Almost Nothing: How the Neutrino Won Its Fifth Nobel Prize

Long-form physics feature · 29 min read

The 2026 Nobel Prize in Physics went to a man who persuaded several hundred people to freeze 5,160 light sensors into the ice under the South Pole. It worked. The official wording, announced on October 6, honors Francis Halzen, an 82-year-old Belgian-born theorist at the University of Wisconsin–Madison, “for decisive contributions to the IceCube Neutrino Observatory and the discovery of high-energy neutrinos of astrophysical origin.”

Eleven years earlier, to the day, the same prize went to Takaaki Kajita of Japan and Arthur McDonald of Canada, “for the discovery of neutrino oscillations, which shows that neutrinos have mass.” Their instruments were a tank of water deep in a Japanese zinc mine and a ball of heavy water deep in a Canadian nickel mine.

Same particle both times. By my count this is the fifth physics Nobel whose citation names the neutrino: 1988, 1995, 2002, 2015 and now 2026. That is a remarkable record for something with no electric charge, almost no mass, and so little interest in the rest of the universe that about a hundred trillion of them pass through your body every second, and in any given second not one of them stops.

I have a small personal stake in this. I finished my physics degree in 1997, and the neutrino in my textbooks had a mass of exactly zero. One year later a water tank in Japan announced that this was wrong. It is humbling to have your education go out of date faster than your computer.

These days I write software for a living, so I can’t help reading the whole story as an engineering one, with two bug reports that stayed open for decades and a production system that nobody will ever be able to touch again.

Meet almost nothing

A neutrino is an elementary particle, like the electron, with three differences that matter.

It has no electric charge. Its mass is at least a million times smaller than the electron’s, and for decades it was assumed to be exactly zero. And of the forces that hold the world together it feels only gravity and the one physicists call the weak force, which is well named.

There are three kinds, called flavors: the electron neutrino, the muon neutrino and the tau neutrino. The muon and the tau are heavier cousins of the electron, and each neutrino is named after its partner. On the rare occasion that a neutrino hits something, it can turn into that partner. An electron neutrino leaves an electron behind, a muon neutrino leaves a muon, and that is how a detector knows which kind it has caught. (Showing that the muon’s neutrino really is a different particle from the electron’s is what the 1988 prize was for.)

They are everywhere. The Sun makes them as a by-product of fusion, and about 65 billion solar neutrinos cross every square centimeter of you every second. At night too: they come up through the Earth and through your mattress. You make your own as well. The radioactive potassium in a human body emits about 4,000 of them a second, which makes each of us a small, steady neutrino source, and makes a banana a smaller one.

And nearly all of them go through whatever is in their way. What stops a particle in ordinary matter is almost always electric force, charge pulling on charge, and a neutrino has no charge to pull on. In 1934 Hans Bethe and Rudolf Peierls calculated that a neutrino should be able to cross more than ten light-years of solid matter, and concluded, in the journal Nature, that there was “no practically possible way of observing the neutrino.”

Keep that sentence in mind. Everything below is a way around it.

A letter from a man on his way to a ball

The neutrino was invented before it was discovered, by a man who was embarrassed about it.

Around 1930 a kind of radioactivity called beta decay was misbehaving. An atomic nucleus spits out an electron, and the electron should carry off the same amount of energy every time. It didn’t. Sometimes it carried a lot, sometimes a little, and the rest of the energy was simply gone. Conservation of energy, the most trusted law in physics, appeared to be broken, and some serious people, Niels Bohr among them, were prepared to let it go.

On December 4, 1930, Wolfgang Pauli wrote to a meeting of physicists in Tübingen. His letter opens “Dear Radioactive Ladies and Gentlemen” and proposes what he calls “a desperate remedy”: a new particle, electrically neutral and very light, that leaves the nucleus together with the electron and carries away the missing energy without being noticed. He regretted that he could not present the idea in person. He was, he wrote, “indispensable here in Zurich because of a ball on the night of 6/7 December.”

