PHYSICS
KATRIN’s 0.45 eV Limit and the Meaning of a Direct Neutrino Mass Measurement
KATRIN’s latest verified direct mass bound, its statistical meaning, and the distinct sterile-neutrino sensitivity study released in March 2026.
Research checked
Newest cited publication
A CHIPS neutrino-detector deployment provides general neutrino-experiment context. KATRIN instead infers a mass limit from the beta-decay spectrum.
Jerry Meier (University of Minnesota), via UCL Mathematical and Physical Sciences · Image source · CC BY 2.0
.jpg){kind=link}
A particle can be known to have mass long before anyone knows how much it weighs. Neutrinos occupy exactly that awkward position. Oscillation experiments establish differences between their squared masses, while the overall mass scale remains a separate experimental question. The Karlsruhe Tritium Neutrino experiment, KATRIN, attacks that question through radioactive decay. Its result published in Science in April 2025 placed the effective electron-antineutrino mass below 0.45 electronvolts divided by the speed of light squared, at 90 percent confidence. As of this article’s October 18, 2026 research cutoff, the sources checked here do not establish a newer absolute-mass bound. The important 2026 development is a sensitivity study for a different search, involving hypothetical sterile neutrinos. Keeping those two stories separate makes both much easier to understand.
A scale made from an energy balance
Imagine a decay in which a fixed energy budget must be divided among several participants. If one invisible participant needs some energy simply to exist, slightly less remains available to the visible participants. KATRIN examines that possibility in tritium beta decay, where an electron and an antineutrino emerge as the nucleus changes. It measures the electron distribution near its maximum permitted energy. The experiment therefore infers mass from the shape of a spectrum rather than catching and weighing individual neutrinos. This is a kinematic measurement: conservation of energy connects the missing share of the budget to the particle’s rest mass.
A useful idealized comparison is two distributions that agree almost everywhere but taper differently at their rightmost edge. Counting all events together would largely hide the difference. Concentrating on the edge increases sensitivity to the feature of interest, while sacrificing the enormous statistics available farther away. That tradeoff explains why a huge apparatus can be justified by a tiny region of a decay spectrum. The decisive quantity is the information carried by selected electrons, not the total number of decays occurring in the source.
Which neutrino mass is being bounded?
An electron-flavour neutrino is a quantum mixture of different mass states. A beta-decay experiment that cannot resolve their tiny individual spectral endpoints measures an effective mass assembled from those states and their electron-flavour weights. The squared effective mass is a weighted sum of squared masses. This differs from adding the masses together, and it differs again from the combinations tested in searches for neutrinoless double beta decay. A precise number needs the name of the quantity attached to it.
For a deliberately fictional two-state example, suppose the squared-mass weights were one quarter and three quarters, and the masses were one and two arbitrary units. The weighted squared mass would be one quarter times one, plus three quarters times four: 3.25 units squared. Taking the square root gives about 1.803 units. The ordinary weighted average would instead be 1.75. Neither calculation describes KATRIN’s measured parameters; the example shows why apparently similar mass summaries cannot be interchanged. Comparison requires translating through the mixing model rather than lining up headline numbers.
A confidence limit does not put the mass at the boundary
The published analysis used five measurement campaigns, with 36 million electrons recorded during 259 measurement days. Its fitted mass-squared parameter was slightly negative, with an uncertainty overlapping zero. That fit does not establish a negative physical mass. A statistical estimator can wander across the physical boundary when the true effect is smaller than the experiment can resolve. The reported upper limit is the physical inference obtained using the collaboration’s statistical procedure, rather than a claim that neutrinos weigh exactly 0.45 eV/c².
Consider an unrelated instrument estimating a faint positive light signal after subtracting a background. Repeated estimates can be positive or negative even though the lamp’s actual intensity cannot be negative. Discarding every negative estimate would distort the instrument’s average response. The same general issue appears whenever a small nonnegative physical quantity is inferred from noisy data. A confidence level describes the procedure’s behavior under repeated experiments and its stated assumptions; it should not casually be rewritten as a personal probability that the true mass lies in a particular interval.
