Illustration of a rotating pulsar sweeping radiation across space. NASA/JPL-Caltech.
Science / The story

How 80 photons led astronomers to a pulsar spinning 261 times a second

A new search found a neutron star in 15 years of Fermi observations using about ten CPU-hours. The crucial step was working out how many turns the detector missed.

GeekRadar3 min read

A neutron star spins 261 times a second, yet a gamma-ray telescope can go days between catching photons from it. How do you measure a rotation you almost always miss? A team working with NASA’s Fermi telescope found a way to connect those sparse arrivals. In about ten hours of processor time, it identified a previously unknown pulsar in 15 years of observations. An earlier search had missed the same object after consuming the equivalent of roughly 600 years on one processor core. The study in Physical Review Letters

The pulsar, PSR J1622−7204, turns once about every 3.8 milliseconds. Its beam sweeps past Earth like the light from a rotating beacon. The trouble comes at the detector: the authors say a gamma-ray pulsar search typically sees only one photon for roughly every hundred million rotations. Other sources contribute photons to the same part of the sky. A sequence of detection times looks little like a clock ticking.

One year of Fermi observations above 300 million electron volts. Brighter regions send more gamma rays towards the telescope. NASA/DOE/Fermi LAT Collaboration.
One year of Fermi observations above 300 million electron volts. Brighter regions send more gamma rays towards the telescope. NASA/DOE/Fermi LAT Collaboration.

How 80 photons gave the team a lead

Suppose two photons really did arrive at the same point in the star’s rotation. Between them the star must have turned a whole number of times. Which whole number? An error of just one turn can spoil a prediction carried across years. The star’s spin gradually changes, and a small error in its position on the sky also changes how the arrival times line up. Trying every combination is where a conventional search spends its computing budget. NASA’s explanation of the timing problem

The researchers turned the missing revolution counts into a problem in geometry. A lattice is a regular arrangement of points. In this case, the points encode possible combinations of whole turns. Tools developed for cryptography can identify unusually short vectors in such a lattice. Here that property points to a combination with small timing discrepancies across the observations. It gives astronomers a candidate worth testing against the sky.

For individual trials, the team selected just 80 photons from those considered most likely to come from the source. A trial took about five minutes on one core. Once a promising pattern emerged, the researchers brought back the full Fermi dataset and checked whether the events clustered at consistent points in the rotation. The much stronger signal in that final test confirmed the discovery. The researchers’ account at the Max Planck Society

What else is hiding in the archive?

The first find is an isolated star. A pulsar orbiting a companion introduces another timing shift: as it moves towards or away from us, its pulses travel slightly different distances. An orbit whose parameters are unknown adds more possibilities to an already large search. The team wants to extend the method to these systems, potentially drawing on the volunteer computing network Einstein@Home. That remains a goal.

Fermi had already recorded the photons needed for this discovery. The result published on November 19 shows how much more information those arrival times still hold. For the next search, a promising place to point the new algorithm is a patch of sky that an older search has already looked at and passed over.