GFU Cluster Alert System

• The GFU Cluster Alert System is a real-time framework that detects spatial and temporal clusters of muon-neutrino events to flag potential neutrino flares.
• It couples a time-dependent clustering algorithm with a curated source list and multi-threshold significance testing to differentiate signal from background.
• The system disseminates alerts via NASA GCN and a web platform, enabling timely multimessenger follow-up by gamma-ray, X-ray, and other observatories.

The Gamma-ray Follow-Up (GFU) Cluster Alert System is the IceCube real-time alert stream dedicated to identifying temporally and spatially clustered excesses of muon-neutrino–candidate events—operationally, neutrino flares—from predefined source directions and, in related modes, from the wider sky. It belongs to IceCube’s multimessenger program, which exploits full-sky coverage, $\sim 99\%$ uptime, and low-latency processing to turn neutrino information into triggers for electromagnetic facilities. In its contemporary form, the system couples a time-dependent clustering algorithm to a curated target list and is transitioning from private distribution to Imaging Air Cherenkov Telescopes (IACTs) toward public dissemination through NASA’s General Coordinates Network (GCN) and a continuously updated web platform (Meneguolo et al., 8 Jul 2025, Mancina et al., 10 Jul 2025).

1. Programmatic role and historical scope

Within IceCube’s alert ecosystem, GFU occupies the branch devoted to clustered activity rather than isolated, exceptionally energetic events. The program issues alerts through two main strategies: one selects neutrino candidates with exceptionally high energy, while the other identifies spatial and temporal clusters of neutrino events; the cluster-alert literature explicitly focuses on the latter strategy (Meneguolo et al., 8 Jul 2025). This distinguishes GFU cluster alerts from the public single-event Gold and Bronze streams, even though the broader IceCube real-time framework uses overlapping classes of high-quality track-like events (Blaufuss et al., 2019).

The broader GFU program has shared cluster alerts with partner IACTs since 2006, and later descriptions characterize the cluster stream as one of the longest-operating neutrino follow-up programs, operating since 2012 (Meneguolo et al., 8 Jul 2025, Satalecka et al., 2021). In the private IACT phase, GFU-cluster alerts were distributed under memoranda of understanding to H.E.S.S., MAGIC, and VERITAS since 2019, and to LST-1 since 2023 (Meneguolo et al., 8 Jul 2025). That operational history established GFU as a neutrino-selected target-of-opportunity system for very-high-energy gamma-ray follow-up, while more recent work generalizes the same infrastructure into a broader multimessenger service (Mancina et al., 10 Jul 2025).

A recurrent source of confusion is the term “Gamma-ray Follow-Up” itself. Historically it reflected the program’s IACT-centered use case, but the current system is no longer restricted to gamma-ray–bright blazars or to gamma-ray observatories alone. The updated source-selection strategy explicitly incorporates X-ray–bright active galactic nuclei and Galactic systems, reflecting the possibility that some neutrino-emitting environments are gamma-ray opaque rather than gamma-ray loud (Meneguolo et al., 8 Jul 2025).

2. Cluster-detection methodology and statistical logic

GFU operates on a real-time stream of track-like events selected by an online reconstruction and filtering chain optimized for muon neutrinos. In the public-cluster description, the GFU filter yields high-quality track-like events with direction resolution “on the scale of $1^\circ$,” and the events are transferred from the South Pole to the northern computing center for the cluster analysis (Mancina et al., 10 Jul 2025). Earlier descriptions of the real-time multimessenger framework give typical alert latencies of $30$–$40$ s for IceCube online processing, which sets the basic timing scale for the arrival of events into GFU (Meneguolo et al., 8 Jul 2025).

