First Double Cascade Tau Neutrino Candidates in IceCube and a New Measurement of the Flavor Composition

Juliana Stachurska (for the IceCube Collaboration)

REVIEW 3 major objections 3 minor 15 references

First Double Cascade Tau Neutrino Candidates in IceCube and a New Measurement of the Flavor Composition

T0 review · 3 major / 3 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read IceCube reports its first two double-cascade events as astrophysical tau-neutrino candidates.

desk verdict First IceCube tau-candidate double cascades are a real result, but the event-level tauness numbers are not as robust as the headline suggests; the paper is honest about that, and it deserves a serious referee. read the letter →

arxiv 1908.05506 v1 pith:NHTYLWSB submitted 2019-08-15 astro-ph.HE hep-ex

Juliana Stachurska (for the IceCube Collaboration) This is my paper · ORCID

classification astro-ph.HEhep-ex

keywords tauneutrinodoublecascadeIceCubeflavorcompositionHigh-EnergyStartingEventsastronomyCherenkovdetector

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

This paper reports two events in IceCube's High-Energy Starting Events sample that are reconstructed as double cascades: two separated light depositions connected by a dim track, the signature of a tau neutrino interacting via charged current and its tau lepton decaying. It calls these the first astrophysical tau neutrino candidates and assigns them a "tau-ness" of about 75% for the event named Big Bird and greater than 97% for Double Double. Using a three-topology classification of the 7.5-year sample, it also measures the flavor composition on Earth, finding a best fit of $\nu_e:\nu_\mu:\nu_\tau = 0.29:0.50:0.21$, consistent with both the 1:1:1 expectation from pion decay and with a zero astrophysical tau component. The interest is that a confirmed double cascade breaks the degeneracy between $\nu_e$ and $\nu_\tau$ that limits lower-energy flavor measurements.

What carries the argument

The central object is the double cascade topology: a $\nu_\tau$ charged-current interaction produces a first cascade at the interaction vertex and a second when the short-lived tau lepton decays, separated by a track whose length scales as roughly 50 meters per PeV of tau energy. Because the mean sensor spacing in IceCube is 125 meters, this topology becomes resolvable only above about 100 TeV. The paper identifies double cascades with a maximum-likelihood fit and three observables: the double cascade length, the energy asymmetry $A_E = (E_1 - E_2)/(E_1 + E_2)$, and the energy confinement (the fraction of light deposited near the cascade vertices), with cuts of length greater than 10 meters, $A_E < 0.30$, and confinement in $[0.99, 1.0]$. To convert candidate events into probabilities, the paper resimulates each event in a restricted parameter space and uses a Gaussian-kernel density estimator with the Rodeo algorithm to estimate the ratio of $\nu_\tau$-CC to non-$\nu_\tau$-CC differential rates at the observed observables, defining the "tau-ness" $\tau$.

What would settle it

Generate enough Monte Carlo events in Double Double's region of interest to resolve its tau-ness beyond the current greater-than-97% lower bound; if the value drops below 97%, or if recalibrating the single-cascade energy-asymmetry tail raises the expected background at Big Bird's observables ($A_E = 0.29$) above roughly 25%, the double-cascade interpretation would be in doubt.

Watch

Extended reading notes

Core claim

The paper's central claim is that two IceCube events, Big Bird and Double Double, are the first events classified as double cascades in the High-Energy Starting Events sample, and that both are strong candidates for astrophysical tau neutrinos. It states explicitly: "We have for the first time classified two of the IceCube data events as double cascades, thus finding promising candidates for astrophysical tau neutrinos." The paper quantifies this with an a posteriori analysis: under the best-fit spectrum and a 1:1:1 flavor prior, the probability that Big Bird came from a $\nu_\tau$ charged-current interaction is approximately 75%, while for Double Double it is greater than 97%. It also performs the first flavor-composition measurement using all three event topologies, with a best-fit composition of $\nu_e:\nu_\mu:\nu_\tau = 0.29:0.50:0.21$.

Load-bearing premise

The result stands or falls on whether the Monte Carlo simulation faithfully reproduces the rare tails in which a true single cascade, mostly from an electron neutrino, is reconstructed with a large separation and a large energy asymmetry; if those tails are wrong, the two events may not be tau neutrinos at all.

