Planar waveguide integration adds flexibility, improves performance
G. Ferris Lipscomb and Marc Stiller
Planar lightwave circuits can improve the cost/performance ratio of components because multiple functions can be integrated onto a single substrate. Manufacturing processes are based on semiconductor industry techniques and can be applied to optics through changes to the photo-mask design, automated design software, and semiconductor tools.
A leading approach to reduce costs and improve performance of fiber components is the use of planar lightwave circuits (PLCs) to integrate multiple optical functions on a single substrate that leads to an optical module in a single package. Compared to traditional discrete components, integrated devices dramatically reduce size and cost, while maintaining or improving optical performance.
Planar lightwave circuit components are manufactured using tools and techniques of the semiconductor industry, and design changes, within established design rules, can be rapidly implemented through mask changes. For example, using automated design software, a 40-channel arrayed waveguide grating (AWG), operating in the S-band, was manufactured within six weeks. To accomplish this using discrete technology would require 39 different production runs.
Similar flexibility applies in more highly integrated components, such as a 40-channel variable optical attenuator (VOA) multiplexer. Here the operating band of the AWG (S-, C-, or L-band), the AWG passband shape (wideband or Gaussian), the channel plan, and the performance of the VOA array can be quickly and efficiently customized through simple mask changes.
Planar lightwave circuits are optical circuits laid out on a silicon wafer, made using tools and techniques developed to extremely high levels by the semiconductor industry. First, a lower cladding layer of index ncl is deposited, followed by a core layer of index nco, with nco typically approximagely 1% larger than ncl (see Fig. 1). The core layer is then patterned using standard photolithographic techniques, and a channel waveguide pattern is transferred to the core layer using various etching techniques.
Removing all of the material around a waveguide leaves a ridge, or rib, of the waveguiding material extending above the substrate. A rib waveguide is produced by etching away the surrounding core using techniques such as reactive ion etching (RIE).
It is also possible to form a variety of other waveguides using different geometries and etching techniques. A channel waveguide is a structure that acts to confine light in two transverse directions and to allow the light to propagate along the channel, analogous to an optical fiber, but laid out on a silicon wafer. The channel waveguide is then covered by a top cladding layer, usually of the same index as the lower cladding, ncl. The end result is a buried channel optical waveguide on the substrate wafer.
To achieve the high levels of performance required by DWDM systems, extreme precision and control must be exercised in each of these manufacturing steps. Imperfections in the waveguide, such as edge-wall roughness and fluctuations in the indices, lead to severe limitations in the final device performance.
Starting with simple waveguide structures, very complex optical circuits can be fabricated, such as a 6-in. diameter silicon wafer containing eight 16-channel, 100-GHz spacing, reconfigurable add/drop multiplexers (see Fig. 2). Many other functions also have been demonstrated in integrated form on PLCs. Active control of light signals can be achieved through the incorporation of thermo-optic elements. Some of the functions that have been demonstrated in PLC form are splitters, couplers, taps, switches, gratings, interleavers, variable attenuators, and amplifiers, as well as subsystem-level modules such as optical add/drop multiplexers and dynamic gain flattening filters.
Once the basic manufacturing process described above has been extended to include active functions, such as thermo-optic devices, each of the elements and subelements mentioned above becomes a basic building block. Each of these building blocks is implemented within a set of design rules that ensure that it can be fabricated using a standard fabrication process. So, the final function of the device is determined by the pattern and interconnection of these building blocks on the masks used in production, and not on changes in the fabrication process itself.
As we shall see, such mask-based production is very flexible, and can respond very rapidly to new demands. A new design on a new mask can be implemented in a few days, much faster than a new manufacturing process.
This description is, of course, somewhat oversimplified. A real production line consists of a set of released production modules, which can be invoked to build up the final circuit. The order and sequence of execution of these modules, in addition to the masks used, determine the final device function and performance. Often, modules are omitted. For example, an AWG does not require thermo-optic drive electrodes, so that module is omitted. The modules are themselves production building blocks, which, when properly deployed, enhance the flexibility of the overall fabrication process.
Fairly complex, integrated optic PLC components have already been developed and are in widespread commercial use. Since AWGs were first proposed by Smit in 1988, extensive work has gone into applying them to dense wavelength-division-multiplexing applications.1, 2, 3
Arrayed waveguide gratings represent, in and of themselves, a fairly high degree of integration. Commercial AWGs are available with 40 channels spaced at 50 or 100 GHz. The same function performed with thin-film filters would require 39 or more different filters, each with its associated beam expansion and refocusing optics, or more than 120 piece parts. Demonstrations of AWGs with 256 channels, spaced at 25 GHz have been made in the laboratory, providing further evidence of the capabilities of AWGs to scale to even higher channel counts.4 Furthermore, the insertion loss of an AWG does not increase linearly with channel count, as it does for thin-film filters and for fiber Bragg gratings. Arrayed waveguide gratings have the best cost/performance ratio at high channel counts.
