A growing effort is underway among researchers to find a way to make spintronics, the manipulation of electrons’ “spin” to do computing, practical. The promise is clear: spintronics could lead to computers that turn on instantly and electronics that use far less battery power, and also overcome the looming limits to Moore’s Law. But the challenges to using spintronics for logic operations are also daunting. Not the least among them is finding the right material to build practical circuits. For several years researchers have been studying semiconductors such as gallium arsenide, doped with elements such as manganese to make them magnetic. But those work only in temperatures below about minus 120°C–hardly an option for everyday computers.
MIT research scientist Jagadeesh Moodera and his team have developed a material that works at room temperature and is easy to create. The material is indium oxide, which is similar to the material used to conduct charge in an ATM’s touch screen, with a small amount of chromium added to make it magnetic. Other materials that might work, Moodera says, include zinc oxide, widely used in sunscreen, and titanium oxide. The magnetic semiconductor would polarize the spin of the electrons, which then flow into the silicon chip where circuits would use them to perform calculations, while a detector, probably made of the same material as the spin injector, would read them as they flow out.
The material needs more development before it can be tested in an actual circuit. But Don Heiman, a professor of physics at Northeastern University, calls the creation of a magnetic semiconductor that works at room temperature “a pretty big breakthrough.”
Indeed, Moodera says, there are several years’ worth of work to be done to build a practical computer chip based on spintronics. For instance, it’s not clear how the silicon and the indium oxide interact at the point where they touch. It’s very difficult to control the shape of the material at that interface, which is about two atomic layers thick, and it’s possible that differences between the two materials can cause the electron spins to lose their polarization. It will take a fair amount of basic research just to understand what happens at the interface, Moodera says, and more work to learn how to control it. And once they’ve built a working spin injector, researchers will still have to design a spin detector and the transistor.
Spin-based circuits are intriguing because they add a new dimension to computing. While electronic computing is based on the negative charge of the electron, rapidly switching current on and off to make the 1s and 0s of the digital world, spintronics relies on scientists’ growing ability to manipulate another quantum mechanical property of the electron–a property known as spin. That’s important to a computer chip industry that is looking to a time when it will not be able to make circuits any smaller. Current computer chips employ silicon transistors inscribed with lines less than 100 nanometers thick. Chipmakers continue to shrink the size of transistors in order to cram more of them onto a chip, but once they get down to a few nanometers in size, they will begin to leak electrons. Moodera and others hope that spintronics will allow them to do more computing with the same number of transistors.
In a normal semiconductor, an electron’s spin is arranged randomly; one points this way, one points that. But when a majority of electrons have the same spin, the result is a magnetic field. That means any computer chip in which the spins are controlled can function as a nonvolatile memory, because spin stays the same until an external force acts to change it. Because the memory is nonvolatile, meaning it doesn’t require a flow of current to keep refreshing the information, it uses far less electrical power, allowing batteries to last longer and reducing their chance of overheating. It also allows a transistor to use both charge and spin and function as both a logic gate and a memory. A computer chip could perform all the same tasks with fewer circuits, leaving room to add additional circuits without having to make the devices even smaller.
Spintronic devices already exist; hard drives store much more information than a decade ago, because Moodera and others developed spintronics-based read heads that can detect tiny changes in magnetism, allowing more-plentiful and smaller chunks of data to be packed into the same space. And one company has started selling magnetic random-access memory, another spin-based technology, that doesn’t require a flow of current to store data.
But to apply spintronics to logic circuits, scientists have to figure out how to put the spin they want into the transistor and how to detect it on the other side once the circuit has manipulated it. They want a thin film they can place on top of the semiconductor and in which they can use the magnetic field to align the spin of the electrons, before transporting those aligned electrons into the semiconductor. Moodera’s indium oxide is a potential candidate for such a film, once he fully understands it and can figure out how to optimize it.
Moodera is unwilling to predict how long it will take to build a practical spintronic circuit, though he estimates that thousands of scientists in the United States, Asia, and Europe are working on it: “It could be 10 years, it could be six. It’s difficult to say.” But he points to how spintronics have improved hard drives from capacities of a few megabytes 15 years ago, when no one had even thought of using spintronics, to current iPods that let you carry a 60-gigabyte drive around in your pocket. “I’m an optimistic researcher,” he says.
How a Russian cyberwar in Ukraine could ripple out globally
Soldiers and tanks may care about national borders. Cyber doesn't.
Meet Altos Labs, Silicon Valley’s latest wild bet on living forever
Funders of a deep-pocketed new "rejuvenation" startup are said to include Jeff Bezos and Yuri Milner.
A horrifying new AI app swaps women into porn videos with a click
Deepfake researchers have long feared the day this would arrive.
Meta’s new learning algorithm can teach AI to multi-task
The single technique for teaching neural networks multiple skills is a step towards general-purpose AI.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.