Pauli called his particle the “neutron.” In 1932 that name went to the heavy neutral particle found inside the nucleus, and Enrico Fermi’s group in Rome gave Pauli’s much lighter one an Italian diminutive: neutrino, the little neutral one. Pauli himself is said to have confessed that he had done a terrible thing, by proposing a particle that could not be detected.

It took 26 years, and the way around Bethe and Peierls was brute force. If almost none of them will stop, go where there are absurdly many. Frederick Reines and Clyde Cowan first planned to catch neutrinos from an exploding atomic bomb: a 20-kiloton blast, with a detector nicknamed “El Monstro” dropped down a shaft 40 meters away at the moment of detonation. Their laboratory at Los Alamos approved the plan. Then they decided that a nuclear reactor would be less spectacular but more useful, set up their detector 11 meters from one at Savannah River in South Carolina, counted about three neutrinos an hour, and in June 1956 sent word to Pauli that his particle existed.

The Nobel Prize for that detection arrived in 1995, thirty-nine years later. Reines collected it without his partner, because Cowan had died in 1974. Neutrino physics rewards patience on a scale that most careers cannot afford.

Swipe sideways to explore the full figure →

Bug report no. 1: two thirds of the Sun’s neutrinos are missing

In the late 1960s a chemist named Raymond Davis Jr. filled a tank with 380 cubic meters of dry-cleaning fluid and parked it a kilometer and a half underground, in the Homestake gold mine in South Dakota. He needed the rock. The surface of the Earth is under constant fire from cosmic rays, fast charged particles from space, and they would have swamped anything a neutrino did in the tank. A kilometer and a half of rock stops them and does nothing to a neutrino.

The fluid was rich in chlorine, and once in a while a neutrino from the Sun would turn one chlorine atom into an atom of argon. Every two or three months Davis flushed the tank and counted the new argon atoms, which is possible because this kind of argon is radioactive and each atom gives itself away when it decays. There were a dozen or two of them, in a volume the size of a swimming pool.

Meanwhile a theorist, John Bahcall, had calculated how many neutrinos the Sun should deliver. Davis kept finding about a third of that.

For the next thirty years both men were told, politely and often, that one of them must have made a mistake. Either the chemistry was off, which seemed plausible to anyone asked to believe that a dozen or two atoms can be fished out of 600 tonnes of liquid, or the model of the Sun was. Davis refined the experiment and ran it until 1994. Bahcall refined the calculation. The answer stayed at one third.

Every developer knows this situation. The test fails. You check the test and it is correct. You check the code and it is correct. What remains is the possibility nobody wants to say out loud, which is that the specification is wrong.

Bug report no. 2: a physicist debugs his own code for a year

The second bug turned up in Japan, in an experiment built for something else.

Kamiokande was a tank holding 3,000 tonnes of water, a kilometer under a mountain, finished in 1983. The last three letters of its name stood for Nucleon Decay Experiment: its job was to watch for protons falling apart. No proton has ever obliged.

But a tank that can see a proton die can also see a neutrino arrive. When a neutrino hits something in the water and turns into its charged partner, that particle flies off faster than light can travel through water. (Nothing outruns light in a vacuum. In water, light is slowed to about three quarters of that speed and can be overtaken.) The result is the optical cousin of a sonic boom, a cone of faint blue light, and the sensors on the walls of the tank see it as a ring. A muon draws a ring with a sharp edge. An electron draws a fuzzy one.

In 1986 a newly minted PhD named Takaaki Kajita was improving the software that tells the two apart and sorts events into electron-like and muon-like. On simulated data his classifier was right 98 percent of the time. Then he ran it on real data, on neutrinos produced when cosmic rays strike the atmosphere, and the muon count came out far too low. In the numbers the group eventually published:

• electron-like: 93 seen, 88.5 expected
• muon-like: 85 seen, 144 expected

His first thought was the one every programmer has: I broke something. In his Nobel lecture he describes suspecting a serious mistake of his own, checking events by eye, and then spending about a year with a colleague, Masato Takita, searching the simulation, the event selection and the reconstruction for the error. They did not find one. In 1988 the group published the deficit, with the cautious remark that neither the detector nor the calculation could account for it, and that a change of neutrino type on the way might.

