SKA: Tooling up down under for the world's most powerful telescope
The Square Kilometre Array will peer to the edge of the universe and generate more daily data than the world's internet traffic. Learning to filter the deluge is already triggering new tools for Big Data. Elizabeth Finkel reports.
On a flat, red mulga plain in the outback of Western Australia, preparations are under way to build the most audacious telescope astronomers have ever dreamed of – the Square Kilometre Array (SKA). Next-generation telescopes usually aim to double the performance of their predecessors. The Australian arm of SKA will deliver a 168-fold leap on the best technology available today, to show us the universe as never before. It will tune into signals emitted just a million years after the Big Bang, when the universe was a sea of hydrogen gas, slowly percolating with the first galaxies. Their starlight illuminated the fledgling universe in what is referred to as the “cosmic dawn”.
“It is the last non-understood event in the history of the universe,” says Stuart Wyithe, a theoretical astrophysicist at the University of Melbourne in Australia.
Like any dream, realisation is the hard part. In 2018, when the first of 130,000 Christmas-tree-like antennae is deployed on the sandy plains of Murchison, an almost uninhabited district of 50,000 square kilometres, it will mark 28 years since its conception.
Epic battles have brought the project to this point – most famously the six-year contest between countries to host the telescope. Australia and South Africa ended up sharing the prize. The SKA’s telescope in South Africa will be built on another flat, red flat plain – the Karoo region of the North Cape. It has somewhat less lofty ambitions – its dishes will probe only halfway to the edge of the universe. Its moniker, SKA-mid, denotes the mid-range frequencies of radio waves stretched across this distance.
Australia’s SKA-low, by contrast, will tune into the low frequencies emanating from the extremities of the cosmos. Together the two telescopes will represent “the largest science facility on the planet,” says SKA director-general and radio astronomer Phil Diamond, who is based at Jodrell Bank Observatory in the UK.
The game-changing technology that will allow us to hear the whispers of newborn stars against the cacophony of the universe doesn’t involve grinding mirrors to atom-thin smoothness or constructing dishes the size of sports fields. The disruptive technology here is supercomputing.
Once SKA-low is running, it will generate more data every day than the world’s internet traffic. Dealing with this deluge is a challenge being tackled by hefty global collaborations of academia and private enterprise – and it is by no means clear how it will be solved. “It’s a scale no one has attempted before,” says Peter Quinn, a computational astrophysicist at the University of Western Australia, and director of the International Centre for Radioastronomy Research (ICRAR) in Perth.
It will generate more data every day than the world’s internet traffic.
While international mega-science projects have been tackled before – think the European Organisation for Nuclear Research (CERN), which operates the world’s largest particle accelerator, the Large Hadron Collider – when it comes to the SKA, the potential world-changing spin-offs have never been so blazingly obvious.
CERN didn’t just find the Higgs boson – computer scientist Tim Berners-Lee created the World Wide Web to manage its information sharing. Wi-Fi was the spin-off when Australian CSIRO astronomers developed ways to realign scrambled radio signals from black holes.
Mega-corporations such as CISCO, Woodside, Chevron, Rio Tinto and Google are already positioned to collaborate with SKA astronomers around the world.
A science project of this grandeur, managed across 10 countries, involving dozens of specialist technical consortia and thousands of people, is challenging enough. The question of how to divvy up the pie for construction contracts and the commercial spin-offs that follow adds a whole new, complicated layer.
But astronomers have a great track record when it comes to teasing their way through gnarly collaborations to deliver triumphs such as the Hubble Space Telescope and the Atacama Large Millimeter Array. After nearly 25 years of wrangling, the signs are that the first binding SKA treaty will be signed early this year, committing the 10 member countries – Australia, Canada, China, India, Italy, New Zealand, South Africa, Sweden, the Netherlands and the UK – to funding and contracts for the 2018 rollout.
Even with the treaty, SKA will remain a confusing beast: not one telescope but two, located in two countries, with headquarters in a third – the UK. Despite the name, neither of the Phase 1 telescopes slated for construction actually boasts a square kilometre of collecting area. That won’t be realised until Phase 2 of the project, negotiations for which have yet to begin.
Nevertheless, as the gears of the vast project slowly grind into action, Australia is bracing to host its first global mega-science project. “It will be our CERN downunder,” says CSIRO astronomer Sarah Pearce, Australia’s science representative to the SKA board. But, she adds, “don’t expect a tour. It’s here precisely because there are very few people.”
