AT2025ulz and S250818k: Investigating early time observations of a subsolar mass gravitational-wave binary neutron star merger candidate

TL;DR

This study analyzes AT2025ulz as a potential electromagnetic counterpart to the sub-threshold GW event S250818k. Using the Nuclear Multimessenger Astronomy framework, it performs Bayesian comparisons between kilonova and shock-cooling models against early optical/NIR data, finding that both can describe the first ~4 days, with a statistical tilt toward shock cooling, while spectroscopic data reveal a bluer continuum than canonical kilonova predictions. Continued monitoring uncovers a SN-like rebrightening incompatible with a standard kilonova, illustrating the difficulty of classifying early-time transients and the risk of mistaking impostors for genuine kilonovae. The results underscore the need for deep, multi-band, long-term follow-up and especially robust near-infrared templates to distinguish kilonovae from other transients at cosmological distances, as well as the possibility of exotic merger channels such as sub-solar-mass compact-object binaries.

Abstract

Over the past LIGO--Virgo--KAGRA (LVK) observing runs, it has become increasingly clear that identifying the next electromagnetic counterparts to gravitational-wave (GW) neutron star mergers will likely be more challenging compared to the case of GW170817. The rarity of these GW events, and their electromagnetic counterparts, motivates rapid searches of any candidate binary neutron star (BNS) merger detected by the LVK. We present our extensive photometric and spectroscopic campaign of the candidate counterpart AT2025ulz to the low-significance GW event S250818k, which had a ${\sim} 29\%$ probability of being a BNS merger. We demonstrate that during the first five days, the luminosity and color evolution of AT2025ulz are consistent with both kilonova and shock cooling models, although a Bayesian model comparison shows preference for the shock cooling model, underscoring the ambiguity inherent to early data obtained over only a few days. Continued monitoring beyond this window reveals a rise and color evolution incompatible with kilonova models and instead consistent with a supernova. This event emphasizes the difficulty in identifying the electromagnetic counterparts to BNS mergers and the significant allotment of observing time necessary to robustly differentiate kilonovae from impostors.

Figures (13)

• Figure 1: A representative color image of the field of AT2025ulz using the $griY$ images obtained on 2025-08-20 with GMOS mounted on Gemini-North. The right panel zooms in on the host galaxy of AT2025ulz. Clumpy star forming regions can be seen across the galaxy. The transient is visible, even without image subtraction, close to the center of the host (though clearly offset from the nucleus). The image is oriented such that North is up and East is to the left. Image credit: NOIRLab CEE Team.
• Figure 2: Top: Optical and near-infrared lightcurve ($grizJ$) of AT2025ulz. We supplemented our photometry (3KK and GMOS) with the data presented by Kasliwal2025sn. The apparent magnitude $m$ is shown on the left axis and the absolute magnitude $M$ is shown on the right. Bottom: Color of AT2025ulz as a function of time in $g-r$, $g-i$, and $g-z$. All photometry has been corrected for Galactic extinction Schlafly2011.
• Figure 3: Comparison of the rest-frame lightcurve of AT2025ulz to the lightcurves of GRB-KN (GW170817/AT2017gfo and GRBs 160821B, 211211A, and 230307A) in the $griz$ bands. The afterglow subtracted lightcurves of the GRB-KN were taken from Rastinejad2025kn, but originally presented in Troja2019bRastinejad2022Troja2022Levan2023Yang2024. The lightcurves have been fit with a spline to guide the eye.
• Figure 4: Color comparison of AT 2025ulz (solid lines) and AT2017gfo (dashed lines) over the first few rest frame days. We note that while the color evolution of AT2025ulz is significant it does not redden to the same degree as AT2017gfo.
• Figure 5: Top Left: Best-fit light curves of the kilonova analysis after 4 days of observation. Top Right: Best-fit light curves of the joint kilonova plus afterglow analysis after 40 days of observation. Bottom Left: Best-fit light curves of the shock cooling analysis after 4 days of observation time. Bottom Right: Best-fit light curves of the SN 1993J template for the same time range, but with a restricted set of passbands. The colored bands in indicate the systematic uncertainty inferred during the analysis. Upper limits are marked as triangles. $J$-band detections that are below the $3\sigma$ detection threshold are used for these fits.