to-dust ratio in a unique way, providing direct determinations of in situ gas and dust columns without recourse to assumptions with large uncertainties. We find that the metals-to-dust ratios in these systems are surprisingly close to the value for the LG, with a mean value of 10(21.2) cm(-2) A(V) mag(-1) and a standard deviation of 0.3 dex, compared to the Galactic value of 10(21.3) cm(-2) A(V) mag(-1) (in units of the Galactic gas-to-dust ratio). There is no evidence of deviation from this mean ratio as a function of metallicity, even down to our lowest metallicity of 0.01 Z/Z(circle dot). The lack of any obvious dependence of the metals-to-dust ratio on column density, galaxy type or age, redshift, or metallicity indicates a close correspondence between the formation of the metals and the formation of dust. Any delay between the formation of metals and dust must be shorter than the typical metal-enrichment times of these galaxies, i.e. shorter than a few Myr. Formation of the bulk of the dust in low mass stars is therefore ruled out by these data at any cosmic epoch. Furthermore, dust destruction must not dominate over formation/growth in virtually any galaxy environment. The close correlation between metals and dust is a natural consequence of the formation of the bulk of cosmic dust in supernovae. Grain growth in the ISM, if it is to be the dominant cosmic dust formation mechanism, is strongly constrained by these data to operate on very short timescales.