KATRIN: measuring neutrino mass through tritium, fields and detector response

The Karlsruhe Tritium Neutrino experiment, KATRIN, placed an upper limit of 0.45 electronvolts divided by the speed of light squared on the effective electron-antineutrino mass, at 90% confidence. The result appeared in Science in April 2025, following a June 2024 preprint. At Karlsruhe in Germany, researchers examined 36 million electrons collected over 259 measurement days. Tritium decay releases an electron and an electron antineutrino; the antineutrino's mass affects the electron spectrum near its maximum energy. Diana Parno of Carnegie Mellon University in Pittsburgh expected mass measurements to continue until about the end of 2025. The laboratory limit comes from decay kinematics and does not require assumptions about cosmic evolution.

Filtering electrons and suppressing spectrometer background

KATRIN guides decay electrons magnetically from its gaseous source toward electrostatic spectrometers. Differential and cryogenic pumping reduce tritium flow before it reaches the spectrometer section. The main spectrometer combines magnetic adiabatic collimation with an electric retarding potential: electrons passing the energy threshold reach a silicon detector divided into 148 pixels. Scanning different retarding potentials produces an integrated energy spectrum. Magnetic fields align the electrons' motion so that the electric filter can separate them by energy.

The shifted-analysing-plane configuration moves the region of highest retarding potential approximately six metres toward the detector. A smaller downstream volume then contributes background electrons. This mode, standard since 2020, reduced spectrometer background by roughly half while preserving resolution below 2.8 electronvolts. Its less uniform fields require direct calibration with conversion electrons from gaseous krypton-83m. Narrow krypton lines improve the determination of transmission parameters, which enter the neutrino-mass analysis; the field configuration's background reduction therefore accompanies a separately measured spectrometer response.

Measuring the electrical conditions inside the source

The tritium source contains a cold plasma produced through beta decay and subsequent ionization, dissociation and recombination. Its electrical potential depends on gas conditions, external fields and the surrounding surfaces. An uneven potential can distort the electron spectrum before the electrons reach the spectrometer. Co-circulating metastable krypton-83m with tritium provides conversion-electron lines for measuring that spatially extended potential under source conditions. A strong L3-32 line supports rapid checks of potential shifts, while the narrow N-line doublet helps determine broadening.

Dedicated campaigns in 2021 and 2023 tested different gas densities and rear-wall voltages. One high-statistics measurement accumulated 180 spectra over 25 days at 75% of the nominal tritium column density. The inferred broadening differed between campaigns, prompting a conservative estimate for the mass analysis. Surface work-function drifts remained a possible explanation. Further krypton measurements were planned for 2025 to investigate the changes. These measurements distinguish source-potential broadening from the separate field calibration of the main spectrometer.

Calibration electrons with selectable energy and angle

A monoenergetic, angular-selective photoelectron source supplies another calibration tool. An upgraded version installed in KATRIN's beamline in February 2022 became the subject of a technical preprint in March 2026. It can provide electrons with energies up to 32 kiloelectronvolts and vary their angle relative to the magnetic field. Energy and direction can therefore be selected when testing how electrons interact with the apparatus, rather than relying entirely on the broad distribution produced by radioactive decay.

These controlled electrons support several measurements: scattering from tritium molecules in the gaseous source, angle-dependent backscattering at the focal-plane detector, and adiabatic transport through the main spectrometer. Each concerns a different part of the electron's passage through KATRIN. The adjustable source allows those effects to be investigated under controlled input conditions. The calibration measurements characterize electron transport and detector response.

Accounting for molecular final states

Tritium's molecular environment also enters the interpretation of its beta spectrum. The calculation of molecular final states contributes uncertainty to KATRIN's mass measurement. Early analyses used a conservative phenomenological estimate of 0.02 electronvolts squared divided by the speed of light to the fourth power for that contribution. Constants, parameters and functions in the final-state calculation, together with its numerical convergence, received a more detailed uncertainty treatment published in October 2023.

