Pith. sign in

REVIEW 2 major objections 6 minor 10 references

All-flavor Time-dependent Search for Transient Neutrino Sources

T0 review · 2 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Combining IceCube's tracks, starting tracks, and cascades into a single time-dependent likelihood improves transient-source sensitivity by 10-15%, and shows that power-law searches can miss short cutoff-spectrum flares.

desk verdict Credible methods paper from IceCube on an all-flavor time-dependent search; the 10–15% gain and cutoff-recovery caution hold up, but the unstated assumption that the three samples are disjoint needs to be made explicit. read the letter →

arxiv 2507.08775 v1 pith:2JS6RUGK submitted 2025-07-11 astro-ph.HE

Jose Carpio , Ali Kheirandish , Hans Niederhausen (for the IceCube Collaboration) This is my paper · ORCID
classification astro-ph.HE
keywords time-dependentneutrinosearchall-flavorIceCubetransientsourcespoint-sourcelikelihoodpower-lawspectrumenergycutoffscascadeevents
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper develops the first all-flavor, all-sky time-dependent IceCube point-source search, merging throughgoing tracks, starting tracks, and cascade events into one likelihood. It claims that for $E^{-2}$ power-law sources at the horizon this combination improves the 90% CL sensitivity and 3-$\sigma$ evidence potential by roughly 10-15% over tracks alone, with the gain largest for flares shorter than about a day. It also introduces a two-sided cutoff spectrum with low- and high-energy exponential cutoffs as a signal hypothesis, and shows that a flare whose true flux has this shape, injected at the 3-$\sigma$ level, is recovered at only about 1-$\sigma$ by a standard power-law fit when the flare lasts less than about a day. If right, previous and ongoing power-law searches could be missing the most plausible transient neutrino signals, and the combined three-sample likelihood is a stronger tool for finding them.

What carries the argument

The engine is an unbinned likelihood $\ln L = \sum_i \ln\left[\frac{n_s}{N}S_i + \left(1-\frac{n_s}{N}\right)B_i\right]$, where the signal pdf $S_i$ factorizes into spatial, energy, and temporal pdfs; spatial pdfs are two-dimensional Gaussians (with a Kernel Density Estimate version for the NT sample), the temporal pdf is a Gaussian centered at $T_0$ with width $\sigma_T$, and the energy pdf encodes the assumed spectrum. The test statistic $\mathrm{TS} = 2\ln\left[\frac{\hat\sigma_T}{T_\mathrm{live}}\frac{L(\hat n_s,\hat\gamma,\hat T_0,\hat\sigma_T)}{L(n_s=0)}\right]$ penalizes short flares by their trial factor. Sensitivity and evidence potential are obtained by scrambling data for background trials and injecting simulations for signal trials. The new spectral ingredient is the two-sided cutoff flux $e^{-E_L/E}\,E^{-2}\,e^{-E/E_H}$, fit with parameters $n_s$, $E_L$, $E_H$, $T_0$, and $\sigma_T$.

What would settle it

Count, in the actual event lists used for the trials, how many events satisfy the selection criteria of more than one of the three samples (throughgoing tracks, starting tracks, cascades), then rerun the sensitivity calculation with any overlaps removed; if overlaps exist, the reported 10-15% improvement is at least partly an artifact of double-counting. The cutoff-spectrum claim can be checked separately by injecting a 3-sigma two-sided cutoff flare with half-width 0.1 days into simulated detector data and measuring its recovered significance under a power-law hypothesis.

Watch

Extended reading notes

Core claim

The central claim is that combining IceCube's three event topologies, throughgoing muon tracks (NT), starting tracks (ESTES), and cascades (DNN), into a single time-dependent likelihood yields the best all-sky sensitivity for transient sources, with a 10-15% improvement in 90% CL sensitivity and 3-$\sigma$ evidence potential for $E^{-2}$ sources at declination zero compared to tracks alone. The analysis also claims that a power-law hypothesis is inadequate for signals with low- and high-energy cutoffs: a 3-$\sigma$ two-sided cutoff flux peaking near 100 TeV is recovered at roughly 1-$\sigma$ by a power-law fit for flares with half-width below about 1 day, because the power-law fit tends to prefer $\gamma > 2$ and biases the fitted number of signal events. The paper frames this as a warning that the standard $E^{-\gamma}$ parametrization can miss the very transient sources IceCube is chasing.

