Neutrino Telescopes

Principles, Results, and Prospects

Dmitry Naumov

A new window on the Universe

Francis Halzen.

Francis Halzen.

2026 Nobel Prize in Physics

IceCube and the discovery of high-energy astrophysical neutrinos.

Dmitry Naumov to Francis Halzen:
“Let me warmly congratulate you and the entire IceCube Collaboration on this thoroughly deserved Nobel Prize in Physics! … I am Dmitry Naumov — a Baikalist, from JINR (Dubna). We have met several times over the years, most recently at the JUNO cafeteria, where we had lunch together and I even ‘predicted’ your Nobel Prize.”

Francis Halzen to Dmitry Naumov:
“Dmitry, I know you since the Naumov flux and remember the lunch in the JUNO cafeteria very well. Thanks for your very kind message.”

IceCube, 6 October 2026

\(1\,\mathrm{PeV} = 10^{15}\,\mathrm{eV}\)

in a single neutrino

a billion times a neutrino from the Sun

over a hundred times a proton in the LHC

Not nuclear fusion. Then what? And where?

Sensor positions: IceCube geometry · light pattern modelled · Bert and Ernie: IceCube 2013 · Big Bird: IceCube 2014, position illustrative

History of Neutrino Telescopes

A telescope inside a natural medium

Moisei Markov speaking at a lectern.

Moisei Markov.

A large target already exists in nature.

  • Place light sensors deep in a lake or sea.
  • Use Cherenkov light to reconstruct charged particles produced by neutrinos.

M. A. Markov, Rochester conference, 1960, p. 578.

An earlier portrait of Moisei Markov from the INR anniversary archive.

Earlier portrait, date unspecified. INR RAS archive.

Moisei Markov on the left with Bruno Pontecorvo on the right.

Markov (left) and Bruno Pontecorvo (right), later photograph, undated. IceCube historical archive.

Frederick Reines at left and Clyde Cowan at the Savannah River experiment in 1956.

Frederick Reines (left) and Clyde Cowan, 1956. Los Alamos National Laboratory.

Frederick Reines, later portrait reproduced in Spiering's historical review.

Later portrait, undated. Reproduced in C. Spiering’s historical review.

Neutrino detection first; cosmic-neutrino astronomy next.

Kenneth Greisen, portrait reproduced in Spiering's historical review.

Kenneth Greisen. Undated portrait via C. Spiering.

Greisen, Reines and Markov discussed cosmic-neutrino detection in 1960.

Markov’s underwater proposal supplied a practical route to a very large target.

The physics was compelling. Could the instrument survive and work?

DUMAND: the first deep-ocean attempt

John Learned at the blackboard, Jed Hirota seated at left and Dick Davisson seated at right, DUMAND 1976 workshop.

DUMAND workshop, 1976. John Learned at the board; Jed Hirota (left), Dick Davisson (right). Photo: Anatoly Petrukhin, via IceCube.

Deep Underwater Muon And Neutrino Detector

  • Reines, Learned and colleagues pursued an ocean array near Hawaii.
  • A detector several kilometres below the surface needed entirely new engineering.

1987: a short prototype string, lowered from a ship, recorded muons in the deep ocean.

Summer 1979: the DUMAND workshop meets in Khabarovsk and at Lake Baikal.

  • Chudakov, Berezinsky, Bezrukov, Dolgoshein, Petrukhin and Zheleznykh plan the deep-ocean telescope together with the American team.
  • December 1979: Soviet troops enter Afghanistan.
  • 1980: the American side ends the cooperation.

“…if perchance they included Russians it would be found that no funding was available.”

Arthur Roberts, DUMAND

1980: Chudakov proposes Lake Baikal for a “Russian DUMAND”.

C. Spiering, Eur. Phys. J. H 37, 515 (2012), arXiv:1207.4952, section 4.1; A. Roberts, Rev. Mod. Phys. 64, 259 (1992).

4800 m below the Pacific. Nearly 500 atmospheres.

  • December 1993: the first string is connected on the sea floor. It sees muons.
  • Over a hundred penetrators lead cables into the pressure housings. One is flawed.
  • Seawater seeps in. About ten hours later: a short circuit.
  • A fuse fails to clear it. The shore loses contact.
  • January 1994: an acoustic signal releases the string from the sea floor.
  • 1995: support ends. The array is never built.

In the deep ocean, one connector can decide the fate of an experiment.

DUMAND Collaboration, astro-ph/9412019; C. Spiering, Eur. Phys. J. H 37, 515 (2012), arXiv:1207.4952.

Lake Baikal: the ice becomes a platform

Winter work at Lake Baikal, in a photograph dated 2005 by the DLNP archive.

Winter expedition, 2005. DLNP JINR archive.

Alexander Chudakov proposed the site.

  • Deep, clear water close to shore.
  • A frozen surface for deploying and maintaining the detector.

Alexander Chudakov in 1946, from the laboratory photo archive.

Alexander Chudakov, 1946.

Alexander Chudakov in 1985, from the laboratory photo archive.

Alexander Chudakov, 1985.

Both dates follow the INR Laboratory of High Energy Leptons photo archive.

Grigory Domogatsky, early portrait labelled 1980 in the original Keynote image filename.

Grigory Domogatsky, 1980

Grigory Domogatsky, later portrait published with the 2006 Markov Prize announcement.

Later portrait

Published with the 2006 Markov Prize announcement; photograph date unspecified.

Led the Baikal neutrino programme from 1980.

Christian Spiering.

Christian Spiering

His Zeuthen group joined in 1988.

  • Scientists, engineers and winter expeditions turned the proposal into a working array.
  • The 2006 Markov Prize recognised Domogatsky and Spiering for NT200+ and high-energy neutrino astrophysics.

INR 2006 prize announcement.

NT36, 1993

Three strings and 36 optical modules at Lake Baikal.

“Congratulations for winning the 3-string race!”

Francis Halzen to Christian Spiering

NT36, DUMAND’s TRIAD and AMANDA: competing routes to a working array.

1980–84A dedicated laboratory, site studies and early strings 1993–94NT36: early neutrino candidates 1996NT96: a clear upward-going neutrino signal 1998NT200 completed with 192 optical modules

Atmospheric neutrinos demonstrated that Markov’s method worked in a lake.

C. Spiering, historical review (2024), slides 29–31; Baikal detector paper.

The path to a cubic kilometre

German AMANDA collaborators with optical modules, from the collaboration's historical archive.

