Improving the detection limit for on-chip photonic sensors based on subwavelength grating racetrack resonators.

Compared to the conventional strip waveguide microring resonators, subwavelength grating (SWG) waveguide microring resonators have better sensitivity and lower detection limit due to the enhanced photon-analyte interaction. As sensors, especially biosensors, are usually used in absorptive ambient environment, it is very challenging to further improve the detection limit of the SWG ring resonator by simply increasing the sensitivity. The high sensitivity resulted from larger mode-analyte overlap also brings significant absorption loss, which deteriorates the quality factor of the resonator. To explore the potential of the SWG ring resonator, we theoretically and experimentally optimize an ultrasensitive transverse magnetic mode SWG racetrack resonator to obtain maximum quality factor and thus lowest detection limit. A quality factor of 9800 around 1550 nm and sensitivity of 429.7 ± 0.4nm/RIU in water environment are achieved. It corresponds to a detection limit (λ/S·Q) of 3.71 × 10-4 RIU, which marks a reduction of 32.5% compared to the best value reported for SWG microring sensors.

Recently, silicon photonic micro-ring resonator, a typical photonic sensor, based on the interaction between the evanescent wave and the analyte, has been reported that it can match and even exceed the detection limit of ELISA (subpicogram/mL) by applying the nanoparticle-clustering methodology to the micro-ring resonator assay [19]. Thus, the overlap between light and matter is larger and the interaction of light and matter is more intense when the evanescent field penetrates deeper into the cladding. Subwavelength grating (SWG) waveguide consists of periodic silicon pillars with a period smaller than the operating wavelength, which can increase the interaction between light and the cladding materials on the top and side of the waveguide, and the space between the silicon pillars on the light propagation path [13]. Intuitively, silicon photonic micro-ring resonator based on SWG waveguide can enhance the evanescent field with respect to a standard silicon strip waveguide resonator with the same cross-section. Therefore, a higher sensitivity can be expected. However, the increased interaction inevitably causes additional absorption loss by analytes and scattering loss induced by surface roughness in the micro-ring region, and the quality factor (Q) of micro-ring resonator would then deteriorate. As a result, for certain wavelength λ, the intrinsic detection limit (iDL = λ/S·Q), which reflects the detection capability of the change of solution concentration may not be enhanced.
In this paper, in order to enhance the photon-analyte interaction to increase the sensitivity without decreasing the qualify factor, and to minimize the detection of limit, we propose a silicon-based transverse magnetic (TM) mode SWG racetrack resonator (SWGRTR) operating at around 1550 nm. The use of TM mode enlarges the overlap between light and sensing medium, thus enhances the photon-analyte interaction and increases the sensitivity. Taking advantage of racetrack-shape resonator brings flexible control of the coupling strength so that the qualify factor of the resonator can be optimized. The experimental results show that a quality factor of 9800 is achieved in water environment when the coupling length and gap are equal to 6.5 μm and 140 nm, respectively. The bulk sensitivity (S) is ~429.7 nm/RIU (refractive index unit), and the enhanced intrinsic detection limit (iDL) of 3.71 × 10 4 RIU are achieved. The intrinsic detection limit is improved by 32.5% compared to the best value reported for SWG microring sensors [20,21]. A schematic of the proposed SWGRTR is shown in Fig. 1(a). The subwavelength waveguide core is formed by periodically interleaving high refractive index (silicon) and low refractive index (analyte solutions) materials [22,23]. Figure 1(b) shows the magnified images (xy plane) of the pillar region in red dashed box in Fig. 1(a). Figure 1(c) shows the top view (xz plane) of the coupling region between the bus waveguide and the racetrack resonator in blue box. Figure 1(d) shows a front view (yz plane) of the SWG waveguide.

