1. Introduction

Integrated optical refractive index (RI) sensors have been broadly studied and used in different biochemical analysis applications for their fascinating features such as high sensitivity, compact size, and electromagnetic interference immunity [1]. In addition, they are promising to be incorporated into a lab-on-a-chip system for the development of low-cost point-of-care (POC) measurements. Several semiconductor optical resonant sensors with detection limit as low as 10⁻⁷~10⁻⁸ refractive index unit (RIU) have been developed [2–4]. Generally, there are two sensing schemes for resonance based sensors: one is by monitoring the resonance wavelength shift and the other is by measuring the output intensity change at a fixed wavelength [2]. Intensity sensor can reach a lower detection limit, however it has a narrow dynamic range [2]. On the other hand, resonator sensor that employs the wavelength-shift scheme has a larger dynamic range and its detection limit is decided by the ratio of the smallest detectable wavelength shift and the sensitivity of the sensor [1]. One way to achieve low detection limit is by using a resonator sensor with a narrow linewidth which gives smaller measurable wavelength shift [5]. For example, a silicon ring resonator sensor demonstrated in [4] has achieved a minimal detectable RI change of ~7.6 × 10⁻⁷ RIU. The sensor has a narrow resonance linewidth with a 3-dB bandwidith of 36 pm, and a not exceptionally high sensitivity of 163 nm/RIU [4]. However, to monitor such a small RI change, a tunable laser system with a wavelength resolution as high as ~0.1 pm is required. High-precision tunable laser is expensive and bulky, which is not suitable for developing a low-cost, portable and highly sensitive sensing system [1,6,7]. Another way to achieve low detection limit is to increase the sensor’s sensitivity [1], which however also has a wide bandwidth [6]. The detection limit of sensors with high sensitivity and large bandwidth is in fact comparable to sensors having low sensitivity and small bandwidth [1,6]. However, high-sensitivity sensor is sometime more desirable since it allows to use cheap and compact low-resolution spectrometers instead of the high-precision tunable laser [6,8]. For example, the resonance shift of a highly sensitive RI sensor can be monitored with an integrated arrayed waveguide grating (AWG) micro-spectrometer [9] together with an array of photodetectors, making it possible to be developed into a portable sensing system on a single chip [8].

Here, we show experimentally an optical RI sensor which was proposed and designed previously by us [7]. The sensor utilizes resonant light coupling between a strip waveguide and a slot waveguide assisted by a grating [7,10]. The resonance wavelength where the strongest light coupling occurs is used for the RI measurement. The dependence of the sensitivity on the slot-waveguide parameters have been investigated and discussed in detail in [7]. According to prior study, the sensor can give high sensitivity which is attributed to the use of the slot waveguide and the grating-assisted coupler mechanism. In this study, we fabricated and demonstrated the proposed sensor with silicon nitride waveguides. Obvious grating induced band-rejection and band-pass characteristics were observed. NaCl solutions at different salt concentrations were used to characterize the sensitivity of the sensor and sensitivity of −756.1 nm/RIU was achieved. In addition, by covering the strip waveguide with the silicon dioxide cladding, the sensitivity can be 2.6 times further enhanced and measured to be as large as −1970 nm/RIU. Good agreement was found between the experimental and theoretical results. The temperature dependence of the sensors was also characterized.