AMANDA collaborators and optical modules. AMANDA Collaboration; photograph undated.

Halzen, Learned, Price, Morse, Barwick and colleagues explored polar ice.

  • Early shallow deployments suffered strong scattering from bubbles.
  • Deeper ice made directional reconstruction possible.

Deployment of the final IceCube optical module in December 2010.

Final optical module, December 2010. IceCube/NSF, via the IceCube timeline.

1999: the cubic-kilometre proposal.

2010: the last of 86 strings.

2013: an astrophysical neutrino flux.

The IceCube construction team in the 2004–2005 season, in front of a drilling tower.

First IceCube construction season, 2004–2005. IceCube/NSF historical archive.

Artist's illustration of ANTARES and the Nautile submersible, not a photograph.

Artist’s illustration of ANTARES and Nautile. ANTARES material via IceCube.

2008: the 12-line detector is complete.

  • Long-term operation demonstrated precision neutrino detection in the deep sea.
  • ANTARES, NEMO and NESTOR supplied experience for KM3NeT.

Detection Principles

Transparent Universe, opaque Earth

A \(1\,\mathrm{PeV}\) muon neutrino in today’s mean baryonic matter

\[ \bar n_{b,0}\simeq0.25\,\mathrm{m^{-3}} =2.5\times10^{-7}\,\mathrm{cm^{-3}} \]

\[ \sigma_{\nu N}^{\mathrm{CC+NC}} \simeq9.5\times10^{-34}\,\mathrm{cm^2} \]

About one nucleon per four cubic metres.

\[ \lambda_{\nu N,0}=\frac{1}{\bar n_{b,0}\sigma_{\nu N}} \simeq4.2\times10^{39}\,\mathrm{cm} \]

\[ \lambda_{\nu N,0}\simeq4.4\times10^{21}\,\mathrm{ly} \]

Observable diameter today: \(D_{\rm obs}\simeq93\times10^9\,\mathrm{ly}\).

\[ \frac{\lambda_{\nu N,0}}{D_{\rm obs}}\simeq5\times10^{10} \]

Present-epoch \(\nu N\) estimate ignoring the Universe evolution. PDG baryon density; CSMS cross sections, Table I.

Ratio of the neutrino flux reaching a detector 1 km deep to the flux entering the Earth, versus zenith angle from 90 to 180 degrees and energy from 10 GeV to 100 PeV, with neutral-current regeneration. Near the horizon the Earth is transparent at all energies; through the centre half of the flux survives up to about 20 TeV and 10 percent up to about 170 TeV. Dashed contours without regeneration lie slightly lower in energy.

Full transport: V. A. Allakhverdyan (CT10nlo, PREM, detector 1 km deep), redrawn · V. A. Allakhverdyan, dissertation (2026), Fig. 4.20 · Allakhverdyan, Zavyalov & Naumov (2026), Fig. 9

\[ 1\,\mathrm{PeV}=10^6\,\mathrm{GeV} \]

Horizontal axis: zenith angle.

\(90^\circ\): near the horizon.

\(180^\circ\): through Earth’s centre.

Colour: flux at the detector over flux entering the Earth.

Dashed: without regeneration,

\[ e^{-N_A\,\sigma_{\rm tot}(E)\,X(\theta)}. \]

Neutral currents do not remove a neutrino; they move it to lower energy. With regeneration the ratio depends on the incident spectrum.

\[ \lambda_{\nu N}(\rho)=\frac{1}{N_A\rho\sigma_{\nu N}} \simeq3.2\times10^3\,\mathrm{km} \left(\frac{5.5\,\mathrm{g\,cm^{-3}}}{\rho}\right). \]

Earth’s diameter is \(12742\,\mathrm{km}\); its dense core matters.

\[ X_{\rm diam}=\int_{\rm diameter}\rho(\ell)\,d\ell \simeq1.1\times10^{10}\,\mathrm{g\,cm^{-2}} \quad\text{(PREM)}. \]

\[ \tau_{\rm diam}=N_A\sigma_{\nu N}X_{\rm diam}\simeq6.3, \qquad P_0=e^{-\tau_{\rm diam}}\simeq1.8\times10^{-3}\simeq0.2\%. \]

  • Almost no PeV neutrinos cross the full diameter without interacting.
  • Near the horizon, the path is much shorter.

\(P_0\): zero CC or NC interactions, not the regenerated flux fraction \(\mathcal T\). PREM column depth: Gandhi et al.

One year of IceCube up-going muons weighs the Earth

  • IC86, 2011–2012: 20 145 up-going muons, 0.4–20 TeV; the zenith distribution follows absorption in the Earth.
  • Earth’s mass: \(M_\oplus^\nu=6.0^{+1.6}_{-1.3}\times10^{24}\,\mathrm{kg}\); gravity gives \(5.9722\times10^{24}\,\mathrm{kg}\).
  • Core mass: \(M_{\rm core}^\nu=2.72^{+0.97}_{-0.89}\times10^{24}\,\mathrm{kg}\).
  • Moment of inertia: \((6.9\pm2.4)\times10^{37}\,\mathrm{kg\,m^2}\); gravity gives \(8.017\times10^{37}\).
  • The core is denser than the mantle: a denser mantle has \(p=0.011\).

Weak interactions alone, without gravity, see the Earth’s dense core.

68% credible intervals, default atmospheric-flux model · A. Donini, S. Palomares-Ruiz, J. Salvado, Nature Physics 15, 37 (2019), arXiv:1803.05901

Left: CC survival probability against zenith angle at 10, 30 and 100 TeV for PREM (solid) and a homogeneous Earth of the same mass (dashed); paths through the core are darker in PREM. Right: relative difference between PREM and the homogeneous Earth over energy and zenith angle.

  • Same mass, different profile: PREM against a homogeneous Earth of \(5.51\,\mathrm{g\,cm^{-3}}\).
  • Paths through the core (beyond \(147^\circ\)) lose \(\sim20\%\) more neutrinos at \(30\,\mathrm{TeV}\); mantle paths lose fewer.

Computed: PREM, CSMS cross sections, detector 1 km deep · CC absorption only; NC and \(\tau\) regeneration not included

Tracks and cascades

A track — KM3NeT

KM3NeT event KM3-230213A: PMT signals along a reconstructed long muon trajectory. The displayed cone is an artistic illustration.

A long muon path through the array.

KM3NeT Collaboration, KM3-230213A.

