Integrated photonic circuits have been established in a wide range of applications, e.g., in communication engineering, quantum computation, or sensing. Especially in biosensing, small concentrations of substances such as pathogens can be detected. For instance, in [1], a bimodal interferometer (BMI) combined with a functional cladding is used for COVID-19 detection. The functional cladding on the waveguides adsorbs the target molecules, leading to a change in the refractive index that can be extracted by the BMI. Other promising candidates for refractometry are resonant structures like ring resonators [2] or photonic crystals [3]. Compared with Mach–Zehnder interferometers (MZIs) [4] or BMIs [5], they achieve the same sensitivities at a smaller footprint. However, they are also sensitive to temperature changes and fabrication tolerances. Interferometers can be designed to be more temperature stable and to be more robust against fabrication inaccuracies. They have low losses and advanced readout systems can be applied. In [6], our group proposes a 90° hybrid as a phase readout system. Compared to the single-ended phase detection, this scheme enables single wavelength operation with an unambiguity over $2\pi$, unsusceptibility against input power fluctuations or unknown modal attenuation, and a sensitivity that is independent of the operation point. In [7], a single-waveguide polarization mode interferometer (PMI) is demonstrated. A large waveguide width of 1 mm results in low propagation losses and high fabrication robustness. Despite such wide dimensions, sensitivity remains comparable to that of MZIs [7]. Thus, dual-polarization operation enables low-loss sensing by maintaining high sensitivity. However, the excitation of different polarized signals as well as their readout is performed externally requiring a polarization insensitive coupling interface. This work presents a novel fully integrated single-waveguide PMI in the 220-nm silicon-on-insulator (SOI) platform that generates the interfering signals on a chip. The result is a compact device that is scalable and can be manufactured using standard semiconductor technologies. In the presented system, converters transform a single-polarization input signal into the fundamental modes per polarization in the sensing waveguide. After the sensing area, the signals interfere by passing the same converters in reverse order. Depending on the interference pattern, the induced mutual phase difference caused by the refractive index of the cladding can be determined. Moreover, to improve the phase readout, a PMI is combined with the 90° hybrid proposed in [6]. This work experimentally demonstrates its functionality for the first time.

The single-ended architecture is shown in Fig. 1(a). The incident transverse-electric fundamental mode (TE0) enters a mode converter described in [5]. Its modified taper-structure [Fig. 1(b)] is optimized for balanced excitation of the TE0 and the transverse-electric mode of first order (TE1) at the butt-coupled interface with a simulated excess loss (EL) of 0.3 dB. The subsequent tapered rib waveguide with a 150-nm-high slab implements a polarization converter, in which the signal passes through waveguide cross sections that have equal effective refractive index for the TE1 and the fundamental transverse-magnetic mode (TM0). In these regions, those modes are strongly coupled. With further propagation, the TE1 converts to the TM0, whereas the TE0 remains unchanged [Fig. 1(c)]. The simulated EL for the transition from TE1 to TM0 is 0.02 dB with an extinction ratio (ER) of 26 dB. The TE0 passes the converter with less than 0.01-dB EL.

 

Fig. 1. (a) Top view of the polarization interferometer, (b) mode conversion from TE0 to balanced TE0 as well as TE1, and (c) the polarization converter.

Download Full Size | PDF

In the sensing region, the TE0 and TM0 modes propagate independently. The analyte on top of the sensing arm affects the corresponding phase velocity per polarization mode in a different way. Therefore, the signal has an arbitrary polarization state at the end of the sensing region of length $L$. The TE0 and TM0 interfere to form an output signal in TE configuration by applying the same converters. The difference of the modal effective refractive indices of the interfering signals $\Delta n$ can be extracted from transmission $T$:

Figure 2 shows the transmission spectrum of the system, fabricated at IMS CHIPS in Stuttgart with a 900-nm-thick $\mathrm {SiO_2}$ cladding. The EL is 1.5 dB at a wavelength of 1544 nm with an ER up to 30 dB, while the sensing region extends over 5 mm. With a 100-$\mathrm{\mu}$m-long reference structure, the combined loss of the mode and polarization converter is identified to be 0.4 dB. If different modal losses in the presence of an absorbing material are expected, the first mode converter can be designed for unbalanced mode excitation maintaining or enhancing the ER [5].

