1. Introduction

Control and manipulation of the polarization state of light have a diverse range of applications including metrology [1], polarimetry [2], time of flight [3] and polarization mode division multiplexing [4]. Currently, such systems are assembled using bulk optical components such as wave plates or retarders, electro-optic modulators, polarizers and/or polarising beam splitters. Photonic Integrated Circuits (PICs) offer the prospect of replacing the manual alignment and assembly of components with lithography, providing advantages in terms of volume of production, yields, form factor and ruggedness. Furthermore, the miniaturisation of the modulation element allows for higher-speed, lower power and lower drive voltage requirements, for example, as obtained with a p-i-n junction. Applications of polarization manipulation in PICs include: all optical computing and ultra-high speed data routers [5] and “lab on a chip” sensing systems [6]. In addition to such applications where the polarization is modulated from a particular defined state, there also exist numerous applications in which the polarization is manipulated from a random polarized state into a pre-defined state appropriate for a polarization-sensitive element, for example alignment to a TE polarization state prior to a semiconductor optical amplifier in an optical communications network.

In this letter we base the polarization conversion on mode-beating in an asymmetric profile semiconductor waveguide [7]. This typically allows for polarization mode conversion in a shorter waveguide length than the alternative of an adiabatic mode converter. Additionally it provides a transformation of a general polarization state involving both fundamental polarization modes with a similar functionality to a bulk optical waveplate. We have also identified that with a suitable fabrication methodology, the asymmetric waveguide section can be tapered to a standard symmetric waveguide, thus minimizing optical losses at transitions between waveguide sections [8].

It has been established that a wide range of waveguide polarization function devices can be based around a universal mode-beating polarization mode converter (PMC) designed to provide 3dB mode conversion in a half-beat-length (equivalent to a half-wave-plate with an optic axis at 22.5° or 67.5° to the wafer normal) [9]. A waveguide polarization modulator/ controller can be devised by combining PMC sections with differential phase shift (DPS) sections that provide a variable birefringence in a symmetric waveguide. For example, a combination of 2 × 3dB PMC sections with 2 × DPS sections is sufficient to provide a polarization modulator that can translate TE- (or TM-) polarized mode into any defined polarization state [9,10].

Using III-V semiconductors as a PIC platform additionally facilitates the monolithic integration of a polarization modulator/controller with diode laser and amplifier sections. It has been previously demonstrated that the monolithic integration of a PMC section with a gain section enables the emission of TM-polarized light from an interband transition in an unstrained p-i-n diode laser [11], and emission of TE-polarized light from an inter-subband transition in a cascade laser [12]. In this letter we demonstrate the feasibility of a dynamic polarization-modulated emitter by monolithically integrating the three essential waveguide elements: laser, PMC and DPS. Furthermore, the characterization of the performance of each of these elements provides the required data for the design of multi-element waveguide devices.

2. Device design and operation

Our integrated laser with polarization controller is shown schematically in Fig. 1 and consists of 3 sections. Devices were fabricated in commercially available InGaAlAs strained multiple quantum-well material which emits TE polarized light at a wavelength of 1550nm [13]. The laser comprises a 700µm long, 2.4µm wide, shallow-etched (to the top of the waveguide core) gain section located between a cleaved facet mirror on the left of the device and a 120µm long shallow- to deep-etched (1µm into the lower cladding) taper, terminating at a deep etched slot (373nm wide). The emitted radiation from this section is polarized in the plane of the wafer, known as the transverse electric (TE) direction.

 

Fig. 1 Schematic of the semiconductor laser with active polarization control consisting of an tapered F-P laser, waveguide slot, polarization mode converter (PMC), differential phase shifter and angled facet output.

