The Telescope in the Ice

by Mark Bowen

  They flew to meet the boat somewhere on the coast of the Red Sea, boarded, and headed north. But just as they were entering the Suez Canal, the first Gulf War broke out and they had to turn around. They flew back to Germany to wait for the Sonne to round the Cape of Good Hope and enter the Atlantic, then rejoined it and attempted to deploy their string somewhere in the vicinity of the Canary Islands. The expedition was not a success. They got a first taste of the hard lessons DUMAND had learned all too well about ocean deployments, and Bosetti drank heavily and fought constantly with little Julia’s mother.

  Since Bosetti was not a particularly supportive mentor, Christopher worked out a complicated arrangement where he joined the DUMAND collaboration, moved to Zeuthen, wrote a thesis related to DUMAND with Christian Spiering as his adviser, and submitted the thesis to Aachen for his doctorate. At Zeuthen, he also got involved in the Baikal project, quite naturally, so he carried both its DNA and DUMAND’s into AMANDA. His specialties were Monte Carlo simulation and data analysis, and he was very much into the details.

  The analysis tools that Christopher brought to the table, including some of the exact software Baikal was using at the time, were folded in with John’s and Serap’s. Sometime in 1996, he, John, and a Zeuthen grad student named Stephan Hundertmark embarked on an intercontinental search for up-going muons that they named MFH for “Make Francis Happy.” The project did not live up to its name.

  * * *

  Francis’s immediate concern was a SAGENAP review that was scheduled for early March 1997. He invited Christopher to come to Madison for two or three weeks in February and asked him to work with Serap on yet a fourth search for up-going neutrino candidates. (John J. was in the process of moving to California.) They found a few, but couldn’t be sure that they weren’t down-going muons masquerading as up-going, because about 10 percent of this million-to-one-fold background was still sneaking through their cuts. Ever the loving mother, Francis did some hand calculations to convince himself that they were up-going and presented them to SAGENAP, but the committee was not convinced. They pronounced them “a bit ‘too hot off the press’” and suggested the collaboration strengthen its abilities on the analysis side, even if it meant bringing in new people. They were very impressed, on the other hand, with progress on the construction side, and recommended not only continued but increased funding. In that sense the review could not have gone better, though it did nothing to ease Francis’s growing concern.

  * * *

  As was often the case, the dark horse in the race to find the first neutrino (and you can bet it was a race) was the Swedish contingent. They even managed to get a jump on the other institutions by volunteering one of their members to carry the tapes from AMANDA-B4’s only year of operation, 1996, directly to Stockholm from Pole. (Of course they shared the data, but it took some pestering to get it from them.)

  It happened that Adam Bouchta, the grad student who had blanched in Venice when Francis had told him about the bubbles, was in a sweet spot in his career just then, since it had finally come time for him to write his thesis. This meant he had time to focus deeply on a project of relatively wide scope, and he and his adviser, Per Olof Hulth, chose to look at the B4 data. Sometime in the spring or summer of 1997, Adam found the very first credible up-going muon tracks ever detected by AMANDA—two of them. While they would never be called gold-plated, they were the first indication that it might be possible to build a neutrino telescope in Antarctic ice.

  This relieved Francis greatly, but it was far too early to broadcast the news. In fact, just what, and how little, Adam’s discovery actually proved helps underscore the immense challenge of neutrino astronomy. For one thing, there is virtually no doubt that the neutrinos (if such they were) that had given birth to his up-going muons did not come from outer space. They had almost certainly been created in cosmic ray interactions on the far, that is, northern, side of the planet. They were atmospheric neutrinos, which are just another form of background, similar to the manmade light pollution that makes it harder to see stars in the night sky. A true telescope would be capable of distinguishing astrophysical neutrinos, which originate beyond the northern atmosphere, from the atmospheric background.

  Second, Adam was flying without a net. He had picked out his events by eye and hadn’t had a chance to run them through the sanity check of Monte Carlo simulation. Like the ones Francis had shown to SAGENAP, they might have been mis-reconstructed down-going muons.

  A lot was known about atmospheric neutrinos, since cosmic ray physicists had been studying them ever since Menon and Reines had first detected them in 1965. Their average flux, for instance—how many passed through a square meter of the Earth’s surface every second—and their energy spectrum were pretty well nailed down. This meant that they could be used as a “test beam” for figuring out how well AMANDA was working and for improving the instrument, and the best way to do that was by running Monte Carlos. Remarkably minute details can be investigated in this way, but the first and broadest check is simply to run a simulated atmospheric neutrino beam through a simulated instrument and see if your real instrument is detecting the expected number of neutrinos. Gary Hill was working on this problem at precisely this time during his spare moments under the aurorae at Pole.

  The Baikal collaboration demonstrated their skill in this regard at the International Cosmic Ray Conference that took place in Durban, South Africa, in the summer of 1996. They presented the three gold-plated events they had obtained with their four-string detector and showed that they matched up well with a Monte Carlo prediction of 2.3 events. There were six Amandroids from Zeuthen among the authors of this paper, including Albrecht Karle, Christian Spiering, and Christopher Wiebusch, so it’s a bit surprising that they were outpaced by the Swedes on the AMANDA side. Maybe Baikal distracted them.

