JUNO-TAO: A Reference Reactor Spectrum at 30 Metres

JUNO — the Jiangmen Underground Neutrino Observatory — is a 20-kiloton liquid-scintillator detector positioned 53 kilometres from each of two reactor complexes in southern China. Its core physics goal is the determination of the neutrino mass ordering through the spectral interference between $\Delta m_{31}^2$ and $\Delta m_{32}^2$ in reactor antineutrino disappearance. The mass-ordering signature is a small but distinctive shift in the fast-oscillation phase that depends on the sign of $\Delta m_{31}^2$, and resolving it requires sub-3% energy resolution at the far detector — the best ever achieved in any large scintillator experiment.

That precision target imposes a corresponding requirement on the input reactor flux. If the predicted unoscillated spectrum from the reactor cores is uncertain at the few-per-cent level, the predicted oscillation pattern carries the same uncertainty, and the mass-ordering signal can be obscured by the flux modelling rather than by the underlying physics. The Huber-Mueller standard reactor flux model carries a quoted 2.5% uncertainty — borderline for JUNO’s needs — and the confirmed 5 MeV bump in the measured reactor spectra shows that real reactor flux deviates from Huber-Mueller in ways the model does not predict.

JUNO-TAO is JUNO’s solution to this problem. A small 2.8-tonne gadolinium-loaded liquid scintillator detector positioned just 30 metres from one of the Taishan Nuclear Power Plant reactor cores, TAO measures the reactor antineutrino spectrum essentially before any oscillation has set in. Its energy resolution at 2% at 1 MeV is the best of any reactor antineutrino experiment ever built. The resulting empirical reference spectrum anchors the JUNO far-detector analysis with systematic uncertainty far below what any model prediction can provide.

This post is about why TAO is needed, how its technology achieves the required precision, and what role it plays in JUNO’s mass-ordering measurement.

The flux-precision problem

JUNO’s mass-ordering measurement is built around the survival probability of reactor antineutrinos at the 53-km baseline. Writing it in the standard form:

$P({\bar{\nu }}_e\to {\bar{\nu }}_e)=1-{\sin }^{2}2{\theta }_{12}{\cos }^4{\theta }_{13}{\sin }^2\left(\frac{\Delta m_{21}^{2}L}{4E}\right)-{\sin }^{2}2{\theta }_{13}{\sin }^2\left(\frac{\Delta m_{ee}^{2}L}{4E}\right)$

The dominant feature at JUNO’s baseline is the fast oscillation governed by $\Delta m_{ee}^2$, which sits underneath an envelope set by the slower solar-mass-splitting oscillation. The sign of $\Delta m_{31}^2$ — the mass ordering — modulates the phase of the fast oscillation through the relation $\Delta m_{ee}^2=\Delta m_{31}^2-(1-\cos {\delta }_{\mathrm{CP}})\Delta m_{21}^2{\sin }^2{\theta }_{12}/2$. The shift between normal-ordering and inverted-ordering predictions is at the level of one to two per cent across the spectrum, concentrated in the 3-7 MeV positron-energy range where the fast oscillation has the most contrast.

JUNO’s predicted event rate at the far detector is the convolution of (a) the reactor antineutrino flux, (b) the inverse-beta-decay cross-section, (c) the propagation factor including oscillation, and (d) the detector response. Each step contributes systematic uncertainty. The dominant input uncertainty is the reactor flux itself: if Huber-Mueller’s 2.5% normalisation uncertainty propagates through, JUNO’s mass-ordering sensitivity drops from approximately $3\sigma$ to closer to $2\sigma$ after six years of operation.

TAO removes this dependence by providing an empirical reference. Since TAO measures the same reactor’s antineutrino spectrum at 30 metres — well before any oscillation has occurred — the TAO-measured spectrum is essentially the unoscillated flux that JUNO would see if there were no oscillation. The far-detector prediction is then obtained by applying the standard oscillation factor to TAO’s measured spectrum, with TAO’s measurement uncertainty replacing the Huber-Mueller modelling uncertainty.

The TAO detector

TAO’s central technical choices reflect its precision-reference role.

The active volume is 2.8 tonnes of gadolinium-loaded liquid scintillator (Gd-LS), with the gadolinium at 0.1% by mass providing the delayed-coincidence neutron-capture tag essential for clean inverse-beta-decay event identification. The Gd-LS is the same recipe used in Daya Bay, scaled down to a much smaller fiducial volume.

The photodetectors are silicon photomultipliers (SiPMs) rather than the conventional photomultiplier tubes used in Daya Bay and JUNO’s own far detector. SiPMs provide higher photon detection efficiency (about 50%), better timing, and complete coverage of the inner surface. They are cooled to $-50°$C to suppress dark-count noise and to push the energy resolution as low as possible.

The energy resolution at 1 MeV is approximately 2% — substantially better than Daya Bay’s 7% and JUNO’s far-detector design value of 3%. The factor-of-two improvement over JUNO’s far detector means TAO’s measured spectrum can resolve the same spectral features at finer detail, and any artefacts in JUNO’s far-detector response can be characterised against the TAO reference.

