After fifteen years, pulsar timing yields evidence of cosmic gravitational wave background

Gravitational wave astronomer here! (Admittedly at the other end of the frequency spectrum – I work with LIGO / Virgo / KAGRA.) I can't overstate how excited I am for my pulsar-timing colleagues. This is a direct result of decades of hard work: starting with with scientists like Sazhin and Detweiler in the late 1970s, through Hellings and Downs in the early 1980s (of particular note for this work), all the way up to the hundreds (thousands?) of scientists who have worked on one or more of the five (!!!) pulsar-timing teams that collaborated on this result.

And I do want to take a moment to emphasize that this an international and multi-collaboration achievement. It includes independent (but cross-checked) results from NANOGrav (featured in this particular story), the Parkes Pulsar Timing Array, the European Pulsar Timing Array and Indian Pulsar Timing Array, and the Chinese Pulsar Timing Array. It perilously easy to over-streamline the language we use in referring to these results and lose sight of the fact that it really is a team effort, spanning dozens of countries, and multiple continents.

I'm also really pleased to see how positively folks tend to be reacting to the contents of the story itself. The results here are compelling evidence for a gravitational wave background – the jumbled up signal you get when you take all of the individual signatures from the universe's many (many!) supermassive black-hole binaries and overlap them. Unlike the smaller (and faster) binaries that LIGO, Virgo, and KAGRA detect, this "stochastic" (meaning "random") background that the pulsar timing arrays have observed blurs together in a way that doesn't quite have the same "wow! look at that gravitational wave!" clarity that the original LIGO result famously had. Compare the first plot from the LIGO binary-black hole detection to figure 1 from the NANOgrav result – both figures are compelling, but a layperson might have a hard time picking out the key pulsar timing results in the same way they can look at the LIGO waveform and say "oh yeah, wow! I can see the signal in the data with my naked eyes!"

That why the numbers provided in the papers are so useful, even if they can sometimes be difficult to explain to a non-expert audience. For example, the NANOgrav paper tests a few different models (including the aforementioned Hellings-Downs pattern that's expected to describe these types of backgrounds). They measure the statistical significance of those tests in two different ways:

• the "Bayesean" way (which in the crudest sense measures the ratio of evidence for a model to the evidence against a model), and...
• the "Frequentist" way (which results in the usual σ-measures of statistical significance you might be familiar with, e.g. 3σ, 4σ, 5σ, etc).

NANOgrav's "Bayes factors" range from a few hundred all the way up to a hundred-trillion (for context, a Bayes factor of ~100 or so is usually considered "decisive"). This translates to a frequentist equivalent of somewhere between 3σ and 4σ (which is very strong support, though not quite the overwhelming 5σ that is sometimes taken to mean "beyond any reasonable doubt"). Each of the other collaborations claim similar levels of statistical significance in support of the gravitational wave background, ranging from ~2σ from the Parkes results, up to ~4.6σ from the Chinese Pulsar Timing Array, with the European / Indian results comfortably in between at better-than-3σ. Suffice to say, all of these measurements lie in the same general range, and they all strongly support the hypothesis that this gravitational wave background is very much real.

But the best part about pulsar timing is that the more time passes, the more data scientists get, and the better our statistics will be! I strongly suspect that subsequent analyses (e.g. using 18 years of data instead of 15 years) will only strengthen these already significant results. It's only a matter of time before this crosses the mythical 5σ threshold (which, by the way, isn't quite as essential outside of particle physics).

But as a gravitational wave astronomer who doesn't work in pulsar timing, the most exciting thing for me, personally, is that this discovery represents the opening of a brand new portion of the gravitational wave spectrum. Scientists like me who work with data from Earth-based interferometers (like LIGO, Virgo and KAGRA) study high frequency (small wavelength) gravitational waves. We'll never be able to detect supermassive black hole binaries with our detectors, because those systems evolve too slowly (low frequency) and on size-scales that are much too large (long wavelength). In the same way that radio telescopes (like the VLA) observe different astronomical phenomena than optical telescopes (like James Webb or Hubble), gravitational wave astronomy with pulsar timing arrays opens a window into a whole new family of astrophysical sources! If we can combine that with space-based detectors like LISA (which will bridge much of the frequency gap between ground-based and pulsar-timing observations), we'll have a much more complete picture of the gravitational wave cosmos than we do today! (To say nothing of the exciting multi-messenger astronomy that will happen along the way.)

Suffice to say, I agree with the other commenters here: this is an awesome result, with really cool applications of previously developed techniques, and is a testament to the ingenuity and persistence of the scientific community. I look forward to the hearing the scientists themselves discuss the results later today (tune into the NSF livestream if you're interested) and can't wait to sink my teeth into these results in more detail.