For tens of millions of years—a mere sliver of astronomical time—the night sky above Earth may have been a bit more populous than it is today. For that brief period, our planet may have had not one but two moons, which soon collided and merged into our familiar lunar companion. No one would have been around to see the second moon—the lunar merger would have occurred nearly 4.5 billion years ago, shortly after Earth had formed.

The two-moon hypothesis, put forth in a study in the August 4 issue of Nature, would help explain why the moon's two hemispheres are so different today. (Scientific American is part of Nature Publishing Group.) The familiar hemisphere facing Earth is covered by low, lava-filled plains (seen as the darker gray areas on the moon's "face"), whereas the far side, which is never visible from Earth, is a collection of rugged, mountainous highlands. Those highlands, according to the new hypothesis, would be the remains of the smaller, short-lived satellite following its collision with the moon that now hangs overhead. The key is that the moonlet's impact would be slow enough to pancake its material across one face of the moon rather than excavating a large crater.

"Usually when you think of two objects colliding, one of them leaves a big hole," says Erik Asphaug, a planetary scientist at the University of California, Santa Cruz, who co-authored the new study with Martin Jutzi of the University of Bern in Switzerland. "At a low enough impact speed, you actually deposit material." The effect of two moons coalescing at subsonic speeds is an almost fluid merger, "like you literally threw a cow pie on the ground and there it is," Asphaug says. "The physics are basically the same."

The leading hypothesis for the moon's creation itself involves an impact, this one a higher-speed crash of a Mars-size body into the nascent Earth. That collision, as the story goes, packed enough punch to kick up a ring of debris around Earth that coalesced into the moon. If an accompanying moonlet formed in the aftermath of that collision, simulations have shown, the system would be unstable, pushing the moonlet into a sudden demise in a collision with the dominant moon or with Earth. But certain orbital safe havens known as Trojan points, leading or trailing the moon in its orbit around Earth, would allow a moonlet to hang around for tens of millions of years before meeting its end.

By that time, the two objects would be at very different stages of evolution: a moonlet roughly one third the diameter of the moon would have cooled and solidified, whereas an ocean of magma would persist on the larger moon. In Jutzi and Asphaug's computer simulations, the pancaking of a solid moonlet against a partly molten moon would provide enough material to create the elevated highlands on one hemisphere and would displace huge amounts of magma to the opposite hemisphere.

One attractive feature of the new hypothesis is that it tidily explains why the near and far sides of the moon are not only topographically but compositionally different. Several sites on the near side sampled by Apollo astronauts had rocks enriched with KREEP—for potassium (K), rare earth elements (REE) and phosphorus (P)—which resists crystallization from magma and hence remains in a molten state until the entire magma ocean has solidified. But KREEP is scarce on the lunar far side. The hypothesized moonlet pushing a cooling magma ocean and its KREEP to the lunar near side would explain the dichotomy. "The momentum of that impact squashes the KREEP onto the other hemisphere," Asphaug says. "There's this compositional puzzle that we did not set out to explain but our model does help to explain."

The planetary scientist who, along with Asphaug, helped vault the giant-impact mechanism for the moon's formation into wide acceptance, sees value in the new hypothesis. "It's an old problem of trying to understand why there's this elevation dichotomy on the far side of the moon," says Robin Canup of the Southwest Research Institute in Boulder, Colo., whose 2001 study with Asphaug identified a Mars-size impactor as the likely moon-yielding culprit. "A nice new explanation that seems plausible for an old problem in planetary science is always a great thing."

A NASA lunar mission scheduled to launch in September, the Gravity Recovery and Interior Laboratory (GRAIL), may help settle the question of whether the moon once had a smaller companion. GRAIL, which will map the moon's gravitational field to expose variations in its near-surface density, may be able to detect the residual effects of a long-lost moonlet pancaked across the lunar far side. "I think it may be testable with time," Canup says. "The type of alteration that they're predicting, you might be able to see some evidence for or against that in something like future GRAIL data."