Neutrinos interact only very weakly with matter. That makes them hard to detect, but it also makes them worth the trouble even for purposes far removed from fundamental neutrino physics. Neutrinos can travel straight to astrophysical observers from remote sources otherwise veiled by obscuring material or magnetic fields. And since the 1960s, it’s been argued that they could do as much for geophysics.

Earth’s crust and mantle abound with unstable isotopes of thorium, uranium, potassium, and lesser contributors to the planet’s radioactivity. In principle, one could detect the “geoneutrino” flux from their decay chains and thus learn much about Earth’s composition and its heat sources. But it was not until 2005 that the KamLAND collaboration reported the first detection of geoneutrinos.1 Now, with five times as much data in hand, the international collaboration has reported its first substantive geophysics result.2 

The principal finding is that present-day radioactivity accounts for about half of the total heat flux from Earth’s interior. That’s no great surprise to geologists. But it does supply the first explicit measurement of the radiogenic heat flux—albeit still with large error bars—for comparison with the broad range of detailed models of Earth’s formation and the energy sources of plate tectonics and the geodynamo.

KamLAND, the Kamioka liquid-scintillator antineutrino detector, shown in figure 1, was completed in 2001, a kilometer underground in the disused Kamioka zinc mine in the mountains west of Tokyo. Its kiloton of liquid scintillator, monitored by thousands of photomultiplier tubes, reveals the interactions in the detector of electron antineutrinos (νes) from the beta-decay chains of 238U and 232Th in Earth’s crust and mantle—and from dozens of power reactors in and around Japan.

Figure 1. The underground KamLAND antineutrino detector in Kamioka, Japan, detects an energetic electron antineutrino (νe) from a reactor or from a radioactive decay in Earth’s interior by the scintillation light produced when the νe interacts with a hydrogen nucleus in the detector’s kiloton of scintillator oil. An array of phototubes inside the 18-m-diameter oil-containment vessel records scintillation pulses. To veto scintillation pulses from incident charged particles that mimic νe interactions, the containment vessel is immersed in water monitored by phototubes that detect an interloper’s Cherenkov radiation.

Figure 1. The underground KamLAND antineutrino detector in Kamioka, Japan, detects an energetic electron antineutrino (νe) from a reactor or from a radioactive decay in Earth’s interior by the scintillation light produced when the νe interacts with a hydrogen nucleus in the detector’s kiloton of scintillator oil. An array of phototubes inside the 18-m-diameter oil-containment vessel records scintillation pulses. To veto scintillation pulses from incident charged particles that mimic νe interactions, the containment vessel is immersed in water monitored by phototubes that detect an interloper’s Cherenkov radiation.

Close modal

KamLAND’s ability to detect geoneutrinos is a byproduct of the facility’s primary particle-physics purpose: monitoring the flavor-oscillation disappearance of νes that emanate from reactors. In fact, KamLAND’s purposeful placement in the midst of so many reactors creates a troublesome background for geoneutrino searches.

When an incident νe of sufficient energy interacts with a hydrogen nucleus in the scintillator fluid, the resulting inverse-beta-decay reaction

νe + p → e+ + n

manifests itself by a robust, double-barreled signal: A prompt scintillation signal comes from the positron’s brief travel and annihilation in the liquid. It provides a good approximation of the incident νe’s energy. Then after about 200 µs, a delayed monoenergetic scintillation pulse signals the 2.2-MeV gamma created when the neutron is finally captured by another proton to form a deuteron. Unfortunately, there’s no information about the direction from which the νe came.

The threshold νe energy for the inverse-beta-decay reaction is 1.8 MeV. That renders the detector blind to lower-energy νes from the 238U and 232Th decay chains, and to the entire decay spectrum of 40K, the third most important geoneutrino source. At the high-energy end, the geoneutrino spectrum extends up to 3.3 MeV and the reactor spectrum continues out to 8 MeV. Separating the two spectra in their overlap region can’t be done event by event. It requires maximum-likelihood fits of a model with free instrumental and geological parameters that incorporate detailed information about the crustal geology around Kamioka and the day-to-day operations of all the relevant reactors.

