Skip to content

Analysis · Neutrino Physics

Neutrino Mass Tension: DESI, KATRIN and the Oscillation Floor

Some cosmological fits push the total neutrino mass towards or below the minimum required by oscillations. KATRIN measures a different effective mass. Understanding the tension means keeping the datasets, confidence conventions, statistical boundaries and dark-energy assumptions beside each quoted number.

Abstract illustration of three translucent forms between a cosmic web and ordered light planes, representing neutrino mass studies in cosmology and the laboratory.

The neutrino mass tension is a mismatch between some cosmological analyses and the minimum total mass implied by neutrino oscillations. Its strength depends on the datasets, cosmological model and statistical method. KATRIN provides a separate laboratory upper limit on a different mass combination; that limit does not itself contradict cosmology.

What is the neutrino mass tension?

Neutrinos have at least two nonzero masses, but their absolute mass scale remains unknown. Experiments that observe flavour changes establish differences between squared masses. Cosmologists instead infer the total mass from the universe’s expansion and structure. The interesting question is whether those independent descriptions fit together. For the underlying particles, start with What Are Neutrinos?.

The best-known DESI DR2 combination gives a total-mass upper limit just above the minimum allowed by normal ordering. Other statistical treatments place the limit below that minimum or report a stronger preference for an effective parameter outside the physical mass range. Those statements have different assumptions and should retain their labels.

This makes the tension a test of the whole inference: data calibration, cosmological modelling, statistical choices and neutrino physics. A small fitted mass is informative, but it is not an unconditional measurement that overturns oscillation experiments.

Four measurements, four different mass quantities

The Particle Data Group review distinguishes four quantities that often appear together in neutrino headlines. The labels m₁, m₂ and m₃ refer to mass states. Electron, muon and tau refer to flavours, which are mixtures of those states.

The mixing-matrix elements Uei specify the electron-flavour contribution of each mass state. A beta-decay measurement weights the squared masses by those mixing probabilities. A cosmological mass sum simply adds the three masses. Neither is the mass of a single flavour treated as an isolated mass eigenstate.

For the equations below, masses are written in electronvolts using c = 1, as is customary in particle physics. The corresponding rest-mass unit with c displayed is eV/c².

Four mass measurements compared

MethodQuantityWhat it can establish
OscillationsΔm²ij = m²i − m²jMass-squared gaps and mixing; no absolute mass scale by themselves.
Beta decaym²β = Σi |Uei|² m²iA kinematic effective mass, inferred from the endpoint spectrum.
CosmologyΣmν = m₁ + m₂ + m₃A total-mass constraint conditional on cosmological assumptions.
Neutrinoless double-beta decaymββ = |Σi U²eimi|An effective Majorana mass for the light-neutrino exchange mechanism.
The last expression includes phases and can exhibit cancellations. The four observables are not interchangeable.

What did KATRIN actually measure?

KATRIN measures the electron spectrum near the endpoint of tritium beta decay. An emitted antineutrino affects the energy available to the electron. The 2025 Science result used 36 million electrons recorded over 259 measurement days and reported mβ < 0.45 eV at 90% confidence.

That is an upper limit, not a detection at 0.45 eV. It allows a much smaller effective mass. A cosmic analysis favouring a small total mass can therefore be compatible with it. There is no valid calculation in which dividing 0.45 by a cosmological upper limit measures a disagreement between experiments.

The laboratory method avoids assumptions about the universe’s expansion history and dark energy. It still needs a model of the beta spectrum, molecular final states, energy loss, the detector response and backgrounds. Calling it independent of cosmology is more precise than saying it assumes nothing.

The 2026 KATRIN review by Drexlin and Weinheimer, revised on 8 October, describes the larger dataset and a target sensitivity below 0.3 eV. A sensitivity projection is not a newly measured limit and should not replace the published 0.45 eV result.

How oscillations impose a minimum total mass

Oscillations measure two independent mass-squared gaps, approximately 7.5 × 10−5 eV² and 2.5 × 10−3 eV². Their meaning and experimental evidence are explained in How We Know Neutrinos Have Mass. Once an ordering is specified, these gaps determine the other masses as functions of the lightest one.

In normal ordering, take the smallest possible lightest mass, m₁ = 0. Then m₂ ≈ √(7.5 × 10−5) = 0.0087 eV and m₃ ≈ √(2.5 × 10−3) = 0.050 eV. Their sum is about 0.059 eV. Increasing m₁ raises the sum.

