Discriminating Exotic Objects from Black Holes

No single currently available observable is a model-independent horizon detector. A horizonless object can mimic black-hole multipoles, photon rings, or part of a ringdown, while accretion and waveform uncertainties can mimic a surface signature. Robust discrimination therefore compares named black-hole and exotic-object models against versioned data, marginalizes environmental and calibration nuisance parameters, and states population prior odds. Datasets and evidence on this page are current through 2026-08-18.

Required background. The Einstein equations supply the black-hole and horizonless model predictions.

Helpful background. Stress–Energy and Conservation supplies response and absorption budgets; Relativistic Matter and Compact Objects supplies alternative interiors; Black-Hole Geometry supplies horizon and photon-region properties; and Computational Relativity supplies waveform and image-model validation.

A portfolio of possible discriminants

Inspiral probes exterior multipoles, spin-induced structure, tidal response, and absorption. A measured nonzero material Love response can disfavor a vacuum Kerr model within specified assumptions, but spin, matter environment, waveform truncation, and the exotic model’s response convention must be controlled. Absence of a measurable tide is not proof of a horizon because a very compact material object’s deformability can be below sensitivity.

Ringdown tests whether damped frequencies and amplitudes are consistent with one Kerr remnant. Horizonless objects may have additional modes or delayed echoes, but echo templates require a reflection law, cavity, excitation, and look-elsewhere accounting. A mismatch can also come from nonlinear merger content, overtone choice, start time, precession, eccentricity, calibration, or noise. Cardoso and Pani compare multipoles, tides, absorption, ringdown, surfaces, shadows, and environmental degeneracies in Cardoso and Pani 2019, §§ 4.1–4.5 and 5.

Horizon-scale imaging resolves a bright, lensed emission structure surrounding a central brightness depression. Its scale can test compact geometry, but the image depends on mass-to-distance ratio, inclination, plasma emissivity, electron thermodynamics, magnetic structure, variability, scattering, array coverage, and imaging regularization. A photon ring or central depression does not logically require an event horizon. Surface-emission arguments similarly depend on accretion rate, thermalization, redshift, emissivity, outflows, and source selection.

Versioned gravitational-wave and imaging data

For gravitational waves, the Gravitational Wave Open Science Center (GWOSC) GW150914-v4 event release, strain files and posterior version is an event-level example; it identifies the release, event version, detector strain, DOI, and parameter-estimation products. Abbott and collaborators report the original detection and binary-black-hole interpretation in Abbott et al. 2016, Methods, Figures 1–4, and Supplemental Material. Any ringdown or exotic-object reanalysis must return to the strain and calibration version rather than copy a catalog point estimate.

Population selection can start from the GWOSC GWTC-5.0 data release, calibration versions and release notes, published 2026-05-26 with DOI 10.7935/bk00-6a89. The cumulative catalog defines a candidate sample, not event likelihoods for an exotic model. Event-level strain, data-quality segments, calibration, and injection-based selection are still required, and alternative astrophysical populations must use the search selection function.

For imaging, the Event Horizon Telescope Collaboration publishes calibrated visibility products and pipeline information. The EHT 2017 L1 data release, product 2019-D01-01 and DOI is a versioned example tied to the M87 observing campaign. The collaboration states the image-scale and interpretation in Event Horizon Telescope Collaboration 2019, Paper I, §§3–6 and Figures 1–5. Reproduction must also use the calibration paper and release-specific gains, flags, array configuration, visibility products, and imaging choices.

Likelihood, priors, and population selection

For model class ${\cal M}$, the evidence is

$$Z_{\cal M}=p(d\mid{\cal M}) =\int {\cal L}(d\mid\theta,\lambda,{\cal M}) \pi(\theta,\lambda\mid{\cal M})\,d\theta\,d\lambda,$$

where $\theta$ contains object parameters and $\lambda$ contains nuisance astrophysics and calibration. A Bayes factor $K=Z_{\rm exotic}/Z_{\rm BH}$ does not include prior odds for the model populations. Those odds must describe formation and abundance consistently with the equilibrium and stability assumptions.

