A galaxy up close

It was an early Christmas for astronomers all over the world when the Gaia Data Processing and Analysis Consortium (DPAC) released a new batch of data from the European Space Agency’s astrometry mission called Gaia on 3 December 2020.

The mission, in space since late 2013, has revolutionised the study of the Milky Way and enabled astronomers to gain unprecedented insights into large-scale processes that have shaped the galaxy’s turbulent life over billions of years. Collisions and mergers with other galaxies, bursts of star formation that had given rise to stars including our Sun, the behaviour of the galaxy’s disc and many other details have emerged from the terabytes of data that Gaia has acquired since 2014.

“Only four or five years ago, I was still asked repeatedly by one of the senior professors here at Cambridge University, ‘What are the proper motions of the Magellanic Clouds?’,” says Floor van Leeuwen, the former leader of the Gaia data processing team at the Institute of Astronomy in Cambridge, who has worked on the mission from its conception in the early 1990s. “‘Have you measured them yet?’ ‘No, we haven’t,’ I had to admit.

“Today, we not only know the proper motions of the Magellanic Clouds with extreme accuracy, but we also know what is happening inside them,” he adds. “The progress that we have achieved thanks to Gaia in such a short period of time is simply immense.”

Magellanic Clouds are two small galaxies orbiting at the distance of more than 150,000 light years away from the centre of the Milky Way. Their proper motion is the apparent movement at which they move on the plane of the sky over time. Proper motion is one of the key parameters that Gaia measures for nearly two billion stars, which enable astronomers to create the most accurate three-dimensional map of the stars in the galaxy and play it backwards and forwards in time, revealing the past and future of this immense galactic organism.

The other parameters that Gaia measures are the stars’ positions on the sky in two dimensions, their distances from the Sun, the radial velocities at which they move either towards or away from the observer, their brightness levels and the spectra of the emitted light that indicate the stars’ composition, age and origin.

The two billion stars that Gaia sees represent only about 1 per cent of the galaxy. With the help of computer modelling, astronomers can, however, reconstruct the galaxy as a whole.

The field that has sprung up from Gaia data is called galactic archaeology, since it enables astronomers to dig through the details of the galactic past in a similar way to archaeologists sifting through the layers of an ancient archaeological site.

Before the wider scientific community can get its hands on the data and start unravelling the galaxy’s mysteries, the approximately 450 researchers and software engineers scattered all over Europe that work as part of the DPAC consortium have to put in a lot of work to create the ultraprecise stellar catalogues that Gaia has become known for. It took more than two and a half years since the previous data release to make the latest one, the Early Data Release 3 (EDR3), ready to go.

“Each star gets measured on average 75 times every five years,” explains Stefan Jordan, one of the researchers of the DPAC team based at the Centre for Astronomy of the University of Heidelberg in Germany. “That means that when we start our work, we have many trillion individual star observations. We have to process this data into a catalogue that for each of the nearly two billion stars that we measure contains only five parameters for astrometry, and then additional parameters for brightness and spectra.”

Astrometry, a discipline that involves measuring precise positions of stars and their movements, is Gaia’s key purpose. The catalogue contains two parameters for position, two parameters for proper motion and one parameter for parallax per star, Jordan explains. Parallax is the apparent annual displacement of a star that reflects the Earth’s orbit around the Sun. This effect, in essence the basis of stereoscopic vision, enables the astronomers to calculate the star’s distance from the Sun with the help of simple trigonometry.

“We are condensing a huge number of measurements per star into a few numbers,” adds Jordan. “But the quantity of the data, together with the clever analysis, is what guarantees the quality and accuracy of the resulting catalogue.”

Before each release, six powerful computing centres in Europe need tens of millions of CPU hours to sort out the thousands of gigabytes of data that pour down from the spacecraft. Every day, the mission sends to the ground about 20 gigabytes of data representing about 850 million individual star observations.

“At the moment, after roughly 2,400 days of operations, we have something like 87,000 gigabytes of compressed data downloaded from the spacecraft,” says Jordan’s colleague Michael Biermann, also from the Centre for Astronomy of the University of Heidelberg. “If you uncompress it, it’s roughly a factor of two larger.”

The processing work requires solving billions of mathematical equations with many unknowns.

“First, we have to organise the images by every individual star,” says Biermann. “That is a rather big process already and roughly takes a quarter of a million CPU hours. It’s also quite complicated when we are dealing with some of the dense areas of the galaxy where we can’t easily tell the stars apart.”

Thousands of rounds of calibrations filter out inaccuracies caused by the motions of the spacecraft or the thermal behaviour of the instruments. As the spacecraft rotates every six hours, the scientists also have to use the data to compute the exact attitude of Gaia (where the spacecraft is pointing) to ensure the stars’ positions entered into the catalogue are precise to the level of a few microarcseconds.

Even with all this processing, the catalogues are huge. According to the German Aerospace Centre (DLR), the mission’s final catalogue will comprise over a petabyte of data which, if printed on paper, would stack up to 100km high.

