In order to chart the particles'paths, the scintillator oil is parceled out into 66 cells—acrylic-walled rectangles nine meters long by 12 centimeters wide and 25 centimeters high, with a photomultiplier on each end. The cells are wrapped in copper foil so that a flash in one won't trip the photomultiplier in a neighbor. But the gamma rays normally fly through several cells before petering out, so the computer continuously digitizes the arrival time and energy of all the flashes picked up by all the photomultipliers in the array, and scans the lot for "coincidences"—signals from photomultipliers in blocks of up to 15 adjoining cells at the same time—and says, "Ah. Triple pulse with the right energy distribution. That's a keeper."

Meanwhile, the neutron plows through the oil, gradually losing steam until it gets absorbed by an atom of gadolinium, which soaks up neutrons like a sponge. (Persuading gadolinium to dissolve in mineral oil is no small feat—like most metals, it's soluble in acids, but uninterested in oil. Some pretty harsh things used to have to be done to the gadolinium to get it into solution, and the result was a dark, nasty liquid that blotted out all light passing through it within half a meter or so. The solution also went bad in just a few months, meaning that the detectors were constantly in the shop for an oil change. So Piepke and Novikov, in collaboration with Bicron, a leading manufacturer of radiation detectors, developed a new recipe for dissolving gadolinium that results in a fluid as clear as water that remains stable for at least two or three years. Bicron now sells the stuff, which has become the industry standard.) Upon catching a neutron, the gadolinium atom emits a fresh cascade of gamma rays at energies of up to 8 MeV. Because these gamma rays are so hopped up, the computer looks for them in coincidences of up to 35 cells at once. A couple of hundred microseconds (millionths of a second) separates the posi-tron's demise and the neutron's capture, and the three-one flash pattern with its set of characteristic energies and delay times is the unmistakable fingerprint of a neutrino.

But lifting that print is not trivial. A bazillion other particles are also zipping through the detector, and they all leave their mark. Says Boehm, "Our detector registers 20 neutrino inter-actions a day, but we get about 2,000 hits per second from the cosmic-ray flux, plus other background radiation, so it's a very difficult experimental problem. We have to use lots of clever tricks." The Super Kamiokande detector is buried a kilometer deep in a zinc mine to screen out as much background radiation as possible. "Unfortunately," says Boehm, "the Arizona desert has no commercial mineral deposits, so there are no deep mines." Instead, the Palo Verde Neutrino Detector is buried about 25 meters (82 feet) deep—as far down as Caltech could afford to dig. In lieu of a kilometer of rock, the scintillator cells are surrounded by a bank of muon detectors that register cosmic-ray hits. Also called veto detectors, the muon detectors when they go off tell the computer, "Any data you are getting right now is from a cosmic-ray shower. Ignore all inputs for the next 10 microseconds." To help keep costs down, the muon detectors were spares from the MACRO (Monopole, Astrophysics and Cosmic Ray Observatory) project, lent by Linde Professor of Physics Barry Barish and then-Division Chair Charles Peck (PhD '64). Between the veto and neutrino detectors, a 100-ton, one-meter-thick wall of water absorbs neutrons, the other chief byproduct of cosmic rays. And, finally, the computer filters the data through screening programs that reject flashes that aren't energetic enough or otherwise don't look promising.

An end-on view of the detector array (main drawing). The gray circles are the photomultipliers. The muon veto detectors are plastic boxes filled with a liquid scintillator sans gadolinium–muons make gamma rays just like positrons do. There's about an inch of headspace over the liquid in each box, so there are two layers of overlapping boxes along each side wall to ensure complete coverage. At the bottom is a side view of a single scintil-lator cell. The last half meter on each end of the cell is partitioned off and filled with pure mineral oil to act as an additional buffer. A series of light-emitting diodes (LEDs) down the length of the cell provide standardized pulses of light to calibrate the photomultipliers and, with the optical fibers, keep tabs on the scintillator oil's clarity.