Skip to main content
Sandbox Physics

Neutrinos · D11 · Catch the first message

Supernova Neutrino Burst Observatory

When should a sudden jump in neutrino events trigger a supernova alarm? When background alone would rarely produce it, even after you count every window you searched. Fix the expected background rate first. If you slide a time window through the data and pick the busiest stretch, the chance for that one window understates the false-alarm risk. Bound the chance that any window gets that busy; the single-window chance times the number of windows is one such bound.

Catch a simulated neutrino burst in a spherical scintillation detector. Freeze the event times, scan a trigger window across them, and follow one interaction from its flash of light to the collection sphere.

Interactive modelSupernova Neutrino Burst Observatory
Recorded interactions—\text{—}
Peak window count—\text{—}
Scan tail upper bound—\text{—}

The same forty-second record

The detector is quiet, then suddenly busy.

Recorded event timeline

Mint bars: sampled counts. Shading: peak window. Solid line: selected event; dashed line: replay clock. Optional gold line: injected expectation, not recovered luminosity. Zero is the prescribed arrival onset at Earth.

The sliding-window scan

Every full window is evaluated on a fixed quarter-second grid. The first crossing is a simulated decision time, not a fitted stellar collapse time.

Look at one interaction

From a point flash to the sphere

Ideal prompt isotropic light; 64 rays are geometric samples, not photoelectron yield. The full 5 m collection sphere and fixed index 1.5 omit scintillation delay, absorption, scattering and sensor response.

Model assumptions & primary sources

Illustrative mean: 120 detected candidates at 10 kpc and exposure one, scaled by exposure and inverse square of distance. Exposure changes candidate yield only; it does not resize the schematic sphere or scale the independently controlled background. Cooling is a truncated exponential over 30 seconds; Fast onset mixes 30% of a 0.4 second component and 70% cooling. Background is stationary Poisson. Uniform vertices fill the inner 4 m sphere, carry no sky direction and do not enter the trigger. No energy, cross-section, efficiency or flavor model; no SN 1987A events are reproduced.

Read SN 1987A →Explore the global history →

Physics tutorial

A trigger is a decision, not a discovery.

BackgroundIn SN 1987A, independent underground instruments each recorded a short burst of neutrinos, and these bursts were linked to a stellar explosion seen in the sky. Today SNEWS combines reports from many detectors to give early warning, but each detector first has to decide whether its own stream of events looks unusual.

Why it mattersA nearby source can produce a rush of interactions; a distant one may leave only a few. A lower count threshold catches more weak bursts, but steady background can cross it too. This Lab keeps sampling, replay and that decision apart.

Start with the essentials

Focus question
How much does a convincing-looking cluster depend on your window, threshold and assumed background?
One-sentence intuition
The same timestamps can lead to different trigger decisions, and a peak found by scanning many windows needs a probability bound for the whole scan, because its single-window tail alone is not the false-alarm probability.

Core mathematical model

Expected signal count from distance and exposure

Nˉs=120 a(10 kpcd)2\bar N_s=120\,a\left(\frac{10\,\mathrm{kpc}}{d}\right)^2

The normalization is illustrative and does not predict counts for any named instrument. Relative exposure changes only the signal yield; the background has its own control, and the schematic chamber keeps the same size.

Assumed cooling profile

F(t)=1−e−t/τ1−e−30 s/τ,0≤t≤30 sF(t)=\frac{1-e^{-t/\tau}}{1-e^{-30\,\mathrm{s}/\tau}},\qquad 0\le t\le30\,\mathrm{s}

Signal events arrive as a Poisson process whose rate changes in time; the expected number up to any moment is the mean yield times this cumulative profile. Fast onset adds a short component that carries 30% of the signal. There is no hydrodynamic supernova or flavor-evolution calculation behind it.

Count the events in one full window

Cj=∑i1{uj−W<ti≤uj},Cj≥k ⇒ candidate triggerC_j=\sum_i\mathbf 1\{u_j-W<t_i\le u_j\},\qquad C_j\ge k\ \Rightarrow\ \text{candidate trigger}

Window end times sit on a fixed quarter-second grid. The rule uses only the observed timestamps; the model's origin labels, the event vertices and the injected onset time do not enter the trigger.

