Laser-Quick Data Transfer
Researchers learn how to make lasers directly on microchips—the result could be computers that download large files much more quickly.
For the first time, researchers have grown lasers from high-performance materials directly on silicon. Bringing together electrical and optical components on computer chips would speed data transfer within and between computers, but the incompatibility of the best laser materials with the silicon used to make today’s chips has been a major hurdle.
By growing nanolasers made of so-called exotic semiconductors on silicon, researchers at the University of California, Berkeley, have surmounted this hurdle. With further development, the Berkeley lasers could provide ways to transfer more data more quickly, speeding up computing within supercomputers and making it faster to download large files.
“Getting data on and off your laptop is becoming a bottleneck,” says Mario Paniccia, director of Intel’s Photonics Technology Lab. It’s difficult to push data through today’s copper wiring at rates higher than 10 gigabits per second. This slows data transfer between components of a computer, such as the CPU and the memory, and imposes limitations on design. Designers must put components as close together as possible so that data doesn’t have to travel too far, generating heat and slowing down the system.
Data encoded in light pulses can travel farther faster and with lower losses. But the only way to get optical components onto today’s chips is to do it using materials and manufacturing methods that are compatible with the silicon systems used in today’s fabs. “The future of photonics is based on silicon,” says John Bowers, Kavli chair of nanotechnology at the University of California, Santa Barbara. The problem is that silicon itself is s a poor laser material, wasting a lot of energy and making little light.
The most efficient lasers are made out of a group of materials called “III-V” semiconductors, whose numerical name comes from the columns of the periodic table where the elements used to make them are found. Like silicon, these materials are crystalline. But the crystal lattices of silicon and of these materials do not line up with one another because the atoms are different sizes. When researchers grow III-V materials on top of silicon, the III-V crystal strains to align with the silicon crystal, leading to defects that degrade performance.
Connie Chang-Hasnain, professor of electrical engineering and computer science at Berkeley, has overcome this incompatibility between silicon and laser materials by taking advantage of the properties of nanostructures and by carefully controlling the growth process. The Berkeley researchers start by placing a silicon substrate inside a chemical growth chamber, raising the temperature to 400 °C, and flowing in gases containing indium, gallium, and arsenide. By controlling the ratios of the gases and their flow rates, Chang-Hasnain has found, it’s possible to direct the growth of these III-V materials so that it starts from a small point called a “seed.” As an indium-gallium nanopillar sprouts up from the seed, it forms a defect-free crystal. The seed seems to protect the rest of the structure from the influence of the underlying silicon. The researchers then flow in a second set two gases to make a gallium-arsenide shell around the pillar.
When the nanopillar is pumped with light from another laser, the light spirals around inside the pillar, as if running up and down a spiral staircase. The difference in materials between the core and the shell encourages this effect, trapping the photons in this spiral until they reach a high enough energy threshold and are emitted. This spiraling effect is something that hasn’t been seen in other types of lasers before. These results are described in the journal Nature Photonics. The next step is to demonstrate that the laser can be pumped with electrical energy, key to making a compact laser. Chang-Hasnain is confident the Berkeley researchers will make an electrically pumped laser. In another publication in Nano Letters her group demonstrated exotic semiconductor diodes on silicon, which they’re now adapting to pump the nanolasers.
Another key to making lasers on silicon chips is not to let the temperature get too high. Chang-Hasnain says that her process could eventually be used to grow high-quality lasers on otherwise finished silicon chips patterned with transistors and optical components, giving them the capability of encoding data into pulses of light. Depositing high-quality III-V semiconductor crystals usually requires higher temperatures—instead of 400 °C, these materials are usually grown at 700 °C, a temperature that would destroy a microprocessor. Chang-Hasnain says it’s the nanostructure of the lasers that makes this possible: high-quality nanostructures can generally be grown at lower temperatures than large films made from the same materials.
“A lot of progress has been made on silicon optical components,” says Intel’s Paniccia. However, progress on lasers that are compatible with silicon chips has lagged behind. Researchers have made various optical components from silicon using materials and processes already present in chip-manufacturing lines. But they have to add on the lasers afterward. Intel, IBM, and other companies have been developing such workarounds.
Chang-Hasnain acknowledges that the group has many more things to prove, from electrical pumping of the lasers to proving they provide enough light of the right wavelengths and can couple it to other optical components. But Intel’s Paniccia says the demonstration that these laser materials can be made compatible with silicon is “a big step.”
Katherine Bourzac is the materials science editor for Technology Review.