 

Fig. 2. Measured transmission spectrum of a PMI with a 5-mm-long and 1.5-$\mathrm{\mu}$m-wide sensing waveguide.

Download Full Size | PDF

The sensing waveguide is another key component of the device. In an ideal case, it offers high sensitivity with low propagation losses. Typically, these propagation losses are mainly caused by scattering processes at the interface between the silicon waveguide and its surroundings. For TE modes, the roughness of the sidewalls is the main loss factor, whereas in TM configuration, the roughness of the top and bottom surface is more dominant. In standard multi-project wafer processes, the sidewalls are much rougher than the top/bottom surface [8]. Thus, propagation losses in TM mode are inherently lower compared with those in TE [9]. To reduce TE losses, wider waveguides are required [8]. However, to get a high sensitivity $S$ of the system, the waveguide should be narrow to increase the evanescent coupling between the modes and the analyte. This system sensitivity $S$ can be defined by the change of the transmission depending on the refractive index unit (RIU) of the cladding $n_\mathrm {c}$:

The term $S_{\mathrm {bulk}}$ is the so-called homogeneous bulk sensitivity and describes the phase response to a refractive index change of the cladding. It depends on $\frac {\partial \Delta n}{\partial n_\mathrm {c}}$, the so-called intrinsic bulk sensitivity, which includes the impact of the sensing waveguide geometry [10]. Figure 3 shows the simulated intrinsic bulk sensitivities for different single-waveguide interferometers. As expected, a larger width leads to lower sensitivity for interferometers using the same polarization for both modes. However, operating with different polarization modes, $\frac {\partial \Delta n}{\partial n_\mathrm {c}}$ converges to 40$\%$. This is based on the fact that the evanescent field of the TM0 mainly protrudes on the top and bottom surface, which is a function of thickness, rather than width.

 

Fig. 3. Intrinsic bulk sensitivity over the waveguide width for single-waveguide interferometers for a change of $n_\mathrm {c}$ from 1.44 to 1.45 simulated in FIMMWAVE from PhotonDesign.

Download Full Size | PDF

The use of wide waveguides not only reduces the influence of scattering processes, but the signal remains in its defined modal and polarization state, because modal conversion induced by roughness is also reduced. This is especially important for long waveguides, whose cross sections have similar effective refractive indices for different modes, as investigated in [11]. As depicted in Fig. 3, the choice of the width in a PMI is more flexible and thus can be adapted to prevent this problem.

The free spectral range ($FSR$), defined as

A common metric for the performance of refractometers is $S_{\mathrm {bulk}}$ that is expected to reach approximately 8000 rad/RIU. The extended specification of $S$ in Eq. (2) also includes losses and the dependency on the operation point, which allows the comparison of readout systems. For biosensing applications, single-wavelength operation is aspired for cheap lab-on-a-chip solutions. That means the measurement of the output power at one given and constant wavelength must be sufficient for phase extraction. This poses several challenges. As Eq. (1) reveals, the modal losses $t_1$ and $t_2$ have to be known. Also, input power fluctuations directly affect the result. To overcome these problems, in [12], a BMI with two differential output ports is demonstrated. The same principle can be applied on the PMI by adding another port to the single-ended mode converter. Nevertheless, the measuring points are only injective over a phase range of $\pi$. Due to the periodic behavior of the transmission, there are ambiguous solutions. Additionally, the sensitivity depends directly on $\Delta \phi$. In the worst case, the operation point ranges around $\Delta \phi \approx 0$, resulting in complete insensitivity [see Eq. (2)]. So, keeping ambiguity and sensitivity, the system can only operate in a strongly limited range. A 90° hybrid offers a solution to these challenges. Replacing the single-ended mode converter in Fig. 1(a) with the device proposed in [6] or Fig. 4(a), the PMI shows the desired behavior. Figure 4(a) also depicts the light propagation in the 90° hybrid. First, a mode converter separates TE0 and TE1. While the TE0 propagates to the centered output of the multi-mode interferometer (MMI), the TE1 mode is asymmetrically split to the upper and lower single-mode waveguides with a mutual phase shift of $\pi$. A set of three identical $2\times 2$ MMIs generates the 90° hybrid behavior depicted in Fig. 4(b). The balanced detection cancels out any input power fluctuations, like the relative intensity noise (RIN) of the laser source. It transforms the four 90° hybrid transmission curves to two photocurrents [see Fig. 4(c)]. In the ideal case, it represents a complex signal $\underline {s}=(T1-T2)+j(T3-T4)$ resulting in