2. Principle of operation

Figure 1(a) illustrates the top view of the sensor, including a silicon nitride (Si₃N₄) slot waveguide and a strip waveguide placed in parallel on a silicon dioxide (SiO₂) substrate [7]. Both ends of slot waveguide are connected with a mode coupler to convert light into a strip waveguide. S bends are incorporated to separate two waveguides and vertical grating couplers are formed at the waveguide endfaces for light coupling to the optical fibers. The black dashed rectangle in Fig. 1(a) denotes the sensing window where the top SiO₂ cladding on the strip and slot waveguides is removed. The cross sections of these two waveguides are shown in Fig. 1(b). The distance between the two waveguides is s and the strip waveguide width, slot waveguide width and slot gap are denoted as W, W_s and g, respectively. The height and slab thickness of the strip and slot waveguides are denoted as h and t. The refractive indices of Si₃N₄, SiO₂ and external medium are n_Si3N4, n_SiO2 and n_ex, respectively. Generally, there is no evanescent light coupling between the two waveguides for the TE polarization because they are not synchronous in phase [7,11,12]. However, as shown in Fig. 1(a), with a corrugation grating of grating pitch Λ formed on the strip waveguide surface, strongest light coupling can be induced between the two waveguides at the resonance wavelength λ₀ when the phase-matching condition ${\lambda }_0=(N_{\text{eff}}^{strip}-N_{\text{eff}}^{slot})\Lambda =\Delta N_{\text{eff}}\Lambda$is satisfied [7,13]. $N_{\text{eff}}^{strip}$and$N_{\text{eff}}^{slot}$are the mode indices of the strip and slot waveguides. Since the mode indices of both waveguide modes change as the RI of the external medium n_ex changes, the resonance wavelength shifts accordingly in the output spectrum.

 

Fig. 1 (a) Layout of the sensor. Cross sections for sensors (b) without SiO₂ isolation and (c) with SiO₂ isolation on the strip waveguide.

Download Full Size | PDF

From the phase-matching condition, the RI sensitivity of the sensor is derived as [7]

3. Sensor fabrication

The sensor was fabricated with standard complementary metal–oxide–semiconductor (CMOS) processes. First, we deposited 3 μm-thick oxide layer on a silicon wafer using a plasma-enhanced chemical vapour deposition (PECVD), followed by deposition of 400 nm-thick Si₃N₄ layer with low pressure chemical vapour deposition (LPCVD). The waveguide shapes were defined with 248-nm deep UV lithography and dry-etched by the reactive ion etching (RIE). Then the etched core layer was covered by a 2 μm-thick top cladding silicon dioxide layer by PECVD. Finally, the sensing area was defined by the lithography process and the sensing area was exposed, using a combination of RIE dry etching and wet etching process. The nominal widths of the strip waveguide (W), slot waveguide (W_S) and slot gap (g) were 1 μm, 420 nm, and 200 nm, respectively. The separation between two waveguides (s) was 1 μm. The thickness of the Si₃N₄ waveguide (h) was 400 nm with a 50-nm thick slab. The thin slab layer was used to protect the buried oxide layer during the etching process for the fabrication of sensing window. The vertical Si₃N₄ grating coupler (17.5 μm in width, grating period Λ = 1.2 μm, etch depth = 400 nm, filling factor = 0.5, number of grating periods N = 16) in our sensor was connected to the strip waveguide via an adiabatic taper (from 17.5 μm wide to 1 μm wide, 200 μm length) [14]. The mode converter between the strip and slot waveguides was 20 μm long. Figure 2 shows the top-view scanning electron microscope (SEM) images for the grating coupler, mode converter, and two parallel slot and strip waveguides with a corrugation grating on the strip waveguide. Gratings with different periods from 14 μm to 18 μm (with step of 0.1 μm) were etched on the surface of the strip waveguides, which were expected to give band-rejection or band-pass bands within the C-band [7,13]. The length and corrugation depth for all sensors were 2000 μm and 50 nm which were chosen to give a suitable grating coupling strength [7,13]. Reference channels without grating on the strip waveguide were also fabricated. Sensors with both types of sensing window were fabricated by fully or partially etching away the SiO₂ cladding on the two waveguides and the SEM top view images and TEM cross section photos are shown in Figs. 2(d)-2(g), respectively.

 

Fig. 2 SEM top view images for (a) the grating coupler, (b) mode converter and (c) two parallel slot and strip waveguides with a corrugation grating on the strip waveguide. SEM top view images and TEM cross section images of the fabricated sensors (d,e) without and (f,g) with SiO₂ isolation on the strip waveguide.

