The integrated circuit has made such steady strides over the past 40 years that it’s easy to believe in a sort of “manifest destiny” for electronics. How could a year go by without the introduction of cool new gadgets boasting previously unimagined capabilities at amazingly affordable prices?
But the chip industry is approaching a crisis. After decades of progress, continued improvement in power efficiency has begun to stall. If we want to continue proliferating cheaper, smaller electronics and usher in what many in the chip industry call an Internet of Things—a future full of billions of always-on, always-connected devices and sensors—we will have to look beyond the CMOS transistor to find a less power-hungry technology.
The future may lie in the past: Looking back to the earliest days of electrically driven computing, we’ve found a surprisingly attractive alternative. It’s the electromechanical relay. As a switch, the relay is about as fundamental as you can get—it uses a voltage to physically open and close a circuit. Early relays were far too slow and power hungry to compete with vacuum tubes, let alone transistors. But by using modern CMOS production processes, we think the relay can get a microscopic makeover.
These miniature moving switches—or nanorelays—aren’t as speedy as the solid-state devices on today’s chips. But what the tiny mechanical switches lack in speed they make up for in energy efficiency. Nanorelays don’t leak current when they’re off, and they can change states with just a fraction of the energy that’s needed to turn a transistor on or off. These qualities make the microscopic switches ideal for ultralow-power chips that can run off scavenged energy from acoustic vibrations, light, or ambient radio signals. With some clever engineering, it may even be possible to make nanorelays fast enough to drive the core logic inside cellphones, tablets, and other portable electronic devices.
Mass-produced chips full of moving parts aren’t as far off as they might seem. After years of small-scale experiments, we’re now on the cusp of demonstrating fully functional, complex integrated circuits that are entirely mechanical. The long-retired relay could soon be reborn.
For decades, every time engineers reduced the size of the transistor, they were rewarded with a faster and more energy-efficient switch. But a little more than a decade ago, chipmakers realized that simply shrinking transistors wouldn’t improve their energy efficiency the way it used to.
The problem is that transistors are imperfect switches; they leak current even when they’re supposed to be shut off. This leakage is fundamental to the way the transistor operates, and so there’s no easy way to eliminate its effects. If you reduce the operating voltage of the transistor, less energy will be needed to switch the device. But the amount of time that this lower-power transistor takes to switch will balloon, and meanwhile the other transistors in the circuit will leak more current while they’re waiting for the operation to be completed. As a result, there’s a fundamental limit to the energy efficiency of a CMOS circuit, and we’re fast approaching a point where we won’t be able to keep boosting the performance of a chip without increasing power consumption.