"It's like being in the Bermuda Triangle," says Rodger Hart of the iThemba Laboratory for Accelerator Based Science in South Africa. I take the compass to see for myself. At first the needle points in a steady direction, which for all I know could be magnetic north. But then I take a step forward, and the needle swings to a completely different quadrant. Another step, and yet another direction. Next I put the compass down against the large rock outcropping we are standing on. Now as I move the compass across the rock, with every few centimeters of motion the needle swings around.

The location is the center of the Vredefort Crater, about 100 kilometers southwest of Johannesburg. Vredefort is the oldest and largest impact remnant on the planet, created about two billion years ago when a 10-kilometer-wide asteroid slammed into the earth. Evidence of older collisions exists elsewhere, in South Africa and in Western Australia, but in those cases no geologic structure has survived the ravages of time.

Vredefort itself is not obviously a crater to the untrained eye. Geologists estimate the total crater size at 250 to 300 kilometers across, but the rim has long since been eroded away. The most obvious structure remaining is the Vredefort Dome, which is the crater's "rebound peak"--where deep rocks rose up in the crater's center after the impact.

According to Hart, the probable source of Vredefort's weird magnetism was a strong and chaotic magnetic field generated by currents flowing in the ionized gases produced at the height of the collision. Laboratory experiments confirm that impacts cause intense magnetic fields in that fashion. Scientists have calculated that a mere one-kilometer-wide asteroid, one tenth the size of Vredefort's, would create a field 1,000 times that of the earth's at a distance of 100 kilometers.

Vredefort's intense but random magnetism was not apparent from aerial surveys. Those analyses showed anomalously low magnetism over the crater, like a hole punched in the prevailing magnetic field. All the magnetic madness on the ground averages out to nothing when seen from too high up.

The results could have implications not only for earth geology but also for studies of Mars. The immense Martian basins Hellas and Argyre displayed virtually no magnetism when measured by the orbiting Mars Global Surveyor. The conventional explanation runs like this: When these craters formed around four billion years ago, the impacts wiped out the preexisting magnetization of the rocks. Therefore, at the time of their creation Mars must not have had a magnetic field, because that field would have been preserved in the magnetization of the basins' rocks when they cooled. Mars does not now have a magnetic field, but long ago it did. Thus, the standard explanation implies that Mars lost its field very early on.

But as Hart points out, if the Hellas and Argyre basins show the same properties as the Vredefort Crater, one cannot conclude anything about Mars's magnetic field when they were formed--it may have still been going strong. Mario Acua, a principal investigator on the Mars Global Surveyor project, however, points out that data from smaller Martian craters of about Vredefort's size do not agree with Hart's scenario.

Back on earth, Hart has proposed a high-resolution helicopter survey of Vredefort's magnetism, from an altitude low enough to see the magnetic variations. That would produce a complete magnetic map--and make some sense of the crater's weirdness.