An idea takes root
The curious thing about astronomy is that telescopes, as they grow more powerful, turn into time machines. When Galileo peered at Jupiter, he saw it as it appeared some 42 minutes earlier – the time it took for its light to reach him. Hubble’s iconic image of the Horsehead Nebula in the constellation of Orion is a snapshot of how it looked 1,500 years ago.
The astronomers who conceived the SKA had their sights set way beyond the 100,000-light-year dimensions of our own galaxy. The faint signals they seek began their journey more than 13 billion years ago, just a few million years after the Big Bang.
At that point, the hot plasma of electrons and protons had cooled enough to fuse and form the simplest atom – hydrogen. Except for a slight ripple here or there, our universe was a featureless sea of it. Today, things are different – the sea is dotted with galaxies. But how did these galactic islands form? To find out requires a telescope that can look back to the rippling hydrogen sea of 13 billion years ago. “That’s why the SKA was originally called the ‘hydrogen telescope’,” Quinn says.
Those who imagined the SKA had a lust for hydrogen. Their appetite had been whet by the Very Large Array (VLA), 27 dishes lying 80km west of Socorro, in New Mexico.
Now known as the Jansky VLA, the telescope generated some of the first detailed maps of atomic hydrogen. The bond between hydrogen’s electron and proton emits a unique 21-centimetre radio wave. Because the universe is expanding, the waves emitted from outer space have stretched by the time they reach us. The futher away, the greater the stretching; hydrogen waves emanating from the edge of the universe measure 1.5m by the time they reach Earth. It’s known as the Doppler effect; on Earth, we experience it when we hear the sound of an ambulance siren deepen as it speeds away, its sound wave stretching as it goes.
In 1990, on the 10th anniversary of the VLA, the world’s radio astronomers met to celebrate one of the New Mexico facility’s crowning achievements – mapping hydrogen in nearby galaxies. Ron Ekers, an Australian former director of the array, recalls that “everyone was on a high”.
Not content to rest on their laurels, a small group of visionary astronomers wondered how far the technology could be pushed. Egged on, Peter Wilkinson from the University of Manchester in Britain pitched the idea of reaching out to galaxies at the edge of the universe. A total collecting area of one square kilometre, he figured, should do the job.
The audacity of the proposal was amazing, Quinn says: “Most telescope improvements aim for a two-to-three-fold increase; this proposal represented a 10,000-fold increase.” That figure reflected a 50-fold increase in sensitivity multiplied by a 200-fold increase in field of view. “The goal was to see a milky way at the edge of the universe,” Quinn adds – and to scour the entire southern sky.
The breakthrough technology needed to enable this leap did not lie in fancy new telescope designs, but in the explosion of computing power and techniques able to handle massive amounts of data.
The receivers themselves could be little more than antennas. Tuned to radio wavelengths, they would pick up the extra-long waves of distant hydrogen – coincidentally the same wavelength used by many FM radio stations. “This is where the early universe is broadcasting,” says Quinn. “You just can’t hear it because it’s buried in the crackle.”
The more antennae, the greater the sensitivity – hence the planned one square kilometre of collecting surface. But the antennae don’t need to be all in one spot. Indeed, the more spread out they are, the sharper the focus.
How does a forest of radio antennae figure out where in the sky a signal has come from? Interferometry, a technique developed by British and Australian radio astronomers in the 1940s, is the key. It relies on the principle that each antenna in an array receives a signal at a slightly different time. For instance, radio waves coming from the easterly part of the sky hit the eastern-most antennae earlier than those lying further west. By electronically tweaking the delay on each, the entire forest could be made to point in a particular direction of the sky.
But using interferometry to tune into signals from the edge of the universe would have required filtering astronomical amounts of data; and that was a challenge yet to be mastered.
In 2000, a SKA steering committee led by Ekers invited proposals for a home for the telescope. Five countries responded. To help their bid, some built serious prototypes known as “pathfinders”. It resulted in an astronomical bonanza. Australia built the majestic dishes of the Australian Square Kilometre Array Pathfinder (ASKAP) and the antenna forest of the Murchison Widefield Array (MWA). South Africa built the seven dishes of KAT-7 and is building the larger MeerKAT. China began work on a prototype which paved the way for the Five-hundred-metre Aperture Spherical radio Telescope (FAST), the largest single radio dish in the world.