I find this the most encouraging detail in the whole story. A Nobel Prize began as a number that looked like a bug, reported by people who had spent a year trying and failing to prove that the bug was theirs.

Kamiokande, incidentally, may be the most successful failed experiment in physics. In February 1987 it caught about a dozen neutrinos from a star that had exploded 160,000 light-years away, out of roughly ten thousand trillion that swept through the tank. That earned its founder, Masatoshi Koshiba, a share of the 2002 Nobel Prize, alongside Davis. Koshiba had been Kajita’s doctoral advisor. The mine has now produced one Nobel Prize for the teacher and another for the student.

To settle the muon question, Japan built the same detector again, much bigger. Super-Kamiokande is a steel cylinder about 40 meters across and 40 meters tall, holding 50,000 tonnes of very pure water and watched by more than 11,000 light sensors the size of beach balls. It began taking data in April 1996.

The trick: use the whole planet

Cosmic rays strike the atmosphere all around the globe, so atmospheric neutrinos reach the detector from every direction, including from below, since the Earth is no obstacle to them. A neutrino coming from straight overhead has traveled about 15 kilometers. One coming from straight below was made in the air on the far side of the planet and has traveled about 13,000.

The source, the detector and the particle are the same in both directions. Only the length of the trip differs. If nothing happens to a neutrino on the way, the two directions must give the same count.

They didn’t. In June 1998, at a conference in Takayama, Kajita presented 535 days of data. For every 100 energetic muon neutrinos arriving from above, only about 54 were arriving from below. Electron neutrinos showed no such difference. The effect was more than six standard deviations from anything chance would produce, which is how physicists say “this is real.”

Nearly half of the muon neutrinos that crossed the planet had gone missing on the way.

Swipe sideways to explore the full figure →

Root cause: the type changes in flight

The missing muon neutrinos had changed flavor.

A neutrino is born with a definite flavor, say muon. But in quantum mechanics a particle is also a wave, and a muon neutrino turns out to be three waves traveling together, one for each of three slightly different masses. Picture three guitar strings tuned very nearly to the same note. Plucked together they sound like one note at first. Then they drift out of step, and the sound begins to swell and fade. In a neutrino what fades is the muon flavor, and what swells in its place is mostly tau. A neutrino that gets caught is always one flavor or another, and the trip only shifts the odds. Catch it a few hundred kilometers from where it was born and there is a real chance that it answers to a different name.

A programmer might put it this way: everyone assumed type was a field, and it turned out to be a method. Its argument is the distance traveled, divided by the energy.

Bruno Pontecorvo had suggested as early as 1957 that neutrinos might do something of the kind. The effect is called oscillation because the probability swings back and forth with distance, and in the simplest case, with two flavors, the whole thing fits in one line of code:

from math import sin

def still_muon(km, gev, dm2=2.5e-3):
    """Chance that a muon neutrino with energy `gev` (in GeV)
    is still a muon neutrino after `km` kilometers."""
    return 1 - sin(1.27 * dm2 * km / gev) ** 2

still_muon(15, 1)       # 0.998  from the sky above: nothing has happened yet
still_muon(495, 1)      # 0.000  500 km later it is certainly something else
still_muon(12_700, 1)   # 0.755  through the Earth: many swings later

Distance is in kilometers and energy in GeV, billions of electron-volts, the everyday unit of particle physics. A typical atmospheric neutrino carries about one. The 1.27 takes care of the units, and I have assumed that the two flavors mix completely, which is close to what is measured. The interesting constant is dm2, the difference between the squares of two of the masses. If it were zero, the sine would be zero and no neutrino would ever change. They do change. So those two masses differ, and two masses that differ cannot both be zero.

Averaged over thousands of kilometers and a spread of energies, the swings blur into “about half,” which is the 54 out of 100. The missing ones had become tau neutrinos, which Super-Kamiokande could barely see.

Swipe sideways to explore the full figure →

Neutrino energy: 1.0 GeV

The Standard Model of particle physics, the specification that had passed every test for a quarter of a century, said that neutrinos have exactly zero mass. On that line the specification was wrong.

