2017ApJ...851L..21V

We present the first effort to aggregate, homogenize, and uniformly model the combined ultraviolet, optical, and near-infrared data set for the electromagnetic counterpart of the binary neutron star merger GW170817. By assembling all of the available data from 18 different papers and 46 different instruments, we are able to identify and mitigate systematic offsets between individual data sets and to identify clear outlying measurements, with the resulting pruned and adjusted data set offering an opportunity to expand the study of the kilonova. The unified data set includes 647 individual flux measurements, spanning 0.45-29.4 days post-merger, and thus has greater constraining power for physical models than any single data set. We test a number of semi-analytical models and find that the data are well modeled with a three-component kilonova model: a "blue" lanthanide-poor component (κ=0.5 cm² g^–1) with M_ej~0.020 M_☉ and v_ej~0.27c; an intermediate opacity "purple" component (κ=3 cm² g^–1) with M_ej~0.047 M_☉ and v_ej~0.15c; and a "red" lanthanide-rich component (κ=10 cm² g^–1) with M_ej~0.011 M_☉ and v_ej~0.14c. We further explore the possibility of ejecta asymmetry and its impact on the estimated parameters. From the inferred parameters we draw conclusions about the physical mechanisms responsible for the various ejecta components, the properties of the neutron stars, and, combined with an up-to-date merger rate, the implications for r-process enrichment via this channel. To facilitate future studies of this keystone event we make the unified data set and our modeling code public.