1. Introduction

Integrated optical devices are remarkable tools for analysis and fast data acquisition in sensing and telecommunication applications. In detection and sensing areas, they possess potential features such as label-free detection of target molecules, integration of arrays of compact structures on a single chip for a simultaneously detection of several analytes, and the ability for high volume fabrication. As the fabrication process is compatible with CMOS technology even electronics integration is feasible. Several optical circuits have been developed so far such as waveguides [1], Mach-Zehnder interferometers [2], directional couplers [3], photonic crystals [4] and microring resonators (MRRs) [5,6]. Among them, MRRs are a favored candidate due to their high sensitivity and ultra-compactness. In addition, an array of MRRs for multiplexed detection has been demonstrated [7–9].

MRRs exploit the interaction of the waveguide’s evanescent field with the external medium. In this manner, the refractive index change caused by an analyte on the waveguide surface can be monitored by the resonance wavelength shift. Bulk refractive index sensitivities (BRISs) from 70 to 200 nm/RIU (refractive index unit) and limits of detection (LODs) from 10⁻⁴ to 10⁻⁶ RIU were reported using MRRs [5–9]. A meaningful progress in the development of MRR-based sensors can be recognized during the last years, especially the high interest in order to improve sensitivity and LOD of MRRs.

Cascaded microring resonators (CMRRs) exploiting the Vernier effect (VE) are novel structures to achieve maximum sensitivity. Firstly, CMRRs were investigated theoretically [10,11] and later experimentally [12–14] based on a Silicon-on-Insulator (SOI) platform. An BRIS of 2169 nm/RIU was obtained for NaCl concentrations in DI water solutions compared to 76 nm/RIU for a single MRR [12]. Later, a low cost system for an intensity interrogation method of CMRRs was investigated by using a low-cost broadband source [15]. All these devices operated at a wavelength of 1.5 µm.

Here, we investigate CMRRs based on a silicon nitride (SiN) platform in the spectral windows of 1.3 µm and 1.5 µm, respectively. Three structures with two different buffer layers −benzocyclobutene (BCB) polymer and thermal silicon oxide (SiO_x) − were fabricated and characterized optically. The experimental results are compared with the simple theoretical analysis reported previously [16]. Tapered grating couplers were integrated at the ends of the waveguides for the investigation of CMRRs at 1.5 µm. Gratings enable a reliable coupling of the light between the chip and optical fibers at moderate alignment tolerances. First devices have been demonstrated operating in a wide operating range to be utilized as optical sensors for biological and chemical applications.

Additionally, polymers like BCB possess good thermal transmission properties being an excellent candidate for the incorporation of individual micro-heater electrodes in MRRs as suggested in [16]. This approach permits the spectral tuning of the resonances compensating the deviations of the fabrication process [8]. In this paper we present the design and experimental results of first CMRR circuits based on BCB buffer layers. BCB can be an alternative material compared to the standard SiO_x material offering unique properties, practical handling and low cost.

2. CMRRs design, fabrication and characterization method

The principle of operation of CMRR circuits is based on the VE. It is defined as the superposition of two signals with different free spectral ranges (FSRs) as shown in Fig. 1(a). Here, the periodicity of the resulting beat signal defines the new FSR, which is larger than the FSR of the individual signals. Thereby, VE can be generated simply by cascading two systems whose signals display slightly different FSR. Firstly, this principle has been exploited only to expand the FSR of filter devices [17,18]. Thereafter, its application has been focused on BRIS enhancement in optical sensors [10–15].

 

Fig. 1 (a) Vernier principle of two cascaded systems. (b) Configuration of CMRRs to exploit the VE. (c) Scheme of the cross section of a SiN ridge waveguide.

Download Full Size | PDF

Here, we propose CMRRs-based devices formed by cascading two MRRs as is illustrated in Fig. 1(b). The two MRRs are connected via a bus waveguide and two additional waveguides are used to couple light in and out of the device. For sensing operation, the structure—filter element (MRR_f) and waveguides—is passivated by using a top cladding layer except the sensor element (MRR_s). Here the evanescent field of the MRR_s interacts with the external medium causing a small refractive index change and resulting in a large wavelength shift of the Vernier spectrum.