Editorial extensions

If this is right

If the two events are true $\nu_\tau$ charged-current interactions, IceCube has detected astrophysical tau neutrinos for the first time and broken the $\nu_e$/$\nu_\tau$ degeneracy that limits lower-energy flavor measurements.
The three-topology flavor fit gives a best-fit composition of $\nu_e:\nu_\mu:\nu_\tau = 0.29:0.50:0.21$, consistent with the 1:1:1 expectation from pion decay and also with zero astrophysical tau neutrinos, so the two candidates alone do not yet force a nonzero tau component.
The expected correlation between event energy and double cascade length for $\nu_\tau$ interactions becomes a testable relation as more double cascade events accumulate.
Double Double sits in a signal-dominated region while Big Bird sits near the edge of a background-dominated region, indicating that the double-cascade channel is viable but still limited by Monte Carlo statistics and background reconstruction tails.

Reading between the lines

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

If these are true tau neutrinos, the physical consequence is that the astrophysical neutrino flux retains a tau component after propagation over cosmic distances, implying flavor mixing from the source; the paper does not spell this out, but it is the direct implication of a confirmed $\nu_\tau$ charged-current event.
A testable extension would be to treat the two double cascades as a small sample of tau energy measurements: the tau decay length in each event should track the reconstructed energy, so a future third double cascade with high energy but short length would be hard to fit into the standard picture.
The paper's own statement that Monte Carlo statistics cap Double Double's tau-ness at greater than 97% points to the decisive next step: a larger dedicated resimulation of that event's region could push the lower bound above 99%, or reveal a background tail that lowers it.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

What you should know: this is the paper that classifies two IceCube HESE events as double cascades and calls them tau neutrino candidates, plus a flavor fit using three topologies. That is genuinely new. The double cascade selection was proposed before and applied to simulations, but this is the first time events in IceCube data are classified that way and given per-event tauness values. The flavor fit breaking the e/tau degeneracy is also a step forward relative to earlier single-cascade + track analyses.

The paper does several things well. The candidate events are described concretely with their reconstructed observables. The authors explicitly say Big Bird sits close to the energy asymmetry cut in a background-dominated region, and that MC statistics limit the Double Double tauness to a lower bound. They did a dedicated resimulation in a restricted parameter space to get better PDFs, and they quote the method's assumptions. That is the right way to present a candidate-level result.

The soft spots are real but not hidden. The tauness numbers come from Eq. 3.3 with a prior taken from the collaboration's own best-fit spectrum and flavor composition, so they are not model independent. Uncertainties in the fit spectrum and flavor composition are not propagated into tauness. The resimulation omitted down-going events that could contribute a few percent to signal and background. Big Bird's asymmetry 0.29 is 0.01 below the cut, and the MC in that region is sparse. Double Double's tauness >97% is a lower bound, not a measured value. Under small errors in the background PDF, the 75% for Big Bird is not stable. None of this is an internal contradiction; the authors state these limitations. But the headline claim 'first tau neutrino candidates' is stronger than the demonstrated support, as the stress-test note says.

Citation pattern looks fine: this is IceCube internal work, citing prior IceCube flavor papers and the tau search methods. No invented entities. No code/data public, but that's normal for conference proceedings.

Who is this for: neutrino physicists and anyone tracking IceCube's path to tau neutrino discovery. The specific tauness numbers should not be quoted without caveats. The paper deserves a serious referee, but the referee should ask for propagated systematics, a less self-referential prior, and validation of the MC tails, or a clear statement that these numbers are placeholder candidates.

Recommendation: accept for peer review as a candidate-level result, with major revisions or follow-up paper. It is worth engaging with.

Referee Report

3 major / 3 minor

Summary. The paper reports on IceCube's 7.5-year High-Energy Starting Events (HESE) sample, for which an algorithmic topology identifier classifies events into single cascades, tracks, and double cascades. The authors present two events, dubbed 'Big Bird' and 'Double Double', as the first double-cascade candidates and therefore as promising astrophysical tau-neutrino candidates. They perform a flavor-composition fit over the three topology samples, obtaining a best-fit composition of nu_e:nu_mu:nu_tau = 0.29:0.50:0.21, consistent with both the 1:1:1 expectation and a zero astrophysical nu_tau component. An a posteriori resimulation is used to estimate the 'tauness' of each candidate: approximately 75% for Big Bird and greater than 97% for Double Double, with the latter limited by Monte Carlo statistics. The paper explicitly acknowledges that fit-spectrum and flavor-composition uncertainties are not yet propagated into the tauness values, and that down-going events were omitted from the resimulation. The central claim is that these two events constitute the first double-cascade tau-neutrino candidates, rather than a definitive detection.

