Researchers at IBM have found a way to create a new kind of binary switch that can retain its state by adding or removing oxygen atoms from a thin film of metal oxide. The discovery, they say, could pave the way for circuits that act like neural connections in the human brain. However, the work also seems to contradict earlier experiments indicating that the oxide switch might someday make a near-ideal replacement for silicon transistors.
The material in question, vanadium dioxide, exhibits strong interactions among its own electrons. This gives the metal oxide a rather peculiar property: It is one of the few known materials that acts as an insulator at low temperatures and as a metal at high temperatures. In principle, this phase transition could also be induced by applying an electric field that would tug on electrons in the material. And that could mean faster, less-power-hungry electronics.
In traditional semiconductor-based transistors, voltage at the gate electrode generates an electric field that modulates the flow of current in the channel below it. Electrons can travel only through a thin part of the channel closest to the gate, and most must migrate to that region from other parts of the transistor before they can be carried across it. But in a metal oxide device, the entire thickness of the material would become conductive, and electrons could move through it as soon as they are freed from the atom they were bound to.
“If you can throw a switch and have these materials change from conducting to nonconducting and back, you would have a qualitatively different transistor,” says Andrew Millis, a theoretical physicist at Columbia University, in New York City, who studies oxides and other materials. “The electrons would always be there, and you could wave a magic wand and get current to flow.”
Vanadium dioxide is among the most promising of these metal oxides because it switches phase at around 340 kelvins, not too far from room temperature. That suggests that the energy needed to induce a phase transition with an electric field might be low enough to be practical for electronics.
In an experiment published last year in Nature, a team at the RIKEN Advanced Science Institute, in Wako, Japan, reported that they had indeed achieved “electrostatic switching.” To achieve high electric fields close to the material, they placed a droplet of ionic liquid—a salt that can be polarized in the presence of an electric field—on top of it. When a voltage was applied to the liquid, charges were drawn to opposite sides of the droplet. The positive charges, which accumulated close to the metal oxide film, created an electric field that seemed to cause the material to turn from an insulator into a metal.
However, research published in March in Science casts doubt on that explanation. Stuart Parkin, manager of the magnetoelectronics group at the IBM Almaden Research Center, in San Jose, Calif., set up a similar experiment. But in addition to measuring the conductance and resistivity of the films, the team also looked to see if there were any chemical changes by placing vanadium dioxide devices in a vacuum chamber. After applying a positive voltage to turn the films metallic, the team introduced oxygen-18, a heavier isotope of oxygen, to the chamber. They then reversed the voltage to turn the film back to its insulating state. When they later analyzed the films, they found oxygen-18 atoms in them in a greater abundance than normal.
“What we’re finding is a different effect is taking place,” Parkin says. “The electric fields you create are so large, they actually cause the atoms or ions to move from the surface of an oxide into the liquid. It’s a permanent switch, until you put the ions back into the surface.”
Switching by altering chemicals is likely to be slower than altering the electron states, and it isn’t likely to pave the way for a replacement of solid-state transistors. But Parkin sees other applications. As a counterpart to IBM’s work on software that functions like the brain, he says, “we need to think of hardware or devices that are also cognitive in the way they operate, that fundamentally change their properties as you use them.” One potential avenue might be a nanofluidic system that could make paths more conductive or more insulating, reinforcing or weakening the level of conductivity much like a connection between two brain cells, which can be strengthened by activity or weakened by neglect.
This sort of application may sound familiar. Memristors have also been tapped for their brain-mimicking properties. A memristor acts like a resistor, but its resistance can vary according to the current passing through it, and it can remember that resistance value after the current stops. Oxygen migration is the way that memristors (such as HP’s titanium dioxide devices) change states, Millis points out, and some researchers have reported that vanadium dioxide shows potential as a memristive system. But Parkin says the system he and his colleagues have studied is unique, since the entire oxide film switches state. When a memristor first forms, filaments of metal atoms migrate in from metal electrodes, and then just a small portion of oxide surrounding the filament changes state during operation.
Does the latest research really disprove the RIKEN team’s results? The answers from experts are mixed. Some say it is likely that the Japanese team’s results were due to oxygen migration. That would mean it has still not been established whether an applied voltage can be used to induce a phase transition in the material. “To me, it’s not clear,” says Jochen Mannhart, a physicist at the Max Planck Institute for Solid State Research, in Stuttgart, Germany. “My hypothesis right now is that if you increase [the] voltage, one gets more and more oxygen depletion.” However, the IBM team may be exposing their films to a somewhat higher voltage than the RIKEN team did, and sorting out exactly what has occurred “may take a few years,” he says.
Shriram Ramanathan of Harvard University says it’s possible that both teams are seeing a mix of electrochemical and electrostatic effects. Regardless, the door is still open to other kinds of metal oxide switching. Ramanathan and others are working on different transistor control structures based on solid insulators and electrodes—instead of ionic liquids—that will be less likely to alter the chemical structure of the oxide.
“[The IBM] paper does a good service to the community because it proves that you have to be very careful how you deal with this ionic gating,” says Ivan Schuller, a physicist at the University of California, San Diego, who is working on understanding the metal-insulator transition. There is much left to be done to see if vanadium dioxide might work as the next big transistor material. “That is a major quest,” Schuller says. “The story is definitely not over.”