The Telescope in the Ice

by Mark Bowen

  “And it would be ten years away,” I responded.

  “Or more.… By the way, that’s an interesting point. I always raised this when we were, you know, begging for money. If instead of a neutrino telescope you called this now a sterile neutrino experiment, a hundred million? That’s nothing in the particle physics context. The problem was, we were a cosmic ray experiment. You know, the smallest Fermilab experiment costs many, many times that.”

  Francis, too, had been let out onto the playground. This was just one of his recent excursions into theory and phenomenology—and not just about IceCube. He still visited CERN for about one month a year, where he worked with data from the Large Hadron Collider between bicycle rides in the foothills of the Alps.

  It seems as though the Madisonian powers-that-were had been waiting until IceCube succeeded before establishing it as an official research center on the campus. When they did, however, they did it in style, establishing a fellowship in honor of John Bahcall. One of the first John Bahcall Fellows was Markus Ahlers, who’d been lead author on the “GRBs on Probation” paper. And Ahlers, Francis, Maria Gonzalez-Garcia, and two other theorists were also collaborating on another game in the IceCube playground: cosmogenic or GZK neutrinos.

  The G is none other than Kenneth Greisen, father of the Kamiokande strain of neutrino detector and so much else in cosmic ray physics. The Z is Georgii Zatsepin, who conceived of underground Cherenkov detection as early as Greisen and Moiseĭ Markov, but never bothered to write it down. The K is Vadim Kuzmin, an esteemed Russian cosmic ray theorist and cosmologist. In the 1960s, these three, along with Ven Berezinsky, the erudite Russian who later surprised the Americans in DUMAND with his knowledge of George Orwell, had realized that the interaction of ultrahigh-energy cosmic rays with the soup of photons that makes up the cosmic microwave background should produce neutrinos by the usual mechanism: a proton interacts with a photon to produce a charged pion, which decays into either a muon or an electron and one or two, in this case, ultrahigh-energy neutrinos.

  The flux and energy spectrum of GZK neutrinos should depend upon both the cosmic ray spectrum and the density of the cosmic microwave soup. Ahlers, Francis, and their theorist colleagues employed the latest data from various cosmic ray instruments to estimate that these neutrinos should have energies between about 1015 electron-volts, aka a peta-electron-volt or PeV, and about a million PeV. That’s 1021 electron-volts: ten times the energy of the Oh-My-God particle. And the peak in GZK neutrino production should occur at between a hundred or a thousand PeV. The advantage of probing such stupendous energies is that the atmospheric neutrino spectrum cuts off at a few tenths of a PeV, so you’re way past that and there should be very little background, as in the case of GRBs.

  * * *

  Aya Ishihara has been interested in the extreme universe since her graduate school days at the University of Texas at Austin, where she worked on the STAR experiment at the Relativistic Heavy Ion Collider at Brookhaven (which Jim Yeck built). The idea behind STAR (Solenoidal Tracker At RHIC) was to create a kind of mini Big Bang. In 2005, she shifted to a post-doc with Albrecht Karle in Madison, where she began searching for neutrinos at the very highest energies and the GZK variety in particular. After three years in Madison, she moved back home to Japan, where she took up a second post-doc in the IceCube group at Chiba University. Like so many of the people on this project, Aya perseveres. By 2012, she had been searching for ultrahigh-energy neutrinos unsuccessfully for about seven years.

  Over time and working on successive incarnations of the instrument, she developed simulations and fashioned a set of “cuts” aimed at picking out the kind of neutrino event she was looking for. By now, the collaboration was much more sophisticated and, for better and worse, bureaucratic than it had been in the AMANDA days. She was a member of the “diffuse-atmospheric” working group, which batted around emerging ideas and preliminary studies of neutrino populations that bombard us from all angles in the sky. Once an idea has ripened in a working group it is elevated to the “analysis call,” a phone/video conference that takes place once a week. Here, the presentations are a bit more formal and representatives from all over the collaboration have a chance to weigh in. One of the purposes is to hone the question being asked in an analysis in order to give it the best possible chance of yielding a meaningful answer, and one of the best ways of doing that from a statistical standpoint is to ask the question only once. The petitioner might do a preliminary study on a so-called burn sample, a small portion of the dataset, and if the results are promising, get permission to “un-blind” the full thing and see what she’s got. Then a paper might be written, at which point the publications committee kicks in.

