IceCat-2: IceCube Alert Track Catalog

• IceCat-2 is a comprehensive catalog of muon track-like neutrino alerts from IceCube, covering data from May 2011 to January 2025.
• It employs a hybrid follow-up reconstruction using the September 2024 update, substantially improving angular resolution and statistical coverage.
• Enhanced localization reduces containment areas by 4-5 times, enabling more efficient cross-correlation with gamma-ray and X-ray source catalogs.

Searching arXiv for the target paper and closely related IceCube catalog context. to=arxiv.search 天天中彩票有人기 大发快三官网 彩经彩票={"query":"(Zegarelli et al., 8 Jul 2025) OR IceCat-2 Updated IceCube Event Catalog of Alert Tracks", "max_results": 5}【อ่านข้อความเต็มanalysis to=arxiv.search code 下载彩神争霸json {"query":"(Zegarelli et al., 8 Jul 2025) OR IceCat-2 Updated IceCube Event Catalog of Alert Tracks", "max_results": 5} Looking up the exact IceCat-2 preprint record. Checking for the first IceCat catalog to anchor the historical comparison. IceCat-2 is the second public catalog of IceCube Alert Tracks: a uniformly reprocessed compilation of all public track-like neutrino alerts from the full-detector era, spanning May 2011 through January 2, 2025, and integrating both the historical sample that predates real-time operations and all real-time alerts issued since 2016. Its defining update is the replacement of earlier event localizations and uncertainties with the September 2024 follow-up reconstruction, which substantially improves angular resolution and statistical coverage for muon track alerts. In preliminary form, the catalog contains 365 track-like alerts and is designed to support systematic multi-messenger studies, including renewed tests of correlations with gamma-ray and X-ray source catalogs (Zegarelli et al., 8 Jul 2025).

1. Scope, event class, and relation to IceCat-1

IceCat-2 preserves the basic scope established by IceCat-1. IceCat-1 gathered all track-like alerts issued in real time since 2016 together with “retro-alerts” reprocessed back to the start of full IceCube operations in 2011, so that the alert stream could be used systematically in multi-messenger analyses. IceCat-2 keeps that time coverage and updates it comprehensively by including all additional alerts since the last IceCat-1 update in October 2023, up to IC-250102A on January 2, 2025.

The catalog is explicitly restricted to “Alert Tracks,” meaning muon track-like events, i.e., charged-current $\nu_\mu$ interactions that produce long muon tracks in the ice. Tracks are emphasized because they afford sub-degree directional precision that improves with energy. Cascades, including those arising from $\nu_e$, $\nu_\tau$, and neutral-current interactions, are not included.

A central curatorial change is the removal of likely cosmic-ray muon contaminants through the IceTop surface-veto introduced in October 2022. Under this cleaning step, 340 events from the original IceCat-1 compilation remain revalidated, and 25 more alerts are added thereafter, yielding a preliminary total of 365 track-like alerts. The sample is distributed all over the sky and corresponds to an average rate of $26.8\ \mathrm{yr}^{-1}$.

2. Reconstruction strategy and uncertainty modeling

The principal methodological change in IceCat-2 is uniform reprocessing with the September 2024 reconstruction. IceCube adopted a hybrid follow-up reconstruction strategy that applies different algorithms depending on energy deposition and topology, with the explicit goals of maximizing directional precision and enforcing robust statistical coverage of the reported uncertainty regions (Zegarelli et al., 8 Jul 2025).

At a high level, the reconstruction seeks agreement between observed DOM photon times and charges and the predictions of a muon-track hypothesis parameterized by a directional unit vector $\theta$, a position/time reference, and an energy proxy $E$. The standard likelihood formulation is

$$L(\theta, E) = \prod_i p(t_i \mid \theta, E, M_{\mathrm{ice}}, M_{\mathrm{DOM}}),$$

or equivalently,

$$-\ln L = \sum_i -\ln p(t_i \mid \theta, E, M_{\mathrm{ice}}, M_{\mathrm{DOM}}),$$

where $t_i$ are photon arrival times or time residuals in DOM $i$, $\nu_e$0 denotes the ice optical model, including depth-dependent absorption, scattering, and anisotropy, and $\nu_e$1 encodes DOM angular acceptance together with timing and charge calibration.

The hybrid strategy selects the most appropriate fit, kernel, and starting seeds across energies and topologies in order to reduce biases and local minima. The reconstruction uses updated calibrations and modern ice/DOM response models, including B-spline-based approximations for light propagation in layered, anisotropic ice. Although the reconstruction framework is used across tracks and cascades, IceCat-2 reports the track results.

Uncertainty estimation is performed by localized scans or approximations around the best fit to construct a directional posterior on the sphere. Reported contours are intended to have nominal frequentist coverage, meaning that a quoted $\nu_e$2 or $\nu_e$3 region contains the true direction in the corresponding fraction of repeated realizations. Achieving this requires a realistic photon timing and propagation model, robust treatment of systematics, and avoidance of underestimation from over-constrained fits.

