Results

Results, release by release

The newest published values first — δm² and sin²θ₁₂ after the first JUNO results — then the most recent full release in detail: best-fit values and allowed ranges for the six oscillation parameters, in both mass orderings, with the paper and its figures beside them. Every earlier release is traced parameter by parameter on the parameter history.

Latest — the six parameters after the first JUNO results

δm² and sin²θ₁₂ revised in 2026, on the March 2025 full release

The first JUNO results and the latest SNO+ data, added to the 2025 global analysis, revise the (1, 2) sector alone. The table below therefore carries the newest published value for each of the six parameters — the two marked 2026 from the update, the other four as they stand in the full release below.

The update is by F. Capozzi, E. Lisi, F. Marcone, A. Marrone, A. Palazzo; the 2025 release the other four rows come from adds W. Giarè and A. Melchiorri.

newest value per parameter two papers, not one fit JUNO + SNO+ 2025 δm² known to 1.3%
The newest published value for each parameter — not a single fit. δm² and sin²θ₁₂, marked 2026, are the final row (“w/ SNO+ & JUNO 2025”) of Table I of Phys. Rev. D 114, 016026 (2026), which adds the first JUNO results and the latest SNO+ data to the (1, 2) sector. The other four are Table I of Phys. Rev. D 111, 093006 (2025), the most recent full release, reproduced whole below — the update quotes no new values for them. Ranges at N σ = 1, 2, 3; the last column is the formal “1σ parameter accuracy”, defined the same way in both papers, 1/6 of the 3σ range divided by the best-fit value, in percent. δ/π is cyclic (mod 2). Values are transcribed from the published tables; nothing here is estimated or rounded further. Δm² is signed, so the inverted-ordering rows read negative: the papers print the magnitude, |Δm²|, and the sign here is the convention stated below, not a different measurement.
ParameterOrderingBest fit1σ range2σ range3σ range“1σ” (%)From
δm² / 10⁻⁵ eV²NO, IO7.487.39 – 7.587.30 – 7.687.21 – 7.781.32026
sin²θ₁₂ / 10⁻¹NO, IO3.0853.010 – 3.1562.939 – 3.2302.866 – 3.3032.42026
Δm² / 10⁻³ eV²NO2.4952.475 – 2.5152.454 – 2.5362.433 – 2.5580.82025
IO−2.465−2.485 – −2.444−2.506 – −2.423−2.527 – −2.4030.82025
sin²θ₁₃ / 10⁻²NO2.232.17 – 2.272.11 – 2.332.06 – 2.382.42025
IO2.232.19 – 2.302.14 – 2.352.08 – 2.412.42025
sin²θ₂₃ / 10⁻¹NO4.734.60 – 4.964.47 – 5.684.37 – 5.815.12025
IO5.455.28 – 5.604.58 – 5.734.43 – 5.834.32025
δ/πNO1.201.07 – 1.370.88 – 1.810.73 – 2.03182025
IO1.481.36 – 1.611.24 – 1.721.12 – 1.8382025

What the update changed: against the 2025 analysis — δm² = 7.37 known to 2.3%, sin²θ₁₂ = 3.03 to 4.5% — the two additions move both best fits up and roughly halve both uncertainties. The paper's own intermediate row, SNO+ added but not JUNO, gives δm² = 7.44 at 2.1%: most of the tightening is JUNO's. δm² is now the second oscillation parameter, after Δm², known to better than 2%. The ordering preference is untouched — normal ordering is favoured at 2.2σ, as in the release below.

Conventions used throughout this site

The two squared mass gaps are δm² = m₂² − m₁² > 0 and Δm² = m₃² − ½(m₁² + m₂²), with α = sign(Δm²) distinguishing normal ordering (NO, α = +1) from inverted ordering (IO, α = −1). Equivalently, our Δm² is the half-sum of the two larger splittings, Δm² = ½(Δm²₃₁ + Δm²₃₂), which is the form the conversion is easiest to read from: since Δm²₃₁ − Δm²₃₂ = δm², passing to either of them is a shift of ±δm²/2.

