Planar waveguides can be based on different materials and created by photolithography techniques. Wafer-scale processing enables automation, integration of multiple functions, and customization to individual requirements.
Bob Shine and Jerry Bautista
Wavelength-division-multiplexing systems and components must deliver increased performance at lower costs. And all this performance must fit into limited space, either at collocation terminals, central offices, or in designated huts along the network. Planar lightwave circuits (PLCs) are a key technology that can help resolve money and space issues.
Essentially, PLCs consist of waveguide structures patterned on a substrate. The waveguide structure is defined using photolithography techniques, allowing wafer-scale processing, automation, integration of multiple functions, and customization to individual requirements. Lower costs can be achieved through high-volume wafer-scale processing, while allowing integration of multiple functions on a single substrate to increase functionalities with reduced footprints. In many ways, the transition to PLCs from labor-intensive discrete optical components mirrors the transition that occurred in the semiconductor industry in the 1970s when products evolved from using discrete transistors to the use of integrated circuits.
While silica-on-silicon is the most commonly used PLC platform because of its excellent index match to silica fibers and the maturity of processing equipment, waveguide components have been fabricated using polymers, silicon-oxynitride (SiON), pure silicon, and indium phosphide (InP). This processing of passive waveguides is a significant benefit in itself and allows a wide range of products. The PLC platform also allows the integration of active devices to create an even wider range of products. With the integration of multiple active and passive functions, individual components can be fabricated to meet unique customer demands, almost as ASIC chips are fabricated in the semiconductor industry.
While the analogy to the semiconductor industry is valid in many ways, there are many differences between manipulating electrons and photons. Unlike the semiconductor industry, which is dominated by a single material and an almost singular focus on making linewidths smaller, optical components can be made from a variety of materials and require a range of factors to be considered. Issues such as what functions are to be addressed, cost, what performance criteria are most critical, and what size is allowable for the device all factor into the technology decision. A look at the different material choices can identify the relative benefits and disadvantages of each (see table).
Silica waveguides (silica-on-silicon) are by far the most common material choice for PLCs because of the refractive index match to the fiber itself, minimizing the insertion loss of the components. Two major classes of deposition process are employed broadly today: chemical-vapor deposition (CVD) and flame hydrolysis (FHD). The chemical-vapor-deposition approach is a modification of standard semiconductor processing and is compatible both with clean-room processes and the production of high volumes of wafers. As an example, roughly a hundred wafers can be loaded at a time into deposition chambers (see Fig. 1).
The flame-hydrolysis-deposition process is markedly different from CVD. Glass precursor chemicals are introduced into a flame that hydrolyzes the chemicals to form the appropriate glass composition. The porous layer deposited in this way on a substrate must then be heated to consolidate the porous layer into a solid, clear-glass layer free of bubbles or other defects.
Exploiting existing techniques
Using process equipment developed for the semiconductor industry allows rapid turnaround of custom prototypes, a key step in reducing the overall system-design time. The mask design for a custom circuit, production of the mask, wafer processing, and packaging can be achieved in a number of weeks. In many cases, the custom product meets the customer's reliability requirements because the waveguide materials and packaging have been tested on other devices, allowing qualification through similarity. Finally, the use of multibeam and multistage waveguide features allows for the significant reduction in component size compared to components made from discrete elements.
From a product standpoint, the use of silica waveguides has a number of significant advantages. An index of refraction roughly equal to that of the fiber minimizes losses at the fiber-chip interface. The CVD process is also very mature and allows low-cost volume production using wafer-processing technology and equipment similar to that developed for the semiconductor industry. The rather large dimensions of arrayed waveguide (AWG) devices—roughly 1 x 3 cm—makes it imperative that clean-room conditions prevail to avoid incorporation of particulates, a factor in favor of CVD processing.
Use of a silicon substrate allows active components to be added to deal with products that integrate active and passive components. One disadvantage of silica waveguides is the limited range of refractive index differences that can be achieved, ultimately limiting the size reduction of individual devices.
Adding structural choices
In a polymer waveguide structure, a polymer layer is spun onto a substrate and patterned to create a waveguide. One advantage of polymer waveguides is that the chemistry of the waveguide can be almost continuously varied to control the desired properties such as refractive index, thermal response, or dopant levels. Polymer waveguides have been used for switches due to their greater degree of sensitivity to thermal variations than silica. In addition, a dopant can be added to a polymer waveguide in much higher concentrations than would be possible in a stable glass structure, a fact of interest in waveguide amplifiers.
