It is not a Pandora’s box that science opens; it is, rather, a treasure chest. We, humanity, can choose whether or not to take out the discoveries and use them, and for what purpose.
—JOHN SULSTON, THE COMMON THREAD
On George Hobbs’s computer at his CSIRO office near Sydney was something that appeared to be a colorful QR code but was actually a simulated data set. It was a square, perhaps a hundred blocks tall and another hundred long, the pixels yellow, gold, orange, or brown.
“This shows you the number of observing channels,” he explained, pointing at the y-axis, which was labeled “frequency.” He motioned to the x-axis and said, “This is time going across.”
He clicked “play,” and the not-QR-code scrolled leftward, and more pixelated columns came into the frame. The movie was not particularly interesting. Individual pixels changed from one color to another, but there was no bright dot or stripe or pattern of any kind that stood out.
“We can’t find any algorithm to say that there’s anything in this data set at all,” he said, and for a moment, I wasn’t sure what I was missing.
Then, a cat came into the frame.
Not a real cat, of course, but a cartoon cat made of bright yellow pixels in a sea of random brown, orange, gold, and other yellow ones. It was as plain as day, and I laughed out loud.
Hobbs smiled. “A human can easily see something in this data set, but because I was specifically asking for pulsar shapes, the computer is blind to this.”
I was reminded of a 1999 video demonstration of something called “selective inattention.” “Count how many times the players wearing white pass the basketball,” read the instructions. And so, being quite sure that you are up to the challenge, you faithfully watch, making sure not to count passes between players wearing anything other than white. After the video, you are asked about the gorilla.
The gorilla? You weren’t looking for one of those. But once you go back to the beginning of the video, you see the obvious. After that, it’s impossible not to see it.
And there’s the rub. If we want to observe things that are different from those that we are deliberately seeking, we need to be able to look for cats and gorillas. But how do you know those things are there if you’ve never seen them?
“Who knows what’s in the actual data?” Hobbs shrugged. “That’s the challenge: finding the unknown unknowns in large data sets.”
So how do you find them? How many haystacks do you have to look through to find a single needle? Or a cat?
I wandered down the hall to get part of the answer from Lawrence Toomey who, as the Parkes Radio Telescope’s data archives project officer, was the person best able to address that last question. In his office was a meter-tall stack of large paper sleeves, each about 40 centimeters (16 inches) on a side. In each sleeve was a plastic photographic plate from the European Southern Observatory. Thinner and lighter than the glass panes housed at the Harvard College Observatory, each plastic plate is a negative image of some patch of sky.
“They were going to just chuck them out,” he said with the apologetic shrug of someone caught hoarding, “and I’d never seen these sorts of things.”
So he adopted them.
Toomey described admiringly how astronomers like Henrietta Leavitt used to find transient events in the images, finally admitting, “Of course, these have all been digitized, but I just keep thinking I’m going to do something with them.”
Somewhere in that stack of plates in the corner of an unassuming office in Marsfield, Australia, there might actually be something new and interesting waiting to be discovered, but even Toomey admitted he probably won’t ever get around to looking through them. Instead, the pile serves mostly as a physical metaphor of a growing problem in transient astronomy. Too much data. Not enough time.
During the Harvard computers’ heyday at the turn of the twentieth century, a given telescope would churn out perhaps ten photographic plates per night. Digitized, a night of observing at that telescope becomes around a hundred megabytes of data, not too much different from the photo total of a twenty-first-century teenager at lunch. Data analysis back then was done manually. Computers such as Henrietta Leavitt—not Leveret—neatly recorded sizes, brightnesses, and spectral features, plotted trends, and inferred cosmic properties. Fundamental discoveries from their analyses propelled the field forward, but even with dozens of human computers scrutinizing plates six days a week, it was overwhelming.
In modern terminology, the Harvard College Observatory’s plate stacks—spanning over a century of observations—contain about ten megabytes of data per plate, give or take. There are half a million plates. The entire collection amounts to something in the neighborhood of a few terabytes of data, approximately enough to fill a generous external hard drive for a home computer.