A cascade — IceCube

IceCube Big Bird recorded cascade: a compact concentration of optical-module signals.

A compact shower compared with module spacing.

IceCube Collaboration / IceCube–NSF, Big Bird.

Measured PMT signals; reconstructed track; illustrative cone. © KM3NeT. Official event display.

IceCube event displays of Bert and Ernie: hundreds of optical modules lit by two PeV cascades.

Bert and Ernie, the first PeV cascades: deposited energies \(1.04\pm0.16\) and \(1.14\pm0.17\,\mathrm{PeV}\). IceCube Collaboration / IceCube–NSF · IceCube (2013)

Muon-neutrino interaction, 1 TeV · NTSim/Geant4 secondaries, LightHit radiative transfer in water · glow = expected photoelectrons, colour = first-light time · not sampled hits, not observed data · light front illustrative

Electron-neutrino interaction, 100 GeV · NTSim/Geant4 secondaries, LightHit radiative transfer in water · glow = expected photoelectrons, colour = first-light time · not sampled hits, not observed data · light front illustrative

Cosmic Sources

Where jets get their power

Computed debris orbits, schematic magnetic engine.

  • Gas feeds an accretion disc around a supermassive black hole.
  • Magnetic fields can extract rotational energy and carry it into a jet.
  • The gas supply can sustain activity beyond a single stellar disruption.
  • Blazars: the jet points close to our line of sight.

Blandford, Meier & Readhead, Relativistic Jets in AGN

Scenario Fuel for the central engine
Massive-star collapse Stellar material falls onto the compact remnant.
Compact-object merger Debris forms a temporary accretion disc.
  • Accretion onto a black hole or the spin of a strongly magnetized neutron star can power a relativistic outflow.
  • The engine operates during a transient event.

P. Mészáros, Gamma-Ray Bursts

  • Charged cosmic rays bend in magnetic fields.
  • Photons can scatter or be absorbed.
  • Neutrinos can escape dense environments and preserve directional information.

How jets produce neutrinos

Scenario and calculation: Barkov, Aharonian & Bosch-Ramon (2010) · The film illustrates the idea: simplified gas model, schematic protons and neutrinos

Accelerated protons collide with nuclei in dense gas.

\[ pp\longrightarrow\pi^{\pm},\,\pi^0+X \]

Possible targets: a stripped stellar envelope or a gas cloud in the jet.

The gas density and the proton residence time control the interaction yield.

Wang et al., jet–cloud/star scenario

Accelerated protons interact with a radiation field.

\[ p\gamma\longrightarrow\pi^{\pm},\,\pi^0+X \]

Possible targets: synchrotron photons in the jet, or photons from the accretion disc and its surroundings.

The pion-production threshold depends on photon energy and collision angle.

Murase & Stecker, High-Energy Neutrinos from AGN

The jet supplies energy. The star creates a shock and a dense target.

\[ r_L=\frac{p}{ZeB}\lesssim R \]

  • A source must retain particles long enough to accelerate them.
  • Energy losses and the available time also limit the maximum energy.

Schematic of one converter acceleration cycle: a proton in a relativistic jet becomes a neutron, leaves the jet without feeling the magnetic field, becomes a proton again outside and is turned back into the jet by the field, gaining energy of order Gamma squared. Each conversion also emits a neutrino. Paths are illustrative.

Charge changes let particles cross the field · Gain \(\sim\Gamma^2\) per cycle · Derishev et al. (2003)

Neutrinos from pion decay

\[pp\ \text{or}\ p\gamma\ \longrightarrow\pi^\pm,\,\pi^0+X\]

Charged pion and muon

\[ \begin{aligned} \pi^+&\to\mu^++\nu_\mu\\[0.5em] \mu^+&\to e^++\nu_e+\bar\nu_\mu \end{aligned} \]

Neutral pion

\[\pi^0\to\gamma+\gamma\]

Neutrinos and photons share the pion-production energy budget.

Charge-conjugate chain for \(\pi^-\) · Ahlers & Halzen (2018)

\[ E_p\ \xrightarrow{\ \sim1/5\ } E_\pi \ \xrightarrow{\ \sim1/4\ } E_\nu \qquad\Longrightarrow\qquad E_p\sim20E_\nu \]

Neutrino Characteristic parent proton
\(10\,\mathrm{TeV}\) \(\sim200\,\mathrm{TeV}\)
\(100\,\mathrm{TeV}\) \(\sim2\,\mathrm{PeV}\)
\(1\,\mathrm{PeV}\) \(\sim20\,\mathrm{PeV}\)

Same frame · Negligible pion/muon cooling · A scale estimate, not an event-by-event conversion

\[pp:\quad N_{\pi^+}:N_{\pi^-}:N_{\pi^0}\simeq1:1:1\]

\[ E_\nu^2\Phi_\nu(E_\nu)\simeq\frac32 E_\gamma^2\Phi_\gamma(E_\gamma), \qquad E_\gamma\simeq2E_\nu \]

At production: \(pp\) benchmark, all neutrino flavours and both signs.

Photon absorption and cascading modify the comparison with observed spectra.

Ahlers & Halzen (2018), Eq. 3

Three messengers, one energy budget

Gamma rays, neutrinos and the most energetic cosmic rays carry similar power.

Energy flux E squared times flux versus particle energy from 0.1 GeV to 200 EeV. Fermi-LAT isotropic gamma-ray background points from 0.1 to 820 GeV, the IceCube astrophysical neutrino spectrum per flavour from about 14 TeV to 5 PeV with its 68 percent band, and Pierre Auger cosmic-ray points above 3 EeV. A dashed band shows the Waxman-Bahcall bound per flavour, 0.45 to 2.25 times 1e-8; the neutrino flux reaches it near 100 TeV.

Neutrinos per flavour · Fermi-LAT (2015) · IceCube (2026), band digitized from Fig. 3 · Auger (2020) · Bound: Waxman & Bahcall (1999)

Production environments

Environment Target for energetic hadrons Observational test
Galactic disk Interstellar gas Spatially extended emission
AGN jet Gas or radiation Direction and variability
AGN corona (hidden core) Gas and X-ray photons Neutrinos brighter than GeV–TeV gamma rays
Star-forming environment Dense gas Population and diffuse-flux constraints
GRB or tidal disruption Gas or radiation Time-dependent searches

Many faint sources: the brightest neutrino sources seen so far supply only a small part of the diffuse flux, so most of it should come from many faint emitters.