‘What we are doing is equivalent to measuring the width of a human hair from the distance of about 2,000km’.

EDR 3, made accessible in early December 2020, contains information about 1.8 billion stars of the Milky Way and, according to Jordan, represents a massive improvement in quality and precision compared to the previous Data Release 2 (DR2), issued in April 2018.

Thanks to these improvements, scientists have already been able for the first time to measure the acceleration of the Solar System towards the centre of the Milky Way and observe the movement of stars between the two Magellanic Clouds.

It was, however, already DR2 that unlocked the golden age of galactic archaeology. On average five scientific papers based on the DR2 catalogue have been published every day since 2018, which makes Gaia the most successful space mission of all time.

“We are not producing such striking images as, for example, the Hubble Space Telescope, but scientifically, we are in the same category, and in fact often even a little bit better,” adds Jordan.

This scientific bonanza, which will probably not slow down any time soon, has supercharged the study of the Milky Way after centuries of very slow progress.

To put things into perspective, in the early 1990s astronomers knew the accurate positions of only about 8,000 stars. Improving telescope technology enabled them to keep steadily increasing the size of the stellar catalogues and improving their accuracy, but the effects of the atmosphere set firm limits on what was achievable. Things only started properly moving forward after the launch of Gaia’s predecessor, the European Space Agency’s mission called Hipparcos, in the early 1990s. Yet, Hipparcos charted the positions of only about 100,000 stars, compared to Gaia’s two billion.

“Hipparcos was built on 1980s technology,” says van Leeuwen, who was involved with that project as a young researcher. “There was very little memory on board, and the satellite had to communicate with the ground station constantly while observing. The detectors were much simpler, it couldn’t see fainter stars, and the whole mission was much less flexible.”

The step change between Hipparcos and Gaia, van Leeuwen says, was mostly due to the improvements in digital technology. Still, from today’s perspective, Gaia’s detectors and cameras might seem obsolete, though they were the state of the art of what was possible in the early 2000s when the mission was being developed. For the next few decades, however, what Gaia allows will have to be enough, as no successor is planned after it finishes its observations at some point in the mid-2020s.

“That’s always the case with these missions,” says van Leeuwen, “but there is no urgent need for a new mission. We will probably only release the final catalogue in 2030 and that will contain the full ten years of mission data. Enough to keep astronomers busy for decades to come.”

How exactly does Gaia perform its magic? The 2.3m-diameter spacecraft, with its attached 10m-diameter deployable circular sunshield that gives the satellite its trademark flying saucer look, is fitted with two telescopes that sit 106 degrees apart. These telescopes, with a focal length of 35m, are folded inside through an elaborate system of mirrors. The telescopes’ focal plane, at 2,800cm2 the largest focal plane flown in space to date, is covered with 106 tailor-made CCD detectors. These CCDs, each about 5cm by 4cm in size but only a few microns thick, together make up almost one billion pixels that enable the telescope to capture the tiniest details of the surrounding Universe.

“This is all needed in order to get to the accuracies that we require,” adds van Leeuwen. “For example, to measure the parallaxes, we use the Earth’s orbit as a baseline, but the Earth’s orbit is really tiny compared to stellar distances. What we are doing is equivalent to measuring the width of a human hair from the distance of about 2,000km.”

The spacecraft orbits around the so-called Lagrange point 2 (L2), one of the five points in the Sun-Earth system where the gravitational forces of the two bodies are in balance. From L2, some 1.5 million kilometres behind the Earth from the Sun, Gaia has a constant view with minimal thermal changes and other disruptions that could affect her observations.

The CCDs are laid out in seven rows and divided into five groups. The first two columns are sky mapper CCDs that identify bright objects in the field of view and send information to the on-board computer to choose which stars to follow. The main astrometric field consists of 62 CCDs that track the selected stars as they move across the focal plane.

“Gaia is continuously taking data and looking where the image of the star is located in the focal plane at any given moment,” says Jordan. “This gives us the position of the star in the sky as long as we know precisely how the satellite is oriented.”

The 20 gigabytes of data that the mission sends ‘home’ every day represent only about 1 per cent of the amount of data it actually acquires. As Biermann says, without Gaia’s ability to process data on board and select only the valuable to send to the ground, the mission would be impossible.

“The on-board computer predicts where a star should be observed on a CCD and then it cuts a little window around that area,” says Biermann. “We only transfer to the ground these little cut-out windows.”

The focal plane contains further CCDs that measure the brightness levels and light spectra. Together, these sensors enable what is one of the most ground-breaking astronomy space projects of all time.

After seven years in space, the spacecraft and its instruments are still in excellent condition. The mission, originally expected to end in 2019, was extended to 2022 and is set for a final extension up to 2025, by which time it will likely run out of the gas that keeps it stabilised.

For the teams of van Leeuwen, Biermann and Jordan that means no slowing down any time soon. In fact, the DPAC teams are already busy working on the next batch of data that will be made public in 2022.