In inverted ordering, m₃ is lightest. Setting it to zero leaves two states near 0.05 eV each, producing a minimum sum near 0.10 eV. These are approximate consequences of measured splittings within the three-neutrino framework, with uncertainties inherited from those splittings. They are not direct measurements of the lightest mass.

The minimum mass spectra for normal and inverted orderingCalculated example with the lightest neutrino mass set to zero. The approximate normal-ordering masses are 0, 8.7 and 50 meV; the inverted-ordering masses are 50, 50.7 and 0 meV. m 1 m 2 m 3 0 25 50 Mass (meV) 0 8.7 50.0 m 1 m 2 m 3 0 25 50 Mass (meV) 50.0 50.7 0 How oscillations set a minimum total mass Illustration: the lightest mass is zero. Normal ordering Total ≈ 0.059 eV Inverted ordering Total ≈ 0.101 eV Computed with Δm²₂₁ = 7.50 × 10⁻⁵ eV² and |Δm²₃₁| ≈ 2.50 × 10⁻³ eV². Approximate spectra; no uncertainty bands.
Original calculation using Δm²₂₁ = 7.50 × 10⁻⁵ eV² and |Δm²₃₁| ≈ 2.50 × 10⁻³ eV². The lightest mass is set to zero for illustration. Values are rounded; these are possible limiting spectra, not measurements of the individual absolute masses.

How cosmology constrains neutrino mass

Relic neutrinos contribute to cosmic energy density. Their transition from relativistic to nonrelativistic behaviour depends on mass, affecting the expansion history. Their motion also suppresses the growth of matter structure on some scales. Cosmological inference compares those predicted effects with observations while fitting other parameters at the same time.

DESI’s baryon acoustic oscillation measurements constrain distances and expansion. The cosmic microwave background supplies early-universe information, and its gravitational lensing adds another constraint. The DESI cosmology paper combines these probes. A bound from such a combination is not a measurement made by DESI alone.

Different probes constrain different aspects of neutrino physics. Distance-based information and the full shape of the matter power spectrum are related but not identical tests. Their combination can be powerful precisely because the assumptions and degeneracies differ.

The standard ΛCDM calculation assumes a cosmological constant, cold dark matter and a standard relic-neutrino population, with neutrinos stable over cosmological times. Changing those assumptions changes how an observed expansion history or amount of structure translates into a neutrino mass.

Why DESI’s 0.0642 eV and 0.053 eV limits differ

The DESI neutrino-physics paper reports Σmν < 0.0642 eV at 95% credibility for its baseline DESI DR2 BAO plus CMB combination in ΛCDM, with three degenerate masses and a positive-mass prior. This lies slightly above the approximate normal-ordering floor.

Using a frequentist profile and the Feldman–Cousins construction to handle the physical boundary gives a 95% upper limit of 0.053 eV for that baseline combination. The smaller number is a different statistical result, rather than an extra observation that simply supersedes the first.

The equal-mass approximation is a cosmological modelling choice, not a claim that oscillation experiments found equal masses. A physical ordering fit must incorporate the measured splittings.

Near a boundary such as zero mass, the treatment of parameter space matters. Naredo-Tuero and colleagues’ critical analysis examined the effects of statistical methods, CMB likelihoods and data selection using an earlier DESI dataset. It illustrates why a quoted limit needs its full specification.

The DESI paper also reports about 3σ tension with the oscillation minimum in an extended analysis allowing a signed effective cosmological mass parameter. That significance belongs to that analysis. It is not a universal significance for a disagreement between KATRIN and the sky.