Other groups quote Δm²₃₁ or Δm²₂₃ instead: those numbers are not comparable with ours at a glance. Wherever this site compares analyses, it states the conversion it applied.

And the errors move differently from the values. Because the offset carries its own uncertainty, the converted interval is wider than the published one: the central value moves by more than one standard deviation while the uncertainty grows by a fraction of a percent. Both numbers are computed from the register and quoted on the parameter-history page, which also records that the propagation still omits the δm²–Δm² correlation, and gives the formula — the published papers do not carry the joint information a rigorous reprojection would need. The exported register carries every interval in both conventions, and an interval_method column saying which treatment produced it.

March 2025 — Entering the era of subpercent precision

Updated global analysis of the known and unknown parameters of the standard three-neutrino framework, using data available at the beginning of 2025: solar, KamLAND, atmospheric, short-baseline reactor and long-baseline accelerator data.

F. Capozzi, W. Giarè, E. Lisi, A. Marrone, A. Melchiorri, A. Palazzo. Received 12 March 2025; accepted 21 April 2025; published 19 May 2025.

both orderings 1σ · 2σ · 3σ ranges NO favoured at 2.2σ Δχ²(IO−NO) = +5.0
Table I of Phys. Rev. D 111, 093006 (2025): best-fit values and allowed ranges at N σ = 1, 2, 3, for either NO or IO. The last column is the formal “1σ parameter accuracy”, defined as 1/6 of the 3σ range divided by the best-fit value, in percent. δ/π is cyclic (mod 2). Values are transcribed from the published table; nothing on this page is estimated or rounded further. Δm² is signed, so the inverted-ordering rows read negative: the papers print the magnitude, |Δm²|, and the sign here is the convention stated below, not a different measurement. † δm² and sin²θ₁₂ have since been superseded by the (1, 2)-sector update at the top of this page.
ParameterOrderingBest fit1σ range2σ range3σ range“1σ” (%)
δm² / 10⁻⁵ eV² †NO, IO7.377.21 – 7.527.06 – 7.716.93 – 7.932.3
sin²θ₁₂ / 10⁻¹ †NO, IO3.032.91 – 3.172.77 – 3.312.64 – 3.454.5
Δm² / 10⁻³ eV²NO2.4952.475 – 2.5152.454 – 2.5362.433 – 2.5580.8
IO−2.465−2.485 – −2.444−2.506 – −2.423−2.527 – −2.4030.8
sin²θ₁₃ / 10⁻²NO2.232.17 – 2.272.11 – 2.332.06 – 2.382.4
IO2.232.19 – 2.302.14 – 2.352.08 – 2.412.4
sin²θ₂₃ / 10⁻¹NO4.734.60 – 4.964.47 – 5.684.37 – 5.815.1
IO5.455.28 – 5.604.58 – 5.734.43 – 5.834.3
δ/πNO1.201.07 – 1.370.88 – 1.810.73 – 2.0318
IO1.481.36 – 1.611.24 – 1.721.12 – 1.838

Δχ² projections, all data included

Fig. 3 of the paper · NO in blue, IO in red

Six panels
showing Nσ as a function of δm², |Δm²|, δ/π, sin²θ₁₂, sin²θ₁₃ and sin²θ₂₃, for
normal and inverted ordering, with all oscillation data included.

Figure 3 from F. Capozzi, W. Giarè, E. Lisi, A. Marrone, A. Melchiorri and A. Palazzo, Neutrino masses and mixing: Entering the era of subpercent precision, Phys. Rev. D 111, 093006 (2025), doi:10.1103/PhysRevD.111.093006. Published by the American Physical Society under the terms of the Creative Commons Attribution 4.0 International license; reproduced here under those terms.

Where θ₂₃ and θ₁₃ meet

Fig. 4 of the paper · increasingly rich datasets

Regions allowed in the
sin²θ₂₃–sin²θ₁₃ plane for increasingly rich datasets, in normal ordering (top)
and inverted ordering (bottom).

Figure 4 from the same paper, same citation and licence as above.