Device simulation and analysis
To make the subwavelength waveguide operating in the subwavelength regime (λ/Λ>2n eff (λ = 1550 nm)), which does not diffract light in the far field, and propagates between the voids and the silicon blocks (affected by Λ/λ) like the conventional strip waveguide, a period of 200nm is chosen [24]. In Fig. 1(c), Λ represents the period of the SWG structure and equals 200nm. L, W and H are the length, width and thickness of silicon (Si) pillars, respectively. L c and G represent the coupling length and the gap between the bus waveguide and racetrack resonator, respectively. The radius of racetrack is 10μm. The thickness of the buried oxide is 3μm. The upper cladding is the analyte solutions to be detected.  Fig. 2, TM mode electric field extends into the upper cladding and lower substrate in xy plane, TE mode electric field extends into the surrounding materials in xy plane. It shows that the evanescent field penetrates deeply into the cladding and substrate. The more presence of electrical field in analytes is expected to increase the photon-analyte interaction and subsequently enhance both the bulk and surface sensitivity.
In order to quantify the photon-analyte interaction and the potential of the proposed SWGRTR structure for sensing applications, we calculate the mode volume overlap integral f between the optical field and the analytes for TM and TE mode while varying the size of pillars. The overlap integral is defined as [25]: Vol. 25, No. 9 | 1 May 2017 | OPTICS EXPRESS 10530 Here, the volume integral in the numerator is the electric field energy outside of Si pillars.
The plot of the f in relation to the duty cycle and width of pillars is shown in Fig. 3(a). The f increases as the Si duty cycle and the width of pillars decrease, meaning the light confinement in the core is decreased and thus the sensitivity is enhanced. However, the optical loss will increase significantly. The photonic modes approach the light line of the cladding materials and are subject to more radiation loss resulting in quasi-guided modes when the waveguide width and Si duty cycle are both located in the green dashed region in Fig. 3(a) [26]. Considering the fundamental mode propagating in the waveguide, we choose a pillar size (η = 0.7, L = 140 nm, W = 600nm, and H = 220nm) labeled in Fig. 3(a) considering the tradeoff between sensitivity and optical loss. The overlap 39.7% of TM modes is larger than 30.2% of TE modes for pillars with the same size. Thus, a higher sensitivity for TM mode SWGRTRs is anticipated compared to the TE mode SWGRTRs with the same geometry. However, the increased interaction inevitably causes additional absorption loss by analytes and scattering loss induced by surface roughness. Therefore, a tradeoff between the photon-analyte interaction and optical loss need to be considered in design. In order to optimize the coupling strength and maximize the quality factor, we analyze the coupling efficiency between the coupling length (L c ) and the gap (G) of the racetrack waveguide and subwavelength bus waveguide with the same dimensions in Fig. 3(a) to minimize the influence of light in the racetrack [27]. We simulate the coupling efficiency shown in Fig.  3(b) changing with L c and G by RSoft BeamPROP software. The monitors are created to measure the total power along the light propagation direction in the racetrack waveguide. As shown in Fig. 3(b), larger coupling length is needed when the gap increases to achieve a maximum coupling strength from bus SWG waveguide to the racetrack waveguide; because the coupling strength is weakened as the gap is increased. The optimized coupling strength can be satisfied when gap is 140nm and the coupling length is approximately equal to 6μm according to simulations, and thus the maximum quality factor can be expected [28]. Specially, the racetrack resonator becomes a ring resonator when the L c equals 0. However, the footprint of racetrack resonator is less than that of circular ring resonator with the same circumference.

Device fabrication and results
Based on the simulation results in Fig. 3, we fabricated a group of SWGRTRs with different coupling length marked in Fig. 3(b). All devices are fabricated on SOI with a 220nm thick top silicon layer and a 3 μm thick buried oxide layer. The SWGRTR structures are firstly patterned with E-beam lithography. The patterns are transferred to the silicon dioxide (SiO 2 ) hard mask, and then to the silicon layer to form the silicon pillar layer through the reactiveion-etching. Figure 4(a) shows the optical microscope image of the fabricated SWGRTR with the coupling length of 6.5um, in which light is coupled into the strip waveguide by the TM grating coupler, then coupled into SWG waveguide by the taper, and coupled out by the other taper and grating coupler. Figure 4(b) is the scanning electron microscopy (SEM) image of the fabricated SWGRTR labeled in green dashed rectangular in Fig. 4(a). Figures 4(c), 4(d) and 4(e) are the TM mode grating coupler, the magnified SEM images of the left taper between strip waveguide and SWG waveguide, and the coupling region of the SWG bus waveguide and SWG racetrack waveguide, respectively.