Only a few events per week pass the double-scintillation-pulse test and survive all other selection cuts—and the majority of those come from the reactors. The recorded time of each event is important because the likelihood function constrains the geoneutrino flux to be constant (within statistical fluctuations), wheras the reactor flux varies with power-station schedules and shutdowns. Ironically, reactor shutdowns due to earthquake damage help the geoneutrino search in two ways: They lower backgrounds, and they provide time tags that help distinguish signal from background.

Figure 2a shows the spectrum of the 841 inverse-beta-decay candidate events accumulated by KamLAND from 2002 through 2009 in the energy range relevant to geoneutrinos. The best fit, shown in the figure, attributes 106 ± 29 of them to geoneutrinos, and the rest to reactor νes, radioactive contaminants in the detector, and accidental scintillation coincidences. With those estimated backgrounds subtracted from the data, figure 2b compares what remains with the sum of the 238U and 232Th decay-chain geoneutrino fluxes expected from the team’s reference model. (The contributions of all other radioisotopes to KamLAND’s geoneutrino signal are thought to be negligible.)

Figure 2. Geoneutrino spectrum. (a) The best fit to the spectrum of all 841 geoneutrino candidate events recorded by KamLAND estimates the contributions from true geoneutrinos and from backgrounds due to radioactive contaminants in the detector, accidental scintillator coincidences, and νes from reactors in and near Japan. (b) With the estimated backgrounds subtracted, the residual measured spectrum is compared with the geoneutrino flux prediction of a reference model.3 The uranium-238 and thorium-232 decay-chain spectra that together make up the prediction are shown, respectively, as yellow and red curves. (Adapted from ref. 2.)

Figure 2. Geoneutrino spectrum. (a) The best fit to the spectrum of all 841 geoneutrino candidate events recorded by KamLAND estimates the contributions from true geoneutrinos and from backgrounds due to radioactive contaminants in the detector, accidental scintillator coincidences, and νes from reactors in and near Japan. (b) With the estimated backgrounds subtracted, the residual measured spectrum is compared with the geoneutrino flux prediction of a reference model.3 The uranium-238 and thorium-232 decay-chain spectra that together make up the prediction are shown, respectively, as yellow and red curves. (Adapted from ref. 2.)

Close modal

The reference model belongs to the widely accepted class of so-called bulk-silicate-Earth models. BSE scenarios for Earth’s formation assume that primordial Earth was a rather homogeneous rocky accumulation of silicates, with the relative abundances of metallic and rare-earth elements close to those of the most primitive meteorites.

About 50 million years later, the scenarios posit, radioactive heating accumulating on top of the planet’s residual heat of formation initiated the “iron catastrophe” that created the core when temperatures became hot enough to melt iron. Percolating down through the silicate rock, the liquid iron carried with it most of the siderophilic (iron-loving) elements like nickel and gold, leaving behind lithophilic (rock-loving) elements like U and Th. Thus BSE models assume that very little radioactivity now emanates from Earth’s core.

KamLAND’s reference model is based on a particularly comprehensive and widely cited 1995 BSE model,3 augmented by detailed upper-crustal data, especially from the Kamioka region. The model also takes account of the significant fraction of geophysical and reactor neutrinos rendered invisible to KamLAND by flavor oscillation en route.

To find the best geoneutrino fit, shown in figure 2a, the team treated the overall terrestrial abundances of 238U and 232Th as independent, free parameters. But the best-fit parameters turned out to be in good agreement with those of the reference model. The expected geoneutrino spectrum, shown in figure 2b, is calculated directly from the reference model without free geological parameters

From thermal gradients measured in thousands of deep boreholes all over the globe, geologists conclude that the total heat flux from Earth’s interior is about 44 terawatts. Some of that flux is surely of “primordial” origin. As the word is used by geologists, it refers to what’s left of the heat from the dissipation of gravitational energy in the formation of the planet and the later infall of its core, and from ancient radioactivity. The remainder, attributable to current radioactive heating, is what the KamLANDers were seeking to determine.

With geoneutrino data from only a single detector site, that’s a poorly constrained inverse problem. So the team availed itself of data from Borexino,4 the only other detector already in the geoneutrino business. Borexino is significantly smaller than KamLAND. But, sitting in a tunnel under Gran Sasso d’Italia, the highest mountain in the Apennines, it provides an excellent complement to KamLAND’s site, which straddles the margin of continental and Pacific occean crust. The floor of the Mediterranean around Italy, by contrast, is thoroughly continental. Furthermore, Borexino, which was originally built to study solar neutrinos, doesn’t have to contend with dozens of nearby reactors.

