Polarization control and manipulation are crucial to the design and the operation of optical devices. The polarization sensitivity is in general an undesirable effect because it causes a polarization-dependent wavelength shift and phase shift in all interferometric devices, such as spectrometers, Mach–Zehnder interferometers, and ring resonators. One common solution is to separate the two orthogonally polarized TE and TM components and process them individually. Polarization splitters are key elements in this approach. They can find applications in signal processing, network monitoring, polarimetric sensing, imaging, and data storage. Different configurations of polarization splitters such as the Y-branch [1], multimode interference coupler [2], directional coupler [3], Mach–Zehnder interferometer [4, 5, 6], Bragg grating [7], photonic crystal [8], and arrayed waveguide grating (AWG) [9] have been reported. All of these devices have limitations in the fabrication tolerances, in the operating bandwidth, or in the density of integration.

In this Letter, we report the design, fabrication, and experimental results for a novel passive polarization splitter based on an AWG configuration using a silicon-on-insulator (SOI) platform. AWGs have been used before as polarization splitters [9]. The novelty of this device lies first in the zero-order grating configuration to eliminate the wavelength dependence. Second, cladding stress engineering is used for polarization control. The stress created by the upper oxide cladding layer effectively modifies the overall waveguide birefringence. Consequently, this method offers the freedom to decouple the birefringence constraints from the waveguide geometry design. Waveguides can be designed to optimize their insertion loss, coupler performance, and bend loss, independent of the birefringence requirements. Finally, the high index contrast between the waveguide core and cladding layers in a SOI system ensures the compact size of the devices. To the best of our knowledge, this device is the first reported wavelength-independent SOI polarization splitter.

The cross-section of a typical ridge SOI waveguide structure is shown in Fig. 1a . The waveguide consists of a silicon ridge core $(n=3.476)$ and a buried ${\mathrm{SiO}}_2$ cladding $(n=1.444)$ layer. An upper ${\mathrm{SiO}}_2$ cladding layer with a thickness $t_c$ and a film stress ${\sigma }_{\mathrm{film}}$ is the key element for birefringence modification. The parameter ${\sigma }_{\mathrm{film}}$ defines the in-plane stress component that is present in a two-dimensional uniform thin oxide film. The detailed investigation of the stress engineering for SOI waveguides can be found in our recent studies [10, 11]. In this approach, the stress birefringence created by a cladding layer can be used to compensate for the existing geometrical birefringence in an SOI waveguide and modify the overall modal birefringence. The cladding stress is due to a combination of grown-in stress that depends on deposition conditions and thermal expansion mismatch among the waveguiding and the cladding layers. Since the thermal expansion coefficient of silicon is approximately seven times higher than that of ${\mathrm{SiO}}_2$, this results in a compressive stress in the oxide layer as the deposited oxide film cools down to the room temperature on a silicon wafer. Through the photoelastic effect, stress modifies the material refractive index of the silicon core, thus affecting the overall effective index of the mode.

The schematic layout of a zero-order AWG-based polarization splitter is shown in Fig. 1b. This device consists of an input waveguide that is coupled to a waveguide array through a slab waveguide free propagation region (FPR). The array is then coupled to the output waveguides through a second FPR. Unlike a conventional AWG, all of the arrayed waveguides have identical path lengths. Only two output waveguides are used, one to capture TE-polarized light and the other to capture the TM-polarized light.

Incident light, with an intensity of $I_{\mathrm{in}}$, enters the input FPR region and is coupled into the array waveguides. A triangular patch of oxide cladding is selectively deposited in the arrayed waveguide section, as indicated in Fig. 1b. The stress in the waveguide cladding induces a polarization-dependent phase difference in the light beams propagating in the waveguide grating section. Since all waveguides have the same length, the phase change between adjacent waveguides in the arrayed section depends only on the cladding stress and the patch-length increment. This stress-induced phase shift $\left(\mathrm{\Delta }\varphi \right)$ for each polarization is

At the output FPR combiner, the light propagating in the waveguide array recombines and converges to a focal point on the image plane. Each polarization experiences a different stress-induced phase shift, resulting in a spatial separation between the focal points of the outgoing TE- and TM-polarized light beams. Since the stress-induced index change in the TE and TM polarization modes has opposite signs for silicon waveguides, the two modes are displaced in opposite directions relative to the FPR centerline, as shown in the inset of Fig. 1b. The lateral separation d between the TE and TM output ports and the reference can be calculated as follows:

Since the waveguide array has zero order, the polarization beam displacement given by Eq. (2) does not depend on wavelength. Hence unlike other recently reported SOI polarization splitters [3, 5, 6], the zero-order AWG devices should achieve wavelength-independent polarization splitting.