Three cosmological upper limits with different assumptionsDESI DR2 BAO and baseline CMB give total-mass upper limits of 0.0642 eV in Bayesian Lambda CDM, 0.053 eV with a Feldman-Cousins construction, and 0.163 eV in a Bayesian evolving-dark-energy model. Normal and inverted oscillation floors are approximately 0.059 and 0.10 eV. 0.00 0.05 0.10 0.15 Total neutrino mass, Σ mν (eV) 0.00 0.05 0.10 0.15 Total neutrino mass, Σ mν (eV) 0.00 0.05 0.10 0.15 Total neutrino mass, Σ mν (eV) Why the quoted upper limit changes DESI DR2 BAO + baseline CMB data ΛCDM · Bayesian 95% Σ mν < 0.0642 eV ΛCDM · Feldman–Cousins 95% Σ mν < 0.053 eV Evolving dark energy · Bayesian 95% Σ mν < 0.163 eV – – Normal-ordering floor ≈ 0.059 eV ···· Inverted-ordering floor ≈ 0.10 eV Upper limits; different assumptions. Source: Elbers et al., arXiv:2503.14744.
Original figure from the baseline results in W. Elbers et al., arXiv:2503.14744, Phys. Rev. D 112, 083513 (2025). Arrows denote upper limits, not measured masses or probability distributions. Bayesian rows are 95% credible limits; Feldman–Cousins is a 95% confidence limit. These cosmological bounds use a degenerate-mass parametrisation without imposing an oscillation lower bound, and differ in statistical or cosmological assumptions.

What does a negative effective neutrino mass mean?

Ordinary positive neutrino masses produce a specific pattern of changes in the expansion and clustering predictions. Researchers can extend a fitting parameter to negative values to ask whether the data prefer effects in the opposite direction. That extension is a diagnostic of the fit, not evidence that a laboratory neutrino has negative rest mass.

Green and Meyers explored how CMB lensing contributes to this preference. The interpretation is conditional on the data and model used. A best fit outside a physical parameter range can motivate tests of model assumptions and systematics without uniquely identifying which assumption needs revision.

KATRIN offers a useful caution from a different setting. Its fitted mass-squared value can fluctuate below zero even though the physical mass is nonnegative. The reported spectrum fit and physical upper limit therefore require a stated statistical procedure. This laboratory situation is not the same parameter extension used in cosmology, but it shows why an unphysical best fit should not be read literally.

Can changing dark energy relieve the tension?

Yes, in some analyses. Elbers and colleagues’ dark-energy study shows how an evolving dark-energy model can remove a preference for negative effective neutrino mass. Both dark energy and neutrino mass influence expansion, so their inferred values can be correlated.

In the DESI neutrino paper, allowing the dark-energy equation of state to evolve in the w₀wₐCDM model relaxes the baseline total-mass upper limit to 0.163 eV at 95% credibility. That model admits sums above either approximate oscillation floor. It is an example of how much a bound can depend on the model.

A complementary frequentist study led by Chebat also investigates expanded cosmological models and combinations including matter-power measurements. The useful lesson is to compare complete analyses rather than select the smallest limit from each.

This does not establish that dark energy evolves. The required model must also describe other observations and survive systematic checks. Our dark-energy guide explains how measurements of cosmic expansion test that possibility.

Does the tension rule out inverted ordering?

A sufficiently tight cosmological bound in ΛCDM can strongly disfavour the larger minimum sum of inverted ordering. It is a conditional cosmological preference. It should not be presented as an assumption-free laboratory exclusion of that ordering.

The first JUNO oscillation measurement published in June 2026 determined solar-sector parameters using 59.1 days of data. That result improves the oscillation picture, but it was not a determination of the mass ordering or a measurement of an individual absolute mass.

An ordering determination from oscillations provides a valuable comparison because it tests different physics from a cosmological fit. The experimental strategy is described in our JUNO mass-ordering explainer. Establishing an ordering would fix which approximate minimum applies; the lightest mass would still require further information.

Where does neutrinoless double-beta decay fit?

Neutrinoless double-beta decay tests lepton-number violation and the Majorana character of neutrinos. Interpreting a half-life as mββ assumes a mechanism, usually exchange of light Majorana neutrinos, and nuclear matrix elements. Unknown phases can cancel contributions to that effective mass.

The complete KamLAND-Zen search reports a half-life lower limit of 3.8 × 1026 years at 90% confidence. For the light-neutrino mechanism, its corresponding mββ upper limits span 28–122 meV depending on nuclear calculations. That range is not an uncertainty interval for the total mass.

A positive signal would be important, but identifying the dominant decay mechanism would still matter. A null result does not establish that neutrinos are Dirac particles. This is an additional question alongside the cosmological tension, rather than a third direct measurement of the same mass.

What would make the comparison more decisive?

The next steps include a completed KATRIN mass analysis, independent ordering measurements and cosmological constraints tested against alternative likelihoods and model extensions. A sensitivity goal becomes a result only after the relevant data and systematic analysis are published.

