Open a real neutrino event file and read it carefully

Pick the release before making a plot

Start at IceCube’s official release page for the two-year contained-event search. It was posted on 21 November 2013 and marked as updated on 21 December 2015. Its archive holds eventsummary.txt, effective_areas.zip and eventviews.pdf. These files come from one past analysis; they are not the observatory’s full current data.

Keep the release URL, the download date and the archive checksum with the file, because a corrected upload can carry the same filename. For this exercise the file was checked on 11 October 2026; our scientific note lists the checksums of the archive and the text file.

The archive describes the 28 candidates selected in the 2010–2012 analysis. It does not contain every trigger or the raw sensor waveforms. First check that the IDs are unique and that the number of rows matches the sample definition. Each row is a candidate, not a confirmed astrophysical neutrino.

Sources: [1] · [2]

Read the units and qualifiers

The columns give the event ID, the deposited energy in TeV with asymmetric errors, the time as a fractional Modified Julian Date, declination and right ascension in degrees, the median angular resolution and the topology. The time column is a date written as a number. It is not a pulse delay or a length of observing time.

Right ascension and declination give the reconstructed direction on the sky, not the name of a source. For tracks, the angular field also has less-than entries. A parser that drops the less-than sign turns a reported bound into an exact value.

Shower and track describe the shape of an event. They do not pin down the neutrino flavor, because different interactions can look alike. Keep each number with its label, errors and units; do not shrink the table to one energy column and a set of sky points treated as exact.

Sources: [1] · [2] · [3]

Event 14: what one row can tell you

In the downloaded eventsummary.txt, event 14 is a shower with a deposited energy of 1,040.7 TeV, lower and upper errors of 144.4 and 131.6 TeV, and a median angular resolution of 13.2 degrees. These values come straight from the release; we did not simulate or refit them.

The row supports one precise description: the release reports a very large energy deposit in a shower-like event. It does not fix a unique energy for the parent neutrino, its flavor or its host galaxy. An angular resolution marks out a region of sky; it cannot say which object in that region sent the particle.

Event 6 has a best-fit deposit of 28.4 TeV, although the release title says energies above 30 TeV. The energy threshold is approximate, and the methods section explains how events were actually selected. A sharp cut of your own on the printed energies would quietly change the sample.

Sources: [1] · [2]

Describe the sample before making a physics claim

A good first exercise is to count the seven tracks and twenty-one showers while keeping all 28 IDs. Plot the released deposits with their errors, or compare the two groups. Label the result as a description of the selected sample.

Next, ask what you would divide by. A flux estimate needs the useful observing time and an effective area that depends on energy and direction. Predicting the reconstructed quantities also needs the detector response. The nested archive gives effective areas after the analysis cuts, split by flavor and direction; keep those splits.

The table alone cannot reproduce the published significance against an atmospheric origin; that needs the background model and the full test procedure. Nor can the event-view PDF stand in for the original sensor hits in a new fit. The data are still useful: these limits show which public-data exercises make sense.

Sources: [1] · [4] · [5]

A newer release is a separate dataset

For a current point-source analysis with tracks, use IceCube’s IceTracks-DR2 release from May 2026 and its linked Dataverse record. It covers observations from April 2008 to May 2022 and replaces the earlier track release as the recommended dataset. It does not extend the 28-row contained-event table.

The release includes response files and published benchmark comparisons made with public tools, and the versioned SkyLLH tutorial walks through the matching analysis. Keep the event selection, response files and software version together. Where old and new samples overlap, joining them would count the same exposure twice.

This exercise checks the historical file’s fields and counts and points to the newer source-search workflow. We have not rerun the full DR2 likelihood or verified any source significance ourselves. That would need the matched data products, a documented model and validated background trials.

Sources: [6] · [7] · [8] · [9]

Try it in the Lab

Primary sources & revision

1. IceCube · Two-year contained-event public release (2013; updated 2015)
2. IceCube Collaboration · 2013 discovery paper, arXiv v2 (16 December 2013)
3. IceCube · IceCube and Neutrinos
4. Particle Data Group · Statistics (2025)
5. IceCube Collaboration · Instrumentation and Online Systems (2017; arXiv v3)
6. IceCube · IceTracks-DR2 release page (21 May 2026)
7. IceCube Collaboration · IceTracks-DR2 (18 May 2026, v1)
8. IceCube SkyLLH · Public 14-year track tutorial, v26.0.0
9. IceCube Collaboration · IceTracks-DR2 Dataverse record

2026-10-11 / 2026-10-13 · Text and method diagrams are original to Sandbox Physics, and the diagrams are schematic. Some numbers are quoted from the named IceCube historical release. We have not reproduced sensor hits or refitted events. This is an independent educational publication, not an official Nobel or collaboration publication.