Spintronic Memristors

16 March 2009—Last May, researchers at Hewlett-Packard stunned the electronics world with the demonstration of a fourth fundamental circuit element to add to the classical three. The memory resistor, or memristor, joined the resistor, the capacitor, and the inductor, closing a theoretical gap in the physics of electronic circuits. Now other researchers have created new types of memristors that rely on the magnetic properties of electrons, potentially leading to entirely new kinds of circuits that should be easy to integrate with existing electronics.

Current flowing through a memristor can alter its electrical resistance, and it retains that altered state even after the current is turned off, making it a natural for nonvolatile memory. The memristor promises the introduction of much tinier circuits, instant-on computers, and the ability to mimic the function of neurons in the human brain.

Shortly after the demonstration of memristance, researchers began looking for this property in spintronics—a relatively new branch of electronics itself. Spintronics is at the heart of recent advances in hard-drive data density and the niche nonvolatile memory known as MRAM. Whereas electronics works by manipulating the movement of electrons, spintronics works by manipulating a quantum mechanical property of electrons known as spin. (Imagine the electron as a spinning ball.) Spin is the property that is responsible for magnetism—materials are magnetized when a majority of their electrons have their spins pointing in the same direction. Melding memristors and spintronics yields devices whose resistance changes according to the spin of electrons passing through it, and those devices will remember that resistance.

Yiran Chen and Xiaobin Wang, researchers at disk-drive manufacturer Seagate Technology, in Bloomington, Minn., described three examples of possible magnetic memristors this month in IEEE Electron Device Letters . In one of the three, resistance is caused by the spin of electrons in one section of the device pointing in a different direction than those in another section, creating a ”domain wall,” a boundary between the two states. Electrons flowing into the device have a certain spin, which alters the magnetization state of the device. Changing the magnetization, in turn, moves the domain wall and changes the device’s resistance.

The different designs can be flipped between high- and low-resistance states at different rates, from picoseconds to microseconds, each preferable in different applications. For reading a hard drive, for instance, you’d want to sense changes in a magnetic field in a few picoseconds, whereas for something like a radiation sensor, you’d want a response time measured in microseconds.

And the devices are all relatively easy to construct. ”We can easily integrate a magnetic device on top of a CMOS device,” says Chen.

Wang believes that a spin memristor can be more finely tuned and is more flexible than the device HP described, which was based on the movement of ions in a material. ”It’s more broad in our opinion, more controllable,” he says. In part, that’s because of the variable switching rates, but it’s also because spin is not a binary condition—neither up nor down but rather existing along a continuum. So a device doesn’t need to make a complete change from magnetized to nonmagnetized to register a change in the resistance.

Massimiliano Di Ventra at the University of California, San Diego, and Yuriy Pershin at the University of South Carolina described a spin memristive device in a paper in Physical Review B last September. Unlike the Seagate devices, theirs combines a ferromagnetic metal and a semiconductor. If the metal is a perfect magnet, all of its electrons are in the spin-up state, and only spin-up electrons will flow from the semiconductor into the ferromagnet, leaving spin-down electrons clustered in a growing cloud at the interface between the materials. Eventually, the cloud grows to the point where it begins to block the flow of further electrons, resulting in spin-dependent conductivity.

Di Ventra says the device is memristive but that it’s not an ”ideal” memristor, because it loses the memory of its state after the power is turned off. Still, the new device could have several advantages, he says. For one, some materials can hold a spin state far longer than they can hold charge, and that might allow engineers to design spin-based circuits that consume less power.

Stanley Williams, the HP researcher who first described a memristor last May, says he’s glad to see other researchers getting into the area. ”I am delighted that people are now playing the game of finding different physical representations of memristance,” he says. ”In general, I think linking memristance and other phenomena such as spin transport is a very excellent path forward to putting a lot of functionality into a small package. The thing that really differentiates a memristor is the fact that it has and remembers a state. That is tremendously powerful for a passive device, and the implications of that have barely been explored.”