3. Localization performance and statistical coverage

The catalog reports directional uncertainties through the containment radii $\nu_e$4 and $\nu_e$5 of the event’s directional posterior on the sphere and the corresponding containment areas $\nu_e$6. For small angles,

$\nu_e$7

A common local approximation models the directional posterior near its mode as a von Mises–Fisher distribution on $\nu_e$8,

$\nu_e$9

where $\nu_\tau$0 is the best-fit direction and $\nu_\tau$1 is the concentration. In the high-$\nu_\tau$2 limit, the distribution becomes approximately Gaussian on the tangent plane with $\nu_\tau$3, so that

$\nu_\tau$4

Relative to IceCat-1, the median $\nu_\tau$5 and $\nu_\tau$6 containment areas shrink by approximately $\nu_\tau$7 and $\nu_\tau$8, respectively, while the spread of the area distributions narrows by factors of roughly $\nu_\tau$9–$26.8\ \mathrm{yr}^{-1}$0. Catalog-wide medians in IceCat-2 are $26.8\ \mathrm{yr}^{-1}$1 and $26.8\ \mathrm{yr}^{-1}$2 (Zegarelli et al., 8 Jul 2025). Using the small-angle scaling $26.8\ \mathrm{yr}^{-1}$3, these area reductions correspond to typical radius improvements of about $26.8\ \mathrm{yr}^{-1}$4 for $26.8\ \mathrm{yr}^{-1}$5 and $26.8\ \mathrm{yr}^{-1}$6 for $26.8\ \mathrm{yr}^{-1}$7.

The reprocessing does not primarily move best-fit directions by large amounts. When IceCat-1 events are reconstructed again with the new method, the best-fit directions differ by less than $26.8\ \mathrm{yr}^{-1}$8 for $26.8\ \mathrm{yr}^{-1}$9 of events and less than $\theta$0 for $\theta$1 of events. This indicates that the principal gain is tighter and better-calibrated localization rather than wholesale displacement of event centroids.

In this context, statistical coverage is not merely a formal property of the posterior approximation. Poor coverage implies under-confident or over-confident contours; IceCat-2 is explicitly tuned for robust coverage, and the improved localization is presented together with improved agreement between nominal and realized containment.

4. Alert streams, selection metrics, and catalog contents

Real-time operations began in 2016 and were upgraded in 2019. The alert stream is partitioned into Gold and Bronze channels so as to balance rate against astrophysical probability. In IceCat-2, roughly 9.9 Gold and 17 Bronze alerts per year are included. Under an $\theta$2 prior, Gold and Bronze typically correspond to average astrophysical probabilities of about $\theta$3 and $\theta$4, respectively. Future versions are expected to revise these probabilities to reflect the softer muon-track spectrum measured by IceCube.

The signalness $\theta$5 is defined as the posterior probability that an event is astrophysical given its reconstruction features, including directional quality, energy proxy, topology, and zenith:

$\theta$6

The false alarm rate (FAR) is estimated by comparing a test statistic $\theta$7 to its background distribution:

$\theta$8

where $\theta$9 is the live time used to accumulate background trials. Thresholds are tuned to satisfy real-time rate goals.

Typical IceCat-2 event records mirror and extend the IceCat-1 fields with updated reconstructions and calibrations.

Field	Content
event_id	e.g., IC-YYMMDDX
time	UTC and MJD
coordinates	best-fit RA, Dec (J2000)
uncertainties	$E$0, $E$1 and/or $E$2, $E$3
stream	Gold or Bronze
signalness	average $E$4 for Gold, $E$5 for Bronze under $E$6
energy proxy	e.g., muon energy at detector or $E$7 proxy
reconstruction tag	September 2024 hybrid follow-up version
quality metrics	fit likelihood, TS, number of DOMs/hits
veto status	e.g., IceTop veto
localization product	contour polygon or HEALPix probability map

IceCat-2 remains a preliminary release pending peer-reviewed publication, after which a public data release is expected to follow. IceCat-1 is available through VizieR as a guide to content and formats, and IceCat-2 is intended to mirror and extend those fields with updated reconstruction outputs.

5. Cross-correlation analyses and revised source associations

A primary use of IceCat-2 is renewed cross-correlation against external source catalogs. For each alert, sources from 4FGL-DR4, 3FHL, 3HWC, TeVCat, and Swift-BAT are checked against the updated $E$8 containment contour. The expected number of chance coincidences is estimated by scrambling alert right ascensions 1000 times and repeating the count. An empirical $E$9-value may be written as

$$L(\theta, E) = \prod_i p(t_i \mid \theta, E, M_{\mathrm{ice}}, M_{\mathrm{DOM}}),$$0

or, if the background expectation $$L(\theta, E) = \prod_i p(t_i \mid \theta, E, M_{\mathrm{ice}}, M_{\mathrm{DOM}}),$$1 is known, by a Poisson tail,

$$L(\theta, E) = \prod_i p(t_i \mid \theta, E, M_{\mathrm{ice}}, M_{\mathrm{DOM}}),$$2

The observed coincidences are reported as consistent with the median expectation from chance in all tested catalogs. Illustrative examples are 4FGL-DR4, with 93 observed versus 89 expected, and Swift-BAT, with 35 observed versus 32 expected (Zegarelli et al., 8 Jul 2025). The reduced localization areas in IceCat-2 lower the expected number of chance associations relative to IceCat-1 and thereby sharpen future association tests.