  Again, Adam’s events were credible, but not gold-plated. Mostly they whetted the appetite—especially since, by the time he presented them to the full collaboration at a meeting in Berkeley in the fall of 1997, the data from their new ten-string array, AMANDA-B10, the largest in the world, was soon to be released.

  * * *

  “In my mind, ’98 was when it really happened, and it was messy, as I guess these things usually are,” writes John Jacobsen. “There was a lot of back and forth discussion about how to archive, filter and distribute the fairly large B10 data set—it was more data than we’d dealt with before as a collaboration.”

  The “back and forth” overheated at times, as the emotions aroused by the possibility of a discovery brought out the aggression in several of the harsher members of the collaboration. One of the Swedes, for example, publicly abused John in collaboration-wide e-mails while they were hashing out the details of how to run the data through the long processing chain.

  The data came north in the form of 246 digital tapes the size of cigarette packs, comprising about 500 gigabytes of information all told—a lot for those days. Once tempers cooled, it was decided that the Swedes would supply the algorithms and code for the first level of filtering and John, who had taken a half-time position at Lawrence Berkeley Laboratory (so he could spend the rest of his time making art), would load the data and software onto a massively parallel array of Cray T3E supercomputers at the National Energy Research Scientific Computing Center at the Lab—one of the most powerful such arrays in the world. (AMANDA and IceCube have been near the cutting edge in the crunching of large datasets ever since. Many of the students and post-docs who enter industry after working on the project take jobs writing code.)

  There was a convenient “tape robot” connected to the Crays, which could be used to upload a few tapes at a time automatically, but each tape took several hours to read, and the entire collection took somewhere around a month. Two grad students at UC Berkeley, along with a post-doc named Kurt Woschnagg, platooned on that tedious task.

  Kurt is a tall, strong, upbeat individual, with pretty much the perfect background for working in AMANDA. He was born in
Sweden to German and Austrian parents, and everyone in Sweden learned English in grade school, so he spoke all three of the major languages in the collaboration. He had obtained his doctorate at Uppsala under a woman named Olga Botner, who was married to a member of the collaboration named Allan Hallgren. (Olga would join in the middle of 1998.) By the time Kurt finished graduate school, he was tired of the constrained and competitive atmosphere at CERN, where he had done his research, and was thinking of leaving physics altogether; but just then an unusual scholarship opened up, which paid for students from Uppsala or Stockholm to work in Berkeley. The scholarships had nothing to do with AMANDA, all he needed was a host, and Buford Price was more than happy to have someone work for him for free. Kurt arrived in California in April 1996. During his first conversation with his new boss, he recalls, Buford “immediately started talking to me, hah, hah, about the properties of the ice. And so … I got roped into that, and that has lasted for many years.” Kurt went to Pole in his very first year to help deploy AMANDA-B10 and has been one of the mainstays on and off the Ice ever since. He still works at Berkeley.

  Once the tapes had been read into the Crays, it took about two and a half months of processor time to filter the data. (Since the processors worked in parallel, the elapsed time was some fraction of that.) The end result was an enriched dataset of one hundred gigabytes that was small enough for the larger workstations in most physics departments to work with.

  * * *

  Again one is astonished at how much this small group of people, about sixty in all, was accomplishing at that point in time. Over the 1997–98 Antarctic season they made great progress on the Ice.

  This was when I learned about the project. I had met Bruce Koci on the top of the Bolivian mountain in the summer of 1997. He went from there to Tibet and then to Pole, where he proceeded to demonstrate that he had the drilling for AMANDA pretty much completely figured out. Despite major conflicts with PICO, he managed to drill three holes to 2,400 meters, 500 meters deeper than they had ever drilled before and as deep as they would need for IceCube. With that accomplishment, one could say that the drilling shifted from research to development, and the first step in that transition, as Bruce saw it, was to wean AMANDA off the habit of relying on him to make heroic efforts in the field every season. He began training other people to use the drill, while he shifted his focus to the broad aspects of the design. He and Bob Morse, his main partner in that effort, knew that a larger and more streamlined “production” drill would be needed if they were ever to build a kilometer-scale instrument.

  Bruce and Bob were not working in a vacuum. Francis had been building the case for a larger instrument at various conferences around the world for about three years, and he and John Jacobsen had already come up with a name. (“I am convinced that he came up with the name, and he’s convinced I came up with the name, so we’ll never know,” says Francis inscrutably. “So the first thing I did was go into Google and type in ‘IceCube.’ And what came up, of course, was an actor and a rapper who had already used the name.… And so then the question was whether he was going to sue us, and I said, ‘I hope so, because then I’m really famous,’ and he never did.”)

  Based on the bare hint of feasibility that had been provided by AMANDA-B4, John Lynch at NSF agreed to support an IceCube Neutrino Detector Workshop to take place in March 1998 in conjunction with an AMANDA collaboration meeting at UC Irvine. As I mentioned in the introduction, I was excluded from the collaboration meeting but allowed into the open workshop. Thus, unknowingly, I attended the very first meeting ever dedicated to IceCube.

