Astronomy in this energetic new frontier calls for new ways of seeing. Neutrinos are almost ideal. While intervening dust clouds and the inner regions of many cosmic accelerators absorb or scatter photons -- from visible light through the highest energy gamma rays and every wavelength between -- neutrinos are immune to both electromagnetism and the strong nuclear force; they can escape from anything except the inside of a black hole itself.

Unfortunately neutrinos are just as hard to catch as they are to stop. Robert Stokstad of Lawrence Berkeley National Laboratory's Nuclear Science Division notes that "resolving point sources of neutrinos at cosmic distances will require detectors that encompass about a cubic kilometer of ordinary water or ice."

Stokstad credits David Nygren of the Physics Division with encouraging the Lab to enter high-energy neutrino astrophysics after attending a Caltech conference in 1994. Stokstad was one of a group of scientists who formed the Km3 group that year, within the Lab's multidisciplinary Institute for Nuclear and Particle Astrophysics. Research into the requirements of an effective cosmic-neutrino "telescope" has been supported by the Department of Energy through the Lab's Directed Research and Development program, and recently through the Physics Division and the National Science Foundation as well.

Two relatively small precursors of such a kilometer-scale telescope are now operating: AMANDA in the South Polar ice and BAIKAL in Siberia's Lake Baikal. Two more, NESTOR in Greece and ANTARES in France, are under development in the Mediterranean.

All four use arrays of photomultiplier tubes, either suspended from cables or supported on underwater structures, to detect flashes of faint blue Cerenkov radiation emitted when a muon streaks past at faster than the speed of light in water.

Muons are generated in great numbers when cosmic rays strike atoms in the Earth's atmosphere above the detector. When these penetrate the ice or water above the detector they leave downward tracks. Rare, upward-going muon tracks, produced by neutrinos traveling all the way through the planet, must be distinguished from the millions-of-times more numerous down-going muons.

By accurately reconstructing a muon's track from the arrival times of its Cerenkov radiation at several modules, the neutrino that created it can be traced back to its area of origin.

"Expanding an existing detector array to a cubic-kilometer telescope will not be a straightforward scale-up," Stokstad points out. In present technology, the analog signals from a photomultiplier tube are sent by optic fiber or electrical cable to a central location for data processing. A typical AMANDA string, for example, carries more than three dozen optical modules -- photomultipliers housed in glass spheres -- spaced along the bottom half of a two-kilometer-long cable.

"By the time a copper cable carries the signal to the data processor, the electrical pulse has spread out and attenuated," says Stokstad. "Fiber optics solves this -- the signal goes in with a sharp rise and comes out that way -- but a fiber cable is expensive, and the connection to the module can deteriorate or break when the glass sphere becomes frozen in the ice."

IN BERKELEY LAB'S DIGITAL OPTICAL MODULE, THE EVACUATED GLASS SPHERE CONTAINS NOT ONLY A PHOTOMULTIPLIER TUBE TO DETECT FLASHES OF CERENKOV RADIATION (BOTTOM HALF) BUT ELECTRONICS THAT CAN BEGIN PROCESSING THE SIGNALS IN PLACE, BEFORE DATA IS SENT TO THE SURFACE (TOP HALF).

The Km3 group came up with a new architecture for processing signals in a neutrino telescope, inspired by the example of a custom chip that had been developed by Stuart Kleinfelder of the Engineering Division. Rather than sending torrents of raw data to the surface in analog form, the data processing could begin down in the ice, done by fast electronics inside the module.

The guts of the new optical module, built around a chip whose final version would be called an "analogue transient waveform digitizer," would slice the photomultiplier's output pulse into sections of one to 10 nanoseconds each, then stamp a time signal on the resulting waveform. Only if modules nearby on the string recorded similar pulses within a microsecond of one another would the pulse be considered a genuine hit and subsequently processed. Data transmission to the surface would occur only through robust digital protocols using tough, relatively inexpensive copper cable.

Stokstad explains that most individual pulses are just noise, caused in part by random emission of photoelectrons from the cathode of the photomultiplier tube. "False signals come from this 'dark current' and also, as we discovered, from radioactive decay of potassium in the glass of the sphere" -- and even from the temporary fluorescence of the plastic circuit board itself.

Since the noise is random, however, it is easily eliminated by comparing detection times of signals among modules near each other on the string. This comparison, called a "local coincidence," can be done on the surface, says Stokstad, "but with digital modules the electronics leave the noise buried in the ice."