For each incoming event, GFU evaluates an event-level signal-to-background measure with respect to a source hypothesis. In the public formulation this preliminary event weight is written as

$$w_i=\mathcal{S}(\vec{x}_{\mathrm{src}},\vec{x}_i,\sigma_i,E_i,\gamma=-3)\,/\,\mathcal{B}(\theta_i,\phi_i,E_i),$$

where the signal term combines spatial consistency with the source direction and an energy term based on a power-law spectrum, while the background term is data-driven (Mancina et al., 10 Jul 2025). In the source-list formulation summarized in the target-list paper, the same idea is stated operationally: if any tested source yields $S/B>1$, the likelihood analysis is triggered (Meneguolo et al., 8 Jul 2025). This threshold defines an analysis-triggering event.

Once triggered, the algorithm scans retrospective time windows ending at the current event and extending back up to 180 days. The best window is selected by maximizing a time-dependent likelihood over the number of signal events, the spectral index, and the flare start time; in the public system, the end time $t_1$ is fixed by the current analysis-triggering event, and the candidate start times $t_0$ are restricted to previous analysis-triggering events within the 180-day lookback (Mancina et al., 10 Jul 2025). The best cluster is then assigned a test statistic and a pre-trial $p$-value using background-only trials constructed from time-scrambled data (Mancina et al., 10 Jul 2025).

GFU runs in both source-list and all-sky modes. The all-sky mode evaluates many more trial positions and therefore incurs a much larger trials factor. The source-list mode instead restricts the hypothesis space to a finite set of monitored objects and thereby “significantly improve[s] the discovery potential” (Meneguolo et al., 8 Jul 2025). This statistical asymmetry has direct operational consequences. In the older IACT-centered implementation, source-list alerts used thresholds around $3\sigma$–$1^\circ$0, while all-sky alerts required $1^\circ$1 pre-trial because of the larger number of tested directions (Satalecka et al., 2021). In the public platform, the source-list mode is organized around three pre-trial thresholds—low, medium, and high—at $1^\circ$2, $1^\circ$3, and $1^\circ$4, corresponding to $1^\circ$5 background alerts per year across the catalog, slightly more than $1^\circ$6 background alert per year, and $1^\circ$7 false alert per 20 years, respectively (Mancina et al., 10 Jul 2025).

A further interpretive caution is explicit in the public-alert literature: the displayed significance curves are not flux light curves. They plot cluster $1^\circ$8-value against time, adjacent points are strongly correlated, and the ordinates are pre-trial rather than catalog-corrected significances (Mancina et al., 10 Jul 2025).

3. Target-source architecture and the unified source list

The defining astrophysical component of the modern GFU cluster system is its unified source list. Earlier GFU source lists were telescope-specific and optimized for IACT detectability; the new compilation removes IACT-specific observability constraints and is maintained by IceCube according to criteria focused on IceCube sensitivity and scientific relevance (Meneguolo et al., 8 Jul 2025). The current published list is built from GeV-motivated and keV-motivated selections, with a TeV-motivated selection still under construction (Meneguolo et al., 8 Jul 2025).

The update reflects both statistical and astrophysical arguments. Statistically, monitoring a few hundred predefined coordinates sharply reduces the trials factor relative to an all-sky pixel scan. Astrophysically, the source classes were broadened in response to evidence from TXS 0506+056 and NGC 1068: the first reinforced interest in gamma-ray–bright blazars, while the second motivated the inclusion of X-ray–bright, gamma-ray–opaque environments such as Seyfert galaxies (Meneguolo et al., 8 Jul 2025).

Selection	Input catalog or basis	Sources
GeV $1^\circ$9-rays (AGN)	4FGL/4LAC-DR3	200
keV X-rays (AGN)	BASS II	44
keV X-rays (Galactic)	Swift/BAT 157-month	10
Microquasars	Model-driven	12

Accounting for overlaps, these subsamples produce 259 unique monitored objects; a future TeV-driven addition is expected to bring the total to $30$0 (Meneguolo et al., 8 Jul 2025). The composition includes 40 Seyfert galaxies, 19 Galactic sources—particularly binary systems and microquasars—and the Crab Nebula; four AGN are common to both the GeV and keV selections: NGC 1275, Mrk 421, 3C 454.3, and 3C 273 (Meneguolo et al., 8 Jul 2025).