Gravitational-wave selections include event significance, detector duty and quality, mass and spin sensitivity, analysis band, ringdown or echo window, template bank, vetoes, and nondetection treatment. Imaging selections include which source, epoch, array, scans, baselines, polarization products, and quality cuts are used. X-ray or surface samples add telescope response, flux limits, absorption, distance, inclination, accretion state, and source classification.

A combined likelihood may use inspiral, merger–ringdown, imaging, timing, and accretion data only if shared mass, distance, inclination, calibration, or population variables are treated coherently. Multiplying published Bayes factors with incompatible priors or reused data double counts information. Posterior-predictive checks should ask whether the favored model reproduces features not used to tune it.

Systematics can imitate the signature

Waveform systematics include missing eccentricity or precession, inaccurate high modes, merger and overtone modeling, ringdown start time, exotic boundary conditions, environmental matter, and finite numerical resolution. Search systematics include strain calibration, nonstationary noise, data-quality cuts, template flexibility, time–frequency windows, and look-elsewhere effects.

Imaging systematics include station calibration, sparse Fourier coverage, scattering, variability, imaging prior, regularization, and the accretion emission model. A ring-size agreement tests a combination of geometry and mass-to-distance calibration; it does not separately establish a horizon. Surface constraints depend on unobserved accretion power and how efficiently a surface reradiates it. Different horizonless interiors can also be mutually degenerate, so rejecting one echo template does not reject all exotic objects.

The strongest claim identifies which combination of observables drives the result and repeats the analysis under alternate waveform, accretion, image, and population priors. Nondetections enter through efficiency and likelihood, not as zero signal. Exclusion curves must remain inside the model’s stability and formation domain.

Reproduction boundary

No strain, visibility, image, spectrum, or timing data were reanalyzed for this page. A reproducible comparison would pin dataset DOI and event or product version, file checksums, strain or array calibration, instrument response, quality flags, selected detectors or stations, time–frequency or imaging window, software repository and environment, likelihood normalization, priors, population selection, sampler settings, convergence diagnostics, and posterior samples.

It should reproduce at least one Bayes factor or exclusion curve under alternate astrophysical and population priors, perform injections or synthetic-data recovery, and show which observable combination changes the odds. New releases, calibration revisions, or retractions after 2026-08-18 may change the conclusions; any new comparison must identify those changed inputs.

Check prior odds

An analysis reports Bayes factor $K=20$ favoring an exotic model over a black-hole model. The population prior odds are $p({\cal M}_{\rm exotic})/p({\cal M}_{\rm BH})=1/100$. What are the posterior odds and exotic-model probability in this two-model toy problem?

Solution

Posterior odds equal Bayes factor times prior odds:

$$\frac{p({\cal M}_{\rm exotic}\mid d)} {p({\cal M}_{\rm BH}\mid d)} =20\times\frac{1}{100}=0.20.$$

With only two models, the exotic probability is

$$p({\cal M}_{\rm exotic}\mid d) =\frac{0.20}{1+0.20} =\frac16\simeq0.167.$$

The data favor the exotic model’s prior-averaged likelihood (its evidence), yet the stated population prior leaves the black-hole model more probable. This demonstrates why a Bayes factor alone is not a population conclusion; the formation prior and its uncertainty must be reported. A ratio of maximum likelihoods would not give the same answer: the evidence also accounts for how much of each model’s parameter prior fits the data.

References

• Abbott, B. P., et al. “Observation of Gravitational Waves from a Binary Black Hole Merger.” Physical Review Letters 116 (2016): 061102. DOI.
• Cardoso, Vitor, and Paolo Pani. “Testing the Nature of Dark Compact Objects: A Status Report.” Living Reviews in Relativity 22 (2019): 4. DOI.
• Event Horizon Telescope Collaboration. “First M87 Event Horizon Telescope Results. I. The Shadow of the Supermassive Black Hole.” Astrophysical Journal Letters 875 (2019): L1. DOI.
• Gravitational Wave Open Science Center. “GW150914-v4 Event Release.” LIGO Laboratory, 2022, accessed 2026-08-18. Data release.