Ice Neutrinos, Helices and a Magnetic Moon

The 2026 Nobel Prize in Physics goes to Francis Halzen, a Belgium-born physicist at the University of Wisconsin–Madison. He is principal investigator of the IceCube Neutrino Observatory at the South Pole. The citation honors his "decisive contributions to the IceCube Neutrino Observatory and the discovery of high-energy neutrinos of astrophysical origin." He is the first person to win the physics prize alone since Georges Charpak in 1992.

How you catch a ghost particle

Of the 17 known fundamental particles, neutrinos are the least understood. They pass through matter almost as if it weren't there, so the only way to study them is to watch a huge volume of material for their very rare collisions with atoms.

IceCube does this with ice. Builders melted holes through a cubic kilometer of Antarctic ice using jets of hot water. They then lowered long strings of light sensors into the holes, about a mile below the surface, where the ice is very clear. Nearly every high-energy neutrino passes straight through. Once in a while, one strikes an atom in a water molecule and sets off a shower of light. From the pattern of light the sensors record, physicists work out the neutrino's energy and the direction it came from. Some of these neutrinos are thought to be accelerated near the centers of distant galaxies.

The usual story of DNA's double helix treats Photograph 51 as a stroke of luck. Rosalind Franklin took the X-ray image in May 1952. Maurice Wilkins then showed it to James Watson, and Watson saw the helix that Franklin supposedly hadn't spotted. A new archival study in the Journal of the History of Biology (published October 5) says she knew much more than that story gives her credit for.

What the archives show

Historian Alistair Sponsel of the Science History Institute and molecular biophysicist Brian Sutton of King's College London went through a newly acquired collection of papers from Franklin and her graduate student Raymond Gosling. They also searched the King's College archives. Their main finding is about how two images relate to each other.

• Photograph 49 came first. It was an exposure of a sample of DNA in its B form. In her notes on it, Franklin recorded that the structure might be helical and that the image held a lot of structural data.
• Photograph 51 was a new exposure of the same sample. She started it the same day she saw Photograph 49.

Sponsel says Franklin saw them as two different kinds of image. Photograph 49 was the working result that turned up the new data. Photograph 51 was meant as the showpiece, the sharpest and clearest version, taken with publication in mind. "She took it because she wanted to get a really refined version," he told Live Science. He reads the decision to retake it as "a sign she knew she had something really important on her hands."

Franklin wrote almost nothing about Photograph 51 itself, and that silence has been used as evidence that she missed what it meant. The study argues that her interpretation is in her notes on Photograph 49.

A month after its model produced what Scientific American calls the biggest math breakthrough in two decades, a solution to the Navier-Stokes problem, OpenAI has posted 372 more results to a GitHub repository. The company says each one solves, or makes substantial progress on, a major open question in mathematics or theoretical computer science.

The claims include a solution to the four-dimensional Kakeya conjecture, improvements to some of the world's most important computer algorithms, and progress toward the Riemann hypothesis. This is a breaking story, and none of it has been peer reviewed yet.

How much can be trusted

Two different questions apply here. The first is whether the proofs are correct. Many have already been checked in Lean, a programming language that verifies each logical step of a proof by machine. The article says this makes those proofs all but certain to be correct.

The second question is whether they matter: do they contain new, important ideas, or are they mostly mash-ups of existing techniques? Mathematicians will need months to work that out. An OpenAI spokesperson said that many of the results are not yet understood even by the company's own mathematicians.

The bigger claim concerns how the results were made. OpenAI says its unreleased model produced almost all of them from a single prompt given to a single AI agent. The Navier-Stokes result, by contrast, took a swarm of 10,000 agents and millions of dollars in computing. If the claim is true, this level of mathematical power could soon be available to anyone. The spokesperson also conceded that some results may have taken several attempts. MIT mathematician Andrew Sutherland says claims of "one-shotting" problems should be treated as unverified until the model is released and others can reproduce them: "We should ask for receipts."

The Moon has no magnetic field today. Researchers have argued for decades over whether it ever had one, because Apollo samples and orbital measurements give conflicting answers. A new paper by X. Yang and colleagues at ETH Zurich, DLR and the Institute of Applied Geosciences argues that it did. Their evidence is a volcano hidden on the lunar far side.

What's under the Dewar swirl

The Dewar swirl lies northwest of the South Pole–Aitken basin. Lunar swirls are bright, brushstroke-like markings that almost always sit on top of magnetic anomalies. Dewar is unusual because it stands alone and sits over a geochemical hotspot rich in iron oxide, titanium dioxide, thorium and pyroxene. It had already been classed as a cryptomare, an ancient basalt deposit buried under impact debris. It shows both a magnetic anomaly and a gravitational one.

The team combined gravity and magnetic data from GRAIL, Lunar Prospector and Kaguya. They then used an information-theory method called variation of information to run an inversion: working backward from the measured signals to the shape of the body that produces them. The result is a dense, strongly magnetized body about 60 km wide, lying 5–9 km deep. They interpret it as a buried volcano.

That magnetized basalt also accounts for the swirl. Its field runs horizontally, so it deflects the solar wind away from the soil above it. Shielded soil weathers more slowly and stays brighter than the darker, weathered basalt around it. The authors suggest this is why some magnetic anomalies have no swirl: a vertical field funnels solar-wind particles straight into the surface, which darkens it and erases any pattern.

A team in Hillel Adesnik's lab at UC Berkeley has built a two-photon holographic mesoscope. Using only light, it can activate chosen groups of neurons in one region of the mouse cortex while recording thousands of cells in several other regions, at close to single-cell resolution. The work appeared in August in Nature Neuroscience.

How it reads and writes

The "write" side uses optogenetics: neurons are engineered to fire when hit with light. Spatial light modulators (SLMs) shape the laser into precise 3D holographic patterns, so only the targeted cells are activated. The "read" side uses genetically encoded indicator proteins that glow when neurons fire.

The obstacle was size. Holographic stimulation covers less than about 2 mm², with the exact figure depending on the SLM's pixel count. That has kept earlier studies to local circuits, or in two 2023 papers, to neighboring areas. The team designed an optimized holography system and fitted it into an off-the-shelf mesoscope that can record across up to 5 mm². A set of fast mirrors moves the stimulation spot anywhere in that field. According to postdoc Lamiae Abdeladim, this made the area they could stimulate about ten times larger.

What they showed

The researchers stimulated neuron groups in four visual areas in turn while recording from six surrounding areas. In one test, they trained a classifier on activity evoked by striped patterns at four orientations. They then used light alone to activate V1 neurons tuned to each orientation. From the activity this produced downstream, the classifier picked the correct orientation significantly above chance. That suggests the stimulation sent visually meaningful signals from one area to the next.