Load-bearing premise

The analysis assumes that no single neutrino event can pass more than one of the three sample selections, so multiplying the sample likelihoods counts every event exactly once.

Editorial extensions

If this is right

  • If the combined three-sample likelihood works as reported, it becomes the natural configuration for future IceCube time-dependent point-source searches and can be applied to real data without methodological changes.
  • Short-duration flares benefit most from the combination, because the merged sample approaches the background-free regime for flares shorter than about 0.1 days.
  • Transient sources with cutoff spectra peaking near 100 TeV can be missed by a power-law fit, so future searches should fit low- and high-energy cutoffs for individual source candidates.
  • The improvement is driven partly by cascades and starting tracks extending sensitivity in the Southern Sky, giving a genuinely all-sky transient search rather than one dominated by throughgoing tracks.
  • Significance claims based on power-law fits to short flares should be treated as lower bounds, since a true cutoff-spectrum flare can be recovered at roughly 1-sigma when its injected strength is at the 3-sigma level.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper does not state whether the three event selections are mutually exclusive; if a single event can pass more than one selection, the product likelihood would count it multiple times and inflate the reported sensitivity gain, so an explicit overlap audit is needed.
  • The cutoff-spectrum result suggests that archival IceCube flare candidates, including blazar and tidal-disruption-event alerts, could be re-evaluated with low- and high-energy cutoff hypotheses to see whether their reported power-law significances underestimate the true signal.
  • A natural extension is a joint fit that leaves the power-law index free while also allowing cutoffs, which would test whether the recovered-significance loss persists when the spectral shape is less constrained.
  • The same all-flavor time-dependent likelihood could be adapted for joint neutrino and electromagnetic time-domain searches, since the temporal pdf is generic and the spatial and energy pdfs can be reweighted for any external alert.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 6 minor

Summary. This paper presents the first all-flavor, all-sky time-dependent point-source search in IceCube, combining throughgoing tracks (NT), starting tracks (ESTES), and cascades (DNN). The analysis uses an unbinned likelihood (Eq. 1) with spatial, energy, and temporal pdfs, a test statistic with a trial-factor penalty (Eq. 2), and scrambled-data background trials with simulated signal injection to compute 90% CL sensitivities and 3-sigma evidence potentials. For E^-2 sources at the horizon the combined analysis improves sensitivity and evidence potential by 10-15% relative to tracks alone for flares shorter than about a day. The paper also implements a two-sided cutoff signal model (Eq. 3) and shows that a 3-sigma two-sided cutoff flux is recovered at only about 1-sigma by a power-law fit for flares with sigma_T < 1 day.

Significance. If the sample-disjointness condition is confirmed, this is a meaningful methodological result: it is the first demonstration that combining all IceCube event topologies improves time-dependent point-source sensitivity across the sky. The analysis follows the standard IceCube trial framework, with background from scrambled data and signal from simulation, and the definitions of sensitivity and evidence potential are explicit. The two-sided cutoff recovery study is a useful caution for interpreting power-law searches and is quantified across flare durations. The manuscript is a compact proceedings contribution; it needs a few technical clarifications, but the core methodology is sound and the reported improvements are plausible.