Significance. If the interpretation holds, this is the first plausible astrophysical tau-neutrino signal and the first flavor-composition measurement using three event topologies in IceCube, with implications for neutrino source models and beyond-standard-model physics. The paper is transparent about its limitations: it discloses the MC-statistics limits, the omitted down-going events, and the use of a prior from the same collaboration's best-fit. The two complementary tau searches mentioned in Section 4 that also find Double Double provide useful cross-checks. However, the quantitative tauness values are presented without systematic uncertainties and depend on assumptions, so the result is best regarded as a candidate-level finding that needs further scrutiny before being treated as a robust measurement.

major comments (3)

[Section 3.1, Eq. (3.3)] The tauness definition in Eq. (3.3) uses the differential expected rates N_nu_tau P_nu_tau and N_non-nu_tau P_non-nu_tau, where the N values are taken from a simulation normalized to the best-fit flavor composition with the 1:1:1 assumption from reference [3]. Because [3] is a fit to the same HESE dataset, the tauness values are prior-dependent rather than model-independent, and Section 3.2 states that uncertainties in the fit spectrum and flavor composition are not propagated. For Big Bird, which lies in a background-dominated region (Section 2, Fig. 3), a modest systematic shift in the background normalization could change the reported 75% substantially. The quantitative claims in Section 4 ('at most one event for every 30' and 'one out of every four') should therefore be presented with an explicit caveat that they carry unquantified systematic uncertainty and are conditional on the assumed prior.
[Section 3.2] The reported tauness for Double Double is '& 97%' because 'the statistics of the generated MC are not sufficient to evaluate the tauness to a higher precision.' This is thus a lower bound set by limited Monte Carlo statistics, not a measured confidence interval. The same section also notes that the resimulation omitted very down-going events that could contribute up to about 3% to the signal and about 6% to the single-cascade background classified as double cascades, and Section 3 states that the restricted primary-energy ranges were determined by trial and error. These limitations mean the background estimate in the double-cascade tail is not yet robust enough to support the strength of the candidate claim without an additional systematic study, particularly because both events sit near the selection thresholds.
[Section 2, Tables 1 and 4] Big Bird's energy asymmetry of 0.29 is only 0.01 below the selection cut of 0.30, in a region where true single cascades dominate the background (Fig. 3). The double-cascade classification therefore relies on the Monte Carlo simulation reproducing the tail of the reconstructed energy-asymmetry distribution for single cascades. The paper does not provide a validation of these tails against data or a systematic uncertainty on the cut efficiency. Given also the ~2 m length resolution and the 10 m length cut, with both events at lengths of 16-17 m, a small error in the simulated tails could change the classification of one or both events. I recommend adding closure tests or explicitly framing the candidate status as conditional on the MC tail model.

minor comments (3)

[Tables 3 and 4] Both tables share the identical caption 'Observables of the two Double Cascades', which is confusing; the second table should have a distinct caption, and the in-text reference to 'Table4' (missing space) should be corrected.
[Section 3.1] In Eq. (3.1), the non-nu_tau terms appear with a placeholder or rendering artifact ('P_nu_tau' and 'N_non-nu_tau'); please ensure the notation is typeset correctly.
[Figure 2] The figure contains the visible label 'WORK IN PROGRESS'; if this label refers to the flavor-composition measurement, it should be removed or explained in the caption, since the text presents the result as final.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: tauness is an explicitly prior-dependent posterior and the flavor composition is a direct fit.

full rationale

The paper's two main outputs are (i) a ternary-topology flavor-composition fit and (ii) per-event 'tauness' values for the two double-cascade candidates. The flavor fit is a direct maximum-likelihood fit of f_alpha to the three topology samples; the contours in Fig. 2 are fit outputs, and no fitted parameter is renamed as a prediction. The tauness posterior in Eq. (3.3) is explicitly computed under stated assumptions, namely 'for the best-fit spectrum and 1:1:1 flavor composition as given in [3]', so it is a conditional probability with an acknowledged prior, not a model-independent measurement and not a quantity extracted from the same fit. The double-cascade classification rests on independent maximum-likelihood reconstructions (single cascade, track, double cascade) and fixed cuts; the two candidate events are not used to define the topology class. The acknowledged limitations -- Big Bird's asymmetry 0.29 sits just below the 0.30 cut, Double Double's tauness is MC-statistics-limited at >97%, and down-going events and spectral/flavor uncertainties are not propagated -- are robustness and correctness concerns about the conditional claim, not circular reductions. Citations [3], [4], [5], [14], and [15] are collaboration self-citations, but they supply priors, algorithmic context, and complementary searches rather than a uniqueness theorem or a fitted target result. No equation in the paper is equivalent to its input by construction.