Geography worked against some of the contestants. The Chinese site wasn’t flat enough. The joint Brazilian-Argentinian bid was let down by a turbulent ionosphere – the uppermost layer of the atmosphere – which distorted the sought-after low frequency radio waves.
By 2006, Australia and South Africa were the last countries standing. Both laid claim to vast unpopulated regions, free of radio wave interference and with relatively placid ionospheres.
The South African site’s higher elevation was in its favour. Australia, on the other hand, had an impressive track record in radio astronomy. It boasted some of the world’s first interferometers, built in the 1940s at Dover Heights south of Sydney, and the iconic CSIRO Parkes telescope, operating since 1961.
algorithms developed to pursue SKA’s goals may be the next wi-Fi.
The contest was fierce, and for good reason: SKA’s benefits clearly stretched far beyond astronomy. “The devices and algorithms developed to pursue SKA’s goals may be the next Wi-Fi, the next multi-trillion dollar technology market,” says Steven Tingay, the former director of the MWA who now leads Italy’s SKA involvement. Whichever country hosted the SKA would be at the heart of the action, attracting and training the next generation of engineers and scientists in advanced manufacturing, telecommunications and high-performance computing.
Accompanied by the sort of media attention usually reserved for a football grand final, a competitive and secretive bidding process ensued.
In May 2012, members of the SKA organisation voted to split the array between the Australian and African sites. The South African telescope would observe radio waves from 350 MHz to 14 gigahertz, enabling it to detect signals up to six billion light-years away – a still sparse chapter in the universe’s life story. It would use dishes like those of the JVLA, but dramatically increase speed and sensitivity.
Australia’s array would detect frequencies in the range of 50 to 350 MHz – ideal for detecting hydrogen signals from the edge of the universe.
Both would rely on the development of disruptive new computation techniques.
“We believe we know how to do it, but I’m not hiding the fact that it’s a challenge,” Diamond says.
Getting to the Australian site of the SKA gives the words “isolation” and “quiet” whole new meanings. First, you make your way to Perth, itself one of the most isolated cities in the world. Then it’s another one-hour flight to the 35,000-strong port town of Geraldton. From there, bump around for four dusty hours in a four-wheel-drive until finally, on the horizon, you see a succession of towering white 12-metre telescope dishes.
You have arrived at the Murchison Radio-astronomy Observatory. The 36 dishes comprise ASKAP. Despite the name, they are not the prototype for SKA-low. That honour goes to the MWA, a rather less majestic affair that lies hidden in the nearby mulga scrub: 2,048 squat, wiry antennae, resembling a swarm of giant spiders. Unlike ASKAP, the MWA has no moving parts to point to different parts of the sky. That’s because this is a software telescope. It relies on a computer to program different delays into the antennae so signals from the same patch of sky are collected at the same time.
Amid great fanfare, MWA first came online in mid-2013. According to director Randall Wayth, it has blazed the trail for SKA-low. It is tuned to receive signals from the early universe within the bandwidth of 80 to 300 MHz. It does not have the sensitivity to detect features of the cosmic dawn, but its impressive 30-degree field of view allows it to map the entire visible sky over a few nights. The Galactic and Extragalactic All-sky MWA (GLEAM) survey, for instance, mapped bubbles of ionised hydrogen gas and quasars from up to six billion light years away.
Two trail-blazing aspects of its operation are key to SKA-low. The first is that it has pioneered methods to adjust for the distorting effects of the ionosphere above Murchison. “It’s like trying to see something at the bottom of a rippling pool,” explains Wayth. “Luckily for us, it’s usually just small ripples.”
Filtering out the ripples of the ionosphere is just one step in a multipronged data-processing operation whose ultimate aim is to deliver sharp images of the ancient universe.
Another early step reduces the noise inherent in the system. The heart of every radio telescope is an onsite computer known as a correlator. Developed through a partnership with IBM and Cisco, the MWA’s correlator compares signals from each of the 2,048 antennae. Noise is random; real signals are correlated. By accepting only correlated signals, this step reduces the data to a manageable 1% of the initial deluge.
