The typical propagation loss of ~5-10 dB/m in a planar silica waveguide is nearly five orders-of-magnitude larger than that in low loss optical fibers. This wide gap in loss performance has prevented planar waveguide technology from bringing potential improvements in stability, footprint, energy consumption, and fabrication cost to photonic applications requiring long propagation lengths or ultra-high-quality-factor resonators. In this letter, we report our progress toward fiber-like losses with a planar silica waveguide technology. We have demonstrated world record low loss below 0.1 dB/m, which makes it feasible to put hundreds of nanoseconds of delay and planar resonator structures with quality factors greater than 100 million on a chip. By grouping these structures with the passive and active building blocks of a planar integration platform, complex systems such as optical gyroscopes, true-time-delay networks, and optical buffers can be fabricated on a silicon substrate.
     Fig. 1 (left) shows three planar waveguide spiral delays with 3, 1, and 2 m lengths in series on a chip. Fig. 1 (right) gives a timeline of our progress toward fiber-like losses in such chips. Scattering loss due to sidewall roughness is a large contributor to the total propagation loss in planar waveguides. The timeline demonstrates how scattering loss has been reduced by using a very thin and wide, or high-aspect-ratio, Si₃N₄ core geometry. As we reduced scattering losses, other contributions such as substrate leakage and hydrogen impurity absorption losses became important. These contributions were reduced by using 15-μm-thick thermally grown SiO₂ in both the top and bottom cladding layers. As shown in Fig. 2, the core is deposited directly onto the bottom cladding layer. The top thermal SiO₂ cladding layer is transferred via wafer-bonding. In addition to low loss, waveguides with wafer-bonded upper claddings also have lower dielectric film stress than those with thick PECVD upper claddings. They can also be fabricated more quickly since the thick thermal oxide wafers can be batch-processed in parallel with the other waveguide processing.

Figure 1. (Left) Three spiral planar waveguide delays in series on a chip.(Right) Timeline of our progress toward fiber-like propagation loss on a chip. The picture shows a 20x1-meter spiral delay line with red laser light injected.

Figure 2. Schematic cross-section of the planar waveguides, an SEM micrograph of a waveguide facet, and simulated fundamental TE and TM modes.	Figure 3. (Top) A planar waveguide coil with red laser light injected. (Bottom) Reflectivity with respect to propagation length measured for infrared light.

     Fig. 2 shows schematic and SEM cross-sections of the buried waveguide structure. The stoichiometric core has a fixed index contrast of ~25% with the SiO₂ cladding. The core dimensions are tuned from 40 to 100 nm thick and from 2 to 15 μm wide, depending on the desired minimum bending radius of the waveguide. The high-aspect-ratio core geometry minimizes the total modal intensity at the sidewall core/cladding interfaces to reduce scattering loss. Simulated fundamental TE and TM mode profiles are also shown in the bottom of Fig. 2. These demonstrate the low core confinement and large effective area of the modes. They also show that the core geometry yields a large birefringence. The large birefringence and the resulting polarization dependent loss are useful for achieving a large (> 35 dB) polarization extinction with compact polarizers and spiral waveguide delays. These benefit systems that require single polarization operation for high performance, such as optical gyroscopes.

Figure 4. Spectral response (top) and layout (bottom) of a 16-channel AWG (de)multiplexer.

     Fig. 3 shows red laser light propagating in a planar waveguide Archimedean spiral. The spiral begins with a bending radius of 10 mm, decreasing by 0.5 mm per round trip before ending at a bend radius of 0.5 mm. The total propagation distance in the structure is around 0.45 m per waveguide in a ten-waveguide bus. Since scattering loss scales proportionally with λ^–4, the small relative decrease in red light intensity over 0.45 m of propagation is a qualitative indicator of low propagation loss. The data in Fig. 2 shows the reflectivity measured with respect to propagation distance into the waveguide spiral for infrared light near 1550 nm. A linear taper is used at the beginning of the spiral in order to avoid the excitation of any higher order modes that could influence the data. One can extract information about a waveguide’s bending capability by fitting a bend radius dependent equation to the data. Since bending loss increases exponentially with respect to the bending radius, it is important to know the radius at which bend loss begins to dominate the total propagation loss.
     Fig. 4 shows the spectral response of a 16-channel, 200-GHz-channel-spacing arrayed-waveguide grating (AWG) (de)multiplexer demonstrated with the ultra-low-loss waveguide technology. The layout of the 2 cm × 1.5 cm chip is also shown. The total on-chip loss of the device is below 0.5 dB. Detailed information is given in reference [4].

Figure 5. Spectral response of planar ring resonators having Q as high as 30 million.

     Fig. 5 shows the transmission spectra for various planar ring resonators measured near 1060, 1310, and 1550 nm wavelengths [5]. Intrinsic Q-factors as high as 30 million are obtained from fits to the spectra. The data demonstrates the transparency of the silica waveguide technology, allowing for device operation over a large wavelength range from the infrared to the visible.
     These low loss waveguides may find applications for stabilized low linewidth lasers, low jitter mode locked lasers, integrated optical gyroscopes, and true time delay elements for phased array antennas. The high-Q resonators can be used for ultrastable microwave oscillators. Our goal is to integrate this low loss waveguide technology with active elements, including lasers, modulators and photodetectors.

Acknowledgements
The authors thank Scott Rodgers, Bill Jacobs, James Adleman, and Di Liang for helpful discussions. This work is supported by DARPA MTO under iPhoD contract No: HR0011-09-C-0123. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing official policies of the Defense Advanced Research Projects Agency or the U.S. Government.

References

1. J. F. Bauters, M. J. R. Heck, D. John, D. Dai, M.-C. Tien, J. S. Barton, A. Leinse, R. G. Heideman, D. J. Blumenthal and J. E. Bowers, Opt. Expr. 19, 3163 (2011).
   
2. J. F. Bauters, M. J. R. Heck, D. John, D. Dai, M.-C. Tien, J. S. Barton, A. Leinse, R. G. Heideman, D. J. Blumenthal and J. E. Bowers, “Ultra-low-loss Single-mode Si₃N₄ Waveguides with 0.7 dB/m Propagation Loss”, European Conference on Optical Communications, Geneva 2011.
   
3. J. F. Bauters, M. J. R. Heck, D. D. John, J. S. Barton, C. M. Bruinink, A. Leinse, R. G. Heideman, D. J. Blumenthal and
   J. E. Bowers, “Ultra-Low-Loss (<0.1 dB/m) Planar Silica Waveguide Technology”, IEEE Photonics Conference, Arlington, 2011.
   
4. D. Dai, Z. Wang, J. F. Bauters, M.-C. Tien, M. J. R. Heck, D. J. Blumenthal and J. E. Bowers, Opt. Expr. 19, 14130 (2011).
   
5. M.-C. Tien, J. F. Bauters, M. J. R. Heck, D. T. Spencer, D. J. Blumenthal and J. E. Bowers, Opt. Expr. 19, 13551 (2011).

Martijn Heck, Jared Bauters, Demis John and John Bowers (l to r) contributed the winning image and are pictured with the measurement system.