Pulsar timing arrays are poised to reveal gravitational waves

In February 2016 the Laser Interferometer Gravitational-Wave Observatory (LIGO) and Virgo collaborations announced to great fanfare the first-ever direct detection of gravitational waves (see Physics Today, April 2016, page 14); two more observations have since been reported. Ground- and space-based interferometers must achieve extremely precise alignment so that length changes of less than a nanometer can be detected. With the pulsar timing arrays, by contrast, the experiment is “kindly and beautifully provided by nature,” Siemens says. The challenge lies in identifying and interpreting teensy changes—a few tens of nanoseconds—in the times of arrival of pulsar signals.

The two approaches are sensitive in different regimes. LIGO detected the death throes of stellar-mass black holes as they spiraled toward each other. The pulsar timing arrays can detect vastly more massive objects that are moving slowly. They are looking for events involving objects on the order of a billion solar masses, such as the supermassive black holes that inhabit the centers of galaxies.

Pulsar timing arrays monitor millisecond pulsars in the Milky Way. As a pulsar rotates, it emits radio waves that sweep by Earth with the period of rotation. The shorter the period, the more and sharper the incident radio-wave ticks, so the better the clock. A gravitational wave passing between a pulsar clock and an observer distorts spacetime and causes the signals to arrive either later or earlier. “That’s the fingerprint,” says EPTA member Alberto Sesana of the University of Birmingham.

The distortions due to gravitational waves are faint; if they arise from galaxy mergers, they occur with periods of decades. Data are collected for each pulsar for about a half hour every few weeks. The hundreds of thousands of pulses from a given observation are summed to extract the signal from the noise.

Extragalactic gravitational waves wash over all the pulsars in the Milky Way. Because the pulsars are independent, and each has its own timing and its own interstellar medium, the main giveaway for detecting a gravitational wave is a correlated signal between the pulsars. “Pulsars do their own thing, and it’s hard to dig out a tiny signal from a large number of sources of noise,” says Sesana. “The main way to overcome this is by timing an array of pulsars.”

Each experiment keeps tabs on around 50 pulsars. Often an individual astronomer is responsible for specific clocks. For example, NANOGrav member Maura McLaughlin of West Virginia University monitors five. Assessing the data “is not completely deterministic,” she says. “It’s a bit of an art form.”

A key complication is that gravitational waves are not the only source of correlations. “You need to mark the time of arrival exactly,” says Verbiest. To minimize timing errors, which can introduce false correlations, the observatories have their own hydrogen maser clocks. Correlations can also come from incorrectly estimating the positions of Earth and other planets in our solar system. “If the Earth is not where we expect it to be,” he says, “all pulsars are affected, but not all in the same way.” The location of the solar system’s center of mass varies by a couple hundred meters in different models, explains Siemens. At the sensitivity now reached by the timing arrays, those differences matter. “We have to account for the variations in our models,” he says.

Fortunately, because the correlations from timing errors, positioning errors, and gravitational waves take different forms—gravitational-wave correlations, which are quadrupolar, are the most complex—they can be disentangled.

Other complications arise from inhomogeneities in the interstellar medium, from pulsars slowing down as they lose energy, and from combining data from different observatories. The interstellar medium spreads out the signal in a wavelength-dependent way, such that high frequencies arrive before low frequencies, says McLaughlin. The interstellar medium introduces other noise too. “It’s important to remove those effects.”

The EPTA uses dishes at five observatories—the Radio Telescope Effelsberg in Germany, the Lovell Telescope in the UK, the Nançay Decimetric Radio Telescope in France, the Sardinia Radio Telescope in Italy, and the 14-dish Westerbork Synthesis Radio Telescope in the Netherlands. Combining raw data from the observatories before the analysis affords greater sensitivity than combining postanalysis. Although the dishes range from 64 m to 100 m in effective diameter, the combination “becomes comparable to Arecibo,” says Cees Bassa of the Netherlands Institute for Radio Astronomy. “It allows us to observe more and fainter pulsars.”

The 305-m-diameter Arecibo Observatory in Puerto Rico is the largest and therefore most sensitive of any radio telescope. (China’s Five-Hundred-Meter Aperture Spherical Radio Telescope, known as FAST, is not yet in full operation.) NANOGrav has won about 20% of observation time on it and 10% on the Robert C. Byrd Green Bank Telescope in West Virginia. The PPTA in Australia uses the venerable 64 m Parkes radio dish in New South Wales.

But Arecibo, Green Bank, and Parkes are on shaky financial ground and have been threatened with closure (see Physics Today, February 2017, page 26). For now, Parkes has external funding. And the NANOGrav collaboration is looking for private donors. “We want to buy all the available time at Arecibo and Green Bank,” says Siemens. “That would be $12 million a year. It would save both telescopes and greatly increase our sensitivity.”

Looking ahead, pulsar timing array science has been identified as a key observational project for MeerKAT, which will see first light in South Africa next year. MeerKAT, FAST, and the Square Kilometre Array “will be a game changer in our business,” says Sesana. “We’ll be able to time pulsars better and to discover more pulsars. And with more pulsars, the sensitivity to [gravitational waves] gets better.”

In 2008 the three projects teamed up to form the IPTA. The first joint data set came out last year and contained a cumulative 559 years of data amassed from 49 pulsars. “It’s hard to quantify how much sensitivity you gain,” says Verbiest, who headed the effort. “A quick and dirty estimate says we gained by a factor of two.” The combining process was “a sobering experience,” he adds. “As much as we are a tight-knit community, each group still has its own methods and habits.”

Each team also wants to be able to make claims on its own, in part to increase its chances of getting funding. Sharing data is a sticking point, says Sesana. “When you share, you open the database of interesting physics to everybody. You might have students working on projects on specific pulsars, and you want to protect your students.” Still, with scientists circulating among the projects and closer collaborations in the works for future facilities, says Verbiest, “the pulsar timing arrays are slowly growing to accept a more global approach.”

The primary targets of the pulsar timing arrays are supermassive black hole binaries. “We think they form when galaxies merge,” Sesana says. “We have circumstantial evidence and predictions, but the way to nail them and to understand them is by observing their gravitational waves.” Scientists hope to learn how frequently they form, how many they are, how they evolve, and more.

The pulsar timing array experiments may also detect “exotic” sources such as erratic gravitational waves from when the universe was undergoing inflation and from superstrings—topological defects in the universe with huge concentrations of energy that could shake up spacetime with relativistic oscillations. In any case, Sesana says, “By looking at longer wavelengths than have previously been probed, we will unveil completely different phenomena.”

Many in the community expect that the first detection won’t be from a single event but from the stochastic background created by thousands of black hole mergers. “It would look like a long-period rumble,” says McLaughlin.

The collaborations have not yet plucked a gravitational wave from their data. “But even nondetections place stringent constraints on the theory,” says Verbiest. “The data have already been useful.” There are a host of sources that might be detectable, he says. “The difficulty is that we don’t know what to expect. We have to be careful as we get sensitive to things we are not aware of. We are probing uncharted territory.”