The molecular wave functions are expanded in a basis set. Changing the size of that set tests whether the calculated final-state distribution has converged sufficiently. Uncertainties from the calculation are propagated into the experimental observable, the squared neutrino mass. Under the conditions of KATRIN's first measurement campaign, the revised procedure constrained the final-state contribution to 0.0013 in the same squared-mass units. That figure applies to the specified campaign conditions and this uncertainty component, rather than representing the uncertainty of every later neutrino-mass result.

Removing tritium activity from the rear wall

A gold-coated stainless-steel disk forms the physical boundary at the rear of the tritium source and defines a reference potential. Exposure to tritium causes adsorption and absorption, leaving activity that can contribute to the measured spectrum. During maintenance, with the source emptied to less than one hundred-thousandth of its nominal gas density, researchers monitored accumulated activity using beta-induced X-ray spectrometry and a silicon detector that observed emitted beta electrons. The activity's dependence on accumulated tritium throughput was also investigated.

Several cleaning approaches produced only small reductions when used separately. They included heating during evacuation, flushing with different gases, and ultraviolet illumination. A combination of ultraviolet light, a heated surface and air at pressures from five to one hundred millibars generated reactive ozone and reduced surface activity by three orders of magnitude in less than a week. The comparison concerned in-situ rear-wall decontamination under the tested maintenance conditions. It reduced a particular source of unwanted beta activity, while the experiment's other background and calibration contributions required their own measurements.

Electrons can lose energy by scattering from molecules as they cross the dense gaseous source. Those losses alter the observed beta spectrum and must be characterized accurately. Measurements in 2022 and 2023 used a newly installed high-rate photoelectron source with a narrow energy and angular distribution to investigate losses in tritium at different electron energies. Sonja Schneidewind and colleagues presented improved methods for analysing these data at the March 2024 DPG meeting in Karlsruhe. The work extended earlier energy-loss measurements. The controlled-electron measurements and their analysis address the spectrum's scattering response, complementing the source's separately studied electrical potential and molecular final-state calculation.

Testing detectors before installation

TRISTAN is a silicon drift detector system developed for a later KATRIN programme covering the full tritium electron spectrum. That broader measurement would search for hypothetical neutrino mass states in the kiloelectronvolt range. Before installation in KATRIN's beamline, detector modules require characterization with electrons of known energy. A custom thermionic electron gun supplies such electrons, with kinetic energies up to 25 kiloelectronvolts and rates around 100,000 electrons per second. Its operation uses thermionic emission, rather than the photoelectron mechanism of the angular-selective calibration source.

The gun's design and commissioning accompanied a first measurement of the TRISTAN detector's electron response. A monoenergetic input allows the measured detector signal to be compared with a selected incident energy. This is a detector-characterization measurement before beamline installation. Its stated electron rate describes the test source; it is separate from the intended incident rate per pixel in the eventual multi-channel readout. The hardware test contributes to understanding how the proposed detector system would record the beta spectrum.

Tracking electrons that scatter from silicon

Electron backscattering from a silicon drift detector's surface changes the shape of the observed energy spectrum. TRISTAN's proposed sterile-neutrino search therefore requires a detailed model of this part of the detector response. Measurements used a tandem silicon drift detector arrangement and a custom electron source to investigate backscattering. The measured response and backscattering probability were compared with dedicated simulations built using the Geant4 toolkit. Both comparisons showed good agreement under the tested conditions.

The investigation supported a proposed search targeting a mixing amplitude of one part in a million. That number was the collaboration's sensitivity goal, rather than an observed sterile-neutrino signal or an exclusion established by the detector test. A distortion associated with a hypothetical particle must be distinguished from a distortion caused by electrons interacting with the detector itself. The tandem-detector measurements constrain the latter effect, while simulations connect the measured response with the model used for electron spectroscopy.

Mapping charge collection and signal timing

A seven-pixel TRISTAN prototype was also characterized with a focused laser beam. High-resolution scans across its surface measured charge collection and the timing of generated signals. The measurements showed how both quantities depend on the detector's geometry. A laser scan provides localized illumination, allowing the response to be investigated at different positions rather than only through an average over the whole detector. The resulting experimental data were compared with simulations, with good agreement.

