wo papers by Harvard and Cornell researchers in the June 13 issue of the journal Nature described a spectacular breakthrough in miniaturization: researchers have now created transistors whose switching components are literally single atoms.

After nearly a year of topsy-turvy excitement and puzzlement over the now discredited findings of Dr. J. Hendrik Schön at Bell Labs, the field of molecular electronics is still very much alive. Researchers are making steady progress at work whose practical prospects are promising, if uncertain.
"Honestly, there's a river flowing here," Dr. Thomas N. Theis, director for physical sciences research at I.B.M., said. "The Schön thing is like throwing a big rock in there. It makes a big splash, but the river keeps flowing on."

Advertisement

At first glance, the findings from June look similar to the fabricated research. Dr. Schön, who was fired last week after an independent investigatory panel found that he had manipulated and fabricated data, had claimed transistors with single molecules as switches. Had they proved real, those transistors might have catapulted the young field of molecular electronics from research laboratories to factories in a few years, transforming the computer chip industry.

What astonished scientists was that the transistors appeared to be the fastest yet made, and they supposedly worked in the same way as silicon transistors in today's computer chips. Like silicon transistors, they exhibited "gain," amplifying the strength of the input electrical signal. For use in computer processors, gain is essential. Otherwise, the signal becomes weaker at each calculation step and fades away before the answer is complete.

The atomic-scale transistors described in Nature, similar devices made by two research groups working independently, remain real advances. But they lack gain and work just at very low temperatures. At present, they are useful only for studying the physics of how electrons flow through molecules. Any applications, far off, would be more modest. They may prove useful as sensors.

The atomic transistors work somewhat differently from silicon ones.

A transistor is just an electric switch. In the off position, no current can flow through. To switch a silicon transistor to the on position, an electric field is applied from the side, injecting electrons. The electrons then carry current across the switch. The atomic transistors have specially designed molecules wedged between two microscope electrodes. To cross from one electrode to the other, electrons have to hop across atomic islands at the center of the molecules.

Because negatively charged electrons repel one another, there is enough room on the island for only one electron, which sits there unmoving. But applying a positive electric field attracts additional electrons. If the field is tuned so that there is an average of one and a half electrons on the island, a parade of electrons starts hopping on and off — an electric current flowing through the transistor, turning it on.

The Cornell researchers, led by Dr. Paul L. McEuen and Dr. Daniel C. Ralph, physics professors, used a single cobalt atom at the center of the molecules. The Harvard group, led by Dr. Hongkun Park, a professor of chemistry, used two vanadium atoms.

"As of now," Dr. Park said, "we are doing these studies to really understand how the electrons move through the molecules."

His team has examined 12 transistor variations. That could allow the researchers to design molecules that work at room temperature. Such transistors might be useful as sensors, by adding binding sites for specific molecules like DNA and proteins that carry electric charges. When the molecules stick to the transistors, they would turn on the transistors, setting off an alert.

For such applications, the lack of gain is not a problem. "For sensing," Dr. McEuen said, "you don't care. If you can detect a single electron, you don't care. For doing physics, it doesn't matter."

The next application of molecular electronics will most likely be for computer memory. In the last five years, scientists at Hewlett-Packard and U.C.L.A. have designed molecules that act like switches by shifting between two shapes. In one configuration, the molecule lets current flow easily. In the other, electrical resistance is high.

A grid of such molecules can be used for computer memory. About one volt of electricity flips the molecule from one shape to the other, allowing data to be written to the switches. To recall data from memory, about one-tenth of a volt is used to read the current positions of the switches without disturbing them.

In the longer term, scientists are still thinking how to use their molecular circuits for performing the logic operations of computer chips. "Logic is a harder target," Dr. James M. Tour, a chemistry professor at Rice University in Houston, said. "You have to have gain."

Two teams of scientists, one at I.B.M. and the other at the Delft University of Technology in the Netherlands, have made transistors and simple logic circuits out of rolled-up carbon molecules known as nanotubes. In May, I.B.M. announced that its nanotube transistors outperformed silicon transistors. But no one has a good idea how to produce them in quantity. Current nanotube manufacturing produces a jumble of tubes of different diameters and twists. Separating the ones needed for the transistors is laborious, and so is guiding the nanotubes to the right places.

Another promising approach is using rods of crystalline silicon. With the rods, known as nanowires, Dr. Charles M. Lieber, a professor of chemistry at Harvard, has not only made transistors, but also electronic devices like light-emitting diodes, sensors and logic circuits.

At Hewlett-Packard, scientists say they think that they may be able to pair simple logic circuits made out of their molecular switches with silicon transistors that boost the signal before sending it to another molecular circuit.

"We've also shown we can do logic with these switches," said Dr. R. Stanley Williams, director for quantum science research at Hewlett-Packard Labs in Palo Alto, Calif. "The answer to a simple logic query ends up as an output resistance."

Dr. Tour has an even more novel architecture. He and his collaborators throw molecular switches together onto a small wafer with small gold particles. The switches and gold link up at random. By applying voltages, the team theorizes that it can program the circuit to program a certain logic function. "It's very much a like a biological system, like a brain," Dr. Tour said.

The researchers have shown in computer simulations that the strategy can work, and they have built a prototype of the chip, but have not programmed it.

They have time. Most electrical engineers say silicon transistor technology has at least another decade to run before it runs into fundamental laws of physics that will prevent further miniaturization.