DONUT and the Tau Neutrino: Completing the Lepton Family

In the late 1950s and early 1960s, the electron and muon neutrinos were established as physically distinct particles through accelerator experiments at Brookhaven. A neutrino beam produced from pion decay was shown to produce only muons, not electrons, when it interacted with matter. The electron and the muon each had their own neutrino partner, and these were not interchangeable.

For the next quarter century, a standard expectation was that any further charged lepton would also have its own neutrino partner. When Martin Perl discovered the tau lepton at SLAC in 1975, finding a fourth elementary neutral fermion almost certainly had to follow. The tau’s decay products, like the muon’s, consistently included the kinematic signature of a missing neutral particle.

But the tau neutrino would not be directly observed for another 25 years. Not because it didn’t exist — its existence was never seriously doubted — but because of the particular experimental challenge of actually catching it. The DONUT experiment at Fermilab finally did so in 2000, closing the three-flavor Standard Model picture that had been assumed for over three decades.

This is the story of that delay, what DONUT actually measured, and why the tau neutrino still occupies a different experimental status than its two siblings.

The tau lepton arrives

Martin Perl’s 1975 discovery of a heavy charged lepton at the SPEAR electron-positron collider was itself a surprise. At 1.78 GeV, the tau was vastly heavier than the muon (106 MeV) and the electron (0.511 MeV). The identification came from anomalous $e\mu$ events — the products of $e^{+}{e}^{-}\to {\tau }^{+}{\tau }^{-}$ where one tau decayed leptonically to an electron and the other to a muon.

For each such event, two undetected neutrinos were kinematically required. These could, in principle, have been electron or muon neutrinos — but the simplest interpretation was that they were the tau’s own partner, the tau neutrino. Conservation of lepton number would then require that tau decay conserve “tau-ness” the way muon decay conserves “muon-ness”.

Perl received the 1995 Nobel Prize for the tau discovery. By that time, indirect evidence for the existence of a third neutrino was overwhelming: LEP’s precision measurement of the Z-boson decay width (1989) had determined the number of light active neutrino flavors to be $N_{\nu }=2.9840\pm 0.0082$ — consistent only with three, not two or four. The Super-Kamiokande atmospheric neutrino result (1998) required oscillations into a non-${\nu }_e$ flavor, most likely ${\nu }_{\tau }$. The community accepted the tau neutrino’s existence almost universally.

But no experiment had ever seen one.

The detection problem

Why so long? Three compounding difficulties.

First, tau neutrinos are rare at most accelerator sources. Standard pion decay $\pi \to \mu \nu$ produces muon neutrinos. Kaon decay produces mostly muon neutrinos. Charmed meson decay produces tau neutrinos via $D_s\to \tau {\nu }_{\tau }$, but the production rate is suppressed by the high mass of the charm quark. A dedicated tau-neutrino experiment needs a beam specifically enriched in ${\nu }_{\tau }$ — which means a proton beam with enough energy to produce charm.

Second, the charged-current interaction of a tau neutrino produces a tau lepton. The tau lives for only $2.9\times 10^{-13}$ seconds before decaying — a lifetime that, even at relativistic velocities, corresponds to a track length of about a millimetre at GeV-scale energies. The tau decay then produces a burst of other particles (pions, electrons, muons, or hadronic jets) but no externally tagged signature.

Third, and most severe: the distinctive feature of a tau neutrino interaction is the decay kink — a sudden change of direction where the tau ends and its decay products begin. Resolving that kink requires a detector with sub-millimetre spatial resolution. Standard electronic detectors (time projection chambers, Cherenkov arrays, wire chambers) had resolutions of centimetres or worse. Only one technology could routinely image a 1-mm kink with confidence: nuclear photographic emulsion.

Emulsion is a heavy photographic film, thick enough and chemically active enough that charged particles passing through leave sub-micrometre tracks when developed. It had been used for cosmic-ray physics since the 1930s (the pion was discovered in emulsion at Bristol in 1947). Its resolution is unmatched, but it has one practical drawback: every event must be located and analyzed by microscope, slide by slide, event by event. Analysis is extraordinarily labor-intensive.