Download Full Size | PDF

Immediately after the slot is a 250µm long, 1.8µm wide, deep-etched Polarization Mode Converter. The PMC section, as shown in the SEM cross-section in Fig. 2 , is based upon a deep-etched waveguide with lateral asymmetry defined with a single offset 100µm long slot (with 4µm input and output tapers) [14]. The slot depth just reaches the graded index region of the waveguide core, and is fabricated reproducibly because of the large difference in etch rate between the phosphide upper cladding and the aluminum containing core layers. A key feature is the tapering of the slot width (and depth due to etching lag) at the transition to the symmetric waveguide, which reduces the optical losses that usually arise from the mode-mismatch and Fresnel reflection that are present at a discontinuous waveguide transition [8]. The light in the symmetric waveguide following the PMC will, in general, consist of both TE- and TM-polarized modes. The target universal PMC design is that the power should be split equally between these two polarization modes, and the modes should be in-phase.

 

Fig. 2 SEM cross section image of a sample cleaved through the center of a PMC (Left) and a profile view of the transition between PMC and symmetric guide section (Right).

Download Full Size | PDF

Following a 120µm long deep- to shallow-etched tapered transition section, the light enters the 2.4µm wide shallow-etched, 1mm long Differential Phase Shifter. Using current injection, the birefringence of this symmetric waveguide section can be modified as the TE-polarized mode will experience a greater magnitude of refractive index change in comparison to the TM-polarized mode due to the presence of the heavy-hole bandedge resonance. The output of the DPS section is angled at 10° to the facet normal, to reduce inter-cavity reflections, which additionally ensures the DPS current is always below the lasing threshold for this section.

3. Characterization

The devices were mounted on a thermo-electric cooler (TEC), set at 20 °C. The laser sections of the devices were tested first and the outputs at the cleaved facet side were measured at several currents above threshold and found to consist of between 97.5% and 98% TE-polarization purity.

The characterization of the output polarization purity at the PMC & DPS sections of the devices is complicated by the presence of anisotropic bandedge absorption/gain due to the quantum wells and the addition of ASE in the DPS section. The anisotropic loss/gain in the DPS can be accounted for by identifying the conditions for transparency operation. It is assumed that the anisotropic gain/loss in this section can be attributed entirely to the interband carrier absorption/stimulated emission. As the transparency current is wavelength dependant, and the emission wavelength of the laser section also varies with injection current, the transparency point was determined versus wavelength, by injecting light from a modulated (1kHz) tuneable laser into the rotator/phase shifter section whilst adjusting the bias current and monitoring the voltage across the devices with a lock-in amplifier until at the transparency point, where the voltage modulation induced by the optical signal is at a null. Once determined, the laser section of the devices was biased above threshold (first at 60mA) whilst the wavelength and the TE- and TM- polarized output powers of the devices (P_TE and P_TM respectively) were monitored as the current applied to the DPS section was varied from 0 to 20mA in steps of 0.2mA. This was then repeated for several values of injection current into the laser section from 60mA and 140mA in steps of 20mA. At 100mA drive current, the laser output from the back facet was 3.18mW. TE and TM output powers at the DPS facet were 78µW and 98µW respectively, with the DPS section biased at the transparency current. Assuming 30% and 40% reflectivity for the cleaved facet and the slot mirrors of the laser respectively, the total loss of the PMC and DPS combination is of the order of 11dB. 6dB of this loss can be attributed to the passive loss (14cm⁻¹) of the DPS section at transparency. The remaining ~5dB is attributed to absorption in the unbiased PMC section, diffraction losses in the slot mirror and residual losses at the PMC/DPS waveguide transition.

The output TM-polarization purity [P_TM/(P_TE + P_TM)] obtained is plotted as a function of DPS current at each laser injection current for the device with a 100µm trench length, with the values obtained for transparency current at each of the laser injection currents indicated (Fig. 3 ). The trend of decreasing TM-polarization fraction is consistent with the phase shifter section acting as an SOA and producing increased TE output as the current is increased. The value at the DPS transparency condition yields the power fraction in each polarization mode exiting from the PMC. A range of devices with various trench lengths were characterized in this fashion, and the result closest to the that for the desired half-beat-length 3dB polarization coupler is the ~40% TM-polarization purity shown for the 100µm trench length.

