What is NMMA-Astro-COLIBRI?
-
Classifying transients from their light curves
NMMA-Astro-COLIBRI is the light-curve fitting service that the Astro-COLIBRI team builds and operates on top of the open-source NMMA code. It is used to analyze the photometric light curves of a wide range of transients detected by optical surveys such as the Zwicky Transient Facility (ZTF), the Asteroid Terrestrial-impact Last Alert System (ATLAS), the Legacy Survey of Space and Time (LSST), and others. The goal is to discriminate the nature of a transient by fitting its observed light curve against a set of astrophysical models and computing Bayesian evidence for each.
When a new transient is detected and its photometry uploaded to Astro-COLIBRI, users can request light-curve fits using models from the SNCosmo library as well as dedicated templates, covering the following phenomena:- Type Ia (SNIa) — Thermonuclear supernovae from white dwarfs in binary systems exceeding the Chandrasekhar limit. Fitted with SALT3 and nugent-sn1a (Nugent template). SNIa are remarkably homogeneous in luminosity and serve as standard candles for cosmological distance measurements.
- Type IIP / IIL (SNII) — Core-collapse supernovae from massive stars (> 8 M☉) showing prominent hydrogen lines. IIP show a characteristic plateau in their light curve; IIL decline linearly. Fitted with nugent-sn2p and nugent-sn2l.
- Type IIn (SNIIn) — Interacting core-collapse supernovae with narrow hydrogen emission lines, signature of ejecta colliding with a dense circumstellar medium. Fitted with nugent-sn2n.
- Type Ib/c (SNIb/c) - Hypernovae — Stripped-envelope supernovae lacking hydrogen (SNIb) and sometimes helium (SNIc) lines, from massive stars that lost their outer envelopes. Hypernovae are extremely energetic variants associated with long GRBs, calibrated on SN 1998bw. Fitted with nugent-sn1bc and nugent-hyper.
- Type IIb (SNIIb) — Transitional events showing hydrogen lines early on that fade over time, eventually resembling SNIb spectra. A notable example is ZTF21abotose (SN2021ugl), initially ambiguous between kilonova and supernova, later confirmed as a SNIIb by spectroscopy.
- Shock Cooling — Early-time emission from shock-heated extended stellar material, preceding the main supernova peak. Characteristic of Type IIb and IIP supernovae. Fitted with Piro2021.
- Kilonovae (KNe) — Electromagnetic counterparts to binary neutron star (BNS) and neutron-star black hole (NSBH) mergers, powered by r-process nucleosynthesis.
For each model, NMMA computes the Bayes factor — a statistical measure of how well the model explains the data relative to other hypotheses. The model with the highest Bayes factor provides the most likely physical interpretation of the transient. This allows Astro-COLIBRI users to quickly assess the nature of an alert and decide whether to trigger follow-up observations.
How the analysis works
-
From photometry to a ranked classification
Once an optical transient is selected, its archived photometry is quality-filtered and converted to the format the fitting engine expects. The light curve is then fitted by nested sampling against the models you choose, and the job runs asynchronously on Astro-COLIBRI's servers: submission to results typically takes a few minutes, with no computing resources required on your side.
Each fit returns the best-fit light curve, a corner plot of the marginalized posterior distributions, the maximum a posteriori parameters with a per-filter reduced χ2, and the Bayesian evidence. Results are delivered straight into the Astro-COLIBRI web and mobile clients, and every job is kept in your history so past fits stay available across sessions.
-
Ranking competing subtypes with Bayes factors
Submitting two or more models on the same transient turns the service into a classifier. The back-end computes the log Bayes factor between the fitted models, ln B12 = ln Z1 − ln Z2, and ranks them accordingly. Because the evidence integrates the likelihood over the whole prior volume, extra free parameters that do not improve the fit are penalized by an Occam factor — the comparison is therefore complexity-aware, not a plain goodness-of-fit ratio. Values are read on the Jeffreys scale, where |ln B| > 4.6 is decisive.
The model library spans eleven supernova models — thermonuclear (Type Ia), hydrogen-rich core-collapse (IIP, IIL, IIn), the transitional Type IIb, stripped-envelope (Ib/c), and an analytical shock-cooling model — together with five kilonova models that serve as an adversarial benchmark against the shock-cooling supernovae which mimic their early-time behaviour. Photometry is combined across ZTF, ATLAS and LSST in a single standardized pipeline.
-
Why it matters: the kilonova impostor problem
Wide-field surveys detect thousands of supernovae for every genuine kilonova, and some subtypes are actively deceptive: Type IIb supernovae with a shock-cooling tail fade on day timescales and stay nearly featureless early on, exactly like a kilonova. Telling them apart photometrically, before spectroscopy settles the question days or weeks later, is a prerequisite for any credible kilonova search.
The service was demonstrated on SN 2021ugl (ZTF21abotose), a Type IIb supernova initially flagged as a kilonova candidate by automated real-time pipelines. Using only the first ~6 days of photometry in two ZTF bands — ten days before spectroscopic confirmation — the empirical Type IIb template already recovered the correct classification, favored over both the kilonova template and the kilonova-mimicking shock-cooling model. On the full 47-day, three-band light curve it again reached the highest evidence of every competing supernova and kilonova model.
-
An open-source code
NMMA is open source, publicly available on GitHub and PyPI. It is actively developed by an international collaboration of astrophysicists. You will find NMMA on GitHub/nuclear-multimessenger-astronomy/nmma and PyPI.org/project/nmma. Full documentation is available on ReadTheDocs.
-
What Astro-COLIBRI provides — and what it does not
Astro-COLIBRI provides the cloud computing resources, the web API and the integration layer of this service, and maintains them. The Astro-COLIBRI frontends (web, Android, iOS) provide a convenient interface to submit light-curve fitting jobs and visualize the results.
Astro-COLIBRI does not maintain, govern or represent the NMMA project, and this service is not endorsed by the NMMA collaboration. The NMMA code is developed independently, and the project runs its own website and documentation at nuclear-multimessenger-astronomy.github.io/nmma.
-
R. W. Kiendrébéogo et al., NMMA--Astro-COLIBRI: An Automated Light-Curve Supernovae Classification Service in the Multi-Survey Era (2026) arXiv
-
P. T. H. Pang et al., Nature Communications 14, 8352 (2023) Nature, arXiv
-
H. Rose et al., nmma: An extended Bayesian framework for Nuclear Multimessenger Astronomy in the Era of Next-Generation Detectors (2026) DOI, arXiv
-
dynesty — J. S. Speagle, MNRAS 493, 3132 (2020) DOI, arXiv
-
PyMultiNest — J. Buchner et al., A&A 564, A125 (2014) DOI, arXiv
-
Systematics error — S. Jhawar et al., Phys. Rev. D 111, 043046 (2025) DOI, arXiv
-
Likelihood-free inference — M. Desai et al., MNRAS (2024) DOI, arXiv
-
Observing scenarios O4/O5 — R. W. Kiendrebeogo et al., ApJS (2023) arXiv
Scientific publications
If you use this service for your research, please cite the following publications:
Service paper (required)
Main NMMA papers (required)