A century later, the entire set of archived data from the Parkes Radio Telescope for the three decades beginning in 1991 amounts to a few petabytes, a word that sounds a bit like a brand of interactive toy animal. A petabyte is 1,000 terabytes, meaning that in 30 years this single radio telescope archived about a thousand times more astronomical data than the Harvard College Observatory’s global network of telescopes collected in a century. The archiving came only after the scientific wheat had been sifted from the chaff, so to speak. Toomey estimated that the team at Parkes collected twice as much data as it kept between 1991 and 2009, and in more recent times, given the finite limits on storage, Parkes dumps far more than it can keep.
Even so, there can be plenty of surprises in those archives. Just ask Maura McLaughlin.
On the door of McLaughlin’s office at West Virginia University was a small poster of the original pulsar, albeit modified to have cat profiles in place of the stacked pulses made famous on the Joy Division album cover. I hesitated entering, not sure how to feel about the fact that I was wearing a T-shirt with the exact same pulsar cats, but I was soon swept into a standard academic office adorned with stacks of papers, an enormous computer monitor, and children’s art. Our meeting would have to be cut short, she explained, because she was traveling that afternoon. What she didn’t say was that she would be presenting the prestigious Gordon Lecture at Cornell University, an event whose previous presenters include Jocelyn Bell Burnell.
McLaughlin studies neutron stars, the havoc that they wreak on the material around them, and their usefulness as timing tools to detect any ripples in the fabric of spacetime. Her journey to studying these bizarre objects began early. “I read A Brief History of Time when I was a sophomore or junior in high school,” she explained enthusiastically, “and I thought, ‘This is the coolest thing I’ve ever read.’ It was all about black holes and gravity and gravitational waves and spacetime.”
As an undergraduate at Pennsylvania State University in the 1990s, McLaughlin was intrigued by the research of a new faculty member, Alex Wolszczan, who had discovered what had been thought impossible: planets around a pulsar. The pulsar’s variable pulse had betrayed the presence of orbiting companions, which pulled the pulsar back and forth every few months, ever so slightly changing its distance and ever so slightly changing the times that we received its pulses.
This was very interesting to McLaughlin, but what really sealed the astronomical deal for her was a chance to observe some of these pulsar planets with the enormous radio telescope at the Arecibo Observatory in Puerto Rico. As she recalled her first trip there, she smiled. “It was just so, so amazing, and from that moment on, I knew I was going to study pulsars.”
What set her next project apart from the mainstream pulsar research that had gone on for three decades was an interest in finding pulsars outside our Galaxy, a lofty goal for a graduate student. The tools to find nearby pulsars within just a few thousand light-years from Earth rely on their faithful timekeeping. Sure, the signal of each pulse is virtually indistinguishable from the background radio hum of the universe, but countless perfect, faint heartbeats add up to an obvious blip. For pulsars hundreds of thousands or even millions of light-years away, though, the heartbeats are too faint, and McLaughlin knew the usual tools wouldn’t work. Instead, she tried to find pulsars that occasionally spat out a single stupendously bright pulse for no apparent reason. Automated digital filters could not be counted on to flag these, so McLaughlin had to return, at least in spirit, to the original method of discovery involving individual pulses. She had to look for the cats in the data herself.
Fortunately, there was already a mountain of data that had been obtained for something called the Parkes Multibeam Pulsar Survey. Parkes was where the data were obtained; “multibeam” indicated that it looked at several (13, specifically) small patches of sky simultaneously, and the term “pulsar survey” signified that this was a project that was going all-out. Astronomical surveys are never small undertakings, and entire software packages have been created to allow astronomers to sort through the data sometime before the next geological era dawns. The survey succeeded in its goal, single-handedly doubling the number of known pulsars, and it also succeeded in generating as a by-product far more data than anyone will ever fully analyze.
In this colossal haystack, McLaughlin would hunt nonrepeating radio needles.
She and her colleagues scoured four years of archived data from the survey, creating their own algorithms and flags along the way. All McLaughlin needed to do was sift through all the archived data to find a signal that was not only temporary, but also cosmic. That meant telling the computer to try out a fleet of different dispersion measures and then let her know if anything stood out from the background noise.