DONUT

The Direct Observation of the NU Tau (DONUT) experiment at Fermilab was approved in 1994 for the specific purpose of catching ${\nu }_{\tau }$ events. Its design was straightforward in principle and extraordinarily demanding in practice:

A proton beam of 800 GeV hit a tungsten target, producing charmed mesons. The $D_s$ mesons decayed to $\tau {\nu }_{\tau }$ pairs; the taus themselves decayed to produce additional ${\nu }_{\tau }$. Downstream of the target, an iron filter absorbed all other particles except neutrinos. The surviving neutrino beam was dominated by muon neutrinos from kaon decays but contained a small tau-neutrino component, roughly 5% of the total flux.

The detector, 36 metres from the target, consisted of:

• Emulsion target stacks — about 260 kg of nuclear emulsion, providing the target mass for neutrino interactions and the spatial resolution to resolve tau decay kinks
• Scintillating fiber trackers — downstream of the emulsion, these provided tracking information that pointed back to specific regions of the emulsion, allowing the post-run scan to focus on the right few square centimetres rather than the entire 260 kg
• Drift chambers and calorimeters — for rough identification of the products of the neutrino interaction
• Magnetic spectrometer — for muon momentum measurement

The heart of the analysis was the combination: electronic detectors provided real-time triggers and localized the interaction vertex to within a few millimetres; emulsion, scanned after the run, provided the final resolution.

The analysis

DONUT’s beam exposure ran from April through September 1997. The emulsion was then developed and shipped to Nagoya University in Japan and the laboratory of the collaboration’s leader for scanning — a process that took approximately two years and involved several tens of thousands of microscope-hours distributed across multiple institutions.

From the triggered event sample, 203 neutrino interactions were identified in the emulsion. Of these, the scanning teams identified four events with a distinctive topology: a neutrino interaction vertex, a single emerging charged track approximately one millimetre long, and then a sudden change of direction with one or more additional charged particles emerging from the kink. This was the ${\nu }_{\tau }$ signature.

The background estimate — events that could mimic this topology from charm production or other sources — was less than one event. Four observed events with backgrounds below one constitute a $\gtrsim 5\sigma$ significance, and DONUT announced the first direct observation of the tau neutrino in July 2000. The paper was published in Physics Letters B in 2001.

The measured event rate was consistent with Standard Model predictions using the estimated ${\nu }_{\tau }$ flux in the beam. No precision cross-section measurement was extracted — the sample was too small and the beam uncertainties too large — but the signal was unambiguous.

Why it matters

DONUT’s discovery, unlike many experimental results, was not a surprise. The existence of the tau neutrino had been assumed for a quarter century and confirmed by indirect measurements of the Z-width and of atmospheric oscillations. What DONUT contributed was a direct observation — a photographic record of a specific neutrino-nucleus interaction producing a tau lepton that decayed after a millimetre of flight.

The result closed the Standard Model lepton triad as an observational fact. Every charged lepton now had a directly observed neutrino partner. The subsequent OPERA experiment at Gran Sasso, using the same emulsion technique, observed tau-neutrino appearance in an oscillated muon-neutrino beam from CERN (2010–2015), bringing the tau-neutrino count to a few tens of events across all of experimental particle physics.

As of 2026, the total number of directly observed tau-neutrino interactions in the history of physics remains of order 25. Compare that to the trillion-plus electron-neutrino and muon-neutrino events that Super-Kamiokande alone records per year. The tau neutrino is, quantitatively, the least-studied particle in the Standard Model — not because it is disputed, but because it is so hard to catch.

What remains

Tau-neutrino physics continues to be a frontier. Three active programmes:

IceCube-Gen2 at the South Pole will, by the 2030s, accumulate hundreds of astrophysical tau-neutrino events at TeV to PeV energies. The flavor ratio of astrophysical neutrinos at Earth — predicted to be near 1:1:1 after oscillation averaging — will be tested with these events, probing new physics at scales inaccessible to accelerators.

FASERnu and SND at the LHC are small emulsion-based detectors downstream of the LHC interaction points at CERN, catching tau neutrinos from LHC hadron collisions. In LHC Run 3, they are expected to observe roughly 10 ${\nu }_{\tau }$ events — a factor-of-two increase in the world sample.

DUNE will record tau neutrinos produced by oscillation in its long-baseline beam. The tau neutrino’s hadronic decay channels are indistinguishable from neutral-current interactions in liquid argon, making appearance-channel identification difficult, but the electron-channel decays remain clean.

Every one of these programmes will slowly, painfully, add events to the historical sample. The tau neutrino remains the flavor of patience: rarely produced, briefly observed, and always catching up to its two more abundant siblings.