Murase & Stecker · Coronae: Kheirandish, Murase & Kimura (2021) · Populations: Groth & Ahlers (2025)

Observations

The astrophysical neutrino flux

Redrawn IceCube 2013 deposited-energy distribution of 28 starting events, compared with atmospheric muon and neutrino backgrounds and the historical best-fit astrophysical component.

28 events · 662 live days

\[30–1200\,\mathrm{TeV}\]

Deposited energy.

White points: data.

Blue + red: atmospheric backgrounds.

Orange: signal + background.

\[4.1\sigma\]

Benchmark atmospheric hypothesis.

2010–2012 sample and fitted spectrum · Redrawn from the vector artwork of IceCube, Science (2013), Fig. 4

Original IceCube side-view diagram of its outer veto and inner fiducial volume, including the dust layer.

Entering muon: early light at the boundary.

Contained interaction: first light inside.

Atmospheric neutrinos can arrive with muons from the same air shower.

Veto and fiducial volume · IceCube (2013), Fig. 1, side view

Redrawn IceCube 2013 zenith distribution after the starting-event selection, compared with atmospheric backgrounds and the fitted astrophysical component.

Direction, energy and veto response constrain the background hypothesis together. · Redrawn from the vector artwork

An independent measurement with Baikal-GVD

Redrawn Baikal-GVD 2018–2024 upward-going cascade-energy histogram: 18 data events, atmospheric neutrino and muon backgrounds, and the fitted astrophysical contribution.

April 2018–March 2024

18 upward-going cascades

\[E_{\mathrm{sh}}>15\,\mathrm{TeV}\]

Expected atmospheric background:

\[2.8\pm1.0\ \text{events}\]

\[5.1\sigma\]

Points: data · Stacked: fitted signal and backgrounds · Digitized from Baikal-GVD (2025), Fig. 2

Redrawn reconstructed-zenith histogram of the same 18 GVD cascades and stacked signal/background expectations.

Same selected sample · \(\cos\theta<-0.25\), additional quality cuts · Baikal-GVD, Fig. 2 · Digitized from the original

Redrawn GVD spectrum comparison: fitted per-flavour GVD spectrum in green, IceCube starting-track points in orange and preliminary combined-fit points in blue.

Per flavour, \(\nu+\bar\nu\) · Published 2025 comparison · Blue IceCube points are preliminary · Digitized from Fig. 4

Redrawn power-law normalization/index confidence contours for GVD and three IceCube samples. GVD has 68% and 90% contours; IceCube has 68% and 95%.

Single power law · Normalization at \(100\,\mathrm{TeV}\) per flavour · Different contour levels retained · Fig. 3 · Digitized from the original

Structure in the diffuse spectrum

Redrawn finalized IceCube segmented per-flavour neutrino-flux estimates and broken-power-law bands: Combined Fit orange, MESE blue, with published uncertainties and upper limits.

CF (orange): tracks + cascades.

MESE (blue): starting events.

Slope changes near

\[30\,\mathrm{TeV}.\]

Single power law disfavoured:

\[>4\sigma\]

Curvature is also tested.

Per flavour, \(\nu+\bar\nu\) · Segmented estimates and fitted bands · Redrawn from the vector artwork of IceCube, PRL (2026), Fig. 3

Redrawn IceCube low- and high-energy spectral-index profile-likelihood contours. The diagonal corresponds to equal indices, a single power law.

\(\gamma_1=\gamma_2\): single power law · \(68\%\) and \(95\%\) regions · Fig. 2 · Redrawn from the vector artwork

Redrawn Combined Fit track and cascade energy-proxy distributions, compared with background and a broken-power-law astrophysical component. Track and cascade count axes differ.

Reconstructed proxies and separate count axes · CF panel of Fig. 1 · Redrawn from the vector artwork

IceCube visualization of the December 2016 Glashow-resonance candidate: schematic of early muon light, the event view across the detector and photon arrival times at the two nearest sensors.

Glashow resonance

\[\bar\nu_e+e^-\to W^-\to\text{hadrons}\]

\[E_{\bar\nu}\simeq6.3\,\mathrm{PeV}\]

One shower, December 2016. Only antineutrinos can do this.

Tau neutrinos

Seven candidates: all three flavours arrive.

Why it matters

\(pp\): \(\ \bar\nu_e:\nu_e\approx1:1\)

simple \(p\gamma\): \(\ \approx1:3.5\)

IceCube, Nature 591, 220 (2021), Fig. 1 · Tau neutrinos: IceCube, PRL 132, 151001 (2024)

The Milky Way in neutrinos

IceCube's five Galactic-coordinate views: optical, gamma rays, predicted neutrino emission, analysis expectation and observed pre-trial significance.

IceCube Collaboration, Science (2023) · Diffuse Galactic emission: \(4.5\sigma\)

ANTARES, 2023: a mild hint of TeV neutrinos from the inner Galactic ridge.

Redrawn sky map in Galactic coordinates: the eight Baikal-GVD cascades above 200 TeV as orange 90% containment circles; five of them lie in or next to the shaded band within 10 degrees of the Galactic plane.

8 cascades above 200 TeV

They crowd towards the Galactic plane.

Median absolute Galactic latitude of the eight cascades compared with skies whose right ascensions are randomized: the data line lies in the low tail of the random-sky histogram.

Baikal-GVD: \(p=1.4\times10^{-2}\)

With IceCube: \(p=3.4\times10^{-4}\)

V. A. Allakhverdyan et al. (Baikal-GVD) with Y. Y. Kovalev, A. V. Plavin, D. V. Semikoz, S. V. Troitsky · ApJ 982, 73 (2025) · Redrawn from Table 1; test of Fig. 1 recomputed

Model Galaxy (illustrative) · Baikal-GVD cascades above 200 TeV and the published shuffling test

Redrawn published spectra of diffuse gamma rays from the Galactic plane between 10 TeV and 1 PeV: LHAASO-KM2A inner and outer Galaxy points from its tables, and Tibet AS-gamma points digitized from its Figure 4, shown as E to the 2.7 times the differential flux.

The same collisions

\[\pi^0\to\gamma\gamma,\qquad \pi^\pm\to\nu\ \ldots\]

The Galactic plane glows in gamma rays up to \(\sim1\,\mathrm{PeV}\).

Tibet AS\(\gamma\), 2021: first diffuse gamma rays above \(100\,\mathrm{TeV}\).