The next phase takes place off-site. An 800 km optic fibre ferries the pre-filtered data from the desert to the Pawsey Supercomputing Centre in Perth. A mirror link also takes it to collaborators at the Massachusetts Institute of Technology in Boston, and Victoria University of Wellington in New Zealand, to be used by some 35 different science projects.
Just as a human brain must process vast amounts of data into a meaningful representation of the world, these supercomputers turn the MWA radio wave signals into pictures of the universe. There are data from across different regions of the sky, and across tens of thousands of frequencies. It is sifted by setting windows to extract “cubes” of information. Like pixels on a screen, they provide an image of the universe.
The MWA’s coarse resolution means its cubes can’t produce a sharp image. SKA, with its 100-fold greater sensitivity and 40-fold increase in resolution, will provide more cubes to show us what is actually there. But in order to do that, it must solve the data deluge problem.
Scaling up to SKA
The antenna design selected for SKA is not the MWA’s squat spider, but one that resembles a pine tree – the so-called log periodic design. Different rung lengths on the tree enable it to resonate in a wide range of frequencies – from 50 to 650 MHz. (MWA typically manages 80 to 300 MHz.) SKA will deploy 130,000 of them.
But that won’t deliver the eponymous square kilometre of collecting area. The €650 million (about US$690 million) funding for phase 1 will only deliver four-tenths of that. Nevertheless, it should have the sensitivity to detect primordial galaxies across large patches of sky.
The first phase of SKA-low will churn out raw data at a daily rate greater than the world’s internet traffic; impossible to store, or for human minds to process in real time. Ingenious algorithms will be needed to sift valuable nuggets from the deluge.
The University of Cambridge leads a consortium of 23 organisations, including Perth’s ICRAR, to develop new hardware and software systems for the task. One of ICRAR’s major software contributions goes by the name DALiuGE – an acronym for “data activated logical graph engine”. It’s also a bilingual play on the word deluge: “liu” is a Chinese character meaning “flow”.
Last June, an ICRAR team successfully ran the prototype of DALiuGE on the second-most powerful supercomputer in the world, Tianhe-2, in Guangzhou, China. Next, the team hopes to test it on the most powerful, Sunway TaihuLight in Wuxi, eastern China.
The computing challenges may be huge, but it’s not the first time the global community has taken on something so big. To solve CERN’s problem of distributed processing and information sharing, its researchers ended up developing the World Wide Web. “That changed our world forever,” Quinn says. “I suspect the SKA will do the same.”
THE WORLD WIDE WEB DEVELOPED BY CERN CHANGED THE WORLD FOREVER. ‘I SUSPECT THE SKA WILL DO THE SAME.’
SKA’s rewards are already reaching beyond science into industry. Besides CISCO and IBM, other big-name collaborators on the project include British-Australian mining giant Rio Tinto, international gas and oil company Chevron, Amazon and Intel. All are highly attuned to new ways of solving their big data problems – whether it is crunching data to make images of oil, gas and mineral deposits below the ground, or finding patterns in vast databases.
Construction in the middle of nowhere
The computational challenges of the SKA are formidable; so too are those involved in building and rolling out the infrastructure in the middle of the Australian outback. It’s a perfect job for a former army tank officer.
Tom Booler has been project manager for the MWA and part of the SKA-low team since 2011. His mission is to plan the construction and deployment of 130,000 antennas in the desert – absent a local workforce, with no construction equipment and no power grid. And that’s only the first phase of SKA-low. The second, slated for the mid-2020s pending funding, will see the number of antennas swell to about a million. The scale, cost and remoteness of the site make it one of the toughest science projects ever undertaken.
Supplying power is a major hurdle. The MWA is powered by a 1.6 MW hybrid solar-diesel power station, parts of which must be shielded to stop the radio waves it creates from interfering with the telescope. Phase 1 of SKA-low will need 2.25 MW. Phase 2 will need the power supply of a small city.
Extreme weather also has to be factored in. In 2015, nearby Milly Milly Station bore a year’s worth of rain in five months. While cattle grazers welcomed it, road closures disrupted plans at the observatory. Besides sudden downpours, Booler also has to reckon with temperatures soaring over 40 ºC in the summer months – and then there’s the desert death adder.
But one thing the shire of Murchison has going for it – and a reason it won the bid for the SKA – is the quiet. Population density is extremely low – just 115 people spread over an area the size of the Netherlands. There are no mobile phone towers or radio and television transmitters. The shire is also hushed by regulations enforced by the Australian Communications and Media Authority.