Closing the first ticket

That left Davis’s missing solar neutrinos, and now there was an obvious suspect. The Sun produces only electron neutrinos. Davis’s tank could detect only electron neutrinos. If two thirds of them changed flavor somewhere between the Sun’s core and his tank, he would see exactly what he had been seeing since the sixties.

To prove it you need a detector that can count the same neutrinos twice: once counting electron neutrinos only, and once counting all three flavors alike. In 1985 a physicist named Herb Chen pointed out how. Use heavy water, in which each hydrogen nucleus carries an extra neutron. One reaction in heavy water responds to electron neutrinos only. Another, in which a neutrino of any flavor simply knocks that loosely bound nucleus apart, treats all three flavors the same. Chen died of leukemia in 1987, before the detector was built.

Canada happened to own a great deal of heavy water, because its nuclear reactors are built around it. The Sudbury Neutrino Observatory borrowed 1,000 tonnes, worth about 300 million Canadian dollars, and hung it in a transparent acrylic sphere 12 meters across, two kilometers down a working nickel mine in Ontario, surrounded by some 9,500 light sensors. Arthur McDonald directed the project.

The data taken between November 1999 and May 2001 gave two numbers. SNO could see only the most energetic solar neutrinos, roughly one in ten thousand of the 65 billion, so the unit here is millions per square centimeter per second:

• counting electron neutrinos only: 1.76
• counting all three flavors: 5.09

Bahcall’s prediction for what the Sun sends: 5.05.

Nothing had ever been missing. The Sun was shining exactly as calculated. Two thirds of its neutrinos were arriving in disguise, and for thirty years the detectors on duty had been looking for a single costume.

The ticket that Davis opened in 1968 was closed in 2002. Not a bug. Works as designed. Documentation updated. Davis received his Nobel Prize that same year, at 88. Bahcall, whose calculation had been right all along, was not on the list.

Swipe sideways to explore the full figure →

Stockholm, 2015

On October 6, 2015, Kajita and McDonald shared the prize, each standing in for a collaboration of hundreds.

Someone was missing from the stage. Yoji Totsuka had led Super-Kamiokande through the discovery and then through the worst day in its history. In November 2001 one of the light sensors, each a large glass vacuum bulb under tens of meters of water, imploded, and the shock wave cracked its neighbors, which imploded and cracked theirs. About 6,600 sensors were lost in a chain reaction, more than half of the detector’s eyes. Anyone who has watched one failed node take down a cluster will recognize the pattern. The detector was rebuilt, this time with each sensor inside a protective shell. Totsuka died in 2008, and the Nobel Prize is not awarded posthumously.

And what did the discovery say about the mass itself? Annoyingly little. Oscillations reveal only the differences between the squares of the masses. From those we know that the heaviest neutrino weighs at least 0.05 electron-volts or so. (The electron-volt is a unit of energy. Particle physicists weigh things in it too, on the authority of E = mc².) From a very different experiment in Germany, KATRIN, we know that it weighs less than 0.45. KATRIN repeats Pauli’s bookkeeping with a spectrometer ten meters across: it measures the energy of the electrons that fly out of decaying tritium and looks for the sliver that the neutrino’s mass holds back. The electron, the next lightest particle that has any mass, weighs 511,000. So the neutrino’s mass sits in a narrow window more than a million times below everything else, and nobody knows why.

A telescope that looks down

So far the neutrino has been the thing under study. The 2026 prize is for turning it into an instrument.

The cosmic rays that drove Davis underground were discovered by Victor Hess in 1912. A few of them carry energies far beyond anything our accelerators can reach, and more than a century later we still don’t know where most of those come from. The trouble is that cosmic rays carry electric charge, and the magnetic fields of the galaxy bend their paths into knots. By the time one arrives, its direction says nothing about its origin.

Whatever cosmic machine can accelerate those particles should also produce neutrinos, because some of the particles are bound to smash into the gas and light around it, and neutrinos come out of the wreckage. A neutrino has no charge and almost never interacts. It travels in a straight line from its source, through gas, dust, galaxies and magnetic fields. Catch one, work out which way it was going, and you are looking straight at the machine.