major comments (2)
  1. [Sec. 2, Eq. (1)] The total likelihood is written as a product over all events, which is statistically valid only if the three event selections (NT, ESTES, DNN) are mutually exclusive and the event sets are statistically independent. The manuscript does not state that the selections are disjoint, and the descriptions of the samples do not rule out that a single physical event could pass more than one selection and be counted twice. Since every sensitivity and evidence-potential curve in Figures 1-3 is built on this combined likelihood, the reported 10-15% improvement could be inflated if any overlap exists. Please state explicitly how the samples are made disjoint (e.g., event-level exclusion after reconstruction) or quantify the overlap and propagate it through the likelihood.
  2. [Sec. 3.2, Fig. 3] The conclusion that a power-law fit can 'miss' a two-sided cutoff signal is based on injecting the 3-sigma evidence flux for the two-sided cutoff hypothesis and then fitting with a power law. For flares with sigma_T less than about 1 day, the injected signal is only 3-4 events, and the authors attribute the loss to a bias in the fitted n_s. This is plausible, but the claim would be more robust if the recovered significance were shown as a function of injected flux for a fixed flare width, so that the reader can see how quickly the recovery approaches 3-sigma as the flux increases. As written, the 'miss' is quantified at only one injected flux point, which makes the headline statement somewhat sensitive to the chosen definition of evidence potential.
minor comments (6)
  1. [Abstract and Sec. 1] The phrase 'unbound E^-gamma power-law sources' should read 'unbroken E^-gamma' or 'unbounded E^-gamma'; 'unbound' has a different meaning.
  2. [Sec. 3.1] The text says the reported quantity is 'fluence (time-integrated fluxes) as Delta T dN/dE, evaluated at 1 TeV,' but Figure 1's axis label is 'Delta T E^2 dN/dE | 1TeV'; please align the text and the axis definition.
  3. [Sec. 2] The sentence describing the KDE enhancement of the NT sample is ambiguous about whether the KDE enters the spatial pdf, the energy pdf, or both; please clarify.
  4. [Sec. 2] For the combined dataset, the choice of sigma_T,max as half the longest livetime deserves a one-sentence justification, since the individual samples may have different livetimes.
  5. [Sec. 3.2] The statement that the TS distributions are not significantly different between signal hypotheses would be easier to assess if the distributions, or a quantitative comparison, were shown.
  6. [Sec. 4] To substantiate the 'first all-flavor, all-sky time-dependent search' claim, please cite explicitly that the earlier searches [4,5] used tracks only, or state how this search differs from them.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the sensitivity and evidence curves are Monte-Carlo simulation outputs from stated signal models, and the two-sided-cutoff recovery study is an honest model-mismatch cross-check.

full rationale

I walked the derivation chain in Sections 2 and 3. The likelihood in Eq. (1) is a standard unbinned product over events, and the signal pdf is explicitly factorized into spatial, energy, and temporal terms, with Gaussian temporal and spatial assumptions and a KDE enhancement cited to [9]. No parameter is fitted to data and then renamed as a prediction. The 90% CL sensitivity and 3-sigma evidence potential are defined via injected simulated signals against scrambled background TS distributions: "We define the 90% CL sensitivity as the signal flux required for the signal TS to be larger than the background median 90% of the time." These are outputs of Monte Carlo trials, not the same quantities they claim to report. The recovered-significance study in Section 3.2 injects a flux that is 3-sigma under the two-sided cutoff hypothesis and then evaluates the same events under a power-law hypothesis; the paper reports that the signal is recovered at only about 1-sigma for short flares. This is a model-mismatch check, not a circular argument, because the two hypotheses have distinct signal pdfs and the comparison is computed with newly generated TS distributions. The only structural worry is the implicit mutual exclusivity and independence of the three event samples in the product likelihood, which could affect statistical validity if violated, but that is a correctness assumption rather than a circular reduction: no equation equates a claimed result to an input by construction. Citations [8] and [9] provide the likelihood formalism and KDE method and are prior methodology, not an imported uniqueness theorem that forces the conclusions. The paper's quantitative claims rest on its own simulations and trial definitions, not on a self-citation chain, so the circularity score is 0.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The central sensitivity claims rest on standard LLH machinery, the KDE enhancement, the Gaussian time profile, the scrambling-based background, and the trial-factor penalty. The analysis has four fitted search parameters (n_s, gamma, sigma_T, E_L/E_H) whose ranges and priors shape the result. No new particles or physical entities are introduced. The most fragile premise is the assumed independence and disjointness of the three event samples, which is not explicitly justified.

