Birefringent crystals find new niche in WDM networks
Until now, low yields have been acceptable for most optical-crystal applications. New approaches to crystal growth and fabrication must now be coupled with manufacturing methodologies to deliver high volumes with unprecedented performance at previously unachievable prices.
The push to achieve faster and more-efficient WDM optical networks has been fueled by some dramatic breakthroughs in component technology. It is therefore ironic that some of the most critical network building blocks rely on one of the most simple and well-known optical phenomena: birefringence or "double-refraction," which is exhibited in optically anisotropic crystals such as calcite and yttrium vanadate.
In birefringent crystals, the refractive index varies depending on the direction of propagation and the electric field component (polarization). Light incident on the crystal will be split into two polarizations unless the beam is polarized along one of the crystal axes. Within the crystal, different polarizations (beams) are separated and "walk off" (see Fig. 1).
The ability to spatially displace beams of different polarizations is the crucial function that has led to widespread deployment of birefringent crystal-based components in WDM systems. The most basic birefringent crystal application is in an optical isolator, which acts as a one-way light valve to eliminate detrimental back reflections at sources and amplifiers. Optical circulators, another application, are versatile micro-optic devices that have application in a number of key network functions such as multiplexing and dispersion compensation. These nonreciprocal devices route light between different ports (see Fig. 2). Birefringent crystals also find applications in a range of other components, including polarizing beamsplitters and combiners, and certain designs of interleavers and gain equalizers.
In February 2001, the market research firm RHK reported that 1.6 million isolators and almost 200,000 circulators were sold in 2000. RHK also predicted a requirement of 8.4 million isolators and 1 million circulators in 2001. Although there has been a significant correction associated with the slowdown in network deployment, these data, combined with the requirement for birefringent crystals in other kinds of components, translate to a potential future requirement for tens of millions of crystal elements annually—three orders of magnitude higher volume than for solid-state lasers, the principal market for most optical-crystal manufacturers.
CRYSTALS FOR COMPONENTS
The key crystal properties for applications in fiberoptic components are reflected in the respective merits of potentially viable oxide crystal options (see Table 1). High transparency across a wide wavelength range and high optical uniformity are essential. From a component functionality standpoint, in a crystal with higher birefringence, the propagation distance required within the medium to achieve a prescribed walk-off separation between the two polarizations is lessened, allowing a significant reduction in component size. This is primarily why yttrium vanadate (YVO4; Δn = 0.204 at 1.55 µm) and rutile (TiO2; Δn = 0.256 ) are now being deployed in next-generation passive components at the expense of calcite (CaCO3; Δn = 0.156) and lithium niobate (LiNO3; Δn = 0.07).
Compared with calcite, the chemical and mechanical stability of rutile and yttrium vanadate also make these materials more suited to the rigors of fabrication into small elements, subsequent integration into tightly packed micro-optic assemblies, and, most important, durable performance in a range of environments. However, despite the importance of high physical strength, the mechanical hardness of rutile is very high, meaning that fabrication into small chips is more challenging and costly than with yttrium vanadate.
In the future, the optical-crystal industry will deliver tens of millions of crystal parts annually. Until now, neodymium: yttrium aluminium garnet (Nd:YAG) laser rods have been the highest volume product for most leading manufacturers, with some 20,000 units sold in 2000, according to Strategies Unlimited. Pricing will also be a critical driver, especially considering the significant price pressures many passive components are currently experiencing. For example, a comparable crystal element for miniature solid-state laser applications is priced 5 to 10 times higher than an equivalent part for an isolator. This observation, coupled with more stringent material and dimensional specifications required for emerging crystal applications, such as in interleavers, suggests that a paradigm shift will be required in the optical crystal manufacturing industry.
Ultimately, the cost and scalability of the manufacturing processes (crystal growth and crystal-element fabrication) involved will be the key determinants of which material succeeds. The Czochralski method is the most conventional and scalable manufacturing technique for growing large, high-quality single crystals. Semiconductor crystals such as silicon and gallium arsenide are grown by this method, where a small rotating seed crystal is "dipped" into a liquid melt and then slowly translated away (usually via computer control), gradually increasing the crystal diameter. Before stable crystallization can be realized, the ideal crystal growth environment/conditions (thermal geometry) must be created. These conditions are achieved by designing a high-temperature (typically <2000°C) furnace that must be specifically tailored to each crystal.
Maximizing revenue and profitability from a crystal-growth cycle is largely dependent on achieving the maximum diameter and uniformity (yield) across the crystal boule. Uniformity is dependent on the stability of the growth process (through precise computer control of furnace conditions) and by how well-designed the crystal growth furnace is in the first place. Translation of process understanding into furnace design is also the primary impediment to maximizing the crystal size and scaling the process.
Clearly, the approach to process development is vital. Most optical crystals have been developed over a long period of time through a more empirical methodology than that seen in the semiconductor industry. Silicon crystals are now grown in 12-in. diameters because of a more scientific and technology focused approach to process development and crystal size scale-up. Quantitative process scaling methodologies and excellent manufacturing disciplines have also been applied to produce high-quality optical crystals such as lithium niobate with a 5-in. diameter.
Yttrium vanadate is also produced using the Czochralski method. However, yttrium vanadate presents a rather different set of scaling challenges. The growth process has so far been developed through a largely empirical approach, and the standard YVO4 crystal size is approximately 1 in. in diameter. The current level of process control adopted by the leading YVO4 manufacturers is reported to be immature.
