13 mins read 23 Sep 2020

Astronomy’s Other Sense: Gravitational Waves

Gravitational-wave astronomers are making rapid progress in their study of the Universe, powered by new detector hardware and sophisticated computer models. With discoveries of merging neutron stars and black holes now commonplace, when will they find the first gravitational wave signal from supermassive binary black holes - and how long until they unravel the secrets of the Big Bang?

A still image from a numerical simulation of two black holes that inspiral and merge, emitting gravitational waves. The black holes have large and nearly equal masses, with one only 3% more massive than the other. The simulated gravitational wave signal is consistent with the recent GW190521 observation made by the LIGO and Virgo. Credit: N. Fischer et al. (Max Planck Institute for Gravitational Physics)/Simulating eXtreme Spacetimes (SXS) Collaboration

Looking for gravitational waves is hard. Our perception of the universe has relied on optical astronomy for centuries, and only recently has gravitational-wave astronomy become possible. The small distortions in space-time (that are no bigger than a fraction of the diameter of a proton by the time they reach Earth), created by the violent mergers of massive, distant objects, mean that scientists need to go to extraordinary lengths to be certain that the signals we attribute to these events are not something more mundane like a passing train or a farmer ploughing a field.

The first direct detection of gravitational waves came in 2015 courtesy of the most accurate ruler ever made, but LIGO, the Laser Interferometer Gravitational-Wave Observatory whose two detectors picked up the signal, had been attempting this enormous science undertaking since 2002.

It took an upgrade accompanied by a 10-fold increase in sensitivity for the two merging black holes over a billion light years away to be detected on Earth, but since that time there have been numerous detections culminating in the recent discovery of GW190521, a signal from the most massive black hole merger yet.

The merging black holes that formed the GW190521 event. This was the first time that intermediate-mass merging black holes had been detected. Credit: LIGO/Caltech/MIT/R. Hurt (IPAC).

Identifying a gravitational wave signal is only the first step in making a legitimate scientific discovery. Once the signal is confirmed, supercomputers around the world begin crunching the data, looking to classify the progenitors by mass and rotation rate. The gravitational wave signal that would be seen on Earth is computed for countless combinations of progenitor parameters, and only when one closely matching the actual signal is found can scientists say with any confidence what it was that caused the cataclysmic event that has been detected.

Just like light, gravitational waves exist on a spectrum, and observatories must be built to detect a specific range of frequencies. Neutron stars are at the high frequency end of the spectrum, while colliding black holes have somewhat lower frequencies.

Scientists deciphering signals from GW190521 were able to identify the two progenitors as black holes, but found that they had unexpectedly large masses. Perhaps they were the products of other, smaller, black hole collisions eons ago, or maybe they are the results of some as yet unknown physical processes.

By studying the evolution of stars scientists can better understand how they end up as compact remnants, and this is an important part of predicting what the gravitational waves will look like when the objects collide. This type of research then guides the design and construction of the next generation of gravitational wave detectors.


The evolution of stars from the main sequence to a stellar remnant. Massive stars fuse heavier and heavier elements until exploding as a supernova and spreading those elements throughout space. Credit: NASA/ESA/L. Hustak (STScI)

Modelling the evolution of stars is typically done using computer code to determine stellar structure at any point in time. That involves solving differential equations for mass, momentum, energy conservation, and energy generation and transport for each star, and needs to be sophisticated enough to include physical processes that can be unique to stars of different masses.

As an example, convection and rotation become particularly important in more massive stars, and so they require more sophisticated modelling methods. The problem is that all of this sophistication is computationally expensive, and it can take several hours to compute the evolution of just a single star. This quickly becomes impractical when modelling stars in dense globular star clusters, precisely the environments where you might expect stellar remnant mergers to be taking place.

To address this problem, a team of scientists led by the ARC Centre of Excellence for Gravitational Wave Discovery (OzGrav) developed a stellar evolution code called METhod of Interpolation for Single Star Evolution (METISSE).

METISSE is able to evolve 10,000 stars in under 3 minutes and can use different sets of stellar models to predict the properties of stars. It does this using interpolation to quickly calculate the properties of a star at any instant, and can simulate stars from their time on what is known as the main sequence to the ends of their lives as stellar remnants.

It is a natural successor to the Single Star Evolution (SSE) package that has been in use for about 20 years, and which is unable to incorporate the latest understanding of the evolution of stars along the main sequence. Lead researcher and PhD candidate Poojan Agrawal from OzGrav and Swinburne University of Technology in Melbourne explained how METISSE has improved upon SSE.

