Integrated polarization rotator for ultra-high-speed optical communication systems

Modulation schemes based on polarization multiplexing are a very promising approach to providing the data rates required by the next generation of optical communication systems (100 Gigabit Ethernet, or 100GE). Polarization multiplexing allows for the doubling of spectral efficiency when compared to conventional schemes, as information is sent on both transversal electric (TE) and transversal magnetic (TM) polarizations. A complete receiver for this kind of modulation can be divided into three subsystems: the phase diversity circuitry, the balanced photodetectors, and the polarization diversity circuitry. Each of these subsystems has different technological requirements that are not easily matched together. This is the reason why, in many practical designs, photonic subsystems are fabricated using different technologies and integrated together by fiber interfaces. Monolithic integration of the full receiver, although challenging, aims to reduce the number of costly package connections and interfaces.

Monolithic frontends comprising phase diversity and photodetector subsystems have already been demonstrated,¹ but integrating the polarization diversity circuitry (polarization rotators and splitters) remains an issue. In the pursuit of full receiver integration, many passive polarization rotators have been proposed. These rotators are based on waveguide geometries that support hybrid modes, which are tilted with respect to the canonical TE (and TM) polarization in the interconnection waveguides. These proposed rotators are based on waveguide geometries that have one² or two³ slanted walls, or are asymmetric waveguides with one⁴ or two⁵ trenches. Here, we discuss how to implement a polarization rotator that can be integrated with a demonstrated monolithic indium phosphide (InP) frontend for coherent 100GE,¹ including phase diversity and photodetectors. The most typically implemented rotator for InP circuits is that based on a waveguide with one slanted wall. A receiver including this rotator requires three etching steps: one dry etch for the interconnection waveguides, one dry etch step for the vertical wall of the waveguide rotator, and one wet etch step for the slanted wall. In this work, we propose a rotator based on a two-step waveguide that allows us to fabricate the full receiver using just two dry etch steps.⁶

Figure 1. Polarization rotator: 3D structure and transversal geometry of the interconnection (interc.) and rotator waveguides (Wg). TE: Transversal electric polarization. TM: Transversal magnetic polarization. T_a, T_b: Thicknesses of the rotator waveguide.W_a, W_b: Widths of the two steps in the rotator waveguide.

The complete polarization rotator comprises a rotator waveguide in the middle, with rib interconnection waveguides at the input and output (see Figure 1). Adiabatic transitions are included between both types of waveguides to reduce insertion loss. The geometries of the interconnection and rotator waveguides in the proposed rotator, together with the orientation of the two higher-order modes (dashed lines), are depicted in Figure 1. If the hybrid modes in the rotator are tilted by 45° and the rotator waveguide section has a half-beat length (the length that produces a π phase shift between the two higher-order modes in the waveguide), it can be shown that any polarization launched into the rotator will be rotated by 90° at the output. This behavior is schematically depicted in the case of launching a TE polarization into the rotator (see Figure 1). When the TE-polarized fundamental mode of the rib waveguide (horizontal green line at input) is launched into the rotator waveguide, its hybrid modes are equally excited (45° red lines). These modes travel with different propagation constants along the waveguide and are π-shifted at a distance that equals the half-beat length. When the π-shifted modes are launched into the rib waveguide, the TM-polarized fundamental mode (the vertical green line at output) is excited and the desired rotation is produced.

The thicknesses of the rotator waveguide (T_a and T_b) will be the same as those used in the rest of the receiver.¹ This way, the thickness, T_a, is defined in the same dry etch step as the interconnection rib waveguides, and only one extra step is required to define the fully etched walls in the rotator waveguide.

Figure 2. The electrical field of the two higher-order hybrid modes in the rotator waveguide, .

As the thicknesses of the rotator waveguide are fixed, the widths (W_a and W_b) are designed to achieve the required 45° angle.⁶ Figure 2 shows the electric field— —of the two higher-order modes of the rotator waveguide. Adiabatic tapers allow a transit between the interconnection waveguides and the rotator waveguide while minimizing insertion loss. This transition can be fabricated at no additional cost, as it is the same thickness as the rotator and interconnection waveguides.

The propagation along the 3D structure is simulated using the fully vectorial commercial tool FIMMPROP.⁷ State-of-the-art performance is achieved for the proposed device with a calculated extinction ratio (ER) of 40dB and a 0.04dB insertion loss at a wavelength of 1550nm, covering the full telecommunications C-band (1530–1565nm) with an ER better than 20dB.

This work proposes a passive polarization rotator waveguide geometry for monolithic integration with an InP frontend for coherent 100GE. Adiabatic tapers minimize insertion loss. A simulated insertion loss of 0.04dB and ER of 40dB are achieved at the central wavelength (ER>20dB in the C-band). The proposed polarization rotator can be integrated together with the frontend at the cost of only one extra dry etch step, in contrast to the typical approach (slanted waveguide rotator²) that requires an extra dry and wet etch step. In the near future we expect to proceed with fabrication and characterization of the rotator to validate the feasibility of the proposed design.

This work has been partially funded under the Spanish Ministry of Science and Innovation under project TEC2009-10152, by the project AVANZA I+D (Celtic) TSI-020400-2008, and the European Commission 7th Framework Programme (FP7) project MIRTHE ICT-2009-5 contract 257980.