The typical deposition process for polymers is spin-coating. For many polymers, this process produces a preferential orientation of the polymer chains, which creates significant birefringence in the waveguide—a serious limitation of the device. Polymer waveguides also tend to be very moisture-sensitive, leading to problems of reliability or requiring hermetic sealing. Finally, both the interface and waveguide losses are higher in polymer waveguides than in silica waveguides, limiting the overall performance that can be achieved.
In the silicon-oxynitride process, silica (SiO2) is exposed to ammonia (NH3) to create silicon-oxynitride. With this material, the index-difference limitation of silica waveguides is overcome and very-high-contrast waveguides are possible. High-contrast waveguides allow very tight bend radii and hence the possibility of very small devices. As the index difference increases, however, the waveguide size decreases, creating challenges in fiber coupling to the waveguide and creating high interface loss.
A large refractive-index difference can also produce higher waveguide-propagation losses due to scattering, as well as the typically higher propagation losses inherent in the SiON processing. Typically a silica overcoat is added to reduce the refractive index difference at the waveguide boundaries.
Silicon can also be used as the waveguide material. A key benefit is use of very-well-developed process technologies built for the semiconductor industry. Automated assembly techniques can be used to integrate active components such as lasers or photodetectors, creating low-cost integrated devices. As with SiON, however, interface and propagation losses can be a challenge for products based on silicon.
A key attraction to an indium-phosphide system is that many active devices can be made directly using this material. So, rather than patterning passive waveguides and bonding active devices to the substrate, a full range of functions theoretically can be made directly on a wafer. While the number of wafer-processing steps would be increased significantly, problems such as alignment of the bonded active device to the patterned waveguides and interface insertion loss can be avoided.
Indium phosphide unfortunately is a difficult material system to work with. The wafers are limited in size (<3 in. in diameter) and must be handled very delicately to avoid breaking. While there may be advantages to using one system to do everything, often a compromise in performance must be made. In many cases, independently optimizing different functions and then combining the results can achieve better performance. A simple example of this case is yield: from a manufacturing point of view, it is better to throw out individual pieces that failed than to reject entire components when one smaller piece fails.
Network demands and PLC products
While the benefits of planar lightwave circuits have been discussed in terms of rapid design cycles, customization capabilities created by the wafer processing and size reductions gained by integrating multiple functions in a single component, it also is instructive to examine specific products that are—and will be—manufactured with this technology.
High-channel-count DWDM systems allow network designers to greatly increase capacity on a fiber but require efficient and cost-effective filtering components. While thin-film filters are currently the most popular filter type, they do not scale well to high channel counts or dense channel spacings. An arrayed waveguide—a PLC component in which multibeam interference allows the simultaneous filtering of 40 channels or more—can provide the desired cost, size, and functionality needed for high-bandwidth systems (see Fig. 2).
Another component required as channel counts increase and optical wavelengths are routed in a mesh network is a dynamic gain equalizer. This product balances individual channels to compensate for amplifier nonlinearities and to avoid receiver saturation. While single-channel devices such as variable optical attenuators exist to perform this function, the same inefficiencies occur as with thin-film filters for high channel counts. Dynamic gain equalizers can be created on a PLC by combining an AWG with an interferometric waveguide structure and controlling the relative phase of one arm. Attenuation is achieved by heating one arm of the interferometric structure to create a controlled amount of interference at the output. The solid-state construction of the design (no moving parts) addresses concerns about reliability.
Switching is another function that is required in an optical network, either for protection routing or for crossconnects. Low-port-count switches can be made with PLC structures very similar to the device previously described, but with the output controlled to be either completely on or off. Other technologies, such as microelectromechanical systems (MEMS) or bubble jets are better suited for the high-port-count switch applications. As with other devices, MEMS can be integrated on a PLC platform to combine functionality.
Finally, bonding active elements to the PLC platform can create hybrid products. An example of such a product is an optical channel monitor, created by bonding a detector array to the output of an AWG to monitor individual channels. This type of product is needed as more functionality is created in the optical layer, for example, to monitor the signal quality and detect if wavelengths are switched in a mesh network.
While only a few specific products have been discussed in this article, it is clear that PLCs create a robust, flexible technology platform that will meet the needs of next-generation optical networks. Integration of multiple functions reduces the overall size and cost; wafer processing allows for rapid prototyping of design with scalability to very high volumes; and careful package design allows full Telcordia requirements to be achieved.
Bob Shine is the director of marketing and Jerry Bautista is the chief technology officer and senior vice president of technology at WaveSplitter Technologies, 46430 Fremont Blvd., Fremont CA 94538. Bob Shine can be reached at firstname.lastname@example.org, or 510-580-8800 x274.