And there are other subtleties. Samples of all the construction materials had to be vetted by an exquisitely sensitive radiation detector in the subbasement of Caltech's Bridge Laboratory. Ordinary rebar contains a trace amount of radioactive cobalt 60, added to help monitor the production process. "This cobalt 60 is weak for all practical purposes," says Boehm. "But not for neutrino detectors! We had to request special batches of low-cobalt steel to be shipped to us. And concrete is always slightly radioactive, and there's nothing you can do about it. Normally, it contains about one part per million of uranium and thorium, which would have emitted enough gamma rays to choke our detector." These trace elements are found naturally in Earth's crust, so that when you crush rock into gravel, or quarry limestone for cement, they come along. In fact, the product from the local gravel plant was particularly bad. The rock was volcanic, so that it had lots of heavy elements from Earth's interior, and relatively young, so that the hot stuff hadn't had much time to decay.

"The USGS helped us find a marble deposit near Phoenix that was 10 times radio-cleaner than the local stone. It was trial and error: they sent us lots and lots of rock samples, and we tested them here at Caltech." The Phoenix gravel plant crushed this marble for them, adding about 6 percent to the construction costs. Marble is a soft rock„that's why sculptors use it„and it had never been used in concrete before, so nobody had any idea whether the stuff would be strong enough to support the thousands of tons of dirt that was going to be backfilled onto the roof of the detector chamber. A lot of special testing had to be done at the cement plant before the first batch was trucked to the site. These two factors make the Palo Verde Marble Mix probably the most expen-sive concrete ever poured„with the possible exception of the night that Jimmy Hoffa disappeared„and certainly the fanciest.

Even with all these precautions, the computer records some 600 megabytes' worth of flashes per day. The data is stored on hard drives at the site, and gets shipped once a day over a fast, dedicated Internet connection to computers at Caltech and Stanford that tease out the fingerprints of the 20 neutrinos a day the collaboration is hoping not to see. These computers also reconstruct the neutrinos' trajectories and energy distributions.

	The detector chamber awaits reburial. The access road will soon become a tunnel. To save money, the tunnel was made from the eight-foot-diameter corrugated steel sewer pipe normally seen on highway drainage projects.

The analysis of the first 70 days' worth of data is now complete. The results are bad news for the neutrinos-as-dark-matter folks, says Vogel. "To be blunt, we do not see oscillations, so the mixing angle cannot be large. And we have moved the mass parameter by a factor of 10 toward smaller masses." These results have almost completely closed the window in parameter space that Super Kamiokande had allowed for electron neutrinos. Thus, it appears that muon neutrinos may mix, but electron neutrinos don't„at least, not to within Palo Verde's detection limits.