  In one of our conversations on the periphery of the collaboration meeting, Francis told me that the most important scientific contribution was made by Serap Tilav, who had actually left Madison a year earlier for Pasadena and Caltech, to work on another project that stretches the definition of a telescope. LIGO, the Laser Interferometer Gravitational-Wave Observatory, was and is a visionary attempt to detect the gravitational waves predicted by Einstein’s general theory of relativity. Serap, who had been around since the early days, later told me that she “felt like a mother raising a child called AMANDA.” She couldn’t leave it behind; she kept working on it in Pasadena. In a similar manner to the way she had used muon air showers tagged by SPASE, the South Pole Air Shower Experiment, to determine that the deep ice in AMANDA was incredibly clear, she now used muon showers to demonstrate that there were some problematic dust layers at different depths in the instrument. This was relatively bad news, but not a showstopper, and it again took a while for the rest of the collaboration to come around. As before, she was eventually proven correct.

  She also eventually realized that she could not leave her child behind. She returned to IceCube in 2001 and is still there.

  At the 1998 IceCube workshop, Ariel Goobar, an assistant professor from Stockholm, presented the two up-going events that Adam Bouchta had found and showed that they dovetailed well with a Monte Carlo prediction of three. Goobar did point out that they couldn’t be sure the events weren’t fakes, but be that as it may, AMANDA was upping its game.

  You may recall Francis telling me in the bar of our hotel in Irvine that I had come too late, the best part of the story was already over. About eight months later, he would happily eat those words.

  * * *

  A week or two after the workshop, John Jacobsen fired the starting gun in the race to find the first gold-plated events by releasing the enriched AMANDA-B10 dataset to the participating institutions.

  At the next collaboration meeting, which took place in Zeuthen in July, the competition was intense enough that even I noticed it. The home team, for example, led by Christopher Wiebusch and Stephan Hundertmark, placed themselves in direct competition with the Swedes by presenting an analysis of the B4 data with their own Monte Carlos, in which they discovered a third event to complement Adam’s two.

  This was before personal computers were commonplace, but this particular demographic was at least ten years ahead of the rest of the world in terms of access to the internet and addiction to computer screens. It was crucially important for the host institution to set up a room with a dozen or so workstations, and exactly where that room was and what the passwords were was always one of the first announcements at a meeting. At Zeuthen, at different times, I saw different young scientists leave the meeting room in the midst of a talk, run to the computer room, log in to their home institution’s network, run a competing analysis, make up a few transparencies, and race back to the meeting room in time to ambush the speaker. Christopher covered the most miles.

  The three strongest institutions, Madison, Zeuthen, and Sweden, were developing many tools in triplicate, ostensibly as checks on each other, but also, obviously, to position themselves in the race. Each was writing its own Monte Carlos and developing its own set of cuts and reconstruction algorithms, although they did discuss their methods openly for the most part and share many of them. Gary Hill tried to inject some sanity by arguing for separating the tasks and assigning them to different institutions, but that went nowhere.

  There was much discussion of the B10 data: how to calibrate it, strain out noise and false signals, and conduct basic tests to make sure it was sound. And this was the first time the collaboration as a whole ever addressed the question of double-blindness: it would be dangerous to use the entire year’s data to develop their methods, as this posed the risk of tuning the methods to a specific dataset and finding things that weren’t actually there. They made the major decision to split the data into odd and even days and keep the even days blind. They would develop their methods with just the odd days, freeze them, and then “open the box” to see if they got similar results with the even days.

  * * *

  Christian Spiering, the head of the Zeuthen group, had decided to use this opportunity to convene another workshop “as a step towards closer interaction between the present four projects”: Baikal, NESTOR, AMANDA, and a new one, ANTARES, which had been started
by a French group that had splintered off from NESTOR and come up with an equally strained acronym: Astronomy with a Neutrino Telescope and Abyss environmental RESearch. Christian was being perhaps unduly optimistic in calling the four projects “present,” however. NESTOR and ANTARES didn’t really exist—neither had any strings in the water—and Baikal, sadly, was twisting in the wind. In March, after ten years of work in increasingly difficult circumstances, the Russians had finally fulfilled their dream of building NT-200, their eight-string detector, but future prospects looked grim. Their country was in such dire economic straits that the government had essentially stopped supporting basic research, and Zeuthen, in consequence, had stopped infusing cash. NT-200 would limp along for another decade or so, plagued by reliability issues, but it would never grow. AMANDA had become the clear front-runner.

  The Baikal collaboration could still do science with their instrument, and several of its members attended the workshop. To save money, they had traveled by train and bus, some all the way from Irkutsk, 4,000 miles away. All but the dapper Leonid Bezrukov, who hailed from Moscow and showed up in a fine tailored suit, arrived looking disheveled. (Bezrukov, who had been Baikal’s lead experimentalist, referred to himself as an administrator and said he was no longer attached to the project. The Soviet Academy of Sciences had barely enough money to keep the lights on.) The others wore what was then the universal garb of the physicist: blue jeans and t-shirts or flannel work shirts—in their cases, threadbare. In spite of their manifest exhaustion, they presented many excellent papers, in English.