  In those days, the analysis call took place at an awkward time for those who lived in the Far East, “half-past midnight in Japan or something,” says Aya with a laugh, “so it’s a kind of uniting us, how the things develop in IceCube” (another indication of her stick-to-itiveness—and her generosity: she spreads the credit around).

  Since the Earth happens to be relatively opaque to ultrahigh-energy neutrinos, Aya optimized her cuts to pick out muon neutrinos coming from directions just below the horizon, that is, just barely up-going. This way she would still have shielding from the down-going cosmic ray background, but the neutrinos would traverse a relatively short distance inside the planet and thus have a better chance of reaching the detector. She also cut on the total brightness of the event: how many photomultiplier tubes were lit up and just how lit up they were. This is a measure of the total energy of the corresponding neutrino. She placed this cut to select events above about a tenth of a PeV, the top end of the atmospheric background.

  Aya wasn’t expecting much in May 2012 when she received permission to un-blind the previous two years’ data. She’d already done this three times; it was becoming routine. However, the last set of data she had un-blinded had come from IC40, the forty-stringed instrument, and now she was dealing with IC79 and IC86, true kilometer-scale detectors. As far back as the writing of the IceCube proposal a decade earlier, Francis and several others had suspected that they would need the full detector in order to see cascades.

  Aya was searching only for muon tracks, and again she saw none. But two absolutely beautiful cascades snuck by her cuts. They snuck by in a larger sense as well, since they were down-going: they came from the southern sky. The simple algorithm she had used to determine the direction of what she assumed would be muon tracks had been confused by these two huge, globular flashes of light. It reconstructed them as traveling upward when they were actually traveling down. As is so often the case in science, this mistake was a stroke of luck. Had they been reconstructed as down-going she never would have seen them.

  The two events had such personality and were examined in such detail that rather than refer to them by their official designations (“18545/63733662” doesn’t roll off the tongue), a grad student in Madison by the name of Jakob Van Santen decided to name them after the Sesame Street duo Bert and Ernie. Bert had exploded in the detector in early August 2011, just three months after the full instrument had taken flight, and Ernie had rung in the New Year a little late, on January 3, 2012. Both were enormous. They filled significant portions of the array. They were larger than the football stadium on the campus at Madison, where football is a big deal (see photograph 32).

  So, why, after focusing on up-going events for almost three decades, were they now so excited about two down-going ones? Because they knew they were neutrinos and they suspected that they were extraterrestrial or astrophysical, the first neutrinos originating outside our solar system that anyone had detected since the miraculous supernova of 1987—and the first high-energy astrophysical neutrinos ever detected. They knew they were neutrinos for two reasons. First, IceTop, the surface array for detecting and vetoing down-going cosmic ray air showers, was silent during both events. Second, both were fully contained in the detector. This meant that the interactions that caused them had to have oc
curred inside the array and therefore had to have been initiated by neutrinos. It isn’t clear, incidentally, which flavor of neutrino causes a cascade. It could arise from the straightforward inverse beta decay of an electron or tau neutrino with a nucleon, or it could arise from a so-called neutral current interaction in which any of the three flavors interacts without losing its life. (Neutral currents are what gave Dave Cline the nickname “A.C.D.C.” back in the early 1980s, when he first thought he found them, then that he hadn’t, and finally that he actually had.)