The hard part is catching one. Energetic cosmic neutrinos are rare, and the estimates all pointed the same way: the detector would have to be about a cubic kilometer. That is a billion tonnes of something transparent, in complete darkness, wired with sensors. Nobody is going to build a tank that size. You have to find one.

The first idea was the ocean. The Soviet physicist Moisey Markov proposed it in 1960, and an American project called DUMAND spent two decades trying to hang strings of sensors 4.8 kilometers down in the Pacific off Hawaii. Its first string went into the water in December 1993, leaked, and short-circuited. The funding was withdrawn in 1995. A Soviet and later Russian team did better in Lake Baikal, on a much smaller scale.

Francis Halzen is a theorist. He had written papers on particle physics and, with Alan Martin, a textbook, Quarks and Leptons, that physics students have been carrying around since 1984. In the late 1980s he heard that Soviet scientists were trying to detect neutrinos with radio antennas in the ice at Vostok station in Antarctica. He worked out that their method would not be sensitive enough, and then wondered about ordinary light instead. Ice is transparent. At the South Pole it is nearly three kilometers thick. The cubic kilometer already existed. It was dark, it was frozen in place, and it was free.

He wrote to John Learned, who led DUMAND, and in 1988 the two presented a paper called “High energy neutrino detection in deep polar ice.”

On the day of the prize, speaking by video from Italy to reporters gathered in Madison, Halzen described how the idea was received. “To summarize, people always thought this was like a cute idea, right?” he said. “But it would never work.” And when it was funded anyway: “I must say I became very worried.”

The first attempt failed

He had reason to worry. The prototype, called AMANDA, lowered its first strings of sensors to between 800 and 1,000 meters in the Antarctic summer of 1993–94 and found the ice full of tiny air bubbles. A flash of light could not travel a meter before being scattered. It was a telescope made of frosted glass.

But bubbles shrink under pressure, and the deeper ice is squeezed harder. Two years later the team went down to between 1,500 and 2,000 meters, and there the bubbles were gone. The deep ice turned out clearer than anyone had dared to hope: light travels about 20 meters there before it is scattered and about 100 before it is absorbed. By January 2000 AMANDA had 677 sensors on 19 strings and was recording neutrinos.

“Some were overcome with pure luck,” Halzen told the reporters in Madison, speaking of the obstacles, “but that’s science.”

IceCube

The real thing was built in seven Antarctic summers, from 2004 to the end of 2010.

• 86 holes, each 60 centimeters wide and two and a half kilometers deep, melted into the ice with about 5 megawatts of hot water.
• About two days and 750,000 liters of meltwater per hole.
• Into each hole, before the water froze again, a cable carrying 60 sensors, spaced out between 1,450 and 2,450 meters below the surface. That makes 5,160 sensors.
• A total cost of 279 million dollars.

Together they make a three-dimensional grid of light sensors filling one cubic kilometer of ice.

When a neutrino does strike an atomic nucleus in that ice, the collision throws out charged particles, and they trail the same cone of blue light that the water tanks relied on. Its name is Cherenkov radiation. The sensors record when the light reaches them, to within a few billionths of a second, and from the arrival times and the amount of light software works out which way the particle was going and how much energy it had.

And the telescope looks down. Every day about 275 million particles from cosmic rays rain into the detector from the sky above, and against those it records about 275 neutrinos. One in a million. The cleanest way to pick them out is to look for particles coming up from below, because a neutrino is the only thing that can cross the whole Earth and arrive from underneath. IceCube sits at the bottom of the world and uses the planet as its filter.

Swipe sideways to explore the full figure →

Notes on a system nobody can ever touch again

This section is for my own tribe.

Each of the 5,160 sensors is a pressure-proof glass sphere 33 centimeters across, holding a light detector and a small computer of its own that digitizes the signal on the spot and sends it up the cable. Consider the deployment constraints.