Why collecting more electrons is only part of the job
The KIT release describes a roughly 70-metre beamline and a main spectrometer about ten metres across. Such dimensions are consequences of demanding source control, electron transport and energy discrimination. However, size alone does not protect the measurement against a small systematic offset. Electrons can lose energy before reaching the analysing region. Molecular excitations share part of the decay energy. Background electrons add counts. Electric potentials determine which electrons pass. Each effect can change the same spectral region used to infer neutrino mass.
In an illustrative measurement with 10,000 independent counts, the characteristic counting fluctuation is about 100 counts. Increasing exposure fourfold produces 40,000 counts with a fluctuation of about 200, improving the fractional fluctuation from one percent to half a percent. A calibration error that remains one percent has not improved at all. This simple square-root calculation explains the value of repeated operating campaigns and improved models. The experiment needs a stronger statistical signal and a defensible explanation of how the instrument transforms that signal into a spectrum.
Why the laboratory result matters beside cosmology
Cosmology asks how neutrinos influence the evolution of structure and radiation across the Universe. Oscillation measurements ask how flavour probabilities change during propagation. KATRIN asks what the decay spectrum permits. These are different observations of related physics. A cosmological constraint may become numerically tighter while depending on a chosen model of cosmic evolution. A laboratory constraint can remain valuable because its assumptions are tied to a much smaller physical system. Describing the latter as direct does not mean it is assumption free; it means the mass enters the local decay kinematics.
The productive comparison is a consistency test. Start with a proposed set of masses and mixing weights, compute each experiment’s relevant quantity, and check the corresponding likelihoods. If the predictions disagree with observations, investigate the full inference chain before declaring new particles or a failure of cosmology. An uncertain energy response and an uncertain cosmological parameter are not interchangeable uncertainties. Independent methods provide protection precisely because their possible mistakes have different causes. Agreement among them is more informative than repeatedly measuring the same combination with similar equipment.
The newer sterile-neutrino story is a different measurement
KATRIN also published a sterile-neutrino search in Nature in December 2025. A hypothetical extra mass state could add a kink to the electron spectrum, allowing the experiment to test a combination of extra-state mass and mixing strength. This is separate from establishing the absolute masses of the three familiar states. The absence of an allowed kink in a specified region excludes that region within the tested model; it does not exclude every conceivable sterile particle.
The collaboration’s March 24, 2026 TRISTAN preprint, revised April 15, studies a future extension to kiloelectronvolt masses using a new detector approach. It presents projected sensitivity from simulations, not a sterile-neutrino discovery or a new sub-electronvolt mass limit. Its discussion of systematic effects is particularly instructive: an attractive statistical reach can weaken substantially once instrument uncertainties are included. The scientific question shifts from the endpoint’s tiny overall deformation to rare admixtures that would leave a feature elsewhere in the spectrum.
What would change the interpretation next?
A convincing nonzero mass determination would require a reproducible spectral deformation, uncertainty small enough to separate it from zero, and convincing control of competing instrumental explanations. A stronger upper limit would instead exclude more of the possible mass scale without selecting a unique value. Both outcomes are useful. Their meanings differ, and a future result should be judged by the actual fitted quantity, confidence construction and data exposure rather than the smallest number in a press release.
My inference from this programme is that neutrino mass research is becoming a test of compatibility between several kinds of evidence. The next important step need not be a dramatic announcement. It may be a calibration improvement that makes a future measurement trustworthy, or a sterile-search analysis that rejects an otherwise tempting explanation of another experiment. KATRIN’s existing limit already demonstrates the value of asking a narrowly defined question exceptionally carefully. The wider neutrino puzzle becomes more tractable when every experiment states exactly which piece it has measured.
Sources & further reading
Primary papers, project documentation, and research releases used in this article. Publication dates below refer to the sources themselves.
- Direct neutrino-mass measurement based on 259 days of KATRIN dataKATRIN Collaboration / arXiv; published in Science in 2025Published Checked
- Astroparticle physics: Neutrinos weigh less than 0.45 electronvoltsKarlsruhe Institute of TechnologyPublished Checked
- Sterile-neutrino search based on 259 days of KATRIN dataKATRIN Collaboration / NaturePublished Checked
- KATRIN Sensitivity to keV Sterile Neutrinos with the TRISTAN Detector UpgradeKATRIN Collaboration / arXivPublished Checked
FOLLOW THE QUESTION
Explore physics