LHAASO, 2023: the spectrum from \(10\,\mathrm{TeV}\) to \(1\,\mathrm{PeV}\).

Redrawn: LHAASO, PRL 131, 151001 (2023), Tables S2–S3 · Tibet AS\(\gamma\), PRL 126, 141101 (2021), digitized from Fig. 4 · Different sky regions and source masks

Neutrino spectra and the gamma-ray budget

Redrawn Zathul et al. 2026 figure 2: measured IceCube neutrinos and Fermi isotropic gamma-ray background with neutrino fits and their calculated electromagnetic cascades. White denotes a single power law, orange a broken power law, blue a log parabola.

Right: measured neutrinos and fitted spectra.

Left: accompanying cascade photons.

White: unbroken continuation.

Orange: break near \(30\,\mathrm{TeV}\).

The low-energy continuation changes the gamma-ray burden.

Dashed: neutrinos · Solid: calculated photons · \(p\gamma\) benchmark · Digitized from Zathul et al. (2026), Fig. 2, CC BY 4.0

\[ \pi^0\to\gamma\gamma,\qquad \gamma\gamma_{\mathrm{EBL}}\to e^+e^-,\qquad e^\pm\gamma_{\mathrm{bg}}\to e^\pm\gamma \]

Escaping high-energy photons feed a lower-energy electromagnetic cascade.

Input Constraint
Measured neutrino spectrum Pion production and its energy dependence
Fermi isotropic gamma-ray background A ceiling on the accompanying extragalactic cascade
Internal photon absorption Redistribution of photon energy before escape

Other source populations also contribute to the gamma-ray background.

Conversion: Ahlers & Murase (2014) · absorption computed with the EBL of Domínguez et al. (2011) and the CMB · cascade shape schematic, energy conserved · Murase, Ahlers & Lacki (2013), Murase, Guetta & Ahlers (2016)

Compact regions and hidden gamma rays

Causal diagram of a compact source: energetic protons meet gas or photons, pions produce neutrinos that escape and gamma rays that can be converted to electron-positron pairs and reprocessed photons. The diagram is schematic and has no calculated flux.

\[ f_{pp}\simeq1-e^{-t_{\mathrm{res}}/t_{pp}},\qquad t_{pp}^{-1}=\kappa_{pp}\,n_{\mathrm{gas}}\,\sigma_{pp}\,c \]

Gas-target energy-loss estimate · Schematic geometry · Murase & Stecker

\[ \gamma+\gamma\to e^++e^-,\qquad E_\gamma\epsilon\gtrsim(m_ec^2)^2\quad\text{(head-on)} \]

Condition Meaning
Above pair-production threshold Absorption is kinematically allowed
\(\tau_{\gamma\gamma}\gtrsim1\) Photon density and path length make it effective
Pair cooling and further interactions Energy emerges at different photon energies

Neutrino escape depends on the matter column as well.

NGC 1068 and X-ray-bright AGN

Actual Hubble ultraviolet and optical observations of the spiral galaxy Messier 77, NGC 1068, showing its bright nucleus and dusty spiral arms.

ESA/Hubble & NASA, L. C. Ho, D. Thilker

Redrawn latest IceCube fine scan around NGC 1068: local significance map, 68% and 95% contours and the galaxy position. The significance of a preselected catalogue source differs from the significance of an all-sky search.

13.1 years

\[4.0\sigma\]

After catalogue trials.

Seyfert 2

A dust-obscured active nucleus.

Hubble (2025) · IceCube Collaboration · Redrawn from 2026 analysis, Fig. 2

Redrawn NGC 1068 spectral energy distribution from the 2022 IceCube paper: archival photon measurements, Fermi and MAGIC points or limits, the 2022 muon-neutrino fit and two theoretical core models.

Blue: 2022 neutrino fit.

Orange: gamma-ray data and limits.

Violet: core models.

The photon spectrum constrains where pion production can occur.

Historical comparison · \(\nu_\mu+\bar\nu_\mu\), \(95\%\) fit band · Redrawn from IceCube, Science (2022), Fig. 4

Redrawn IceCube all-flavour spectra of the strongest X-ray-bright AGN, including NGC 1068, compared with the all-sky diffuse neutrino flux. Source spectra and the diffuse spectrum retain the publication's normalization conventions.

All flavours · NGC 1068: \(\Phi_\nu(E)\propto E^{-\gamma}\), \(\gamma=3.4\pm0.2\) · IceCube Collaboration · Redrawn from the vector artwork

Redrawn IceCube binomial test of 47 X-ray-bright AGN, with NGC 1068 excluded: ordered local source p-values and the binomial probability of source subsets, with the most significant subset of 11 marked.

47 X-ray-bright AGN

NGC 1068 excluded.

11-source ensemble

\[3.3\sigma\]

After population trials.

Population evidence, not 11 individually established sources · IceCube Collaboration · Redrawn from the vector artwork

Blazars: TXS 0506+056 and radio-bright jets

22 September 2017

IceCube-170922A

Most probable energy

\[E_\nu\sim290\,\mathrm{TeV}\]

Fermi-LAT + MAGIC

A flaring blazar in the neutrino direction.

\(\sim3\sigma\) association

IceCube Collaboration · Multimessenger observations, Science (2018)

Redrawn sky localization of the Baikal-GVD 2021 cascade, IceCube-170922A and TXS 0506+056; the Baikal angular uncertainty is broad and includes the blazar.

18 April 2021

A cascade in Lake Baikal, from below the horizon.

\[E_{\rm sh}\simeq224\,\mathrm{TeV}\]

The blazar lies inside its error circle.

A candidate, not a proof.

Baikal-GVD Collaboration · Redrawn from TXS 0506+056 candidate, ICRC 2023, Fig. 3; IceCube contour digitized

Redrawn light curves of TXS 0506+056, 2010–2022: Fermi-LAT gamma-ray flux as white points and RATAN-600 11 GHz radio flux as orange diamonds with a fitted sum of flares. Dashed vertical lines mark IceCube-170922A and GVD210418CA; a shaded band marks the 2014–2015 IceCube low-energy excess.

Both neutrinos arrived as a strong radio flare began.

White: Fermi-LAT gamma rays · Orange: RATAN-600, 11 GHz, and fitted flares · Dashed: neutrino arrivals · Digitized from Baikal-GVD, ICRC 2023, Fig. 4

Recoloured sky map in equatorial coordinates: IceCube events above 200 TeV as blue error ellipses, VLBI-selected active galaxies inside them as stars coloured by 8 GHz parsec-scale flux density, other sample members as grey dots; four strongest associations and TXS 0506+056 labelled.

