Distinguishing Kilonovae from Binary Neutron Star and Neutron Star-Black Hole Mergers

Abstract

Kilonovae from compact binary mergers are most informative when accompanied by a gravitational-wave signal, which can help identify the source as a binary neutron star (BNS) or a neutron star-black hole (NSBH) merger. However, future events will also be discovered serendipitously or through follow-up of other transients, without a confident identification of the progenitor. Hence, we ask whether the kilonova light curve alone can distinguish between these two progenitor channels. Using simulated BNS and NSBH populations together with semi-analytic light curve models, we compare their post-peak evolution across the optical $ugrizy$ bands. The strongest contrast appears in the blue $u$ band 2 days after peak and in the redder $i$ band 10 days after peak. In the $u$ band, typical BNS kilonovae decline by only $\sim 1$ mag within 2 days of peak, whereas NSBH kilonovae typically decline by $\gtrsim 3$ mag over the same interval. In the $i$ band, the trend reverses for most of the population, with NSBH kilonovae evolving more slowly than BNS kilonovae. We attribute this behavior to differences in ejecta mass, opacity, and diffusion timescale between the two merger classes. Although the quantitative overlap is model-dependent, the qualitative distinction persists across model variations, identifying post-peak decline to be a viable diagnostic for inferring whether the source was a BNS or an NSBH merger.

Figures (7)

• Figure 1: Distribution of the post-peak decline $m_{\rm AB}^{\rm peak} - m_{\rm AB}^{{\rm peak}+\Delta t}$ for KNe from BNS (red) and NSBH (blue) mergers. Left: $u$ band with $\Delta t = 2$ d. Right: $i$ band with $\Delta t = 10$ d. Each histogram combines samples from all three NS EOS considered. We find the two populations to be well separated in the $u$ band, with some overlap in the $i$ band.
• Figure 2: Same as Figure \ref{['fig:fid_diff_eos']}, but varying the astrophysical population priors for BNS and NSBH mergers while fixing the EOS to DD2. The separation in the $u$ band plot is largely insensitive to the adopted population. In the $i$ band, the "Gaussian + Uniform" NSBH prior shows slightly higher overlap with the BNS distributions.
• Figure 3: Same as Figure \ref{['fig:fid_diff_eos']}, but shown separately for each EOS. Varying the EOS produces only a small shift in the post-peak decline distributions and does not significantly change the separability between BNS and NSBH KNe in either band.
• Figure 4: Same as Fig. \ref{['fig:fid_diff_eos']}, but illustrating the impact of variations in the ejecta-model assumptions for the fiducial "Uniform" populations with the DD2 EOS. For BNS mergers, we vary the allocation of the total unbound ejecta mass, $m_{\rm ej}$, among the blue, purple, and red components from the fiducial fractions $f_m=\{0.2,0.6,0.2\}$ to $f_m=\{0.26,0.60,0.14\}$, and also modify the characteristic ejecta velocities, $v_{\rm ej}$, of these components. For NSBH mergers, we instead vary the electron fraction of the disk wind ejecta, changing the fiducial value $Y_e=0.3$ (corresponding to $\kappa\sim 3.6\,{\rm cm}^2\,{\rm g}^{-1}$) to $Y_e=0.1$ ($\kappa\sim 37\,{\rm cm}^2\,{\rm g}^{-1}$) and $Y_e=0.5$ ($\kappa\sim 2\,{\rm cm}^2\,{\rm g}^{-1}$).
• Figure 5: Comparison of the post-peak decline distributions obtained using two different prescriptions for the BNS ejecta masses, for the fiducial "Uniform" BNS and NSBH populations with the DD2 EOS. The alternative BNS fits in Eqs. \ref{['new_BNS_dyn']} and \ref{['new_BNS_disk']} remove the bimodality introduced by the $\tanh$ disk mass fit used in the main text, but also impose an effective floor on the disk ejecta for sufficiently compact primary NSs. The new fits shift the BNS distribution toward slower-decaying light curves in the $i$ band for low-mass systems. In the $u$ band, they produce a bimodal BNS distribution: slowly declining light curves from low-mass NS binaries, and a rapidly decaying branch from systems with $M_1 \gtrsim 1.6\,M_{\odot}$ whose disk ejecta masses collapse to the fit floor. Since this floor reflects numerical uncertainty in the underlying simulations Kruger:2020gig, the rapidly decaying $u$ band branch should not be interpreted as physical.