Hydrogen can peel atoms off layer by layer.

By harnessing hydrogen's tendency to be selective about the chemical bonds it breaks, chemists can now peer at the underlying structure of semiconductors and control the growth and quality of thin films more precisely.

Recent advances in scanning tunneling microscopy made it possible to resolve atomic details of a material's surface, but little of what lies below. Now, chemist John J. Boland of the IBM Thomas J. Watson Research Center in Yorktown Heights, N.Y., has used hydrogen to remove or shove aside atoms on the surface of semiconductors so that he can train his microscope on atoms one or more layers down. By peeling layers away, "we get an in-depth view, almost a cross-section," Boland says.

Other scientists have had to probe underlying structure by trying to look between atoms on the surface, he adds.

Within any material, atoms tend to arrange themselves into a stable, low-energy configuration. Because they lack an upper layer with which to bond, however, those on the surface must arrange themselves differently. The germanium atoms on a semiconductor surface, for example, form three strained bonds with like atoms below, leaving one unlinked "bond" protruding upward.

In Boland's studies, incoming hydrogen atoms first link up with any free bonds sticking out of the semiconductor. Then hydrogen goes after the most strained bonds, which exist between the surface atoms and those directly below. Sometimes four hydrogen atoms surround -- and free -- a germanium atom, releasing it as a volatile compound. In other cases, hydrogen forces the surface atoms to clump, exposing patches of an underlying layer.

Bonds between those underlying germanium atoms are less strained and therefore strong enough to resist being snipped free by hydrogen. Consequently, hydrogen can link with those atoms' free bonding sites, but does not break bonds within the semiconductor.

Using this approach, Boland made very clear scanning tunneling microscope images of germanium. In the Jan. 10 SCIENCE, he shows that the layer directly below the surface structurally matches the bulk of the semiconductor. This contrasts with silicon, in which the penultimate layer still contains some strain and thus differs from the bulk structure below.

To study silicon, Boland and his IBM colleagues controlled the hydrogen-reaction conditions so that they could break bonds of specific strengths -- peeling off silicon atoms one layer at a time. "You can actually pace the chemistry, and pacing the chemistry is very important," says Boland.

This approach also lets the IBM group modify thin films as they are made. When chemists deposit a material on silicon--to make computer chips, for example -- some atoms link up to the silicon via strained bonds, while others form stable connections. The strained bonds represent weak spots. But by halting deposition of that substance midway and allowing the hydrogen to snip out weak bonds, "we can make very much better high-quality films," says Boland. This approach also allowed the IBM scientists to make thin-film transistors in fewer steps than usual.