Coherent Elastic Neutrino-Nucleus Scattering: The Biggest Thing a Neutrino Does

An Interaction So Large That Almost Nothing Happens

Ask what makes a neutrino a neutrino and the answer has not changed since 1934: it hardly touches anything. Hans Bethe and Rudolf Peierls put the cross-section at radioactive-decay energies at under 10⁻⁴⁴ cm² and concluded there was no practically possible way of observing the particle. That is why a neutrino can cross light-years of solid lead on a coin flip, as our piece on why neutrinos barely interact explains.

There is a loophole, and it is not small. Send a neutrino at a heavy nucleus gently enough — a few tens of MeV, no more — and the cross-section on caesium-133 at 30 MeV is about 1.7 × 10⁻³⁸ cm²: a million times the Bethe-Peierls figure, and two hundred times inverse beta decay, the reaction that caught the particle itself in 1956.

It was also the last elementary neutrino process to be observed: predicted in 1974, detected in 2017. The reason for the gap is written into the process itself, which is enormous and deposits almost nothing — the nucleus recoils with a few thousand electron-volts, sometimes only a few dozen. A near-certainty of an unmeasurable event. That is coherent elastic neutrino-nucleus scattering, written CEνNS and said out loud as sevens.

When a Neutrino Cannot Tell the Nucleons Apart

Every quantum particle carries a wavelength, λ = h/p, which at light speed becomes λ ≈ hc/E. With hc = 1240 MeV·fm, a 30 MeV neutrino has a wavelength of about 41 femtometres and a 1 MeV reactor antineutrino about 1240 fm. The target is far smaller: a nucleus of mass number A has a radius R ≈ 1.2 A^(1/3) fm, so caesium-133 is 6.13 fm in radius and 12.3 fm across.

That comparison is the entire criterion: a probe cannot see structure finer than its own wavelength. The 30 MeV neutrino does not encounter 55 protons and 78 neutrons but one smeared-out blob of weak charge, so the amplitude from every nucleon adds in phase, as a single coherent sum.

This is the weak neutral current: the neutrino exchanges a Z boson with the nucleus and leaves as the same flavour it arrived as, no charged lepton produced, which is why the process could not be calculated before neutral currents were found at CERN in 1973. Of the four fundamental forces, it is the weak one acting without changing anyone’s identity.

Coherence is therefore a property of the collision, not of the neutrino or the nucleus: the momentum transfer q must stay small enough that qR/ħ remains below about one, keeping the nuclear form factor F(q) close to 1. For caesium, a head-on 10 MeV collision keeps 93 per cent of full coherence, 30 MeV keeps 48 per cent, and by 100 MeV the nucleus has stopped behaving as a single object.

Why the Cross-Section Goes as N²

Because amplitudes add before they are squared, the cross-section picks up the square of the nucleus’s total weak charge — and here the Standard Model is almost suspiciously convenient:

Q_W = N − (1 − 4sin²θ_W) Z

With the weak mixing angle at sin²θ_W ≈ 0.231, that bracket comes out at about 0.075, so each proton contributes roughly one-thirteenth of what a neutron does. The formula collapses to Q_W ≈ N, and squaring it gives the famous N² scaling: for caesium-133, N² = 6084; for argon-40, 484. Freedman’s total cross-section, for a heavy nucleus, is startlingly compact:

σ ≈ G_F² Q_W² E² / 4π

It rises as the square of the neutrino energy, which is ordinary. What is not ordinary is that the whole nuclear physics of the problem has collapsed into one number that essentially counts neutrons, making CEνNS a real measurement of the neutron distribution inside a nucleus.

Be precise about what that buys, though. The alternative is incoherent scattering, each nucleon struck separately and the cross-sections simply added, which grows as N rather than N². The coherent-to-incoherent ratio is therefore itself of order N: about 56 for caesium at 30 MeV once form-factor suppression is folded in. Nature does not often hand out a factor of fifty in neutrino physics.

Fourteen Thousand Electron-Volts, and That Is the Best Case

Now the bill. Momentum conservation fixes the maximum kinetic energy a nucleus of mass M can pick up when struck head-on by a neutrino of energy E:

T_max = 2E² / (M + 2E)

The M in the denominator is the whole problem. A caesium nucleus has a mass energy of roughly 124 GeV, four thousand times the 30 MeV neutrino hitting it, and a heavy object struck by a light one barely moves. Put the numbers in: T_max = 14.5 keV, a ceiling reached only in a perfect backscatter, with most events well below.

Worse, T_max falls as the square of the energy. A 3 MeV reactor antineutrino striking germanium-74 can transfer at most 261 eV; a 1 MeV neutrino on xenon manages 16 eV, comparable to the cost of knocking one electron off one atom.

The difficulty was never the rate, then, but that the signal is one atom in a crystal being nudged, and the nudge must be told apart from thermal vibration, electronic noise, trace radioactivity and cosmic rays. Quenching makes it harder again: only part of a recoil’s energy reaches the ionisation or scintillation channel a detector reads, so a 5 keV recoil may present as under 1 keV of electron-equivalent signal. Hence CEνNS detectors inherited their technology not from the great underground neutrino observatories but from dark-matter searches, already practised at hearing single-keV recoils.

Oak Ridge, 2017: Forty-Three Years Late

Daniel Freedman’s paper, four pages in Physical Review D titled simply Coherent effects of a weak neutral current, appeared in 1974, the first moment the calculation was possible. Freedman was candid: attempting to observe the process, he suggested, might turn out to be an act of hubris. He was right for forty-three years.