Our devices were designed and fabricated on SOI $⟨100⟩$ wafers, with a $2.2\mathrm{\mu }\mathrm{m}$ thick silicon core layer and a $0.40\mathrm{\mu }\mathrm{m}$ thick buried oxide layer. The ridge waveguides were first patterned by wet etching through an optical mask with a nominal width of $1.5\mathrm{\mu }\mathrm{m}$ and a target etch depth of $1.5\mathrm{\mu }\mathrm{m}$. After the initial processing of the devices, a silicon dioxide $\left({\mathrm{SiO}}_2\right)$ upper cladding layer was deposited by plasma-enhanced chemical vapor deposition at $390°\mathrm{C}$. The deposited oxide cladding film had a thickness of $1\mathrm{\mu }\mathrm{m}$, and the cladding film stress was measured to be approximately $-340\mathrm{MPa}$. The oxide cladding was then patterned and etched to produce the desired patch size.

Our fabricated zero-order AWG-based polarization splitters have 100 waveguide channels in the arrayed grating section, with 13 inputs and 17 output waveguides. The overall device size is $\sim 12\mathrm{mm}\times 4\mathrm{mm}$. Figure 2 shows the fabricated AWG devices with triangular cladding regions of varied patch sizes. With the deep etch depth, the ridge waveguides themselves are not single mode. However, the waveguide bends in the arrayed section (with a minimum bending radius of $500\mathrm{\mu }\mathrm{m}$) provide an efficient way to filter out the higher-order modes. Based on the measured cladding film stress and thickness, the cladding patch-length increment $\mathrm{\Delta }L$ was set to $16.4\mathrm{\mu }\mathrm{m}$ to produce the required polarization-dependent phase shift at the output of the array section. The stress-induced index changes, $\delta n_{\mathrm{stress}}^{\mathrm{TE}}$ and $\delta n_{\mathrm{stress}}^{\mathrm{TM}}$, are usually of the order of ${10}^{-4}$ to ${10}^{-3}$, which introduces a small diffraction order to the AWG. Assuming that the patch-length increment between two adjacent waveguides is $\mathrm{\Delta }L=16.4\mathrm{\mu }\mathrm{m}$ and the center wavelength is ${\lambda }_0=1550\mathrm{nm}$, the AWG would have a stress-induced order of $m=\delta n_{\mathrm{stress}}\mathrm{\Delta }L∕{\lambda }_0<0.011$. Since stress-induced effects are independent of wavelength, and the AWG operates close to zero order, this splitter is expected to be almost entirely independent of the operating wavelength, providing an ultrabroadband polarization splitting function.

A tunable laser with a wavelength range of 1460 to $1580\mathrm{nm}$ was used to scan the input wavelengths. The output of the laser was fed through a polarization-maintaining fiber to a half-wave plate such that the input beam could be polarized in either the TE or the TM direction. The beam was then coupled into the input facet of the test device. The output light was collected with another single mode fiber with a tapered tip and the power was measured by a photodetector. The on-chip loss for these splitters is approximately $-8\mathrm{dB}$. Similar losses were obtained for single ridge waveguides, indicating that this is almost entirely due to waveguide loss from sidewall scattering. Four measurements were performed for each AWG device: the TE outputs in channels I and II (see Fig. 1) with a TE input and the TM outputs in channels I and II with a TM input.

The extinction ratio of the polarization splitter was always better than $-10\mathrm{dB}$ for both output polarizations over the entire tuning range of our laser (1465 to $1580\mathrm{nm}$). Figure 3 shows the measured transmission data for a functioning polarization splitter from $\lambda =1537$ to $1557\mathrm{nm}$. Over this range, the device exhibited at least $-15\mathrm{dB}$ splitting extinction ratio. The highest extinction ratio achieved in Fig. 3 was $-20\mathrm{dB}$. The measured waveguide spatial separation between the output TE and TM polarizations ($d_s^{\mathrm{TE}}$ and $d_s^{\mathrm{TM}}$) in the device presented in Fig. 3 was $6.0\mathrm{\mu }\mathrm{m}$. Using Eq. (2), the resulting stress-induced index changes, $\delta n_{\mathrm{TE}}$ and $\delta n_{\mathrm{TM}}$, are $-1.1\times {10}^{-3}$ and $+1.5\times {10}^{-3}$, respectively, in good agreement with the simulations.