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

The central claim rests on fitted flavor fractions, hand-set double cascade cuts, and assumed inputs from the same collaboration's best-fit in reference [3] (spectrum and 1:1:1 flavor prior). No new entities are introduced. The tau-ness estimate is therefore not a self-contained derivation from external benchmarks.

free parameters (4)

Flavor fractions f_e, f_mu, f_tau = 0.29, 0.50, 0.21 (best fit)
The three flavor fractions are the free parameters of the binned maximum likelihood fit to the 60 HESE events, with constraint f_e + f_mu + f_tau = 1 (Section 2). They are fitted to the data, not derived.
Double cascade selection cuts = length >10 m; energy asymmetry -0.98 to 0.30; energy confinement 0.99 to 1.0
Hand-set thresholds in Table 1 that define which events are classified as double cascades. The central claim of two double cascade events depends on these choices; Big Bird sits 0.01 below the asymmetry cut.
Best-fit astrophysical spectrum parameters from reference [3] = not stated in this paper (single power law, best fit from [3])
The tau-ness values are evaluated at the best-fit spectrum from reference [3], a fit to the same HESE sample. These are inputs, not derived here, and their uncertainties are not propagated.
Prior nu_tau content in the double cascade topology subsample = taken from best-fit with 1:1:1 composition in [3]
Used in Equation 3.3 as N_nu_tau / (N_nu_tau + N_non-nu_tau). This is a fitted and assumed quantity from the same collaboration's analysis, not an external calibration.

assumptions (6)

standard math Wilks' theorem is used to draw flavor composition contours.
The contours in Figure 2 rely on Wilks' theorem; the asymptotic approximation is standard but not verified in the small-sample regime here.
domain assumption Pion decay in astrophysical sources produces roughly 1:1:1 flavor composition on Earth.
Used as the reference expectation and as the composition assumed for the tau-ness prior; this paper does not test it.
domain assumption IceCube detector simulation accurately models reconstruction tails and topology classification.
All PDFs, background estimates, and tau-ness values come from Monte Carlo; no data based closure test is shown.
ad hoc to paper The tau-ness prior is taken from the best-fit flavor composition with a 1:1:1 assumption in reference [3].
Adopted in Equation 3.3 to evaluate the posterior; it is an input from the same collaboration's analysis of the same HESE sample.
ad hoc to paper Restricted resimulation ranges cover the relevant parameter space, and down-going events can be neglected.
The authors state the simulation did not include very down-going events, which could contribute about 3% to signal and 6% to single cascade background classified as double cascades.
ad hoc to paper Primary neutrino energy ranges for resimulation, determined by trial and error, are adequate.
Section 3 says the primary energy range had to be determined by trial and error; if too narrow, the tau-ness KDE would be biased.

how reviews work

0 comments

Cite this review

Pith. "Pith review of First Double Cascade Tau Neutrino Candidates in IceCube and a New Measurement of the Flavor Composition." pith.science (2026). https://pith.science/paper/NHTYLWSB

@misc{pith2026190805506,
  author       = {Pith},
  title        = {Pith review of: First Double Cascade Tau Neutrino Candidates in IceCube and a New Measurement of the Flavor Composition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NHTYLWSB}},
  note         = {Machine review of arXiv:1908.05506}
}

abstract

The IceCube Neutrino Observatory at the South Pole, which detects Cherenkov light from charged particles produced in neutrino interactions, firmly established the existence of an astrophysical high-energy neutrino component. The expected neutrino flavor composition on Earth is $\nu_e:\nu_{\mu}:\nu_{\tau}$ of about 1:1:1 for neutrinos produced in astrophysical sources through pion decay. A measurement of the flavor composition on Earth can provide important constraints on sources and production mechanisms within the standard model, and can also constrain various beyond-standard-model processes. Here the measurement of the flavor composition performed on IceCube's High-Energy Starting Events sample with a livetime of about 7.5 years is presented. IceCube is directly sensitive to each neutrino flavor via the single cascade, track and double cascade event topologies. In IceCube, $\nu_{\tau}$-CC interactions above $\sim$ 100 TeV can produce resolvable double cascades, breaking the degeneracy between $\nu_e$ and $\nu_{\tau}$ present at lower energies. IceCube's first two identified double cascades are presented and the properties of the two $\nu_{\tau}$ candidates are discussed.