  Both Bert and Ernie packed enough punch to make it unlikely that they were atmospheric. They were, in fact, the highest-energy neutrinos that had ever been detected. They came in at about one PeV each, roughly ten times the atmospheric cut-off. They were not energetic enough to be GZK neutrinos, so in a sense Aya had struck out again. However, she remembers stumbling across those two events as a once-in-a-lifetime experience, the kind physicists live for.

  * * *

  Isn’t it wonderful that IceCube’s first real surprise came from cascades and from “behind the telescope”?

  * * *

  It is hard in this internet age to keep an intriguing result quiet. Francis began getting e-mails right away. Being a fan of the pithy correspondence, his favorite was the four-word inquiry he got from Luis Anchordoqui, a colorful Argentinian who had collaborated with him and Markus Ahlers on several of the recent theory papers. “Is it fucking true?” asked Anchordoqui. “Yes,” Francis responded.

  The crucial question was whether Bert and Ernie were extraterrestrial. When Aya asked that question of her data, employing statistical techniques hammered out through lengthy negotiation in the analysis call, she found that there were about three chances in a thousand that the two events might simply have been unusually energetic atmospheric neutrinos. That may sound good, but it’s only 2.8 sigma, not significant enough to claim discovery—and not even significant enough to claim “evidence.” That word is used when you reach three sigma—and even then, there is an adage that half of all three sigma results are wrong!

  Francis holds the opinion, incidentally, that Aya had at least found “evidence” for and possibly even “observed” (equivalent to “discovered”) the very first high-energy extraterrestrial neutrinos. Aside from seeing Bert and Ernie as being far too energetic to be atmospheric, he realized when he inspected them in the event viewer that they were in a “different league” altogether from the most energetic atmospheric neutrinos they had seen before. Pointing out that the labored statistical reasoning that had emerged from the analysis call ignored the fact that the veto from IceTop had been silent during both events, he adds wryly that “statistics is a lot easier than physics.” If neutrinos of such high energy had been created by cosmic rays hitting the atmosphere, other particles, including muons, would almost certainly have been created along with them.

  So Francis landed in the optimistic camp, as usual. Naturally, of course, there were pessimists as well.

  * * *

  Nathan Whitehorn was the kind of graduate student that passes through a professor’s life once or twice in a career. Francis says that “if somebody is going to make it big in this experiment, I predict it’s him.”

  One of the reasons Nathan enjoyed working on the project was that “especially by comparison with these large accelerator experiments, IceCube is actually a very simple detector.” (One has the impression that it might seem simpler to him than to most people.) He liked the fact that as a student he could not only grasp all the steps in the analysis chain, from the moment light was detected to the point where he had a dataset he could do physics with, he could actually play a role in ensuring that they worked and improving them. He contributed to what you might call the “guts” of the experiment in several ways, not necessarily related to his thesis work. He wrote a software program called millipede, for example, for estimating the energy of muons passing through the detector, that has become a standard tool. (When he was working on millipede, he came across a series of papers that the DUMAND collaboration had produced when they were trying to do the same thing two decades earlier. They had never had a detector to work with, of course, but now, with IceCube built, Nathan told me, he could “do the thing that they meant to do.… You see this ghost of the other experiment.”)

  Nathan was in an excellent position to capitalize upon Aya’s discovery, since, even before she made it, he had been working on ways of using the veto power of IceTop to open the view toward the southern sky. He was also free, since he’d defended his thesis by then. He was staying on as a temporary post-doc in Madison, while his wife, a biologist, completed her thesis.

  He and Claudio Kopper, a John Bahcall Fellow in Madison who specialized in Monte Carlos, began thinking out loud together, sketching ideas on whiteboards and the like. “It was a pretty great opportunity,” remembers Claudio. “After we saw these two high-energy cascades, both Nathan and I thought, ‘Well, this is exciting and there have to be more and we have to find those.’”

  They settled upon a disarmingly simple idea.

