The nucleosynthetic site of the rapid ( r ) neutron-capture process is currently being debated .
The direct detection of the neutron star merger GW170817 , through gravitational waves and electromagnetic radiation , has confirmed such events as important sources of the r -process elements .
However , chemical evolution models are not able to reproduce the observed chemical abundances in the Milky Way when neutron star mergers are assumed to be the only r -process site and realistic time distributions of such events are taken into account .
Now for the first time , we combine all the available observational evidence of the Milky Way and its dwarf galaxy satellites to show that the data can only be explained if there are ( at least ) two distinct r -process sites : a quick source with timescales comparable to core-collapse supernovae , t _ { \textsl { quick } } \lesssim 10 ^ { 8 } yr , and a delayed source with characteristic timescales t _ { \textsl { delayed } } \gtrsim 4 Gyr .
The delayed r -process source most probably originates in neutron star mergers , as the timescale fits well with that estimated for GW170817 .
Given the short timescales of the quick source , it is likely associated with massive stars , though a specific fast-track channel for compact object mergers can not be excluded at this point .
Our approach demonstrates that only by looking at all the available data will we be able to solve the puzzle that is the r -process .