About 40 km thick, Earth’s continental crust, formed over eons by repeated vertical migration and remelting, is thought to have 100 times the U and Th concentrations of the 3000-km-deep mantle on which it floats. The U and Th concentrations of the thinner ocean-floor crust are only about 10 times those of the mantle, from which it is continually refreshed.

To estimate the worldwide contribution of the 238U and 232Th decay chains to Earth’s heat flux, the KamLAND team fitted its data and Borexino’s with a global model incorporating detailed crustal geology near both detectors. Some geological parameters were left free, but the overall ratio of Th to U atoms was held fixed at 3.9, a global number that geologists take to be well determined from meteorite compositions.

For simplicity, the model assumes that Th and U are uniformly distributed throughout the mantle and that their presence in the core is negligible. There are speculations, however, that a local uranium concentration somewhere in the solid inner core might be functioning as a natural fission reactor.5 The new data put an upper limit of 3 TW on the power of such a putative reactor. Whether the turbulent geodynamo action of the liquid outer core requires ongoing local radioactive heating remains an open question.

Figure 3 compares the total geoneutrino fluxes measured at KamLAND and Borexino with the fitted model’s expectations, given separately for the crust and mantle at the two detectors and also at Hawaii, a possible mid-ocean setting for a future detector.

Figure 3. Contributions of Earth’s crust and mantle to the geoneutrino fluxes at KamLAND (in Kamioka, Japan) and Borexino (in Gran Sasso, Italy). The contributions are estimated from the best fit of a global model to the fluxes measured by the two detectors. The model prediction is also shown for Hawaii, a typical oceanic-crust site. (Adapted from ref. 2.)

Figure 3. Contributions of Earth’s crust and mantle to the geoneutrino fluxes at KamLAND (in Kamioka, Japan) and Borexino (in Gran Sasso, Italy). The contributions are estimated from the best fit of a global model to the fluxes measured by the two detectors. The model prediction is also shown for Hawaii, a typical oceanic-crust site. (Adapted from ref. 2.)

Close modal

The KamLAND–Borexino fit yields a global 238U plus 232Th heat flux of 20.0 ± 8.7 TW. About 13 TW of that total is attributed to the mantle. “Radioactive heating of the mantle is of particular interest,” says KamLAND spokesman Kunio Inoue (Tohoku University, Sendai, Japan), “because it’s thought to contribute significantly to mantle convection, which drives plate tectonics and thus earthquakes.”

Adding the roughly 4 TW of heating estimated for the 40K decays neither detector can see, one gets a radiogenic contribution of about 24 TW to Earth’s total 44 TW heat flux. The data exclude, with a confidence level of 97%, the notion that Earth’s primordial heat is already exhausted.

In 1897, knowing that Earth’s interior was still giving up heat but not yet aware of radioactivity, Lord Kelvin proclaimed that the planet couldn’t be older than 40 million years, much to the annoyance of geologists and Darwinians. If Earth were any older, he argued from a naive conduction model, its heat of formation would already have radiated away.

Even before the geoneutrino results, the notion that the primordial heat of the 4.6-billion-year-old planet might already be exhausted had few adherents. But serious conjectures about the radiogenic fraction of Earth’s heat flux range from 30% to 70%. So, more geoneutrino detectors at geologically varied sites are clearly important for probing Earth’s interior with increasing precision. Directional sensitivity and lower detection thresholds would also help.

1.
T.
Araki
 et al (KamLAND collaboration),
Nature
436
,
499
(
2005
).
2.
A.
Gando
 et al (KamLAND collaboration),
Nat. Geosci.
(
2011
), .
3.
W. F.
McDonough
,
S. S.
Sun
,
Chem. Geol.
120
,
223
(
1995
).
4.
G.
Bellini
 et al (Borexino collaboration),
Phys. Lett. B
687
,
299
(
2010
).
5.
5.
J. M.
Herndon
,
D. A.
Edgerley
, http://arxiv.org/abs/hep-ph/0501216.