IceCube events above 200 TeV

Blazars with bright parsec-scale radio cores sit inside their error regions more often than chance allows.

Redrawn comparison of the average 8 GHz VLBI flux density of AGN inside IceCube error regions, an orange triangle, with the 68% interval for randomly shifted events, a blue bar.

Post-trial \(p=2\times10^{-3}\)

Still debated: an IceCube search with its alert events finds no significant correlation.

A. V. Plavin, Y. Y. Kovalev, Y. A. Kovalev, S. V. Troitsky · Redrawn from ApJ 894, 101 (2020), Figs. 1 and 3 · ICRC 2021 · IceCube (2023)

Redrawn IceCube archival flare plot: event weights versus time in 2012–2015, coloured by muon-energy proxy, with the fitted box and Gaussian flare profiles. This is the historical 2018 analysis.

\(13\pm5\) excess events · \(3.5\sigma\)

No matching strong gamma-ray flare.

Historical 2018 analysis · Redrawn from IceCube Collaboration, Science (2018), Fig. 2

Original processing

Event directions, one-sigma angular uncertainties and reconstructed-energy proxies around TXS 0506+056 using the original Point Source Tracks processing, redrawn from the 2023 IceCube reconstruction comparison.

Updated processing

The corresponding TXS 0506+056 event directions, one-sigma angular uncertainties and reconstructed-energy proxies with Northern Tracks v5 processing. Two cascade-like events are removed.

Two cascade-like events removed; a smaller fitted excess.

Same flare, revised event sample · IceCube, ICRC 2023 · Both panels redrawn from the vector artwork

KM3-230213A: an ultra-high-energy event

13 February 2023

ARCA · 21 detection units

Parent neutrino

\[E_\nu\sim220\,\mathrm{PeV}\]

  • Nearly horizontal.
  • No identified counterpart.

KM3NeT Collaboration · Nature (2025)

Energy intervals of KM3-230213A drawn from the published numbers: the muon energy at the detector, 120 PeV with a 68 percent interval of 60 to 230 PeV, and the inferred parent-neutrino energy, median 220 PeV with a 68 percent interval of 110 to 790 PeV, assuming an E to the minus 2 spectrum. Whiskers show 90 percent intervals.

Muon at the detector

\[E_\mu=120^{+110}_{-60}\,\mathrm{PeV}\]

\(68\%\) confidence interval

Parent neutrino

Median: \(220\,\mathrm{PeV}\)

Central \(68\%\): \(110–790\,\mathrm{PeV}\)

Assuming \(\Phi_\nu(E)\propto E^{-2}\).

Drawn from the numbers in the text of KM3NeT, Nature (2025) · Their Fig. 2 (CC BY-NC-ND) is kept unmodified for questions

Observatories

What improves source sensitivity?

Illustration: an invented source of NGC 1068’s strength, simulated events, computed significance · Not a forecast for any detector

Point sources · background-dominated limit

\[ N_s\propto\mathcal E,\qquad N_b\propto\mathcal E\,\sigma_\theta^2, \qquad Z\simeq\frac{N_s}{\sqrt{N_b}} \propto\frac{\sqrt{\mathcal E}}{\sigma_\theta} \]

Change Signal Background Significance
\(4\times\) exposure \(4\times\) \(4\times\) \(2\times\)
\(2\times\) sharper directions unchanged \(\div4\) \(2\times\)
Both \(4\times\) unchanged \(4\times\)
  • Calibration controls pointing and energy systematics.
  • A global network adds sky coverage and rapid follow-up.

Idealized scaling, fixed spectrum and event selection. Gen2 science case · Upgrade calibration

Neutrino observatories

South Pole, Antarctic ice

IceCube: \(1\,\mathrm{km}^3\), operating.

Upgrade: 5 new strings, installed in 2026.

Gen2: \(8\,\mathrm{km}^3\) optical array, proposed.

Measurement: more high-energy events; improved calibration and a radio extension.

Official visualization of the proposed IceCube-Gen2 expansion, including optical, surface and radio arrays around the existing IceCube detector.

Gen2 design illustration. DESY, Science Communication Lab.

Mediterranean Sea

ARCA: \(\sim1\,\mathrm{km}^3\) for neutrino astronomy.

ORCA: \(\sim7\,\mathrm{Mt}\) for neutrino oscillations.

Full design: 230 / 115 strings.

Construction and data taking.

Measurement: source searches with ARCA; oscillations with ORCA.

Artist's impression from inside a KM3NeT building block on the sea floor: vertical detection units carrying spherical multi-PMT optical modules, the yellow buoy of a nearby unit, a close-up optical module and anchors on the seabed. An illustration, not a photograph.

Artist’s impression · Edward Berbee/Nikhef · KM3NeT

Lake Baikal, fresh water

2026: 16 clusters, operating.

Reported effective volume: \(0.8\,\mathrm{km}^3\).

Modular growth through winter deployment.

Measurement: independent diffuse spectra and northern-site source searches.

Baikal-GVD's official 2026 configuration: sixteen numbered physics clusters, the separate experimental cluster, laser strings and deployment years. Original Russian labels are retained.

2026 configuration. Baikal-GVD.

Pacific Ocean, British Columbia

Full design: \(\sim1\,\mathrm{km}^3\), 70 strings.

  • Ocean Networks Canada’s NEPTUNE infrastructure.
  • Prototype and deployment engineering.

Measurement: neutrino astronomy from another northern ocean site.

P-ONE collaboration illustration of seven planned clusters and a ten-string cluster with kilometre-long mooring lines. This is a design drawing, not a deployed full detector.

Proposed configuration. K. Holzapfel / P-ONE.

South China Sea

Design: \(7.5\,\mathrm{km}^3\), 1,211 strings.

Sensors: hybrid PMT + SiPM modules.

R&D and pathfinder programme.

Design goal: photon timing and directional information from hybrid sensors.

TRIDENT's proposed Penrose-tiling array layout from figure 2 of Ye et al. Black points mark strings, red triangles junction boxes, dashed paths allow ROV maintenance, with a 500 metre scale bar.

Proposed layout. Ye et al., Nature Astronomy (2023), Fig. 2.

China + Russia

Design: \(\sim30\,\mathrm{km}^3\).

Sensors: 20-inch PMTs.

