Crystal growth techniques underlie component performance
Lithium niobate and yttrium vanadate are critical materials for many active and passive WDM components. Crystal manufacturers and component designers must understand the current capabilities and limitations of these materials and develop methods to extend their usefulness.
Unprecedented demand for optical materials, driven by rapid expansion in the optical telecommunications network, creates a paradigm shift for the optical-crystals industry. Many of the materials used in telecom components have been commercially available for many years and in many applications. Until recently, each of these applications has had unique performance specifications and generally required a relatively low volume of elements. Market demands for optical crystals could be met with only a moderate number of crystal growth stations. Individually handling, sorting, and selling crystals relative to their operational demands could offset variability in the quality of the resulting crystals. Low yields were offset by higher prices, supported by relatively small numbers of high-value products or systems that employ these optical crystals.
In contrast, the crystalline optical elements to be used in the telecommunications components now being designed and manufactured must be produced in quantities of hundreds of thousands with resulting crystal sales in tens of millions of dollars for each component. However, for this market to be fully realized, optical crystals and the telecom components employing these crystals must be manufacturable in large volume at reasonable costs.
FIGURE 1. Typical commercially available YVO4 single crystal boule. The top (seed end) and bottom of the crystal have been oriented and ground (perpendicular to the c-axis) removing 3 to 4 mm of material from each end. Although the crystal shown has been annealed, yellow coloration remains near the seed due to vacancies and/or reduced vanadium.
A key prerequisite to large-scale manufacturing is the availability of adequate quantities of optical crystals of sufficient quality and regularity. Equally important is the development of components and scalable manufacturing techniques that realistically consider the limitations of currently available crystals and anticipate the progress in crystal supply and quality. Two of the most widely used optical crystals in telecom products are lithium niobate (LiNbO3) and yttrium vanadate (YVO4).
Lithium niobate has been used for many bulk optic and waveguide applications. The broad application of LiNbO3 is primarily due to its large electro-optic effect (r33 = 30.8 pm/V), nonlinear optic effect (d33 = -33 pm/V), and moderate birefringence (ne=2.15, no=2.20). It has a high optical transmission over a broad spectral range (0.45 to 4.5 µm) and is thermally, chemically, and mechanically stable. In addition to being electro-optic, nonlinear optic, and birefringent, LiNbO3 is also acousto-optic, photoelastic, photorefractive, pyroelectric, piezoelectric, and ferroelectric. These additional properties can be advantageous or deleterious depending on the application.
FIGURE 2. YVO4 viewed along the z-axis from the seed end. Crystal is the same as shown in Fig. 1. Note the horizontal crack perpendicular to the x-axis.
Lithium niobate boules are usually grown along the crystal's z-axis and can be up to 15 cm in diameter and several tens of cm in length. Hundreds of tons of lithium niobate are produced each year and 3- and 4-in. wafers are readily available. However, the vast majority of this material is used in surface acoustic wave (SAW) devices and is of insufficient optical quality to be used for optical applications. One reason is that particular care must be taken in the growth of optical grade LiNbO3 to reduce scatter and stress-induced changes in refractive index.
Lithium niobate crystals are most commonly grown using the Czochralski technique from a melt composed of 48.6% Li2O and 51.4% Nb2O5.1 This composition is referred to as the congruent composition and is chosen because the melt freezes to form a crystal of the same composition. Congruently grown LiNbO3 crystals are deficient in Li ([Li]/[Nb] = 0.94) and give rise to Li vacancies, Nb anti-site (Nb ions on Li sites), and oxygen vacancies. These intrinsic stoichiometry-related point defects, along with extrinsic impurities (most notably iron contamination), give rise to undesirable crystalline properties including multidomains, high internal stress, and photorefractive damage susceptability (light induced change in refractive index).
Photorefractive damage is more commonly observed at wavelengths shorter than 1 µm. However, the high optical intensity within a waveguide can lead to photorefractive damage at optical powers as low as 100 to 200 mW (100 MW/cm2). The susceptibility to photorefractive damage is significantly reduced by adding 5% MgO to the melt.2
Czochralski-grown crystals can be made single domain after growth by applying an electric field along the crystal's z-axis at elevated temperatures. While poling can produce single-domain material, the procedure can introduce cracks and/or scattering centers. Complete poling of heavily magnesium oxide (MgO) doped crystals is more difficult because of the increase in Curie temperature and inhomogenous distribution of Mg dopant.
