Very Thin Lasers Could Speed Computers
Slimmer vertical lasers could link microchips with light
Image: Dr. Hongjun Yang
Holy Mirrors! Photonic crystal reflectors are made from a patterned piece of silicon, to keep the profile of a new kind of laser slim.
1 August 2012—A new type of ultrasmall laser could bring optical communications onto computer chips, breaking a bottleneck that limits computing speed.
The laser is small enough to allow an array of hundreds of them to be placed on a chip, the researchers who devised it say in a July online paper for Nature Photonics. “Once you make it so flat, thin, on-chip integration becomes very simple,” says Zhenqiang Ma, professor of electrical and computer engineering at the University of Wisconsin–Madison. Data transfer between processors is limited by the capacity of copper wires to carry signals, but this can be overcome by using beams of light, just as optical fiber is being used to replace copper cables in the telephone system.
Previous designs for this type of laser led to devices that were between 15 and 30 micrometers tall, toweringly high on a chip on which other features are measured in tens of nanometers. The new laser, designed by Ma and Weidong Zhou, a professor of electrical engineering at the University of Texas at Arlington, stands just 2 μm high.
The device is a vertical-cavity surface-emitting laser (VCSEL), which emits light through its top surface. Chip designers will most likely prefer VCSELs to more traditional, edge-emitting lasers, in which the beam comes out of one end. VCSELs are easier to manufacture, easier to align with elements like optical fibers, and easy to build into arrays of multiple lasers. Every laser needs a mirror at each end to bounce photons back and forth through the laser cavity, thus building up the light beam. Most VCSELs use distributed Bragg reflectors—alternating layers of material with different refractive indexes—to act as mirrors, but it’s these layers that make the laser so tall.
Instead of Bragg reflectors, Zhou and Ma used much thinner structures, photonic crystal mirrors made of a particular pattern of silicon and air-filled gaps. They built the active part of the laser out of indium gallium arsenide phosphide, a compound semiconductor.
Assembling the device required transfer printing, a technique developed by John Rogers, a materials scientist at the University of Illinois at Urbana-Champaign. In that process, Ma and his colleagues pressed a sticky polymer gel onto each piece they wanted to transfer, such as the mirror. Lifting the gel rapidly pulled the piece underneath up with it, in much the same way ripping off a Band-Aid pulls the hair on your arm. They then placed the piece where they wanted it and pulled the gel away slowly, leaving the piece behind.
In this case, the researchers built the first mirror directly on the chip and the laser cavity and second mirror each on separate wafers. Then they used transfer printing to place the laser cavity on top of the first mirror and put the second mirror on top of the cavity. Thin layers of silicon oxide between the layers held the whole thing together. The researchers say the technique is compatible with processes for making silicon logic chips, which can’t tolerate the high temperatures needed for other laser-manufacturing approaches. “It’s the only way to make a laser in a low-temperature process,” Ma says.
The laser they built shines with a wavelength of 1550 nanometers, which is common in communications systems. Zhou says they could produce other wavelengths by altering the design of the photonic crystals or by using different semiconductor materials, allowing the arrays to send multiple streams of data over different wavelengths simultaneously and thus increasing bandwidth.
“This is an interesting piece of research and is novel in that they propose to use vertically emitting devices,” says Jeffrey Kash, a senior researcher in the Center for Integrated Science & Engineering at Columbia University. “As a first step, it is a very nice result.”
Kash adds, though, that a number of key improvements are needed to make the lasers practical. For one, the lasers will need to run on electricity; right now, they depend on another laser beam as a power source. The researchers also need to efficiently increase their power to at least 1 milliwatt. [[different antecedents]]
Zhou believes the research team will have an electrically pumped version of the laser within a year or two and that they will also be able to improve its other characteristics. “In theory, we believe our laser should have much better performance than conventional VCSELs, but, yes, we do have some ways to go,” he says.
Zhou and Ma say their laser could be made commercially viable within about five years. They’ve formed a company, Semerane, to commercialize the technology.
About the Author
Neil Savage, based in Lowell, Mass., writes about strange semiconductors and amazing optoelectronics. In the April 2012 issue of IEEE Spectrum, he reported on molybdenum disulfide, a potential rival to graphene in nanoelectronics.