From the four years of archived data, she and her team found a grand total of 11 objects. Unfortunately, all of them are happily residing within the confines of the Milky Way Galaxy, each popping off intense radio waves for a handful of milliseconds at a time. They aren’t the things she was seeking, but that was fine. It’s not every day that a graduate student discovers a completely unknown class of object. Some of these new things repeat, but not on time scales typical of pulsars. Further data suggested that, like pulsars, these are rotating neutron stars with extreme magnetic fields. Because they are rotating radio transients, they were dubbed RRATs, pronounced “rats.”
And the galaxy must be absolutely infested with them! Out of the 86,400 seconds in a day, a given RRAT might show up for as much as one of those seconds, and it would be observed only if the telescope happened to be pointing in the right direction. Given that the Dish was looking at patches of sky less than a square degree—which is to say less than 1% of 1% of the sky—at any given time, finding any single RRAT was incredibly improbable. To find 11 of them in the archived data meant that there must be countless RRATs, far more than traditional pulsars.
Radio astronomers wondered what else the archives were hiding. “Single pulses suddenly became interesting again,” declared Simon Johnston. He added ominously, “And I’ve got stories about those that will curl your hair.”
Keen to try a new hairstyle, I went straight to the source of the controversy.
Honestly, Duncan Lorimer didn’t seem like the type to stir up trouble. “I’ve moved into administration for the time being,” he said almost apologetically as he welcomed me to his office. His roles as acting chair of the West Virginia University Physics and Astronomy Department and associate dean for research would typically be enough for anyone. “But,” he said. “I told them I had to keep my research going.”
Lorimer had initially gotten into the millisecond pulsar game in the 1990s. “At the time, there weren’t that many millisecond pulsars, and you could actually know all the names by heart,” he explained, sounding somewhat nostalgic. “I was fortunate to be involved in the discovery of 0437-4715. It could only be seen by Parkes, and it was just astoundingly bright! It was so bright, we almost flagged it as interference.”
Just over 500 light-years away, 0437-4715 is the closest millisecond pulsar and one of the most rapidly rotating, whipping around 174 times per second. As impressive as it is, this was not why I wanted to talk to Lorimer. I wanted to know about the field that opened up with the discovery of the Lorimer burst. It had been the first of an entirely new class of objects known as fast radio bursts (FRBs, sometimes pronounced “Furbies” by those in the business). This was the research that he felt compelled to continue even while wrestling with administrative duties. This was why Johnston had stories that would curl my hair.
To be fair, it was Lorimer’s undergraduate student David Narkevic who made the discovery in 2007. Bolstered by the successful hunt for RRATs, astronomers were again sifting through the Parkes archives, but this time Lorimer was specifically hoping to find any signals out of the ordinary in the general neighborhood of the Small Magellanic Cloud. Narkevic found a singular strong radio flash hiding in the 2001 Parkes data, so strong that it had saturated the detectors for a whopping five milliseconds. Forget PSR J0437-4715. This radio burst was the definition of “astoundingly bright.” There was no way of telling exactly how bright it was, but it was brighter than the maximum the sensors could take.
In and of itself, that would have made the burst a spectacular find. But what made it even more spectacular was just how much time elapsed between the earliest part of the signal—the high-frequency waves—and the lagging low-frequency waves. The dispersion measure was off the charts, putting the source far beyond the scant 200,000 light-years to the SMC. Far beyond millions of light-years.
Once again, it was time for the next verse of a familiar song. To saturate the detectors from such a distance, the Lorimer burst had to be immensely powerful, more powerful than anyone had imagined.
Astronomers scrambled to see if other bursts had been detected from that same location. Alas, Lorimer and his team reported, “No further bursts were seen in 90 hours of additional observations, which implies that it was a singular event such as a supernova or coalescence of relativistic objects.”
No problem. The universe contains countless objects, and even though they had just one FRB under their belts, astronomers estimated that there could be literally hundreds of such events each day. All they had to do was search tirelessly for them, preferably without stopping for lunch.
Sarah Burke-Spolaor was just starting her PhD in 2007 when her thesis supervisor, Matthew Bailes, swept into her office. “He basically brought the Lorimer paper to my desk and said, ‘I want you to find more of these things,’ ” she recalled. “So I wrote some software to find more.”
Easy peasy.
Once again, it was time to scour the data, but this time for a different sort of profile. Burke-Spolaor’s programs sifted through hundreds of hours of observations, flagging anything that met certain requirements. They had to be significantly above the noise threshold; they had to have particular pulse characteristics; and they had to have a dispersion measure consistent with cosmic distance.
