A New Method Finds Hidden Pulsars a Million Times Faster Using Cryptography Math
A technique borrowed from code-breaking found a spinning dead star in public Fermi data that a 600-computer-year search had missed, in about 10 hours of computing.

An international team of scientists has developed a way to find pulsars in publicly available NASA data about a million times faster than before, opening the door to discovering many more of these rapidly spinning stellar remnants. The results were published in Physical Review Letters. Using 15 years of data from NASA's Fermi Gamma-ray Space Telescope and a cryptography-based method, the team found PSR J1622−7204, an isolated pulsar spinning 261 times per second. The search took the equivalent of 10 hours on a standard computer processor.
By comparison, the previous best method failed to find the pulsar even though it used enough computing power that a standard processor would have needed 600 years to run it. A pulsar is a rapidly spinning neutron star that sweeps beams of radiation toward Earth, much like a lighthouse. Finding them in gamma-ray data is extraordinarily hard because Fermi detects only a tiny number of photons from any given pulsar, mixed with background noise from other sources and the wider universe.
To pull out the signal, researchers must correctly guess several unknowns at once: how fast the star spins, how quickly that spin is slowing and exactly where it sits in the sky. "We don't see every rotation from the pulsar, and we can't just count the rotations to work out where the photon came from," said lead author Daniel Pook-Kolb of the Max Planck Institute for Gravitational Physics in Germany. "We have a four-dimensional grid, and we must search each direction with millions of trials in it. There's just not enough computing power on Earth to do it."
Traditionally, scientists have split Fermi data into smaller chunks and crowdsourced the computing from volunteers around the world, and even then it typically took the equivalent of hundreds of computer-years to identify one pulsar. Dotan Gazith and Barak Zackay of the Weizmann Institute of Science in Israel realized the search could be recast as a geometry problem: finding the shortest vector in a high-dimensional lattice, a regularly spaced array of points. The approach draws on algorithms developed by cryptographers trying to build, or break, encryption designed to resist future quantum computers.
"We have nothing to do with cryptography in our analysis here, but those algorithms can help us solve the problem of searching for pulsars in our data," Gazith said. Each trial run took about five minutes on a single core of a computer, easily done on a laptop. Once the lattice method flagged a candidate, the team confirmed it using the full Fermi data set, which showed a signal too strong to be chance.
The next target is harder: pulsars in binary systems, where the neutron star orbits a companion. The companion's gravity distorts the signal in ways that have made these objects essentially impossible to find with traditional methods. "We know that there are probably tens or maybe even hundreds of these pulsars in the data set that we just couldn't find," said Colin Clark, an astrophysicist at the Max Planck Institute and co-author. The team plans to combine the new algorithm with volunteer computing from the Einstein@Home project.
The work also depended on openness. Fermi, launched in 2008, scans the whole sky every three hours in gamma rays, the highest-energy form of light, and all of its data are public. "The nice thing about NASA data is it's all open," Clark said. That let researchers from outside the Fermi mission apply a wholly new idea to the same archive.