Download Full Size | PDF

4. Sensor performance and discussion

The transmission spectra of the sensors were measured with a C-band amplified spontaneous emission (ASE) source and an optical spectrum analyzer (OSA). TE polarized light was coupled into and detected from the sensor through the vertical grating couplers. A plastic chamber was adhered onto each sensor surface and DI water or different concentration of NaCl solutions were applied. The experiment was performed at room temperature.

The spectra of a sensor without a grating on the strip waveguide (reference channel) were first measured and shown in Fig. 3(a). The sensor surface was covered with water. Light was coupled into the strip or slot waveguide and detected from both strip and slot waveguides, respectively. It is noted from Fig. 3(a), when the light was coupled into strip waveguide and detected from strip waveguide (Strip in, Strip out), the total insertion loss was found to be 35−40dB (the C-band light source had an output power close to 0 dBm at the same OSA setting) which mainly came from the coupling loss between the grating coupler and optical fiber and propagation loss of the strip waveguide. The insertion loss was similar when light was coupled into slot waveguide and detected from slot waveguide (Slot in, Slot out). As shown in Fig. 3(a), in the case that light was coupled into strip or slot waveguide and detected from slot or strip waveguide (Strip in, Slot out or Slot in, Strip out), no obvious light coupling was observed in the coupled waveguide and the extinction ratio was as large as ~40dB as compared to the cases when light was coupled and detected in the same waveguide (Strip in, Strip out or Slot in, Slot out). The results verify that there is no obvious evanescent light coupling between the strip and slot waveguides. Next, another sensor with a 2000-μm long grating (Λ = 16.9 μm) formed on the strip waveguide was measured and the spectra are shown in Fig. 3(b). The resonance wavelength for the case that light was coupled into strip waveguide and detected from strip waveguide (Strip in, Strip out) agrees reasonably well with that for the case that light was coupled into slot waveguide and detected from slot waveguide (Slot in, Slot out). Meanwhile, the contrasts at the resonance wavelengths for the two cases are significantly different. Figure 3(b) also illustrates the spectra obtained from the coupled cores, showing clear band-pass characteristics. The significant difference in both band-rejection contrast and band-pass coupling efficiency at various launching and detection conditions needs further investigation and will be discussed in another paper. In the following sensor characterization experiments, only spectra for the case that light was coupled into and detected from the strip waveguide were measured.

 

Fig. 3 Transmission spectra for sensors (a) without and (b) with a grating formed on the strip waveguide. Black: light is coupled into the strip waveguide and detected from the output of the launching strip waveguide; Red: light is coupled into the slot waveguide and detected from the output of the launching slot waveguide; Blue: light is coupled into the strip waveguide and detected from output of the coupled slot waveguide; Green: light is coupled into the slot waveguide and detected from output of the coupled strip waveguide. The grating pitch and length for the sensor with the grating is 16.9 μm and 2000 μm.

Download Full Size | PDF

The bulk sensitivity was characterized by the NaCl solutions with concentrations of 1%, 3%, 5% and 10% added to the sensor chamber. The measured normalized spectra for a sensor without an isolation cladding on the strip waveguide are shown in Fig. 4(a). The spectra were normalized with respect to that of a reference channel without grating. The grating period was 17.1 μm. It is seen from the figure that the band-rejection bands shifted to longer wavelength as the salt concentration decreased. The RI of NaCl solution is obtained from $\text{n}(\text{p}\%)=1.3105+0.17151\times \text{p}\%$ [15], where p% is the concentration. A linear dependence between the resonance wavelength and RI of NaCl solution is obtained, as shown in Fig. 4(b), with a sensitivity of −756.1 nm/RIU (R² = 0.96). The RI increase of NaCl solution induced larger increase of $N_{\text{eff}}^{slot}$ than $N_{\text{eff}}^{strip}$, resulting in the decrease of $N_{\text{eff}}^{strip}-N_{\text{eff}}^{slot}$and therefore a blue shifted resonance wavelength.