We have described and demonstrated what we believe to be the first wavelength-independent polarization splitter based on a zero-order AWG configuration with SOI ridge waveguides. We used cladding stress to generate the polarization-dependent phase shifts required to achieve polarization splitting at the output channels. In the fabricated device the polarization extinction ratio was always better than $-10\mathrm{dB}$ for both polarizations over a broad wavelength range (1465 to $1580\mathrm{nm}$). The main advantage of this device is that fabrication is relatively simple. The device consists only of standard ridge waveguides, and waveguide shape is not critical. The only critical design parameters are the cladding patch dimensions and the lateral placement of the two output waveguides along the FPR image plane. These parameters are easily calculated once the oxide cladding stress and thickness are known. The achievable polarization extinction ratio is determined first by the intrinsic cross-talk of the AWG. Commercial AWG devices regularly achieve cross-talk values better than $-30\mathrm{dB}$. The second limit is set by the hybrid nature of the waveguide modes. Each mode supports a small amount of power in the orthogonal polarization to the primary polarization direction. However, the minor polarization component should be negligible due to the deeply etched ridge waveguides used throughout the study. The current device configuration was chosen for ease of design and fabrication. The device size can be reduced by a factor of two simply by optimizing the device layout, and much smaller devices may also be obtained using smaller silicon waveguides.

 

Fig. 1 Cross-section of a typical SOI waveguide. (b) Schematic layout of a broadband zero-order AWG polarization splitter. Inset, geometry of the output FPR with a typical Rowland mounting configuration.

Download Full Size | PDF

 

Fig. 2 Left, a top view of our fabricated wavelength-independent zero-order AWG polarization splitter. Right, a scanning electron microscope image of the oxide cladding.

Download Full Size | PDF

 

Fig. 3 Measured wavelength dependence of a zero-order AWG-based polarization splitter.

Download Full Size | PDF

1. Y.-P. Liao, R.-C. Lu, C.-H. Yang, and W.-S. Wang, Electron. Lett. 32, 1003 (1996). [CrossRef]  

2. J. M. Hong, H. H. Ryu, S. R. Park, J. W. Jeong, S. G. Lee, E.-H. Lee, S.-G. Park, D. Woo, S. Kim, and B.-H. O, IEEE Photon. Technol. Lett. 15, 72 (2003). [CrossRef]  

3. I. Kiyat, A. Aydinli, and N. Dagli, IEEE Photon. Technol. Lett. 17, 10 (2005). [CrossRef]  

4. L. B. Soldano, A. H. de Vreede, M. K. Smit, B. H. Verbeek, E. G. Metaal, and F. H. Groen, IEEE Photon. Technol. Lett. 6, 402 (1994). [CrossRef]  

5. T. K. Liang and H. K. Tsang, IEEE Photon. Technol. Lett. 17, 393 (2005). [CrossRef]  

6. W. N. Ye, D.-X. Xu, S. Janz, P. Waldron, and N. G. Tarr, in Proceedings of IEEE Group IV Photonics’06 (IEEE, 2006), p. 249.

7. L. L. Soares and L. Cascato, Appl. Opt. 40, 5906 (2001). [CrossRef]  

8. L. Zhang and C. Yang, IEEE Photon. Technol. Lett. 16, 1670 (2004). [CrossRef]  

9. A. R. Vellekoop and M. K. Smit, J. Lightwave Technol. 8, 118 (1990). [CrossRef]  

10. D.-X. Xu, P. Cheben, D. Dalacu, A. Delâge, S. Janz, B. Lamontagne, M. Picard, and W. N. Ye, Opt. Lett. 29, 2384 (2004). [CrossRef]   [PubMed]  

11. W. N. Ye, D.-X. Xu, S. Janz, P. Cheben, M.-J. Picard, B. Lamontagne, and N. G. Tarr, J. Lightwave Technol. 23, 1308 (2005). [CrossRef]