free parameters (4)
  • Signal normalization n_s = fit parameter (n_s >= 0)
    Maximized in the likelihood; defines the signal strength.
  • Power-law spectral index gamma = fit parameter, gamma in [1,4]
    For the power-law hypothesis; the recovery study attributes the missed significance to overfitting gamma > 2.
  • Flare half-width sigma_T = fit parameter, 0 to half livetime
    Scanned in the search; also enters the trial-factor penalty sigma_T/T_live in Eq. 2, so the reported sensitivity depends on the prior over sigma_T.
  • Cutoff energies E_L and E_H = E_L in [100 GeV, 10 TeV], E_H in [10 TeV, 100 PeV]
    Fitted for the two-sided cutoff hypothesis; the recovery study injects E_L = 1 TeV, E_H = 100 TeV.
assumptions (6)
  • standard math Unbinned two-component Poisson likelihood (Eq. 1) with product over events
    Foundational TS from Braun et al. [8]; the paper uses it directly without re-derivation.
  • domain assumption Signal spatial PDF is a 2D Gaussian whose width is the per-event reconstruction error, or a KDE sample
    Assumes Gaussian reconstruction errors and that the KDE, taken from ref [9], is a faithful estimator of the spatial/energy PDF.
  • domain assumption Temporal PDF is a Gaussian centered at T0 with width sigma_T
    Flares are assumed to be Gaussian in time; sensitivity is optimized for this shape and may differ for other temporal profiles.
  • domain assumption Background is built by scrambling data times
    Assumes background arrival times are uniform and stationary; the analysis relies on scrambled data for TS distributions.
  • standard math Trial-factor penalty sigma_T/T_live in Eq. 2 is the correct look-elsewhere correction
    Adopted from Braun et al. [8]; this penalty directly shapes the TS and the reported sensitivity.
  • domain assumption The three event samples are independent and non-overlapping
    The total likelihood is a product over all events; the paper does not state whether an event can pass multiple selections.

how reviews work

0 comments
Cite this review

Pith. "Pith review of All-flavor Time-dependent Search for Transient Neutrino Sources." pith.science (2026). https://pith.science/paper/2JS6RUGK

@misc{pith2026250708775,
  author       = {Pith},
  title        = {Pith review of: All-flavor Time-dependent Search for Transient Neutrino Sources},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2JS6RUGK}},
  note         = {Machine review of arXiv:2507.08775}
}
abstract

Transient sources are among the preferred candidates for the sources of high-energy neutrino emission. Intriguing examples so far include blazar flares and tidal disruption events coincident with IceCube neutrinos. Here, we report the first all-flavor, all-sky time-dependent search for neutrino sources by combining IceCube throughgoing tracks, starting tracks and cascades. Throughgoing tracks provide the best sensitivity in the Northern Sky, while cascades have worse angular resolution but yield better sensitivity in the Southern Sky than tracks. The relatively new starting tracks sample has reduced contamination from atmospheric muons. This analysis takes advantage of the strengths of each of the datasets, combining them for increased statistics and obtaining the best accessible all-sky sensitivity for transient searches. In this search, we look for unbound $E^{-\gamma}$ power-law sources, as well as $E^{-2}$ sources with low and high-energy exponential cutoffs, optimizing the sensitivity for the duration of the flares.

Figures

Figures reproduced from arXiv: 2507.08775 by the authors.

