Imagine a sphere the height of a thirteen-storey building, hollowed out of clear acrylic, suspended inside a swimming-pool-sized cavern of ultrapure water, and filled with 20,000 tonnes of a faintly golden liquid that glows whenever a subatomic particle whispers through it. Bury that sphere roughly 700 metres below a hill in southern China, line it with tens of thousands of light-sensitive eyes, and you have the Jiangmen Underground Neutrino Observatory, JUNO. On 10 June 2026, the JUNO Collaboration published the first physics results from this instrument on the cover of Nature [1], and in doing so sharpened humanity's grip on one of the most stubborn puzzles in particle physics: how the three neutrinos arrange themselves in mass.
Neutrinos, the particles that barely exist
A neutrino is, in almost every sense, a near-nothing. It carries no electric charge, has a mass at least a million times lighter than an electron, and interacts with ordinary matter so rarely that roughly 65 billion of them pass through every square centimetre of your body each second, unimpeded, on their way out of the Sun. They come in three flavours, paired with the three charged leptons: electron, muon, and tau. In the 1990s and 2000s, experiments at Super-Kamiokande in Japan and SNO in Canada showed that neutrinos produced as one flavour will, given enough distance, reappear as another [2]. This flavour-switching, called oscillation, can only happen if the neutrino has mass. That single, slippery fact is enough to break the Standard Model of particle physics, the otherwise ruthlessly successful catalogue of fundamental particles and forces, which originally assumed neutrinos were massless.
What the oscillation experiments actually measure are not masses directly, but differences between the squares of masses. Two of those differences, denoted Δm²₂₁ and |Δm²₃₂|, govern how rapidly neutrinos swap identities as they travel, and the JUNO first result refines both at once [1].
The mass-ordering question
Because neutrinos are born and detected in flavour states but travel as a quantum superposition of three mass states, the three masses themselves can be arranged in only two distinct ways. In the normal ordering, two of the masses are light and cluster close together, with the third sitting slightly heavier above them. In the inverted ordering, the situation is flipped: two sit on top, one sits at the bottom. Knowing which arrangement nature has chosen is not a matter of taste. The answer reshapes predictions for almost every other neutrino experiment, from those hunting for an extremely rare kind of nuclear decay to those searching for cosmic CP violation, the subtle asymmetry between matter and antimatter that may explain why the universe contains something rather than nothing.
By 2025, the long-baseline accelerator experiments NOvA at Fermilab and T2K in Japan had published a joint analysis in Nature that constrained the mass-squared differences to better than 2% precision, but stopped short of declaring an ordering winner [3]. The preference tilted gently toward normal, yet remained statistically inconclusive.
Inside a 13-storey aquarium
JUNO is built to break that impasse by exploiting a deceptively simple trick. Nuclear reactors are prodigious sources of electron antineutrinos, and JUNO sits roughly 53 kilometres from a cluster of reactor cores in Guangdong, a baseline deliberately tuned to the point where the two candidate orderings imprint slightly different wiggles on the measured energy spectrum [2]. To see those wiggles, the detector has to be enormous, exceptionally clean, and precise about how much light each neutrino event produces.
The instrument's heart is a 35.4-metre-diameter acrylic sphere filled with 20,000 tonnes of an organic liquid scintillator, a transparent fluid that emits a flash of light whenever a charged particle deposits even a small amount of energy [4]. When a reactor antineutrino enters and meets a proton in the scintillator, it undergoes inverse beta decay, producing a positron and a neutron. The positron fires immediately, the neutron captures roughly 200 microseconds later, and the two successive flashes give physicists both the antineutrino's energy and a powerful tag against backgrounds.
Catching those flashes are more than 43,000 photomultiplier tubes (PMTs) pressed against the inside of the acrylic sphere: roughly 20,000 large 20-inch PMTs supplemented by 25,600 smaller 3-inch PMTs that fill in the gaps [4]. The whole assembly is held by a 41.1-metre stainless-steel truss and lowered into a 44-metre-deep pool of ultrapure water, which shields the scintillator from natural radioactivity in the surrounding rock and from stray cosmic rays that survive even 700 metres of granite overhead [4].
What 1.6× sharper actually means
The first JUNO physics analysis used 59 days of valid data collected between 26 August and 2 November 2025, a deliberately short commissioning run designed to validate the detector and test the analysis chain [1]. Even on that sliver of data, the collaboration reduced the combined uncertainty on Δm²₂₁ and |Δm²₃₂|/Δm²_ee by a factor of 1.6 relative to all previous reactor-neutrino experiments combined [1][5]. Put differently, two months of JUNO have done the work of more than a decade of global effort, and the detector has been running smoothly for roughly nine months as of publication, with new results expected to arrive in sequence through the summer and beyond [4].
The measured values are consistent with prior best estimates, which is itself useful. The exotic new-physics explanations some theorists floated to explain earlier discrepancies are now under even tighter pressure, and the 2022 sensitivity forecasts for the full six-year run still look achievable [6][7]. A News & Views commentary by Patricia Vahle and Zoya Vallari, published alongside the paper, frames the moment as the start of a precision era for neutrino oscillation physics, with JUNO taking its place alongside the long-baseline accelerator experiments as a defining instrument of the field [8].
What comes next
The mass-ordering question is still open. JUNO's first 59 days are not yet enough to pick a winner, and the collaboration's projections suggest that several more years of data will be required before the two hypotheses separate cleanly [9]. Complementary experiments, T2K, NOvA, Super-Kamiokande, and IceCube, will weigh in with their own channels, with the global picture expected to firm up around the end of the decade [8][3].
JUNO is also designed as a multi-purpose observatory. Over its lifetime, it expects to measure three of the six neutrino-mixing parameters to better than 1% precision, to detect neutrinos from a nearby galactic supernova, to map radioactive decays inside the Earth via geo-neutrinos, and to weigh in on the long-running solar-neutrino tension, the persistent disagreement between solar neutrino measurements and standard models of how the Sun shines [5]. The June 2026 paper is, in effect, a ribbon-cutting: the instrument is running, and the data are arriving faster than the analysis teams can digest them.
What genuinely remains uncertain, in the honest sense physicists use that word, is whether the neutrino mass ordering is normal or inverted, and whether JUNO's two-month glimpse is the leading edge of a steady improvement or the prelude to some quieter surprise. For now, the sphere glows in the dark under a Chinese hill, counting flashes, and the count is finally precise enough to matter.