When 16 signals passed all the tests, Burke-Spolaor and her colleagues shared an instant of triumph.
Except, there was a problem. There were 13 detectors simultaneously picking up signals from adjacent patches of sky. If something interesting were to go off in a distant galaxy, it would be visible only to the receiver beams that happened to be pointing directly at it. The Lorimer burst had triggered 3 of the 13 detectors, something not totally out of the question for a sudden bright source in space. The new bursts—whatever they were—showed up in all 13.
Nothing could be in that many places at once.
Their disappointment was palpable. The source had to be terrestrial. But why had the high-frequency waves reached the telescope before the low-frequency ones? How did the source mimic something beyond our own Galaxy while also clearly coming from Earth? And what did this mean for the Lorimer burst?
Because of their hybrid nature—seemingly cosmic, seemingly terrestrial—the 16 new signals were dubbed “perytons” after the mythical hybrid between a stag and a bird. Partly grounded, partly airborne, peryton might have been a fitting name, but it offered no solutions. Over three years had passed since the discovery of the first FRB, and astronomers were no closer to explaining it. Worse yet, all the new evidence traced the culprit squarely back to Parkes.
Operations scientist John Sarkissian held up a cockatoo feather and dropped it. As it drifted casually to the floor of the Parkes Radio Telescope control room, he began, “All the energy gathered by all the radio telescopes on planet Earth since the dawn of radio astronomy is less than the energy of that feather hitting the floor.” I looked at the feather. I wasn’t about to argue with someone who had been involved with the Dish since 1996, but it still seemed unfathomable to me. The telescope had six decades of nearly continuous observing under its belt, and it is just one of hundreds of radio telescopes on the planet. Sure, radio waves have low energies, but they add up, right?
Despite appearances, even the brightest objects in the radio universe aren’t all that bright. Places like Parkes and the Green Bank Observatory ask you to turn off electronics when you visit because the universe is indeed whispering to us, and the radio ears we have created are incredibly sensitive.* A single cell phone on the Moon, Sarkissian told me, would be one of the strongest radio signals in the sky.
Given this sensitivity, it shouldn’t be surprising that radio astronomers have to contend with countless non-astronomical sources. A passing satellite or airplane, a garage door opener, a Wi-Fi network, a cell phone, and even a robotic vacuum cleaner can interfere with radio telescope observations. Internationally, astronomy is granted certain frequency bands, but as technology marches along and as demand increases, these bands get squeezed more and more tightly. Astronomers are nearly powerless to halt the encroachment, and as a result, more radio frequency interference leaks into their observations, sometimes from unexpected sources.
With all these extraneous sources lighting up radio frequencies, it’s completely forgivable that the Lorimer burst was initially met with quite a bit of skepticism. It could have been lightning, some argued, or some manmade source that hadn’t been classified yet. They would defer judgment until more were found. When it seemed like more had been found—and found wanting—many in the astronomical community turned their backs on FRB research. When a second FRB showed up in only a single beam of the Parkes Multibeam Pulsar Survey in 2012, the reception was cool at best, suspicious at worst. After all, there were plenty of radio telescopes on the planet, but only one was spotting these odd events.
A handful of new FRB discoveries at Parkes did little to allay the skepticism. McLaughlin recalled the bitterness. “We were writing proposals for grants, and people would come back and say they’re not real. And we just pushed through it.”
Simon Johnston echoed the frustration. “There were a few influential people who put out a paper that never got published saying that FRBs couldn’t possibly be cosmological, that they were just [interference] or rubbish in the data. My view is that they set the field back five years because it was hard to get around these guys saying this.”
Finally in 2014, an FRB was spotted in the archived data from the Arecibo Observatory, and the celestial nature of FRBs was almost begrudgingly acknowledged.
That was all well and good, but it still didn’t explain the Parkes perytons. There had to be a way to disentangle the mythical beasts from the true cosmic events. So, in the spirit of true scientific investigation, a PhD student and member of the Centre of Excellence for All-Sky Astrophysics (CAASTRO), Emily Petroff, along with Burke-Spolaor, Sarkissian, and others, began gathering all the clues they could, no matter how small. Was it raining on the day a peryton was observed? What time of year was it? What time of day? In which direction was the telescope pointing at the time? Anything that would help pinpoint the source.
