Strategy for identifying Vera C. Rubin Observatory kilonova candidates for targeted gravitational-wave searches

TL;DR

The paper tackles how to identify Vera C. Rubin Observatory kilonova candidates in the Rubin/Fink alert stream to enable targeted gravitational-wave searches for binary neutron star mergers. It combines Rubin LSST simulations (Kasen and Bulla models) with broker filtering strategies to estimate detection rates and contamination, and assesses the feasibility and resource needs of follow-up and targeted GW analyses. The findings show Rubin could detect tens of kilonovae per year, but only a small fraction will emit high-SNR alerts suitable for triggering GW searches; contamination from other transients remains high, demanding robust, multi-criterion filtering and substantial follow-up. The work demonstrates a path toward coordinated multi-messenger discoveries, highlighting practical constraints and proposing concrete filtering and observational strategies to maximize scientific return.

Abstract

Since the observation of the binary neutron star merger GW170817 and the associated kilonova AT2017gfo, the next joint gravitational-wave/optical kilonova has been highly anticipated. Overlapping observations between the Vera C. Rubin Observatory and the international gravitational-wave detector network are expected soon. Wide-field survey facilities, such as Rubin, can serve dual roles in gravitational-wave astronomy: conducting dedicated optical counterpart searches following gravitational-wave triggers and, through surveys such as the Legacy Survey of Space and Time (LSST), providing opportunities for fortuitous kilonova discoveries during routine operations. We use simulations to develop a strategy for identifying kilonova candidates observed by Rubin and processed by the Fink broker. These candidates can be used as astrophysical triggers for a targeted gravitational-wave search. We simulate kilonovae light-curves for the first year of Rubin with the latest observing strategy for the Wide-Fast-Deep and the Deep Drilling Fields. Assuming a kilonova rate of 250 Gpc$^{-3}$ yr$^{-1}$, we find that Rubin brokers should observe $\sim 4$ kilonovae per year with at least one alert above a signal-to-noise ratio of 5 within the gravitational-wave detector horizon ($\sim 350$ Mpc). Most of these will be faint, and detected 1-2 days following the neutron star merger. Photometric and spectroscopic follow-up will be limited to large telescopes. Using archival data from the Zwicky Transient Facility (ZTF) and our proposed selection criteria, we estimate a minimum contamination of at least 30 events per month from other transients and variables, even under our strictest selection criteria. A deep gravitational-wave search targeting Rubin kilonova candidates may lead to the next multi-messenger discovery.

Figures (7)

• Figure 1: Examples of simulated kilonova light-curves using Rubin observing strategy v5.0. We show examples both for Kasen and Bulla models at different redshifts (z). Circles indicate SNR$>3$ and lower triangles SNR$<1$ measurements for difference image photometry. A black vertical line indicates the simulated kilonova peak.
• Figure 2: Well sampled simulated kilonova light-curves for the Kasen model (a) and Bulla model (b) in the Wide Fast Deep fields and (c) the Bulla model in the Deep Drilling Field. The first three columns show the data that the Fink broker will receive. The different columns show: 1. the first alert received by the broker, 2. the second alert plus forced photometry if the new alert has been received more than one day after the first alert, 3. all data received by the broker, 4. the full light-curve detected by Rubin (including proprietary data). Detections with SNR higher than 5 are shown in opaque circles while SNR between 1 and 3 are semi-transparent and are equivalent to the forced photometry received by the broker. We indicate the time of maximum brightness of the kilonova with a black grey line and the range of data received by the broker in grey.
• Figure 3: Magnitude histograms for simulated Kilonovae with the Bulla (Blue) and Kasen (Orange) models. We show the nominal 5-sigma source detection limits for each band-pass for Rubin as a vertical line, and shade the region for which photometry is fainter in light grey.
• Figure 4: Broker detection efficiency as a function of simulated redshift. We show the ratio of light-curves with an emitted alert ($\mathrm{SNR} > 5$) as a fraction of all simulated light-curves. Results are generated using two different kilonova lightcurve models, Bulla (blue) and Kasen (orange). The error bars are from propagated Poisson uncertainties.
• Figure 5: Distribution of duration between the first detection by the broker and: (a) simulated kilonova peak or (b) simulated explosion time. Most kilonovae are detected $\approx 1.5$ days after maximum light and between one and two days after explosion.