Figures

Figures reproduced from arXiv: 1908.05506 by the authors.

**Figure 2.** Measured flavor composition of IceCube HESE events with ternary topologyID and sensitivity at the best fit spectrum. Contours obtained using Wilks’ theorem [9]. likelihood fit is performed on the three topology samples using two-dimensional PDFs obtained from Monte Carlo simulations, as shown in [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** Distribution of the reconstructed energy asymmetry in the double cascade sample split by true topology for the best-fit astrophysical and atmospheric spectra based on simulation. The two double cascades (“Double Double” at -0.80, “Big Bird” at 0.29) are shown. 3. Detailed study of candidate events The MC statistics available at the time of the unblinding were absolutely insufficient to incorporate the full informat… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

15 extracted references · 11 canonical work pages

[3]

IceCube Collaboration, PoS(ICRC2019)1004 (these proceedings)
[1]

IceCube Collaboration, M. G. Aartsen et al., JINST 12 (2017) P03012

2017
[2]

IceCube Collaboration, M. G. Aartsen et al., Science 342 (2013)

work page 2013
[4]

IceCube Collaboration, PoS(ICRC2017)974 (2018)

work page 2018
[5]

Stachurska, https://doi.org/10.5281/zenodo.1301122, (2018)

IceCube Collaboration, J. Stachurska, https://doi.org/10.5281/zenodo.1301122, (2018)

work page doi:10.5281/zenodo.1301122 2018
[6]

IceCube Collaboration, M. G. Aartsen et al., JINST 9 (2014) P03009

work page 2014
[7]

Ahrens et al., Nucl

AMANDA Collaboration, J. Ahrens et al., Nucl. Instrum. Meth. A524 (2004) 169–194

2004
[8]

Patrick Hallen, On the Measurement of High-Energy Tau Neutrinos with IceCube, Master’s thesis, RWTH Aachen, 2013

work page 2013

Show all 15 references

[9]

S. S. Wilks, Annals Math. Statist. 9 (1938) 60–62

1938
[10]

C. A. Argüelles, A. Schneider and T. Yuan, Journal of High Energy Physics 2019 (2019) 30

2019
[11]

IceCube Collaboration, M. G. Aartsen et al., Astrophys. J. 809 (2015) 98

2015
[12]

IceCube Collaboration, M. G. Aartsen et al., Phys. Rev. D99 (2019) 032004

2019
[13]

H. Liu, J. Lafferty, and L. Wasserman in Proceedings of the Eleventh International Conference on Artiﬁcial Intelligence and Statistics, vol. 2, pp. 283–290, (2007)

2007
[14]

IceCube Collaboration, PoS(ICRC2019)960 (these proceedings)
[15]

IceCube Collaboration, PoS(ICRC2019)1036 (these proceedings). 7

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [3]

IceCube Collaboration, PoS(ICRC2019)1004 (these proceedings)

[2] [1]

IceCube Collaboration, M. G. Aartsen et al., JINST 12 (2017) P03012

2017

[3] [2]

IceCube Collaboration, M. G. Aartsen et al., Science 342 (2013)

work page 2013

[4] [4]

IceCube Collaboration, PoS(ICRC2017)974 (2018)

work page 2018

[5] [5]

Stachurska, https://doi.org/10.5281/zenodo.1301122, (2018)

IceCube Collaboration, J. Stachurska, https://doi.org/10.5281/zenodo.1301122, (2018)

work page doi:10.5281/zenodo.1301122 2018

[6] [6]

IceCube Collaboration, M. G. Aartsen et al., JINST 9 (2014) P03009

work page 2014

[7] [7]

Ahrens et al., Nucl

AMANDA Collaboration, J. Ahrens et al., Nucl. Instrum. Meth. A524 (2004) 169–194

2004

[8] [8]

Patrick Hallen, On the Measurement of High-Energy Tau Neutrinos with IceCube, Master’s thesis, RWTH Aachen, 2013

work page 2013