The Baikal option, B-HUNT, is the next slide.

Official HUNT concept illustration showing large single-PMT optical modules on underwater strings. This is a design illustration, not a photograph of an operating array.

Design illustration. HUNT / IHEP, CAS.

South China Sea

Design: \(10\,\mathrm{km}^3\), 1,200 strings.

Sensors: multi-PMT optical modules.

Proposal and hardware R&D.

Design goal: high-energy source searches with a large multi-PMT array.

NEON's proposed Fibonacci-spiral string layout and multi-PMT module from the upper panel of figure 3 in Zhang et al. The illustration includes the original 1,790 metre radius and 1,000 metre height labels.

Proposed layout. Zhang et al., Fig. 3.

Baikal-GVD → B-HUNT

Milestones: 2024 first HUNT prototype string in Lake Baikal with 12 modules of 20-inch photomultipliers; 2025 a full-scale string with 20-inch modules operating in Lake Baikal, and four prototype detector units with an LED calibration unit at 1600 metres in the South China Sea; 2026 Baikal-GVD with 16 clusters and about 0.8 cubic kilometres effective volume; about 2028 Baikal-GVD at about 1 cubic kilometre; about 2040 B-HUNT at about 30 cubic kilometres. Two cubes of 1 and 30 cubic kilometres are drawn to volume scale.

Russia + China · Baikal-GVD Collaboration with IHEP (Beijing) · JINR (2026) · IHEP

The ice is a platform: strings are deployed from the frozen lake every winter.

Water: light scatters less than in ice, so cascade directions are sharper.

Northern site: the Galactic Centre is below the horizon \(75\%\) of the day.

A gamma-ray partner: LHAASO watches the same northern sky.

Alerts and follow-up: work in progress.

Official HUNT concept illustration showing large single-PMT optical modules on underwater strings. This is a design illustration, not a photograph of an operating array.

Design illustration. HUNT / IHEP, CAS.

Prospects

What we want to discover

Open questions placed on a neutrino-energy axis from 1 TeV to 1 EeV over the measured cosmic neutrino spectrum: Seyfert galaxies beyond NGC 1068 at 1.5 to 15 TeV; the spectral break near 25 TeV, pp or p-gamma; the Milky Way and PeVatrons above 200 TeV; the neutrino cross section from Earth absorption above 40 TeV; the Glashow resonance at 6.3 PeV; and KM3-230213A at about 220 PeV, source or cosmogenic. Transients through alerts and new physics can appear at any energy.

Questions on the energy axis · Spectrum: IceCube (2026) · Glashow: IceCube (2021) · KM3-230213A: KM3NeT (2025)

What we lack

What is missing Why it matters
Northern exposure beyond \(10\,\mathrm{km}^3\) Faint sources and the Galactic Centre seen through the Earth
Sharper cascade directions All flavours point back to sources
More events above \(1\,\mathrm{PeV}\) The spectrum’s end, Glashow events, the 100 PeV puzzle
Global alerts, joint analyses, open data One sky, several telescopes
Calibrated optics of water and ice Energies and directions we can trust
Population theory within the gamma-ray budget Turning many faint hints into sources

Symbolic chain: the programme ФНАФЧ, neutrino physics and particle astrophysics, with courses, seminars and the Baikal school; people, students who become researchers and lecturers; results, data analysis, for example the Milky Way in Baikal-GVD data with a first author from the programme; and the next experiment, B-HUNT, Russia and China.

Telescopes are built and understood by people.

ФНАФЧ — Физика нейтрино и астрофизика частиц · Courses on teach-in.ru

Where are the sources?

Equatorial sky map. The Galactic plane with the Galactic Centre is marked as diffuse neutrino emission from the Milky Way. NGC 1068 stands for Seyfert galaxies, TXS 0506+056 for blazars, and an open circle marks the direction of KM3-230213A, whose origin is open. Caption: most of the flux comes from many faint sources, not yet resolved.

Found: our Galaxy · NGC 1068 and Seyfert galaxies · blazar candidates.

Still open: most of the flux — a crowd of faint sources.

Next: bigger telescopes and networks — B-HUNT, KM3NeT, IceCube-Gen2.

We have found where to look. Next we must resolve the crowd.

Positions: Baikal-GVD source list (2026) · Evidence: result slides of this lecture

Backup

What creates the light?

\(\nu_\mu+N\longrightarrow\mu^-+X.\)

\(\nu_e+N\longrightarrow e^-+X.\)

Cherenkov light

Gold: charge. Blue circles: wavefronts. White: common front. Cyan arrows: light.

Light gets ahead — \(v<c/n\)

The particle gets ahead — \(v>c/n\)

Same-age wavefronts: below threshold the circles are nested; above threshold they share a tangent front behind the particle.

The charge disturbs the medium. A fast charge leaves a common wavefront behind.

Light travels along a radius of a spherical wavefront, perpendicular to the common tangent front. The emission angle is measured from the forward particle track.

Light goes across the front — not along it.

\[ \cos\theta_C=\frac{c}{nv} \qquad\Longrightarrow\qquad \theta_C\simeq41^\circ\ \text{in water, for }v\simeq c. \]

One photon, a measurable pulse

IceCube

An actual original IceCube DOM during deployment, with one large PMT inside a glass pressure sphere.

One \(10\,\mathrm{inch}\) PMT.

IceCube/NSF.

KM3NeT

An actual KM3NeT digital optical module with many small PMTs pointing in different directions.

\(31\) three-inch PMTs.

Courtesy KM3NeT.

Baikal-GVD

A real Baikal-GVD optical module above an ice hole during deployment, from DN's 2021 lecture.

One \(10\,\mathrm{inch}\) PMT.

Baikal-GVD; supplied by DN.

TRIDENT

An actual TRIDENT hybrid optical-module prototype in the laboratory, combining PMTs with small SiPM assemblies.

PMTs + SiPMs. Prototype.

TRIDENT / TDLI, SJTU.

Reconstructing an event

Next event:Fit model:

Loading the synthetic event…
Synthetic water detector · crossing point is known · fit direction and time, not vertex or energy

Predict a time for each sensor

\[ \cos\theta_C=\frac{1}{\beta n_{\rm phase}},\qquad t_i^{\rm pred}=t_0+\frac{s_i}{\beta c}+\frac{n_{\rm group}D_i}{c}. \]

\(s_i\): muon distance from the known crossing point to emission.

\(D_i\): photon path from emission to the sensor.