The same table, seen at a glance

best fit · 1σ · 3σ · one shared scale, in % of each best fit

-20% -10% best fit +10% +20% δm² / 10⁻⁵ eV² 3σ: 6.93 – 7.93 (−6% / +8%) 1σ: 7.21 – 7.52 (−2% / +2%) best fit 7.37 7.37 |Δm²| / 10⁻³ eV² (NO) 3σ: 2.433 – 2.558 (−2% / +3%) 1σ: 2.475 – 2.515 (−1% / +1%) best fit 2.495 2.495 |Δm²| / 10⁻³ eV² (IO) 3σ: 2.403 – 2.527 (−3% / +3%) 1σ: 2.444 – 2.485 (−1% / +1%) best fit 2.465 2.465 sin²θ₁₂ / 10⁻¹ 3σ: 2.64 – 3.45 (−13% / +14%) 1σ: 2.91 – 3.17 (−4% / +5%) best fit 3.03 3.03 sin²θ₁₃ / 10⁻² (NO) 3σ: 2.06 – 2.38 (−8% / +7%) 1σ: 2.17 – 2.27 (−3% / +2%) best fit 2.23 2.23 sin²θ₁₃ / 10⁻² (IO) 3σ: 2.08 – 2.41 (−7% / +8%) 1σ: 2.19 – 2.3 (−2% / +3%) best fit 2.23 2.23 sin²θ₂₃ / 10⁻¹ (NO) 3σ: 4.37 – 5.81 (−8% / +23%) 1σ: 4.6 – 4.96 (−3% / +5%) best fit 4.73 4.73 sin²θ₂₃ / 10⁻¹ (IO) 3σ: 4.43 – 5.83 (−19% / +7%) 1σ: 5.28 – 5.6 (−3% / +3%) best fit 5.45 5.45 δ/π (NO) 3σ: 0.73 – 2.03 (−39% / +69%) 1σ: 1.07 – 1.37 (−11% / +14%) best fit 1.2 1.2 −39% +69% δ/π (IO) 3σ: 1.12 – 1.83 (−24% / +24%) 1σ: 1.36 – 1.61 (−8% / +9%) best fit 1.48 1.48 width = 3σ range as a percentage of the best fit

Every row is centred on its own best fit and measured outward in percent of it, on one shared axis, so the width of a row is how well that parameter is known: |Δm²| is measured to ±2.5% at 3σ, while δ in normal ordering runs off the axis at −39%/+69% and is marked where it leaves. Absolute values are printed at the right. Values are those of Table I above; the figure is generated from them by tools/make_figures.py, not drawn by hand.

What changed

Δm² is now constrained at the 0.8% level — the first oscillation parameter to enter the subpercent precision era — against 1.1% in the previous update; the paper underlines in the same breath that there are issues about systematics which might affect that error estimate. The uncertainty on sin²θ₁₃ falls to 2.4%, from about 3%. For sin²θ₂₃ the two quasi-degenerate minima are closer than before, differing by roughly 15% against about 25% previously. Constraints on δ are similar to before within large uncertainties, with a weaker rejection of CP conservation in NO. The overall offset between the orderings is now Nσ = √5.0 = 2.2, down from 2.5σ.

Not included

The first published SNO+ reactor results are not included: they constrain δm² with an error larger than the one above by a factor of about 6, and only under a prior on θ₁₂. SNO+ is nevertheless expected to surpass the present δm² accuracy on (δm², θ₁₂) within a few years.

Since this release: the (1, 2) sector, updated

δm² and sin²θ₁₂ were revised by the partial update at the top of this page, which adds the first JUNO results and the latest SNO+ data to this analysis. It quotes no new values for Δm², sin²θ₁₃, sin²θ₂₃ or δ, which stand as in the table above. Both releases are on the parameter history, with the table each number was transcribed from.

Machine-readable data

planned for this release

Δχ² profiles and the tables above will be published here as documented files at stable URLs — the kind of thing you can fetch from a script and cite — under data/<release>/. They are exported from the analysis by a dedicated step; the analysis code itself stays where it belongs and is never part of this site.

Nothing is linked from this section yet: the export runs when the release material is prepared, and a link that does not resolve is worse than no link. What does exist today is the parameter register — every value on this site, with its source — published as JSON and CSV on the parameter history page.