IceCube: how do you turn a cubic kilometre of ice into an instrument?

Sensors placed in natural ice

The South Pole already has an enormous volume of glacial ice, so an observatory can put light sensors inside it instead of building a giant tank. The catch is that the natural ice must be measured as carefully as any built detector.

The 2017 instrument paper describes the original complete in-ice array: 5160 digital optical modules on 86 strings, at depths from about 1450 to 2450 metres. These numbers describe that configuration and do not include later upgrades.

The modules sample light at many points spread through the volume; they do not form a wall that stops neutrinos. Most neutrinos pass straight through without interacting. The large target makes the rare interactions that do happen more useful.

Sources: [1] · [2]

A flash becomes a timed electrical signal

A neutrino interaction can produce charged particles that emit Cherenkov light. In each module a photomultiplier turns the arriving photons into electrical signals, and the electronics digitize the waveforms. The data keep both the arrival times and the amount of light collected.

One sensor usually cannot describe a whole event. Reconstruction uses the pattern across the array: which sensors lit up first, which collected more light, which stayed dark. A dark sensor tells you something only if you know it was working and how sensitive it was.

The measured charge reflects how the sensor responded; it is not a direct reading of the neutrino’s energy. Turning a spread-out light pattern into a physical event needs models of the interaction and of how light is produced, travels and is detected.

Sources: [1] · [3]

Clear ice still changes the light

Absorption removes photons; scattering bends their paths and delays them. The glacial ice is not equally clear everywhere: its optical properties change with depth and direction, and the refrozen ice around each drill hole also changes how light reaches the sensors.

Calibration light sources send known flashes, and researchers compare what the sensors see with models of light propagation. Using several wavelengths and paths separates effects that one flash could not tell apart. The ice is part of the apparatus and has to be measured like the rest.

A wrong optical model can shift a reconstructed direction or change an energy estimate. Better calibration can therefore improve a result even from the same events. That makes better use of existing data; it does not add new independent observations.

Sources: [4] · [5]

One observatory, several observing tasks

DeepCore packs sensors more densely into part of the in-ice array to observe lower-energy interactions. IceTop, on the surface, measures air showers from cosmic rays. Each layout serves a different question.

When an analysis looks for interactions that start inside the array, light in the outer sensors helps reject charged particles coming in from above. Such a veto misses some particles and suits only certain analyses. Events that pass it are not automatically astrophysical, because atmospheric neutrinos can get through too.

Synchronized clocks, triggers, data transmission and monitoring keep all these measurements comparable. A useful burst of light has to be stored so that researchers can reprocess it later and still know the detector’s state and calibration at the time.

Sources: [6] · [1] · [7]

The work goes on after construction

Once sensors are frozen into deep ice, servicing them is nothing like opening a laboratory instrument. Reliable electronics and constant monitoring matter. Staff on site and the wider collaboration keep the observatory running through the Antarctic seasons.

A 2026 collaboration announcement describes the 2025–2026 Upgrade as installed, with first science data expected later in 2026. That was the status on that date. Installation, first data and proven improvement are separate milestones, and the expected performance is not yet a measured result.

Our Lab shrinks the apparatus to a small teaching array with a simplified light model. It lets you see why arrival times pin down a hidden track; its assumptions are listed separately. Its sensor count, fit precision and simulated data do not represent IceCube’s calibrated performance.

Sources: [8] · [9] · [10]

Try it in the Lab

Primary sources & revision

1. IceCube Collaboration · Instrumentation and Online Systems (2017; arXiv v3)
2. IceCube · Observatory overview
3. IceCube · IceCube and Neutrinos
4. IceCube · The end of the tenth polar season (5 March 2014)
5. IceCube Collaboration · Evidence of optical anisotropy of South Pole ice (ICRC 2013)
6. IceCube · Observatory science and subarrays
7. IceCube Collaboration · Evidence for High-Energy Extraterrestrial Neutrinos (2013)
8. IceCube · Working at the South Pole
9. IceCube · Francis Halzen, 2026 Physics Nobel Prize
10. IceCube · Public data releases

First published 2026-10-10; last revised 2026-10-10. Original explanatory text and diagrams by Sandbox Physics. Illustrations are schematic; no experimental event records are reproduced here. This is an independent educational publication, not an official Nobel or experiment collaboration publication.