Neutrino Atlas
EN
  1. Learn
  2. The mass ordering

Chapter 3: Oscillation and mass · Lesson 3.4

The mass ordering

Open

Oscillation experiments have measured the gaps between the three neutrino masses with per-cent precision, yet they cannot tell which way the large gap points. Matter effects in the Earth, a fine ripple in reactor spectra and the cosmic mass budget are now closing in on the answer.

Beginner

Three masses, one open question

Neutrinos come in three flavours, but the states that carry a definite mass are three different ones, simply numbered 1, 2 and 3. Neutrino oscillation has shown that these three masses are not equal, and it has measured how far apart they are with remarkable precision. Yet one basic fact is still missing: we do not know which of the three is the heaviest. This question is called the mass ordering, in older texts also the mass hierarchy.

Two possible floor plans

Picture a small house with three flats. Two of them lie very close together, almost on the same floor; the third lies far higher or far lower. Oscillation experiments have measured both gaps, the small one between flats 1 and 2 and the large one to flat 3. What they have not found out is whether flat 3 is the penthouse or the basement.

  • In the normal ordering the close pair is at the bottom and the third neutrino sits far above it: two light neutrinos and one heavier one. This resembles the charged particles, where each family is heavier than the one before.
  • In the inverted ordering the close pair is at the top and the third neutrino lies far below it: two heavier neutrinos and one very light one.

What we already know

Part of the order is settled. Neutrinos from the Sun pass through its dense interior, and the way matter changes their behaviour there, the MSW effect, works only if neutrino 2 is slightly heavier than neutrino 1. So the order within the close pair is known. What is missing is the position of the third neutrino relative to that pair. Ordinary oscillation in empty space depends only on the size of the large gap, not on its direction, so on its own it cannot answer the question.

Three ways to find out

Physicists are approaching the question from three directions.

  • Matter on the way: neutrinos that travel hundreds or thousands of kilometres through the Earth are nudged by its electrons. The nudge works differently for the two orderings, and differently for neutrinos and antineutrinos. Beam experiments such as NOvA and the future DUNE use this, as do detectors for neutrinos made in the atmosphere.
  • A fine ripple: the Chinese detector JUNO, which began measuring in 2025, records antineutrinos from nuclear power plants about 53 kilometres away. The ordering leaves a tiny shift in a fast ripple on their energy spectrum.
  • The universe: the Big Bang left behind enormous numbers of neutrinos, and their total mass slightly slows the growth of galaxies. In the inverted ordering the three masses together cannot be below about 0.10 electronvolts; in the normal ordering they can be as small as about 0.06.

Why it matters

The answer has practical consequences. A very rare nuclear decay, neutrinoless double beta decay, would come within reach of the next generation of experiments if the ordering is inverted and neutrinos are their own antiparticles. In the normal ordering the decay could be far rarer. The ordering also fixes how small the total neutrino mass can be, a number cosmologists need, and it is a clue for theorists who try to explain why neutrino masses are so tiny, a question treated in the deep dive on where neutrino masses come from.

Where things stand

Current data lean towards the normal ordering, but not strongly enough to call the question settled. Several large experiments expect a clear answer within roughly the next ten years.

Quiz

Test your knowledge

4 questions on this level · Print as worksheet

  1. 1What does the mass ordering question ask?
  2. 2Which part of the order is already known from neutrinos passing through the Sun?
  3. 3What does the inverted ordering look like?
  4. 4How does JUNO try to find the ordering?

Intermediate

Defining the two orderings

The three mass eigenstates ν₁, ν₂ and ν₃ are labelled by their electron-flavour content: ν₁ contains the largest share of νe (about two thirds), ν₂ about a third and ν₃ only about 2 %. Oscillations measure two independent mass-squared differences: the solar splitting Δm²21 ≈ 7.5 × 10⁻⁵ eV² and the atmospheric splitting |Δm²31| ≈ 2.5 × 10⁻³ eV², about thirty times larger1. In the normal ordering (NO) m₁ < m₂ < m₃; in the inverted ordering (IO) m₃ < m₁ < m₂.