The cone uses the phase index; flight time uses the group index.

Find the direction that agrees with the signals

\[ S=\frac{1}{2\sigma_t^2}\sum_{i,k} \left(t_{ik}^{\rm obs}-t_i^{\rm pred}\right)^2 \longrightarrow\min_{\theta,\phi,t_0}. \]

Every detected photoelectron contributes a measured time.

Here: known crossing point, straight muon, homogeneous medium. Only direction and time offset are fitted — not vertex or energy.

How does a wrong model of the medium change the inferred direction?

Synthetic teaching model, not an experiment’s reconstruction or an angular-resolution forecast. The arrow is the muon’s direction of travel; the source direction is opposite.

Backgrounds

Telescope Optical background Pulses / PMT Muon-dominated samples
IceCube PMT + glass noise \(0.56\,\mathrm{kHz}\) · 10-inch \(2.7\,\mathrm{kHz}\) · IC86 DAQ
KM3NeT \(^{40}\mathrm K\) + bioluminescence \(\sim7\,\mathrm{kHz}\) · 3-inch \(\sim1\,\mathrm{Hz}\) · ARCA6 reconstructed
Baikal-GVD Water luminescence \(20–100\,\mathrm{kHz}\) · 10-inch \(\sim3\,\mathrm{Hz/cluster}\) · reconstructed

Pulses, DAQ triggers and reconstructed events are different counting levels. IceCube · KM3NeT · Baikal-GVD

Calculated neutrino-nucleon interactions per year per energy decade in one cubic kilometre of water from the downward hemisphere, before event selection. Conventional atmospheric neutrinos fall more steeply than the astrophysical component. A dashed prompt-charm prediction has a shaded theoretical band. Astrophysical best-fit extrapolations outside the sensitive range are dotted.

\(1\,\mathrm{km^3}\) water · downward \(2\pi\) · all flavours, \(\nu+\bar\nu\) · CC+NC · before selection.

From events to a neutrino flux

Illustrative IceCube point-source effective-area curves versus true neutrino energy for three zenith intervals, redrawn from curves digitized from the published review figure. Selection suppresses low-energy downgoing events; Earth absorption suppresses the highest-energy vertically upgoing events.

Near the horizon: short Earth chord.

Upward: Earth absorption at high energy.

Downward: atmospheric-muon rejection.

\(A_{\mathrm{eff}}\) includes interaction, propagation and event selection.

\[ N=T\sum_\alpha\int dE_\nu\,d\Omega\, \Phi_\alpha(E_\nu,\Omega)\,A_{\mathrm{eff},\alpha}(E_\nu,\Omega) \]

Illustrative point-source selection · Redrawn, digitized from Spiering (2026), Fig. 6

IceCube 2013 all-sky-averaged effective areas by flavour for contained starting events, in square metres versus neutrino energy in TeV, redrawn from the step histogram digitized from the published figure. The electron-flavour peak at 6.3 PeV is the Glashow resonance.

A specific sample

  • Start inside the fiducial volume.
  • Reject early light at the boundary.
  • The accepted events determine the response used in the flux fit.

Historical HESE selection, 2010–2012 · Redrawn, digitized from IceCube (2013), Fig. 7

Event Light constrains Parent-neutrino inference
Contained \(\nu_e\) CC Electromagnetic + hadronic deposits \(E_{\mathrm{dep}}\simeq E_\nu\), after response calibration
Through-going \(\mu\) Muon energy losses inside the array \(E_\mu\) at the detector, then a spectrum-dependent \(E_\nu\) distribution
Neutral current Hadronic deposit \(E_{\mathrm{dep}}\sim yE_\nu\); outgoing neutrino is unseen

Historical HESE reconstruction (2013): tracks \(<1^\circ\); cascades \(10–15^\circ\).

Quoted median angular scales above about \(50\,\mathrm{TeV}\). Neutrino–muon kinematics adds a separate spread.

\(y\): inelasticity · IceCube reconstruction (2013)

Point-source searches: Baikal-GVD

988 tracks · 92 directions · April 2019–March 2024

The 92 Baikal-GVD E^-2 source flux limits from published Table 1, drawn versus sine of declination at 90% confidence. NGC 1068, TXS 0506+056 and the KM3-230213A direction are highlighted. This graph contains only observed source limits, not a new sensitivity calculation.

\(E^{-2}\) spectrum · \(90\%\) CL · No significant catalogue excess · Baikal-GVD (2026), Table 1

Real-time alerts and multimessenger follow-up with Baikal-GVD: work in progress.

Redrawn Baikal-GVD figure 2: 90% source limits and median expected sensitivities for spectral indices 2, 2.5 and 3.2. Symbols are limits; solid curves are sensitivities.

Symbols: source limits · Lines: expected sensitivity · Baikal-GVD (2026), Fig. 2 · Limits from Table 1, sensitivities redrawn from the vector artwork

Redrawn comparison of Baikal-GVD and ANTARES 90% point-source limits for E^-2 and E^-2.5 spectra, with the declination axis of the original. This compares selected analysis samples, not equal-volume instruments.

Matched spectral assumptions · Baikal-GVD Collaboration (2026), Fig. 3 · Redrawn from the vector artwork; Baikal-GVD limits from Table 1

A global view of the neutrino sky

Mollweide world map of operating IceCube, KM3NeT ARCA and ORCA, and Baikal-GVD sites, with open markers for P-ONE's Cascadia region and the TRIDENT design site. Open markers denote next-generation projects, not completed operating arrays.

Site overview · October 2026 · coastlines: Natural Earth

Calculated fraction of a sidereal day each source spends below the horizon of IceCube, ARCA and Baikal-GVD versus declination. At the Galactic Centre the fractions are 0%, 63% and 75%, respectively. This is geometry, not effective area or sensitivity.

Geometry only, not sensitivity · IceCube also observes the southern sky with starting and contained events.

ANTARES–IceCube joint-search E^-2 discovery-flux comparison, redrawn from digitized published curves: the combined gold curve is lower than the separate seven-year IceCube and nine-year ANTARES curves in the southern sky. The ordinate is the five-sigma discovery flux, not a detected source flux.

IceCube · 7 years

ANTARES · 9 years

  • Gold: joint analysis.
  • Lower flux = better discovery potential.
  • Assumed spectrum: \(E^{-2}\).

Published joint analysis, not a next-generation projection · Redrawn, digitized from ANTARES & IceCube (2020), Fig. 2