A pulsar is one of the most reliable clocks nature makes, yet finding one can take longer than a human lifetime. The difficulty is not brightness but bookkeeping. NASA’s Fermi Gamma-ray Space Telescope catches a gamma-ray photon from a pulsar only about once every 100 million rotations, and nobody knows how many turns the star made in between. A team in Germany, Israel and Sweden has now recast that bookkeeping as a geometry problem borrowed from cryptography, and used it to find a new gamma-ray pulsar that the previous best search had missed.
The result, published on 19 November 2026 in Physical Review Letters, comes from Daniel Pook-Kolb and Colin Clark at the Max Planck Institute for Gravitational Physics in Hannover, working with Dotan Gazith and Barak Zackay of the Weizmann Institute of Science and colleagues at Chalmers University of Technology. Their quarry was PSR J1622-7204, an isolated pulsar in the southern sky that spins 261 times per second. It had been hiding in about 15 years of public Fermi data.
Pulsars are the collapsed cores of massive stars, neutron stars that sweep beams of radiation across space like a lighthouse. Fermi, launched in 2008, scans the whole sky every three hours but sees only a sprinkling of each pulsar’s photons, mixed in with background light from unrelated sources. To prove a pulse is real, astronomers must guess the spin rate, how quickly that spin is slowing, and the star’s exact position, then check whether the photon arrival times line up. That is a four-dimensional grid with millions of trials along each direction. The volunteer computing project Einstein@Home has found more than 80 pulsars this way, often at a cost of hundreds of computer-years each.

The new approach starts from a simple observation. If the guessed spin rate is right, the number of rotations between photons comes out as a whole number; if it is wrong, the counts land awkwardly between integers. Searching for the hypothesis that brings those counts closest to whole numbers turns out to be a version of the shortest vector problem: finding the shortest step between points in a regular, many-dimensional grid called a lattice. Cryptographers have spent years building fast algorithms for exactly this task, because some encryption schemes designed to resist quantum computers rely on it being hard. Gazith and Zackay adapted one of the strongest of those algorithms to pulsar timing.
Rather than feed in the whole dataset, the team worked with small samples. Each run used just 80 photons judged most likely to come from the source and took about five minutes on a single processor core. Out of 20 runs, each with a slightly different set of photons, one produced the right answer, and loading the full Fermi record then showed a very strong signal that confirmed the discovery. The previous best method had already been tried on this source and failed, despite needing the equivalent of about 600 years on one core. According to NASA, the new search is roughly a million times faster.
Speed matters because the Fermi archive is far from exhausted. The telescope has picked out almost 1,000 regions of gamma-ray sky with pulsar-like characteristics, each a potential target. Fast-spinning pulsars are prized as clocks: timed precisely, they can test general relativity, help detect low-frequency gravitational waves and, in principle, guide spacecraft. The team says it has already widened the search and is finding more.
There are limits. The method is probabilistic, and theory suggested about 300 runs would be needed for a 95 per cent chance of success, so finding the pulsar within 20 owed something to luck. It has so far been demonstrated on a single isolated pulsar. The bigger prize is pulsars in binary systems, whose orbital motion distorts the timing in ways conventional searches cannot easily handle and which demand far more computing. Clark estimates there are “probably tens or maybe even hundreds” of these hiding in the data. Showing that the lattice method can catch them is the next test, not yet a result.
Mathematics built to keep secrets safe from future computers has been turned to the opposite job, coaxing a secret out of the sky. PSR J1622-7204 has been turning 261 times a second all along. It simply took the right kind of maths to start counting.