For the AWG to operate properly, each of the more than 100 optical elements on the chip must perform precisely. Such is the power of photolithography and semiconductor manufacturing techniques that, once the design is correct and the manufacturing process stable, these integrated optical circuits can be manufactured in high volume. Performance data from an AWG device, with a wide band or "flat-top" passband, with 40 channels in 100-GHz channel spacing, shows the results (see Fig. 3).
A real-time parametric data collection system, which is directly linked to manufacturing execution (MES) and statistical process control (SPC) systems, can provide real-time feedback and analysis on new product and mask designs. This real-time feedback provides the basis for utilizing the powerful and flexible design analysis system, which allows for rapid design of new, high-performance AWGs.
This analysis system, coupled with the flexibility of mask-based manufacturing, allows very rapid response to customer requirements. For example, AWGs can be designed and fabricated to match a desired channel plan. Typically, this involves selecting 40 channels on the ITU grid in the C-band. However, many systems use interleavers to combine two 40-channel 100-GHz spaced AWGs to produce 80 channels spaced at 50 GHz. This requires one standard C-band AWG and one offset by 50 GHz (for example, on the 50-GHz ITU grid), typically termed a C+ AWG. To implement the equivalent of a C+ AWG in thin-film filters or fiber Bragg gratings requires 39 new designs and 39 separate production runs.
FIGURE 5. A single, 40-channel VOA array/ multiplexer cascade plus test structures, can be fabricated on a 6-in. silicon wafer. The AWG can be seen clearly in the upper section of the device. Below that is a bright area resembling a block letter "P." This is the set of metal traces connecting the heater electrodes to the bonding pads near the bottom of the device. Because of the high density of metal lines, the individual lines are not resolved in this photo, except very near the heater electrodes.
A new AWG design for the C+ band design can be modeled automatically. This is accomplished by being able to collect and analyze product parametric data in a real-time mode and using this same data for process and equipment control out on the manufacturing floor by utilizing the MES system's integrated real-time SPC system. It can also be included on a mask containing other AWG designs. Similarly, the L-band can be covered by designing L and L+ AWGs, instead of 78 separate filters.
As an example of the design and manufacturing capability of PLCs, we consider the case of an OE, which had a requirement for S-band AWGs. The design was completed in one week, wafers were completed in two weeks, new test protocols were developed and implemented in one week, and parts were packaged in two weeks. Thus, having no previous experience with S-band components, it took six weeks from initiating the project to shipping finished products to a customer. Data results from an S-band AWG show the capabilities (see Fig. 4).
The flexibility of PLC manufacture is even more pronounced when implementing highly integrated devices. Research demonstrations of complex, multifunction devices monolithically integrated on the same substrate have already been reported.5, 6, 7
The first multifunction PLC likely to see widespread use is the cascade of a high-performance AWG filter with an array of thermo-optic VOAs for balancing the power levels among the individual channels. The primary use of such a device is to balance and/or predistort the amplitude spectrum to accommodate the response of subsequent amplifiers on a channel-by-channel basis. An engineering evaluation wafer containing a single 40-channel VOA-array/MUX cascade has been fabricated and tested (see Fig. 5).
As was the case with AWGs alone, mask-based PLC manufacturing allows the functions to be tailored to system needs. By changing the design cell on the mask, the channel plan or AWG performance in a VOA-multiplexer can be altered without changing the VOA performance. This flexibility allows the VOA multiplexer to be optimized for multiplexing or demultiplexing performance across the C-, L-, and S-bands (see Fig. 6).
G. Ferris Lipscomb is vice president of marketing and Marc Stiller is director of product marketing with Lightwave Microsystems, 2911 Zanker Road, San Jose, CA 95134. Ferris Lipscomb can be reached at email@example.com
- M. K. Smit, Elect. Lett. 24, 385 (1988).
- H. Takahashi, Elect. Lett. 26, 87 (1990).
- C. Dragonne, IEEE Phot. Lett. 3, 896 (1991).
- Y. Hida et al., IEE Elect. Lett. 36 (9) (April 27, 2000)
- K. Okamoto et al, IEE Elect. Lett. 32, 1471 (1996).
- C. R. Doerr et al, IEEE Phot. Tech. Lett. 11, 581 (1999).
- A. Ticknor and K. Purchase, Proc. SPIE Conf., 3949, Photonics-West 2000.