⚛️ phenomenology

Correlations in oscillation data and their impact on neutrino mixing models

This paper introduces a multidimensional $\tilde{\chi}^2$ framework to rigorously test theoretical neutrino mixing models against full parameter correlations in global-fit data, finding that Inverted Ordering schemes (particularly $\mu1$ -- $\tau1$ ) provide superior fits compared to Normal Ordering and that the $\text{TM}_2$ scheme is excluded at over $3\sigma$ confidence level.

Original authors: Janusz Gluza, Biswajit Karmakar, Szymon Zięba

Published 2026-10-21

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Janusz Gluza, Biswajit Karmakar, Szymon Zięba

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Neutrinos are the most abundant massive particles in the universe, yet they remain among the most elusive. These ghostly particles zip through everything, from the core of the sun to the human body, rarely interacting with the matter they pass through. For decades, scientists have known that neutrinos come in three distinct types, or "flavors": electron, muon, and tau. What makes them truly fascinating is their ability to change identity as they travel. A neutrino born as an electron type can transform into a muon or a tau type by the time it reaches a detector. This phenomenon, known as oscillation, proves that neutrinos have mass, a discovery that overturned the standard model of particle physics. To understand how this transformation happens, physicists measure specific angles that describe how the flavors mix and the differences in the squares of their masses. These measurements are incredibly precise, but they are also deeply interconnected; changing the value of one angle inevitably shifts the likely values of the others.

A team of researchers in Poland has developed a new way to test theoretical ideas about how these neutrinos mix. For years, scientists have proposed elegant mathematical patterns, often based on hidden symmetries in nature, to explain why the mixing angles take the values they do. These theories predict that the angles are not independent but are linked by strict rules, much like the sides of a triangle are linked by geometry. To test these ideas, researchers usually compare the theory's predictions against the allowed ranges for each angle individually. However, the new study shows that this traditional method is too loose. Because the experimental data is so rich, the angles are tied together in complex, multi-dimensional ways that a simple, side-by-side comparison misses. By ignoring these hidden connections, previous studies might have accepted theories that are actually wrong or rejected ones that are correct.

The researchers addressed this by building a sophisticated statistical tool that maps out the full, tangled relationship between all the mixing parameters at once. Instead of looking at the data as a collection of separate lines, they reconstructed a complete, six-dimensional landscape of possibilities based on the latest global experimental results. They then overlaid the predictions of several popular theoretical models onto this landscape to see if the models' strict rules could pass through the regions where the data is most likely to be found. This approach acts like a high-resolution filter, revealing whether a theory's specific recipe for mixing angles is compatible with the actual, correlated behavior of neutrinos observed in experiments.

When they applied this rigorous test to several leading theories, the results were decisive. One specific model, which predicted a particular relationship between the mixing angles, was found to be incompatible with the data. The researchers determined that this model is excluded with a high degree of statistical certainty, effectively ruling it out as a description of nature. Another model, which allowed for a slightly different relationship, fared much better, particularly in scenarios where the neutrinos follow an "inverted" mass ordering. In this scenario, the model's predictions aligned almost perfectly with the experimental landscape, suggesting it is a strong candidate for explaining the underlying physics. The study also examined how these new, tighter constraints affect predictions for the absolute mass of neutrinos, which is a key goal for future experiments searching for a rare type of radioactive decay.

The implications of this work extend beyond just rejecting one idea. The new method demonstrates that as experiments become more precise, the correlations between different measurements become the most powerful tool for discovery. By accounting for these connections, the allowed space for theoretical models shrinks dramatically, forcing theories to be much more precise to survive. The researchers found that for some models, the inclusion of these correlations narrowed the range of possible values so significantly that the models moved from being merely plausible to being either strongly favored or completely excluded. This shift is crucial for the future of neutrino physics, as upcoming experiments will provide even more precise data. Without a framework that respects these deep correlations, scientists might misinterpret the results of these future experiments. The study provides a ready-made toolkit for the scientific community to test any new theory against the full weight of current data, ensuring that the next breakthrough in understanding the neutrino is built on a foundation of statistical rigor.

Technical Summary: Correlations in Oscillation Data and their Impact on Neutrino Mixing Models

Problem Statement
Neutrino oscillation experiments have established the existence of non-zero neutrino masses and flavor mixing with high precision. However, theoretical models predicting specific mixing patterns (e.g., based on discrete flavor symmetries) often rely on algebraic relations between oscillation parameters (sum rules). Traditionally, the viability of these models has been assessed by comparing their predictions against independent 1D or 2D confidence intervals of oscillation parameters derived from global fits. The authors argue that this approach is flawed because it ignores the genuine multi-dimensional correlations encoded in experimental data. Treating parameters as independent can lead to overly optimistic conclusions about model viability, accepting parameter combinations that are actually disfavored when correlations are accounted for, or rejecting viable combinations. Furthermore, complete correlation information (up to 6D) is not publicly available in a format directly usable for testing arbitrary theoretical relations.

Methodology
To address the lack of publicly available full-dimensional correlation data, the authors develop a methodology to reconstruct a conservative estimate of the full six-dimensional (6D) $\chi^2$ function, denoted as $\tilde{\chi}^2$ , using only available 2D and 3D correlation datasets from the NuFIT global fit (versions 6.0 and 6.1).

