From K2K to DUNE: The Long-Baseline Generation

The 1998 Super-Kamiokande paper announcing the atmospheric neutrino anomaly contained one of the cleanest oscillation signals in particle physics: the muon-flavour deficit grew with zenith angle in just the way three-flavour oscillation predicts. By 1999 the question had shifted from “do neutrinos oscillate?” to “how precisely can we measure the mixing parameters?” — and the only way to answer it was to build dedicated experiments with controlled neutrino sources at known energies and baselines.

The K2K experiment, running from 1999 to 2004, was the first such facility. It sent a 1.3 GeV muon-neutrino beam 250 km from the KEK proton synchrotron to Super-Kamiokande, demonstrating the technique that has dominated neutrino-oscillation precision physics ever since. Twenty-five years later, DUNE is being built to do essentially the same thing with a fifty-fold more intense beam, a five-fold longer baseline, and a detector two orders of magnitude larger.

This post is about that progression. The five-step climb from K2K through MINOS, T2K, and NOvA into DUNE is one of the cleanest stories of generational progress in modern experimental physics, with each generation building deliberately on the lessons of the previous one. Looking at the sequence as a whole reveals what works, what does not, and what the long-baseline program is collectively trying to measure.

K2K: proving the principle (1999–2004)

K2K — “KEK to Kamioka” — pushed a 12-GeV proton beam from the KEK 13-GeV synchrotron onto a graphite target, producing positive pions that decayed in flight to give a ${\nu }_{\mu }$ beam with a peak energy around 1.3 GeV. The beam direction was tuned to point at Super-Kamiokande, 250 km away in the Kamioka mine. A near detector at KEK measured the beam composition and rate before oscillation, and Super-K detected the oscillated flux.

The geometry — 250 km at $\sim 1$ GeV — gave $L/E\approx 200$ km/GeV, near the first oscillation minimum for the atmospheric mass splitting. K2K’s central observation was a deficit of muon-neutrino events at Super-K compared to the near-detector-anchored prediction: 112 events observed where 158 ± 9 were expected, a 4.3-sigma confirmation of muon-neutrino disappearance.

K2K’s $\Delta m_{32}^2$ measurement, combined with Super-K’s atmospheric data, anchored the field’s understanding of atmospheric mass splitting at the few-percent level. The experiment’s main legacy beyond the measurement itself was demonstrating that long-baseline beam neutrino physics was possible: that one could deliver a coherent neutrino beam over hundreds of kilometres and reconstruct an oscillation pattern with a controlled source.

MINOS: precision at long baseline (2005–2016)

MINOS — the Main Injector Neutrino Oscillation Search — was the next major step. It used the Fermilab NuMI beam, a 120-GeV proton-induced neutrino beam with a peak energy of around 3 GeV, sent 735 km from Fermilab to the Soudan mine in Minnesota. The far detector was a 5.4-kiloton magnetised iron-scintillator tracker, with 484 alternating layers of 2.5-cm-thick steel and plastic scintillator.

The MINOS magnetic field — running at 1.3 Tesla through the iron — provided particle-charge identification, allowing ${\nu }_{\mu }$ and ${\bar{\nu }}_{\mu }$ events to be distinguished. This is critical for separating neutrino and antineutrino oscillation rates as a probe of CP violation and the mass ordering. The detector also had reasonable energy resolution thanks to the alternating layers, allowing differential measurements of the disappearance probability across the beam energy spectrum.

MINOS’s principal results were precision measurements of $|\Delta m_{32}^2|$ and ${\sin }^{2}2{\theta }_{23}$, with central values consistent with the atmospheric-neutrino picture but at substantially higher precision than K2K. The experiment also searched for ${\nu }_{\mu }\to {\nu }_e$ appearance — the channel that would later be the bread-and-butter of T2K and NOvA — and reported a hint of ${\theta }_{13}\ne 0$ that was eventually confirmed by the reactor experiments Daya Bay, RENO, and Double Chooz.

MINOS’s two-detector design (a near detector at Fermilab measured the unoscillated beam) became the template for every subsequent long-baseline experiment. The near-far ratio cancels many cross-section and beam-flux systematics that would otherwise dominate the measurement.

