Surface Emitting Lasers Open New Roads
VCSELs may become the CMOS of the photonics world.
Ruth DeJule, Associate Editor -- Semiconductor International, 6/1/1998
For the past 20 years, semiconductor laser diodes have attained only a small fraction of the success and widespread applications of the silicon-based transistor, despite impressive modulation speeds, low power consumption and long lifetimes. Today, vertical cavity surface emitting lasers (VCSELs), some believe, are the CMOS of the photonics world, the key enabler for VLSI photonics. One reason is that the laser output is perpendicular to the surface, which allows the use of standard semiconductor processing techniques and testing and eases integration. Circular, low divergence output beams provide optical fiber coupling efficiencies approaching 100%, and arrays of hundreds to thousands of lasers operating in parallel can pave the way to a wide range of applications, from photoelectronic sensors and laser pointers to terabit (1x1012) data communi cations transceivers.
Laser printers, for example, currently use one or two of the established edge-emitting laser diodes with data speeds up to ~100Mbit/sec to scan printed information onto a photoreceptor. One constraint to this strategy, however, is that individual scanned laser spots must be precisely positioned relative to each other to achieve high resolution. Alternatively, a single beam can be split into two beams and modulated separately. Monolithic arrays of multiple lasers can provide a simplified solution, noted Dr. Robert Thornton, principal scientist at Xerox PARC (Palo Alto, Calif.).
Edge-emitting laser diodes and VCSELs are made of GaAs or InP based materials, depending on the operating wavelength, and are typically grown by metal organic chemical vapor deposition (MOCVD). Both consist of a central optical cavity surrounded by layers that act as optical waveguides (Fig. 1). However, instead of using low refractive index materials to ensure longitudinal optical confinement, VCSELs employ high-reflectivity distributed Bragg reflector (DBR) mirrors. These mirrors, which form the vertical optical cavity, are also transparent at the laser wavelength for light extraction from the top (or bottom) surface of the laser structure. They must also exhibit minimum resistance for optimal current injection.
|
| Fig. 1. VCSELs may require ~50 epitaxial layers for the semitransparent mirrors that form the vertical optical cavity. (Source: Honeywell) |
 |
| Fig. 2. From four levels, VCSEL light is emitted at the tip of the blue lines, surrounded by a hexagonal pattern of holes. (Source: Xerox PARC) |
The primary application for VCSELs, however, is for data communications where the capability of these miniaturized, high-speed modulators can be fully realized. VCSELs are already used in serial fiber transceivers such as gigabit ethernet, fiber channel and asynchronous transfer mode (ATM) applications. Though recently suspended, Motorola's parallel fiber transceiver, Optobus, also employed VCSELs.
Some estimate that the density of free-space optical interconnects will eventually reach several thousand I/O channels/cm2, each operating at 1Gbit/sec. To achieve this level of I/O density requires the integration of optical and electronic components. Currently, the most economical approach is using hybrid chips: electronics on silicon and photonics on GaAs. The monolithic integration of lasers and photodetectors, however, would be ideal, eliminating the need to stack and package two different die and the associated ball bonding required. This has been demonstrated at Honeywell's Technology Center (Minneapolis, Minn.), where a VCSEL, photodetector and integrated circuits were fabricated on a single GaAs substrate, stated Jim Guenter, senior staff engineer at Honeywell's Micro Switch Division (Richardson, Texas).
The question remains: Can the monolithic integration of optical and electronic components be done economically? While the hybrid approach currently provides the necessary level of functionality at a lower cost, the chip is only part of the equation; system costs must also be factored in. As miniaturization requirements continue, the monolithic strategy may become economical even at the current cost levels, Guenter noted. Until that time, however, work proceeds in hybrid chips to meet the growing optoelectronics market.