This characterization concerns a small prototype and its response to the laser system. It supports understanding of the silicon drift technology proposed for TRISTAN's future electron spectroscopy. The intended sterile-neutrino measurement follows completion of KATRIN's neutrino-mass campaign. Position-dependent charge collection and timing are therefore detector-development questions within that later programme. The prototype scans do not constitute a full-array neutrino measurement: they examine how geometry influences signals that the larger system's acquisition electronics would eventually have to identify and process.

Digitizing and processing a high-rate spectrum

The TRISTAN readout is designed for more than a thousand silicon drift detector pixels, each recording incident rates of about 100,000 electrons per second. Its remote analog-to-digital conversion architecture separates early digitization from subsequent processing. Front-end boards close to the vacuum chamber minimize analog signal paths, accommodate the magnetic-field and high-voltage environment, and send digitized waveforms through optical links. Back-end field-programmable gate arrays operate outside the detector's stray magnetic fields at earth potential, with a server handling data management.

Digital processing reconstructs event timing and deposited energy from detector pulses. A fast trigger filter and an energy filter have different tasks; longer energy-filter responses improve resolution but increase overlap between nearby events. Event flags allow affected measurements to be treated during analysis. Coincident depositions within the detector's rise time can remain unresolved. The system identifies several effects that can bias the spectrum:

• Pileup from successive energy depositions in one pixel.
• Charge sharing between neighbouring pixels.
• Crosstalk and distorted waveforms.

Forecasting a search for heavier sterile neutrinos

A hypothetical sterile neutrino with a mass of several kiloelectronvolts could produce a kink-like distortion in the beta-electron spectrum. TRISTAN would investigate energies well below the endpoint used for the light-neutrino mass measurement, using differential spectroscopy at high rates. The proposed search's sensitivity was evaluated with a dedicated simulation framework in work first posted in March 2026. With four months of detector live time, the statistical reach extended to mixing amplitudes around one part in a million for sterile-neutrino masses between four and thirteen kiloelectronvolts.

Experimental systematic uncertainties weakened that projected sensitivity by factors of ten to fifty across the same mass range. The statistical estimate therefore cannot stand alone as a forecast of the complete experiment's performance. Detector response and other experimental uncertainties remain part of the sensitivity calculation. These are simulated capabilities for a hypothetical particle search after KATRIN's mass programme, rather than a reported detection or a measured exclusion at the projected mixing amplitude. A sterile particle's possible role as dark matter also remains conditional on its existence and properties.

A separate search close to the endpoint

KATRIN has already used endpoint-region data to investigate a different sterile-neutrino mass range. An analysis first posted in March 2025 searched the final forty electronvolts below the tritium endpoint for a discontinuity in the spectrum's curvature. This kink signature depends on spectral shape rather than absolute normalization. The approach provides a laboratory comparison with reactor and radioactive-source experiments whose anomalous results motivated possible additional neutrino states. The proposed sterile state does not interact through the weak force.

The resulting constraints excluded a substantial part of the parameter space associated with the gallium anomaly and challenged the Neutrino-4 claim. Combined with other neutrino-disappearance experiments, KATRIN probed sterile-to-active mass splittings from fractions of an electronvolt squared to several hundred electronvolts squared, excluding mixing angles above a few percent. Those statements describe the tested parameter ranges. They do not exclude every possible sterile-neutrino model or replace the separate kiloelectronvolt-range sensitivity projections for the TRISTAN upgrade.

Testing interaction terms with the same spectroscopy

Precision beta spectroscopy also allows constraints on general neutrino interactions. An effective field-theory description incorporates theoretically allowed interaction terms and their effects on the tritium beta spectrum. Specific terms produce characteristic spectral changes, allowing the measured distribution to constrain their effective parameters. KATRIN's second measurement campaign, conducted in 2019, supplied four million electrons for the first such constraints. The analysis was initially posted in October 2024 and subsequently published in Physical Review Letters in 2025.

Selected interaction cases included a right-handed W boson, a charged Higgs and leptoquarks. These were possibilities examined through their effects on the spectrum, rather than newly detected particles. The same precision electron measurement can therefore be analysed with different physical questions and corresponding models. A direct effective-mass fit, a search for a sterile-neutrino kink and constraints on interaction terms use distinct signatures and assumptions. The interaction analysis yields parameter constraints within its effective description; it does not establish any of the proposed new interactions.