The timing and location of dust formation in the remnant of SN 1987A

The discovery with the {\it Herschel Space Observatory} of bright far infrared and submm emission from the ejecta of the core collapse supernova SN\,1987A has been interpreted as indicating the presence of some 0.4--0.7\,M$_\odot$ of dust. We have constructed radiative transfer models of the ejecta to fit optical to far-infrared observations from the literature at epochs between 615 days and 24 years after the explosion, to determine when and where this unexpectedly large amount of dust formed. We find that the observations by day 1153 are consistent with the presence of 3$\times$10$^{-3}$M$_\odot$ of dust. Although this is a larger amount than has previously been considered possible at this epoch, it is still very small compared to the amount present in the remnant after 24 years, and significantly higher dust masses at the earlier epochs are firmly ruled out by the observations, indicating that the majority of the dust must have formed at very late times. By 8515-9200 days after the explosion, 0.6--0.8\,M$_\odot$ of dust is present, and dust grains with radii greater than 2\,$\mu$m are required to obtain a fit to the observed SED. This suggests that the dust mass increase at late times was caused by accretion onto and coagulation of the dust grains formed at earlier epochs. These findings provide further confirmation that core collapse supernovae can create large quantities of dust, and indicate that the reason for small dust masses being estimated in many cases is that the vast majority of the dust forms long after most supernovae have been detectable at mid-infrared wavelengths.