NA62 measures ultra‑rare K+→π+νν decay with precision under 20% and finds agreement with the Standard Model
Scientists working on the NA62 experiment at CERN report a new measurement of how often a positively charged kaon (K+) decays into a charged pion (π+) and a neutrino–antineutrino pair (νν̄). This decay is extremely rare: the chance is about one in 10 billion. Combining data taken from 2016 through 2024, NA62 measures the branching ratio — the probability of this decay — as (9.6 +1.9 −1.8) × 10^−11. That value is compatible with the Standard Model prediction and has a precision better than 20%.
The NA62 team studies kaons produced by the CERN Super Proton Synchrotron (SPS). They use a “decay‑in‑flight” method: a beam of 75 GeV particles that contains about 6% kaons is sent through a vacuum decay region, and detectors downstream measure the charged pion and veto any extra activity. Key detector components include a Cherenkov counter for kaon identification (KTAG), a silicon pixel beam spectrometer (GTK), a magnetic straw tracker (STRAW), a ring‑imaging Cherenkov counter (RICH), calorimeters (including a liquid krypton calorimeter, LKr), and a set of photon veto detectors (LAV, IRC, SAC). The analysis identifies signal candidates by computing the square of the missing mass from the measured kaon and pion momenta and placing strict signal regions that suppress the main kaon decay modes.
The 2023–2024 data brought measurable improvements. NA62 doubled its signal sample while keeping backgrounds down in proportion. Operational changes included running at 75% of the previous maximum beam intensity and switching the KTAG radiator gas from nitrogen to hydrogen to reduce material in the beam. New analysis tools were also introduced: a transformer‑based four‑dimensional GTK tracker reduced the chance of a missed kaon track from 6% to 4%, and a convolutional neural network (CNN) using calorimeter information improved separation of muons from pions. The combined RICH and calorimetric particle identification suppresses muon→pion misidentification by about 10^7, and the photon veto system rejects neutral‑pion (π0) backgrounds by about 10^8.