• No hardware access, ever. Once the hole refreezes, the sensor is locked in for good. It cannot be repaired, swapped or even looked at. Software is another matter: it is updated remotely, over the cable.
• The install is violent. As the water in the hole turns back into ice it squeezes whatever is in it, so every sphere was tested to 690 bar, nearly 700 times atmospheric pressure.
• Uptime. A status report published in 2017 found 98.4 percent of the sensors in the ice still working, and the detector as a whole taking data 99 percent of the time.
• Every sensor has a name. Each cable has a theme, and its 60 sensors are named to match. The last one assembled at Stockholm University is called Sista, which is Swedish for “last,” and the crew signed the final sensor before it went into the ice in December 2010.
• Bandwidth is the bottleneck. The detector produces about a terabyte of raw data a day. Roughly a tenth of that, after filtering, goes north by satellite. The rest is written to disks, and the disks leave once a year.
• On-call is two people. They are called winterovers. They stay at the Pole for a full year, including the months of darkness when it is too cold for routine flights, and keep the data flowing.

And the entire installation is moving. The ice sheet at the Pole slides about ten meters a year, and it carries the detector along as a single block.

Bert, Ernie and the neighbors

At first IceCube saw what was expected, which was neutrinos made in our own atmosphere, the same kind Kajita had studied. Particle physicists have an old saying that yesterday’s discovery is today’s calibration and tomorrow’s background. Kajita’s Nobel-winning signal had become Halzen’s noise.

Then, in 2012, the team found two events in their data, recorded in August 2011 and January 2012, that did not fit. Each had released about 1,000 trillion electron-volts in the ice. That is roughly 150 times the energy of a proton in the Large Hadron Collider, carried by a single neutrino, and the atmosphere almost never makes them like that. The collaboration named the two events Bert and Ernie, after the characters on Sesame Street.

A dedicated search through two years of data then turned up 26 more, 28 in all, with energies and arrival directions that the atmosphere could not explain. The paper appeared in Science in November 2013 under a careful title, “Evidence for High-Energy Extraterrestrial Neutrinos at the IceCube Detector.” Halzen’s comment at the time: “This is the dawn of a new age of astronomy.”

Where do they come from? That is taking longer, and the answers are still arriving one at a time.

• September 22, 2017. A single neutrino of about 290 trillion electron-volts arrives from a point just off the left shoulder of Orion. IceCube sends the coordinates to astronomers around the world, and about twenty observatories on the ground and in orbit turn to look. They find a known galaxy with a giant black hole at its center, TXS 0506+056, about 4 billion light-years away, in the middle of an outburst. It is the first time a high-energy cosmic neutrino has been matched to a likely source.
• November 2022. NGC 1068, a galaxy 47 million light-years away in the constellation Cetus and visible with large binoculars, turns out to have sent about 80 neutrinos.
• June 2023. The first picture of our own Milky Way made with neutrinos instead of light, drawn from 60,000 events collected over ten years and sorted with the help of machine learning.

The ocean idea had not died with DUMAND. In February 2025 KM3NeT, a telescope under construction at the bottom of the Mediterranean and about one tenth complete at the time, reported a neutrino it had recorded two years earlier, with an estimated energy of 220,000 trillion electron-volts. That is more than 30,000 times the energy of a proton in the Large Hadron Collider. Put differently, one subatomic particle was carrying about the energy of a ping-pong ball dropped from chest height. Nobody knows yet what produced it.

Swipe sideways to explore the full figure →

One name on the prize

This year’s physics prize went to a single person, for the first time since 1992. IceCube is the work of about 450 scientists at 58 institutions in 14 countries, and that is before counting the drillers and the people who have spent winters at the Pole. Halzen’s statement that day was about them: “It’s a great relief for me to finally deliver the recognition that this great collaboration deserves.”

The chair of the Nobel Committee for Physics, Mark Pearce, gave the committee’s reasoning in one sentence: “His tenacity and scientific vision has paved the way for a new kind of astronomy.”

A month after Kajita and McDonald’s prize was announced, the Breakthrough Prize in Fundamental Physics honored the same discovery and divided its three million dollars among five experiments and 1,377 physicists. A Nobel Prize can go to three people at most, and this one went to one. A cubic-kilometer detector is built by hundreds of people. It also needed one person to keep saying, for the twenty-two years between a conference paper in 1988 and the last sensor going into the ice in 2010, that the cute idea would work.

