Pulsar Timing Arrays

In June 2023, four collaborations on three continents announced the same finding within a week: spacetime itself is humming. The hum has a period of years, not milliseconds. The instrument that heard it was not a laser interferometer but a fleet of dead stars whose rotation we have tracked, in some cases, since the 1980s.

How the trick works

A millisecond pulsar is a neutron star spinning hundreds of times per second, sweeping a radio beam past Earth like a lighthouse. The fastest known, PSR J1748-2446ad, rotates 716 times per second. Their pulse arrival times are so regular they rival atomic clocks over decade timescales. PSR B1937+21, discovered in 1982, has a measured period stable to roughly one part in 10^15.

A gravitational wave passing between Earth and a pulsar stretches and compresses the intervening spacetime, shifting the pulse arrival time by tens of nanoseconds. One pulsar alone cannot distinguish this from clock drift or interstellar plasma. But a wave passing Earth affects every pulsar in the sky in a correlated pattern: pulsars 90 degrees apart on the sky get anti-correlated timing shifts; pulsars at the same sky position get correlated shifts. This signature is called the Hellings-Downs curve, predicted in 1983.

A pulsar timing array is the experiment of monitoring 50-100 millisecond pulsars for 15+ years and hunting for that specific angular correlation in the residuals.

The June 2023 announcements

Four collaborations released results within days:

NANOGrav (North America): 15 years of data, 67 pulsars, evidence at ~3.5-4 sigma.
EPTA (European Pulsar Timing Array): 25 years of data, 25 pulsars, ~3 sigma.
PPTA (Parkes, Australia): ~2 sigma.
CPTA (Chinese Pulsar Timing Array, using FAST): first major release, ~4.6 sigma local significance.

None of them individually claim discovery (5 sigma is the conventional bar). Combined, the evidence for a nanohertz stochastic gravitational-wave background is strong. The signal sits in the period range of months to decades, which corresponds to wavelengths of light-years.

What is making the hum

The leading candidate is the merging population of supermassive black holes across cosmic history. When two galaxies merge, their central black holes (millions to billions of solar masses) spiral toward each other over hundreds of millions of years. In the final pre-merger phase, lasting tens of thousands of years, they emit gravitational waves at exactly the nanohertz frequencies pulsar timing arrays are sensitive to. Summed across the observable universe, the population produces a stochastic background.

The measured amplitude is roughly consistent with this prediction, but somewhat higher than most pre-2023 models expected. Which leads to the contested part.

What is contested

The signal's amplitude is at the upper edge of supermassive-black-hole population models. Several alternative sources fit equally well:

A network of cosmic strings from a phase transition in the early universe.
A primordial gravitational-wave background from inflation, if the inflaton had unusual properties.
A first-order phase transition near the QCD scale in the early universe.
Domain walls or other topological defects.

Distinguishing these requires more data and, crucially, detecting the anisotropy of the signal. A supermassive-black-hole-merger background should be slightly clumpy (some galaxies are closer than others). A cosmological background should be smooth. With current data, both look smooth.

There is also a quieter contested question: the 3-4 sigma significance is not yet discovery-level. Systematic uncertainties in solar-system ephemerides and interstellar medium effects could be hiding real signal or producing apparent signal. The next data release, expected around 2027-2028 from the combined International Pulsar Timing Array, should settle it.

Why this has to do with other realms

LIGO detects gravitational waves at frequencies of tens to thousands of Hertz — wave periods of milliseconds. Pulsar timing arrays detect waves at periods of years. The two experiments are sensitive to entirely different sources: stellar-mass black hole mergers for LIGO, supermassive black hole mergers for pulsar timing. Future space-based detectors like LISA will fill the middle of the spectrum. Together they are opening gravitational-wave astronomy the way radio, infrared, and X-ray opened the electromagnetic spectrum across the 20th century. See concept gravitational waves.

The pulsars themselves are also probes of concept neutron stars physics, and the precision required pushes against limits set by concept interstellar medium dispersion.

An open question

If the signal is supermassive-black-hole mergers, the anisotropy map should eventually pick out individual loud sources, the way LIGO picks out individual mergers. The first "resolved" supermassive black hole binary in the pulsar timing data would be a galaxy we could point a telescope at, knowing two billion-solar-mass black holes are spiraling at its center. Which galaxy will it be?

Key sources

Agazie et al. (NANOGrav Collaboration), The NANOGrav 15-Year Data Set: Evidence for a Gravitational-Wave Background, ApJL 2023 — the announcement paper.
Hellings & Downs (1983), Upper limits on the isotropic gravitational radiation background from pulsar timing analysis, ApJL — the foundational correlation prediction.
Burke-Spolaor et al. (2019), The Astrophysics of Nanohertz Gravitational Waves, A&A Review — the pre-detection state of the field.
to verify: Sesana (2013) on expected supermassive-black-hole-merger background amplitudes.

Abhishek