In an early collaboration with NASA's Jet Propulsion Laboratory, the Km3 group tested the idea at AMANDA's South Pole location using two modules, which confirmed the principle by sending digitized waveforms to the surface. In 1997 the Km3 group teamed with the local AMANDA group, led by Buford Price of the University of California at Berkeley and supported by the NSF.

AMANDA was planning to deploy six strings of modules during the Antarctic summer of 1999-2000, which would bring the total number of strings to 19. In November 1998, AMANDA and the Km3 group decided to include a full string of digital optical modules, or DOMs, in this deployment schedule.

"Our initial schedule called for a prototype by February of 1999," says Stokstad, who became the DOM project manager. "We were a little optimistic."

The first prototype was finally finished in April; after intensive testing, a second was finished in late June. Remarkably, by mid-October, 50 production printed circuit boards were finished, and by the end of November the circuit boards, wiring, and photomultiplier tubes -- "potted" in clear gel -- of all 42 modules needed for Antarctica had been assembled and sealed in their spheres at AMANDA's facilities at the University of Wisconsin. The finished DOMs were on their way south almost immediately.

Stokstad delivered the first two personally, arriving December 13, two weeks ahead of the main shipment, on his first trip to the South Pole. Other members of the Km3 group were soon on hand to send the full string into the ice, including the DOM's principal designer, Jozsef Ludvig of Engineering, along with Gerald Przybylski of the Physics Division, who brought the final two modules which had been left at Berkeley Lab for further tests, and Chuck McParland of Information and Computing Sciences.

"The first tests with the actual electrical cable were only possible after arriving at the Pole. After we hooked up several modules to this cable, we decided that some adjustments were necessary to improve the digital communication," says Stokstad. "That meant all the spheres had to be unsealed and opened, the boards had to be removed, modified, and replaced, and the spheres pumped out again and vacuum-sealed" -- all while working in a cramped shack near the drill site. With long hours and steady work, the 41 modules slated for the string had been reassembled by the time the AMANDA drilling crew was ready.

TO BURY STRINGS OF DETECTORS DEEP IN THE ANTARCTIC ICE,  HOT WATER IS FORCED THROUGH A NOZZLE TO MELT A TWO-KILOMETER-DEEP SHAFT, A DRILLING PROCESS THAT TAKES THREE OR FOUR DAYS.

Hot water forced through a nozzle melts a two-kilometer-deep hole in the ice, a process that takes three or four days. The water standing in the hole freezes solid in 36 hours or less, and the whole string of modules must be in place before any part of the hole gets too narrow.

"As the string went in we hooked each module into position, made the electrical connections -- using a hot-air gun to warm the plastic -- plugged it in, booted it up, and checked that it talked," says Stokstad. "Because we also had the standard optical fiber hook-ups on the string, it took us 15 or 20 minutes for each module, but in 12 hours the whole string was in place. We made 42 modules and deployed 41 in the ice."

The ice hardened and pressure increased on the glass spheres -- the same kind used for deploying sensors at depths up to four kilometers in the open ocean -- while, Stokstad says, "we waited with bated breath to see how many would talk."

Eventually only one module failed, not because of its digital electronic components but because of a bad photomultiplier tube base. There was also one broken link in the fiber optics, which had been added to the string at AMANDA's request so that data from the digital modules could be processed in the same way as all the other AMANDA data. By extraordinary luck this was the same module, the one with the failed tube base.

Unlike other AMANDA strings, each module on the DOM string is separately programmable, and -- because of communication systems designed by McParland -- not just from Antarctica: not only the operating software but also the more fundamental "firmware" of each buried computer can be reprogrammed right from Berkeley Lab.

John Jacobsen and Azriel Goldschmidt of the Physics Division were also at the Pole and implemented the remote communications and operations; since their return they have been collecting and analyzing a steady stream of data from the digital modules now permanently frozen into the Antarctic ice.

While some features, such as local coincidence, still await implementation, the so-far unbounded success of the DOMs makes the future of neutrino astronomy look that much brighter.

The Lab's Km3 group, in collaboration with other astrophysicists and astronomers, have recently put forth a proposal to the National Science Foundation for an expansion of AMANDA dubbed "IceCube." In addition to those named above, Km3 includes senior members William Chinowsky, Dave Nygren, and George Smoot of Physics Division, Doug Lowder of UCB and SSL, Jodi Lamoureux of NERSC and Howard Matis of Nuclear Sciences.

Photo credits:  The photos are by Robert Stokstad.

Additional Information:

• Sidebar: Jet fuel and long johns -- living and working at the South Pole
• Overview of neutrino astrophysics (George Smoot group)
• Antarctic Muon and Neutrino Detector Array (AMANDA)
• IceCube neutrino detector proposal