Supernova Multimessenger Signals

A nearby core-collapse supernova can be observed through neutrinos, gravitational waves, electromagnetic radiation, and the composition and kinematics of its ejecta. The channels probe different regions and clocks: neutrinos escape from near the proto-neutron star, gravitational waves trace changing quadrupoles, and photons emerge only after propagation through opaque matter. Distinguishing bounce, turbulence, black-hole formation, and explosion geometry therefore requires a joint forward model, not coincidence by eye. The observational record and release statements here are current through 2026-08-18.

Helpful background. Stress–Energy and Conservation supplies source budgets; the Einstein equations supply relativistic gravity; Relativistic Matter and Compact Objects supplies neutrino and stellar physics; Black-Hole Dynamics supplies collapse geometry; Computational Relativity supplies waveform forward models; and Black-Hole Perturbation Theory supplies damped-mode interpretation.

Four channels and four response models

For a neutrino detector $d$, the expected reconstructed event density is

$$\lambda_d(t,E_{\rm rec})= N_{{\rm target},d}\sum_\alpha \int dE\,R_d(E_{\rm rec}\mid E)\, \epsilon_d(E,t)\,\sigma_{\alpha d}(E)\, \Phi_\alpha(t,E)+b_d(t,E_{\rm rec}),$$

where $R_d$ is the energy-response matrix, $\epsilon_d$ the efficiency, $\sigma_{\alpha d}$ the cross section, $\Phi_\alpha$ the flavor-dependent flux at Earth, and $b_d$ the background. Detector clock, live time, threshold, dead time, energy calibration, and response version are part of the dataset. The published Kamiokande-II event list for SN 1987A remains a concrete measured benchmark; Hirata and collaborators give event times, energies, directions, background, and detector response in Hirata et al. 1988, Tables I–II and §§II–IV.

Gravitational-wave data are calibrated strain time series $d_k=h_k(\vartheta)+n_k$ with detector response, data-quality flags, and a time-dependent noise spectrum. Electromagnetic data require telescope bandpass, zeropoint, cadence, image subtraction, extinction, distance, and radiative-transfer assumptions. Ejecta abundances require line formation, atomic data, ionization state, mixing, and nuclear reaction networks. These responses cannot be replaced by one common “signal amplitude”.

As a public reproducibility example, the Gravitational Wave Open Science Center provides versioned GWOSC O1 and O2 strain data releases. The optically targeted search by Abbott and collaborators illustrates source windows, waveform injections, detector coverage, efficiency, and null-result interpretation in Abbott et al. 2020, §§II–V and Tables I–IX. That search is evidence about its selected supernovae and waveform families, not a generic exclusion of core-collapse emission.

Which physical features can survive inference

The sharp neutronization burst in electron neutrinos marks shock passage through the neutrino-decoupling region in a forward model. A rapidly rotating core can produce a comparatively coherent gravitational-wave bounce feature, whereas convection and SASI generate stochastic, time-varying emission and may modulate neutrino luminosity. Abrupt termination of all neutrino flavors can be evidence for black-hole formation if detector dead time, threshold drift, and flavor conversion cannot explain it. An electromagnetic shock-breakout clock depends on progenitor radius and envelope propagation and need not coincide with the central collapse time.

Explosion geometry can affect gravitational-wave polarization, neutrino anisotropy, continuum polarization, line profiles, remnant kicks, and the spatial distribution of radioactive material. None is unique in isolation. Rotation, viewing angle, stochastic turbulence, progenitor interfaces, and magnetic fields create degeneracies. Nucleosynthetic yields integrate the thermodynamic histories of ejecta and so test the explosion model, but nuclear rates and unobserved material add systematic uncertainty.

The SuperNova Early Warning System combines neutrino triggers so that follow-up can begin before optical discovery. Al Kharusi and collaborators describe the SNEWS 2.0 design, detector inputs, coincidence strategies, and multimessenger objectives in Al Kharusi et al. 2021, §§5–6. An alert is a selection and communication product; scientific inference still returns to detector-level event streams, calibration, and likelihoods.