T2K: off-axis pioneering (2009–present)

T2K — “Tokai to Kamioka” — was the first experiment to deliberately exploit the off-axis beam technique. The neutrino beam from the J-PARC accelerator complex at Tokai was pointed not directly at Super-Kamiokande 295 km away, but 2.5 degrees off-axis. As discussed in our off-axis beam post, this geometry produces a sharply peaked neutrino energy spectrum around 600 MeV, optimised for the first oscillation maximum at the chosen baseline.

The narrow energy spectrum concentrates events at the relevant $L/E$ for ${\nu }_{\mu }\to {\nu }_e$ appearance, with the peak energy sitting just below the threshold for neutral-current pion production that would mimic the ${\nu }_e$ signal. The combination of narrow-band beam and water-Cherenkov far detector turned out to be exceptionally well matched to the discovery of electron-neutrino appearance, which T2K announced in 2011 at 2.5-sigma significance and confirmed at higher significance in subsequent updates.

T2K’s near-detector complex includes the ND280 detector — a magnetised tracking spectrometer with multiple subdetectors — that provides detailed cross-section measurements and unoscillated flux determinations. The combination of ND280 with Super-K’s far-detector imaging gives T2K some of the best-controlled systematics in the field.

By the mid-2020s, T2K has set the world’s tightest constraint on the CP-violating phase ${\delta }_{\mathrm{CP}}$, with the data favouring values near $3\pi /2$ that correspond to maximal CP violation in the direction that enhances ${\nu }_e$ appearance relative to ${\bar{\nu }}_e$ appearance. The constraint is at the 2-3 sigma level — suggestive but not discovery-quality.

NOvA: complementary baseline (2014–present)

NOvA — the NuMI Off-axis ${\nu }_e$ Appearance experiment — was designed as a long-baseline complement to T2K. It uses the same NuMI beam that MINOS used, but at a different geometry: 14 milliradians off the beam axis, with a 14-kiloton liquid-scintillator-and-PVC far detector at Ash River, 810 km from Fermilab.

The longer baseline of NOvA relative to T2K means the matter effect through Earth’s mantle plays a stronger role, contributing more directly to the asymmetry between neutrino and antineutrino appearance. The matter-induced asymmetry depends on the mass ordering, so NOvA’s data has more direct mass-ordering sensitivity than T2K — at the cost of less direct CP sensitivity, because the matter effect partially masks the genuine CP-violating contribution.

The NOvA far detector is a 14-kiloton liquid-scintillator-filled extruded-PVC structure, with the cells oriented perpendicular to the beam direction for good shower reconstruction. The technology was new at this scale and represented a deliberate choice to optimise for electron/photon discrimination in the appearance channel.

NOvA’s data has some tension with T2K’s preferred parameter values, with NOvA slightly favouring CP-conserving values closer to ${\delta }_{\mathrm{CP}}=\pi$ while T2K favours ${\delta }_{\mathrm{CP}}\approx 3\pi /2$. The tension is not yet decisive, but it has motivated joint fits of the two datasets and increased attention to cross-section systematics that affect them differently.

DUNE: the flagship (~2030 onward)

DUNE — the Deep Underground Neutrino Experiment — represents the largest step yet in the long-baseline programme. The Fermilab Long-Baseline Neutrino Facility will deliver a 1.2-MW (upgradable to 2.4 MW) wide-band neutrino beam pointing 1,300 km to the Sanford Underground Research Facility in South Dakota. The far detector will be a 40-kiloton liquid-argon TPC arranged in four 10-kiloton modules deep underground.

Three central design choices distinguish DUNE from its predecessors:

Wide-band on-axis beam. Unlike T2K and NOvA, DUNE uses an on-axis beam whose energy spectrum covers both the first oscillation maximum at $\sim 2.5$ GeV and the second at $\sim 0.8$ GeV. The wide spectrum allows DUNE to map the oscillation pattern across multiple energies in a single dataset, with self-consistency between the two maxima providing a powerful cross-check on the extracted oscillation parameters and on systematic uncertainties.