FIGURE 3. Near field Maltese cross observed along z-axis in YVO4 boule. For a uniform crystal, the pattern should consist of perfectly circular rings centered on a well defined cross.
Lithium niobate crystals also have been grown using solution growth techniques from melts with excess Li2O (58% Li2O -42% Nb2O5; [Li]/[Nb] = 1.38) and from solutions with stoichiometric compositions ([Li]/[Nb] = 1) in an 11% K2O flux.3 Solution growth has produced nearly stoichiometric crystals with [Li]/[Nb] ratios as high as 0.994. As expected from point defect models, crystal stoichiometry is strongly related to photorefractive damage in LiNbO3 crystals. For near-stoichiometric lithium niobate crystals, the threshold for MgO doping required to dramatically reduce the susceptibility to photorefactive damage is less than 2% (in contrast to 5% for congruently grown crystals).3, 4
A significant advantage of stoichiometric crystals is that they are typically single domain after growth and thus require no post-growth poling, leading to a significant reduction in the fabrication costs.5 One of the most common drawbacks of solution growth is the changing composition of the melt as the crystal grows. The effect of the changing melt composition can be reduced by crystallizing only a small fraction of the melt (<10%). As a result, very large melts are required to produce large single-crystal boules. An alternative is to add Li2O and Nb2O5 to the melt as the crystal grows, thereby replacing the Li and Nb removed from the melt by crystallization. While this technique eliminates changing melt compositions it adds significant complexity to the growth system and procedure.
LITHIUM NIOBATE COMPONENTS
A variety of telecom components have been designed and produced using waveguides in lithium niobate, including modulators, routers, splitters, and multiplexer/demultiplexers, mostly based on Mach-Zehnder interferometer or directional coupler configurations. Waveguides in LiNbO3 are most commonly produced by either titanium (Ti) in-diffusion or annealed proton exchange (APE) Several devices are fabricated on a single wafer using standard photolithographic techniques. The wafer is then processed to form the waveguides before dicing and polishing into individual elements. The choice of x-, y-, or z-cut wafers and propagation direction depends on the choice of in-diffusion or APE process and on the application. Waveguide width must be maintained to +0.1 µm, thus surface flatness of wafers is critical to controlling waveguide width during photolithography.
The in-diffusion of Ti is performed at 950°C to 1050°C and care must be taken to prevent the out-diffusion of Li2O from the surface, which results in an unwanted planar waveguide for z-polarized light. The diffusion process temperature is only slightly below the Curie temperature (1150°C for undoped congruent crystals, and 1190°C to 1200°C for undoped nearly stoichiometric crystals). As a result of the high process temperature additional precautions must be taken to prevent ferroelectric domain reversal during diffusion. Diffusion rates can differ depending on the crystal stoichiometry and process times may vary depending on growth process and/or crystal source.
The APE process is a much lower temperature process with proton exchange conducted in a hydrogen rich source (usually benzoic acid) at temperatures of 150°C to 250°C followed by thermal annealing at 350°C to 400°C. APE waveguides support only the extraordinary polarization and are therefore polarizing waveguides.
Several novel devices have been proposed and demonstrated based on recent advances in periodic poling of lithium niobate, including optical switches, routers, multiplexer/demultiplexers, broadband amplifiers, wavelength-channel switches, and polarization-mode dispersion compensators. Periodically poled structures have been produced in bulk crystals as well as both Ti-diffused and APE waveguides. Stoichiometric LiNbO3 is particularly attractive in these applications because of its much lower poling voltage as compared to congruent LiNbO3 (4 kV/mm for stoichiometric vs. 16 to 21 kV/mm for congruent).
Yttrium orthovanadate (YVO4) is a birefringent crystal that is well suited for polarizing components and beam displacers. Compared to calcite, YVO4 has higher birefringence (DnYVO4 = 0.204, Dncalcite = 0.156 at 1.55 µm), is more mechanically and chemically stable, and is easier to polish than the relatively soft calcite. While rutile has a higher birefringence it is generally more expensive and is very difficult to polish because of its hardness, approaching that of sapphire.
Yttrium orthovanadate is most commonly grown using the Czochralski technique. The congruent melting temperature is approximately 1810°C, requiring RF induction heating of iridium crucibles in an inert atmosphere. Boules typically measure 30 mm in diameter by 35 to 50 mm in length. While YVO4 is becoming more readily available, the quality and uniformity of currently available crystals vary greatly.
FIGURE 4. Far- field Maltese cross in YVO4 shows distortions in birefringence over short distances.
