Optical integration has been a hotly debated subject over the past few years, but now many industry observers are recognizing the need for multifunction optical integration on a highly scalable, low-cost platform.
In the last two years much discussion has been devoted to the problem of how to satisfy soaring demand for high-speed communications bandwidth. But recent market events have thrown this demand into sharp relief: Now more than ever, the industry needs to deliver bandwidth at the right price, at the right time.
Industry analysts RHK have said that the optical-infrastructure industry as a whole needs to achieve a 50x to 100x improvement in cost structure, while Fechtor, Detwiler & Co. have pointed to the labor-intensive, nonautomated hand-assembly process inherent in current component manufacturing as the Achilles' heel for the whole optical-communications industry.
In fact, the components industry has been searching for some time for an optical-integration technology that will stand the test in the cost-constrained world of mass-production. As much as anything else, this represents an engineering challenge.
At least 40 GHz
To satisfy bandwidth requirements, successful technologies need to function at the 40-GHz speeds required in next-generation networks, and must cater to wave-division multiplexed systems with channel counts well in excess of 40. Miniaturization is required to help operators reduce space requirements at their central offices. Infrastructure providers need to reduce costs by using commercial assembly techniques, and by integrating various optical functions into a single module to dramatically reduce the fiber content and connections necessary in advanced networking equipment.
To best meet those requirements, the successful optical-component platform must provide both parallel integration (many duplicates of the same function within a single device) and serial integration (several associated components within a single device), thus delivering a full range of custom-integration capabilities. And the unit must provide all of that capability on a scalable production platform at a low cost per channel.
In contrast to discrete optical components, silica and the uncertainties surrounding pure monolithic integration in exotic III-V materials, Bookham Technology's solution to optical integration is based on one of the best-understood, most highly engineered and manufactured high-technology materials on Earth: silicon. The technology is called ASOC, for Active Silicon Optical Circuit.
The technology is based on single-mode rib waveguides formed on silicon-on-insulator substrates. Silicon is practically "transparent" at telecommunications wavelengths, and such waveguides have low losses (less than 0.2 dB/cm) and relatively small birefringence (less than 10-3). The refractive index is high (3.5), which allows compact optical circuits to be made.
But the key differentiating feature for silicon is in materials technology and the ability to integrate a broad array of optical functions using very scalable production processes. ASOC involves silicon device fabrication techniques developed for electronic device processing: It has the key advantage of working in a well-characterized materials system and using mature processing technology. It is this differentiator that convinced Bookham Technology that it was worth investing time in solving the related engineering challenges.
An important aspect of the initial ASOC offering was the development of a set of waveguide-based elements that could be assembled via CAD models into practical integrated optics devices. These fundamental passive elements included bends, couplers and fiber-waveguide interfaces. In addition, engineers developed other elements such as doped structures and waveguide gratings. Discrete III-V devices-lasers, photodetectors and others-are conveniently hybridized onto the surface of the silicon, to complete the range of optical functions.
The physical properties of silicon rib waveguides-in particular, the effective-mode index-are largely determined by their dimensions, and designing for optimum operation is a classic engineering trade-off. In general, the waveguides need to be designed for single-mode operation, suggesting that the waveguides need to weakly confine the light traveling through them.
Most practical functions require the waveguides to include bends, which should have as tight a radius as possible without high losses. That requires strong confinement.
The bend radius largely determines the chip size. There is thus a strong incentive to reduce bend radius, and hence chip size and cost. The trade-off is that tighter bends exhibit higher losses.
To tackle that problem, Bookham Technology engineers have developed proprietary processing techniques and waveguide topologies that have allowed them to produce bend radii at 500 microns.
Another consideration is that external light signals need to be coupled into the device from an optical fiber, and the waveguide size determines the efficiency and tolerance at that interface.
To overcome this, a fiber-waveguide interface was developed. In ASOC technology the fiber is located in a V-groove that is wet-etched into the silicon, using the same photolithography mask that is used to define the rib waveguide on the device. Lateral alignment is therefore tightly controlled by the mask itself. Vertical alignment is defined by the depth of the V-groove etch, and again can be controlled tightly (to within 0.4-micron standard deviation).
The waveguide itself extends out over the V-groove, and is tapered toward the fiber, to maximize coupling. Light must also be coupled to the waveguide from lasers hybridized onto the ASOC device, and again the coupling efficiency and tolerance are key parameters.
A key initial attraction of silicon as an optical manufacturing process was its ease of hybridization: Mounting other semiconductor elements, such as lasers, on the surface of the silicon device can be done via established processes. Unlike alternatives such as silica, there is little thermal mismatch between the hybridized device and the silicon substrate, thereby increasing the reliability of operation.
There is also a good match between the refractive indices of the various semiconductor materials, making the coupling task easier. Lasers, for instance, can be located in etched holes in the silicon, allowing the laser stripe to be positioned against the waveguide facet. Since the waveguide and laser location holes are defined in the same etch process, the placement accuracy using automated pick-and-place equipment in high-speed assembly processes is assured.
Finally, the overall waveguide dimensions determine the accuracy with which the effective index can be controlled. Bookham Technology's engineers studied those variations extensively to produce a process with optimum control of the effective index.
The development of those basic components allowed the company to enter production in 1997 with a set of bidirectional optical transceivers and transmitters. But, successful though they were, these were simply a step along the way-the optical equivalent of discrete components in the development of microelectronics.
The engineering investment in ASOC as a large-scale optical integration platform began to pay off in 1999, as the company began to build a suite of products based on its arrayed waveguide grating (AWG) offering.
In volume production, silicon AWGs yield nonadjacent channel crosstalk levels of below -45 dB, enabling worst-case specifications of -23 dB for 40-channel Gaussian devices, eliminating the need for "cleanup" filters. The small-dimension ASOC waveguides were combined with 2-D mode expanders to achieve insertion losses of less than 8 dB for flattened devices and less than 6 dB for a Gaussian.
The next step toward higher-level integration was to utilize the ability to inject charge into the silicon waveguide, to enhance the range of optical functions intrinsically available through monolithic integration (rather than by hybridization). Charge injection into the silicon changes both the refractive index and linear absorption coefficient of the material. Thus, an electrical signal can be used to attenuate the light traveling through the waveguides. This is a valuable capability, since it points directly to other functions commonly used in dense wave-division multiplexed networks.