Source ranking is based on a figure of merit,

$30$1

where $30$2 is the catalog flux in the relevant electromagnetic band and $30$3 is IceCube’s one-year point-source sensitivity at the source coordinates, using a proxy sensitivity curve for a spectrum $30$4 with $30$5 (Meneguolo et al., 8 Jul 2025). For the GeV list, the ranked quantity is the $30$6–$30$7 GeV integral photon flux from 4LAC-DR3; photon flux is preferred over energy flux so as not to bias the ranking toward hard-spectrum sources already favored by the TeV-based selection under construction (Meneguolo et al., 8 Jul 2025). For the BASS II AGN selection, the intrinsic $30$8–$30$9 keV X-ray flux is used; for Galactic Swift/BAT sources, the observed $40$0–$40$1 keV flux is used; the microquasar subset is model-driven rather than FoM-ranked (Meneguolo et al., 8 Jul 2025).

The new list is also geometrically different from its predecessors. Only 34% of the sources in the earlier GFU program remain in the new selection, because the updated list is driven by IceCube sensitivity, which favors the horizon region and the Northern sky, rather than by the observability constraints of individual IACTs (Meneguolo et al., 8 Jul 2025).

4. Public transition, muting, and dissemination infrastructure

For most of its operational life, GFU-cluster alerting was private. In that regime, once a source crossed the alert threshold and an alert was sent, the source was muted: no further updates were distributed until the cluster significance fell back below threshold (Mancina et al., 10 Jul 2025). The explicit rationale was twofold. First, muting helped preserve blindness for offline analyses using related catalogs. Second, it reduced alert spam for partner telescopes (Mancina et al., 10 Jul 2025).

The public platform changes this model fundamentally. The new system removes muting and makes all above-threshold information available, so that the subsequent behavior of an active source is no longer obscured (Mancina et al., 10 Jul 2025). In the new terminology, a source is “active” if its most recent cluster exceeds the low threshold of $40$2; every later above-threshold cluster then updates the source’s significance curve and alert status in real time (Mancina et al., 10 Jul 2025).

Public dissemination proceeds through two coupled channels. The first is NASA GCN. GFU notices are designed to carry the minimal operational information required for follow-up: source name and equatorial position, the time $40$3 of the analysis-triggering event that generated the cluster, the best-fit flare start time $40$4, and the false alarm rate corresponding to the cluster significance (Mancina et al., 10 Jul 2025). The multi-threshold scheme determines whether the system sends an initial notice, a stronger update, or—at medium and high thresholds—a human-readable GCN Circular (Mancina et al., 10 Jul 2025).

The second is a dedicated public website implemented with the Python Flask framework. The site contains an active-alerts page listing currently active sources, an archival page listing previous alerts, and source-detail pages showing aliases, coordinates, a dynamic significance curve, and the full alert history for each source (Mancina et al., 10 Jul 2025). These source pages are intended to carry the longitudinal information that would be too bulky for compact GCN notices.

A related refinement introduced in the target-list update is source tagging. When GFU alerts are distributed via GCN, the notices communicate the selection that motivated inclusion of the source—GeV AGN, X-ray AGN, Galactic hard-X-ray source, microquasar, and eventually TeV selection—so that different observatories can apply their own follow-up priorities (Meneguolo et al., 8 Jul 2025). This tagging formalizes the shift from a private IACT trigger stream to a general-purpose multimessenger alert service.

5. Multimessenger follow-up and scientific use

GFU was created as a bridge between IceCube’s all-sky neutrino monitoring and the narrow fields of view and limited duty cycles of high-energy photon facilities. In the historical IACT phase, the final decision to observe was taken independently by each telescope collaboration and typically depended on a combination of flare properties—such as false alarm rate and duration—and local visibility and weather constraints (Satalecka et al., 2021). Typical source-list follow-ups were therefore target-of-opportunity observations of known gamma-ray or X-ray emitters, not blind scans of a large error region.