 

Fig. 3 Output polarization TM purity versus DPS current for a range of laser bias currents in mA. The transparency point for each laser bias is indicated on the graph.

Download Full Size | PDF

4. Polarization modulation

With a constant current of 120mA applied to the laser section, the current to the phase shifter section was varied between 0 and 6mA in steps of 0.2mA, while the output optical power was monitored through a polarization analyzer, set at −45° and + 45° to the plane of the wafer, as shown in Fig. 4(a) . The measurements were repeated with the insertion of a quarter-wave-plate with the fast axis parallel to the plane of the wafer (TE-polarization) immediately prior to the polarization analyzer, as shown in Fig. 4(b).

 

Fig. 4 (a) Output versus DPS current for polarization analyzer angles of 45 and −45 degrees. (b) Output with the addition of a quarter-wave-plate with a fast axis aligned in the plane of the wafer.

Download Full Size | PDF

The total optical power increases with the DPS current due to the reduction in TE-polarization bandedge absorption. The differential phase shift provides the observed oscillation as the output polarization state alternates between linear and elliptical polarization states. Consequently, over the current range shown, the differential phase shift can be deduced to be slightly larger than 2π. As the power splitting of the PMC at ~40% is less than ideal, we do not observe the depth of the oscillations reaching 100% visibility.

The results with the incorporation of the quarter-wave-plate are similar in form, but are advanced by a relative phase of π/2 where the fast axis is aligned with the TE-polarization. It can therefore be deduced that increasing the DPS current decreases the refractive index of the TE-polarized mode relative to the TM-polarized mode. This result is consistent with a refractive index change due to bandfilling affecting the heavy-hole/conduction band resonance.

The temporal response of the polarization state was also investigated by modulating the bias current applied to the DPS section. Using a function generator, the current was modulated between 3mA and 6mA with a pulse of 8ns duration, and the corresponding optical pulse at the output of the polarization analyzer was measured using a 2.5GHz bandwidth, AC coupled photodetector. Figure 5 shows the polarization switched output with the analyzer set at + 45° along with the electrical input driving pulse. The output exhibits a 10/90 rise time of 4.8ns with the trailing edge of the pulse exhibiting a 90/10 decay time of 14.6ns. We attribute the limited bandwidth to the probe system used to drive the device. We note that the carrier lifetime in devices fabricated from the same expitaxial material has been shown to be of the order of 500ps and modulation at higher speeds may be obtained by employing the quantum confined Stark effect with a reverse-bias voltage.

 

Fig. 5 Polarization switched output measured through an analyzer set at 45° after the application of an 8ns input current pulse to the DPS.

Download Full Size | PDF

5. Conclusions

We have demonstrated a semiconductor laser integrated monolithically with elements of a dynamic polarization modulator. The device consists of a Fabry-Perot laser and an asymmetric waveguide polarization mode converter followed by a differential phase shifter. A range of polarization mode converters were characterized and it was deduced that the largest TM-polarized component obtained after the PMC was ~40%. However, it should be noted that this figure is due to a combination of the mode-beating in the asymmetric waveguide and the anisotropic bandedge absorption, and therefore the mode conversion attributed to mode-beating in the asymmetric waveguide alone will be significantly lower. This result indicates a strategy for optimizing the performance of the PMC. First, it will desirable to reduce the bandedge absorption in the PMC section. One option would be to blue-shift the bandedge using quantum well intermixing [15]. Second, a suitable range of trench lengths should allow the half-beat-length to be identified. Third, the trench width and position can be varied to identify the ideal 3dB half-beat-length PMC.

It has been shown that the output polarization state can be substantially modified by varying the current injection in the differential phase shift section. The differential phase shift is attributed to bandfilling in the quantum wells, and we demonstrate an electrically limited decay time of ~14.6ns. Higher speed modulation than the ~1GHz implied by the carrier lifetime in this material may be obtained by employing the quantum confined Stark effect with a reverse-bias voltage. However, this would necessitate a blue-shift of the bandedge to avoid bandedge absorption. Again this could be realized with quantum well intermixing [15].