BROOMFIELD, Colo. A technology for building 1.3-micron electrically pumped vertical-cavity surface emitting lasers (VCSELs) grown on gallium arsenide is being developed by Sandia National Laboratories and Cielo Communications Inc. The technique, scheduled to be unveiled next year, could lower the cost of neighborhood or midrange fiber optic networks.
High-speed communications over emerging midrange nets need to use the 1.3-micron wavelength, because the fiber's minimum chromatic dispersion there keeps losses to a minimum. But until now only expensive edge-emitting lasers were available at 1.3 micron.
"We've uncovered the holy grail for midrange communications networks," John Klem, a Sandia researcher, said of VCSELs.
A Cielo spokesman said, "The important thing is that VCSELs took over the 850-nm market when they became available, and we expect the same thing will happen when 1.3-micron VCSELs become available. We plan on having chips for ATM/Sonet, gigabit and 10-gigabit Ethernet, Fibre Channel and fiber-to-the-home before the end of the year."
Edge-emitting lasers shoot a beam out of a chip's edge, complicating board layouts and connectors. In addition, they are expensive. On the other hand, surface-emitting VCSELs shoot the laser right out the top of a chip, simplifying layouts and lowering the cost of optical alignment. They can also be packed more tightly on a chip itself.
VCSELs bounce photons between tiny mirrors grown on a chip to emit a laser vertically from the wafer's surface. One wafer can house thousands of high speed VCSELS, each of which can power a laser fiber optic communications link. Besides being easier to handle and manufacture, VCSELs are considered more reliable than edge emitters while consuming as little as one-third their power.
"It's really a lot cheaper to manufacture VCSELs," said Klem. "For one thing, you can test them while they are still on the wafer, like a normal circuit, whereas with the edge emitters you have to dice them up before you can even test them."
The problem before with manufacturing VCSELs at 1.3 micron was that the lattice constant of gallium arsenide (GaAs) limited it to about 980-nm wavelength. With indium phosphide substrates, long-range 1.55-micron VCSELs can be produced, but 1.3 micron was too small for indium phosphide. The 1.3-micron wavelength, then, was too big for GaAs, and too small for indium phosphide.
"The reason 1.3 micron was the hold grail for VCSELs was that there was this gap between 980 nm and 1.55 micron that made it hard to get to we did it by growing a layer of indium gallium arsenide nitride (InGaAsN) on top of the GaAs substrates," said Klem. The added layer upped the lattice constant just enough to accommodate 1.3-micron wavelengths. The technique, originally detailed by a Hitachi laboratory in the mid-1990s, was optimized by the Sandia/Cielo work.
"There is substantial strain between the GaAs and the larger lattice of the InGaAsN, but the process has proven to be reliable and repeatable," the Cielo spokesman said. GaAs is cheaper, the wafers are easier to handle and the other materials needed to grow the VCSEL on indium phosphide are simpler for GaAs substrates. Now that VCSELs can be produced on GaAs, the other more-exotic materials will no longer be necessary for 1.3-micron communications chips.
Cielo Communications was formerly the GaAs fab and laser-array component operation of Vixel Corp., a leading manufacturer of VCSELs, but was spun out in 1998 to provide high-performance packaged optical modules for Gigabit Ethernet, Fibre Channel and ATM/Sonet.