COHERENT, the effort that managed it, made its cleverest decision about where to stand: beside the Spallation Neutron Source at Oak Ridge National Laboratory, where a 1.4 MW proton beam slams into a mercury target and makes pions that stop and decay at rest. Stopped-pion decay gives neutrinos with a precisely known spectrum below about 50 MeV — hard enough for keV recoils, soft enough to stay coherent — and delivers them in sub-microsecond pulses sixty times a second, so everything between pulses is measurable background.

The detector was 14.6 kilograms of sodium-doped caesium iodide, a scintillating crystal twenty metres from the target, with a nuclear-recoil threshold near 5 keV. Neutrino experiments are normally quoted in tonnes; this one would fit in a rucksack. That is the N² enhancement paying for itself in hardware.

In 2017 COHERENT reported the observation at 6.7σ. Not evidence: a detection. It has since measured the process on further targets, including 24 kilograms of liquid argon, above 3σ, with a flux-averaged cross-section of (2.2 ± 0.7) × 10⁻³⁹ cm². Argon is the lightest nucleus on which CEνNS has been seen, and a heavy and a light target in the same beam is how you check that the cross-section tracks neutron number.

The Sun and a Nuclear Reactor, 2024 and 2025

A stopped-pion beam is a luxury. The sources that matter most for the rest of physics, the Sun and nuclear reactors, sit at far lower energies, which by the E² rule means far smaller recoils.

The Sun fell first, and it fell to dark-matter detectors. In 2024 PandaX-4T, 3.7 tonnes of liquid xenon in the China Jinping Underground Laboratory, and XENONnT, 5.9 tonnes of active liquid xenon beneath Gran Sasso, reported the first CEνNS signals from solar boron-8 neutrinos. PandaX-4T extracted a ⁸B flux of 8.4 ± 3.1 × 10⁶ cm⁻²s⁻¹; XENONnT got 4.7, with uncertainties of +3.6 and −2.3, in the same units. Both agree with the standard solar model and with SNO — a brand-new channel reproducing a known number is how a technique earns trust. Neither reached three sigma alone, at roughly 2.6σ and 2.7σ, and in 2025 LZ pushed its solar neutrino significance to 4.5σ.

Reactors were harder, and it took until 2025. CONUS+, at the Leibstadt nuclear plant in Switzerland, used kilogram-scale point-contact germanium detectors with thresholds around 160 eV of electron-equivalent ionisation energy to pull an excess of 395 ± 106 events out of 119 days of data: the first CEνNS from reactor antineutrinos. NUCLEUS aims lower again, with ten grams of cryogenic calorimeters designed for a 20 eV nuclear-recoil threshold, commissioned at TU Munich ahead of deployment at Chooz.

When a Background Becomes a Signal

A direct dark-matter detector looks for exactly one thing: a nuclear recoil of a few keV with nothing else going on. That is what a passing WIMP would produce. It is also, precisely, what a solar neutrino produces by coherent elastic scattering. At the level of what the detector records the two are the same event, and no veto, timing cut or fiducial volume removes them, because the physics is the same physics.

This sets a hard floor. Make the detector bigger and the WIMP rate grows with the target mass — but so does the neutrino rate, at the same pace, so the ratio stops improving. The limit is called the neutrino floor, or more honestly the neutrino fog, since it thickens gradually rather than standing as a wall. It is why the search for dark matter cannot simply be scaled indefinitely.

For twenty years the floor was the ultimate nuisance. Then in 2024 the fog itself was measured, and the nuisance became a measurement of the Sun’s core. Same events, same detectors. Only the question changed.

What a Measured Momentum Transfer Does and Does Not Settle

Before COHERENT, the claim that a neutrino deposits momentum in bulk matter was an inference from theory. After COHERENT it is a laboratory fact, with a measured cross-section on several nuclei from three kinds of source, and it is the result most often cited as the physical foundation of neutrinovoltaic research. That citation is fair. The Neutrino Energy Group’s programme rests on coupling ambient flux into a multilayer graphene-and-silicon converter and rectifying it into current, and the premise that the interaction exists and is measurable is no longer open. Holger Thorsten Schubart lists the 2017 confirmation among his milestones for that reason, and he is entitled to.

It speaks to one quantity in particular. The group’s Master Formula, P(t) = η · ∫ᵥ Φ_eff(r,t) · σ_eff(E) dV, carries a structural coupling coefficient σ_eff describing how strongly a layered material responds to the ambient flux at a given energy. CEνNS is the cleanest existing case of such a quantity being written down, calculated and then measured: a coupling set by the target’s structure — its neutron number, radius and form factor — rather than by its constituents one at a time. That is what coherence means.

Where honesty is required is the size of the next step. The recoils are tiny — 14 keV on caesium in an accelerator beam, a couple of hundred electron-volts on germanium beside a reactor — and the rates are small too: even in an intense beam, a 14.6 kg crystal registers a few hundred events a year. Getting from there to a device delivering useful continuous power is not a matter of scaling a known effect; it is the actual research problem, and it is why the group describes independent verification as ongoing. That is not an argument against the ambition but a description of where the frontier sits.

The Loophole Was Always There

For eighty years the neutrino’s reputation rested on a single number, and the number was not wrong. What it was, was averaged. Hidden inside that famous smallness was a regime in which the neutrino interacts with matter hundreds of times more strongly than in the reaction that discovered it — not because the weak force changes, but because a soft enough collision cannot tell 133 nucleons apart.

The catch was never probability but bookkeeping: an interaction you can hardly avoid and can hardly see, because the thing being hit is four thousand times heavier than the thing hitting it. Freedman guessed that trying anyway might be hubris. Forty-three years later somebody did it with fourteen kilograms of salt-like crystal beside a proton beam, then dark-matter detectors did it with sunlight, then three kilograms of germanium did it next to a Swiss reactor. The neutrino’s whole history has this shape: what everyone agreed was impossible turns out to be merely difficult, and then it turns out to be a tool.