Teeny-weeny transistors.

They keep getting smaller, and smaller, and smaller.

Seemingly without limit, electronic circuits and micromachines continue to shrink, zooming toward the utopian technological goal of devices built atom by atom.

Now, using scanning tunneling microscopes and atomic force microscopes, scientists are etching surfaces and building machine parts with features only a fraction of a micrometer in size. They can do so because of these devices' minuscule probes, or single- atom tips, which gently skate over a surface one atom at a time.

Though originally designed for imaging surfaces with atomic resolution, these "proximal probes" are yielding new techniques for making incredibly tiny electronics.

Eric S. Snow and P. M. Campbell, both physicists at the Naval Research Laboratory in Washington, D.C., have devised a way to use an atomic force microscope as "a unique lithographic tool."

"The key is the development of a fast, reliable exposure process that enables us to modify a surface and use it to transfer a pattern by selective etching," Snow says. Using a tiny metal- coated probe, the researchers draw a pattern on a silicon surface. The pattern, marked with oxidized silicon, serves as a "mask" for a subsequent etching process, whereby a corrosive liquid dissolves away silicon around the mask.

The technique yields etched details a few micrometers wide and less than 1 micrometer thick, Snow says.

"A unique feature of this lithography process is that the exposure and imaging mechanisms of the atomic force microscope operate independently," Snow adds. Thus, while making a circuit, someone can make high-resolution images of the etchings without damaging them.

To demonstrate the technique's potential for making tiny electronics, Snow and Campbell have fashioned a key piece of a field-effect transistor only 30 nanometers wide.

Taking this lithography concept a step further, physicists Steven C. Minne and Calvin F. Quate of Stanford University and their colleagues are fabricating arrays of probes to draw circuit patterns in parallel. Since writing with a single probe is slow and tedious, the researchers figured that writing with many tips at the same time would speed up the whole affair.

To date, Quate's group has produced a row of five tips that write in concert. In the long run, the team is aiming for 1,000 tips that draw simultaneously. "In one configuration, there's a single row of 1,000 probes. In a second, there are 10 rows of 100 tips," says Quate. "The first one's easier to make, but the second one is more efficient, though harder to build."

Meanwhile, to confirm that such methods can yield usable microelectronics, Quate's group has made a transistor segment only one-tenth of a micrometer wide.