In their 2026 study, Graham, Green and Meyers explore how changing expansion or CMB lensing could address the cosmological preference and what additional signatures could distinguish explanations. Such proposals are testable possibilities, not established resolutions.

For any future headline, ask which quantity was constrained, which data were combined, whether the result is a detection or an upper limit, and which assumptions produced it. A mass sum inferred in ΛCDM, a beta-decay effective mass and an oscillation minimum answer connected but different questions.

The evidence establishes nonzero neutrino masses and increasingly restrictive, complementary bounds. Whether a particular cosmological fit exposes new physics or a problem in modelling or data treatment remains open. Keeping those qualifications beside the numbers is essential to understanding what the tension actually tests.

References

  1. Particle Data Group, Neutrino Masses, Mixing, and Oscillations, Review of Particle Physics (2026)
  2. KATRIN Collaboration, Direct neutrino-mass measurement based on 259 days of KATRIN data, Science 388, 180–185 (2025) doi:10.1126/science.adq9592; peer-reviewed measurement
  3. G. Drexlin and C. Weinheimer, KATRIN experiment, Encyclopedia of Particle Physics, vol. 4, 438–461 (2026); arXiv v3 revised 8 October 2026 doi:10.1016/B978-0-443-26598-3.00122-X; experiment review
  4. W. Elbers et al. (DESI Collaboration), Constraints on neutrino physics from DESI DR2 BAO and DR1 full shape, Physical Review D 112, 083513 (2025) doi:10.1103/w9pk-xsk7; peer-reviewed analysis
  5. DESI Collaboration, DESI DR2 Results II: Measurements of Baryon Acoustic Oscillations and Cosmological Constraints, Physical Review D 112, 083515 (2025) doi:10.1103/tr6y-kpc6; peer-reviewed analysis
  6. D. Naredo-Tuero et al., Living at the Edge: A Critical Look at the Cosmological Neutrino Mass Bound, research manuscript (2024)
  7. D. Green and J. Meyers, Cosmological preference for a negative neutrino mass, Physical Review D 111, 083507 (2025)
  8. W. Elbers et al., Negative neutrino masses as a mirage of dark energy, Physical Review D 111, 063534 (2025) doi:10.1103/PhysRevD.111.063534; peer-reviewed model analysis
  9. D. Chebat et al., Cosmological neutrino mass: a frequentist overview in light of DESI, research manuscript (2025)
  10. JUNO Collaboration, Measurement of reactor neutrino oscillation with the first JUNO data, Nature 654, 343–348 (2026) doi:10.1038/s41586-026-10538-z; peer-reviewed measurement
  11. KamLAND-Zen Collaboration, Search for Majorana Neutrinos with the Complete KamLAND-Zen Dataset, Physical Review Letters 135, 262501 (2025); arXiv version revised March 2026 doi:10.1103/jkf6-48j8; peer-reviewed search
  12. P. W. Graham, D. Green and J. Meyers, New interpretations of the cosmological preference for a negative neutrino mass, Physical Review D 113, 043514 (2026) doi:10.1103/1bqb-qlrj; peer-reviewed interpretation

What is the neutrino mass tension?

It is the mismatch between some model-dependent cosmological fits and the minimum total mass implied by neutrino oscillations. Its significance depends on the datasets, model, priors and statistical method.

Does KATRIN disagree with DESI?

KATRIN’s published 0.45 eV limit is a 90% upper limit on the beta-decay effective mass. DESI analyses constrain a different quantity, the total mass. The KATRIN limit does not itself establish a contradiction.

What is the minimum total neutrino mass?

For three standard neutrinos, measured mass-squared splittings imply approximate minimum sums of 0.059 eV for normal ordering and 0.10 eV for inverted ordering. The precise values depend on the fitted splittings.

Do neutrinos have negative mass?

A negative effective parameter in a cosmological fit is a diagnostic extension of the model. It does not establish negative physical rest masses for neutrinos.

Has cosmology ruled out inverted ordering?

Some cosmological datasets and models strongly disfavour its larger minimum mass sum. That is a conditional inference, not an assumption-free laboratory exclusion of inverted ordering.

Leave a response

Your email address is not published.

Newsletter

Get new Quantum Nature articles by email.

New research write-ups and explainers, sent only when there is something worth reading — no fixed schedule, no filler.

Browse recent articles

Free. Unsubscribe in one click.