Figure 1
Figure 1. Top left panel: Per-flavor 90% CL sensitivity (solid lines) and 3𝜎 evidence (dashed lines) fluences for an 𝐸 −2 source at declination 𝛿 = 0 ◦ , for different flare half-widths 𝜎𝑇 and 𝐸 −𝛾 signal flux hypothesis. The blue, red, orange and black lines correspond to DNNCascade, NT, ESTES and combined datasets, respectively. Here, the NT sample assumes Gaussian spatial pdfs. Top right panel: Same as top left panel, but … view at source ↗
Figure 2
Figure 2. Per-flavor 90% CL sensitivity (solid lines) and 3𝜎 evidence potential (dashed lines) fluences to a neutrino flare of half-width 𝜎𝑇. The blue, orange and black lines correspond to DNNCascade, ESTES and combined datasets, respectively. 3. Analysis performance 3.1 Power-law fits In this section, our signal hypothesis to calculate TS will be a point source with an 𝐸 −𝛾 power-law spectrum, where 𝛾 ∈ [1, 4], where the fit… view at source ↗
Figure 3
Figure 3. Recovered significance in power-law fits after injecting the 3𝜎 evidence flux for two-sided cutoffs, for different flare half-widths 𝜎𝑇. Left (right) panel corresponds to the DNNCascade (NT) sample. Hence, the loss in significance is mostly tied to the reconstruction of the signal pdf parameters. The power-law hypothesis tends to fit for 𝛾 > 2, which causes an overfitting of the signal parameter 𝑛𝑠. In the case of f… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

10 extracted references · 3 canonical work pages

  1. [1]

    Murase and I

    K. Murase and I. Bartos, http://dx.doi.org/10.1146/annurev-nucl-101918-023510 Ann. Rev. Nucl. Part. Sci. 69 (2019) 477--506

    work page doi:10.1146/annurev-nucl-101918-023510 2019

  2. [2]

    IceCube Collaboration, M. G. Aartsen et al. , http://dx.doi.org/10.1126/science.aat2890 Science 361 no. 6398, (2018) 147--151

    doi:10.1126/science.aat2890 2018

  3. [3]

    IceCube Collaboration, M. G. Aartsen et al. , http://dx.doi.org/10.1103/PhysRevLett.124.051103 Phys. Rev. Lett. 124 no. 5, (2020) 051103

    doi:10.1103/physrevlett.124.051103 2020

  4. [4]

    Abbasi et al

    IceCube Collaboration, R. Abbasi et al. , http://dx.doi.org/10.3847/1538-4357/ad3730 Astrophys. J. 967 no. 1, (2024) 48

    work page doi:10.3847/1538-4357/ad3730 2024

  5. [5]

    Abbasi et al

    IceCube Collaboration, R. Abbasi et al. , http://dx.doi.org/10.3847/2041-8213/ac2c7b Astrophys. J. Lett. 920 no. 2, (2021) L45

    doi:10.3847/2041-8213/ac2c7b 2021

  6. [6]

    IceCube Collaboration, R. Naab, E. Ganster, and Z. Zhang, `` Measurement of the astrophysical diffuse neutrino flux in a combined fit of IceCube's high energy neutrino data ,'' in 38th International Cosmic Ray Conference . http://arxiv.org/abs/2308.00191 arXiv:2308.00191 [astro-ph.HE]

    arXiv

  7. [7]

    Abbasi et al

    IceCube Collaboration, R. Abbasi et al. https://arxiv.org/abs/2406.07601

    arXiv

  8. [8]

    Braun, M

    J. Braun, M. Baker, J. Dumm, C. Finley, A. Karle, and T. Montaruli, http://dx.doi.org/10.1016/j.astropartphys.2010.01.005 Astropart. Phys. 33 (2010) 175--181

    work page doi:10.1016/j.astropartphys.2010.01.005 2010

Show all 10 references
  1. [9]

    Abbasi et al

    IceCube Collaboration, R. Abbasi et al. , http://dx.doi.org/10.1126/science.abg3395 Science 378 no. 6619, (2022) 538--543

    2022 doi

  2. [10]

    write newline

    " write newline "" before.all 'output.state := FUNCTION blank.sep after.quote 'output.state := FUNCTION fin.entry output.state after.quoted.block = 'skip 'add.period if write newline FUNCTION new.block output.state before.all = 'skip output.state after.quote = after.quoted.blo...

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.