Device tests and results
Fabricated devices are tested using a 1550 nm superluminescent LED (SLED) source (covering 1530-1570nm) and an optical spectrum analyzer (OSA). Light from the broadband SLED is guided through one terminal of the fiber array to the TM grating coupler on the chip and excites TM mode in the waveguide. Output light is collected by another terminal of fiber array, and optical spectra are captured by the OSA. The fabricated SWGRTRs with different coupling length are measured in DI water. Figure 5 shows the transmission spectra with 5 different L c (5.5μm, 6μm, 6.5μm, 7μm, and 7.5μm). Inset in green dashed box shows a magnified view of the resonances in the blue box with each resonance shifted for a relative distance to clarity. As seen in the inset, the quality factor and the extinction ratio of the fabricated SWGRTRs are maximized when the coupling length is 6.5μm. The quality factor of the resonance at 1555.9 nm in DI water is about 9800, and the extinction ratio is about 24.6dB. The optimized coupling length is 6.5μm in fabrication which is slightly different from the simulation results (6 μm). The fabrication induced roughness is possibly the reason for the discrepancy between the simulation and the fabrication. The fabrication results show that the fabrication yield of SWG racetrack resonators is better than that of the SWG circular ring resonators due to the large coupling length between the SWG bus waveguide and racetrack waveguide [29]. Fig. 5. The testing transmission spectra of the fabricated 5 SWGRTRs with different Lc (5.5μm, 6μm, 6.5μm, 7μm, and 7.5μm) in DI water. Inset in green dashed box is the magnified resonance dips in blue dashed box, and curves were shifted deliberately to show the difference. The bulk sensitivity of the fabricated SWGRTRs with different L c is characterized by monitoring the resonance shift when different concentration solutions are consecutively injected onto the surface of the fabricated samples through microfluidic channels. The chip stage is kept at 25°C with a temperature controller (Newport, Model 3040 Temperature Controller). In order to monitor the performance of the fabricated SWGRTRs, Fig. 6(a) shows the monitoring results of the fabricated SWGRTR with L c of 6.5μm when solutions with different concentration are applied onto the surface of sample. The vertical green dashed line represents the time period when new concentration of glycerol solution is injected and resonance shifts. Figure 6(b) shows the stabilized transmission spectra after the new concentration solutions are applied. The refractive indices for 0%, 5%, 10%, and 20% concentration solutions of glycerol are 1.333, 1.340, 1.347 and 1.362, respectively [30].  Then, to achieve the detection capability of the change of solution concentration, we analyze the iDL. The iDL is the minimum index change required to shift the resonance wavelength by one linewidth (Δλ 3dB ), which is decided by both the sensitivity and the quality factor of resonator. In view of the racetrack resonator with the change of coupling length from 5.5μm to 7.5μm, the bulk sensitivity depends on the mode volume overlap between the optical field and the analytes. The mode volume overlap is the same when only the coupling length is varied and the silicon duty cycle isn't changed, thus the bulk sensitivity is almost the same, whereas the quality factor is a measure of the photon lifetime in the racetrack resonator, which relates to the total loss of resonator and will be changed significantly with the coupling length. Therefore, we optimized the quality factor of racetrack resonator to achieve the lowest iDL at maximum Q. The iDL and quality factor of fabricated SWGRTRs with different L c are shown in Fig. 7 and fitted with Gaussian curve. It can be seen that the minimum iDL and maximum quality factor are achieved simultaneously when the coupling length equals 6.5 μm. At this coupling length, the minimum iDL of 3.71 × 10 4 was achieved, which is 32.5% lower than the best value reported in SWG resonator sensor (5.5 × 10 4 RIU) [21]. The sensitivity of the SWGRTR is higher than that of TM mode strip waveguide ring resonators with waveguide thicknesses of 150 nm (247nm/RIU) and 220nm (238nm/RIU) [20].
Moreover, the bulk sensitivity can be further improved by decreasing the Si duty cycle and the width of waveguide to increase the mode volume overlap. However, the optical loss, especially absorption loss and scattering loss, will increase inevitably, which is mentioned in the simulation of overlap in Fig. 3. The absorption loss caused by analytes in water can be suppressed by moving to a shorter wavelength operating at 1310 nm, and the devices would suffer about an order of magnitude less of the absorption loss [20]. The scattering loss can be further decreased if better fabrication condition is available to reduce the surface roughness of devices.

Conclusion
In conclusion, we have demonstrated a highly sensitive TM mode SWGRTR without deteriorating the high quality factor to enhance the intrinsic detection limit of sensor. Based on the simulation results, considering the tradeoff between light-analyte interaction and