Why one sign is known and the other is not

In vacuum, oscillation probabilities depend on sin²(Δm² L/4E), which does not change when Δm² changes sign. A sign becomes visible only when a second effect interferes. For the solar pair that effect is matter. Lincoln Wolfenstein showed in 1978 that electron neutrinos feel an extra potential in matter6, and Stanislav Mikheyev and Alexei Smirnov found in 1985 that it can drive a resonant conversion, the MSW effect. The observed suppression of high-energy ⁸B solar neutrinos to about a third requires this resonance, which exists only if m₂ > m₁. For Δm²31 the matter route opened only after Daya Bay and RENO measured a sizeable θ₁₃ in 2012, because the relevant Earth matter effect scales with sin²2θ₁₃.

Method 1: matter effects in the Earth

In νμ → νe appearance over long baselines, Earth matter effects enhance the probability for neutrinos and suppress it for antineutrinos in the normal ordering, and the reverse in the inverted ordering. The effect grows with energy: about ±10 % for T2K (295 km, 0.6 GeV), roughly ±30 % for NOvA (810 km, 2 GeV) and about ±40 % for DUNE (1,300 km from Fermilab to the Sanford Underground Research Facility)3. The complication is that δCP also shifts neutrino and antineutrino rates in opposite directions, so the two can imitate each other (parameter degeneracies). The 2025 joint T2K–NOvA analysis found no strong preference for either ordering.

Atmospheric neutrinos cross the Earth's mantle and core over baselines up to 12,700 km. For the normal ordering a resonance appears for neutrinos at a few GeV, for the inverted ordering for antineutrinos. Super-Kamiokande, IceCube DeepCore, KM3NeT/ORCA and later Hyper-Kamiokande search for it.

Method 2: reactor interference at JUNO

JUNO uses a different handle. Its 20,000-tonne liquid-scintillator detector lies about 53 km from the Yangjiang and Taishan reactors, near the maximum of the slow oscillation driven by Δm²21. On top of that broad dip sits a fast ripple from the two nearly equal frequencies Δm²31 and Δm²32. Which of them is larger depends on the ordering, so the ripple's phase shifts slightly. Seeing this needs an energy resolution of about 3 % at 1 MeV and an energy scale known to well below 1 %. JUNO expects a 3σ median sensitivity after about 6.5 years at full reactor power2, independent of δCP and of matter effects.

Method 3: cosmology

The splittings set a floor on the sum of neutrino masses: Σmν ≥ ≈0.059 eV in NO and ≈0.10 eV in IO. DESI baryon acoustic oscillations combined with the cosmic microwave background give Σmν < 0.0642 eV (95 %)4, below the inverted floor. This disfavours IO only if the standard cosmological model ΛCDM holds; extended models weaken the bound (see “How heavy is a neutrino?”).

What follows from the answer

In IO the effective Majorana mass probed by neutrinoless double beta decay cannot fall below about 0.02 eV; in NO it can be far smaller. KamLAND-Zen now excludes mββ above 28–122 meV5, and tonne-scale experiments aim to cover the whole inverted range.

Quiz

Test your knowledge

4 questions on this level · Print as worksheet

  1. 1Roughly how much larger is |Δm²31| than Δm²21?
  2. 2Why did the Earth-matter route to the ordering open only after 2012?
  3. 3What complicates the ordering measurement at T2K and NOvA?
  4. 4Why does the DESI bound disfavour the inverted ordering only conditionally?

Expert

Conventions and the mass spectrum

With Δm²_ij = mi² − mj² and the mass eigenstates labelled by decreasing |Uei|², solar data fix Δm²21 > 0. The ordering is the sign of Δm²31. NuFIT uses Δm²3ℓ, meaning Δm²31 > 0 for NO and Δm²32 < 0 for IO, and finds Δm²21 = (7.49 ± 0.19) × 10⁻⁵ eV² and Δm²3ℓ = +2.534 (+0.025, −0.023) × 10⁻³ eV² (NO, without Super-Kamiokande atmospheric data)1. For a lightest mass m₀:

NO: m₁ = m₀, m₂ = √(m₀² + Δm²21), m₃ = √(m₀² + Δm²31) IO: m₃ = m₀, m₂ = √(m₀² + |Δm²32|), m₁ = √(m₀² + |Δm²32| − Δm²21)Spectra in the two orderings. For m₀ → 0, Σmν ≈ 0.059 eV (NO) and ≈ 0.10 eV (IO); for m₀ ≫ 0.05 eV both become quasi-degenerate and the ordering ceases to matter for absolute-mass observables.

