global-nu
Global analyses of neutrino oscillation data
The Bari group's global analysis, published openly: best fits, allowed ranges and Δχ² profiles for both mass orderings, with the figures and the data behind them.
Current results
Bari Release 2026.a Preliminary
Preliminary Bari Release 2026.a: the paper is in preparation and these values may still change. Data files will be published with the paper.
Best fit, normal ordering (NO); the inverted-ordering (IO) value below each.
- δm² / 10⁻⁵ eV²
- 7.465±0.076
- IO 7.439
- sin²θ₁₂
- 0.3047±0.0055
- IO 0.3036
- Δm² / 10⁻³ eV²
- 2.4720±0.0134
- IO −2.4810
- sin²θ₁₃
- 0.02218±0.00055
- IO 0.0223
- sin²θ₂₃
- 0.491+0.047−0.021
- IO 0.5446
- δ/π
- 1.19±0.15
- IO 1.52
NO IO · Nσ = √Δχ² up to 5σ, both orderings on the common scale (guides at 1σ and 3σ); shaded: the 1σ range (NO). Profiles of the final combination.
Normal ordering favoured at 3.9σ · Δχ²(IO−NO) = +15.34
% Bari Release 2026.a: paper in preparation — replace with its record once out. @misc{globalnu:2026.a, title = "{Bari global analysis of neutrino oscillation data, release 2026.a}", howpublished = "\url{https://global-nu.org/results.html}", note = "Preliminary; values as of 2026-10-15", year = "2026" }What changed
against the published values (2025 release with the 2026 (1, 2)-sector update)
- δm² 1σ accuracy 1.3% → 1.0%*
- sin²θ₁₂ 1σ accuracy 2.4% → 1.8%*
- Δm² 1σ accuracy 0.8% → 0.5%*
- sin²θ₁₃ 1σ accuracy 2.4% → 2.5%*
- sin²θ₂₃ 1σ accuracy 5.1% → 4.4%*
- δ/π 1σ accuracy 18% → 12.3%*
- Ordering NO at 2.2σ → NO at 3.9σ (Δχ² +5.0 → +15.34)
1σ accuracy, in percent. Bari Release 2026.a: the 1σ error (half the 1σ range) divided by the best fit — its errors are Gaussian to a very good approximation; δ/π too, whose profile is far from Gaussian but not because of the octant degeneracy — except sin²θ₂₃: one sixth of the 3σ range divided by the best fit, since its range spans both octants. Earlier releases: as published, one sixth of the 3σ range divided by the best fit, in percent, as in the tables of our papers.
* computed from the release's unrounded ranges, with its definition stated above.
The six parameters, as measured
current values · normal ordering
Best fit with 1σ and 3σ ranges for the current values in the panel above, normal ordering. All six rows share one scale, in percent of each parameter's own best fit, so the width of a row is how well that parameter is known; a row that runs past the axis is marked with an arrow and its true extent. Absolute values are printed at the right. Full tables and both orderings are on the results page.
How the precision improved
1σ accuracy · logarithmic scale
Each line is one parameter’s 1σ accuracy in each release — a sixth of its 3σ range over its best fit, as in the tables of our papers; for Bari Release 2026.a its 1σ error over its best fit (sin²θ₂₃: a sixth of the 3σ range), see the note under the results. Computed from the same tables the parameter history is built on.
Every day
generated automatically — not results of the group
Today on arXiv
The newest experimental and theoretical preprints, kept in two streams and ranked by a keyword score whose word list is printed at the foot of that page — no model decides what appears there.
The full digest → NewsAI-assistedWhat is happening
Experiments, results and milestones, written from fetched sources: every item on that page carries the link it came from, and an item whose citation cannot be resolved is dropped before publication.
More news → ConferencesautomaticWhere the field meets
Upcoming and recent meetings with dates and links — and the venue wherever the source publishes one — rebuilt each day from the conference indexers rather than kept by hand.
The calendar →