Tapered grating couplers are used for optical interfacing of the optical circuit with feeding optical fibers as these coupling elements enable efficient coupling performance and moderate alignment tolerances. An efficient coupling of 60% at wavelengths of 1.31 µm has been previously reported using grating couplers integrated on SiN waveguides [19].

CMRRs are based on SiN platform technology where waveguides have a ridge shape as illustrated in Fig. 1(c). Three types of SiN structures were designed (c. f. Table 1): Structure I and Structure II consist of SiN waveguides (n_SiN = 1.94) with a buffer layer of BCB (n_BCB = 1.55). Structure III is based on SiN waveguides (n_SiN = 1.98) on SiO_x buffer layer (n_SiOx = 1.46). Here, BCB is utilized as a top cladding layer. The effective (n) and group (n_g) refractive indices of waveguides were calculated as described in [16]. Initially, all structures were designed for TE single mode light propagation at 1.5 µm. In addition, Structures I and II support TE resonances at 1.3 µm. The difference between both radii (R_f and R_s) is defined by ΔR where R_s>R_f, e. g., this value is ΔR = 298 µm - 297 µm = 1 µm for Structure I.

Table 1. Nominal parameters for Structures I, II and III

View Table

The FSR of individual MRRs can be calculated by using FSR_i = λ²/n_g,iL_i, where n_g,i and L_i are the group refractive index and circumference of each MRR. Both FSRs of individual MRRs leading to estimate the VernierFSR (FSR_VE) using the standard expression given by

From (1), FSR_VE values at λ = 1.3 µm were found to be of 105 nm and 41.3 nm for Structures I and II, respectively. Note that the FSR_VE is larger when the difference of individual FSRs is smaller.

The VernierBRIS (S_VE) in CMRR scan be calculated from the BRIS (S) of the sensor element multiplied by an amplification factor (M) that depends on the difference of FSRs [12]:

Here, δn_s and δn_medium are the effective index change and the index change of the sensing medium, respectively. An M factor at λ = 1.5 µm can be expected of 51.6for Structure III with l = 10 µm when water (n_medium = 1.33) covers the sensor element.

CMRRs were elaborated by using a standard lithographic technique [8]. For Structures I and II, a 7 µm-thick BCB polymer was spin-coated on a commercial 4″ silicon (Si) wafer. Afterwards, a 400 nm-thick SiN layer (n_SiN = 1.94) was deposited on top of the BCB layer by using ICP-PECVD followed by a dry etching process to fabricate a ridge depth of 250 nm. Here, the waveguides have no top cladding layer. All MRRs had a racetrack shape with a straight section length (l) of 10 µm, and coupling gap of 1 µm. For Structure III, a 250 nm-thick SiN layer (n_SiN = 1.98) was deposited via LPCVD on a thermal SiO_x layer of 8 µm thickness. A ridge depth of 180 nm was obtained by dry etching. Finally, the complete structure except the MRR_ss was covered with a BCB layer of 4 µm thickness. In this structure, the straight section lengths vary from 10 µm to 50 µm. Figure 2 shows the microscope images of fabricated structures: CMRRs, coupling region of waveguides, and the tapered grating coupler with a period of 1.2 µm – grating width of 0.6 µm and spacing of 0.6 µm – localized at the feeding ends of waveguides in Structure III.

 

Fig. 2 Microscope images for two types of CMRRs fabricated on SiN platform: (a) Structure I with a buffer layer of BCB and (b) Structure III with a buffer layer of SiO_x and top cladding layer of BCB on MRR_fs. (c) Top view of the tapered grating coupler with a period of 1.2 µm. Inset images in Fig. 2(a) and Fig. 2(b) show the coupling region between the bus waveguide and MRR. All MRRs have a nominal gap of 1 µm.

Download Full Size | PDF

The performance of structures was investigated by using two tunable laser sources (TLS) centered at 1.3 µm and 1.5 µm, respectively. Both lasers operate in a dynamic sweep mode and with wavelength steps of 1 pm. For Structures I and II, the emitted light from the TLS was launched into the input waveguide via a tapered optical fiber. A linear optical fiber controller was used to adjust the TE polarization status of the coupled light. The transmission spectrum of CMRRs was measured via a second tapered fiber aligned to the output waveguide and feeding a photodetector. In contrast to the previous structures, Structure III was characterized by using cleaved fibers in order to couple the light via interfacing with the tapered grating couplers. The chips with 10 × 10 mm² size were characterized at constant temperature of25°C.