  On a Saturday, about a month after Aya’s discovery, the two of them, independently and at home, according to Claudio, “just wrote some sample code that would filter out, like, starting tracks and cascades.” There were subtleties to it, but the basic idea was simply to pick out high-energy events that were born, as Bert and Ernie were, inside the detector. Over the next several months, the two post-docs went about presenting their “starting-track analysis” in a working group and then at the analysis call, and as it dawned on the rest of the collaboration that they had come up with a brilliant idea, the old fear of discovery kicked in and they were subjected to ever more intense scrutiny. Two of the more critical members of the collaboration were assigned to review their proposals, Spencer Klein from Lawrence Berkeley Laboratory and Christopher Wiebusch, now a professor at Aachen University, who had played a similar role fourteen years earlier when the Peacock events were discovered. Claudio and Nathan were put through their paces.

  In October 2012, when they finally got permission to un-blind, their analysis revealed twenty-six more events at energies near and above the atmospheric cut-off. Seven were muon tracks, the other nineteen were cascades, and most came from the southern sky. There were now twenty-eight in all, including Bert and Ernie. This was “evidence,” at a hair less than five sigma, for extraterrestrial neutrinos—not quite “observation,” but that would be splitting hairs.

  Since point source fever reigns in the collaboration pretty much all the time, Nathan and Claudio had brought in Madison’s point source expert early on: Naoko Kurahashi Neilson, a post-doc working with Albrecht Karle. Searching for a point source is in some sense the next step after finding the events, so the collaboration had wisely kept this threesome blind to the events’ directions by keeping them blind to the times when they occurred. (Since the Earth spins, you need the time in order to know which way IceCube was pointing.) Nevertheless, this “evidence” ramped the fear of discovery up almost to the level of the AMANDA days. Ty DeYoung, who is now a professor at Michigan State University, observes that it also sparked a comparable burst in creativity. And he should know, since he was responsible for much of the creativity the first time around.

  Claudio says that the real work started after the un-blinding, when they tried to determine the energies and directions of the events (inside the detector, that is; they were still blind to the times).

  Their most impressive accomplishment was to figure out a way to reconstruct the directions of cascades. Claudio and Jakob Van Santen, the grad student who had named Bert and Ernie, took the lead. (Albrecht was adviser to both.) What makes this task especially difficult is that the cigar-shaped shower of electrons and positrons that creates the light in a cascade is only ten or twenty meters long, nothing like the track of a muon, which can run along for more than a kilometer. And since the light detectors in IceCube are so far apart—the strings are separated by sixty meters and the detectors on a stri
ng by seventeen—you don’t have enough pixels, so to speak, to get any sort of image of the original shower; you have to wait until its light reaches your detectors. What you end up with, as in Bert and Ernie, is a nearly spherical blob of light the size of a football stadium, in which the tiny shape of the shower is completely lost.

  Claudio and Jakob discovered, amazingly, that the individual bunches of photons emitted by a shower, wave-packets they’re called, do carry information about its direction. Without going into the details, if the packets are going in the same direction as the shower, they will be deformed in one way as they travel through the ice, and if they’re “swimming upstream” against the direction of the shower, they’ll be deformed in another way. Thus directional information is carried by the shapes of the individual light pulses as they hit the detectors. This is the sort of insight that can make science seem like magic sometimes.

  It’s also where Dave Nygren’s digital optical module comes in. For it is capable of recording the shapes of the pulses and sending them to the surface of the Ice. And it turns out that Francis made an important contribution here as well. Recording and saving the shapes, or waveforms, added complexity to the DOM. It would have been much easier just to take a couple of measures of the height and shape of the pulses and send them up. So at first Dave questioned whether the waveforms ought to be saved. He remembers organizing a software workshop in Berkeley way back in 1998, when they were designing the first fifty DOMs for AMANDA’s string 18. “We tried to ask the question, are these waveforms really useful? Didn’t really know,” he says. “But Francis said, ‘Yes, the waveforms are useful,’ so the decision point was made.”