Matter effects

Charged-current forward scattering on electrons adds the matter potential V = √2 GF Ne to νe; the sign flips for antineutrinos. In the effective two-flavour description of the (1,3) sector:

A = 2√2 GF Ne E ≈ 1.52 × 10⁻⁴ eV² × Ye (ρ / g cm⁻³)(E / GeV) sin²2θ₁₃,m = sin²2θ₁₃ / [(cos 2θ₁₃ − A/Δm²31)² + sin²2θ₁₃]Matter term A and effective mixing angle in matter. The MSW resonance A = Δm²31 cos 2θ₁₃ occurs for neutrinos if Δm²31 > 0 and for antineutrinos (A → −A) if Δm²31 < 0.

For the mantle (ρ ≈ 4.5 g cm⁻³, Ye ≈ 0.5) the resonance lies near 6–7 GeV, for core-crossing trajectories at lower energy, where parametric enhancement adds further structure. Atmospheric detectors must separate ν from ν̄ statistically, through different fluxes and cross sections, unless they have charge identification as planned for INO-ICAL. In accelerator appearance the leading term is

P(νμ → νe) ≈ sin²θ₂₃ sin²2θ₁₃ sin²[(1 − x)Δ] / (1 − x)² + (δCP-dependent terms), x = A/Δm²31, Δ = Δm²31 L / 4ELeading matter-modified appearance probability; x changes sign with the ordering and between ν and ν̄.

At DUNE, x ≈ ±0.2 at the first maximum near 2.5 GeV, which changes the appearance rate by roughly ±40 %. With a wide-band beam sampling the first and second maxima, DUNE expects to determine the ordering at more than 5σ for all values of δCP3. Because the two orderings are discrete, non-nested hypotheses, Wilks' theorem does not apply, and significances are obtained from pseudo-experiments rather than read off as √Δχ².

Reactor interference

Pee = 1 − cos⁴θ₁₃ sin²2θ₁₂ sin²Δ₂₁ − sin²2θ₁₃ (cos²θ₁₂ sin²Δ₃₁ + sin²θ₁₂ sin²Δ₃₂), Δij = Δm²ij L / 4EElectron antineutrino survival in vacuum; matter corrections at 53 km are small and included in fits.

Since cos²θ₁₂ ≈ 0.69 outweighs sin²θ₁₂ ≈ 0.31, the fast oscillation is dominated by Δ₃₁. In NO |Δm²31| = |Δm²32| + Δm²21, so the dominant frequency is the higher one; in IO it is the lower one. The two frequencies differ by only Δm²21/|Δm²31| ≈ 3 %, which is why JUNO needs about 3 %/√(E/MeV) resolution and sub-per-cent control of the energy non-linearity; the near detector JUNO-TAO measures the fine structure of the unoscillated reactor spectrum. The method, proposed by Serguey Petcov and M. Piai in 2002, is independent of δCP and θ₂₃. JUNO's first 59 days already gave sin²θ₁₂ = 0.3092 ± 0.0087 and Δm²21 = (7.50 ± 0.12) × 10⁻⁵ eV². A further gain comes from combination: reactor ν̄e disappearance and accelerator νμ disappearance measure different effective splittings, Δm²ee and Δm²μμ, which agree only under the correct ordering.