Gradually a picture of the guilty party began to emerge. Peryton events overwhelmingly happened around midday, specifically at lunchtime, and only when the Dish was aimed in a particular direction. As eminent Canadian astrophysicist Victoria Kaspi later quipped of the whole affair, “The cosmos ought not know when it’s lunchtime in Australia.”
Meanwhile, SUPERB (Survey for Pulsars and Extragalactic Radio Bursts) was ramping up, and with it came major advances in data processing and analysis. Signals from the telescope could finally be studied in real time before being shuffled to an off-site supercomputer. Now astronomers could see within seconds if a transient meeting certain criteria appeared.
The first and most obvious suspects for the perytons were the microwave ovens at the observatory. But no matter which microwave they used, and no matter in which direction they pointed the Dish, perytons remained stubbornly elusive. Until one day in early 2015, Sarkissian assumed the role of the hungry astronomer.
5 . . . 4 . . . 3 . . .
The microwave counted down.
2 . . .
Like someone impatient for their reheated leftovers, he popped the door open before the bell.
Three perytons appeared.
They tried different microwaves and several telescope orientations to get a more robust data set, but the answer was clear. The mythical beasts were escaping from the microwave ovens. Perytons are a cautionary tale against opening the microwave prematurely. The mechanism that generates the microwave radiation to warm your cocoa takes time—thankfully not much—to cycle down when the door opens. In that split second, it emits lower and lower frequencies until it shuts off completely. In this way, it successfully mimics the dispersion measures of objects billions of light-years away. The only thing that blew the perytons’ cover was that they appeared in all 13 fields of view simultaneously.
It is admittedly a bit awkward to discover one’s own microwave oven in the hunt for astrophysical phenomena. Scientifically speaking, though, the case of the pesky Parkes perytons is a success story. Not only did researchers resolve the issue of the terrestrial interference, but the explanation also lent more weight to the argument that the other FRBs are truly cosmic sources.
Tragically, the public doesn’t always see the value in such a finding.
At first, the news reports were largely positive, if somewhat disappointed that the signals didn’t turn out to be aliens. Petroff was invited to do a number of interviews and television spots, even appearing alongside Nobel Laureate Brian Schmidt. Not bad press for a graduate student. The frenzy then died down, and everyone assumed they could get back to work. Inexplicably, interest in the story surged weeks later, but this time with a more hostile slant.
Things got so bad for Petroff and her colleagues that CAASTRO ultimately felt compelled to issue a statement condemning the sloppy journalism, which was “damaging to the public portrayal of science” and a “gross misrepresentation of Emily’s discovery and previous research.” It went on to state that interview footage “was being used to effectively ridicule both Emily and Brian” and concluded emphatically that “the identification and description of perytons . . . does NOT invalidate previous research into FRBs.”
And what of the research into FRBs? After the peryton kerfuffle, it gathered steam. Evan Keane, Johnston, and a growing team were on a mission to pinpoint the exact location of a Furby with an army of telescopes across every wavelength regime. In 2016, they were pretty sure they had succeeded in linking an FRB to a distant galaxy. In doing so, they also seemed to answer some long-standing questions about the material between galaxies. Startlingly, the same community that had rallied to defend the peryton team from public scorn regarded the latest announcement with scorn of its own.
Simon Johnston recalled, “It was super personal stuff. We didn’t know what we were doing. We didn’t know statistics. They’d say we were pulsar people, so what did we know about anything extragalactic? Then, we were chickens because we wouldn’t go to conferences, chickens because we weren’t on social media to defend ourselves.”
Johnston gazed into the distance and added, “I haven’t done anything in FRBs since.”
The Lorimer burst had opened a Pandora’s box. If not hair-curling, it was certainly hair-pulling and, moreover, deeply distressing. A single energetic event billions of light-years away had managed to open up rifts within the astronomical community, redirect at least one career, and strain the frequently fragile relationship between scientists and the public.
Thankfully, science news was about to be dominated by one of the most colossal events in the history of the universe, an event that came with its own song and dance. Astronomers had waited a century to witness it, and yet it was seen by precisely zero telescopes.