Portland, Ore.-Purdue University researchers have invented a "spin filter" that can sort charge carriers by their spin state. "This component is essential to future quantum computers, which need to be initialized in a known state," said professor Leonid Rokhinson. "It will also be necessary for the many other 'spintronic' devices that people are working on today."

Storing data on the polarized-wave function, or "spin," of a charge carrier- a mobile electron or hole in a semiconductor-holds promise for future single-electron transistors, nanoscale nonvolatile memories, quantum computers and other so-called spintronic devices. But even the simplest components necessary to make spintronics a reality-such as filters-have been lacking.

Using the basic physics of a mass spectrometer as his model, Rokhinson fabricated a chip version that can sort particles based on their spin configuration. "Spin filters are one of the most crucial missing ingredients of spintronic devices because they allow you to create spin polarization," he said.

The device uses magnetic focusing to enable the spatial separation of trajectories of carriers (in a two-dimensional "hole" gas fabricated from GaAs heterostructures) by virtue of their spin state. Two quantum-point contacts, acting as a monochromatic source of ballistic holes and a narrow detector, respectively, were demonstrated to work as a tunable spin filter.

Rokhinson used atomic-force micro-scopy with local anodic oxidation and molecular-beam epitaxy to create oxide lines. He was able to create tracks that separate gallium arsenide charge carriers-either "up" for one or "down" for zero-with 200-millivolt potential barriers. The fabrication created two quantum-point contacts, controlled by their own gates but separated by a central gate. By biasing the gates appropriately and supercooling the device to eliminate friction, the balance between momentum and charge could be fine-tuned by the magnetic field, enabling the devices to separate charge carriers with different spins.

"In the absence of friction, the only forces at work here are the magnetic field and the momentum-to-charge ratio," said Rokhinson. "We put two tiny orifices along the diameter of the circular orbit to sort the particles."

The catch here is that charge carriers all have the same mass. For instance, electrons all weigh the same and have the same charge, so the ratio between the momentum of the particle and its electric charge should always be the same.

"In a semiconductor, the spin is coupled to the motion of the particle-its direction and speed-which is called its spin-orbit interaction. The spin somehow adds to the momentum of the particle in the same direction, but will oppose its momentum if the spin is opposed to the particle's direction," said Rokhinson.

As a result, particles with different spins also differ slightly in momentum, depending on whether the spin is in the direction of the motion or opposed to it. For Rokhinson's prototype, the demonstration was even more profound, since the charge carriers were nonmaterial holes, not electrons, yet the device still sorted them by spin.

"In gallium arsenide, spin-orbit interactions are 10 times stronger for holes than for electrons. In indium arsenide, it is just the opposite. . . . But today only gallium arsenide has the engineering maturity necessary to create a spin filter," Rokhinson said.

Rokhinson modeled the gallium arsenide and aluminum gallium arsenide devices before building the spin filter by inserting a small "delta" magnetic field into the equations governing its charge carrier's orbit size. The small additional magnetic field, said to be caused by the spinning of the charge carrier either in the direction of motion or opposed to it , slightly altered the orbit's size, enabling Rokhinson to calibrate how far apart the orifices should be to separate charge carriers by spin. The result was 800 nanometers.

"We model these spin-orbit interactions as a small magnetic field attached to the particle that has different signs for spin up and spin down. The local fields then either add to or subtract from the momentum of a particle, depending on whether spin is up or down," said Rokhinson. "I used that property to design a spin spectrometer instead of a gas spectrometer, because particles with different spins will have slightly different cyclotron orbits-technically they have different momenta, but the same energy."

Another reason that Rokhinson choose to use "holes" instead of electrons was to simplify the experiment. The resultant elimination of scattering helped ensure the success of the first-pass prototype. Even though a hole is just as good a charge carrier as an electron, in a semiconductor holes do not scatter the way that electrons do when traveling in a cyclotron orbit.

"For quantum computing, we can use our current spin filter for the initialization step, but for the future we want to reduce our device to a single-spin detector-a spin transistor," Rokhinson said. "Now it works only with streams of charge carriers."

Another challenge is temperature. "Right now we need very low temperatures-a fraction of a degree above absolute zero," he said. "We need to reproduce the effect at higher temperatures for chips to become commercially available."