Correct for scanning many windows

pk=Pr⁡{Poisson(bW)≥k},Pr⁡{max⁡jCj≥k}≤min⁡(1,Mpk)p_k=\Pr\{\mathrm{Poisson}(bW)\ge k\},\quad\Pr\{\max_j C_j\ge k\}\le\min(1,Mp_k)

The union bound stays conservative even though the windows overlap. It assumes a known, steady Poisson background. It covers one fixed choice of window and threshold; it does not cover retuning after looking at the record, or continuous monitoring over a year.

Light travel time from one flash

L=−r⋅q^+(r⋅q^)2+R2−∣r∣2,thit=nLcL=-\mathbf r\cdot\hat{\mathbf q}+\sqrt{(\mathbf r\cdot\hat{\mathbf q})^2+R^2-|\mathbf r|^2},\qquad t_{\mathrm{hit}}=\frac{nL}{c}

Each representative ray starts at the selected event vertex and ends on the same ideal 5 m collection sphere. The refractive index is 1.5, coordinates are in meters and optical times are in nanoseconds. The number of rays is not a photoelectron yield. The scintillation emission delay and the detector response are left out.

Common difficulties

One window's tail is not the whole search

Typical misconceptionThe smallest single-window tail is the false-alarm probability for the entire scan.

Better mental modelPicking the peak gives a fluctuation many chances to appear. The scan bound counts every window on the prescribed grid; any further choice of model or settings needs its own correction.

Uniform vertices do not point to a star

Typical misconceptionThe bright pattern inside the sphere shows the direction of the supernova.

Better mental modelThe deposits are prescribed to be uniform, only to illustrate the optics. The model has no directional interaction, no triangulation and no sky reconstruction.

Run the experiment

  1. 01

    Catch a burst

    Acquire Galactic burst and replay the event clock. Watch the half-second counts and the first threshold crossing.

    What to observe: The clock plays back a record that is already frozen; it does not generate new events. Gold marks the selected vertex and recent interactions, and mint marks the events collected so far.
  2. 02

    Keep the record, change the decision

    Change the window duration and the count threshold in the analysis panel.

    What to observe: The event times stay fixed, but the peak count, the candidate crossing and the false-trigger bound all change together.
  3. 03

    Try background alone

    Acquire Background only, restore this record’s background assumption, then lower the threshold.

    What to observe: A background fluctuation can set off the trigger. The probability bound says how rare that is under the stated background; it does not turn the trigger into a stellar discovery.
  4. 04

    Follow one interaction

    Select an event on the timeline, choose Follow this flash in 3D, then drag the optical clock.

    What to observe: Representative light rays stop on the collection sphere, and the histogram is computed from those same path lengths. The burst clock and the optical clock describe different time scales.

Related experiments and articles

Continue exploring

How detectors see a neutrino

A neutrino leaves only light, charge or a pair of signals close in time. How does each detector turn that into a direction, an energy or a source?

  1. LabReactor Antineutrino CoincidenceSelect prompt and delayed signals against accidental pairs.
  2. LabCherenkov Light & Detector RingsTurn charged-particle geometry into a recorded light pattern.
  3. LabLiquid Argon Time Projection ChamberTurn electron arrival times and wire projections into calibrated slice centroids.
  4. LabIceCube: From Sensor Hits to a TrackInfer a track direction from the recorded sensor times.
  5. You are hereSupernova Neutrino Burst ObservatoryScan a frozen burst record, then inspect one interaction at the optical timescale.
  6. LabNeutrino Sky Search ObservatorySearch frozen neutrino directions and compare a local hotspot with paired finite-grid background trials.

Neutrinos · Reading path

Build a new way to observe the sky

Articles on this path that use this experiment, followed by the other experiments they suggest.

  1. ArticleSN 1987A: a supernova seen from underground
  2. LabReactor Antineutrino Coincidence
  3. LabCherenkov Light & Detector Rings

Neutrinos · Reading path

Read evidence, data and worldwide contributions

Articles on this path that use this experiment, followed by the other experiments they suggest.

  1. ArticleWhen can a few neutrino events become evidence?
FeedbackSomething broken, confusing, or missing on this page?Tell us
What kind of feedback?

Sent with this page’s address, its Lab ID, the site language and your device type. Please leave out personal details. Feedback is kept for up to 12 months.Privacy policy