LArTPC far detector. Liquid-argon time projection chambers provide three-dimensional event reconstruction with millimetre-scale spatial resolution and excellent particle identification. This makes electron-neutrino appearance signatures clean against neutral-current pion backgrounds, and opens up rare-channel measurements (tau-neutrino appearance, supernova-neutrino flavour decomposition) that no other technology can match at the scale required.

Long baseline through the Earth. The 1300 km baseline drives substantial matter effects, which DUNE uses to determine the mass ordering with high significance — the matter-induced asymmetry between ${\nu }_e$ and ${\bar{\nu }}_e$ appearance distinguishes normal from inverted ordering by a calculable amount that depends on the baseline.

DUNE’s projected sensitivity includes 5-sigma discovery of CP violation for approximately 50% of the possible ${\delta }_{\mathrm{CP}}$ values within a few years of full operation, 5-sigma mass-ordering determination within 1-2 years of operation, and sub-degree precision on ${\theta }_{23}$. The full DUNE configuration also enables supernova-neutrino observation with thousands of events from a galactic burst, plus proton-decay search at a sensitivity competitive with Super-Kamiokande.

Hyper-Kamiokande as the parallel direction

Hyper-Kamiokande is the other major next-generation long-baseline experiment, designed as the successor to T2K at the same J-PARC beamline. It uses the same 2.5-degree off-axis geometry with a 295-km baseline, but with a 258-kiloton water Cherenkov far detector about 10 times larger than Super-Kamiokande. The beam intensity at J-PARC is being upgraded to 1.3 MW for Hyper-K running.

The strategic logic is to complement rather than duplicate DUNE. Hyper-K’s shorter baseline keeps matter effects small, isolating the CP-violating contribution to the appearance asymmetry more cleanly. The much larger far-detector mass collects $\sim 10$ times more events than T2K at the same energy. The combination of Hyper-K’s per-event statistics and DUNE’s broader-spectrum coverage provides complementary inputs to a joint fit on CP violation and the mass ordering.

What 25 years has taught the field

Several lessons from the K2K-to-DUNE progression apply beyond neutrino physics.

The near-far cancellation principle, first applied at MINOS, is now central to all long-baseline experiments. Cross-section uncertainties that would otherwise dominate the systematic budget are largely cancelled by measuring the same nuclear-target physics at the near and far detectors.

The off-axis trick, pioneered by T2K, has become a standard tool for tuning beam energy spectra to specific oscillation maxima. Its persistence into Hyper-K shows that the trade-off between flux and spectral cleanness favours cleanness for first-oscillation-maximum experiments.

The choice of far-detector technology matters as much as baseline. Iron-scintillator gave MINOS particle-charge ID; water Cherenkov gave T2K and Hyper-K cheap massive volumes; liquid-scintillator-PVC gave NOvA good electron/photon discrimination; LArTPC will give DUNE imaging-detector capability at scale.

The progression from K2K’s 250 km to DUNE’s 1300 km has tracked the field’s growing ambition: starting with confirmation of oscillation, then precision measurement of mass splitting and mixing angles, then ${\theta }_{13}$ discovery, then CP-violation discovery, and culminating in absolute-precision-era determinations of all oscillation parameters with full systematic control.

Summary

The long-baseline accelerator-beam programme has progressed through five generations in 25 years: K2K (1999–2004, 250 km), MINOS (2005–2016, 735 km), T2K (2009–present, 295 km off-axis), NOvA (2014–present, 810 km off-axis), and now into DUNE (2030, 1300 km on-axis wide-band) and Hyper-Kamiokande (2027, 295 km off-axis). Each generation has built on the previous one — the near-far cancellation from MINOS, the off-axis trick from T2K, the LArTPC technology from DUNE — and has pushed the precision on ${\theta }_{23}$, $|\Delta m_{32}^2|$, ${\theta }_{13}$, the mass ordering, and the CP-violating phase ${\delta }_{\mathrm{CP}}$ by a factor of a few per generation. The combined DUNE-plus-Hyper-K endpoint will deliver 5-sigma CP-violation discovery for substantial parameter regions, mass-ordering determination, and per-cent-level precision on all PMNS angles, completing the precision phase of three-flavour neutrino oscillation physics and setting the stage for whatever beyond-Standard-Model neutrino physics will look like.