"METISSE is an improvement over the method of defining fitting formulae to the stellar models and the code package SSE. SSE is known for its speed plus ease of use and has been a popular choice for computing stellar parameters in population synthesis codes – codes that can simulate the evolution of millions of stars while also taking into account the different interactions between them."

"SSE, however, can only make use of stellar data that was computed in 1998 which is a bit outdated now. METISSE, on the other hand, calculates stellar parameters on the fly, and can easily make use of any up-to-date stellar data with improved speed and accuracy. METISSE will help us to test our models of stellar evolution with recent observations of stellar systems," said Poojan.

This will be of interest not only to scientists studying stellar astrophysics, but also in the ongoing hunt for gravitational waves. Unlike low mass stars similar to the Sun, how more massive stars live and die is surprisingly not all that well known.

The detection of GW190521 was evidence that massive black hole collisions are occurring often enough that we can expect regular detections, so understanding the evolution of the progenitor stars is becoming increasingly important.

In a test run of METISSE using two different sets of stellar models (one computed by Modules for Experiments in Stellar Astrophysics, or MESA, and the other by the Bonn Evolutionary Code, BEC) for stars up to 100 times the mass of the Sun it was found that the predictions of remnant masses could differ by as much as 20 solar masses. But as stellar models are refined by future surveys dedicated to the study of massive stars, METISSE, unlike SSE holds a key advantage, in that it will be able to incorporate the new data into population synthesis simulations and make predictions to pave the way for new missions.


Artistic interpretation of the binary black hole merger responsible for GW190521. In this image, spacetime, represented by a fabric on which a view of the cosmos is printed, is distorted by the GW190521 signal. The turquoise and orange grids represent the dragging effects due to the individually rotating black holes. The estimated spin axes of the black holes are indicated with the corresponding coloured arrows. The background suggests a star cluster, one of the possible environments where GW190521 could have occurred. Credit: Raúl Rubio/Virgo Valencia Group/The Virgo Collaboration

Of course, computer models are only one component in the gravitational wave discovery pipeline. The brief signal associated with GW190521, spanning just one-tenth of a second, peaked at a lower frequency than any other binary black hole merger that has been detected to date.

With both the time interval of the signal and the maximum frequency it reaches inversely proportional to the binary’s total mass, it was clear very early on that if this was a gravitational wave detection, it involved two very large black holes.

Answering questions about how and why they merged though is more difficult, in part due to the short duration of the signal. Next generation observatories will have greater sensitivities to low frequency signals meaning we could observe massive binaries in orbit for longer before mergers occur, but researchers have been able to elucidate some interesting properties of GW190521.

Monash University PhD candidate and OzGrav researcher, Isobel Romero-Shaw, led a recent study looking at the GW190521 signal, and explained some of the results.

"The results of both the LIGO and Virgo collaboration analysis and our analysis support the hypothesis that GW190521 formed in a dense stellar environment. An example of a dense stellar environment is a globular cluster: a dense, spherical collection of stars and stellar remnants, with a dark core containing many black holes that interact with each other through their gravitational pulls."

The globular cluster Omega Centauri. Globular clusters are collections of hundreds of thousands of stars bound together by gravity. The cores of globular clusters are very dense and are probably environments where mergers are likely to occur. Credit: Dylan O'Donnell

"In our analysis, we show that GW190521 is consistent with being a highly eccentric merger; the path that the two black holes traced around each other was like a very squashed circle. This kind of highly-eccentric black-hole binary can arise in environments like the cores of globular clusters because the black holes can merge so quickly after they meet – some binaries can merge in as little as four days after meeting up in such an environment!"

There are still some uncertainties about GW190521, but truly characterising massive black hole mergers will require models that incorporate all the complex physics that can change the gravitational waves they emit.

One way to go is to develop supercomputers capable of running full numerical relativity simulations including all of Einstein’s theory of general relativity, but scientists could also focus on less computationally expensive models that nonetheless contain all of the physics.

Next generation hardware running advanced detection and analysis software will give scientists the opportunity to make detections right across the gravitational-wave spectrum, from tiny distortions or ‘mountains’ on the surface of a pulsar, to the merger of a neutron star and a black hole, or even a pair of supermassive black holes locked in a slow motion dance of death.

Supermassive Binary Black Holes

Earth, surrounded by pulsars emitting beams. Measuring the timing of the pulses can be done to detect passing gravitational waves. Credit: AAS.

While the study of stars and gas in nearby galaxies has provided scientists with strong evidence of the existence of monstrous supermassive black holes inhabiting their centres, there has not yet been any conclusive and direct evidence of the binary evolution of these oversized black holes.

