Spectroscopic r-Process Abundance Retrieval for Kilonovae II: Lanthanides in the Inferred Abundance Patterns of Multi-Component Ejecta from the GW170817 Kilonova

In kilonovae, freshly-synthesized $r$-process elements imprint features on optical spectra, as observed in AT2017gfo, the counterpart to the GW170817 binary neutron star merger. However, measuring the $r$-process compositions of the merger ejecta is computationally challenging. Vieira et al. (2023) introduced Spectroscopic $r$-Process Abundance Retrieval for Kilonovae (SPARK), a software tool to infer elemental abundance patterns of the ejecta, and associate spectral features with particular species. Previously, we applied SPARK to the 1.4 day spectrum of AT2017gfo and inferred its abundance pattern for the first time, characterized by electron fraction $Y_e=0.31$, a substantial abundance of strontium, and a dearth of lanthanides and heavier elements. This ejecta is consistent with wind from a remnant hypermassive neutron star and/or accretion disk. We now extend our inference to spectra at 2.4 and 3.4 days, and test the need for multicomponent ejecta, where we stratify the ejecta in composition. The ejecta at 1.4 and 2.4 days is described by the same single blue component. At 3.4 days, a new redder component with lower $Y_e=0.16$ and a significant abundance of lanthanides emerges. This new redder component is consistent with dynamical ejecta and/or neutron-rich ejecta from a magnetized accretion disk. As expected from photometric modelling, this component emerges as the ejecta expands, the photosphere recedes, and the earlier bluer component dims. At 3.4 days, we find an ensemble of lanthanides, with the presence of cerium most concrete. This presence of lanthanides has important implications for the contribution of kilonovae to the $r$-process abundances observed in the Universe.