What we still don’t know

Nobody has yet weighed a neutrino. The mass has only been fenced in: by the window described above, and by astronomers, who read a tighter limit off the way matter clumps in the universe, on the condition that their model of the universe is right. Nobody knows in what order the three masses come, although JUNO, a new detector in southern China, has been taking data since August 2025 and said in July 2026 that it leans toward the arrangement with two light neutrinos and one heavier one, while calling it too early to say. And most of what IceCube sees cannot yet be traced to any source.

The question with the most riding on it concerns antimatter. Every particle has a mirror-image twin, and the Big Bang should have made the two kinds in equal amounts, which would then have wiped each other out. Something tipped the balance. If neutrinos and their twins change flavor at slightly different rates, that may be part of the reason there is anything here at all. (Having no charge, the neutrino might even be its own twin. Experiments are looking, and none has seen a sign of it.) Two enormous experiments are being built to measure those rates. Hyper-Kamiokande in Japan will hold 260,000 tonnes of water and is due to start in 2028. DUNE, in the United States, will fire a beam of neutrinos 1,300 kilometers through the Earth, from Fermilab near Chicago to detectors in South Dakota. Those detectors are going into the old Homestake mine, on the same level where Ray Davis kept his tank of dry-cleaning fluid.

IceCube itself received its first major expansion in fifteen years this past Antarctic summer: more than 600 new sensors and calibration devices, packed closely together in the middle of the array. A successor with eight times the volume, IceCube-Gen2, has been proposed but not yet funded.

On the phone to Stockholm, Halzen mentioned that he was in the middle of writing a proposal and hoped the prize would help to get it approved. He was joking. I doubt he was only joking. Asked what comes next, he said, as Reuters reported it: “This is just an introduction to the science, the astronomy is still to come.”

Almost nothing

So that is the particle. No charge, almost no mass, no interest in us at all. Pauli was embarrassed to have suggested it. Bethe and Peierls concluded that it could never be observed. It has since been observed coming out of a reactor, the Sun, the atmosphere, an exploding star, the surroundings of a black hole four billion light-years away and the disk of our own galaxy, and it has picked up five Nobel Prizes along the way.

The two bug reports are the part I will take back to work. Davis’s number sat at one third for thirty years while people looked for the error in his chemistry or in Bahcall’s Sun. Kajita spent a year looking for the error in his own code. Both of them were right, and the error was in the specification that everybody trusted. When the test is correct and the code is correct and they still disagree, the possibility that remains deserves a hearing, however senior the specification.

Kajita and Halzen both talk about luck. Kajita, in his Nobel lecture: “I feel that I have been extremely lucky.” Halzen, eleven years later, on what science takes: “a combination of perseverance, talent and luck.” I notice that this kind of luck visits people who have spent a year checking their own software, or twenty-two years pointing at an ice sheet.

While you were reading this, something like a hundred thousand trillion neutrinos passed through you. You didn’t feel a thing. Neither did they.

↑ Back to the beginning

Notes and sources

A few places where I simplified, and where the numbers come from.

Fine print for physicists (optional technical details)

