AT2018kzr: the merger of an oxygen-neon white dwarf and a neutron star or black hole

We present detailed spectroscopic analysis of the extraordinarily fast-evolving transient AT2018kzr. The transient's observed lightcurve showed a rapid decline rate, comparable to the kilonova AT2017gfo. We calculate a self-consistent sequence of radiative transfer models (using $\textsc{tardis}$) and determine that the ejecta material is dominated by intermediate-mass elements (O, Mg and Si), with a photospheric velocity of $\sim$\,12000--14500\,$\rm{km}\,s^{-1}$. The early spectra have the unusual combination of being blue but dominated by strong Fe\,{\sc ii} and Fe\,{\sc iii} absorption features. We show that this combination is only possible with a high Fe content (3.5\%). This implies a high Fe/(Ni+Co) ratio. Given the short time from the transient's proposed explosion epoch, the Fe cannot be $^{56}$Fe resulting from the decay of radioactive $^{56}$Ni synthesised in the explosion. Instead, we propose that this is stable $^{54}$Fe, and that the transient is unusually rich in this isotope. We further identify an additional, high-velocity component of ejecta material at $\sim$\,20000--26000\,$\rm{km}\,s^{-1}$, which is mildly asymmetric and detectable through the Ca\,{\sc ii} NIR triplet. We discuss our findings with reference to a range of plausible progenitor systems and compare with published theoretical work. We conclude that AT2018kzr is most likely the result of a merger between an ONe white dwarf and a neutron star or black hole. As such, it would be the second plausible candidate with a good spectral sequence for the electromagnetic counterpart of a compact binary merger, after AT2017gfo.