The coloration of YVO4 changes from boule to boule but is most commonly observed as an uneven yellow coloration that is darkest near the seed end of the boule. The origin of this coloration has not been definitively identified but is generally thought to arise from nonstoichiometry (vanadium and/or oxygen vacancies) or reduced vanadium (V5+ --> V4+, V3+).6, 7
The coloration can be lessened by annealing in an oxidizing atmosphere at temperatures between 500°C and 1500°C. While the crystals are more transparent after post-growth annealing, a coloration gradient persists with darker coloration noted at the seed end of the boule and a radial gradient with darker coloration at the center (see Fig. 1).
Another characteristic defect observed in all Czochralski-grown YVO4 boules evaluated is a hazy, smoke-colored region near the center of the boule that exists along the entire z-axis of the crystal. Increased scattering and other nonuniformities due to this effect may limit the amount of usable material volume for certain applications (see Fig. 2).
The yield of usable optical elements from YVO4 boules is also limited by cracking of the crystals that routinely occurs during fabrication. Cracks usually initiate near the seed and form planes parallel to the principal planes (perpendicular 100 and 010 axes). Cracking presumably results from stress due to varying stoichiometry. Stress-induced cracking may limit the growth of larger boules as longer growth time required is likely to produce greater stoichiometric variation.
Inhomogeneities can also be observed in the birefringence of YVO4 crystals, which can seriously affect device performance. For most applications, the birefringence of the material cannot vary more than 1%. Some current telecom devices under development demand a birefringence uniformity greater than 0.1%.
The birefringence inhomogeneity can be imaged or quantified using a birefringence interferometer. Birefringence inhomogeneities can also be observed qualitatively by noting irregularities in the Maltese cross produced when the crystal is placed between crossed polarizers and viewed along the c-axis (see Fig. 3). Figure 3 shows typical Maltese cross patterns observed in the near field. Ideally, the patterns should consist of perfectly circular rings centered on a well-defined cross. When viewed in the far field, the center of the cross can be made so large that it nearly covers the entire area of the crystal. Sharp irregularities in birefringence are easily seen in the far-field Maltese cross (see Fig. 4).
Although less developed than Czochralski growth, YVO4 has been grown using solution growth techniques with encouraging results.8, 9 Solution growth of YVO4 involves use of a melt consisting of a mixture of YVO4 and a suitable solvent, typically lithium metavanadate (LiVO3). Yttrium orthovanadate can be crystallized from such a solution over a wide temperature range, but generally hundreds of degrees below the melting point of pure YVO4. Because it is grown at significantly lower temperatures, the resulting crystals are much closer to ideal stoichiometry than are Czochralski-grown boules, and are likely to have far less inhomogeneity. However, the solution growth process is considerably slower than the Czochralski method, with growth times being measured in several weeks rather than several days.
The telecommunications industry places new demands on optical crystals, primarily with regard to high volume production of uniform high-quality material. This is reminiscent of the early days of the semiconductor industry and the increased demands on silicon crystal growth. Just as silicon crystal growth techniques advanced to establish high-quality, large-diameter silicon wafers at near commodity status, crystal growth of optical materials such as lithium niobate and yttrium vanadate must similarly advance to capitalize on their potentials
- P. F. Bordui, R. G. Norwood, C. D. Bird, G. D. Calvert, J. Crystal Growth 113, 61 (1991).
- D. A. Bryan, R. Gerson, H.E. Tomaschke, Appl. Phys. Lett. 44, 847 (1984).
- Y. Furukawa et al., J. Crystal Growth 211, 230 (2000).
- Y. Furukawa et al., Opt. Lett. 23, 1892 (1998).
- V. Gopalan, T. E. Mitchell, Y. Furukawa, K. Kitamura, Appl. Phys. Lett. 72, 1981 (1989).
- Y. Nobe, H. Takashima, T. Katsumata, Opt. Lett. 19, 1216 (1994).
- S. Erdei and F. W. Ainger, Mat. Res. Symp. Proc. 329, 245 (1994).
- S. Erdi, J. Crystal Growth 134, 1 (1993).
- T. Katsumata, H. Takashima, H, Ozawa, K. Matsuura, Y. Nobe, J. Crystal Growth 148, 193 (1995).
Mike Scripsick is president of Nova Phase, 435 Route 206, PO Box 366, Newton, NJ 07860-0366. He can be reached at 973-300-3065 or firstname.lastname@example.org