Search For a Counterpart to the Subsolar Mass Gravitational Wave Candidate S251112cm

Abstract

The recent gravitational-wave (GW) alert from a compact object merger involving at least one subsolar mass (SSM) object has prompted questions about their origins. S251112cm is reported by LIGO/Virgo with a false alarm rate of 1 per 6.2 years, nearby luminosity distance $93 \pm 27$ Mpc, probability of containing a SSM object of 100%, and probability of containing a $1-3~M_\odot$ object of just 8%. Such a system likely did not involve the supersolar neutron stars or black holes invoked to explain kilonovae. One must then also invoke hitherto unobserved and speculative models to produce SSM mergers and the resultant electromagnetic (EM) counterparts. We introduce a framework which vets and scores candidate counterparts to SSM GW events to inform follow-up in search of any among the zoo of potential EM transients: kilonovae, kilonovae-within-supernovae, super-kilonovae, or AGN flares from binary black hole mergers. We use a suite of telescopes to perform tiling, galaxy-targeted observations, and photometric/spectroscopic follow-up of promising candidates. In near-real time, we ingest candidates reported by the community, including some of the first observations reported by the Vera C. Rubin Observatory. We vet and score a total of 248 candidates, including 67 from Rubin, but find no likely counterpart. We nonetheless highlight candidates which demonstrate the ability of our framework to distinguish between different transient types and describe strategies to maximize the chances of detecting a counterpart to the next SSM event. Our framework will be implemented in the forthcoming Multimessenger Tool for Rapid Object Vetting and Examination (TROVE).

Figures (10)

• Figure 1: Primary and secondary masses $m_1$ and $m_2$ for varying source frame chirp mass $\mathcal{M}$ in the range $\mathbf{[0.10, 0.87]~M_{\odot}}$. The progenitor system $(m_1, m_2)$ for S251112cm may lie along any of the plotted curves. We adopt the LVK convention that $m_1 \geqslant m_2$ such that the mass ratio $q = m_2 / m_1 \leqslant 1.0$. Depending on the chirp mass, $m_1$ may also be sub or supersolar in mass. We truncate for both $m_1~\mathrm{and}~m_2 \leqslant 0.2~M_{\odot}$ and $q \leqslant 0.1$, given the limits of waveform template banks hanna2025_SSM-templates. We highlight the range $1 - 3~M_{\odot}$. S251112cm was reported by the LVK with a probability of containing a $1 - 3~M_{\odot}$ object (HasNS) of just 0.08.
• Figure 2: GW localization region of S251112cm, with our telescope pointings in search of an optical counterpart, and candidates identified by all teams. Purple and blue contours denote the 50% and 90% confidence localization regions, which span 371 and 1,681 $\mathrm{deg}^2$, respectively. Top left and right: Zoomed-in insets show our telescope pointings in search of a counterpart with DLT40 (magenta), CSS (orange), and T80-S (yellow), in the two lobes of the localization region. Bottom: Points denote the 248 candidate counterparts reported to the TNS and/or in GCNs within 10 days of the GW signal, color-coded by score, where we select the maximum score among KN, KN-in-SN, and super-KN scores (Section \ref{['sec:scoring']}).
• Figure 3: Depths achieved by our CSS and T80-S observations, compared to a prototypical KN, SN Ic-BL, and SN IIb. Models are generated from redback following an AT 2017gfo-like kilonova, SN 1998bw (Ic-BL), and SN 2016gkg (IIb; a class of SESNe but also noted KN impostors), scaled to the GW luminosity distance $93 \pm 27~\mathrm{Mpc}$ to S251112cm (Section \ref{['ssc:obs-search']}). Translucent bands reflect the uncertainty on this distance. The KN model is truncated at 25 days, beyond which it loses reliability. Our CSS and T80-S observations would have likely found any KNe, SNe Ic-BL, or SNe IIb, in our observed fields. We also include median $g$- and $i$-band depths achieved by Rubin-LSST, as the greatest depths achieved by any search (gcn42707gcn43257).
• Figure 4: Spectra of candidate counterparts or the putative hosts of candidate counterparts, with source classification and measured redshift. Observations are compiled in Table \ref{['tab:spectra']}. We obtained SOAR/Goodman spectra of five candidates (AT 2025adgp, AT 2025adhf, AT 2025adhs, AT 2025adim, and AT 2025adiw) and the host of AT 2025adjf, MMT/Binospec spectra of AT 2025adiz, and Bok/B&C spectra of the host galaxies of AT 2025adht and AT 2025adra. All candidates resemble SNe Ia or II at redshifts $z$ too distant to correspond to S251112cm. All host galaxies similarly lie outside of the localization volume.
• Figure 5: Overall framework for scoring and ranking candidate counterparts to a SSM GW event. We ingest GW alerts from the International Gravitational-Wave Observatory Network (IGWN) alert stream, transients and their photometry/classifications from the TNS, galaxies, point sources, variable objects, minor planets, and AGN from relevant catalogs, and forced photometry. For each candidate, GW alerts are used to compute the 2D localization subscore $S_{\mathrm{2D}}$ and in combination with galaxy catalogs to compute distance subscores $S_{\mathrm{dist}}$. Candidates' positions are compared to those of point sources, minor planets, and variable objects to compute $S_{\mathrm{PSMPC}}$. Association of candidates with AGN using an AGN catalog yields the AGN association subscore $S_{\mathrm{AGN}}$. If vetting for AGN flaring, the distance to any associated AGN is used in distance scoring, as represented with a dotted line. Photometry scores $S_{\mathrm{phot}}$ are computed by checking for predetections and comparing candidates' photometry to predictions for peak luminosity $L_{\mathrm{peak}}$, peak time $t_{\mathrm{peak}}$, and decay rate $\Delta m$, dependent on transient type $T$. Our framework introduces a dependence on the EM transient class $T$ in question (KNe, KNe-in-SNe, super-KNe, or AGN flares) for distance, point source/minor planet association, AGN association, and photometry scoring. Taking the product of the five subscores, we produce unique overall scores $S_T$ for each candidate for each class of EM transient $T$ (Equation \ref{['eqn:overallscore']}).