At the South Pole, the sky is not the only direction from which astronomy can be done.
Buried in the ice below the Amundsen-Scott South Pole Station is an observatory with no mirror, no lens and no conventional camera. Its detecting elements are 5,160 glass spheres, each holding a sensitive photomultiplier tube and its electronics. They hang on 86 cables that descend through a cubic kilometre of the Antarctic ice sheet.
The deepest sensors sit about 2,450 metres below the surface. The top of the main array is closer to 1,450 metres down, so it is more precise to say that IceCube extends through the kilometre of ice between those depths than that the whole instrument lies two and a half kilometres underground.
Its most unusual feature is the direction in which it can look. For much of its neutrino work, IceCube uses the Earth itself as a filter. A particle enters the planet in the Northern Hemisphere, travels through thousands of kilometres of rock and ice, then produces a flash of blue light in the Antarctic detector. From the South Pole, that upward-moving signal has arrived from below.
The planet becomes part of the telescope.
A telescope built from empty-looking ice
IceCube does not take a photograph of a neutrino. Neutrinos carry no electric charge and interact with matter through the weak nuclear force, so most pass through the detector without leaving any trace. Countless neutrinos cross your body every second. Almost all continue through the Earth as if it were scarcely there.
That evasiveness is exactly why IceCube must be so large. A cubic kilometre of ice contains an immense number of atomic nuclei, giving a small fraction of passing neutrinos a chance to collide with one. The instrumented mass is roughly a gigatonne.
According to the IceCube Collaboration’s description of the detector, the 5,160 digital optical modules are arranged on 86 strings. On a typical string, 60 modules are spaced 17 metres apart. Most strings stand about 125 metres from their neighbours, while a more tightly packed central region called DeepCore is designed to detect lower-energy neutrinos.
This is sparse instrumentation. The glass spheres do not fill the ice like pixels in a television screen. There can be more than a hundred metres of apparently empty ice between neighbouring cables. That works because the charged particles created in a high-energy neutrino interaction can produce light over long distances.
The setting gives the experiment three things at once. The ice supplies the target in which a neutrino may interact. It provides a transparent medium through which the resulting light can travel. It also locks the sensors into a stable three-dimensional arrangement after the boreholes refreeze.
I recently wrote about Antarctic ice preserving air from a world 1.2 million years ago. IceCube uses the same broad property of deep polar ice in a completely different way. One project reads the ice as an archive of Earth’s atmosphere. The other turns it into an active particle detector.
How an invisible particle produces blue light
A neutrino becomes detectable only on the rare occasion that it strikes a proton or neutron in the ice. The collision can create a charged secondary particle, such as an electron, muon or tau.
That secondary particle may travel through the ice faster than light can travel through ice. This does not violate Einstein’s speed limit. Light moves more slowly through a material than it does in a vacuum, while the charged particle remains below the vacuum speed of light.
As the particle passes through the ice, it disturbs the electromagnetic fields of nearby molecules. The result is a cone of blue Cherenkov radiation, the optical equivalent of the pressure wave behind a supersonic aircraft. IceCube’s photomultipliers convert arriving photons into electrical signals and record their timing and brightness.
The pattern across many sensors allows software to reconstruct what happened. Modules that light up first establish one part of the event’s geometry. The later signals show how the light propagated through the ice. The amount of light helps constrain the energy deposited, while the shape of the illuminated region helps identify the kind of secondary particle involved.
In broad terms, muons leave long tracks. A muon can travel for kilometres, producing light along its path, so the line it draws can be traced backwards towards its source with comparatively good angular precision. Electron-neutrino interactions and neutral-current interactions tend to produce compact showers called cascades. Cascades often give a better measure of energy deposited inside the detector, but their direction is harder to reconstruct. At sufficiently high energy, a tau neutrino may produce two separated showers, the rarely seen topology known as a double bang.
The glacial ice is transparent, but it is not optically simple. Layers of dust and volcanic material scatter and absorb light differently, and the ice crystals themselves have a preferred orientation. IceCube carries calibration LEDs inside its modules so researchers can measure how flashes travel through the detector. The collaboration’s work on modelling dust in the South Pole ice is not a side issue. A mistaken model of the ice could produce a mistaken estimate of a neutrino’s direction or energy.
Why IceCube looks through the Earth
Neutrinos are not the only particles capable of making light in the detector. Cosmic rays striking the atmosphere above Antarctica create showers of secondary particles, including muons. Those downward-moving atmospheric muons reach IceCube in numbers that overwhelm the much rarer astrophysical neutrino events.
Putting the sensors deep underground removes much of that background, but not all of it. The more decisive filter is the planet.
A muon cannot pass through the entire Earth. If IceCube reconstructs a long track moving upwards through the detector, the particle that began the event must have been capable of crossing the planet. A neutrino can do that, interact in the ice or nearby rock, and create an upward-moving muon just before detection.
Geometry makes this particularly elegant at the South Pole. The downward direction through Earth opens onto the northern celestial sky. Instead of turning a telescope towards a star in the north, IceCube waits for a neutrino from that direction to pass through the planet and approach the detector from below.
This filtering method does not remove every background. Cosmic rays also create atmospheric neutrinos, and those can cross the Earth just as astronomical neutrinos do. Researchers must separate the populations statistically, using energy, direction, event shape and whether several events cluster around a possible source.
The reason for accepting all this difficulty is that neutrinos preserve information that other particles lose. Cosmic rays are electrically charged, so magnetic fields bend their paths while they travel to Earth. The arrival direction of a charged particle may bear little resemblance to the direction of the object that accelerated it.
