A theoretical study says an ultralight axion-like particle, or ALP, background could reduce the visible contrast of neutrino oscillations in a way that is independent of neutrino energy at leading order. When the calculation is matched to the first JUNO damping constraint, its aligned coupling benchmark reaches about 0.90 GeV⁻¹ for a local ALP density of 0.4 GeV/cm³.
The quoted 90% confidence level belongs to the reference energy-independent damping model used for the comparison. The mapped number is therefore an experimental scale matched to JUNO's baseline, rather than an independently calibrated confidence interval for the ALP theory.
The proposed effect is carried by a phase
The authors study a bottom-up derivative interaction between active neutrinos and an ultralight ALP background. They restrict the interaction to couplings that are diagonal in the neutrino mass basis, then derive the propagation phase with a first-order forward-scattering and WKB expansion, an approximation used to follow the wave as it travels.
At leading ultrarelativistic order, the ALP contribution can be retained as a factor tied to the field at the production and detection endpoints. A field redefinition can move pieces of the effect between the interior propagation phase and those endpoints, but it does not change the complete production-to-detection amplitude.
Once the endpoint field difference is averaged over unresolved configurations, each interference term is multiplied by the background's characteristic function, the mathematical object that records how a distribution of phases reduces coherent overlap. For a fixed-amplitude coherent mode sampled uniformly in phase, the factor becomes the order-zero Bessel function J₀.
The answer depends on what the experiment can sample
The leading result is simple only in a restricted regime. If the neutrino crosses the background slowly enough and the induced dephasing is small, the fixed-mode and Gaussian-halo descriptions agree at quadratic order. The resulting suppression scales as baseline squared, L², with zero leading power of energy, E⁰.
The calculation separates two clocks that are easy to confuse. Temporal averaging over the exposure is governed by the product of the ALP angular frequency and the observation time, written ωₐT. Changes in the field during one neutrino flight are governed by |ΩₐTflt|. The study notes that both conditions can hold at the same time.
For the first JUNO exposure, described as 59.1 live days spread across 69 calendar days, the nominal mass ranges run from a quasi-static regime below about 6.9 × 10⁻²² eV, through a finite-time transition up to about 6.9 × 10⁻²¹ eV, and then a coherent many-cycle range extending to about 1.1 × 10⁻¹⁶ eV. A many-patch halo regime runs up to about 4.1 × 10⁻¹² eV, above which finite-flight corrections become important.
A baseline match, not a finished ALP search
To make contact with JUNO, the authors compare the ALP pattern with a reference damping benchmark in which the damping is energy-independent, the 21 and 31 mass splittings share the same damping parameter, and the 32 term is left undamped. Under the aligned choice g₂ = g₃, the ALP model has the same pattern: the 21 and 31 interference terms receive the same J₀ suppression, while the 32 term is unchanged.
The reference analysis reports γ < 3.4 × 10⁻²² GeV at 90% confidence. Matching that suppression at JUNO's characteristic baseline gives the low-mass plateau |Δg₂₁| = |Δg₃₁| ≲ 0.90 GeV⁻¹ for the stated density. The unrounded fixed-mode value is 0.9016 GeV⁻¹, while the fixed-mode and Gaussian-halo prescriptions differ by 1.1%, so both round to the same approximate scale.
That agreement has a clear boundary. It applies to the low-dephasing plateau, not to the oscillatory large-argument behavior of the fixed-mode Bessel function or to a realistic high-mass exclusion curve for a virialized halo. Near the coherence transition, the boundary depends on event-time weighting and coherence conventions.
The electroweak completion points to a different scale
The study also examines what happens if the interaction is embedded in an electroweak-gauge-invariant dimension-five construction. Its charged-lepton translation gives |Δg₂₁| ≲ 4.4 × 10⁻⁹ GeV⁻¹, more than eight orders of magnitude below the GeV⁻¹ scale reached by the reactor-neutrino recast. The comparison depends on the electroweak and flavor assumptions, and the paper notes that a dimension-seven embedding still brings correlated interactions and matching effects.
What the number does and does not mean
The mapped result should be read as a baseline-matched experimental scale, not as a full ALP-specific confidence interval. Establishing that interval would require a likelihood analysis built directly in the ALP mass and coupling parameters.
The authors identify the endpoint-phase and characteristic-function treatment as the central analytical results, and regard the low-mass plateau as stable at roughly the percent level across the fixed-mode and Gaussian-halo descriptions. Their result is therefore a target for a more complete ALP-specific analysis, rather than its final statistical interpretation.
The document is arXiv:2608.25069v1, a hep-ph preprint dated 25 August 2026.
Paper data and sources
Original title: Energy-independent neutrino dephasing from an ultralight axion-like background
Authors: B. A. Couto e Silva, B. L. Sánchez-Vega
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-25
DOI: Not available
Original paper · Full text