Rutile, an alternative to yttrium vanadate, is currently grown by the more unconventional Verneuil crystal-growth process. This process is even more challenging to scale up and generally results in lower quality material relative to crystal pulling.
Assuming it will at some point be possible to produce large, high-quality birefringent crystals, the next key cost and performance barrier will be fabrication of the crystal boule into small chips. Here the fabrication process involves x-ray orientation, slicing, lapping, polishing, coating dicing, cleaning, final testing, packaging and the associated handling through the various crystal steps. The elements are generally cuboid or trapezoid in shape, with two optically polished and antireflection-coated faces (see Fig. 3).
The optical functionality of the chip requires a very high geometric specification. In addition, the crystallographic axes must be aligned at specified angles to the polished faces, requiring precise orientation of the crystals using x-ray diffraction.
Because these chips will be closely packaged alongside other optical elements in a sealed component, it is essential that the physical dimensions and optical characteristics—flatness, parallelism, and axes orientation—of the chip are well-controlled. If the geometry is not correct, stress can result at interfaces between adjacent optical elements, producing wavefront distortions (higher bit-error rate) and higher polarization-dependent loss. This is a major issue for smaller trapezoidal chips, where control of the wedge face (angle, flatness, and orientation) is a key quality challenge, especially for components with epoxy-free designs.
Although these chips are already produced in relatively high volume, no current suppliers can offer manufacturing tolerances or even basic statistical process control data that would facilitate the transition to fully automated alignment and assembly for most crystal-based components. The current necessity to manually pair wedges and align with other optical elements is largely a consequence of inconsistencies in conformance to crystal chip-fabrication specifications (see "Standardizing optical testing and material specs," below).
The optical-crystal fabrication has evolved in a similar way to crystal growing. High-volume manufacturing processes have not been required, and most fabrication processes are highly skill-based. Most suppliers are relatively small in size, and many of the fabrication processes developed to manufacture low-volume, high-unit-value components such as laser crystals are not readily scalable. Since the unit fabrication cost has exceeded $100 for such laser products, achieving high yields has not been as important for crystal elements for passive components, where the unit price will eventually have to be about $1.
Trends in the fiberoptic component industry are driving a revolution in the fabrication of optical crystals with exceptionally challenging geometric specifications and new chip shapes (see Table 2). The new requirements have significant consequences for crystal-fabrication technology. For example, birefringent crystal waveplates will be deployed in interleavers. The target 1-µm length tolerance required on these cuboid crystal chips for interleavers is more than an order of magnitude higher than the current 50-µm tolerance for cuboid parts used in beam displacers. However, no optical-crystal fabrication technology currently exists to achieve this specification on a manufacturing basis. In addition, material quality (loss and uniformity) is a much more important characteristic for the interleaver cuboids because of the greater path length.
Significant technological and cost barriers need to be overcome to achieve future tolerances on wedge chips for higher-volume applications in isolators and circulators. The interleaver crystal waveplates are currently priced at more that 10 times that of the wedge parts. Although there is a proportionately lower volume of crystal, the key cost driver is fabrication.
Clearly, there are several technical challenges resulting from the reduction in aperture. The transition from 2 mm2 to 1 mm2 is a natural progression to shrink component size and to enable compact integrated components such as isolator arrays. However, handling these tiny elements becomes exceptionally problematic. Since these parts have an angled face, it is not possible to simply dice from a wafer of prescribed thickness. Multiple fabrication (and handling) process steps currently have to be performed after they have been formed into their basic shape. The fundamental and obvious handling difficulties with these small elements make it even more difficult to achieve and maintain the new, very challenging length, angle, and orientation specifications.
Despite the challenges, technological solutions will be found. The application of world-class manufacturing disciplines from other industries such as semiconductors and optical disk drives will be essential. Fusion of these disciplines with the innovative engineering expertise that does exist in current low-volume optical-crystal fabrication is a prerequisite.
In the future, the main boundary will be cost per unit to fabricate the small, wedge-shaped chips. Given the unparalleled volumes predicted, it's hard to imagine how manual handling can result in a scalable, robust manufacturing solution, leading to the unavoidable conclusion that automated handling will have to be implemented for certain process steps.
Certainly, the application of existing automation technology will not be routine. A critical hurdle that must be overcome is to avoid scratching and chipping of these rather fragile elements. Assuming a solution can be found, the ability to automatically fabricate, test, and handle these chips presents the exciting opportunity of integrating crystal fabrication and at least the first steps of component assembly into a single manufacturing line. The incentive is clear: to drive costs out of the entire optical-component manufacturing process while at the same time increasing yields.
John Nicholls is CEO of Photonic Materials, Strathclyde Business Park, Bellshill, ML4 3BF, Scotland. He can be reached at firstname.lastname@example.org.
Standardizing optical testing and material specs
Test standardization is a priority for all optical components and systems, and robust performance criteria, such as Telcordia 1221, have been developed. However, a clear disconnect exists between component and crystal specifications. Translating standard component metrics such as insertion loss and polarization-dependent loss (PDL) back to more basic and routinely measured material quality properties such as stress birefringence, absorption and scattering loss, wavefront distortion, and extinction ratio is a problem. The fact that material specifications for the same birefringent crystal chip vary enormously from vendor to vendor is probably related to this disconnect.
To drive down costs, it will be essential for crystal and component manufacturers to work together to target which material or fabrication specifications most strongly impact component performance. The next step is to develop standard crystal-testing methodology, and then finally to automate this as part of the fabrication process. In general, a key differentiator in the future will be the ability to guarantee crystal-chip performance compliance through reliable test methods.