The same logic can be cast in likelihood-ratio form:

$$L(\theta, E) = \prod_i p(t_i \mid \theta, E, M_{\mathrm{ice}}, M_{\mathrm{DOM}}),$$3

In this framework, the signal term evaluates the event directional PDF at the source location, while the background term uses local source densities or isotropic expectations. Smaller $$L(\theta, E) = \prod_i p(t_i \mid \theta, E, M_{\mathrm{ice}}, M_{\mathrm{DOM}}),$$4 improves signal-to-background contrast by reducing the chance-coincidence area.

Several previously discussed associations are re-evaluated under the tightened localizations. TXS 0506+056 remains inside the revised localization of IC-170922A, but it now lies within the $$L(\theta, E) = \prod_i p(t_i \mid \theta, E, M_{\mathrm{ice}}, M_{\mathrm{DOM}}),$$5 contour rather than the $$L(\theta, E) = \prod_i p(t_i \mid \theta, E, M_{\mathrm{ice}}, M_{\mathrm{DOM}}),$$6 contour. Two alerts associated with NGC 7469, IC-220424A and IC-230416A, remain spatially coincident within their updated contours, and the previously quoted $$L(\theta, E) = \prod_i p(t_i \mid \theta, E, M_{\mathrm{ice}}, M_{\mathrm{DOM}}),$$7 joint-coincidence context remains of interest. By contrast, the tidal disruption event candidates AT2019dsg, AT2019fdr, and AT2019aalc are now well outside the tightened IceCat-2 contours, effectively disfavoring those associations.

6. Multi-messenger implications, limitations, and future revisions

The tighter localizations have direct operational consequences for follow-up programs. With median $$L(\theta, E) = \prod_i p(t_i \mid \theta, E, M_{\mathrm{ice}}, M_{\mathrm{DOM}}),$$8 reduced to about $$L(\theta, E) = \prod_i p(t_i \mid \theta, E, M_{\mathrm{ice}}, M_{\mathrm{DOM}}),$$9, an instrument with a $$-\ln L = \sum_i -\ln p(t_i \mid \theta, E, M_{\mathrm{ice}}, M_{\mathrm{DOM}}),$$0 field of view typically requires $$-\ln L = \sum_i -\ln p(t_i \mid \theta, E, M_{\mathrm{ice}}, M_{\mathrm{DOM}}),$$1 pointings to cover the $$-\ln L = \sum_i -\ln p(t_i \mid \theta, E, M_{\mathrm{ice}}, M_{\mathrm{DOM}}),$$2 region, compared with $$-\ln L = \sum_i -\ln p(t_i \mid \theta, E, M_{\mathrm{ice}}, M_{\mathrm{DOM}}),$$3 pointings when using pre-2024 reconstructions with roughly $$-\ln L = \sum_i -\ln p(t_i \mid \theta, E, M_{\mathrm{ice}}, M_{\mathrm{DOM}}),$$4 larger areas. For wide-field X-ray and optical facilities with $$-\ln L = \sum_i -\ln p(t_i \mid \theta, E, M_{\mathrm{ice}}, M_{\mathrm{DOM}}),$$5 fields of view, this enables near-complete coverage in one or two tiles. Pointed facilities likewise benefit from higher on-source probability per pointing, so that follow-up effort can be concentrated more deeply rather than more broadly.

Programmatically, the combination of tighter localizations and improved coverage reduces ambiguous counterpart claims. Fewer chance coincidences and better-calibrated contours improve the reliability of both archival and real-time association tests (Zegarelli et al., 8 Jul 2025).

Important limitations remain. Angular resolution still depends on energy, geometry, and ice/DOM systematics. Some events may retain elongated or multi-modal posteriors, and rare reconstruction failure modes can persist. The current Gold/Bronze signalness values are still tied to an $$-\ln L = \sum_i -\ln p(t_i \mid \theta, E, M_{\mathrm{ice}}, M_{\mathrm{DOM}}),$$6 assumption; IceCube plans to revise them using the softer muon-track spectrum measured by the experiment. IceCat-2 itself is preliminary, and a full public release with DOIs, stable data formats, analysis notebooks, probability maps or contours, and explicit reconstruction versioning is planned following peer-reviewed publication.

A broader implication is that the improved catalog can both rule out earlier associations and strengthen future ones. The disfavoring of some TDE candidates already illustrates the former. The latter remains plausible, but any claim of new association will still require explicit treatment of multiple testing and look-elsewhere effects, including penalties associated with scanning across catalogs, epochs, and source classes.