3. Results and discussion

3.1 Optical devices operating at 1.3 µm

Structures I and II were characterized optically at 1.3 µm. Single MRR_ss from Structure II show resonances with a Q-factor about 10⁴ and an FSR_s of ~0.45 nm. The waveguides on BCB have propagation losses of about 10 dB/cm. Figure 3 shows the transmission spectra of CMRRs with two slightly different radii: (a) ΔR = 1 µm and (b) ΔR = 3 µm. The typical Vernier spectrum (pink line), that is constituted by a series of peaks whose envelope of maxima (purple dots) have a Lorentzian shape, is observed for both cases between 1.27 – 1.335 µm. The experimental analysis of the spectrum is based on a Lorentzian fit of the envelope of the maxima. After an individual fit of the amplitudes of the peaks, the envelope was generated by localizing the maxima. The obtained fit envelopes were centered at 1.27927 µm and 1.32378 µm, their widths (Δλ_VEs) were found to be of 3.53 nm and 2.29 nm, respectively. Concerning the FSR_VE, defined as the spectral distance between two peaks at center of the envelope, they can be estimated by using (1). FSR_VEs about 131.7 nm and 47.21 nm, respectively, were calculated. Only one complete Vernier peak can be observed within the laser tuning range (1.27 – 1.34 µm). In Fig. 3(a), a significant deviation of the amplitudes of individual peaks compared to the fitted envelope can be identified. This discrepancy originates from perturbations caused by mechanical oscillations of the fibers which can cause an error on the fit of the envelope as well. This becomes critical for monitoring small changes in the external medium of the sensor element of CMRRs and finally deteriorates the LOD.

 

Fig. 3 Transmission spectra of SiN CMRRs for (a) Structure I (ΔR = 1 µm) and (b) Structure II (ΔR = 3 µm). Purple dots indicate the positions of the transmission maxima and the fitted envelope is drawn as a black line.

Download Full Size | PDF

The theoretical model presented in [16] has been used to simulate the spectrum of Fig. 3(b). The model is based on an ideal filter MRR generalized for two cascaded MRRs. Initially, the experimental spectrum of each MRR was reconstructed by a simple add-drop filter model, where a finite difference time domain method (FDTD) was utilized in order to obtain the losses (f) and coupling coefficient (α) parameters that were used later as input for the simulations of CMRRs. Afterwards, the spectral position was matched by adjusting the radii (R_f and R_s) until the solutions closely reproduced the experimental Vernier spectrum. This slight adjust in radius can be included within fabrication tolerances. Figure 4 illustrates the normalized experimental spectrum and the theoretical results. The parameters used for such simulations are given in the figure caption. The theoretical envelope is centered at 1.3235 µm having a Δλ_VE of 2.23 nm, while the center of the experimental envelope is shifted around 280 pm with a 60 pm wider Δλ_VE. These values show good agreement in spite of that the calculations were done considering ideal MRRs where the propagation loss was assumed wavelengths independent over the investigated spectral range as well as the fabrication tolerances of CMRRs were not taken into account. The envelope fit in Fig. 4(b) was achieved as explained previously.

 

Fig. 4 Transmission spectra of SiN CMRRs for Structure II: (a) experimental and (b) theoretical results. Simulation parameters following [16]: α₁ = α₂ = α₁₁ = α₁₂ = 0.957, f₁ = f₂ = 0.97, n_g,f = n_g,s = 1.969292, R_f = 297.03 µm, R_s = 299.918 µm, l = 10 µm.

Download Full Size | PDF

3.2 Optical devices operating at 1.5 µm

Structure II was investigated also in order to study the VE at 1.5 µm. In Fig. 5(a), the transmission spectrum has been measured between 1.55 µm and 1.65 µm. Here, two Vernier peaks were centered at 1.57165 µm and 1.63685 µm, respectively, resulting in an FSR_VE of 65.2 nm. The FSR_VE calculated using (1) is 68.2 nm, which is a good approximation to the experimental value. The envelope fit for each Vernier peak showed a Δλ_VE of 6.85 nm and 10.24 nm, respectively. This spectrum was obtained via the theoretical model as explained above. Figure 5(b) shows the calculated Vernier peaks centered at 1.56879 µm and 1.64013 µm using the parameters given in the figure caption as input of the theoretical model mentioned above. A large theoretical FSR_VE of 71.34 nm was found by and the envelope widths differ 142 pm and 2.49 nm, respectively, from the values of the measured spectrum. These theoretical results are based on an analytical model where the absorption of materials and the material dispersion were not considered. Here, this effect has been relevant for longer wavelengths, resulting in a moderately good approximation to the experimental results. Note that the amplitude difference of the two Vernier peaks in Fig. 5(a) originates from the wavelength-dependent coupling coefficient of the directional couplers.