The baseline of 30 metres is short enough that oscillation effects are sub-percent across the entire spectrum, but long enough that the geometric flux suppression from the finite reactor core size is manageable. Designing the detector so close to a power-reactor core required careful engineering of radiation shielding (against neutron and gamma backgrounds from the reactor) and of the local geology (the reactor complex is on the seacoast, with associated radon and humidity controls).

What TAO will measure

TAO’s measurement programme has several components.

The headline product is the per-isotope antineutrino spectrum with 2% energy resolution across the relevant 1-10 MeV range. By recording event rates throughout the reactor fuel cycle and correlating with the reactor-operator’s reported isotope fractions, TAO will decompose the measured spectrum into ${}^{235}$U, ${}^{239}$Pu, and ${}^{241}$Pu contributions much as Daya Bay’s evolution measurement did, but at substantially higher precision. This provides direct empirical input to refine the Huber-Mueller predictions and to identify any sub-features (5-MeV-bump-like) that need flux-model corrections.

The second is the absolute flux normalisation at the 1% level, replacing Huber-Mueller’s 2.5% as the input to JUNO’s analysis. The absolute flux determination requires careful tracking of the reactor’s thermal power output (provided by the operator) and of TAO’s detection efficiency (calibrated by the LED light pulser and radioactive sources injected periodically). The 1% target is achievable with TAO’s design and represents the cleanest absolute reactor-flux measurement ever performed.

The third is detailed search for spectral features at small scales. TAO’s 2% resolution is fine enough to resolve any oscillation or structure at the few-hundred-keV scale, which is too fine for any previous reactor experiment. Substructure in the antineutrino spectrum from specific fission-fragment beta-decay branches becomes accessible for the first time, potentially shedding light on the underlying physics of the 5 MeV bump and on other anomalies.

A fourth, more speculative product is short-baseline sterile-neutrino search. At 30 metres, any sterile-neutrino oscillation with mass-squared splitting in the range $\Delta m^2\sim 0.1$ to $10$ eV² would produce a detectable distortion in TAO’s spectrum, complementary to STEREO, PROSPECT, and BEST searches at the gallium-anomaly mass scale.

TAO and the JUNO timeline

JUNO’s far detector is being filled and commissioned through 2025-2026, with first physics expected in 2026. TAO is being installed in parallel at the Taishan site and will begin data-taking on a similar schedule. The expectation is that TAO will have several years of high-statistics data before JUNO’s mass-ordering analysis matures, providing the empirical reference at the precision required when JUNO is ready for its principal physics output.

The combined JUNO-plus-TAO analysis represents a fundamentally different approach from earlier reactor experiments. Daya Bay’s analysis depended on the Huber-Mueller prediction as an absolute reference; the near-far cancellation handled the bulk of the cross-section and detection-efficiency uncertainties but left the residual flux modelling as a limiting systematic. JUNO replaces this with a direct, in-situ flux measurement from TAO at a precision substantially better than the model prediction, leaving the residual oscillation-physics extraction limited only by the actual physics rather than by external modelling.

Implications beyond JUNO

The TAO measurement programme will have effects beyond JUNO’s own mass-ordering analysis.

For the reactor antineutrino anomaly — the 6% global deficit relative to Huber-Mueller — TAO’s per-isotope decomposition will provide the most precise empirical determination of the per-isotope yields. The current best decomposition comes from Daya Bay’s 2019 evolution analysis with about 3% uncertainty per isotope; TAO will reach 1% per isotope after a few years of running. The combined picture should essentially settle the question of whether the global deficit is a ${}^{235}$U-specific flux-modelling issue (as Daya Bay’s data suggest) or whether some other element of the model is also off.

For future reactor experiments, TAO provides a reference spectrum that subsequent experiments can use as their input flux prediction, with the systematic uncertainty inherited from TAO rather than from the older Huber-Mueller model. This benefits any future reactor monitoring experiment, nonproliferation effort, or theta_13 cross-check.

For nuclear physics, the empirically-extracted per-isotope spectra are direct measurements of the cumulative beta spectra of fission fragments at unprecedented precision, providing input that nuclear-data libraries can incorporate to improve their summation calculations for other applications.

Summary

JUNO-TAO is a small 2.8-tonne gadolinium-loaded liquid-scintillator detector positioned 30 metres from one of the Taishan Nuclear Power Plant reactor cores, designed to measure the reactor antineutrino spectrum at 2% energy resolution before significant oscillation has set in. Its role is to provide an empirical reference spectrum for the JUNO far-detector analysis at 53 km baseline, replacing the Huber-Mueller flux-model prediction with a direct measurement. The 2% energy resolution is the best ever achieved in a reactor antineutrino detector, made possible by silicon photomultipliers cooled to $-50°$C and complete inner-surface coverage. TAO’s per-isotope flux decomposition, absolute flux normalisation, and high-resolution spectral measurement will refine the reactor antineutrino anomaly picture and provide reference inputs for future reactor experiments beyond JUNO. The TAO measurement is the key technical step that enables JUNO to reach its 3-sigma mass-ordering target within six years of operation, replacing the limiting external-model systematic with a directly-measured in-situ reference.