The Kirkendall effect.

At the boundary between two solids diffusing into each other at different rates, for example zinc and copper, their alloy (brass) grows in the direction of the faster-moving species (zinc). Unfilled voids are left behind and coalesce into large pores.

Discovered in 1942, the Kirkendall effect describes what happens when two solids diffuse into each other at different rates. The boundary between two metals, zinc and copper for example, is formed by a growing layer of alloy — brass, in this case — which expands in the direction of the faster-moving species, zinc.

This was the clue to Ernest Kirkendall's discovery that the atoms of the two solids don't change places directly; rather diffusion occurs where voids open, making room for atoms to move in. In the wake of the faster-moving material, large pores or cavities form as unfilled voids coalesce.

On the nanoscale, the Kirkendall effect explains why a fast-diffusing cobalt nanocrystal leaves a hollow center behind as it moves into a surrounding sulfide-compound shell. 

The nanospheres were remarkably uniform: depending on the proportions of the starting materials, their hollow centers were 40 to 70 percent as big as the initial crystal, but hole size varied no more than 13 percent in any given batch. This uniformity and versatility suggested a wide range of applications including drug delivery systems, optics, electronics, and selective chemical reactors, all on the nanoscale.

Ref.
"Formation of hollow nanocrystals through the nanoscale Kirkendall effect," by Yadong Yin, Robert M. Rioux, Can K. Erdonmez, Steven Hughes, Gabor A. Somorjai, and A. Paul Alivisatos in Science, 30 April 2004