Global status and consequences

In NuFIT 6.0 the inverted ordering is disfavoured by Δχ² = 6.1 when the Super-Kamiokande atmospheric χ² table is included, whereas without it IO is marginally preferred by Δχ² = 0.61. Cosmology adds DESI's Σmν < 0.0642 eV (95 %, ΛCDM)4, below the IO floor. For neutrinoless double beta decay, mββ = |cos²θ₁₂ cos²θ₁₃ m₁ + sin²θ₁₂ cos²θ₁₃ m₂ exp(iα) + sin²θ₁₃ m₃ exp(iβ)| lies between about 0.019 and 0.050 eV in IO for m₀ → 0, but only at 1.5–4 meV in NO and can cancel for m₀ of a few meV. KamLAND-Zen excludes mββ > 28–122 meV5; LEGEND-1000 and nEXO target the full IO band. Similarly the kinematic floor is mβ ≳ 0.048 eV (IO) versus 0.009 eV (NO). In a galactic supernova, the νe neutronisation burst survives with a probability near |Ue3|² ≈ 2 % in NO but near |Ue2|² ≈ 30 % in IO, giving a further independent test.

Quiz

Test your knowledge

4 questions on this level · Print as worksheet

  1. 1In NuFIT's Δm²3ℓ convention, which sign describes the inverted ordering?
  2. 2Near which energy does the (1,3) MSW resonance occur for mantle-crossing neutrinos?
  3. 3Why can JUNO's reactor spectrum distinguish the orderings?
  4. 4For m₀ → 0 in the inverted ordering, where does mββ lie?

Open questions

The sign of Δm²_31 itself. Global fits favour the normal ordering at roughly the 2–3σ level, and that preference depends on how atmospheric data are included. JUNO, the atmospheric programmes of IceCube, KM3NeT/ORCA and Hyper-Kamiokande, and DUNE are expected to reach a combined determination at 5σ within about a decade.

Whether the cosmological pressure on the inverted ordering is real: the DESI bound sits even below the normal-ordering floor at some confidence levels, which could reflect systematics, dynamical dark energy or new neutrino physics rather than the ordering. If the normal ordering is confirmed, testing whether neutrinos are Majorana particles will require double beta decay experiments sensitive well below 0.01 eV.

Terms to explore

People

On the timeline

  1. 1978Wolfenstein: matter changes neutrino oscillations
  2. 1985Mikheyev and Smirnov discover resonant conversion
  3. 2012Daya Bay discovers a non-zero θ₁₃
  4. 2025JUNO begins data taking
  5. 2025JUNO's first results set precision records
  6. 2025NOvA and T2K publish a joint analysis

References

  1. I. Esteban, M. C. Gonzalez-Garcia, M. Maltoni, I. Martinez-Soler, J. P. Pinheiro, T. Schwetz, NuFit-6.0: updated global analysis of three-flavor neutrino oscillations, JHEP 12, 216 (2024), arXiv:2410.05380
  2. JUNO Collaboration, Potential to identify neutrino mass ordering with reactor antineutrinos at JUNO, Chin. Phys. C (2025), doi:10.1088/1674-1137/ad7f3e (arXiv:2405.18008)
  3. B. Abi et al. (DUNE Collaboration), Deep Underground Neutrino Experiment (DUNE) Far Detector Technical Design Report, Volume II: DUNE Physics (2020), arXiv:2002.03005
  4. W. Elbers et al. (DESI Collaboration), Constraints on neutrino physics from DESI DR2 BAO and DR1 full shape (2025), arXiv:2503.14744
  5. S. Abe et al. (KamLAND-Zen Collaboration), Search for Majorana neutrinos with the complete KamLAND-Zen dataset (2024), arXiv:2406.11438
  6. L. Wolfenstein, Neutrino oscillations in matter, Phys. Rev. D 17, 2369 (1978), doi:10.1103/PhysRevD.17.2369

Last updated: · Editorial standard

Cite this page

Text

Neutrino Atlas (2026). “The mass ordering”. Retrieved 22 November 2026, from https://neutrino-atlas.com/learn/the-mass-ordering/

BibTeX

@misc{neutrinoatlas2026learnthemassordering,
  title        = {The mass ordering},
  author       = {{Neutrino Atlas}},
  year         = {2026},
  url          = {https://neutrino-atlas.com/learn/the-mass-ordering/},
  note         = {Accessed 2026-11-22},
  language     = {english}
}