 

Fig. 4 (a) Normalized transmission spectra measured at different concentration of NaCl solutions for a sensor without an isolation layer on the strip waveguide. (b) Variation of the resonance wavelength with the RI of NaCl solution.

Download Full Size | PDF

The normalized measured transmission spectra for another sensor with an isolation layer on the strip waveguide and a grating with a period of 15.1 μm are illustrated in Fig. 5(a). Again, the resonance wavelength shifted to longer wavelength as the salt concentration decreased. The results for the resonance wavelength dependence on the RI of NaCl solution are shown in Fig. 5(b). The sensor gives a linear response to RI changes and the sensitivity is −1970 nm/RIU, which is 2.6 times more sensitive than the one without cladding isolation. For an OSA with a minimum measurable wavelength shift of 0.01nm, the sensor can achieve a smallest detectable RI value of 5 × 10⁻⁶ RIU.

 

Fig. 5 (a) Normalized transmission spectra measured at different concentration of NaCl solutions for a sensor with an isolation layer on the strip waveguide. (b) Variation of the resonance wavelength with the RI of NaCl solution.

Download Full Size | PDF

The sensitivity of these two types of sensor were also calculated using the fabricated waveguide parameters obtained from TEM pictures shown in Figs. 2(e) and 2(g). The parameters used in the calculation were W_co = 945nm, t_slab = 44nm, h = 400nm, s = 1 μm, n_co = 1.89, n_s = 1.444, n_ex = 1.3105. The slot gap was wider at top (g_top = 284nm, W_slottop = 360 nm) and narrower at bottom (g_bottom = 195nm, W_slotbottom = 404.5nm). The results are summarized in Table 1. The calculated sensitivity values for both sensors agree well with the experimental results.

Table 1. Sensitivity calculation for the sensor.

View Table

The temperature dependence of the resonance wavelengths for both sensors were also characterized. A peltier heat pump was placed under the sensor to control the temperature. DI water was added into the chamber. The measured results are shown in Fig. 6. As shown in Fig. 6, the resonance wavelengths of both sensors shifted to longer wavelength as temperature increased and their temperature sensitivity is linearly fitted to be 251 pm/°C (with SiO₂ isolation, R² = 0.99) and 160 pm/°C (without SiO₂ isolation, R² = 0.97), respectively. Both the direction and magnitude of the temperature sensitivity agree quite well with the analysis in [7]. To further reduce the temperature dependence of the sensor, a polymer isolation layer with negative thermal-optic coefficient should be used instead of the SiO₂ isolation layer [7].

 

Fig. 6 Dependence of the resonance wavelengths on the temperature for sensors without and with SiO₂ isolation.

Download Full Size | PDF

5. Conclusion

We successfully demonstrate a resonance-based RI sensor with silicon nitride waveguides. The resonance is produced by the grating-assisted light coupling between a strip waveguide and a slot waveguide. The RI change of external medium is measured by monitoring the resonance wavelength shift. Two types of such senor are fabricated: one with SiO₂ upper cladding on the strip waveguide and one without. The sensitivity characterization experiments show that the sensor with cladding isolation on strip waveguide has a sensitivity value as high as −1970 nm/RIU, being ~2.6 times larger than that of the sensor without cladding isolation. The high sensitivity facilitates the use of low cost and compact spectrum analyzer. In addition, unlike the ring resonators that support multiple resonances, there is only one resonance corresponding to each grating period for this demonstrated sensor. Therefore, its spectrum is FSR (free-spectral range)-free which gives a wide dynamic range to the sensor. Furthermore, since the grating is intrinsically wavelength-selective, the sensor can be extended for wavelength multiplex measurement by cascading several gratings with different periods along the strip waveguide. In the output spectrum, resonance peaks corresponding to different grating periods could be used as different sensing channels for multiplexed measurement.