We’re approaching the speed limit for electronic computer chips. If we want to go faster, we’ll need data-carrying photons—and some tiny lasers.
Nearly fifty years ago, Gordon Moore, the cofounder of Intel, predicted that the number of transistors packed into computer chips would double every two years. This infamous prediction, known as Moore’s law, has held up pretty well. When Intel released the first microprocessor in the early 1970s, it had just over 2,000 transistors; today, the processor in an iPhone has several billion. But all things come to an end, and Moore’s law is no exception.
Modern transistors, which function as a computer’s brain cells, are only a few atoms long. If they are packed too tightly, that can cause all sorts of problems: electron traffic jams, overheating, and strange quantum effects. One solution is to replace some electronic circuits with optical connections that use photons instead of electrons to carry data around a chip. There’s just one problem: Silicon, the main material in computer chips, is terrible at emitting light.
Now, a team of European researchers says they have finally overcome this hurdle. On Wednesday, a research team led by Erik Bakkers, a physicist at Eindhoven University of Technology in the Netherlands, published a paper in Nature that details how they grew silicon alloy nanowires that can emit light. It’s a problem that physicists have grappled with for decades, but Bakkers says his lab is already using the technique to develop a tiny silicon laser that can be built into computer chips. Integrating photonic circuits on conventional electronic chips would enable faster data transfer and lower energy consumption without raising the chip’s temperature, which could make it particularly useful for data-intensive applications like machine learning.
“It’s a big breakthrough that they were able to demonstrate light emission from nanowires made of a silicon mixture, because these materials are compatible with the fabrication processes used in the computer chip industry,” says Pascal Del’Haye, who leads the microphotonics group at the Max Planck Institute for the Science of Light and was not involved in the research. “In the future, this might enable the production of microchips that combine both optical and electronic circuits.”
When it comes to getting silicon to spit out photons, Bakkers says it’s all about the structure. A typical computer chip is built upon a thin layer of silicon called a wafer. Silicon is an ideal medium for computer chips because it is a semiconductor—a material that only conducts electricity under certain conditions. This property is what allows transistors to function as digital switches even though they don’t have any moving parts. Instead, they open and close only when a certain voltage is applied to the transistor.
Within the wafer, the silicon atoms are arranged as a cubic crystal lattice that allows electrons to move within the lattice under certain voltage conditions. But it doesn’t allow similar movement for photons, and that’s why light can’t move through silicon easily. Physicists have hypothesized that changing the shape of the silicon lattice so that it is composed of repeating hexagons rather than cubes would allow photons to propagate through the material. But actually creating this hexagonal lattice proved incredibly challenging, because silicon wants to crystalize in its most stable, cubic form. “People have been trying to make hexagonal silicon for four decades and have not succeeded,” says Bakkers.
Bakkers and his colleagues at Eindhoven have been working on creating a hexagonal silicon lattice for about a decade. Part of their solution involved using nanowires made of gallium arsenide as a scaffold to grow nanowires made of the silicon-germanium alloy that have the desired hexagonal structure. Adding germanium to the silicon is important for tuning the wavelength of the light and other optical properties of the material. “It took longer than I expected,” says Bakkers. “I expected to be here five years ago, but there was a lot of fine tuning of the whole process.”
To test if their silicon alloy nanowires emit light, Bakkers and his colleagues blasted them with an infrared laser and measured the amount of infrared light that made it out on the other side. The amount of energy Bakkers and his colleagues detected coming out of the nanowires as infrared light was close to the amount of energy the laser dumped into the system, which suggests that the silicon nanowires are very efficient at transporting photons.
The next step, says Bakkers, will be to use the technique they’ve developed to create a tiny laser made from the silicon alloy. Bakkers says his lab has already started work on this and may have a working silicon laser by the end of the year. After that, the next challenge will be figuring out how to integrate the laser with conventional electronic computer chips. “That would be very serious, but it’s also difficult,” Bakkers says. “We’re brainstorming to find a way to do this.”
Bakkers says he doesn’t anticipate that future computer chips will be entirely optical. Within a component, such as a microprocessor, it still makes sense to use electrons to move the short distances between transistors. But for “long” distances, such as between a computer’s CPU and its memory or between small clusters of transistors, using photons instead of electrons could increase computing speeds while reducing energy consumption and removing heat from the system. Whereas electrons must transmit data serially, one electron after the other, optical signals can transmit data on many channels at once as fast as physically possible—the speed of light.
Because photonic circuits can quickly shuffle large amounts of data around a computer chip, they are likely to find widespread use in data-intensive applications. For example, they could be a boon to the computers in self-driving cars, which have to process an immense amount of data from onboard sensors in real time. Photonic chips may also have more mundane applications. Since they won’t generate as much heat as electronic chips, data centers won’t need as much cooling infrastructure, which could help reduce their massive energy footprint.
Researchers and companies have already managed to integrate lasers into simple electronic circuits, but the processes were too complex and expensive to implement at scale, so the devices have only had niche applications. In 2015, a group of researchers from MIT, UC Berkeley, and the University of Colorado successfully integrated photonic and electronic circuits in a single microprocessor for the first time. “This demonstration could represent the beginning of an era of chip-scale electronic–photonic systems with the potential to transform computing system architectures, enabling more powerful computers, from network infrastructure to data centres and supercomputers,” the researchers wrote in the paper.
By demonstrating its application in the main ingredient in conventional computer chips, Bakkers and his colleagues have taken another major step toward practical light-based computing. Electronic computer chips have faithfully served our computing needs for half a century, but in our data-hungry world, it’s time to kick our processors up to light speed.
All Rights Reserved for Daniel Oberhaus