Joint likelihood, priors, and clocks

Neutrino event lists are naturally modeled with an inhomogeneous Poisson likelihood. Calibrated gravitational-wave strain often uses a Gaussian-noise likelihood or a burst statistic validated on off-source data. Photon counts, fluxes, spectra, and upper limits require their instrument-specific likelihoods. A schematic joint posterior is

$$p(\vartheta,\lambda\mid D) \propto {\cal L}_\nu{\cal L}_{\rm GW}{\cal L}_{\rm EM}{\cal L}_{\rm nuc}\, \pi(\vartheta,\lambda),$$

where $\vartheta$ describes the collapse and $\lambda$ collects distance, orientation, clock offsets, calibration, flavor conversion, extinction, background, and waveform-model nuisance parameters. Conditional independence of the channel likelihoods is an assumption and must be checked when they share distance, sky location, or external calibration.

Priors on progenitor mass and structure, rotation, magnetic field, distance, orientation, neutrino mass ordering and flavor conversion, and explosion mechanism can materially change the posterior. Report prior and likelihood normalization, interval convention, sampler and convergence diagnostics, and posterior-predictive checks. A cross-messenger time delay must include both clock uncertainty and source emission delay; subtracting timestamps alone does not measure a propagation effect.

Selection and nondetections

Selection begins with event windows, detector thresholds and vetoes, duty cycle, sky coverage, optical classification, and the rule associating a transient with a host. A gravitational-wave search triggered by optically discovered supernovae omits optically missed events and has an uncertain collapse-time window. A neutrino alert requires coincidence and therefore has a different completeness. Electromagnetic samples are affected by dust, cadence, limiting magnitude, and host-galaxy targeting.

Nondetections belong in the likelihood through detection efficiency or an upper-limit construction. Injection campaigns must span the stated waveform model class, distance, orientation, and calibration uncertainty. A result conditional on extreme emission models should not be transferred to neutrino-driven waveforms. Likewise, absence of an electromagnetic counterpart does not imply failed explosion without modeling extinction, sky coverage, and survey completeness.

Systematics and reproduction boundary

Dominant systematics include progenitor structure, finite-temperature equation of state, neutrino transport and flavor conversion, waveform stochasticity, gravity treatment, distance and orientation, detector cross sections and thresholds, strain calibration, telescope bandpass and extinction, nuclear rates, and cross-messenger clock offsets. Alternate forward models should be compared or marginalized rather than compressed into an artificially precise engine label.

No event stream, strain, image, or spectrum was reanalyzed for this page. A reproducible analysis would pin archive identifiers, data release and retrieval date, file checksums, calibration or response matrix, quality flags, selected time and energy windows, vetoes, software repository and environment, random seeds, likelihood, priors, injection population, and posterior samples. It would regenerate representative event counts, strain efficiencies or upper limits, and electromagnetic timing while varying progenitor and orientation. New sources or calibration revisions after 2026-08-18 may change the interpretation and require a new analysis with those inputs identified.

Check a neutrino count likelihood

In a one-second window, a detector observes $n=8$ events. The background-only model predicts $b=2$ events; a fixed signal model predicts $s=6$ additional events. Ignoring nuisance parameters and trials, compute the log likelihood ratio for signal plus background against background under a Poisson model.

Solution

For Poisson means $\mu_1=s+b=8$ and $\mu_0=b=2$, terms independent of the model cancel:

$$\ln\frac{{\cal L}_1}{{\cal L}_0} =n\ln\frac{\mu_1}{\mu_0}-(\mu_1-\mu_0) =8\ln4-6\simeq5.09 .$$

The data favor this fixed signal model in the toy window. This number is not a detection significance: a real analysis must include the window-selection trials, uncertain background and efficiency, energy and timing information, calibration, signal-parameter priors, and the probability of the selection under both hypotheses.

References

• Abbott, B. P., et al. “Optically Targeted Search for Gravitational Waves Emitted by Core-Collapse Supernovae during the First and Second Observing Runs of Advanced LIGO and Advanced Virgo.” Physical Review D 101 (2020): 084002. DOI.
• Al Kharusi, S., et al. “SNEWS 2.0: A Next-Generation SuperNova Early Warning System for Multi-Messenger Astronomy.” New Journal of Physics 23 (2021): 031201. DOI.
• Gravitational Wave Open Science Center. “O2 Data Release.” LIGO Laboratory, 2019, accessed 2026-08-18. Data release.
• Hirata, K. S., et al. “Observation in the Kamiokande-II Detector of the Neutrino Burst from Supernova SN1987A.” Physical Review D 38 (1988): 448–458. DOI.