The principal IACT partners have been H.E.S.S., MAGIC, VERITAS, and later LST-1 (Meneguolo et al., 8 Jul 2025). Descriptions of the upgraded private GFU system from 2019–2020 report 27 GFU alerts from 17 distinct sources, with 7 sources followed up by at least one IACT; no significant very-high-energy counterpart was found, and no changes in source flux levels or spectra were detected relative to previous observations (Satalecka et al., 2021). Those null results are operationally important: they show that GFU is not merely an alert generator but also a source of targeted upper limits constraining contemporaneous gamma-ray activity.

The scientific scope, however, is now broader than gamma-ray follow-up alone. The inclusion of X-ray–bright AGN, binaries, microquasars, and NGC 1068–like systems is motivated by the possibility that some neutrino sources are optically thick to GeV–TeV photons and reprocess their radiative power to the keV–MeV band (Meneguolo et al., 8 Jul 2025). This implies that X-ray facilities, optical telescopes, and other observatories may in some cases be better matched than IACTs to the physics of a given GFU alert.

A continuous monitor such as Fermi-LAT occupies a complementary role in this ecosystem. In seven years of follow-up to IceCube real-time high-energy neutrino alerts, the LAT strategy has been to search immediately for gamma-ray activity from known and newly detected sources consistent with the neutrino localization, using time windows from one day to the full mission (Garrappa et al., 2024). That approach is conceptually parallel to GFU’s source-list logic: both systems reduce a large background by conditioning on source hypotheses and time structure, but LAT operates on gamma-ray data after a neutrino trigger, whereas GFU operates on neutrino data before the electromagnetic response.

6. Historical evolution, interpretation, and open issues

The present GFU cluster system is the product of successive redesigns. Early IceCube neutrino-triggered target-of-opportunity work was source-list–based and IACT-driven, with source selections requiring strong variability, $40$5, sufficient elevation at a given telescope, and an extrapolated $40$6 GeV flare flux compatible with that telescope’s $40$7 discovery potential in $40$8–$40$9 h (Collaboration et al., 2016, Satalecka et al., 2021). The post-2019 upgrade broadened the sky coverage, added an all-sky cluster stream, and aligned the event stream more closely with the high-quality track selections used in public IceCube alerts (Satalecka et al., 2021). The 2025 source-list revision then replaced telescope-specific criteria with the unified, IceCube-centered ranking described above (Meneguolo et al., 8 Jul 2025).

Several limitations are explicit in the recent literature. The flare search only considers durations up to 180 days, so longer-lived activity is not represented as a single cluster in the real-time analysis (Mancina et al., 10 Jul 2025). The public significance curves remain pre-trial quantities for individual clusters rather than catalog-corrected discovery significances, and their temporal points are highly correlated (Mancina et al., 10 Jul 2025). The initial public implementation focuses on source-list alerts; public all-sky GFU alerts are still described as under development because of their much higher false alarm rate and the harder localization problem (Mancina et al., 10 Jul 2025).

The public transition also modifies the epistemic status of the system. Muting had previously protected blindness for offline analyses using overlapping source catalogs; the new platform abandons that concealment in favor of richer public information (Mancina et al., 10 Jul 2025). This does not invalidate the alert stream, but it changes how offline analyses must account for prior knowledge of source behavior.

Finally, the literature identifies several directions for future development rather than fixed roadmaps. The source list is expected to evolve as new neutrino associations emerge, IceCube sensitivity estimates change, and the follow-up infrastructure expands (Meneguolo et al., 8 Jul 2025). Public all-sky alerts are planned after the source-list release, and joint real-time cluster alerts involving additional neutrino detectors such as KM3NeT, Baikal-GVD, and IceCube-Gen2 are presented as a plausible next step for multimessenger neutrino astronomy (Mancina et al., 10 Jul 2025). This suggests a system whose core logic—a time-dependent cluster search on high-quality tracks—has stabilized, while its target ontology, dissemination model, and networked role within multimessenger astronomy remain actively in development.