That problem sits behind my earlier article about the extraordinary cosmic-ray particle detected over Utah in 1991. Its energy could be measured, but its charged path through interstellar and intergalactic magnetic fields makes its birthplace difficult to recover. Neutrinos have no charge. Once their direction has been reconstructed, it can point back towards where they were produced.
The Earth is a filter, not a transparent window
The phrase “looking through the Earth” needs an important qualification. The weak force gives a neutrino only a small chance of interacting with matter, but that chance rises with energy. A path through the diameter of the planet contains so much material that the highest-energy neutrinos are increasingly likely to be absorbed before reaching Antarctica.
IceCube has used this effect as a measurement in its own right. In 2017, the collaboration reported its first observation of the Earth absorbing high-energy neutrinos. The result was consistent with the expected neutrino interaction probability, while also demonstrating that the planet becomes less transparent as neutrino energy increases.
At energies between roughly one and ten petaelectronvolts, looking through the Earth can discard much of the signal along with the atmospheric muons. That is why IceCube does not simply ignore everything arriving from above.
For the southern sky, researchers use other strategies. They can demand that an event begin inside the instrumented volume, allowing the outer layers of sensors to act as a veto against incoming atmospheric muons. They can select extremely energetic events, since the atmospheric background falls steeply with energy. IceTop, the surface array above IceCube, can also identify air showers associated with downward particles.
A 2025 southern-sky analysis combined a demanding muon-neutrino selection with information from IceTop. Only two events survived the selection in nine years of data. Neither could be associated with a known astrophysical source, but the analysis showed why the detector must use different methods in different parts of the sky.
IceCube therefore does not have one fixed mode of vision. At many energies, Earth is a shield that makes the northern sky cleaner. At the greatest energies, that shield becomes too thick, and the southern sky may offer the better route despite its much heavier background.
How 5,160 sensors were frozen into place
Building the detector required making temporary shafts through ice that had been compacting for tens of thousands of years. Between 2004 and 2010, crews used a hot-water drill to melt boreholes approximately 60 centimetres wide and as deep as 2,450 metres.
The IceCube Collaboration’s construction figures say each hole took an average of about 48 hours to drill. Lowering a cable carrying 60 optical modules took another 11 hours. The water then refroze around the string.
That last step made the deployment permanent. A damaged sensor cannot be pulled up and replaced. The modules were extensively tested before installation, and their software can be updated through cables leading to the IceCube Laboratory on the surface, but the glass spheres themselves are now inaccessible.
The design also explains why the detector is buried so deeply. The ice above it shields the modules from much of the surface radiation and gives atmospheric particles more material to cross before reaching the array. At those depths, pressure has also squeezed out the air bubbles that would otherwise scatter light severely.
Nothing moves to follow a target. There is no dome to rotate and no observing night to schedule. The detector watches continuously, using the timing relationship among fixed sensors to infer the direction of an event. From the geographic South Pole, the sky circles the horizon while the geometry of the detector remains stable.
Calling IceCube a telescope is therefore an extension of the word, but not an indulgence. A telescope is an instrument that gathers a messenger from a distant source and reconstructs where it came from. IceCube does that without focusing the messenger and without detecting the neutrino directly. It waits for one rare collision, reads the light left by the debris and calculates the incoming path.
What neutrino astronomy has already revealed
IceCube’s scale would be hard to justify if it merely detected occasional particles without learning where they came from. Its first decisive astronomical result was statistical. In 2013, the collaboration reported evidence for a population of high-energy neutrinos from beyond the Solar System. The events arrived from across the sky and had energies and numbers that could not be explained by the expected atmospheric background alone.
The next challenge was to connect those particles to particular objects.
On 22 September 2017, IceCube detected a high-energy neutrino and issued an automated alert. Follow-up observations by the Fermi Gamma-ray Space Telescope and the MAGIC telescopes drew attention to the active galaxy TXS 0506+056, whose jet was in a bright gamma-ray state. An examination of older IceCube data then found an earlier excess of neutrinos from the same direction. The 2018 multimessenger result provided evidence that this active galaxy was a source of high-energy neutrinos and, by implication, a site of particle acceleration.
In 2022, the collaboration reported evidence for high-energy neutrino emission from NGC 1068, also called Messier 77. This nearby active galaxy is especially useful because dust and gas obscure its central region in some forms of light. Neutrinos can escape dense environments that photons may not, giving astronomers information about a hidden galactic core. The NGC 1068 result was based on an excess of events collected over years rather than one individually conclusive neutrino.
Then, in 2023, IceCube produced the first image of the Milky Way in high-energy neutrinos. That map was built largely from cascade events and statistical patterns, not from a crisp succession of point sources. It showed the galactic plane using particles of matter rather than electromagnetic radiation.
This gives a useful counterpoint to my article on Webb finding mature-looking galaxies in the first few hundred million years of cosmic history. Webb depends on photons, mostly infrared ones, that have crossed the expanding universe. IceCube adds a messenger that is not absorbed by dust, is not deflected by magnetic fields and can escape regions from which light struggles to emerge.
Neutrino astronomy remains sparse. Most IceCube events are backgrounds, most astronomical neutrinos cannot be assigned to a named source, and every reconstruction carries uncertainty. The detector does not produce the visual certainty of a planetary photograph.
Its achievement is different. It has made an observatory from a naturally occurring block of ice, used the Earth as both obstruction and shield, and extracted directions from flashes that last only a tiny fraction of a second. A particle can begin on the far side of the sky, pass through the planet beneath our feet and reveal itself only in its final collision, two kilometres below Antarctica.
