The transistor is the basic technology behind, well, everything in modern computing. In digital circuits, a transistor acts like a tiny voltage-controlled switch: it can be on, allowing current to flow, or off, blocking it. Those two electrical states are the basis for representing binary data—1s and 0s—and for building the logic gates that make processors work. Modern CPUs and GPUs are packed with transistors; the base M4 chip in the laptop I’m writing this on contains about 28 billion of them.
But is the transistor’s time in the sun drawing to an end? The humble little switch has served us extremely well, but it does place a limit on our ability to process data. If we want to process more data, then we need more transistors. And if we want to process data faster, then we need transistors that switch from on to off and back again more quickly. And if we want both, then we need to cram more and more transistors—while also making them smaller and faster—onto our silicon wafers.
We’ve spent decades miniaturizing and speeding up our transistors, but eventually this process starts to bump up against fundamental limits imposed by the laws of physics. One of these is heat generation: switching current generates heat, and the faster your transistors are switching, the more heat you end up generating. (There’s a good Explain It Like I’m Five post about this phenomenon here.)
Figuring out how to get around these limits in a manner that’s efficient and practical is the holy grail of computing research, and a new paper published in Science this month describes one promising new idea. The paper describes how a team from the University of Tokyo took a radical approach to the problem: they did without transistors entirely. Instead, their “non-volatile quantum switching element” uses the spin of an individual electron to represent the state of a given bit.
(A brief aside: spin is a quantum mechanical property that’s analogous to the way that a macroscopic ball can spin around a given axis—it can rotate in one of two directions. Electrons don’t actually spin, though, because if they did, their surface would be traveling faster than the speed of light. Trying to wrap one’s head around quantum spin is difficult, but for the purposes of this post, the point is that an electron can have one of two spin states, and those can be used to encode a 1 or a 0.)
It turns out that flipping electrons’ spin states is both faster and more energy efficient than flipping transistors on and off. As per the paper, processing a single bit of information with the quantum switching element takes 40 picoseconds. (A picosecond is one trillionth of a second, or 1 x 10-12 seconds.) This is an insanely short amount of time; for comparison, even today’s fastest computers need somewhere in the order of a nanosecond, which is 1 x 10-9 seconds, to do the same thing—so we’re talking several orders of magnitude worth of improvement here.
There are a couple of other exciting aspects of this technology. The electrons remain in their assigned spin states until something alters them again, which means information stored this way is non-volatile: data remains stored without power. It also seems to be extremely durable: the paper describes how the switching element remained stable after 100 billion transitions, which is again several orders of magnitude better than current technologies, where heat causes progressive degradation and eventual failure.
Of course, the usual disclaimers about this being essentially a proof of concept apply, and there’s no guarantee that anyone will be able to manufacture chips using this technology in a cost-effective manner. But it does point to one way that we might be able to move past the limits of our current computing technology.
