Volume 3, Issue 19 March 2006
Spinning Out the Future of Computing
The computer industry is headed toward a brick wall. Within a decade or so, Moore's Law–which predicts that the number of transistors that can be packed on a silicon integrated circuit doubles every 18 months–will run up against the laws of physics. To keep processing speed on the ever-increasing fast track, scientists and engineers are experimenting with entirely new kinds of devices that could be the building blocks of tomorrow's computers. UC Berkeley physicist Joe Orenstein's research is in the realm of "spintronics," a field that could lead to computers that store and process information in the spins of individual atoms.
Joe Orenstein, shown here with a grating used in spin spectroscopy, is an expert at employing electromagnetic radiation to probe condensed matter systems. (photo Roy Kaltschmidt, CSO)
"Chip companies are feeling the pressure of Moore's Law," says Orenstein, who is also affiliated with Lawrence Berkeley National Laboratory's Materials Sciences Division. "They're looking for what's next, and spintronics is certainly under consideration."
The angular momentum carried by an electron is referred to as electronic spin. An electron's spin is similar, at least conceptually, to the direction of a rotating top. Labeled "spin-up" or "spin-down," an electron's spin can be used to represent a zero or one, much like a charge of voltage in a traditional transistor. At this small scale though, so-called "quantum weirdness" comes into play. Spin can be actually be "spin-up," "spin-down," somewhere in the middle, or, oddly, both at one time.
That's why if spintronics can be made to work, Orenstein says, the direction of an electron's spin would "make the perfect qubit," or building block of a quantum computer. Each qubit could represent multiple values at the same time. As more qubits are strung together, the power of the quantum processor grows exponentially.
Meanwhile, certain kinds of spin current, generated as electrons move through a semiconducting material like silicon, may dissipate less energy than the charge current of today's transistors. So even if quantum weirdness can't be harnessed to build a quantum computer, spintronics could still help prevent microprocessors from overheating so easily.
"As a practical matter though, can we create currents of spin, store spin, and change spin?" Orenstein says. "And if so, can we detect it? Those are all open questions."
Electron spin is assigned a value of "up" or "down." (courtesy Berkeley Lab)
Recently. Orenstein and his colleagues took a leap forward in uncovering the secrets of spin currents. Specifically, the researchers demonstrated that compared to charge current, spin current moves through the semiconductor at a slower rate. The effect, called "spin Coulomb drag," is caused by the electrons bouncing into each other as they move through the material.
Spin Coulomb drag was first posited in 2000 by scientists at the University of Missouri. The theory wasn't well-received and Orenstein's group hadn't even heard of it when they began their research. Through a series of novel experiments though, Orenstein and his team proved that the 2000 theory was right on target.
"We learned something very definite and concrete at the end of the day," he says.
To conduct their experiments, Orenstein and his colleagues modified a tried-and-true technique called transient grating spectroscopy. They point two lasers at a sample of electron gas held in "quantum wells" of gallium arsenide. The lasers pulse at one-tenth of a picosecond, creating alternating layers of electons with different spin states. Over time, the bands of spin-up and spin-down states blur into each other. The resulting pattern is then "read" by two other lasers. This provides the researchers with the "spin diffusion coefficient" that can be used to calculate the drag on the collective motion of the electrons.
Whether these new findings will hinder or accelerate the development of spintronics-based computers depends on the particular architecture of the processor, Orenstein explains. When high-speed flow of spin current is essential, the drag would be, well, a drag. But if the system calls for the spins to remain isolated for long periods, spin Coulomb drag might be a feature rather than a bug.
Still, the future of spintronics-based computers is far from a certainty. Right now, the experimental devices can only operate at incredibly cold temperatures to prevent the fragile electron spin states from changing willy-nilly. To that end, Orenstein is collaborating with UC Berkeley chemistry professor Peidong Yang to determine whether nanowires, thousands of times thinner than a human hair, could be used to reign in the electrons so their direction could be better controlled.
"With this kind of tabletop science, each day presents the opportunity for discovery, which is different from experiments where the apparatus may take many years to build," Orenstein says. "It's possible to learn things about matter that aren't relative or subjective but can actually be proven to be absolute truths."