Our best models of the cosmos predict that large galaxies are assembled through mergers of smaller galaxies, and so scientists expect to find gravitationally bound supermassive black holes throughout the universe. Should a pair of supermassive black holes collide and merge, the event would produce low frequency gravitational waves.

However, it could take years for the long wavelength of a single one of these gravitational waves (from one wave peak to the next) to completely pass over the Earth. Such low frequency wavelengths would be completely undetectable by Earth-based gravitational wave observatories, and so scientists committed to the search have had to use other methods.

One approach is to analyse the arrival times of signals from millisecond pulsars, checking for slight changes that would be indicative of the passing of a gravitational wave. The Parkes radio telescope in New South Wales has been running one such project, known as the Parkes Pulsar Timing Array project, since 2005, however it is yet to make any promising detections. The project is part of a larger international search for low-frequency gravitational waves produced by in-spiral supermassive black hole binaries.

It may be that all we need is to look at slightly different frequencies, which can be done by adjusting the cadence of the observations, but it does appear likely that within the next few years we’ll (hopefully) be able to detect a gravitational wave background signal caused by the superposition of many gravitational waves from many mergers over a long period of time.

Blazar OJ 287 is thought to be powered by a supermassive binary black hole system. Outbursts every 12 years are caused by the 150 million solar mass companion black hole crossing the accretion disk of the central black hole which weighs in at over 18 billion solar masses. Credit: Dey et al./NASA/JPL-Caltech/Scott Sutherland

Analysing quasars is another promising method in the search for merging supermassive black holes. Quasars are extremely bright and distant objects that are believed to be powered by the accumulation of gas clouds onto supermassive black holes.

If the centre of a quasar contains two black holes orbiting each other, the orbital motion might change the gas cloud accumulation and lead to periodic variation in its brightness. The most prominent example of an object of this nature known to scientists is OJ 287, a blazar 3.5 billion light years away that puts on a spectacular light show roughly every 12 years, the regularity of which has been interpreted as a second black hole crossing the accretion disk of the first.

Dr Xingjiang Zhu, a postdoc at OzGrav and the lead investigator in some new research looking at the identification of colliding supermassive black holes, explained.

"Supermassive black holes live in the centres of galaxies. They supply energy for some extremely luminous objects called quasars by swallowing gas in their vicinity. Most galaxies contain a single black hole, and the brightness of quasars fluctuates in a random way."

"However, if a pair of black holes is spiralling around each other, the quasar brightness varies in a periodic fashion due to the binary orbital motion. Our method helps us detect periodic quasar variability and thus identify possible supermassive binary black hole systems."

"For a supermassive binary black hole, each around 1 billion solar masses, it takes years or even decades for two black holes to complete a full cycle of orbital evolution. However, available data for most quasars are less than a couple of decades long. This leads to a major difficulty in the quasar periodicity search – random quasar variability can easily mimic a periodic pattern when the data span is fewer than several signal cycles," he said.

Using statistical methods they developed for the research, Dr Zhu and his colleague and OzGrav Chief Investigator Dr Eric Thrane looked at a prominent candidate, PG1302-102, to see whether it was likely to be powered by a pair of supermassive black holes.

Their conclusion was 'probably not', as the signal was almost certainly more complex than current models predict, but the framework for their analysis could be used for unambiguous detections in the future with just a few extensions.

"Our results are showing that quasars are complicated," said Dr Thrane. "We’ll need to improve our models if we are going to use them to identify supermassive binary black holes."

And if that happens, it will help scientists to detect very long wavelength gravitational waves, by knowing where to look, and even the orbital frequencies and masses of the black holes involved, with detection sensitivities vastly improved in comparison to current all-sky blind searches.

A New Window To The Universe

Gravitational waves exist across a spectrum, much like light. This diagram shows the gravitational wave spectrum, together with potential sources and relevant detectors. Credit: Institute of Gravitational Research/University of Glasgow

Gravitational wave astronomy has provided astronomers with a view of a universe that is utterly invisible to those working with data from the electromagnetic spectrum. Unlike light that is absorbed, reflected, or refracted, gravitational waves travel the vast distances across the universe unimpeded, carrying with them information about their origins and revealing to us the causes of some of the most energetic events in the universe.

The next generation of Earth and space-based observatories will be better tuned to peer into this invisible world, and the software powering the analysis of the data will be more efficient and yet incorporate better physical models.

The oldest light that we can see from the Big Bang is from some 380,000 years after the event, but gravitational-wave astronomy will allow scientists to peer back in time to just a split second after creation.

A new understanding of the cosmos is beckoning - and multi-messenger astronomy is undoubtedly the way to get it.