 

Fig. 5 Transmission spectrum at wavelengths around 1.5 µm for Structure II: (a) experimental and (b) theoretical results. Simulation parameters following [16]: α₁ = α₂ = α₁₁ = α₁₂ = 0.915, f₁ = f₂ = 0.935, n_g,f = n_g,s = 1.920625, R_f = 297.03 µm, R_s = 300.02 µm, l = 10 µm.

Download Full Size | PDF

Finally, several measurements were performed for Structure III where both MRRs have slightly different perimeters by only varying their straight section lengths from 10 µm to 50 µm. Note that, the radii and the gap were constant. The waveguides on SiO_x have propagation losses of about 3 dB/cm. Figure 6(a) shows five spectra for different straight sections. Note that the Lorentzian shape of the envelope is less pronounced when the straight section length of MRRs increases as the individual peaks become broader, resulting a partial overlapping of peaks. The width of individual peaks (Δλ_singlepeak) increase from 30 pm up 138 pm as a function of the straight section length. Meanwhile, the envelope width of the Vernier spectrum centered at 1.5645 µm broadens up to a factor4.4 times starting with a Δλ_VE of 5.4 nm. The Fig. 6(b) gives the linear relation between the Vernier envelope width and the single peak width. This experimental observation is confirmed with the analytical model shown in [12], where the envelope width is proportional to the width of the individual peak multiplied by a factor related to the FSRs of the individual MRRs. For l = 50 µm, the spectrum tends to be uniform as a spectrum obtained from the single MRR and also the background intensity is slightly higher. Thus, short straight section lengths about of 20 µm for gaps of 1 µm in racetrack SiN MRRs are the optimum parameters in order to generate efficiently the VE using Structure III. However, a complete study of the structural parameters should be necessary, e.g., the gap between the MRR and the waveguides.

 

Fig. 6 (a) Spectra for different straight section lengths (l = 10 - 50 µm) using Structure III. (b) The width of the Vernier envelope (Δλ_VE) as a function of the width of the individual peak (Δλ_{single peak}) for the straight section lengths given. Black dots are the data and the red line is the linear tendency of the obtained data. The radius of the MRR_f is 199 µm with ΔR = 1 µm.

Download Full Size | PDF

Excellent stability in amplitude of the peaks has been measured in each spectrum using tapered grating couplers. The envelope of peaks was not affected by noisy effects due to slightly vibrating fibers. Nevertheless, an optimization of the coupling efficiency of the SiN grating couplers would improve the performance of the Vernier structures.

For sensing operation in aqueous solutions, a VernierBRIS of about S = (51.6)·(190 nm/RIU) = 9804 nm/RIU for Structure III can be expected by using (2)and the BRIS of the single MRR reported previously in [20]. Concerning the LOD, a value about of 2.04 × 10⁻⁶ RIU− with a minimum detectable wavelength shift of 0.02 nm taking into account the noise level of our system −can be expected for the fabricated cascaded SiN MRRs. Then, CMRRs based on a SiN material are promising candidates for broad applications in sensing areas.

4. Conclusion

We investigated an optical device based on silicon nitride cascaded microring resonators (CMRRs) and the Vernier effect (VE). The characterization of the Vernier spectrum was carried out successfully around 1.3 µm and 1.5 µm, respectively. Three structures were fabricated with two different buffer materials and different structural parameters. For CMRRs on BCB, a detailed analysis of the Vernier spectrum was performed showing good agreement with the simulations. CMRRs on SiO_x with different straight section lengths incorporating tapered grating couplers were measured at 1.5 µm, which exhibit a good stability of the Vernier spectrum. This study suggests that CMRRs based on SiN waveguides will be an excellent candidate to realize highly sensitive optical sensors for biological and chemical applications to be operated in a wide wavelength region.