• “For every 100 from above, 54 from below” is my conversion of the published up-down asymmetry, (U − D)/(U + D) = −0.296 ± 0.048 for multi-GeV muon-like events, into a ratio.
• The code uses the two-flavor vacuum formula with full mixing. One mass difference only proves that one mass is not zero. There is a second, smaller difference, measured with solar and reactor neutrinos, and the two together show that at least two of the three masses are not zero.
• For the energetic solar neutrinos that SNO counted, most of the flavor change happens inside the Sun’s dense matter (the MSW effect) rather than on the way to Earth. The bookkeeping in the text is still right: about a third arrive as electron neutrinos.
• SNO’s three numbers are for boron-8 neutrinos, the energetic tail of what the Sun emits, in units of 10⁶ per cm² per second. The prediction is 5.05 with an uncertainty of +1.01 and −0.81.
• Kajita’s four numbers are those of the paper the Kamiokande group published in 1988.
• KATRIN’s limit, below 0.45 eV at 90 percent confidence, is on the effective mass of the electron neutrino. The “at least 0.05 eV” comes from the larger mass splitting, about 2.5 × 10⁻³ eV². The limit from cosmology is on the sum of the three masses, currently below about 0.12 eV.
• The Nobel press release counts twelve supernova neutrinos in Kamiokande; the collaboration’s own papers list eleven. I wrote “about a dozen.” The same press release credits Davis with about 2,000 captured neutrinos in thirty years, which is where “a dozen or two” per extraction comes from.
• Reines and Cowan detected antineutrinos, strictly speaking, and antineutrinos are also most of what the potassium in your body emits.
• Three sums are mine. A 70 kg body holds about 140 g of potassium and with it roughly 4,000 decays of potassium-40 a second, each of which emits one neutrino or antineutrino. 220 PeV is 0.035 joules, which is a 2.7 g ping-pong ball falling 1.3 meters. And 100 trillion a second over a twenty-minute read is about 10¹⁷.

The 2026 prize

• The Nobel Prize in Physics 2026, for the citation.
• IceCube’s announcement, for Halzen’s statement and the size of the collaboration, and the University of Wisconsin–Madison’s.
• Fermilab’s summary of the Academy’s press release, for 1988 and for Mark Pearce’s words.
• Halzen’s spoken remarks are from the Associated Press report, which covers both his phone call to Stockholm and the news conference in Madison, and from the Reuters report.
• Sky & Telescope, for “first single laureate since 1992,” and the American Institute of Physics for the biography.

The 2015 prize

• The Nobel Prize in Physics 2015: press release and its popular account, “The chameleons of space”.
• Takaaki Kajita’s Nobel lecture, for the year of debugging, and Arthur McDonald’s, for SNO and Herb Chen.
• Super-Kamiokande, “Evidence for oscillation of atmospheric neutrinos” (1998).
• SNO, “Direct evidence for neutrino flavor transformation from neutral-current interactions” (2002).
• Super-Kamiokande, the Kamioka Observatory and Yoji Totsuka on Wikipedia, for the detectors, the sharp and fuzzy rings, and the 2001 accident.
• The 2016 Breakthrough Prize in Fundamental Physics, awarded in November 2015.

The early history

• Pauli’s letter of December 4, 1930, in English, and Symmetry magazine on its story.
• Bethe and Peierls, “The ‘Neutrino’”, Nature 133, 532 (1934), as described in Samoil Bilenky’s “Neutrino. History of a unique particle”.
• Reines and Cowan: the bomb plan, the experiment and their 1956 message to Pauli in the CERN archive.
• The Nobel Prizes of 1988, 1995 (which Reines shared with Martin Perl, the discoverer of the tau) and 2002.
• The Homestake experiment and what became of the mine, the Sanford Underground Research Facility.

IceCube

• Christian Spiering, “History of high-energy neutrino astronomy”, for Markov, DUMAND, Baikal, the Vostok story, AMANDA’s bubbles, and Bert and Ernie.
• IceCube’s own pages: the detector, quick facts, frequently asked questions and data and computing.
• “The IceCube Neutrino Observatory: instrumentation and online systems” (2017), for the sensors, the pressure test and the uptime.
• Live Science’s photographs of the last sensor, for Sista.
• The results: 2013, TXS 0506+056, NGC 1068, the Milky Way and the 2026 upgrade.
• “Neutrino oscillations with IceCube DeepCore”, for the 25 GeV in the caption of the curve.
• KM3NeT’s 220 PeV neutrino.

Everything else

• KATRIN’s limit on the neutrino mass.
• JUNO at its July 2026 collaboration meeting, and the detector.
• Hyper-Kamiokande’s cavern.
• The Large Hadron Collider, for 6.8 TeV per proton.
• Wikipedia on the neutrino, for the solar flux and the limit from cosmology, on neutrino oscillation, for Pontecorvo and the mass splittings, and on potassium-40.