Read IceCube’s discovery paper, one figure at a time

Figure 1: how events were selected

Figure 1 of arXiv:1311.5238v2 is a drawing of the detector. It explains how events were selected; it does not show the events found. The shaded outer regions act as a veto: an event whose first light appears there is rejected. The aim is to tell a particle that entered from outside from an interaction that began inside.

Trace an imagined incoming muon with your finger. Which sensors should light first, and what could let it slip past the veto? Look closely at the dusty layer. It is an optical blind spot, so it can change which particle paths stay in the sample. The detector’s geometry is part of the argument.

The caption also names two observing seasons. A detector can change while it is being built, so one drawing may not fit every day of data. Counts and predictions can be compared only if both use the detector layouts and useful observing time of the analyzed sample.

Sources: [1] · [2]

Figure 2: each point is a reconstructed event

Figure 2 plots deposited energy against declination, the north–south position on the sky. Crosses mark track-like events and filled circles mark showers. Read the legend and the error convention first: symbols printed equally sharp do not have equally precise directions.

Compare the directions first and ignore energy; then compare the energies and ignore direction. A scatter plot easily blurs these two questions: where the reconstruction points, and how much energy the detector saw.

The energy deposited in the detector can fall short of the neutrino’s energy, because escaping particles carry some away. Read a large value as energy left in the ice before treating it as a property of the source. Gaps between points need a statistical test too: a small sample leaves empty intervals by chance.

Sources: [1] · [3] · [4]

Figure 3: why plot the square of radius?

Figure 3 compares where the first light appeared with the pattern expected from muons entering from outside. The radial coordinate is squared because rings of equal width hold different volumes: a ring farther from the cylinder axis holds more material.

In a uniform cylinder, equal steps in radius squared enclose equal cross-sectional area. A flat distribution in that coordinate is therefore consistent with interactions spread evenly through the volume. Squaring the radius is a geometric check, not a preference for any part of the detector.

Check which regions were excluded and whether the prediction excludes the same ones. What matters is how the selected events compare with the expected background pattern. An even-looking pattern alone cannot show that every remaining event came from far away.

Sources: [1] · [2]

Figure 4: compare counts with predicted counts

Figure 4 sets the observed energy and direction distributions against predictions: atmospheric components plus an extraterrestrial one. Read each colored component, then the total. A component curve is a prediction, not another set of measured events.

Before judging which curve fits, check what is plotted and how it is binned. A flux from the sky becomes an expected count only after the interaction probability, the detector acceptance and the energy response are applied. Laying a flux curve over an event histogram skips that step.

Then read how the text computes the significance. The result depends on how the two earlier events and the atmospheric model are treated. Subtracting one background number from 28, or putting a Gaussian error bar on the difference, does not reproduce the published result.

Sources: [1] · [4] · [3]

Figure 5: the source search asks a different question

Figure 5 maps the strength of a point-source fit across the sky. After correcting for having searched the whole sky, the strongest cluster had an 8% probability of arising by chance, and the paper identified no neutrino source from it. Even within one paper, the evidence for an extraterrestrial population and the search for a source support different claims.

In the NGC 1068 paper, look for the catalog of sources tested, the fitted signal, and the correction for searching many positions. The fitted signal is a statistical excess; it does not identify which events came from that galaxy. Direction and energy enter the fit together.

For each paper, note the version, the sample dates, the selection, what the axes mean, the competing explanations and the one sentence the evidence supports. Later papers can then be compared without quietly carrying one sample’s assumptions into another. The figure numbers above follow v2 of the 2013 paper.

Sources: [5] · [4] · [1]

Primary sources & revision

1. IceCube Collaboration · 2013 discovery paper, arXiv v2 (16 December 2013)
2. IceCube Collaboration · Instrumentation and Online Systems (2017; arXiv v3)
3. Francis Halzen · Astroparticle Physics with High Energy Neutrinos: from AMANDA to IceCube (2006)
4. Particle Data Group · Statistics (2025)
5. IceCube Collaboration · Evidence for neutrino emission from NGC 1068 (2022; arXiv v2, 2024)

First published 2026-10-11; last revised 2026-10-11. 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.