Surface diffusion with a hop, skip or dip.

On the atomic scale, a smooth metal surface has the undulating appearance of an orderly layer of ball bearings. Scientists have long assumed that individual atoms deposited on such a surface readily shift from place to place, moving about like hard spheres rolling across bumpy terrain.

A series of theoretical studies and experiments now reveals that on certain surfaces, a deposited atom actually trades places with a surface atom. Any motion across such a surface consists of a sequence of exchanges, in which an atom momentarily on top of the surface ends up in the surface layer, and the atom displaced from that layer finds itself on top, a short distance away from the deposited atom's initial position.

"We've discovered a new phenomenon which seems interesting and important," says Peter J. Feibelman of the Sandia National Laboratories in albuquerque, N.M., who used an elaborate computer model to predict this effect.

The discovery adds a new dimension to studies of surface diffusion. Such investigations -- which focus on the motion of adsorbed particles across a surface -- are valuable for materials scientists interested in growing novel crystalline materials layer by layer.

"It's a new mechanism, and that makes it important," says Tien T. Tsong of Pennsylvania State University in University Park. Studies of this mechanism and its consequences have already prompted the publication of several papers in PHYSICAL REVIEW LETTERS, with more papers to follow.

Feibelman started by theoretically determining the energy required for an adsorbed aluminum atom to "hop" over the energy barrier separating one location from another on an aluminum surface. The value he obtained proved significantly higher than expected on the basis of empirical data.

He then turned to an idea that had first surfaced more than a decade before in connection with the diffusion of adsorbed atoms on a grooved crystal surface. In certain cases, adsorbed atoms appeared to move readily from one groove to another instead of following the grooves, suggesting that some kind of exchange was taking place between the adsorbed atoms and the slightly elevated surface atoms defining the grooves.

Applying a similar idea to the smoother aluminum crystal surface he was considering, Feibelman found that an adsorbed aluminum atom actually starts forming a chemical bond with a neighboring substrate atom. The energy required for the entire exchange process proves considerably less than that required for an aluminum atom to roll from hollow to hollow as if it were a ball bearing.

"The key is to think of [the process] as a chemical phenomenon rather than in terms of a hard sphere moving on a bumpy plane," Feibelman says.

He also predicted that the exchange mechanism would produce a distinctive pattern of sites visited by a diffusing atom, and that this pattern should be apparent in data compiled from field-ion microscope observations. His Sandia colleague Gary L. Kellogg found such a pattern for a platinum atom diffusing on a platinum surface, and Tsong and his group observed a similar pattern for an iridum atom on an iridium surface.

More recent observations by Kellogg reveal that individual, adsorbed platinum atoms displace nickel atoms from a nickel surface. A similar process involving rhenium atoms on an iridium surface has allowed Tsong and his co-workers to observe directly the intermediate steps in the exchange process.

"It's very exciting because we can see the exchange taking place step by step," Tsong says. "We can now see the intermediate state -- an adsorbed atom pushing up a substrate atom."

But the price is complicated by the fact that some combinations of atoms, such as palladium on platinum, don't trade places. Furthermore, recent experiments show that although pairs of platinum atoms migrate by a series of exchanges, three-atom clusters move by a combination of exchanges and hopping.

"It depends on what the materials are," Feibelman says. "But at this point, we don't know the rules."