Within the observatory, every appliance gets stripped of wi-fi hardware before it arrives. The observatory control centre, which houses computers that crunch data from the existing telescopes, is the radio equivalent of an airlock, with radio-wave-proof double doors and no windows.
Inside a radius of 70km around the observatory authorities can order mobile phones be turned off. Out to 260km emissions are regulated in key radio frequency ranges.
The entire area is more than six times bigger than the US National Radio Quiet Zone, home to the Green Bank radio telescope and a population running into hundreds of thousands.
The quiet zones do not extend to high altitudes, so planes communicating with air traffic control could present a problem. To tackle that issue, CSIRO researchers have begun to investigate ways to measure the interference and remove it from the telescope observations.
As difficult as building the SKA will be, coming up with the money to bankroll it is trickier. Negotiations with the 10 participating governments for the first phase have been underway since late 2015.
But there’s a new sense of ease pervading the SKA community as it looks to an April 2017 sign-off on a binding treaty known as an International Government Organisation (IGO).
Once signed, ministers of each country will have a year to ratify it. Once ratified, researchers are confident things should roll out smoothly. There is a strong precedent: CERN is governed by an IGO, with 22 member states. “It’s a well-tested model,” says Pearce, who previously worked on computing challenges for the LHC as part of a multinational collaboration.
With SKA-low expected to come online in 2021, and be fully operational in 2024 astronomers are at last allowing themselves to get excited. “Until we can put a radio telescope on the moon, it will be the greatest advance in low-frequency radio astronomy,” says Elisabeth Mills, a radio astronomer at San José State University in California. “With such a great leap in technical capabilities, the most important advances from this telescope may be in areas we cannot even currently predict or imagine.”
Back to the cosmic dawn
We know something of the first few moments after the violent birth of our universe. A split second after the Big Bang, it was a tiny mushrooming fireball, 10 billion degrees hot and filled with a plasma of frenetic charged particles.
Over the next 380,000 years, the expanding universe cooled. Charged particles – electrons and protons – lost enough of their youthful energy to bond with each other and form the first hydrogen atoms. In this more staid universe, light from the Big Bang could at last move in uninterrupted straight lines. As space continued to expand, the light waves stretched to the length of microwaves – which we see today as the cosmic microwave background.
This much we know. The next episode remains a mystery.
At 380,000 years old, the universe was a peaceful sea of hydrogen. A billion years later, most of it was gone. We know a small percentage snowballed under the influence of gravity to form stars and galaxies. But the vast intergalactic sea of hydrogen gas disappeared, reionised into a plasma of protons and electrons. The era is known as “the epoch of reionisation”.
How did this happen? It turns out there are lots of theories, and they are almost completely unconstrained by data.
Traces of dark matter laid down in the Big Bang, slightly denser than their surrounds, may have triggered the snowballing of hydrogen into stars. But why did the intergalactic hydrogen disappear?
The leading theory is ultraviolet radiation from the first hot stars stripped the surrounding hydrogen of its electrons. But there is another contender: quasars. Quasars (quasi-stellar radio sources) are among the brightest and oldest objects in the universe. They are powered by black holes; the source of their light is the radiation emitted by accelerating gases as they are sucked towards the accretion disc. Quasars can be surprisingly ancient, appearing just 770 million years after the Big Bang.
“It’s controversial, but one exciting possibility is that it was quasars that reionised the universe,” says astrophysicist Stuart Wyithe, at the University of Melbourne, who specialises in trying to recreate this unknown period of the history of the universe. The theory also suggests that massive black holes may have played a far greater role in shaping our Universe than previously thought.
In Wyithe’s computer modelling, the ancient universe resembles Swiss cheese. The cheese is neutral hydrogen and the holes are where it has been eaten away, leaving an ionised plasma. Over a period of about 300 million years, the holes grow larger until, by about a billion years after the Big Bang, there’s almost no cheese left.
SKA-low is designed to supply theorists like Wyithe with hard data. It will have the resolution and the wide angle to map the distribution of hydrogen in the early universe and trace how it changed over time. He will combine these images with those from the Hubble telescope to try and detect what’s at the centre of those cheesy holes: stars or quasars.