The phase difference $\Delta \phi$ can be easily computed by

Figure 4(d) shows the calculated phase difference $\Delta \phi _c$ over the real phase difference $\Delta \phi$.

 

Fig. 4. (a) Propagation of the TE0 and TE1 through the 90° hybrid simulated with BeamPROP from RSoft Photonic Device Tools, (b) phase-dependent transmissions of the four output ports, (c) balanced detected signals, and (d) the calculated and real phase difference.

Download Full Size | PDF

The deviation of the transmission over $n_\mathrm {c}$,

Figure 5(a) shows a micrograph of the 90° hybrid that is embedded in a PMI with a 500-$\mathrm{\mu}$m-long and 1.5-$\mathrm{\mu}$m-wide sensing waveguide. Its transmission spectrum is shown in Fig. 5(b) with a measured EL of 1.8 dB. Reduced by the 0.4-dB loss of the other converters, the measured 90° hybrid EL is 1.4 dB. The 0.7-dB difference compared to the simulation is mainly caused by waveguide losses and a reduced waveguide layer height of approximately 210 nm. Therefore, the spectrum is shifted to lower wavelengths. The small ripples in the measurement are artifacts from undesired modes of higher order excited by the first mode converter [see Fig. 1(b)]. They can be filtered out, e.g., by passing a waveguide section that satisfies the single-mode condition and that is then tapered to the sensing waveguide. Nevertheless, the device shows the designed behavior and allows accurate phase computation, which is a result of the device robustness in terms of fabrication. The output powers are detected by a fiber-coupled power meter. In software, the balanced detection is reconstructed mathematically by subtracting the signals of the corresponding differential ports. The result is shown in Fig. 5(c). Figure 5(d) depicts the extracted phase. To achieve higher phase resolution, further digital signal processing (DSP) can be applied for the compensation of fabrication tolerances causing amplitude as well as phase errors [4].

 

Fig. 5. (a) Fabricated system with (b) the corresponding measured transmission, (c) the subtraction of the corresponding differential output powers, and (d) the calculated phase.

Download Full Size | PDF

In conclusion, a novel design of a high-efficiency integrated PMI is presented. The design is realized in 220-nm SOI, but the structure can be fabricated in other material systems with different spectral transmission windows. Due to the operation with two polarization modes, even very wide waveguides can be used to achieve high sensitivity and low propagation losses. With a measured EL of 1.5 dB, a homogeneous bulk sensitivity of approximately 8000 rad/RIU is simulated.

In combination with a designed 90° hybrid, the system is robust against input power fluctuation, varying modal losses, and operates always in a point of high sensitivity. Compared to a single-output PMI, the unambiguity at single-wavelength operation is extended from $\pi$ to $2\pi$. The use of more efficient and compact 3-dB couplers as well as advanced DSP offers further potential for optimization.

The building blocks of the presented PMI can also be used as stand-alone components. The readout system is capable for e.g., polarization state analysis. The low loss excitation of TE0 and TM0 in a single waveguide is also interesting for the generation of quantum states or other applications.