Pulsar timing array

Millisecond pulsars are rapidly rotating neutron stars whose radio pulses can be timed with high precision. Astronomers observe a set of pulsars over many years and fit a timing model to each, accounting for factors such as rotation, motion and, where relevant, binary orbits. The differences between observed and predicted pulse arrival times are called timing residuals. A passing gravitational wave can alter the propagation of radio signals through spacetime, leaving very small, slowly varying deviations in these residuals.

A single pulsar's timing residuals can be affected by intrinsic rotational noise, interstellar propagation, measurement errors and reference-frame or clock uncertainties. A gravitational-wave background is expected to produce a common low-frequency signal whose correlations depend on the angular separation between pulsars. General relativity predicts a characteristic quadrupolar angular correlation, known as the Hellings–Downs pattern. Comparing many pulsar pairs helps distinguish this signature from unrelated noise and other correlated effects.

PTAs probe frequencies around the nanohertz band, far below those accessible to ground-based laser interferometers such as LIGO. Their long observing baselines make them sensitive to waves with periods of years to decades. A leading expected contributor to a stochastic background is the combined emission of many inspiralling supermassive black-hole binaries in distant galaxies. Other astrophysical or early-Universe processes could also contribute, so a measured common signal does not by itself establish a unique source population.

In 2023, several independent collaborations reported evidence for a nanohertz gravitational-wave background in pulsar-timing data. NANOGrav's 15-year analysis used observations of 68 pulsars and found correlated timing variations consistent with the predicted background pattern. The European Pulsar Timing Array's second data-release analysis examined 25 millisecond pulsars and reported different evidence strengths for its full and modern-instrument subsets; it also discussed uncertainties from instrumental and interstellar noise and the need for further characterization. Such findings are evidence built from statistical comparisons, not images of individual waves or definitive identifications of their sources. International collaboration and shared analysis are important because data from different observatories improve sky coverage and can help characterize noise.