Reconstruction of Correlations: The authors utilize NuFIT data, which provides 2D correlation contours for all pairs of parameters and specific 3D contours for each mass ordering. They employ a recursive approach to extend these lower-dimensional datasets to higher dimensions:
- Extension ( $\chi^2_{ext}$ ): To construct a $\chi^2$ for a set of $n$ variables, they take the maximum $\chi^2$ value over all reduced subsystems of size $n-1$ (Eq. 3). This ensures that the constructed $\tilde{\chi}^2$ provides a lower bound on the true $\chi^2$ , avoiding the unjustified exclusion of models.
- Reduction ( $\chi^2_{red}$ ): To project higher-dimensional $\chi^2$ functions down to lower dimensions, they minimize over the unwanted parameters (Eq. 6).
- By combining these operations, they can evaluate $\tilde{\chi}^2$ for any subset of the six oscillation parameters ( $\theta_{12}, \theta_{13}, \theta_{23}, \delta_{CP}, \Delta m^2_{21}, |\Delta m^2_{3\ell}|$ ) using only the provided 2D and 3D data.
Model Testing Framework: The authors apply this framework to specific mixing schemes that impose algebraic relations among parameters. For a given model, the parameters are constrained by the model's sum rules. The authors then evaluate the $\tilde{\chi}^2$ for points satisfying these relations. A model is considered viable if $\tilde{\chi}^2 \leq 9$ (corresponding to a $3\sigma$ level for 1 degree of freedom). The minimum value, $\tilde{\chi}^2_{min}$ , determines the best achievable agreement for the model.
Implementation: The authors provide a Mathematica notebook containing the interpolated correlation functions and the implementation of the extension/reduction formulas to facilitate community use.

Key Results
The framework was applied to several mixing schemes: Trimaximal mixing (TM1, TM2) and partial $\mu-\tau$ reflection symmetry ( $\mu_1-\tau_1$ , $\mu_2-\tau_2$ ).

Impact of Correlations: Incorporating full correlations significantly narrows the allowed parameter regions compared to analyses using uncorrelated 1D/2D intervals. For example, in the $\theta_{13}$ - $\theta_{12}$ plane, the allowed area for TM2 shrinks substantially when correlations are included.
Model Viability (NuFIT 6.0):
- TM2: Performs poorly, with $\tilde{\chi}^2_{min} = 8.1$ for both Normal Ordering (NO) and Inverted Ordering (IO).
- TM1: Performs better, with $\tilde{\chi}^2_{min} = 2.8$ (NO) and $0.9$ (IO).
- $\mu-\tau$ Symmetries: The $\mu_1-\tau_1$ scheme provides the best fit for IO ( $\tilde{\chi}^2_{min} = 0.1$ ).
Impact of JUNO Data (NuFIT 6.1): The inclusion of JUNO data (improving precision on $\theta_{12}$ $θ_{12}$ and $\Delta m^2_{21}$ $Δ m_{21}^{2}$ ) further tightens constraints.
- TM2 Exclusion: Under NuFIT 6.1, TM2 is excluded at $>3\sigma$ C.L. for both orderings ( $\tilde{\chi}^2_{min} \approx 21.4$ ).
- $\mu_1-\tau_1$ Preference: The $\mu_1-\tau_1$ scheme for IO remains the best fit, with $\tilde{\chi}^2_{min}$ decreasing slightly to $0.12$.
Effective Masses: The correlation-respecting constraints significantly reduce the allowed ranges for the effective neutrino mass in tritium $\beta$ -decay ( $m_\beta$ ) and neutrinoless double $\beta$ -decay ( $m_{\beta\beta}$ ). The reduction is most pronounced for NO and for TM1/TM2 schemes at low lightest neutrino masses ( $m_0$ ).
Parameter Shifts: The method reveals shifts in best-fit values and narrower $3\tilde{\sigma}$ ranges for parameters like $\delta_{CP}$ and $\theta_{12}$ , particularly for TM2 where the $\theta_{12}$ - $\theta_{13}$ relation is strongly constrained by the new data.

Significance and Claims
The paper claims that its primary contribution is a practical, data-driven methodology to reconstruct full multi-dimensional correlations from available partial data, allowing for a rigorous test of theoretical mixing schemes without assuming parameter independence.

Conservative Estimation: The constructed $\tilde{\chi}^2$ provides a conservative lower bound on the true $\chi^2$ , ensuring that viable models are not excluded due to approximation errors.
Model Independence: The approach is not tied to specific models; it can test any theoretical framework predicting algebraic relations among a subset of oscillation parameters.
Precision Era Necessity: The authors argue that as experimental precision increases (e.g., with DUNE, T2HK, JUNO), accounting for these correlations becomes essential. They demonstrate that full correlation structures can shift a model from marginal compatibility to exclusion (as seen with TM2), which would be missed by standard 1D/2D analyses.
Community Resource: By providing the code and interpolated functions, the authors aim to enable the community to test arbitrary sum rules and models against the full correlation structure of global fits.

The paper concludes that while $\mu_1-\tau_1$ (under IO) and TM1 remain compatible with current data, the TM2 scheme is strongly disfavored, highlighting the critical role of parameter correlations in distinguishing between competing theoretical models of neutrino flavor.

More like this